Sensitivity of human lung adenocarcinoma cell lines to targeted inhibition of BET epigenetic signaling

William W. Lockwooda,1, Kreshnik Zejnullahua,b, James E. Bradnerc,d, and Harold Varmusa,1

aCancer Biology and Genetics Section, Cancer Genetics Branch, National Research Institute, National Institutes of Health, Bethesda, MD 20892; bHoward Hughes Medical Institute, Chevy Chase, MD 20815; cDepartment of Medical Oncology, Dana Farber Cancer Institute, Boston, MA 02215; and dDepartment of Medicine, Harvard Medical School, Boston, MA 02115

Edited by Peter K. Vogt, The Scripps Research Institute, La Jolla, CA, and approved October 4, 2012 (received for review July 13, 2012)

Bromodomain and extra terminal domain (BET) proteins function as provide potential therapeutic targets for modulating expres- epigenetic signaling factors that associate with acetylated sion programs associated with various human diseases. and facilitate transcription of target . Inhibitors targeting the Selective inhibitors of BET proteins have recently been de- activity of BET proteins have shown potent antiproliferative effects in veloped that competitively occupy the acetyl-binding pockets of hematological cancers through the suppression of c-MYC and down- the closely related bromodomains, resulting in release from active stream target genes. However, as the epigenetic landscape of a cell chromatin and the suppression of downstream signal transduction varies drastically depending on lineage, transcriptional coactivators events to RNA polymerase (8, 10). These compounds have shown such as BETs would be expected to have different targets in cancers potent inhibitory activity against a range of cell lines derived from derived from different cells of origin, and this may influence the hematological malignancies, including multiple myeloma (MM), acute myeloid leukemia (AML), Burkitt’s lymphoma (BL), and activity and mechanism of action of BET inhibitors. To test this hy- – fi pothesis, we treated a panel of lung adenocarcinoma (LAC) cell lines mixed-lineage leukemia (MLL) (8, 10 13). Importantly, the ef - with the BET inhibitor JQ1 and found that a subset is acutely suscep- cacy of these compounds has been attributed mainly to their tible to BET inhibition. In contrast to blood tumors, we show that LAC ability to suppress c-MYC (v-myc myelocytomatosis viral onco- gene homolog) expression and downstream transcriptional tar- cells are inhibited by JQ1 through a mechanism independent of c- gets, upon which these cancers are dependent for their sustained MYC down-regulation. Through gene expression profiling, we dis- growth (8, 13). c-MYC is known to contribute to the pathogenesis covered that the oncogenic transcription factor FOSL1 and its targets fi

