SWI/SNF-Compromised Cancers Are Susceptible to Bromodomain Inhibitors

A growing number of potent, selective small-molecule compounds targeting epigenetic enzymes and transcriptional regulators have been developed, including inhibitors targeting bromodomain and extraterminal motif containing proteins (BETi). This class of bromodomain proteins includes BRD2, BRD3, BRD4, and BRDT (1). Of these, BRD4 has been the most extensively studied, and a wealth of evidence has been acquired, revealing it to be an important therapeutic target (2, 3). BRD4 is considered as a “reader” of lysine acetylation that specifically binds to acetylated histone tails, thereby stimulating the recruitment of transcriptional machinery to promoter and/or enhancer regions of target genes most commonly implicated in tumorigenesis (4). Preclinical studies in diverse cancer types point to BRD4 as the primary target of most BETi (5, 6). However, these inhibitors generally bind other BET family members as well. Importantly, a number of BETi have been proven highly selective, showing minimal off-target binding to bromodomain-containing proteins outside the BET family (5, 7). BETi act as acetyl-histone mimetic compounds that disrupt BRD4 function by competitively inhibiting its binding to chromatin, which, in turn, inactivates transcription of proliferation-related gene networks. In particular, the oncogenic c-MYC network is known to be repressed after exposure to BETi (8). However, not all responses to BETi are dependent on c-MYC repression, and downregulation of additional pathways, such as the PI3K signal transduction cascade, undoubtedly mediates the antiproliferative effects of BETi in some tissues (9, 10).

Selective inhibitors of BET proteins, such as JQ1, OTX015, and I-BET762, have shown anticancer activity against a plethora of tumor types, both in vitro and in murine tumor models of NUT midline carcinoma (NMC), multiple myeloma, myeloid leukemia, pancreatic, lung and breast cancers, among others (5, 6, 11–13). However, the doses of BETi used in these studies were generally quite high, ranging from 500 nmol/L to 1 μmol/L for in vitro experiments and using a typical dose of 50 mg/kg or more per day in vivo against solid tumors. To identify patient cohorts likely to show durable clinical response to BETi, it is likely that biomarkers predicting substantially lower doses will be required. This is especially important considering that numerous trials are currently under way against both solid and hematopoietic tumors, and results thus far have not met expectations (14–16).

Preclinical studies have focused primarily on c-MYC downregulation as the principal mediator of the antiproliferative effects of BETi. In some cases, intrinsic resistance to BETi may be conferred through ectopic expression of c-MYC (17). Although encompassing a limited cohort of patients, preliminary clinical trials have not yet substantiated elevated c-MYC as a strong predictive biomarker of BETi activity (14, 15). Preclinical data clearly show that inhibition of c-MYC–independent targets of BETi may also drive growth arrest. Interestingly, these include signaling pathways such as phosphatidylinositol-3 kinase (PI3K) or Hedgehog signaling, which may mediate the antiproliferative effects of BETi in a tissue-specific manner (9, 10, 18, 19). To date, biomarkers predicting response to BETi are still lacking, but their identity would obviously enhance the therapeutic potential of this class of drugs. Beyond acting as single-agent, the full potential of BETi as anticancer agents may not be realized until their utility as combination therapies is fully explored. In particular, recent data indicate BETi in combination with PARP or MAPK inhibitors hold great promise (20, 21).

SMARCA4 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a, member 4; also known as BRG1) is a core catalytic component of the multisubunit SWI/SNF (SWItch/sucrose nonfermentable) chromatin remodeling complex (22). Using energy generated through ATP hydrolysis, SMARCA4 shifts or evicts nucleosomes, generally leading to transcriptional activation across a spectrum of genes. SWI/SNF complexes may utilize either SMARCA4 or another highly homologous ATPase, SMARCA2 (also known as BRM), in a mutually exclusive manner. Physiologically, these two ATPases are often expressed in a mutually exclusive manner across tissues, but in cancer their relationship is more complex. SMARCA4/A2 may coregulate target genes, and each subunit has been demonstrated to compensate for the loss of the other (23). In cancer, SMARCA4 plays a Janus-like role, acting most commonly as a tumor suppressor, but also as a context-specific oncogene (24, 25). The role of SMARCA4 as a tumor suppressor has been widely studied, and mice heterozygous for the gene are predisposed to tumors of the mammary gland and lung (26, 27). SMARCA4 deletion or loss-of-function mutations are thought to promote tumorigenesis through complementary mechanisms including metabolic reprogramming and an incapacity to evict Polycomb repressors from target genes (28, 29).

Clinical data also support this tumor suppressor role of SMARCA4. SMARCA4 protein is lost in almost 100% of small cell carcinoma of the ovary, hypercalcemic type (SCCOHT), generally because of loss-of-function mutations (30). SCCOHT is a deadly cancer with an overall 5-year survival of only 16% (41 patients out of 257), with slightly better outcomes expected for disease diagnosed at early stages (31). IHC staining shows a complete loss of SMARCA4 protein in nearly 100% of SCCOHT. Notably, SMARCA2 was observed concurrently lost in 100% of SCCOHT, albeit in a limit cohort of 46 patients (32). A similar pathology was observed for a subset of NSCLC patients where ∼15% show a complete loss of SMARCA4 and a concurrent loss of SMARCA2 in 10% of NSCLC (33), again based on a small cohort of 41 tumors. This same study indicated that the concurrent loss of these two chromatin remodelers is associated with a poor clinical outcome. SMARCA4/A2-deficient tumors, especially SCCOHT and NSCLC, have proved challenging to treat, and these cancers respond weakly to the current standard of care. For SCCOHT, treatment generally involves surgery and adjuvant chemotherapy, most commonly platinum-based agents (31, 34). Emerging evidence suggests that high-dose chemotherapy followed by autologous stem cell rescue prolongs the survival of SCCOHT patients (31). Similarly, for NSCLC beyond early stages, radical surgery, followed by platinum-based chemotherapy and/or radiotherapy, is provided. But for both diseases, there is a dire need for new therapeutic approaches.

