BET inhibitors induce apoptosis through a MYC independent mechanism and synergise with CDK inhibitors to kill osteosarcoma cells

Osteosarcoma (OS) survival rates have plateaued in part due to a lack of new therapeutic options. Here we demonstrate that bromodomain inhibitors (BETi), JQ1, I-BET151, I-BET762, exert potent anti-tumour activity against primary and established OS cell lines, mediated by inhibition of BRD4. Strikingly, unlike previous observations in long-term established human OS cell lines, the antiproliferative activity of JQ1 in primary OS cells was driven by the induction of apoptosis, not cell cycle arrest. In further contrast, JQ1 activity in OS was mediated independently of MYC downregulation. We identified that JQ1 suppresses the transcription factor FOSL1 by displacement of BRD4 from its locus. Loss of FOSL1 phenocopied the antiproliferative effects of JQ1, identifying FOSL1 suppression as a potential novel therapeutic approach for OS. As a monotherapy JQ1 demonstrated significant anti-tumour activity in vivo in an OS graft model. Further, combinatorial treatment approaches showed that JQ1 increased the sensitivity of OS cells to doxorubicin and induced potent synergistic activity when rationally combined with CDK inhibitors. The greater level of activity achieved with the combination of BETi with CDK inhibitors demonstrates the efficacy of this combination therapy. Taken together, our studies show that BET inhibitors are a promising new therapeutic for OS.


OS cells are sensitive to BET inhibition.
We treated a panel of genetically diverse human and mouse OS cultures, including long term established human OS cell lines and low passage primary cell cultures derived from genetically engineered mouse models (GEMM) of OS or human xenograft derived material 31,32 , with the BETi (+)-JQ1 (JQ1). The murine OS cultures were derived from two GEMM of human OS that recapitulate the OS fibroblastic and osteoblastic subtypes 33,34 . Primary tumour derived cultures were used to provide a closer representation of the in vivo tumour by limiting potential changes related to long term in vitro culture. JQ1 showed antiproliferative activity in all OS cells assessed at 72 hrs (Fig. 1a). An IC 50 concentration in the low nanomolar range (0.047 μM to 0.406 μM) was observed across subtypes and species (Fig. 1a). No appreciable differences in IC 50 were apparent if cell lines were clustered based on species, histological subtype, tumour site (primary vs metastasis), time in culture, or basal proliferation rate ( Fig. 1a and Supplementary, Fig. 1). To demonstrate specificity, we tested the inactive (-)-JQ1 enantiomer that does not inhibit BET bromodomains 26 . (-)-JQ1 showed no appreciable effects on OS cells at concentrations up to 10 μM (Fig. 1b). Additional testing was undertaken with two structurally unrelated BETi compounds, I-BET762 and I-BET151. Both compounds demonstrated antiproliferative activity in all cultures examined at 72 hrs, with IC 50 values predominantly in the nanomolar range (Fig. 1c), revealing a potent class effect of BETi on OS cells.

BET inhibition induces apoptosis in primary OS.
To identify the cellular mechanism underlying the antiproliferative effects of JQ1 in OS we assessed apoptosis and cell cycle status in representative primary OS cell cultures (494H, primary mouse fibroblastic; 148I, primary mouse osteoblastic; OS17, primary human). JQ1 rapidly induced apoptosis in a dose-responsive manner, evidenced by increased levels of cleaved caspase-3 within 24 hrs (Fig. 2a). A significant increase in cells that were Annexin-V positive compared to DMSO or (-)-JQ1 treatments in primary OS cell cultures was evident following treatment with 1 μM JQ1 for 48 hrs (Fig. 2b). Interestingly, the long established MG63 cell line, despite exhibiting an equally pronounced antiproliferative response to JQ1 (Fig. 1a,b), did not undergo apoptosis after extended treatment times (48-72 hrs) (Fig. 2c,d). To further investigate the antiproliferative effects, we assessed if JQ1 altered the cell cycle distribution of the cells. The mouse OS cultures (494H, 148I) were largely unaffected by treatment with 1 μM JQ1 for 24 hrs (Fig. 2e). Only a marginal decrease in the proportion of cells in the S and G 2 /M phases with a concurrent increase in cells in G 0 /G 1 were observed in the human OS17 primary cell culture (Fig. 2e). In contrast, the established MG63 cell line showed an accumulation of cells in G 0 /G 1 , indicative of a cell cycle arrest.
