SMARCA2-deficiency confers sensitivity to targeted inhibition of SMARCA4 in esophageal squamous cell carcinoma cell lines

SMARCA4/BRG1 and SMARCA2/BRM, the two mutually exclusive catalytic subunits of the BAF complex, display a well-established synthetic lethal relationship in SMARCA4-deficient cancers. Using CRISPR-Cas9 screening, we identify SMARCA4 as a novel dependency in SMARCA2-deficient esophageal squamous cell carcinoma (ESCC) models, reciprocal to the known synthetic lethal interaction. Restoration of SMARCA2 expression alleviates the dependency on SMARCA4, while engineered loss of SMARCA2 renders ESCC models vulnerable to concomitant depletion of SMARCA4. Dependency on SMARCA4 is linked to its ATPase activity, but not to bromodomain function. We highlight the relevance of SMARCA4 as a drug target in esophageal cancer using an engineered ESCC cell model harboring a SMARCA4 allele amenable to targeted proteolysis and identify SMARCA4-dependent cell models with low or absent SMARCA2 expression from additional tumor types. These findings expand the concept of SMARCA2/SMARCA4 paralog dependency and suggest that pharmacological inhibition of SMARCA4 represents a novel therapeutic opportunity for SMARCA2-deficient cancers.


CRISPR-Cas9 screening identifies SMARCA4 as a novel dependency in ESCC.
To identify potential novel drug targets in ESCC, we applied domain-directed CRISPR-Cas9 screens 37 across ten ESCC cell lines. Stable Cas9-expressing cell lines were transduced with a sgRNA library consisting of >1300 sgRNAs targeting 179 epigenetic regulators 38 and negative selection of sgRNA-expressing cells was monitored by next-generation sequencing after 18 population doublings (Fig. 1A). From these screens, we identified SMARCA4 as a strong dependency in a subset of the ESCC cell lines ( Reciprocal to the known dependency of SMARCA4-deficient cancer cell lines on SMARCA2 [25][26][27] , the majority of SMARCA4-dependent ESCC cell lines (KYSE-270, KYSE-30, KSYE-510 and COLO-680N) displayed low or non-detectable levels of SMARCA2 mRNA and protein expression ( Supplementary Fig. S1A). In contrast, T.T and KYSE-410 cells were dependent on SMARCA4 despite readily detectable SMARCA2 expression. SMARCA4 mRNA and protein expression was detected in all ESCC lines ( Supplementary Fig. S1A).
To establish a link between SMARCA2 expression levels and dependency on SMARCA4 beyond the cell models used in this study, we analyzed functional genomic datasets from recent pooled shRNA and CRISPR-Cas9 viability screens 39,40 , including 14 and 16 ESCC cell lines, respectively. In both datasets, sensitivity to SMARCA4 loss-of-function was anti-correlated with SMARCA2 expression levels ( Supplementary Fig. S1B).
We performed depletion studies using individual sgRNAs to corroborate and strengthen the findings of the pooled CRISPR-Cas9 screens ( Supplementary Fig. S2A). A higher frequency of loss-of-function mutations are observed with sgRNAs directed to sequences encoding functionally relevant protein domains 37 . Therefore, 17 different sgRNAs targeting DEXDc-, helicaseC-, bromodomain or N-terminal coding regions of SMARCA4 were tested in the KYSE-30 and KYSE-510 cell lines ( Supplementary Fig. S2B). Both cell lines showed strong depletion effects with sgRNAs targeting the ATPase encoding regions, while efficacious depletion with bromodomain-directed sgRNAs was observed in KYSE-30 cells only. In contrast, N-terminal targeting sgRNAs displayed only mild effects in both models ( Supplementary Fig. S2B).
We then screened a selected panel of SMARCA2-proficient and -deficient ESCC lines in a time-resolved fashion using the three most efficacious DEXDc-and bromodomain targeting sgRNAs ( Fig. 2A). Absence of SMARCA2 protein expression in KYSE-270, KYSE-30, KYSE-510 and COLO-680N cells was confirmed by capillary Western immunoassay (Fig. 2B). No strong depletion effects were observed upon knock-out of SMARCA4 in the SMARCA2-proficient cell models KYSE-450, KYSE-140, KYSE-70 and KYSE-150 over a course of 28 days. In contrast, targeting SMARCA4 in the SMARCA2-deficient cell models KYSE-270, KYSE-30, KYSE-510 and COLO-680N resulted in depletion effects similar or stronger than observed for the POLR2A positive control sgRNA ( Fig. 2A). With the exception of KYSE-30, DEXDc-targeted sgRNAs showed stronger and faster depletion than bromodomain-directed sgRNAs in SMARCA2-deficient cells ( Fig. 2A,B), indicating a requirement of SMARCA4 ATPase activity in the context of SMARCA4 dependence 37 . Of note, SMARCA4 bromodomain-targeting, but not ATPase-directed, sgRNAs were included in the pooled sgRNA library. The effect observed with SMARCA4 bromodomain-targeting sgRNAs in the individual sgRNA depletion studies (Fig. 2) thus might be exacerbated in the pooled negative selection setting (Fig. 1B).