In contrast to its role as a tumor suppressor, SMARCA4 is also a known oncogene and is overexpressed in several cancers. It has been suggested that a switch in SMARCA4 function takes place during the course of tumor progression, allowing it to promote cell proliferation, and possibly confer drug resistance (25, 35). It is currently unclear what molecular events trigger this switch. Mechanistically, there is evidence that SMARCA4 acts as a coactivator for oncogenic transcription factors such as MITF or ZEB1 (36, 37). The oncogenicity of SMARCA4 may further rely on the activation of survival and proliferation-related genes, such as c-MYC, or activation of the PI3K pathway (38, 39). Importantly, in the case of c-MYC, it has been shown that SMARCA4 and BRD4 independently occupy distal enhancer elements and coactivate transcription in a redundant manner (38). Thus, it is reasonable to hypothesize that in SMARCA4-deficient cells, BRD4 is solely responsible for driving an oncogenic network that is otherwise controlled in a redundant fashion by SMARCA4 and BRD4. Consistent with this, a siRNA screen revealed that SMARCA4-mutant esophageal cancer models depend on BRD4 for survival (40). These data led us to test the hypothesis that SWI/SNF-compromised cancers may show sensitivity to BETi. We expected that this approach may be especially relevant to SCCOHT tumors, which carry a light mutational burden and are invariably characterized by the loss of SMARCA4 (41).

Here, we demonstrate that SMARCA4/A2-deficient SCCOHT and NSCLC models are highly sensitive to BETi in vitro at doses of BETi in the low nanomolar range, significantly lower than those previously reported for cells derived from solid tumors (5, 6, 12, 13). In contrast, mutation of SMARCB1 does not predict such sensitivity. Add-back of SMARCA4/A2 leads to acquired BETi resistance. Using an orthotopic xenograft model of SCCOHT, we also show hypersensitivity to BETi in vivo, again at doses well below those utilized to target solid tumors in previous reports. RNA-sequencing revealed that BETi represses the transcription of the receptor tyrosine kinase (RTK) ERBB3 (HER3) and downstream signaling events in BETi-sensitive cells lacking SMARCA4/A2. HER3 downregulation was found to be at least partially responsible for the antiproliferative effects of BETi in SCCOHT models. Overall, our findings suggest that BETi may represent a potential therapeutic strategy against aggressive SMARCA4/A2-deficient cancers.

The SCCOHT1 cell line was provided by Dr. Ralf Hass (Hannover Medical School, Hannover, Germany). OVCAR8, SKOV3, NCI-H522, NCI-H2228, and HEK293T were purchased from the ATCC. OVCAR4 and IGROV-1 cells were obtained from Dr. Edwin Wang (University of Calgary, Calgary, Alberta, Canada) and NCI-H358 was a gift from Dr. Moulay A. Alaoui-Jamali (McGill University, Montreal, Quebec, Canada), originally purchased from the ATCC. The OVK18 cell line was obtained from the RIKEN Cell Bank. The BIN67 cell line was provided by Dr. Barbara Vanderhyden (University of Ottawa, Ottawa, Ontario, Canada). Ovarian and lung cancer cells were cultured in the RPMI-1640 1× medium supplemented with 10% FBS. HEK293T cells were cultured in DMEM/10% FBS. BT12 and CHLA266 cells were provided by the Children's Oncology Group Cell Culture and Xenograft Repository (Texas Tech University Health Sciences Center) and cultured in Iscove's medium supplemented with 10% FBS, 4 mmol/L L-glutamine, 1× ITS (5 μg/mL insulin, 5 μg/mL transferrin, 5 ng/mL selenous acid). The cell lines were maintained in culture for no more than 10 passages from the initial stock (prepared from early passages). All cells were tested for Mycoplasma contamination by DAPI staining monthly and not otherwise authenticated.

Compounds used in the studies included OTX015 (MedChem Express, #HY-15743), JQ1 (MedChem Express, #HY-13030), Decitabine (Selleckchem, #S1200), GSK343 (Selleckchem #S7164), Vorinostat (Cayman Chemical, #10009929), and cisplatin (Accord, #DIN 02355183).

For cDNA expression of ERBB3 (HER3), ERBB3-pReciever-LV120 (GeneCopeia, #EX-M0854-Lv120) was used, along with the control vector pReciever-LV120 (GeneCopeia, #EX-EGFP-Lv120). Inducible SMARCA4 expression was achieved using the vector pInducer-20 carry full-length SMARCA4 cDNA, kindly provided by Dr. Jannik N. Andersen (The University of Texas MD Anderson Cancer Center, Houston, TX). Similarly, a SMARCA2-inducible plasmid was obtained by subcloning full-length SMARCA2 cDNA into pInducer-20 lentiviral vector (Addgene, #44012). shRNAs directed against BRD4, HER3, and SMARCA4 are held in vector TRC1.5/PLKO.1 and were obtained from Sigma (Supplementary Table S1).