of certain cancer types, so these unique ndings provide a thera- MEDICAL SCIENCES are suppressed by JQ1 in a dose-dependant manner. Knockdown of peutic avenue for pharmacologically inhibiting c-MYC. However, BRD4 also decreased FOSL1 levels, and inhibition of FOSL1 phenocop- the activity of BET inhibitors in other cancers, especially carci- ied the effects of JQ1 treatment, suggesting that loss of this transcrip- nomas, remains largely unexplored. tion factor may be partly responsible for the cytotoxic effects of BET The epigenetic landscape of a cell varies drastically depending inhibition in LAC cells, although ectopic expression of FOSL1 alone did on its lineage and differentiation state. Thus, it is reasonable to not rescue the phenotype. Together, these findings suggest that BET assume that chromatin binding factors such as BET proteins will inhibitors may be useful in solid tumors and that cell-lineage–specific have different transcriptional targets in cancers derived from dif- differences in transcriptional targets of BETs may influence the activ- ferent cells of origin, which in turn will influence the consequences ity of inhibitors of these proteins in different cancer types. and mechanism of action of BET inhibition. To test this hypothesis, we used lung adenocarcinoma (LAC) as a representative carci- drug development | chromatin noma and treated a large panel of cell lines with the BET inhibitor JQ1. We establish that a subset of these cell lines are acutely sus- hromatin remodeling is a key epigenetic mechanism for ceptible to BET inhibition and that, in contrast to hematological Cregulating gene expression that contributes to both normal malignancies, this sensitivity is not dependent on c-MYC down- cell phenotypes and cellular transformation. This dynamic pro- regulation. Furthermore, we found that the oncogenic transcrip- cess is controlled through the posttranslational modification of tion factor FOS-like antigen 1 (FOSL1) and its downstream ex- histones by distinct families of enzymes such as deace- pression targets are suppressed by JQ1 treatment and knockdown tylases, methytransferases, and lysine demethylases that of BRD4, implying that FOSL1 may play a role in drug response in fi add or remove functional groups at a variety of residues on LAC. Together, these ndings suggest that BET protein inhibitors histone tails (1). The resulting “marks” are then recognized by may be effective across a wide range of cancers, with the mecha- discrete classes of “reader” proteins that interpret the infor- nism of action dependent on the epigenetic status of the cell. mation and assemble into complexes that facilitate chromatin Results remodeling (1). This recognition step is critical in the regulation of gene expression, as the nuclear reader proteins recruit the Subset of LAC Cell Lines Is Sensitive to BET Protein Inhibition. We necessary factors that initiate mRNA transcription and support treated a panel of 19 lung cancer cell lines for 72 h with a range of concentrations of the cell-permeable, small-molecule bromo- mRNA elongation (2). fi Members of the bromodomain and extra terminal domain domain inhibitor JQ1, which displays high potency and speci city (BET) family of proteins (BRD2, BRD3, BRD4, and the testis- specific BRDT) function as important reader molecules that as- sociate with acetylated histones and govern the assembly of chro- Author contributions: W.W.L. and H.V. designed research; W.W.L. and K.Z. performed re- fi search; J.E.B. contributed new reagents/analytic tools; W.W.L. analyzed data; and W.W.L., matin complexes and transcription activators at speci cpromoter K.Z., and H.V. wrote the paper. sites (3). For example, bromodomain-containing protein 4 (BRD4) Conflict of interest statement: Dana Farber Cancer Institute has licensed drug-like deriv- recruits the positive transcription elongation factor b (P-TEFb) atives of JQ1 from the J.E.B. laboratory to Tensha Therapeutics for clinical development as complex to defined genomic locations in mitotic chromatin, pro- anticancer agents. moting the phosphorylation and activation of RNA Pol II (4). This article is a PNAS Direct Submission. Recently, BETs have been shown to control the expression of Freely available online through the PNAS open access option. fl numerous genes involved in cell cycle, cell growth, in ammation, 1To whom correspondence may be addressed. E-mail: [email protected] or harold. and cancer, suggesting that they function as epigenetic signaling [email protected]. fi proteins that regulate transcription from speci cpromotersin This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. a cell context-dependent manner (5–9). Thus, these proteins 1073/pnas.1216363109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1216363109 PNAS Early Edition | 1of6 Downloaded by guest on October 1, 2021 toward the acetyl-binding cavity of BET proteins (10). [Two 1F). Together, these data suggest that JQ1 treatment decreases nonsmall cell lung cancer cell lines used in this study are not proliferation, blocks cell cycle progression, and induces apoptosis classified as adenocarcinomas. H1155 is a generic nonsmall cell in a subset of lung adenocarcinoma cell lines that are sensitive to carcinoma, and H460 (see Figs. 1 and 3) is a large cell carci- inhibition of BET proteins. noma.] A subset of lung cancer cell lines (8/19–42%) demon- strated half maximal inhibitory concentration (IC50) values of JQ1 Inhibition Does Not Prompt MYC Down-Regulation. Based on the less than 5 μM (range 0.42–4.19 μM, median = 1.26 μM), which recent findings that JQ1 reduces levels of c-MYC in drug-sensi- A we deemed sensitive to JQ1 treatment (Fig. 1 ). All other cell tive cell lines derived from hematological malignancies, we next > μ lines had IC50 values 10 M, the maximum dose used in the determined if JQ1 suppresses expression of c-MYC in drug-sen- treatments. Although all sensitive cell lines were LACs, there sitive LAC cell lines. Comparison of basal c-MYC mRNA and was no apparent association between JQ1 sensitivity and known B protein levels in JQ1-sensitive and -insensitive cell lines revealed driver mutations in the cell lines (Fig. 1 ). To complement the fi c-MYC results from short-term treatments, we performed long-term a signi cant association between high expression and JQ1 sensitivity (Fig. S3 A and B). Furthermore, gene set enrichment colony-forming assays to determine if the inhibitory effects of fi fi JQ1 are sustained over time. Even at low JQ1 concentrations analysis (GSEA) (14) of microarray data identi ed signi cant up- (500 nM), proliferation of sensitive cell lines was severely regulation of c-MYC transcriptional targets in the sensitive cell inhibited (>90% approximate reduction in cell number), whereas lines, suggesting these cancers may be driven by c-MYC activa- the insensitive line showed little effect (Fig. 1C), confirming tion (Fig. S3C). Surprisingly, however, c-MYC mRNA levels ei- that some lung cancer cell lines are inherently sensitive to BET ther significantly increased or remained unchanged after JQ1 protein inhibition. treatment in the majority (6/8) of the sensitive lung cancer cell To characterize the physiological effects of JQ1, we performed lines (Fig. 2A). Of note, c-MYC transcript levels increased more flow cytometry of JQ1-exposed cells to determine the consequence than twofold in H23 cells, although this cell line is the most on cell cycle progression and apoptosis. A549 and H1975 cells, sensitive to JQ1. In contrast, consistent with previous reports (8), which were sensitive to the drug in the proliferation assays, dem- c-MYC levels were dramatically suppressed by JQ1 in the MM onstrated a pronounced decrease in the proportion of cells in S cell line RPMI-8226 (Fig. 2A). c-MYC protein levels, like mRNA phase, with a concurrent increase in cells in G0/G1 during treat- μ levels, were elevated or unaffected by JQ1 exposure in most lung ment with 1 M JQ1 for 24 h, whereas H460 cells showed no cancer cell lines (Fig. 2B). In addition, c-MYC protein levels were significant changes in cell cycle progression (Fig. 1D and Fig. S1). stable after long-term treatment and did not decrease when cells This pattern is consistent with previous studies that demonstrated underwent apoptosis as measured by cleaved poly (ADP-ribose) a critical role for the BET member BRD4 in the transition from C mitosis to G1 and is similar to the effects on cell cycle induced by polymerase 1 (PARP1) (Fig. 2 ). Lastly, there was a dose-de- JQ1 in MM and BL cell lines (4, 13). In addition to cell cycle arrest, pendent increase in c-MYC levels in H23 and H1975 cells, mir- treatment with modest levels (1 μM) of JQ1 also increased the rored by increasing PARP1 cleavage; in contrast, c-MYC levels number of cells undergoing apoptosis after 48 h, as measured by decreased with the appearance of cleaved PARP1 in RPMI-8226 annexin V staining and PARP cleavage in sensitive cell lines (Fig. 1 cells, as anticipated. Collectively, these findings suggest that c- E and F and Fig. S2). In contrast, no evidence of apoptosis was MYC down-regulation is not a requirement for JQ1-mediated observed in H460 cells at 48 h even at high JQ1 doses (5 μM) (Fig. inhibition in LAC.

