The Interaction of SWI/SNF with the Ribosome Regulates Translation and Confers Sensitivity to Translation Pathway Inhibitors in Cancers with Complex Perturbations

The SWItch/Sucrose Non-Fermentable (mSWI/SNF) complexes are ATP-dependent chromatin remodelers that influence chromatin architecture and gene expression (1–3). The three main human complexes, BAF, PBAF, and ncBAF, are composed of common and specific subunits altered by mutations and copy-number variation in more than 40% of cancers (1, 2, 4). Many mutations result in loss-of-function driving disease progression and therapeutic resistance (1–3, 5).

While mSWI/SNF functions are mostly studied through direct interaction with the nucleosome, few studies investigate nonchromatin substrates. In yeast, SWI/SNF evicts Sir3p from the nucleosome (6), and Brg1 (SMARCA4) activates Mec1 kinase (7). In humans, BAF evicts PRC1 from the chromatin (8). Furthermore, mSWI/SNF subunits shuttle between nucleus and cytoplasm (9, 10). Differential localization could be disease-relevant: SMARCA4 is cytoplasmic in some corticotroph adenomas (11), nuclear SMARCA2 correlates with better prognosis in lung cancer (12), and cytoplasmic ARID1B promotes oncogenesis and correlates with advanced pancreatic cancer (13). Together, these studies provide evidence for possible disease-associated functions for mSWI/SNF outside roles on the nucleosome.

In this study, we report mSWI/SNF-mutated cancer cell lines’ genetic dependency to translation factors and sensitivity to translation pathway inhibitors. We demonstrate mSWI/SNF subunits interact with the translation machinery and affect translation. mSWI/SNF inhibition and depletion in combination with mTOR-PI3K inhibitors decrease cell viability, suggesting a possibility for translation-targeted therapy in mSWI/SNF-perturbed diseases. Indeed, expression of SMARCA4 loss-of-function pathogenic mutations decreases translation. Finally, we reveal the substantial alteration of mSWI/SNF subunits’ nuclear and cytoplasmic expression in breast cancers. Together, our data support a direct role for mSWI/SNF in translation with the potential to develop into new therapeutic strategies.

HEK293T and HME-1 cells were obtained from ATCC. HAP1—wild-type (WT) and HAP1-SMARCA2 (HZGHC004055c012), -SMARCA4 (HZGHC002878c004), -ARID2 (HZGHC000907c009), -ARID1A (HZGHC000618c010), -BRD7 (HZGHC000923c010), -BRD9 (HZGHC000934c010) knockout (KO) cell lines were obtained from and characterized by Horizon Discovery. Cells were cultivated following supplier's instructions. HAP1 cell lines were used under 30 passages. Mouse embryonic stem cells genome editing and culture are described in ref. 14. Cells were regularly tested for Mycoplasma (Mycoalert, Lonza).

Cell fractionation was performed as in ref. 15. For the Western blot analysis a comparable fraction of each compartment was loaded on a gel.

See supplements.

The studies were conducted in accordance with recognized ethical guidelines, utilized retrospective tissue samples, and were approved by an institutional review board. The tissue microarray includes a wide range of patient ages, tumor grades, and are composed of estrogen receptor–positive, HER2-positive, and triple-negative cases. It is composed of 46 tumors cases from 2002 to 2019 including primary tumors and nodal and distant metastases. Targeted genomic data are available for these tumors. Clinical information and results are reported in Supplementary Table S1. SMARCA4 was performed on a Roche Ventana Benchmark Ultra instrument using CC1 retrieval and antibody dilution at 1:25. SMARCA2 (1:300), BRD7 (1:400), and PB1 (1:2,000) were performed manually using Tris-EDTA antigen retrieval (pH 9). ARID2 (1:250) was performed manually using citrate antigen retrieval (pH 6). Phospho-RPS6 was performed with VECTASTAIN Elite ABC Universal PLUS kit (Vector Laboratories Inc.) using citrate antigen retrieval (pH 6). Subcellular localizations were identified by comparison with hematoxylin and eosin (H&E) staining from the same samples.

Immunoprecipitation was performed as in ref. 16. +RNAse: RNAse A (1 μg/mL) was added in the lysates prior to immunoprecipitation. Mass spectrometry (MS) experiments were performed using a nanoAcquity liquid chromatography (Waters) coupled to an Orbitrap Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific). For complete MS protocol and data analysis see supplements.