We next examined the specificity of the anti-OS activity of JQ1 by assessing its effects on normal wild-type primary osteoblast cultures. JQ1 treatment for 72 hrs suppressed the proliferation of primary osteoblast cultures at an IC 50 value of 0.196 μM (Fig. 2f) but did not reduce cell viability at time points when apoptosis was induced in JQ1 treated primary OS cultures (24-48 hrs) (Fig. 2g,h). A small decrease in the proportion of osteoblast cells in the S and G 2 /M phases with a concurrent increase in cells in G 0 / G 1 were observed following 24 hrs of JQ1 treatment (Fig. 2i). Taken together, these data demonstrate that JQ1 exerts an antiproliferative activity against primary OS cell cultures by inducing apoptosis rather than preventing cell cycle progression. Importantly normal osteoblasts are protected from the pro-apoptotic effects of JQ1.   (Fig. 3a). There was a striking suppression of proliferation of the OS in JQ1 treated mice (Fig. 3b,c). MicroCT analysis at day 28 showed that tumour bone volume was reduced on average by 70%  in JQ1 treated mice (Fig. 3d,e). Tumour masses were on average 50% smaller than vehicle treated mice (Fig. 3f). Taken together, these data demonstrate JQ1 has potent anti-OS activity in vivo.

BRD4 knockdown phenocopies BET inhibitors in OS.
To assess the contribution of the individual BET isoforms, we first determined that Brd2, Brd3 and Brd4 are expressed in mouse OS cells ( Supplementary Fig. 2a). Similarly, evaluation of microarray data demonstrated that BRD2, BRD3 and BRD4 are also expressed in human OS ( Supplementary Fig. 2b). We confirmed the expression of BRD2, BRD3 and BRD4 in our human and mouse OS cells compared to normal primary osteoblastic cells using QPCR. There was increased relative expression of Brd4, Brd2 and Brd3 in murine OS, with the highest levels in fibroblastic OS cells compared to normal osteoblastic cells ( Fig. 4a and Supplementary Fig.  3a-b). BRD4 expression was only moderately elevated in human OS cells compared to normal human osteoblasts ( Fig. 4a and Supplementary Fig. 3c-d). BRD4 inhibition has been strongly implicated in mediating the activity of JQ1 25,27 . To determine the contribution of BRD4 inhibition, we used siRNA against BRD4 in mouse and human OS. Reducing BRD4 levels (Fig. 4b,c) led to almost a 50% reduction in cell viability in all OS cultures examined (Fig. 4d). The anti-proliferative effect in primary OS cells (494H) was due to the induction of apoptosis, evidenced by increased levels of cleaved caspase-3 (Fig. 4e). Interestingly, consistent with the response to JQ1, the anti-proliferative response to BRD4 loss in MG63 cells was not due to induction of apoptosis (Fig. 4e). JQ1 also inhibits the activities of BRD2 and BRD3. siRNA targeting Brd2 and Brd3 in mouse OS cells also reduced cell viability to some extent (10-30%) ( Supplementary Fig. 4). Therefore BRD4 inhibition is the primary mediator of JQ1 activity in OS, but cumulative effects from the inhibition of all three BET proteins may contribute.

JQ1 suppresses FOSL1, independent of MYC inhibition. To define the molecular mechanisms of
BETi activity in OS, we examined genes transcriptionally downregulated by JQ1 in other cancers 25,27 . We focused on MYC, FOSL1 and RUNX2 as potential targets due to their link with OS or osteoblast biology. Firstly, we examined expression changes in OS cultures after JQ1 treatment (500 nM) for 6-24 hrs to identify early effects of JQ1. In contrast to many reports of MYC suppression in response to JQ1 23,25,28-30,36 , we saw either no reduction or, in some instances, slightly increased MYC expression following 6 hrs of treatment (Fig. 5a). After 24 hrs, MYC expression had only moderately reduced in 148I cells. Treatment with increasing concentrations of JQ1 over 24-48 hrs had negligible effect on MYC protein levels in all OS cells examined (Fig. 5b). Hence, under identical conditions to those that induced apoptosis or cell cycle arrest, we failed to observe changes in MYC expression suggesting MYC inhibition is not mediating the effects of JQ1 in OS.