Fig. 1. JQ1 treatment inhibits the growth of ABIC (µM) DMSO 500 nM JQ1 a subset of lung cancer cell lines. (A) Average JQ1 50 C 012345678910Cell Line Mutation IC50 (µM) H23 G12C IC50 values for the indicated 19 lung cancer cell H23 KRAS 0.42 ± 0.20 H820 H820 EGFRdelE746-E749/T790M 0.91 ± 0.16 lines. Cells were treated with increasing doses of H2087 H2087 BRAFL597V 0.98 ± 0.17 JQ1 for 72 h and the number of viable cells was H2122 H2122 KRASG12C 1.06 ± 0.14 H23 A549 A549 KRASG12S 1.45 ± 0.07 determined using Alamar Blue. Red bars signify H2009 H2009 KRASG12A 1.63 ± 0.43 < μ HCC2279 HCC2279 EGFRdelE746-A750 2.65 ± 0.46 sensitive cell lines (IC50 values 5 M); blue bars H1975 H1975 EGFRL858R/T790M 4.19 ± 0.39 PC9 PC9 EGFRdelE746-A750 >10 mark the insensitive cell lines (IC50 values >10 μM). H1648 H1648 - >10 A549 Error bars denote the SDs of independent experi- H358 H358 KRASG13C >10 H2030 H2030 KRASG12C >10 ments. (B) Putative driver mutations and JQ1 IC50 Line Cell HCC827 HCC827 EGFRdelE746-A750 >10 values for the 19 cell lines tested. Values are pre- H1155 H1155 KRASQ61H >10 H1650 H1650 EGFRdelE746-A750 >10 sented as averages ± SD of two to four experi- H460 H460 KRASQ61H >10 H3255 H3255 EGFRL858R >10 ments. (C) Long-term colony formation assay of cell HCC2935 HCC2935 EGFRdelE746-S752, I ins >10 H1650 lines deemed sensitive (H23 and A549) and in- HCC4006 HCC4006 EGFRdelL747-A750, P ins >10 sensitive (H1650) in the short-term viability assays. Cells (2,000–8,000 per six-well plate) were grown in DEDMSO 1µM JQ1 H1975 FH1975 Q2 Q3 the absence or presence of 500 nM of JQ1 for 10 d, 150 48.3 ± 0.1 200 29.8 ± 5.9 105 24 hrs 48 hrs AV+ = 12.4 ± 2.5 JQ1 (µM ): 015015 stained with crystal violet, and photographed. (D) 150 104 Cell cycle analysis of H23 (sensitive) and H460 (in- 100 PARP 3 *Cleaved PARP μ 100 10

sensitive) cells treated with 1 M JQ1 for 24 h. BrdU DMSO

H1975 50 incorporation (represented by the proportion of 50 102 GAPDH 0 0 0 Q1 Q4 cells stained with an anti-BrdU antibody labeled 1 2 3 4 1 2 3 4 2 3 4 5 fl 10 10 10 10 10 10 10 10 010 10 10 10 Count with APC uorescent dye) indicates cells in S phase. 7-AAD 5 Q2 Q3 H460 200 41.3 ± 7.5 200 39.9 ± 5.0 10 Numbers in red represent the average percentage AV+ = 29.6 ± 1.6 24 hrs 48 hrs 4 of cells in S phase ± SD (n = 2). (E) Induction of 150 150 10 JQ1 (µM ): 015015 JQ1 apoptosis by JQ1 in H1975 cells. Cells were treated 100 100 103 PARP µM μ fl H460H460 *Cleaved PARP with 1 M JQ1 for 48 h and ow cytometry was 2 50 50 1 10 performed using 7-aminoactinomycin D (7-AAD) 0 0 0 Q1 Q4 GAPDH and annexin V staining. Values represent the av- 101 102 103 104 101 102 103 104 0102 103 104 105 + APC-BRDU PE-Annexin V erage percentage of annexin V cells (early (Q4) + late (Q3) apoptotic) ± SD (n = 2). (F) Induction of cleaved PARP1 in JQ1 sensitive (H1975) but not insensitive (H460) cell lines. Cells were treated with the indicated concentrations of JQ1 for 24 and 48 h and assessed for full-length and cleaved PARP1 protein levels by Western blot using an anti-PARP antibody. GAPDH serves as a loading control.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1216363109 Lockwood et al. Downloaded by guest on October 1, 2021 BET Protein Inhibition Represses FOSL1 and Its Targets in LAC Cells. inhibition. Genes repressed by JQ1 treatment were significantly Because c-MYC was not suppressed by JQ1 in drug-sensitive enriched for activator protein 1 (AP-1) DNA binding motifs; these LAC cells, the phenotypic effects of BET protein inhibition are serve as targets for the heterodimeric AP-1 transcription factor likely to be mediated by different transcriptional targets in this complex formed by FOS and JUN protein family members (Fig. type of tumor. To seek the relevant targets, we performed gene 3D and Fig. S4) (15). Thus, both IPA and GSEA demonstrated expression profiling of two sensitive (H23 and H1975) and one a significant enrichment for FOS targets within the JQ1-regulated insensitive (H460) cell line treated with JQ1. To increase the gene set. In contrast, binding motifs for the other transcription likelihood of identifying direct effects of JQ1, we collected sam- factors predicted by IPA to be associated with JQ1 response ples after just 6 h of treatment. The insensitive cell line was used (EGR1, HIC1, and GFI1) did not significantly overlap with the to identify changes in gene expression that did not inhibit growth. gene signature determined by GSEA (Fig. S4). Lastly, IPA or Comparison of significantly differentially expressed genes iden- GSEA did not identify c-MYC targets as being significantly re- tified a set of 298 annotated transcripts that are regulated by BET pressed upon JQ1 treatment, suggesting that BET inhibition is not inhibition exclusively in sensitive cells (Fig. 3A). Ranking these interfering with c-MYC transcription function in a manner in- transcripts by their differential expression score (defined in dependent of c-MYC down-regulation (Fig. 3C and Fig. S4). Materials and Methods) revealed numerous genes previously im- Gene expression analyses highlighted a potential role for the plicated in tumorigenesis that were strongly down-regulated by transcription factor FOS in mediating the response to JQ1 in LAC JQ1 treatment; these included HAS2, ILR7, MDM2, RUNX2, cell lines. Although FOS itself was not differentially expressed TRAF1,andFOSL1 (Fig. 3B and Dataset S1). upon JQ1 treatment, its closely related family member, FOSL1, To determine if the deregulation of a specific transcription was one of the most significantly down-regulated genes (Fig. 3B). factor could potentially explain the changes in gene expression Like FOS, FOSL1 forms dimers with JUN family members to induced by BET inhibition, we performed ingenuity pathway regulate AP-1 target genes (16), and FOS and FOSL1 are thought analysis (IPA) using the JQ1-affected genes. Four transcription to control similar sets of genes, because FOSL1 is able to replace FOS regulators were found to be significantly associated with the JQ1 in genetically altered animal models (16). Although both gene signature, with three predicted to be activated (EGR1, HIC1, proteins have previously been implicated in tumorigenesis, and GFI1) and one inhibited (FOS), based on whether their target FOSL1 is the main FOS family member linked to lung cancer genes were up- or down-regulated (Fig. 3C). To complement the (17). Analysis of RNA-Seq data from The Cancer Genome Atlas IPA analysis, we used GSEA to identify common transcription- revealed that it is the only FOS member commonly overexpressed factor binding sites associated with genes affected by BET in LAC (Fig. S5). Furthermore, the BET protein BRD4 is known to localize to the FOSL1 enhancer where it initiates transcrip- tional initiation and elongation (18). Thus, we predict that de- MEDICAL SCIENCES regulation of FOSL1 might be responsible, at least in part, for the 3.0 A B A549H23 H1975 H2087 H2122HCC2279 FOS gene expression signature induced by JQ1 treatment and JQ1: -++ -+- -++ -+- may represent a key target of BET inhibition in sensitive cell lines. *** MYC Consistent with this hypothesis, FOSL1 mRNA was repressed upon treatment of all JQ1-sensitive cell lines, with 6/8 reaching 2.0 * GAPDH statistical significance (Fig. 3E). In addition, FOSL1 protein levels decreased following BET inhibition in most drug-sensitive H1975 cell lines (Fig. 3F), whereas the insensitive cell line H460 showed C 24 hrs 48 hrs FOSL1 JQ1 (µM ): 015015 no appreciable effects on expression or protein levels 1.0 after JQ1 treatment (Fig. 3 E and F, respectively), consistent MYC fi fi *** with the ndings from the gene expression pro ling analysis (Fig. PARP A B *Cleaved PARP 3 and ). The MM cell line RPMI-8226 also showed decreased *** *** FOSL1 Expression Relative to DMSO Treated Control Treated DMSO to Relative Expression mRNA after treatment, although this did not reach sig- 0.0 GAPDH nificance. This contrasts with c-MYC expression levels (Fig. 3 E H23 A549 H820 F FOSL1 DMSO