Lentiviral particles were packaged in HEK293T cells as described previously (42). For viral transduction, 0.5 × 10⁶/mL of SCCOHT1, OVK18, or H522 cells were seeded in 10-cm petri dishes. To achieve stable ERBB3 (HER3), or inducible, ectopic expression of SMARCA4 or SMARCA2, cells were incubated with the viral suspension at ratios of 1 mL of virus for each 4 mL of cell media, with a final concentration of hexadimethrine bromide (polybrene) at 5 μg/mL. After 24 hours, the media were changed, and stable selection was carried out using appropriate selective antibiotic. ERBB3 (HER3) stable cell lines were obtained after selection with 0.5 μg/mL of puromycin, while 200 μg/mL of neomycin was used to generate SMARCA4- or SMARCA2-inducible cells. For BRD4 or ERBB3 (HER3) knockdown experiments, SCCOHT1 and OVCAR8 cells were plated in 6-well plates and transduced with lentiviral particles at increasing dilution ratios (1:50, 1:20, 1:10, 1:6). Expressing cells were elected for using puromycin. For SMARCA4 knockdown, OVCAR8 cells were incubated with shSMARCA4-containing lentivirus followed by selection with puromycin.

Ovarian or lung cancer cells were seeded in 24-well plates at 10⁴–10⁵ cells/mL and treated for a period of 5 days with the indicated concentrations of 5 compounds (JQ1, OTX015, GSK343, decitabine, and vorinostat). DMSO (0.01%) was used as a control. Culture media were changed each day. For washout experiments, SCCOHT1 and OVCAR8 cells were plated and treated with DMSO, cisplatin, or OTX015 for 5 days. At day 5, the drugs were washed out, and the cells were grown for additional 3 days (day 8 time point). For knockdown of HER3 or BRD4, cells were infected at increasing viral titer and analyzed at day 4 or 5, respectively, after transduction. For SMARCA4 or BRD4 knockdown, cells were exposed to viral particles at a ratio of 1 mL virus per 7 mL culture media, followed by selection using 0.5 μg/mL of puromycin. Next, treatments with BETi were conducted for 5 days. For ectopic expression of HER3, cells stably expressing ERBB3 were plated and treated with the indicated concentrations of BETi. To assess the BETi resistance of SCCOHT1 or H522 cells expressing SMARCA4 or SMARCA2, cells were first seeded in 24-well plates followed by treatment for 1 day with 1 μg/mL of doxycycline prior to the BETi exposure. For quantification of cell viability curves, crystal violet staining was measured as previously described (43).

For clonogenic assays, either 400–500 cells/mL (for fast-proliferating) or 2,500 cells/mL (for slow-proliferating) cells were seeded in each well of 6-well plates. The next day, treatments with BETi were initiated at 10 nmol/L to 100 nmol/L every 3 to 4 days until the cells reached 70% to 80% confluency. The plates were fixed with 4% formaldehyde/1× PBS and stained with 0.1% crystal violet/10% ethanol. A quantification step was performed by application of 10% acetic acid followed by absorbance measurements. Colony formation value was normalized to DMSO.

A total of 0.5 × 10⁶/mL cells were seeded in 6-well plates. Apoptosis analysis was performed with FITC Annexin V Apoptosis Detection kit I (BD Biosciences; cat. #556547), and cell-cycle analysis was assessed via propidium iodide staining as described previously (42). The samples were analyzed by flow cytometry at day 3 post-incubation with DMSO (0.01%), cisplatin (5 μmol/L), or OTX015 (50 nmol/L) and day 5 (washout of compounds at day 3).

Total RNA was extracted using Gene Elute Mammalian Total RNA kit (Sigma, #RTN350) according the manufacturer's procedure. RNA (1 μg) was subjected to reverse transcription carried out with 5× All-In-One RT MasterMix (ABM, #G490) following commercial instructions. The resultant cDNA was subsequently analyzed by qPCR. GoTaq qPCR Master Mix (Promega, #A6002) was used for qPCR reactions according to the manufacturer's protocol. PCR product was amplified with specific primers at 250 nmol/L per reaction (Supplementary Table S2). mRNA levels were normalized to human 36B4 expression.

Cells were exposed to 0.01% DMSO or 100 nmol/L OTX015 for 4 and 24 hours. After treatments, RNA extraction was performed as described above and 1 μg of RNA by replicate was sent to the “Genome Quebec Innovation Center” for RNA-sequencing, with each condition being sequenced in duplicate. mRNA-stranded library preparation was carried out using Illumina TruSeq rRNA-depleted stranded library preparation followed by paired-end 100 bp sequencing with a HiSeq 2500. Using Trimmomatic v0.32, low-quality bases (phred33 < 30) and adapters were removed, the first three bases were clipped and only reads with a minimum length of 35 bases, and in pairs, were kept for the next step. Clean paired sequences were aligned using STAR v2.3.0e, and then the number of reads per gene using UCSC hg19 annotation and featureCount was obtained and used for differential expression analysis with DESeq2. The selection of differentially expressed genes was based on fold change > 0.5 and FDR <0.1. The data are available on the GEO database (GEO accession #GSE102908). Heat maps and volcano plots were generated using XLSTAT 2014. Validation of ERBB3 expression was carried out by RT-qPCR as described above. The Gene Set Enrichment Analysis tool (GSEA; http://www.broad.mit.edu/gsea/) was used to determine the enrichment of the custom 18 gene signature "Receptor Tyrosine Kinase gene signatures." Fourteen of these genes are held within the Gene Ontology_Phosphorylation set (biological processes). The percentage of tumors with concurrently low SMARCA4/A2 was analyzed from The Cancer Genome Atlas (TCGA) data website by using the median expression to cut the data in low and high. For each cancer type, SMARCA2 and SMARCA4 expression levels under the median values were considered as low, and the percentage of patients with "low" expression for these two genes was computed across multiple cancer type. The total number of analyzed samples was 9144. A Kaplan–Meier plot was generated from TCGA data. The total number of patients studied was 239 from the LUAD gene expression (Illumina HiSeq) data set.