We wanted to extend our studies beyond ESCC cell models. Therefore, we selected non-ESCC cell lines with low SMARCA2 mRNA expression 43 and identified four cell lines -HCT 116, SK-CO-1 (both colorectal carcinoma),   HuP-T4 (pancreas carcinoma) and OV-90 (ovarian carcinoma) -with low or non-detectable SMARCA2 protein expression suitable for CRISPR-Cas9 depletion studies ( Supplementary Fig. S5). Similar to the effects observed in ESCC cell lines, strong depletion of SMARCA4 sgRNAs was observed in all four SMARCA2-deficient cell models tested.
In summary, the SMARCA2 reconstitution and knock-out studies demonstrate a strong and hard-wired synthetic lethal interaction of SMARCA2 and SMARCA4 in ESCC cell models. Further, our data highlight SMARCA4 as a universal vulnerability of SMARCA2-deficient cancer cells, irrespective of the tumor type.

Pharmacological targeting of SMARCA4 impairs viability of a SMARCA2-null ESCC model. A
sgRNA-resistant SMARCA4 expression construct in which the SMARCA4 bromodomain is replaced with the bromodomain of BRD9 (SMARCA4 res -BD BRD9 ) was designed to determine whether acute degradation of SMARCA4 represents a therapeutic approach against SMARCA2-deficient cancers (Fig. 5A). The bromodomain substitution renders the SMARCA4 variant susceptible to targeted degradation via a potent and highly selective Proteolysis Targeting Chimera (PROTAC) directed against the bromodomain of BRD9, referred to as dBRD9 44 . SMARCA4 res -BD BRD9 was transduced in SMARCA4-dependent KYSE-30 cells and, subsequently, a monoclonal line harboring knock-out of the endogenous form of SMARCA4 was generated (KYSE-30-SMARCA4 res -BD BRD9 ). Treatment with dBRD9 for four hours led to a decrease in SMARCA4 protein levels in the KYSE-30-SMARCA4 res -BD BRD9 cell line, but not parental KYSE-30 cells (Fig. 5B). Ectopic expression of SMARCA4 res -BD BRD9 rendered KYSE-30-SMARCA4 res -BD BRD9 cells inert to SMARCA4-targeting sgR-NAs, indicating functionality and sufficient expression of the bromodomain substituted SMARCA4 variant ( Supplementary Fig. S6). In line with the genetic depletion of SMARCA4, dBRD9-mediated degradation of SMARCA4 was accompanied by decreased viability of KYSE-30-SMARCA4 res -BD BRD9 cells with a half maximal inhibitory concentration (IC 50 ) of 322 ± 122 nM. In contrast, parental KYSE-30 cells were not affected by treatment with dBRD9 (Fig. 5C).
The results indicate that pharmacological targeting of SMARCA4 might constitute a novel therapeutic approach for the treatment of SMARCA2-deficient tumors.

Discussion
Besides surgery, chemotherapy and radiotherapy regimens constitute the mainstay of clinical therapy of ESCC. Responses to combination chemotherapy regimens are short-lived in advanced ESCC due to the rapid emergence of resistance 45,46 . While recent large scale genomic analyses have provided a detailed landscape of genomic alterations in ESCC [33][34][35][36] , approaches to clinically exploit selective vulnerabilities in ESCC with targeted agents have proven unsuccessful to date 45,47,48 .
The results of this study uncover the BAF complex ATPase SMARCA4 as a novel vulnerability in a subset of ESCC cell lines characterized by low or absent expression of SMARCA2 -a relationship reciprocal to the known synthetic lethal interaction between SMARCA2 and SMARCA4 in SMARCA4-deficient cancer cell lines [25][26][27] . We demonstrate that lack of SMARCA2 is both required and sufficient to induce dependency on the paralogous ATPase SMARCA4. These findings extend the concept of synthetic lethality between SMARCA2 and SMARCA4 in cancer cells. In addition to ESCC cell models, SMARCA2-deficient colorectal, pancreas and ovarian carcinoma cell lines display SMARCA4 dependency, indicating that the interdependence of SMARCA2 and SMARCA4 is hard-wired and context-independent. SMARCA4 dependency of SMARCA2-deficient cell lines is linked to its ATPase, but not bromodomain, function. This result is in line with a strict requirement on SMARCA2 ATPase activity in SMARCA4-mutant NSCLC cell lines 25,41 . Either SMARCA2 or SMARCA4 are sufficient to rescue from loss of SMARCA2 ATPase activity, further emphasizing the context-independent nature of SMARCA2/ SMARCA4 synthetic lethality.