In contrast to MYC expression, robust reductions in FOSL1 and RUNX2 levels occurred within 6 hrs of treatment (Fig. 5a). RUNX2 expression was decreased at 6 hrs, but expression in human OS cells was effectively restored after 24 hrs (Fig. 4a). FOSL1 expression levels were significantly repressed within 6 hrs of JQ1 treatment, except the MG63 cell line, which only showed a marginal reduction in expression at 24 hrs (Fig. 5a). Correspondingly, FOSL1 protein levels decreased with JQ1 treatment (Fig. 5b) suggesting FOSL1 could be a direct target of BETi.
To determine whether FOSL1 suppression may mediate the response to JQ1, we used siRNA to FOSL1. Knockdown of FOSL1 was confirmed (Fig. 5c), and significantly reduced viability up to 45% Enrichment is normalised to the total input DNA and a background region devoid of genes. Mean ± SEM (n = 3). * p < 0.05, ** p < 0.01 students t-test.
Scientific RepoRts | 5:10120 | DOi: 10.1038/srep10120 compared to non-targeting siRNAs (Fig. 5d). While the reduction in cell viability was less than observed with either BRD4 knockdown or JQ1 treatment, the results support a reduction in FOSL1 mediating the effects of JQ1. The rapid reductions of FOSL1 in response to JQ1 suggested JQ1 may be directly displacing BRD4 from the FOSL1 locus. We performed chromatin immunoprecipitation to map BRD4 occupancy at the FOSL1 locus. BRD4 was localised at the FOSL1 promoter and enhancer regions (Fig. 5e). Following JQ1 treatment, BRD4 was displaced, most prominently from the promoter. We observed minimal enrichment of BRD4 at the RUNX2 promoter region and its occupancy was unchanged in response to JQ1 treatment. BRD4 was localised in the coding region of RUNX2, and was displaced following JQ1 treatment (Fig. 5e). Enrichment of BRD4 at known sites surrounding the MYC transcription start site in OS cells was weak and occupancy was not reduced following JQ1 treatment, consistent with a lack of MYC transcriptional response to JQ1 treatment. These results define FOSL1 as a direct target of BRD4 in OS cells. JQ1 enhances standard OS chemotherapy and synergistically combines with CDK inhibitors. Chemotherapy treatment has improved the outcome of OS patients with localised tumours, yet therapy-related morbidity and poor response in patients with metastatic disease still remain significant hurdles 2,4 . We explored whether the addition of JQ1 to standard therapy could increase the efficacy of treatment in vitro. We treated a panel of primary OS cells (494H, 148I, OS17) with a dose range of doxorubicin combined with low dose JQ1 for 72 hrs and monitored cell proliferation responses. The addition of JQ1 was additive to the cytotoxic effects of doxorubicin suggesting JQ1 can chemo-sensitise OS cells (Fig. 6a). To further explore the role of BETi in OS response to chemotherapy, siRNA was used to deplete BRD4 in primary OS cells (494H, 148I, OS17). Consistent with the effects of combining JQ1 with doxorubicin, BRD4 depletion significantly increased the antiproliferative response of 494H, 148I and OS17 cells to doxorubicin (Fig. 6b), suggesting JQ1 could enhance the sensitivity of OS cells to standard therapy.