Proximity ligation assays were performed were performed following supplier's instructions from Duolink PLA (Sigma Millipore). HEK293T cells were seeded 24 hours prior the experiments to reach a confluency of 80%. Cells were imaged with Leica DM400 objective 40×/1.27.

Polysome profiling were performed as in ref. 16. Because of the high number of proteins tested and size restriction, identical fractions from polysome profiles were assessed with several gels. mSWI/SNF inhibitors concentrations: iBRD9 10 μg/ml, PFI-3 5 μmol/L, TP472 5 μmol/L.

Simultaneous overview of trimolecule biosynthesis (SOM₃B) was performed as in ref. 17. Cells were treated with inhibitors, puromycin, IdU, and BrU for 10 minutes. Single-cell mass cytometry datasets are available in flowrepository.org (FR-FCM-Z3XL).

L-Azidohomoalanine (AHA) incorporation were performed as in ref. 16, AHA was added for the last hour of drug treatments. Puromycin incorporation assays were performed as in ref. 18, puromycin was added for the last 30 minutes of drug treatments. Antibodies intensities from Western blot membranes were analyzed using ImageJ. Because of the string smear obtained for these Western blots, loading controls (actin) were run on separate membranes. Quantification panels represent the average of three independent experiments. mSWI/SNF inhibitors concentrations: iBRD9 10 μg/mL, PFI-3 5 μmol/L, TP472 5 μmol/L.

Viability assays were performed following supplier's instructions from the CellTiter-Glo 2.0 Cell Viability Assay (Promega). Luminescence reading was performed using GloMax Explorer Microplate Reader (Promega). mSWI/SNF and translation pathway inhibitors concentrations (unless annotated): iBRD9 10 μg/mL, PFI-3 5 μmol/L, TP472 5 μmol/L, AZD8055 50 nmol/L, BEZ235 50 nmol/L.

Experiments were performed at least three independent times. Data are expressed as mean ± SEM. P values < 0.05 calculated with a two-tailed Student t test were considered significant and annotated with “*” or “#.”

The data generated in this study are available within the article and its Supplementary data files. Single-cell mass cytometry datasets are available in flowrepository.org (FR-FCM-Z3XL).

Analysis of the DepMap PRISM repurposing primary drug sensitivity screen testing 578 cell lines with 4,686 compounds (19) revealed that cells harboring mutations in 11 of 29 mSWI/SNF subunits were more sensitive to inhibitors targeting translation and its pathways when compared with all other cell lines. Further analysis of the DepMap genetic dependency study reporting RNAi against 17,309 genes in 712 cell lines (20) revealed that cells harboring mutations in 15 of 29 mSWI/SNF subunits exhibited genetic dependencies to one or more translation factors. For example, ARID1A and ARID1B mutated cancer cells were more sensitive to AKT and mTOR/PI3K inhibitors (Fig. 1A; Supplementary Fig. S1A) and were genetically dependent to the ribosomal subunit RPL22L1, the RNA helicases DDX27 and DHX9, and SRSF1, a splicing factor with direct role in translation (Fig. 1B; Supplementary Fig. S1B).

To further understand the unexpected association of mSWI/SNF with translation, we assessed mSWI/SNF subunits subcellular localization. Biochemical fractionation from human embryonic kidney HEK293T cells revealed their substantial presence in the cytoplasm (Fig. 1C). We therefore hypothesized mSWI/SNF could be involved in translation. Immunoblots from polysome profile fractions revealed mSWI/SNF subunits enriched in initiating fractions (Fig. 1D–E; Supplementary Fig. S1C). We confirmed mSWI/SNF interaction with translation factors by coimmunoprecipitation, proximity ligation assays (various intensity of cytoplasmic signal was observed in the vast majority of the cells), and MS (Fig. 1F–H; Supplementary Fig. S1D–S1H). Some subunits such as SMARCA4 were detected in the cytoplasm only upon separation of the translation initiation machinery (Supplementary Fig. S1C compare fractions 1 to 2–4), when performing proximity ligation assays (Fig. 1G), or when enriched by immunoprecipitation (Supplementary Fig. S1D and S1E). Consistent with mSWI/SNF subunits forming complexes (21), deletion of BRD7 led to the loss of PB1 but not SMARCA4 nor SMARCB1 from initiating fractions (Fig. 1I). These data suggest that mSWI/SNF complexes could directly impact translation.