H1975 H2122 H2087 H2009 and ). protein also diminished with increased duration

HCC2279 G c-MYC FOSL1

RPMI-8226 H23 H1975 RPMI-8226 of JQ1 treatment in H23 cells (Fig. 3 ); unlike D JQ1 (µM ): 0 0.25 0.5 1 2.5 5 0 0.25 0.5 1 2.5 5 0 0.25 0.5 1 2.5 5 protein appeared to decline in association with cleaved PARP1 MYC (Fig. 3G and Fig. S2), suggesting a functional link between FOSL1 suppression and the induction of apoptosis in this cell PARP FOSL1 *Cleaved PARP line. Lastly, there was a dose-dependent decrease in protein in H23 and A549 cells, demonstrating the dependency of GAPDH FOSL1 levels on BET activity. Together, these data suggest that FOSL1 is a direct target of BET inhibition in LAC cell lines. Fig. 2. Growth inhibition by JQ1 in lung adenocarcinoma cells is not de- pendent on c-MYC down-regulation. (A) Quantitative RT-PCR for c-MYC RNA FOSL1 Is a Transcriptional Target of BRD4 in LAC Cell Lines. BRD4 is levels in JQ1-treated cell lines. The MM control cell line RPMI-8226 (blue) known to localize to the FOSL1 enhancer where it recruits P- and sensitive lung cancer cell lines (red) were treated with 1 μM JQ1 for 6 h TEFb, leading to the activation of RNA Pol II and subsequent before RNA extraction and analysis. Data are presented as the average ratio transcriptional elongation (18). To confirm the requirement of of MYC expression for each cell line relative to its corresponding DMSO- BRD4 for FOSL1 transcription, and thus, demonstrate its via- treated control (mean ± SEM). Asterisks denote the level of statistical sig- bility as a target of BET inhibition in LAC, we reduced expres- nificance (*P < 0.05, **P < 0.01, ***P < 0.005; two-tailed t test). (B) c-MYC sion of BRD4 with siRNA in JQ1-sensitive cell lines and assayed protein levels in JQ1-treated sensitive lung cancer cell lines. Cells were FOSL1 − μ + for mRNA and protein levels. H23, A549, and H1975 treated with DMSO ( )or5 MJQ1( ) for 6 h before lysis and Western blot cells were treated with BRD4 siRNAs and reduced levels of analysis with an anti–c-MYC antibody. GAPDH serves as a loading control. BRD4 were confirmed by quantitative RT-PCR (Fig. 4A). BRD4 (C) c-MYC protein levels are stable over time in response to JQ1 treatment. fi FOSL1 H1975 cells were treated with the indicated doses of JQ1, then c-MYC knockdown signi cantly reduced levels of transcripts in all JQ1-sensitive cell lines analyzed, mimicking the effects of JQ1 and PARP levels were evaluated at 24 and 48 h. Cleavage of PARP occurs in A FOSL1 the absence of c-MYC down-regulation. (D) Dose-dependent effects of JQ1 treatment (Fig. 4 ). In addition, BRD4 siRNA suppressed treatment on c-MYC protein levels in sensitive lung cancer cell lines (red protein in these lines, whereas it had no effect on c-MYC protein bars) and the control MM cell line (blue bar). Cells were treated with the levels, similar to the results produced by JQ1 treatment in these indicated doses of JQ1 for 24 h before analysis. Again, cleavage of PARP cell lines (Fig. 4B and Fig. S6). Knockdown of the BET family occurs without c-MYC down-regulation. members BRD2 and BRD3 had no apparent effects on FOSL1