We further examined the utility of JQ1 in combination therapy approaches. BRD4 recruits the P-TEFb complex to promoters including FOSL1 37 , resulting in changes in RNA polymerase II phosphorylation and gene expression 38 . Cyclin dependent kinase 9 (CDK9) forms part of the P-TEFb complex, which is potently targeted by the CDK inhibitors, flavopiridol and dinaciclib 39 . Knockdown of Cdk9 resulted in a loss of OS cell viability (Fig. 6c,d) and OS cells were highly sensitive to both flavopiridol and dinaciclib with IC 50 values between 0.2 μM-0.33 μM and 0.007 μM-0.03 μM respectively (Fig. 6e,f and Supplementary Fig. 5). Treatment for 24 hrs induced a significant increase in apoptotic Annexin-V positive cells (Fig. 6g). We thus hypothesised that a multi-targeting approach of BRD4 activity through co-treatment with JQ1 and CDK inhibitors would be highly efficacious in OS.
Combinations of JQ1 with flavopiridol and dinaciclib were tested using effector concentration ranges below a 60% inhibition response. Viability was measured and combinatorial synergistic interactions were tested using the Bliss-additivity model. Greatly enhanced cell death was achieved when JQ1 was combined with flavopiridol or dinaciclib, reaching up to 50% more than any drug alone (Fig. 6h). Strong synergism between JQ1/flavopiridol and JQ1/dinaciclib was observed, reaching up to 300% and 80% more inhibition than predicted to be additive respectively (Fig. 6i). To further explore the synergy observed with co-inhibition of BETi and CDKs, we tested whether knockdown of BRD4 would synergise with dinaciclib to kill OS cells. OS cells were transfected with siRNAs to BRD4 or a non-targeting siRNA together with 25 nM dinaciclib and cell viability assessed. BRD4 knockdown combined with dinaciclib was moderately synergistic (15.64-15.06%), with BRD4 siRNA/dinaciclib treated cells being far greater inhibited (44-62%) than non-targeted cells treated with dinaciclib (Fig. 6j). The combination of JQ1 with dinaciclib resulted in a significant increase in the number of Annexin-V positive cells (Fig. 6k). The concurrent inhibition of BRD4 and CDKs is potently synergistic at inducing apoptosis of OS cells.

Discussion
The introduction of chemotherapy to OS treatment protocols resulted in a significant improvement in the survival of patients with localised tumours 2 . However, survival rates have plateaued and improvements in survival have been a result of improved delivery of existing treatments rather than from the rational application of new targeted agents 2,40 . In this study we have used a panel of OS cultures to show a potent class effect of BETi on OS cells. BETi, predominantly via inhibition of BRD4, had potent antiproliferative activity against human and mouse OS irrespective of origin (primary vs metastatic site), subtype, p53 status and proliferation rate, as well as potent in vivo activity in an OS preclinical model. Despite the strong link between BETi activity and MYC suppression in multiple models 23,25,[28][29][30]36 , we identified JQ1 anti-OS activity is MYC independent and was instead driven by FOSL1 suppression, similar to effects in lung cancer 27 . In addition, the anti-OS activity of JQ1 was selective, inducing rapid apoptosis in all primary OS cells, while normal osteoblasts remained protected from pro-apoptotic effects. Importantly, proof-of-principle studies for a BETi combinatorial treatment approach to OS are demonstrated. JQ1 sensitised OS cells to doxorubicin, a chemotherapeutic agent used in routine OS treatment practices. Furthermore, combining JQ1 with CDK inhibitors resulted in synergistic killing of OS, identifying a novel rational combinatorial treatment with agents in clinical trials. At least four BETi are currently being assessed in Phase I clinical trials. These studies provide a strong rationale for the further testing of BETi either alone or in combination clinically in OS. Downregulation of MYC has been reported as the mechanism mediating BETi activity in many tumour models [23][24][25]28,29 . Surprisingly, despite evidence implicating MYC as an oncogenic driver in OS [11][12][13] , MYC was not downregulated by JQ1 in our studies. Instead, we demonstrated FOSL1 is downregulated and is a significant contributor to the effects of JQ1. Although FOSL1 has been shown to be an important regulator of the osteoblastic content of bone by promoting osteoblast differentiation 20,21 , this is the first time FOSL1 has been implicated as a potential therapeutic target in OS. How suppression of FOSL1 transcription factor mediates the antiproliferative effects of BETi in OS remains to be determined. A closely related family member, FOS, is strongly associated with OS 18 , and redundancy between FOSL1 and FOS has been reported 19 . In lung cancer cells where BETi was shown to suppress FOSL1, a downregulated FOS gene signature was detected 27 . A gene signature common to these transcription factors may be important for OS survival.