Cells were seeded in four, 15-cm culture plates and treated with 0.01% DMSO or 100 nmol/L of OTX015 for 24 hours. Chromatin immunoprecipitation (ChIP) was carried out as previously described (42). DNA was extracted by the QIAquick PCR Purification Kit (Qiagen) according to the manufacturer's instructions. The samples were analyzed by qPCR with the results being represented as a percentage of input. The list of primer sequences can be found in Supplementary Table S2. Further information is available to the reader upon request.

Whole-cell lysates were collected, and Western blotting analysis was performed precisely as previously described (44).

Animal studies were conducted in accordance with guidelines of the Canadian Council of Animal Care and approved by the Animal Resources Centre at McGill University. A total of 2 × 10⁷ of SCCOHT1 cells or 3 × 10⁶ of OVCAR8 cells in PBS were injected into the left ovary of 4-week-old female NOD/SCID mice. After 3 weeks for SCCOHT1 and 2 weeks for OVCAR8, mice were randomized into groups of 5. Next, treatments with 20 mg/kg/day of OTX015 dissolved in 2% DMSO, 30% PEG300, 5% Tween 20 or vehicle (same solvent) were carried out by oral gavage, 5 days per week. After 3 weeks of treatment, mice were sacrificed and tumors were collected and weighted. Tumors were measured with caliper, and tumor volume was calculated using the following formula: ½ × ((length in mm) × (width in mm)²). IHC procedures were carried out at the Segal Cancer Centre Research Pathology Facility (Jewish General Hospital) as previously described (45). Slides with tissue samples were incubated in 1:100 dilution of primary Ki-67 rabbit antibody. Sections were analyzed by conventional light microscopy. Each part of the tissue section was analyzed individually, and the average value was used as a final score. The intensity of staining was measured by ImageJ and was scored by assessing five-tiered ranking (0 = no staining, 1 = very weak staining, 2 = weak staining, 3 = intermediate staining, 4 = strong staining, 5 = very strong staining).

Antibodies for Western blotting were as follows: anti-SMARCA4 (Santa Cruz, #sc17796), anti-SMARCA2 (Abcam, #ab12165), anti-SMARCB1 (Bethyl, #A301-087A), anti-Actin (Sigma, #A5316), anti-Lamin A (Santa Cruz, #sc-20680), anti-Parp1/2 (Santa Cruz, #sc-7150), anti-c23-MS3-nucleolin (Santa Cruz, #sc8031), anti-BRD4 (Cell Signaling Technology, #13440s), anti-HER3 (Cell Signaling Technology, #12708s), anti-AKT (Cell Signaling Technology, #2920s), anti-pAKT (Cell Signaling Technology, #4060s), anti-S6 (Cell Signaling Technology, #2217s), anti-pS6 (Cell Signaling Technology, #2215s), anti-ERK1/2 (Cell Signaling Technology, #4695s), anti-pERK1/2 (Cell Signaling Technology, #9101s). For ChIP-qPCR, anti-RNA polymerase II subunit B1 phospho-CTD Ser-2 (Millipore, #04-1571) was used. Anti-Ki-67 (SP6; Abcam, #ab16667) was used for IHC analysis.

XLSTAT 2014 and SigmaPlot 12.0 were used for statistical analysis. All the error bars represent SEM. Significance of results was calculated with the Student t test. Data were compiled from at least three replicates and were considered significant by obtaining a P value of ≤0.05.

SMARCA4/A2 deficiency has been previously characterized in both SCCOHT and NSCLC (30, 32, 33). Based on median RNA-seq values from TCGA data sets, the molecular phenotype of concurrently low SMARCA4/A2 expression appears common across many cancers and may be associated with poor clinical outcomes (Fig. 1A and B; Supplementary Fig. S1A; refs. 30, 31, 46). We predicted that the absence of SMARCA4 may dictate cell sensitivity to BET inhibitors. This hypothesis was based on data from a siRNA screen indicating SMARCA4-mutant esophageal cancer models depend on BRD4 for survival (40) and other work suggesting the proliferation and survival of cancer cells may be controlled in some tumors by coregulation of an oncogenic network by BRD4 and SMARCA4 (38). Thus, in SMARCA4-deficient cells, inhibition of BRD4 with BETi might lead to a shutdown of these coregulated genes, culminating in a loss of proliferation. To test this hypothesis, we first used a panel of control cell lines expressing SMARCA4/A2, including the serous ovarian cancer lines OVCAR8, SKOV3, OVCAR4, and IGROV1, as well as the NSCLC lines H358 and H2228. As a comparison with these, we tested the SMARCA4/A2-deficient SCCOHT cell lines SCCOHT1, OVK18, BIN67, and the lung cancer cell line H522 (Fig. 1B; Supplementary Fig. S1B; Supplementary Table S3), each carrying mutations with the coding region of SMARCA4. Two widely used BETi, JQ1 (5) and the clinically relevant OTX015 (14, 15), were then tested for antiproliferative effects against these cells. We tested the relative sensitivity and resistance to BETi using short-term (5 day) cell viability assays (Fig. 1C; Supplementary Fig. S1C). Here, we see that SMARCA4/A2-deficient cells show IC₅₀ values between 50 and 100 nmol/L for both BETi. Complementing these data, long-term clonogenic assays using a range of drug concentrations revealed that SMARCA4/A2-deficient cells show hypersensitivity to both BETi at IC₅₀ concentrations ≤ 50 nmol/L (Fig. 1D; Supplementary Fig. S1D). These concentrations were substantially lower than thresholds previously established for BETi sensitivity in solid tumors (<750 nmol/L), including ovarian cancer (47, 48). BETi sensitivity was accompanied by an accumulation of sub-G₁ cells and positive staining for Annexin V, indicating BETi is inducing a proapoptotic, rather than a cytostatic, response (Supplementary Fig. S2A and S2B). Consistent with a cytotoxic response, cells continued to undergo apoptosis after cell washout, and cell numbers did not recover (Supplementary Fig. S2C).