Lockwood et al. PNAS Early Edition | 3of6 Downloaded by guest on October 1, 2021 H460-Insensitive (895) H23 H1975 H460 AB CTranscription Predicted Regulation p-value of # of Target Regulator Activation z-score Overlap Molecules Target Molecules in Dataset JQ1: +- -++ - State in Dataset HAS2 CASP3, CLU, FOSL1, 259 ARL14 EGR1 Activated 2.668 0.00285 8 GADD45A, IGF1R, LDLR, SLITRK6 SOCS1, TOE1 GPR87 ADORA2B, CCNE2, NEXN HIC1 Activated 2.143 0.00446 5 RUNX2 CYP24A1, ITPR1, TRIM16 IL7R 152 148 LOC388022 ATF1, BCL3, CREBBP, MDM2 GFI1 Activated 2.012 0.000165 8 E2F5, F2R, IL7R, SMAD3, LOC728377 TNFRSF1A 336 SEMA4B ADNP, BCL9, CASP3, CLU, TRAF1 CREBBP, CTGF, CTH, FOSL1 FOSL1, HAS2, KCNN4, 16 TSKU FOS Inhibited -2.072 0.0094 MDM2, PTPN12, 1124 430 LYPD1 /// GPR39 RAB11FIP1, RARG, CLCF1 SLC19A1, ZHX2 298 SEMA4C E ADORA2B *** * MMACHC 2.5 MTL5 MYC SESN3 CCNE2 H23-Sensitive (1910) H1975-Sensitive (1212) 2.0 FOSL1 GCLC * HIST1H4H V$AP1_Q2_01 HIST2H2BF SLC6A8 /// SLC6A10P D 0.0 LOC93622 1.5 -0.1 SLC10A5 -0.2 ZNF14 HS6ST1 -0.3 PAG1 1.0 -0.4 HIST2H2BE TOB1 * *** -0.5 CTGF *** *** -0.6 Expression Score Differential HIST1H2BJ 0.5 *** DUSP1 Enrichment Score -0.7 C7orf53 *** ***** *** -0.8 ARRDC4 HIST1H1T Control Expression Relative to 0.0 OR2B6

Normalized Expression H23 A549 H820 H460 DMSO H1975 -1.5 +1.5 H2122 H2087 H2009 HCC2279 RPMI-8226 F G H23 H H23 A549 H1975 H2087 H2122 HCC2279 H460 24 hrs 48 hrs H23 A549 JQ1: -++ -+- -++ -+-+- JQ1 (μM ): 015015 JQ1 (μM ): 0 0.25 0.5 1 2.5 5 0 0.25 0.5 1 2.5 5

FOSL1 FOSL1 FOSL1

MYC MYC MYC

GAPDH GAPDH GAPDH

Fig. 3. JQ1 represses expression of FOSL1 and produces a downstream gene expression signature in drug-sensitive lung cancer cell lines. (A) Venn diagram depicting the overlap of significantly differentially expressed genes (Benjamini–Hochberg corrected P ≤ 0.01) after exposure to 1 μM JQ1 for 6 h in two sensitive (H23 and H1975) and one insensitive (H460) lung cancer cell lines. The red font highlights the number of genes differentially expressed in both sensitive cell lines but not the insensitive cell line. (B) Heatmap representation of the top-20 down-regulated (blue) and top-20 up-regulated (red) genes following JQ1 treatment in sensitive and insensitive lung cancer cell lines. Genes are ranked by the differential expression score, which is shown in the Left column (details in SI Materials and Methods). Data presented are mean normalized by row for each cell line. FOSL1 (arrow) is down-regulated by JQ1 treatment. (C) Ingenuity Pathway Analysis of transcription factor programs significantly deregulated (z score ± 2.0, P < 0.05) by JQ1 treatment of drug- sensitive lung cancer cell lines. The 298 genes highlighted in A, along with their differential expression scores, revealed that a gene expression program consistent with inhibition of FOS genes is induced by JQ1 treatment. The number of genes predicted to be regulated by each transcription factor program that are also deregulated by JQ1 treatment are presented along with the corresponding gene symbols. (D) Gene set enrichment analysis plot displaying the down-regulation of genes with AP-1 DNA binding motifs after JQ1 treatment in drug-sensitive cell lines. The 298 genes from A are ranked according to their differential expression score from highest to lowest along the x axis. The overrepresentation of genes with AP-1 sites (represented by the black lines) at the bottom of the ranked gene list suggests that there is a correlation between genes with this binding motif and JQ1 down-regulated genes. The green line represents the running enrichment score. Additional details are provided in Fig. S4 and SI Materials and Methods.(E) Quantitative RT-PCR for FOSL1 (red) and c-MYC (blue) RNA levels in JQ1-treated cell lines. Data are presented as the average ratio of each gene’s expression for each cell line, relative to corresponding DMSO-treated controls (mean ± SEM). All adenocarcinoma cell lines displayed are sensitive to JQ1 except H460. The MM cell line RPMI-8226 is also depicted. Asterisks denote the level of statistical significance (*P < 0.05, **P < 0.01, ***P < 0.005; two-tailed t test). (F) Analysis of FOSL1 and c-MYC protein levels in JQ1-treated sensitive (red) and insensitive (blue) lung cancer cell lines. Cells were treated with DMSO (−)or5μMJQ1(+) for 6 h before assay with anti-FOSL1 and anti–c-MYC antibodies as in Fig. 2. (G) FOSL1 protein levels diminish with the duration of JQ1 treatment. H23 cells were treated with the indicated doses of JQ1, and FOSL1 proteins were assessed at 24 and 48 h. c-MYC proteins are shown for comparison. (H) Dose-dependent effects of JQ1 treatment on FOSL1 protein levels in a drug-sensitive and a drug-resistant lung cancer cell line. c-MYC protein levels are shown for comparison. Cells were treated with the in- dicated doses for 6 h before analysis. GAPDH serves as a loading control for all Western blots.

mRNA or protein levels, implying that the decrease of FOSL1 used RNAi to inhibit FOSL1 in sensitive cell lines, because this upon JQ1 treatment is specifically due to inhibition of BRD4 should phenocopy the effects of BET protein knockdown and JQ1 (Fig. S6). treatment. H23, A549, and H1975 cells were treated with siRNAs for FOSL1 or BRD4, and knockdown was confirmed by quanti- JQ1-Sensitive LAC Cells Are Inhibited by FOSL1 Knockdown. The tative RT-PCR and Western blot (Fig. 4 A and B and Fig. S6). Cell sensitivity of a subset of LAC cell lines to BET inhibition, co- viability was assessed 72 h after knockdown, with nontargeting incident with down-regulation of FOSL1 in these cell lines, sug- (NonT) and KRAS or KIF11 siRNAs serving as negative and gested that FOSL1 might mediate the response to JQ1. positive controls, respectively. Knockdown of FOSL1 significantly Correspondingly, this would also mean that sensitive cell lines are reduced the viability of all three sensitive cell lines (average re- dependent on sustained FOSL1 expression for survival. The duction in viability = 35%), in agreement with a role for FOSL1 in finding that FOSL1 mRNA and protein levels were significantly mediating survival (Fig. 4C). Although FOSL1 siRNA inhibited higher in sensitive compared with insensitive cell lines could be the cells to a lesser degree than JQ1 treatment (average reduction interpreted to support this assumption (Fig. S7). Therefore, we in viability = 64%), the effects were similar to those observed by