The loss of viability from FOSL1 knockdown was less than observed with either JQ1 treatment or BRD4 knockdown suggesting, not unexpectedly, a multi-factorial action of BETi. Multiple targets are likely involved, as previous studies have demonstrated that over-expression of a single putative target such as MYC or FOSL1 was insufficient to completely rescue cells from JQ1 25,41 . Indeed we observed that JQ1 exposure repressed RUNX2, which has been linked to OS 8,9,42 . JQ1 has also been shown to downregulate anti-apoptotic genes such as BCL2 24 . Loss of BCL-2 has been shown to sensitise OS cells to doxorubicin 43 . An underappreciated aspect of the effects of JQ1 may be the upregulation of genes, which may also contribute to BETi activity. Further, JQ1 also inhibits BRD2 and BRD3. Although BRD4 loss was the most potent in reducing viability, knockdown of Brd2 and Brd3 in mouse OS cells also led to decreased survival, suggesting that the antiproliferative effects of JQ1 may not be fully mediated via a single target in BRD4.
JQ1 activity in OS cell lines has recently been reported in two independent studies that utilised long established OS cell lines 41,44 . We have confirmed their findings that RUNX2 is suppressed in response to JQ1 treatment, however several important differences are evident. Both groups reported that JQ1 treatment repressed MYC transcription in a subset of OS cell lines 41,44 . Interestingly, a loss of MYC protein was not evident or failed to be reported in the long established OS cells in response to JQ1 41,44 , in contrast to other tumour models where BETi activity was correlated with significant reductions in MYC protein 25,29,45 . Further, we demonstrated apoptosis was rapidly induced by JQ1 treatment in all primary OS cell cultures examined, whereas the prior studies observed apoptosis to be variable and limited in OS cell lines 41,44 . Why such contrasting findings? It perhaps involves the models used for preclinical testing. Our studies have deliberately made use of primary OS cultures which provide a closer representation of the disease by limiting potential changes from long term in vitro culture. Extended culture conditions may result in alternative BRD4 genomic localisation patterns, rendering cell lines differentially susceptible to treatments. The cell type specific effects mediated by JQ1 is proposed to be controlled by cell specific asymmetrical super-loading of BRD4 at enhancers of genes involved in cell-fate determination and oncogenic drivers 23,46 . JQ1 treatment leads to a preferential loss of BRD4 at these enhancers and subsequent transcriptional repression 23,46 . Genome-wide super-enhancer H3K27ac ChIP-Seq analysis could distinguish differences that segregated diffuse large B cell lymphoma cell lines from primary samples 23 . Indeed, cell specific BRD4 super-loading of oncogenic drivers in OS cells may explain the protection of normal osteoblasts from the pro-apoptotic effects of JQ1 treatment. The protection of normal osteoblasts from the pro-apoptotic OS activity of JQ1 provides compelling rational for clinical testing, and warrants investigation of OS BRD4 super-enhancers to potentially identify oncogenic susceptibilities.
One function of BRD4 is to recruit P-TEFb to gene promoters where it phosphorylates RNA polymerase II to induce transcriptional elongation 38 . We tested whether concurrent inhibition of P-TEFb would co-operate with JQ1 by further compromising BRD4 function. JQ1 combined with the CDK inhibitors dinaciclib or flavopiridol led to synergistic killing of OS in vitro. Synergy was observed with both agents suggesting the effects were not compound specific, rather driven by "on target" inhibition of CDKs. The molecular basis of the synergistic reaction remains to be determined. Knockdown of BRD4 with dinaciclib was able to phenocopy the synergistic effects of the combined pharmacological approach. Dinaciclib has been shown to bind and inhibit the activity of BET proteins, albeit to levels significantly less than CDKs 47 . As synergy was still achieved with knockdown of BRD4, the interactions between the drugs is not likely driven by direct inhibition of BRD4 by dinaciclib. The more potent effect of JQ1 compared to BRD4 knockdown when combined with dinaciclib may reflect contributions from BRD2/3.