In parallel with the testing of BETi efficacy, we undertook experiments to ensure the results observed with BETi could be closely recapitulated upon BRD4 knockdown. Using lentiviral-mediated delivery of two independent shRNAs targeting BRD4, we demonstrated that SCCOHT1 cells respond in a dose-dependent manner to BRD4 knockdown (Fig. 1E). Consistent with our BETi data, OVCAR8 cells were quite resistant to growth inhibition upon BRD4 knockdown. We further found that BETi exposure in BRD4 knockdown cells continued to decrease viability, suggesting BET proteins beyond BRD4 are being targeted by the compounds (Supplementary Fig. S2D).

SMARCA4-mutant SCCOHT tumors clinically and morphologically mirror atypical teratoid rhabdoid tumors (AT/RT; ref. 30). AT/RT cancers are characterized by inactivating mutations to the SWI/SNF subunit SMARCB1. We tested two such cell lines (BT12 and CHLA266) for sensitivity to BETi. In contrast to SCCOHT cells, AT/RT lines displayed a strong intrinsic resistance to BETi exposure (Supplementary Fig. S3A), suggesting that SMARCB1 deficiency confers oncogenic properties distinct from SMARCA4 loss.

Beyond BETi, we tested our ovarian and lung cell line panels with other epigenetic therapies, either FDA approved, or undergoing clinical evaluation. These included the DNA methylation inhibitor decitabine (5-Aza-2-deoxycytidine), the histone deacetylase inhibitor vorinostat and the EZH2 inhibitor GSK343. Over the course of our relatively short-term, 5-day assays, these inhibitors did not exert the robust antiproliferative effects against SMARCA4/A2-deficient SCCOHT1 and H522 cells at concentrations of 50–100 nmol/L as is seen with BETi (Supplementary Fig. S3B). These results underscore the selectivity of SMARCA4/A2-deficient tumor cells for BETi. Among the three additional epigenetic therapies tested, the most potent was vorinostat, consistent with a previous report indicating SCCOHT cells show responsiveness to HDACi in vitro (32).

To extend our in vitro data, we next aimed to test the efficacy of BETi against an orthotopic model of SCCOHT. Considering that OTX015 has been evaluated in numerous clinical trials and demonstrates favorable bioavailability characteristics in preclinical studies, we used this compound for our in vivo studies. Previous murine studies of SCCOHT exclusively used subcutaneous models. Here, to more closely mimic the tumor microenvironment and perhaps to better predict future, clinical drug responses, we developed orthotopic ovarian xenograft models of SCCOHT (SCCOHT1) and serous ovarian carcinoma (OVCAR8). Previous reports showing BETi efficacy against solid tumors in vivo generally administered 50 mg/kg/day or even higher doses (5, 13). However, due to the hypersensitivity of SMARCA4/A2-deficient tumors to BETi, we reasoned that this dose could be considerably reduced. Following tumor development at the ovary, mice were treated with vehicle control or OTX015 at doses of 20 mg/kg/day, by oral gavage, for a period of 3 weeks (Fig. 2A). OTX015 was well tolerated by mice at this dose without body weight loss (Supplementary Fig. S3C). Quantification of tumor growth showed BETi to act as an effective antineoplastic agent against SCCOHT in vivo, consistent with our in vitro results described above. Although SMARCA4/A2-deficient SCCOHT tumors were highly susceptible to OTX015-mediated growth inhibition, OVCAR8 tumors were unresponsive (Fig. 2A). BETi reduced tumor weight and volume by almost 80% across all of the SCCOHT tumors. No antitumor response was observed at this dose in any of the tumors arising from OVCAR8 cells. Consistent with our macroscopic tumor measures, OTX015 nearly ablated the expression of the classic marker of cell proliferation Ki-67 in SCCOHT tumors (Fig. 2B). In contrast, no significant changes were observed in OVCAR8 tumors, again indicating a complete lack of response at the doses suitable for treating SCCOHT tumors.

At this point, our data supported the hypothesis that SMARCA4/A2-deficient cells rely on BET family members for survival, and that the loss of SMARCA4 sensitizes cells to BRD4 inhibition. Therefore, we next further interrogated this hypothesis using complementary approaches of SMARCA4 reexpression and SMARCA4 knockdown. If our model is correct, then ectopic expression of SMARCA4 in deficient cells should impart BETi resistance. Conversely, knockdown of SMARCA4 in resistant cells might lead to inhibitor sensitization. First, we restored wild-type SMARCA4 via a doxycycline-inducible lentiviral vector system in ovarian SCCOHT1 cells and the lung cancer cell line H522 (Fig. 3A). SMARCA4 restoration rendered both cell lines significantly more resistant to both JQ1 and OTX015 across a range of concentrations. Complementary experiments using SMARCA4 knockdown also supported the concept that this protein is partially responsible for intrinsic cell resistance to BETi (Fig. 3B). Here, SMARCA4 knockdown was achieved using a lentiviral system to deliver two independent shRNAs targeting SMARCA4 to the BETi-resistant cell line OVCAR8. Subsequent viability analysis found that SMARCA4 depletion altered the response of OVCAR8 to JQ1 and OTX015, allowing a dose-dependent reduction of cell survival. These data have several implications. First, there is clear indication that SMARCA4 expression confers resistance to BETi. Second, our work shows that reexpression of SMARCA4 partially prohibits the antiproliferative effects of BRD4 inhibition. These data support previous work characterizing coordinated gene expression by BRD4 and SMARCA4 to promote proliferation and survival (38, 40). The data further suggest that small-molecule inhibitors of SMARCA4, if developed, might act in synergy with BETi.