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1216363109 Lockwood et al. Downloaded by guest on October 1, 2021 BRD4 knockdown (average reduction in viability = 40%) (Fig. protect cells from growth inhibition mediated by JQ1. Thus, al- 4C). Importantly, knockdown of BRD2 and BRD3 also inhibited though FOSL1 down-regulation likely plays a role in response to the viability of the sensitive cell lines, by 37 and 27%, respectively the BET inhibitor, it is probable that additional factors also (Fig. S6). Although JQ1 is selective toward binding of BET, and render LAC cells sensitive to inhibition of BET proteins. not other, bromodomains, it has nearly equal activity against BRD2, BRD3, and BRD4. Thus, the effects of JQ1 treatment on Discussion LAC cells may be due to the cumulative inhibition of all BET In this study, we show that a subset of LAC cell lines is highly family members and not a single protein alone. Coupled with the sensitive to the BET inhibitor JQ1, suggesting that such inhibitors finding that BRD4 is the only BET family member that appeared may provide a viable therapeutic strategy for treatment of solid as to control FOSL1 levels (Fig. S6), these data suggest that FOSL1 well as hematological tumors. Down-regulation of c-MYC and its may serve as a downstream factor mediating the BRD4-specific target genes was identified as the main mechanism mediating the effects of JQ1 inhibition and support the conclusion that down- antiproliferative effects of BET inhibition in leukemia and lym- regulation of FOSL1 is at least a partial cause of growth inhibition phoma (8, 11, 13). In these cancer types, BRD4 actively binds to c-MYC induced by BET inhibitors in LAC cells. the promoter and regulates signaling events leading to transcript elongation (13). In response to BET inhibitors, BRD4 is c-MYC Exogenous Expression of FOSL1 Is Not Sufficient to Rescue Cells from released from acetylated histones at the locus, reducing JQ1 Treatment. Next, we asked whether exogenous expression of gene expression (13). We found here that BET inhibition does not FOSL1 could rescue sensitive cell lines from the deleterious down-regulate c-MYC in sensitive lung cancer cells but instead effects of BET inhibition. Two sensitive cell lines (H23 and affects the expression of several other genes potentially involved in the drug response. This was surprising, especially considering the H1975) were stably transduced with retroviral vectors that express FOSL1 fi fact that sensitive LAC cell lines demonstrated many aspects of c- cDNA and expression was con rmed by Western blots c-MYC FOSL1 MYC dependency, including high mRNA and protein (Fig. S8). Exogenous expression of led to stable protein levels and up-regulation of c-MYC target genes (Fig. S3). How- levels in these cell lines even after JQ1 treatment, consistent with ever, as the epigenetic state of a cell is dramatically influenced by the expectation that BRD4 is not required for transcription from cell lineage, it is likely that BET and other chromatin binding the retroviral construct (Fig. S8). Cells were then treated with proteins have different transcriptional targets in different cancer JQ1 and assayed as in the initial screens described in Fig. 1 to types. Indeed, our results imply that c-MYC, although dependent determine sensitivity to BET protein inhibition. Surprisingly, cells on BET proteins for transcription in the hematological cancer cell overexpressing FOSL1 and control cells were equally sensitive to lines, is regulated by BET-independent mechanisms in LAC cells. JQ1 (Fig. S8), indicating that exogenous FOSL1 was unable to These findings have important implications, as they suggest that MEDICAL SCIENCES not all cancers dependent on c-MYC for survival will respond to BET inhibition. Instead, sensitive cancers from different tissue H23 A549 H1975 H1975 types will likely respond to BET inhibition through different ABFOSL1 fi 1.0 1.0 1.0 pathways. As a result, it may be dif cult to predict which patients BRD4 siRNA: NonT GAPDH MYC FOSL1 BRD4 will benefit from therapy with BET inhibitors. Thus, a strategy that * * FOSL1 considers cell lineage, genetic background, and epigenetic status ** will likely be needed to achieve optimal treatment response. Fur- 0.5 0.5 0.5 MYC thermore, it should also be noted that reduction of c-MYC is not Relative Expression Relative Expression Relative Relative Expression Relative solely responsible for the antiproliferative effects of BET in-

*** *** *** *** hibition in hematological cancers, as exogenous expression of c- *** *** GAPDH 0.0 0.0 0.0 MYC from a retroviral vector cannot fully rescue cells from JQ1 siRNA: siRNA: siRNA: NonT BRD4 NonT BRD4 NonT BRD4 treatment (8, 13). Also, in addition to potentially maintaining the FOSL1 FOSL1 FOSL1 C H23 A549 H1975 A549 expression of driver oncogenes, BET proteins serve many general 1.0 1.0 1.0 functions involved in regulating cell cycle progression (9), such as siRNA:

NonT GAPDH MYC FOSL1 BRD4 bookmarking genes for postmitotic reactivation (19). As a result, * ** *** ** the action of additional genes is likely, and the effects of BET ** * FOSL1 *** inhibition may be mediated through multiple factors, even in a 0.5 0.5 0.5 *** *** ** *** MYC single cancer type. Relative Viability Relative *** Viability Relative Viability Relative *** *** *** Down-regulation of the oncogenic transcription factor FOSL1 was identified as a potential cause of the antiproliferative effects 0.0 0.0 0.0 GAPDH of BET protein inhibition in LAC cells. FOSL1 mRNA and siRNA: KRAS NonT KRAS NonT KRAS NonT BRD4 BRD4 KIF11 BRD4 KIF11 KIF11 DMSO FOSL1 DMSO FOSL1 DMSO FOSL1 protein levels were reduced by JQ1, and genome-wide expression 1uM JQ1 1uM JQ1 1uM JQ1 profiling identified a significant enrichment of putative FOSL1 Fig. 4. FOSL1 knockdown phenocopies the effects of JQ1 treatment and targets within the set of JQ1-repressed genes. FOSL1 poses BRD4 knockdown in lung cancer cell lines. (A) Knockdown of BRD4 decreases a reasonable target for mediating the effects of BET protein the expression of FOSL1 in JQ1-sensitive cell lines. FOSL1 (blue) and BRD4 inhibition in LAC: numerous lines of evidence point to a key role (red) mRNA levels were assessed by quantitative RT-PCR 72 h after trans- for this transcription factor in lung tumorigenesis. For example, fection of the indicated siRNAs. Data presented are the average ratio of each FOSL1 is persistently activated by toxicants in cigarette smoke, gene’s expression relative to the levels observed in cells transfected with acts as a downstream transcription effector of signaling pathways nontargeting (NonT) siRNA (mean ± SEM, n = 3). (B) Knockdown of BRD4 commonly mutated in lung cancer, and induces lung epithelial decreases FOSL1 protein levels. H1975 and A549 were transfected with the cell invasion and anchorage-independent growth (17, 20). Fur- indicated siRNAs and FOSL1 and c-MYC levels were assessed by Western blot thermore, overexpression of FOSL1 but not other AP-1 com- after 72 h as described in Fig. 3. GAPDH serves as a loading control. BRD4 ponents, in transgenic mouse models leads to the development knockdown decreases FOSL1 but not c-MYC levels, mimicking JQ1 treatment. fi FOSL1 (C) Knockdown of FOSL1 and BRD4 decreases viability of JQ1-sensitive lung of lung bronchoalveolar tumors (21). Our nding that is cancer cell lines. Cell lines were transfected with the indicated siRNAs and overexpressed in a large proportion of human LAC tumors viability was assessed by Alamar Blue after 72 h. Data are presented as the further supports this conclusion. Lastly, FOSL1 is necessary for average viability of cells treated with each siRNA or JQ1, relative to cells KRAS- and EGFR-induced neoplastic transformation in differ- transfected with the nontargeting (NonT) siRNA or DMSO (mean ± SEM, ent model systems (22). Because many of the JQ1-sensitive LAC n = 3). KIF11 and KRAS siRNAs serve as positive controls. Asterisks denote lines have mutations in KRAS or EGFR, FOSL1 may play a vital the level of statistical significance (*P < 0.05, **P < 0.01, ***P < 0.005; role in maintaining the malignant phenotype of these cells. In- one-tailed, one-sample t test). deed, we found that many cell lines that are sensitive to JQ1

Lockwood et al. PNAS Early Edition | 5of6 Downloaded by guest on October 1, 2021 treatment have elevated FOSL1 levels and that FOSL1 knock- proteins are disrupted by exposure to BET inhibitors. Together, down inhibits their growth, suggesting that these cells are de- our findings, combined with those from previous studies, high- pendent on sustained FOSL1 for survival. However, as some of light the requirement for BRD4 in regulating FOSL1 gene the LAC cell lines most sensitive to JQ1, notably H23, express transcription. very low levels of FOSL1 whereas some that are insensitive to JQ1 The finding that exogenous FOSL1 did not rescue sensitive LAC have high FOSL1 mRNA and protein—as well as mutations in cell lines from the effects of JQ1 treatment might seem to con- EGFR and KRAS—the predictive value of these characteristics tradict our conclusion that this gene is involved in mediating the are clearly limited. Low or high levels of FOSL1 may sensitize drug response. We argue that this finding does not preclude a role LAC cell lines to JQ1, depending on the specificcellularcontext, for FOSL1. Because the knockdown of BRD2 and BRD3, in ad- such as the status of other FOS family members. Furthermore, dition to the knockdown of BRD4, inhibited the growth of drug- JQ1-induced down-regulation of FOSL1 was less common in in- sensitive lung cancer cells, it is likely that the cumulative inhibition sensitive lung cancer cell lines and, unlike sensitive cell lines, their of all BET proteins is responsible for the antiproliferative effects of growth was not significantly inhibited by FOSL1 knockdown (Fig. JQ1 in this tumor type. However, because down-regulation of S9), implying a possible correlation between FOSL1 dependency FOSL1 is attributed specifically to BRD4 inhibition, it is likely that and JQ1 sensitivity in lung adenocarcinoma. Collectively, our data exogenous FOSL1 cannot protect sensitive cells from the delete- show that FOSL1 is down-regulated by BET inhibition and that rious effects of BRD2 and BRD3 inactivation elicited by JQ1. In this event is associated with the decreased viability of drug-sen- addition, because FOSL1 needs to form a hetrodimer to regulate sitive LAC cells. Because FOSL1 has a critical role in lung cancer transcription from AP-1 DNA binding motifs (16), it is possible development, BET inhibitors may provide a therapeutic strategy that overexpressing FOSL1 alone may not be sufficient to rescue for targeting tumors dependent on this gene. FOSL1 target genes without coexpression of an appropriate binding Further support for as a target of BET inhibition is partner. Moreover, as JQ1 leads to the down-regulation of several provided by the observation that BRD4 is required for ex- HAS2 MDM2 FOSL1 genes with potential roles in tumorigenesis, such as , , pression of the gene (18). The mechanism by which this ILR7,andTRAF1, in addition to FOSL1, we cannot rule out the occurs is well established and is stimulated by the phosphory- possibility that other factors may be responsible for mediating lation of serine 10 in histone H3 (H3S10ph) at the FOSL1 the effects of BET inhibition in LAC. Indeed, a recent study on the enhancer by PIM1 kinase. This phosphorylation triggers a cas- fi cade of events leading to histone acetylation, BRD4 binding, effects of JQ1 in acute lymphoblastic leukemia identi ed IL7R as and subsequent enhancement of transcriptional elongation a critical target (24). Even so, our cumulative results support the (18). Thus, it is assumed that treatment with JQ1 would directly conclusion that loss of FOSL1 is at least partly responsible for the interfere with localization of BRD4 to the FOSL1 enhancer by cytotoxic effects of BET inhibition in LAC cells. disrupting the ability of its bromodomain to bind to the acety- Materials and Methods lated histone H3 induced by this signaling cascade. Consistent with this notion, we found that knockdown of BRD4 mimicked Cell lines were cultured under standard conditions in 96-well plates and the effect of JQ1 treatment, leading to decreased levels of treated with JQ1 for 72 h followed by the addition of Alamar Blue cell viability FOSL1 transcripts and proteins in sensitive lung cancer cell reagent for dose–response analysis. Cell cycle analysis, Annexin V staining, lines. Interestingly, c-MYC is also known to localize to the protein extraction, western blots, quantitative RT-PCR, gene expression fi FOSL1 enhancer where it is involved in recruiting PIM1 kinase pro ling, and siRNA transfections were performed using standard methods. and thus initiating the chain of events described above (23). A detailed description of the reagents, protocols, and computational anal- Because JQ1-sensitive LAC cells express high levels of c-MYC yses used in this study can be found in SI Materials and Methods. RNA and protein, in addition to FOSL1 gene products, and c- MYC FOSL1 ACKNOWLEDGMENTS. We thank Nancy Colburn and Matthew Young for and transcript levels are highly correlated across providing the pFB-FRA1 construct. This work was funded by the National lung cancer cell lines (Fig. S7), it will be interesting to explore Institutes of Health Intramural Research Program. W.W.L. is supported by the relationship between c-MYC, FOSL1, and BRD4 in this the Canadian Institutes of Health Research Jean-Francois Saint Denis Fellow- cancer type and investigate how the dynamics among these ship in Cancer Research.