To investigate involvement in translation, we tested mSWI/SNF bromodomain inhibitors (Fig. 2A). A 24-hour treatment decreased translation (Fig. 2B and C) without altering proliferation (Supplementary Fig. S2A). To decipher if these translational defects could be a consequence of direct roles in translation, we shortened treatment to 10 minutes to capture any changes in translation before transcriptional changes would be observed at the protein level, as RNA transcription, export, and translation take at least 20 minutes for small proteins (22, 23). A 10-minute inhibition reduced translation to a similar extent as 24 hours (Fig. 2D). We then performed SOM₃B to measure viability and puromycin and BrU incorporation in single cells (17). We confirmed 10-minute inhibition decreased translation in live cells (Supplementary Fig. S2B–S2C). While we did not detect any changes in transcription (data not shown), 10 minutes is a short timeframe for BrU incorporation and the detected signal was not reliable. Translation remains active during interphase and is reduced during mitosis (24). To assess translation across cell cycle, we examined IdU incorporation, cyclin B1, and histone H3 serine 10 phosphorylation (17). mSWI/SNF inhibition reduced translation across interphase and very modestly in mitosis (Supplementary Fig. S2D–S2E). Finally, we demonstrated that 10 minutes of mSWI/SNF inhibition substantially decreased RPS6 phosphorylation, consistent with a decrease of translation (Fig. 2E; ref. 25).

We then assessed translational defects by polysome profiling. Inhibition for 24 hours increased 80S and decreased polysomes, typical of a translation initiation defect. Inhibition for 10 minutes increased polysomes, which could be reflective of ribosome collision (Fig. 2F; Supplementary Fig. S2F–S2I).

Because mSWI/SNF inhibition decreased translation, we hypothesized mSWI/SNF inhibitors could increase cells’ sensitivity to translation inhibitors. A 72-hour co-treatment of AZD8055 (mTOR) or BEZ235 (mTOR-PI3K) and iBRD9 or TP472 decreased further HEK293T viability than treatments alone (Fig. 2G; Supplementary Fig. S2J).

We confirmed our findings using mouse embryonic stem cells containing an auxin-dependent degron within Smarca4, Arid1A, or Pb1 alleles, allowing a complete degradation of endogenous proteins within 2 hours of auxin treatment (14). Degradation of Smarca4 and Arid1A but not Pb1 decreased translation (Fig. 3A), suggesting the defects observed upon PFI-3 to be the consequences of SMARCA2/SMARCA4 inhibition. Depletion of Smarca4 equally up- and downregulated specific RNAs (14), validating potential independent roles in translation.

We then utilized HAP1-KO for PBAF (BRD7, ARID2), BAF (ARID1A), ncBAF (BRD9), and ATPases (SMARCA2, SMARCA4; Supplementary Fig. S3A). mSWI/SNF subunits presented nuclear and cytoplasmic distributions in HAP1-WT (Supplementary Fig. S3B) and their KO did not alter proliferation (Supplementary Fig. S3C). A 10-minute treatment with iBRD9 and TP472 decreased translation (Supplementary Fig. S3D). Depletion of each subunit decreased translation to various degrees (Fig. 3B) but presented distinct polysome profiles. SMARCA4-, ARID2-, and BRD7-KO increased initiating/early and decreased late polysomes (Supplementary Fig. S3E–S3G); SMARCA2- and ARID1A-KO decreased the entire profile (Supplementary Fig. S3H and S3I); and BRD9-KO, albeit more variable, increased polysomes (Supplementary Fig. S3J and S3K). These distinct profiles suggest specific roles for BAF, PBAF and ncBAF in the cytoplasm as described in the nucleus. Finally, and demonstrating specificity, reexpression of GFP-SMARCA4 in SMARCA4-KO rescued the translational defects (Fig. 3C).

Many mSWI/SNF cancer mutations being loss-of-function (1–3) and some heterozygous missense dominant-negative (26), we hypothesized they could alter translation. Transient overexpression of SMARCA4 pathogenic mutations p.K785R (ATP cleft) and p.E861K (DNA groove) decreased translation without altering proliferation (Fig. 3D; Supplementary Fig. S3L).

Because depletion of mSWI/SNF subunits decreased translation, we hypothesized it would sensitize cells to translation inhibition. ARID2- and BRD7-KO increased sensitivity to AZD8055 and BEZ235 (Fig. 3E; Supplementary Fig. S3M), while SMARCA2-, SMARCA4-, ARID1A-, and BRD9-KO did not (Fig. 3F–G; Supplementary Fig. S3N S3O). Consistent with ARID1A mutations increasing cells’ sensitivity to AKT inhibitors (Fig.1A), ARID1A-KO sensitized cells to these inhibitors (Fig. 3H). Therefore, depletion of mSWI/SNF subunits creates a vulnerability to specific inhibitors.