The SWI/SNF chromatin remodeling complex may integrate SMARCA4 or SMARCA2 in a mutually exclusive manner. However, as mentioned above, evidence demonstrates that these remodelers are capable of compensating for the loss of the other ATPase in a context-dependent manner, indicating a degree of functional redundancy (23). Thus, it is not surprising that a dual repression of SMARCA4/A2 is commonly observed in a number of cancers (Fig. 1A; refs. 32, 46, 49). We noted that the cells showing acute sensitivity to BETi have concurrent SMARCA4/A2 loss (Figs. 1B and 3A). Therefore, we wanted to examine whether SMARCA2 reexpression might mediate resistance to BETi in a similar fashion as SMARCA4. Using the same inducible approach as described above for SMARCA4 restoration, we transduced SCCOHT1 with SMARCA2 and tested for the sensitivity to BETi (Fig. 3C). SMARCA2 restoration rendered cells considerably less responsive to the two bromodomain inhibitors in comparison with the parental controls. Our findings indicate that SWI/SNF activity modulates the response to BETi and that concurrent low expression of both SMARCA4 and SMARCA2 may act as a predictive biomarker for tumor sensitivity to BETi.

BET family members are key activators of genes controlling proliferation and survival. Therefore, using RNA-seq we aimed to identify potential BRD4 target genes whose repression mediates the antiproliferative effect of BETi in SMARCA4/A2-deficient cells. Until now, transcriptomic analysis of BETi targets in cells derived from solid tumors have primarily utilized drug doses ≥ 500 nmol/L or even micromolar ranges (17, 50). Due to the exquisite sensitivity of SMARCA4/A2-compromised cells to BETi, we used treatments of 100 nmol/L to distinguish gene-expression profiles between SCCOHT1- and BETi-insensitive OVCAR8 cells. Such exposures for 4 and 24 hours revealed a distinct transcriptional profile imposed by BETi in SCCOHT1 cells that was absent in OVCAR8, highlighting the differential response of these cell lines to 100 nmol/L OTX015 (Fig. 4A and B; Supplementary Table S4; GEO accession #GSE102908). Analysis of the most strongly 150 downregulated genes from our RNA-seq data (>2-fold decrease) using Kegg pathway analysis revealed a network of genes potentially regulating signal transduction through the PI3K and RAS pathways [Kegg pathway, PI3K–AKT, P < 0.005; MAPK, P < 0.006; GSEA RTK gene set (14 of 18 genes within this custom set are found within the Gene Ontology_Phosphorylation set (biological processes); Supplementary Fig. S4A and S4B]. Notably, the oncogenic RTK ERBB3 (HER3 protein) was strongly suppressed by OTX015 in SCCOHT1 cells (Fig. 4A and B, P ≤ 0.05). HER3 is a key initiator of signal transduction through the PI3K and RAS signaling cascades (51). Thus, suppression of ERBB3 levels represented a rational candidate for a BRD4 target gene whose repression might be responsible for mediating some of the antiproliferative effects of BETi. Therefore, we further explored the regulation of ERBB3 by BETi. We validated our RNA-seq data by qPCR (Fig. 4C, top) and observed a significant downregulation of ERBB3 in SCCOHT1 cells exposed to BETi in both time- and dose-dependent manners. In contrast, ERBB3 mRNA levels in OVCAR8 were largely invariant across the range of conditions tested, and the changes that could be observed at 4 hours were quite transient. Similar results were observed in the H522, BETi-sensitive lung cancer cells, compared with the resistant line H358 (Supplementary Fig. S4C). ERBB3 was considerably downregulated in SCCOHT1 tumors exposed to OTX015, but this regulation was not observed in OVCAR8 tumors (Fig. 4C, bottom). These data suggest that ERBB3 repression may mediate a portion of the antiproliferative effects of BETi in acutely sensitive cells. Consistent with this, knockdown of HER3 led to a significant decrease in the proliferation of SCCOHT1 cells, but a far more muted response was seen in SMARCA4/A2-expressing OVCAR8 cells (Fig. 4D).