The results offer promise for increasing the in vivo efficacy of dinaciclib. Dinaciclib showed in vitro activity in multiple cancer cell lines and was well tolerated in pre-clinical testing 48,49 . We observed that primary OS cells are extremely sensitive in vitro to dinaciclib, with IC 50 values in the low nanomolar range. However in vivo testing of dinaciclib in OS xenografts showed limited activity 48 . Apart from CLL, phase II trials of dinaciclib in multiple malignancies have not yielded encouraging responses 50 . Several Phase I trials have commenced testing dinaciclib in combination with bortezomib, dexamethasone and the AKT inhibitor MK2206. The current findings provide rationale for the testing of dinaciclib and BETi combination treatment.
In addition to dinaciclib, JQ1 improved OS cell response to doxorubicin. We demonstrated that JQ1 and doxorubicin induces additive effects in reducing OS cell viability. Although a synergistic response was not achieved, the results indicate the addition of JQ1 could reduce the dose of doxorubicin required to achieve the same degree of cell death. Given doxorubicin treatment is associated with severe life Scientific RepoRts | 5:10120 | DOi: 10.1038/srep10120 threatening side effects 4 , strategies that can lower the dose of doxorubicin could have significant benefit to patients' morbidity and treatment tolerability.
The ability of BETi to target different oncogenic pathways in different cancers suggests they will have wide ranging clinical applications. Our studies provide strong evidence that BET inhibitors represent a novel therapeutic strategy to treat OS either as monotherapy or in combination with current or novel therapies. We demonstrate that primary tumour derived cell cultures from murine OS models or human patient material are highly sensitive to BETi and undergo apoptosis as the primary antiproliferative response. This contrasts with work in long established cell cultures, where apoptosis was variable and cell cycle arrest and induction of senescence was more prominent. This study emphasises the importance of testing novel agents and defining their mechanism of action in models more closely representative of the disease as it occurs in patients.

Methods
Cell culture. Cells were maintained in αMEM supplemented with 10% FCS. Human cell lines were purchased from ATCC (no authentication performed). Primary human OS cultures were derived from tumours from xenograft models 31,32 (material generously provided by Dr Peter Houghton, Nationwide Children's Hospital, Ohio; no authentication performed). Normal human osteoblasts were derived from bone marrow aspirates from normal healthy individuals 51 (no authentication performed). Primary mouse OS cultures were derived from tumours from fibroblastic or osteoblastic models of OS 33,34 (no authentication performed). Primary osteoblast cells were isolated from C57BL6 mouse collagenase digested bones. Osteoblastic cells were purified by flow cytometry (FACSAria, BD Biosciences) as previously described 52,53 , or directly cultured following collagenase digestion. Flow cytometry antibodies are described in Supplementary Table S1. Cell viability and proliferation assays. Cell were seeded 24 h prior to drug treatment at 2500/well in 96 well plates. Compounds, diluted in DMSO, were applied (0.01 nM to 10 μM) for 48-72 h and viability quantified using CellTitre-Glo (Promega). Half maximal inhibitory (IC 50 ) values were calculated using Prism 6.0e software. Proliferation rate was determined by counting cells on 4-5 consecutive days using a BioRad TC10 counter (Hercules, CA, USA).

Synergy calculations.
Synergy was calculated using the Bliss-Additivity model 54 . Experimental responses > 15% above the predicted combined response value were considered to indicate synergism 54 .
RNA extraction and Real-time PCR. QPCR analyses were carried out using primers described in Supplementary Table S2 using methods as previously described 33,34 .