Several mSWI/SNF subunits are cytoplasmic in tumors and altered localization could be useful markers (11–13). While primarily considered tumor suppressors, evidence starts to shed light on potential oncogenic roles. For example, SMARCA4 is likely an oncogene in breast cancer (27). Mis-expression or localization not being necessarily associated with genomic alteration, we evaluated mSWI/SNF subunits in breast cancer by IHC. Mutations in the PBAF-specific subunits ARID2 and PB1 being less frequent in breast cancers than the BAF-specific ARID1A and ARID1B (1, 3), we investigated expression of PBAF subunits.

PB1 presented various levels of nuclear expression in normal breast epithelium. Forty-one percent of cases presented a strong and 22% a weak cytoplasmic expression, and 30% a loss of expression (Fig. 4A; Supplementary Fig. S4A).

BRD7 presented an overall cytoplasmic with a weak nuclear expression in less than 50% of normal breast epithelium. Sixty-eight percent of cases exhibited increased nuclear and cytoplasmic expression and 16% increased cytoplasmic (Fig. 4B; Supplementary Fig. S4A).

ARID2 resembled BRD7 in normal breast epithelium. Thirty-one percent of cases presented increased nuclear expression, 26% increased cytoplasmic, 11% increased cytoplasmic and nuclear, and 12% cytoplasmic loss (Fig. 4C; Supplementary Fig. S4A).

All considered, most cases presented an increase in PBAF subunits expression. Of note, some patients presented both increased and lost expression and/or variation in compartmental distribution across different samples, highlighting the heterogeneity between primary, recurrent, and metastatic samples (Supplementary Fig. S4B).

We then assessed both ATPases. As previously reported, SMARCA2 exhibited moderate nuclear expression in approximately 50% of normal breast epithelium (28). Sixty-one percent of cases demonstrated loss of SMARCA2, 9% a concomitant increased cytoplasmic, 11% increased nuclear and cytoplasmic, and 7% increased cytoplasmic only (Fig. 4D).

SMARCA4 was very minimally expressed in normal breast epithelium, with occasional cells presenting nuclear expression (28). Fifty-one percent of breast cancer cases presented increased nuclear and cytoplasmic SMARCA4 and 18% a weak cytoplasmic expression that could be considered as background (a potential limitation of IHC; Fig. 4E; Supplementary Fig. S4A). Five cases (11%) harbored a SMARCA4 mutation, three of which with increased expression (Supplementary Fig. S4C).

Taken together, 14% of cases presented an increased cytoplasmic SMARCA4 and SMARCA2, and 66% an increased expression of SMARCA4 with a loss of SMARCA2, suggesting a swap of ATPases within mSWI/SNF complexes (Fig. 4F).

Because 10 minutes of mSWI/SNF inhibition decreased phospho-RPS6 (Fig. 2E), we examined its presence in our breast cancer specimens. Each core with high cytoplasmic SMARCA4 presented phospho-RPS6 (Fig. 4G; Supplementary Table S2), consistent with a correlation between cytoplasmic SMARCA4 and levels of translation in breast cancer.

Finally, assessment of clinical information suggested a correlation between cytoplasmic SMARCA4 and breast cancer progression: SMARCA4 presented a high cytoplasmic expression in 7% of primary tumors, 31% of local recurrences, and 42% of metastases (Supplementary Table S1).

This study reports the cytoplasmic presence of mSWI/SNF and its implication in translation. KO of each tested subunit decreased translation, but cell sensitivity to mTOR-PI3K inhibition increased upon depletion of PBAF-specific subunits. This could reflect complexes’ specific roles in translation. Confirming a potential specificity, PBAF-KOs presented identical polysome profiles. SMARCA4-KO polysome profiles resembled PBAF-KOs and SMARCA2-KO resembled BAF-KO; future studies will determine if cytoplasmic complexes involve specific ATPases, in contrast to the nucleus (21).

mSWI/SNF-mutated cancer cells were genetically dependent to translation factors and sensitive to translation pathway inhibitors. This could be used as a clinical strategy, such as PARP inhibitors for BRCA-deficient cancers (29). However, some subunits such as BRD9 did not present such dependency even though iBRD9 decreased translation. Only specific mutations (i.e., bromodomain) could increase sensitivity to translation pathway inhibitors and be missed when considering all mutations. A systematic study of each subunit's cancer alterations will be necessary to assess consequences on translation and sensitivity to translation pathway inhibitors.