The strong repression of ERBB3 by BETi indicates the gene is a direct target of BET family members. BRD4 activates gene expression through facilitating the recruitment of an elongation complex that phosphorylates RNA POL II on serine 2 of its C-terminal domain. Consistent with ERBB3 being a direct transcriptional target of BRD4, we detect multiple BRD4 binding sites are observed flanking the ERBB3 gene (Supplementary Fig. S4D). Further, using ChIP with an antiphosphoserine2-POL II antibody, we see that OTX015 mitigates the serine 2 phosphorylation of RNA POL II at ERBB3 in SCCOHT1 cells (Fig. 4E). Exposure to 100 nmol/L OTX015 effectively reduced the enrichment of phosphoserine2-POL II throughout the gene body of ERBB3 from +200 bp relative to the transcription start site, through +23 kb, near the 3′ terminus. Again, these data support the conclusion that the reduction of ERBB3 mRNA seen after BETi exposure results from diminished transcriptional activity. Our ChIP data further revealed that BETi did not reduce the enrichment of RNA POL II serine 2 phosphorylation at ERBB3 in BETi-resistant OVCAR8 cells. Overall, the data suggest that a compensatory mechanism, possibly dependent on SMARCA4, maintains ERBB3 expression in OVCAR8 cells after inactivation of BRD4. It also suggests that the sensitivity of SCCOHT1 cells to BETi is reflective of the capacity of low doses of BETi to repress the transcription of proliferation-related genes, including ERBB3.

Further validating our transcriptional data, the reduction of HER3 (encoded by the ERBB3 gene) on the protein level was also confirmed by Western blotting in BETi-sensitive SCCOHT1, OVK18, and H522 cells (Fig. 5; Supplementary Fig. S4C). HER3 protein was diminished at concentrations of only 25 to 50 nmol/L in sensitive cells, but such a decrease was not observed in any of the resistant cell lines.

The HER3 protein is a critical mediator of survival and proliferation in a range of cancers, including ovarian cancer. These protumorigenic activities are carried out by signal transduction through both the PI3K/AKT and RAS/MAPK signaling axes (51). To explore whether these classic effector pathways are repressed by BETi, we exposed a panel of sensitive and resistant ovarian cell lines to OTX015 and evaluated phospho-S6 and phospho-AKT from the PI3K pathway, and phospho-ERK1/2 from the RAS/MAPK pathway. In SCCOHT1 cells, we detected a significant decrease of phospho-AKT and phospho-S6 in response to OTX015 that was not observed in the three resistant ovarian cell lines tested (OVCAR8, SKOV3, and IGROV1; Fig. 5). Likewise, we also observed a dose-dependent reduction in phospho-ERK1/2 in sensitive cells that was not apparent in resistant cells. In BETi-sensitive OVK18 cells we also observed a reduction in these phosphorylation events. However, the reduction in phospho-ERK1/2 was considerably more pronounced, indicating repression of MAPK signaling may be critical for the antiproliferative effects of BETi in these cells.

A majority of the reports investigating the role of BRD4 in cancer progression have focused on c-MYC as a principal target mediating these effects. However, we saw little effect on c-MYC mRNA levels over a 24-hour period after BETi exposure in SCCOHT1 cells (Supplementary Fig. S5A, GEO accession #GSE102908). Our data suggest that BET family members may also regulate PI3K and MAPK signaling, in part, through upregulation of the HER3 oncogene. Conversely, the loss of BET proteins, such as BRD4, would be expected to dampen signaling through the HER3 axis. To validate that the repression of HER3 and downstream signaling cascades by BETi is directly dependent on BRD4, we next knocked down BRD4 via shRNA and again tested for HER3 expression and RTK-dependent signaling events. Consistent with our BETi data, depletion of BRD4 lowered HER3 protein levels and repressed signaling through both the PI3K and MAPK pathways in ovarian SCCOHT cells (Supplementary Fig. S5B).

Collectively, our data support a model where cells lacking SMARCA4/A2 are highly sensitive to BETi, in part, through inactivation of HER3 RTK signaling. If this model has merit, we expect that forced expression of HER3 would confer resistance to BETi. To test this concept, we generated stable expression of the ERBB3 gene in SCCOHT1 and OVK18 SMARCA4/A2-compromised cells, leading to elevated HER3 protein levels (Fig. 6). The cells were subsequently exposed to either JQ1 or OTX015 at increasing doses, and changes to cell viability were quantified. We found that maintaining the pool of HER3 through ectopic expression imparted the cells with partial, but significant, BETi resistance in both cell lines. Thus, our experiments reveal that HER3 are a key target of BETi whose downregulation significantly contributes to the hypersensitivity to BETi in SMARCA4/A2-deficient cancer cells.

All cancers, even those with comparatively low mutational burdens, such as SCCOHT, show deregulated transcription. Thus, targeting transcriptional processes, particularly through “epigenetic therapy,” offers an attractive approach to treat many cancers. Indeed, two such drugs, decitabine and vorinostat, are approved by the FDA. Although the panel of drugs targeting transcriptional and epigenetic processes has grown substantially in the past decade, and many clinical trials are ongoing, biomarkers predicting the efficacy of such drugs are sorely lacking. Included among the list of small molecules aimed at “drugging the genome” being tested in clinical trials, but without clear predictive biomarkers are the BET inhibitors.

Exciting preclinical studies demonstrating potent anticancer activity of BET inhibitors have prompted the rapid clinical development of this class of compounds. So far, phase I trials using BETi against myeloma, acute leukemia, and NMC have not met expectations, but complete responses have been reported (14, 15). These clinical trials have been unable to identify biomarkers predicting patient responses. Further studies of a longer duration may be required to determine whether c-MYC holds the same strength as a biomarker of BETi efficacy in a clinical setting as it does in some preclinical investigations. Other preclinical studies demonstrate that not all cells responding to BETi are reliant on the repression of c-MYC (52, 53), suggesting that BETi may regulate other key targets depending on the cancer type. Our findings reveal for the first time that SCCOHT and NSCLC cells with compromised SMARCA4/A2 show hypersensitivity to BETi. Importantly, the add-back of SMARCA4 or SMARCA2 leads to acquired resistance and SMARCA4 knockdown–sensitized cells to BETi, demonstrating the specificity of SMARCA4 as a predictive biomarker. Our in vivo work validated the exquisite sensitivity of SCCOHT cells that we observed in vitro. Again, using SCCOHT as a relevant in vivo model, we found that only 20 mg/kg of OTX015 delivered orally 5 days per week was sufficient to ablate tumor growth. To our knowledge, this is the lowest dose of any BETi that has shown efficacy against solid tumors in vivo. These data potentially have important clinical implications, but with several caveats.