1. Kouzarides T (2007) Chromatin modifications and their function. Cell 128(4):693–705. 14. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based ap- 2. Arrowsmith CH, Bountra C, Fish PV, Lee K, Schapira M (2012) Epigenetic protein proach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 102 families: A new frontier for drug discovery. Nat Rev Drug Discov 11(5):384–400. (43):15545–15550. 3. Filippakopoulos P, et al. (2012) Histone recognition and large-scale structural analysis 15. Eferl R, Wagner EF (2003) AP-1: A double-edged sword in tumorigenesis. Nat Rev of the human bromodomain family. Cell 149(1):214–231. Cancer 3(11):859–868. 4. Yang Z, He N, Zhou Q (2008) Brd4 recruits P-TEFb to at late mitosis 16. Verde P, Casalino L, Talotta F, Yaniv M, Weitzman JB (2007) Deciphering AP-1 to promote G1 gene expression and cell cycle progression. Mol Cell Biol 28(3): function in tumorigenesis: Fra-ternizing on target promoters. Cell Cycle 6(21): 967–976. 2633–2639. 5. Dey A, Nishiyama A, Karpova T, McNally J, Ozato K (2009) Brd4 marks select genes on 17. Reddy SPM, Mossman BT (2002) Role and regulation of activator protein-1 in toxicant- mitotic chromatin and directs postmitotic transcription. Mol Biol Cell 20(23):4899–4909. induced responses of the lung. Am J Physiol Lung Cell Mol Physiol 283(6):L1161–L1178. 6. LeRoy G, Rickards B, Flint SJ (2008) The double bromodomain proteins Brd2 and Brd3 18. Zippo A, et al. (2009) Histone crosstalk between H3S10ph and H4K16ac generates couple histone acetylation to transcription. Mol Cell 30(1):51–60. a histone code that mediates transcription elongation. Cell 138(6):1122–1136. 7. Nicodeme E, et al. (2010) Suppression of inflammation by a synthetic histone mimic. 19. Zhao R, Nakamura T, Fu Y, Lazar Z, Spector DL (2011) Gene bookmarking accelerates Nature 468(7327):1119–1123. the kinetics of post-mitotic transcriptional re-activation. Nat Cell Biol 13(11):1295–1304. 8. Delmore JE, et al. (2011) BET bromodomain inhibition as a therapeutic strategy to 20. Adiseshaiah P, Lindner DJ, Kalvakolanu DV, Reddy SP (2007) FRA-1 proto-oncogene target c-Myc. Cell 146(6):904–917. induces lung epithelial cell invasion and anchorage-independent growth in vitro, but 9. Wu S-Y, Chiang C-M (2007) The double bromodomain-containing chromatin adaptor is insufficient to promote tumor growth in vivo. Cancer Res 67(13):6204–6211. Brd4 and transcriptional regulation. J Biol Chem 282(18):13141–13145. 21. Jochum W, et al. (2000) Increased bone formation and osteosclerosis in mice over- 10. Filippakopoulos P, et al. (2010) Selective inhibition of BET bromodomains. Nature 468 expressing the transcription factor Fra-1. Nat Med 6(9):980–984. (7327):1067–1073. 22. Milde-Langosch K (2005) The Fos family of transcription factors and their role in tu- 11. Zuber J, et al. (2011) RNAi screen identifies Brd4 as a therapeutic target in acute mourigenesis. Eur J Cancer 41(16):2449–2461. myeloid leukaemia. Nature 478(7370):524–528. 23. Zippo A, De Robertis A, Serafini R, Oliviero S (2007) PIM1-dependent phosphorylation 12. Dawson MA, et al. (2011) Inhibition of BET recruitment to chromatin as an effective of histone H3 at serine 10 is required for MYC-dependent transcriptional activation treatment for MLL-fusion leukaemia. Nature 478(7370):529–533. and oncogenic transformation. Nat Cell Biol 9(8):932–944. 13. Mertz JA, et al. (2011) Targeting MYC dependence in cancer by inhibiting BET bro- 24. Ott CJ, et al. (2012) BET bromodomain inhibition targets both c-MYC and IL7R in high- modomains. Proc Natl Acad Sci USA 108(40):16669–16674. risk acute lymphoblastic leukemia. Blood 120:2843–2852.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1216363109 Lockwood et al. Downloaded by guest on October 1, 2021