Large-scale sequencing studies revealed the cooccurrence of mutations in PI3K and mSWI/SNF, and a cooperative mechanism suggested (2, 3). While biologically this cooccurrence is not understood, roles for mSWI/SNF in translation could be relevant. The decreased translation upon mSWI/SNF-inactivating mutations could need the counteraction of increased translation through PI3K-activating or PTEN-inactivating mutations (3). Altering the translation of specific mRNAs is another route to enable a selective advantage to cancer cells. We do not observe any association between PI3K mutations and mSWI/SNF altered expression/localization in breast cancers. SMARCA4 was overexpressed in more than 90% of cases, confirming a potential oncogene and its low expression as a favorable outcome (3, 27). On the contrary, SMARCA2 expression was extensively lost. It will be interesting to determine if such cases are more sensitive to translation inhibitors, especially when SMARCA4 is not overexpressed and/or mutated. Furthermore, some cases exhibited loss of PBAF subunits, which could sensitize to mTOR/PI3K inhibitors. Finally, we also observed several cases with exclusive gain in cytoplasmic expression or opposite changes within different compartments. The consequences of differences in subcellular functions could have great potential for the discovery of therapeutically targetable events.

Our study supports a direct role for mSWI/SNF complexes in translation, however, we cannot completely exclude that other mSWI/SNF functions are important. Transcription and translation being intertwined, it is challenging to experimentally separate one from the other. Furthermore, mSWI/SNF complexes could regulate the transcription and translation of identical mRNAs. However, understanding the direct and indirect functions of mSWI/SNF in translation is essential to understanding the consequences of their cancer alterations.

In summary, we underscore that mSWI/SNF cancer mutations render cells sensitive to translation pathway inhibitors and genetically dependent to translation factors. We further demonstrate mSWI/SNF interacts with the translation machinery and its inhibition, depletion, and cancer mutations decrease global translation. Overall, this study highlights that considering unappreciated functions for chromatin remodelers could be essential to the discovery of new therapeutic strategies.

G.R. Crabtree reports personal fees and other support from Foghorn Therapeutics outside the submitted work. C. Van Rechem reports grants from The Children's Cancer Research Fund, The Stanford Women's Cancer Center and the Stanford Cancer Institute (Stanford, CA), The Jacob Churg Award, and grants from The Weintz Family COVID-19 Research Fund during the conduct of the study. No disclosures were reported by the other authors.

L. Ulicna: Conceptualization, investigation, writing–review and editing. S.C. Kimmey: Investigation, writing–review and editing. C.M. Weber: Resources, investigation, writing–review and editing. G.M. Allard: Investigation. A. Wang: Resources, data curation, supervision, funding acquisition, writing–review and editing. N.Q. Bui: Resources, formal analysis, supervision, writing–review and editing. S.C. Bendall: Resources, supervision, funding acquisition, writing–review and editing. G.R. Crabtree: Conceptualization, resources, supervision, writing–original draft, writing–review and editing. G.R. Bean: Resources, data curation, supervision, writing–review and editing. C. Van Rechem: Conceptualization, formal analysis, supervision, writing–original draft, writing–review and editing.

C. Van Rechem was supported by the Stanford Women's Cancer Center and the Stanford Cancer Institute, the Weintz Family COVID-19 Research Fund, the Jacob Churg Award, and the Children's Cancer Research Fund. G.R. Crabtree was supported by the Howard Hughes Medical Institute, grants from the NIH (grant nos. CA163915 and NS046789), and a grant from the Department of Defense, Breast Cancer Research fund. C.M. Weber was supported by a grant from GSK, Sir James Black Fellowship. S.C. Kimmey and S.C. Bendall received support from the NIH (grant no. T32GM007276 to S.C. Kimmey; grant nos. 1DP2OD022550-01, 1R01AG056287-01, 1R01AG057915-01, and 1U24CA224309-01 to S.C. Bendall. This work was supported in part by NIH P30 CA124435 utilizing the Stanford Cancer Institute Proteomics/Mass Spectrometry Shared Resource, Vincent Coates Foundation Mass Spectrometry Laboratory (RRID:SCR_017801).

The authors thank Drs. Joshua Black and Johnathan Whetstine for helpful discussions, and Ellen Gomulia and Ivy Mangonon for assistance with SMARCA4 IHC.