Genome wide expression analyses. RNA-Sequencing was conducted on an Illumina HiSeq 2000
with 100 bp paired-end reads. Reads were aligned to the mouse genome build mm9/NCBI37 using Casava 1.7 and Bowtie v0.12.2 mapping software, normalised using Voom linear modeling 55 and transcript abundance measured as reads per kilobase of exon per million mapped reads (RPKM) 56 . Datasets are deposited in GEO (accession number GSE58916). Gene expression data from a publicly available human OS dataset (accession number GSE33382 57 ) were normalised using R/bioconductor. Chromatin Immunoprecipitation. ChIP was carried out on 5 × 10 6 cells as described previously 58 .
Cells were treated with 1 μM JQ1 or 0.1% v/v DMSO for 24 hrs. Cells were fixed with 1% formaldehyde diluted in PBS for 10 min at room temperature. Cross-linking was quenched by incubating cells with 0.125 M glycine diluted in PBS for 10 min at room temperature. Cells were washed twice in PBS before being scraped from plates and pelleted by centrifugation at 400 x g for 5 minutes at room temperature. Cell pellets were diluted in sonication buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl pH8.1) with cocktail protease inhibitors (Roche) and the DNA sheared to lengths between 100 and 400 bp using a UCD-200 Bioruptor (Diagnenode) on high for a total shearing time of 15 min (90 min of 10 s on and 50 s off), maintained at 4 degrees celcius during shearing. Cell debris was cleared by centrifugation for 10 min at 13,000 rpm at 4 degrees celcius and supernatants diluted 10-fold in ChIP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM ETA, 16.7 mM Tris-HCl pH 8.1, 167 mM NaCL) with protease inhibitors. After removing 20 μl of each sample as an input control, protein-DNA fragments were immunoprecipitated with 5 μg BRD4 antibody (Bethyl Laboratories) or control rabbit IgG (Merck Millipore), or no antibody overnight at 4 degrees celcius with rotation. Complexes were collected for 1.5 hrs at 4 degrees celcius with rotation with 60 μl of protein A dynabeads (Invitrogen) that had been pre-blocked for 1 hr in 0.5% BSA in TE buffer. Beads were washed one time each with Low Salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 m Tris-HCl pH8.1, 150 mM NaCl), High Salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH8.1, 500 mM NaCl) and LiCl buffer (0.25 M LiCl, 1% NP40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl pH8.1), followed by two washes with TE buffer. Protein-DNA complexes were eluted from the beads (0.1 M NaHCO 3 , 1% SDS) with 2 hrs incubation at 62 degrees celcius with shaking. Cross-links were reversed overnight at 65 degrees celcius with the addition of 0.2 M NaCl. Protein was digested by incubation with 50 μg of proteinase K at 45 degrees celcius for 1 hr. DNA was purified by phenol:chloroform:isoamyl alcohol (25:24:1) (Sigma) extraction and ethanol precipitation with GlycoBlue (Life Technologies) as a carrier. BRD4 bound and input samples were analysed by QPCR using primers that amplified regions of the FOSL1, c-MYC and RUNX2 loci or a negative control background region (region devoid of genes) (sequences available in Supplementary table S2). Relative enrichment was determined by the 2 (−deltaCT) method normalizing to input controls and background regions.
In vivo studies. All animal experiments were undertaken in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (7th edition, 2004 NHMRC; the "approved guidelines") and under experimental protocols approved by the St Vincent's Hospital Melbourne Animal Ethics Committee. 494H mouse fibroblastic OS cells infected with pMMP-LucNeo virus were used to generate the model. 25,000 tumour cells (increased to 75,000 in a second independent study) resuspended in extracellular matrix (Cultrex PathClear BME Reduced Growth Factor Basement Membrane Extract, Trevigen) were implanted sc on the dorsal hind flank of 6-7 week old Balb/c nu/ nu mice. Tumours were allowed to establish for 1 week and were visualised and quantitated by bioluminescence imaging before treatment commenced with JQ1 (50 mg/kg) or vehicle (10% DMSO in 10% hydroxypropyl beta cyclodextrin) daily by ip injection. Tumour volume was measured using an IVIS Spectrum Imaging System (Perkin Elmer). Mice were injected with XenoLight D-luciferin (15 mg/ml) (Perkin Elmer) intraperitoneally. Bioluminescence was quantitated using Living Images software (Perkin Elmer).