Although our study indicates SMARCA4/A2-deficient cancers may be highly sensitive to BETi and warrants further clinical investigation, our data also reveal potential resistance mechanisms. First, we show that SMARCA2 may compensate for the loss of SMARCA4, as previously described. Based on this, we predict that BETi will show the highest degree of antitumor activity in a setting where both SMARCA4 and SMARCA2 are concurrently compromised by either mutation or loss of expression. In fact, SCCOHT patients are invariably characterized by concurrent loss of SMARCA4/A2 (32). Further, in NSCLC and prostate cancer, concurrent low expression of these proteins has also been reported (46, 49), as well as in numerous other cancers based on TCGA (Fig. 1A). Considering that SMARCA4/A2 compromised cancers are associated with poor clinical outcomes (Supplementary Fig. S1A), BETi may represent a new answer for an unmet clinical need.

The ERBB3 RTK acts as a potent oncogene mediating tumor growth and drug resistance whose expression is elevated in several types of cancers (51). We see that BETi represses ERBB3 mRNA production in a dose- and time-dependent fashion in SMARCA4/A2-deficient cells. This repression is concomitant with reduced signal transduction through the PI3K and MAPK pathways. These data are consistent with previous reports indicating BETi dampens signal transduction through these pathways (19, 47, 48). Mechanistically, our add-back studies confirm that HER3 repression is a key effector of BETi, and whose repression mediates at least a part, of their antiproliferative response. Therefore, our data suggest that the repression of oncogenic signaling observed in BETi-exposed cells is at least partially dependent on HER3 downregulation. Further, our data predict that aberrant activation of PI3K or MAPK signaling may promote resistance to the antitumor effects of BETi. This concept is indeed supported by previous findings. Knockdown of the tumor suppressor LKB1, leading to elevated PI3K signaling, confers resistance to JQ1 (13). Likewise, kinome analysis of ovarian cells selected for acquired resistance to chronic BETi exposure revealed increased signaling through the PI3K/RAS pathways (47). However, further work where constitutively active AKT or MAPK mutants are introduced into SMARCA4/A2-deficient cells will be required to solidify the hypothesis.

Although our data, and those of others, suggest that aberrant activation of PI3K/RAS signaling may impose resistance to BETi, this model also indicates that BETi in combination with HER3/PI3K/MAPK inhibitors represent a rational approach to overcome this resistance. In fact, multiple studies have shown a synergy between inhibitors of PI3K, and more recently MAPK, and BETi against a variety of cancers, but key questions remain (9, 19, 20). Currently, defined molecular pathologies predicting sensitivity to these combination therapies are lacking. Our data suggest SMARCA4/A2-deficient tumors with additional oncogenic hits activating the MAPK pathway may be especially sensitive to such combinatorial therapeutic intervention. Further, it is possible that combination therapy with BETi and drugs targeting the MAPK pathway might represent a more effective therapeutic avenue than BETi with PI3K inhibitors. Our work also indicates that therapeutic antibodies targeting HER3 may hold promise as combination therapy with BETi. Systematic comparison of such combinations will be of great interest and necessary to optimize the clinical efficacy of BETi.

Based on the work described herein and current literature, it is clear that repression of the PI3K and RAS signaling pathways contributes to the antiproliferative effects of BETi. Undoubtedly, ERBB3 is not the only gene responsible for facilitating the antiproliferative effects of BETi, and we predict further targets of BET family members are involved in modulating the activity of the PI3K and RAS signaling pathways. Our RNA-seq data indicate other potential targets include KRAS, BRAF, and upregulation of MAPK repressors such as RASA4 (Supplementary Fig. S4A). Future work will be required to assess whether modulation of these targets also influences the antiproliferative effects of BETi and whether BRD4 and SMARCA4 cooperate to control their transcription.

No potential conflicts of interest were disclosed.

Conception and design: T. Shorstova, M. Witcher

Development of methodology: T. Shorstova, J. Su, M.A. Alaoui-Jamali, M. Witcher

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Shorstova, J. Su, J. Johnston, S. Huang, M.A. Alaoui-Jamali

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Shorstova, M. Marques, J. Johnston, C.L. Kleinman

Writing, review, and/or revision of the manuscript: T. Shorstova, N. Hamel, S. Huang, M.A. Alaoui-Jamali, W.D. Foulkes, M. Witcher

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Shorstova, M. Witcher

Study supervision: W.D. Foulkes, M. Witcher

This study was supported by grants from the Canadian Institute for Health Research (#399061 and #399114) and the Quebec Breast Cancer Foundation to M. Witcher. This work was funded in part by the United States Army Medical Research Department of Defense, under the Ovarian Cancer Research program, contract #W81XWH-15-1-0497 to W.D. Foulkes and S. Huang, and the Canadian Cancer Society Research Institute (#702961) to W.D. Foulkes and M. Witcher. T. Shorstova is supported by a McGill Integrated Cancer Research Training Program fellowship. M. Marques is supported by a fellowship from the Fonds de Recherche du Québec- Santé (FRQS). M. Witcher is supported by a Chercheur Boursier award (#35128) from the FRQS.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.