Of note, two ESCC cell lines, T.T and KYSE-410, display dependence on SMARCA4 despite readily detectable levels of SMARCA2. In various cellular contexts, SMARCA4 has been linked to an oncogenic function 49 and high expression levels of SMARCA4 are associated with poor prognosis in multiple tumor types 50 . A critical requirement on SMARCA4 in promoting MYC-dependent transcription has been described in SMARCA2-proficient acute leukemia cell models 51 . Thus, the dependency on SMARCA4 in the presence of SMARCA2 in T.T and KYSE-410 cells might be attributed to a context-dependent requirement on SMARCA4 function distinct from the paralog interdependence. Further work is required to elucidate the mechanisms underlying the dependency on SMARCA4 in the SMARCA2-proficient ESCC cell models.
In contrast, the function of SMARCA4 in cancer has also been linked to a tumor suppressive role. Inactivating mutations of SMARCA4 are frequently observed in human cancer, most prominently in non-small cell lung cancer (NSCLC) 13,22,[52][53][54][55] . The tumor suppressive function of SMARCA4 is corroborated by Smarca4 knock-out studies in mice. While homozygous deletion of Smarca4 is embryonic lethal, heterozygous loss of Smarca4 predisposes to tumor formation of epithelial origin 56,57 . Lung-specific, conditional ablation of Smarca4 enhances tumor formation in a carcinogen-induced lung cancer model 58   www.nature.com/scientificreports www.nature.com/scientificreports/ occurs with long latency and low penetrance 56,57 , indicating that for treatment of SMARCA4-dependent cancers a therapeutic index for transient inhibition of SMARCA4 might exist without the risk of secondary malignancies.
In contrast, loss-of-function mutations of SMARCA2 are primarily linked to developmental/neurological disorders 59 . Missense mutations or in-frame deletions affecting the ATPase function of SMARCA2 are causative for Nicolaides-Baraitser syndrome, a rare condition characterized by sparse hair, facial and limb abnormalities and intellectual disabilities 60,61 . In cancer, mutations of SMARCA2 are rarely observed 14 . The lack of SMARCA2 expression in cancer cells might therefore be related to cell-specific and epigenetic regulation mechanisms 50,62 . A notable exception is adenoid cystic carcinoma (ACC), a rare type of cancer arising from salivary glands where mutations within the ATPase domain of SMARCA2 are found in approximately 5% of cases 63 , indicating a potential tumor suppressive function of SMARCA2 in this context.
Using a PROTAC approach in an engineered cell model, we demonstrate that pharmacological targeting of SMARCA4 effectively impairs proliferation of a SMARCA2-deficient cell line. Our data suggest that SMARCA4 inhibition would allow for a selective, cancer-cell directed therapy sparing normal, SMARCA2-expressing cells and tissues. Expression of SMARCA2 is predominantly detected in brain, liver, muscle and endothelial cells and has been linked to slow-cycling and differentiated cell states 64 . While it is difficult to define a threshold for expression of SMARCA2 that maintains sufficient BAF functionality in the absence of SMARCA4, the fact that homozygous loss of SMARCA4 is embryonic lethal, indicates that SMARCA2 cannot universally compensate for SMARCA4 function 56 . In adult tissues, dual inactivation of SMARCA2 and SMARCA4 primarily affects viability of vascular endothelial cells 65 . Pharmacological targeting of SMARCA4 would therefore require a thorough analysis of potential toxicities in normal cell and tissues.
In human cancer also concomitant loss of SMARCA2 and SMARCA4 expression has been described 25,[66][67][68] . In small-cell carcinomas of the ovary, hypercalcemic type (SCCOHTs), characterized by bi-allelic inactivating mutations of SMARCA4, loss of SMARCA2 expression is frequently observed [68][69][70][71] . These examples depict a striking exemption of the concept of hard-wired SMARCA2/SMARCA4 paralog dependency and constitute a potential mechanism for resistance to BAF targeted cancer therapies 72 . Further studies are required to decipher whether alternative ATP-dependent nucleosome-remodeling complexes, such as ISWI and CHD complexes 73 , can fulfill a compensatory function in the absence of BAF activity.
Our study highlights the use of functional genomic screening for identification of novel vulnerabilities in cancer cell lines. Given the possibility to develop selective SMARCA4 ATPase inhibitors or PROTACs selectively inducing SMARCA4 proteolysis 72,74,75 and tolerability of systemic SMARCA4 inactivation, the results of this study suggest that targeting of SMARCA4 might be a valuable strategy for the treatment of tumors characterized by low or absent expression of SMARCA2.