BET Inhibition Induces Apoptosis in Aggressive B-Cell Lymphoma via Epigenetic Regulation of BCL-2 Family Members

The bromodomain and extra-terminal domain (BET) family members (BRD2, BRD3, BRD4, and BRDT) comprise a class of epigenetic reader proteins, which bind acetylated lysine residues on histones to facilitate the recruitment of transcriptional elongation complexes (1). BRD4 is associated with almost all active promoters and most active enhancers in both transformed and nontransformed cells (2). Loading of BRD4 onto “super-enhancers” drives oncogenic transcription programs in lymphoma, particularly where immunoglobulin gene switch translocations are juxtaposed to cMYC (3). Thus, small-molecule BET-inhibitors have been proposed as a MYC pathway–targeted therapeutic with preclinical activity demonstrated in multiple myeloma, Burkitt lymphoma, acute lymphoblastic lymphoma, diffuse large B-cell lymphoma (DLBCL), and acute myelogenous leukemia (4–8). BRD4 may also function as a coactivator of MYC- and/or E2F-regulated genes and inhibition of BRD4 activity can affect the expression of MYC-target genes independent of decreased MYC expression (3). BET inhibitors also have activity in models of MLL-rearranged AML where MYC is not the primary oncogenic driver and instead epigenetic modulation of cell-cycle regulators and apoptotic pathways were mechanistically implicated (8–12). Histone-independent roles for BET proteins are also relevant in lymphoma cell survival, as BRD4 chaperones acetylated RELA, augmenting NFκB signaling (13).

As BET inhibitors are now in active clinical development, there is a pressing need to elucidate mechanisms of antineoplastic activity to facilitate effective drug combination strategies and help predict which patients are most likely to derive therapeutic gain. We therefore sought to interrogate determinants of apoptosis induction by the prototypical thienodiazepine-based BET-inhibitor, JQ1 (14), using a well-characterized and genetically tractable murine model of aggressive Myc-driven lymphoma (15). Consistent with previous reports, JQ1 potently induced cytostasis and apoptosis in Eμ-Myc lymphoma cells at on-target nanomolar concentrations (16). Unexpectedly, however (and in contrast to human Burkitt lymphoma lines), JQ1 did not downregulate transgenic cMyc transcription or protein in Eμ-Myc cells. Thus, the Eμ-Myc model provided a unique system for dissecting BET-inhibitor activity in MYC-driven malignancy where expression of MYC is maintained when BRD4 is inhibited. We characterized this response as p53-independent and mediated by epigenetic modulation of BCL-2 family proteins to activate the intrinsic mitochondrial apoptotic pathway. Despite statistically significant prolongation of survival, including p53-defective lymphomas, progressive disease was observed with subsequent in vivo derivation of JQ1-resistant lymphoma. Mutational activation of RAS, as well as perturbations of the balance between proapoptotic BIM and antiapoptotic BCL-2 family proteins mediated resistance to JQ1. These data indicate that MYC downregulation is not an absolute requirement for BET-inhibitor activity, and that lesions antagonizing the intrinsic apoptotic pathway may result in acquisition of BET-inhibitor resistance in the clinic.

Eμ-Myc lymphomas were derived, cultured, and transplanted as previously described (17). Retroviral transduction of freshly isolated Eμ-Myc lymphomas used murine stem-cell virus-internal ribosomal entry site-green fluorescence protein (MSCV-IRES-GFP) vectors according to standard techniques. MSCV-IRES-GFP/Bcl-2 was cloned as previously described (18). MSCV-IRES-GFP/NrasQ61K was generated by subcloning wild-type murine Nras into the MSCV vector and performing site directed mutagenesis using the Agilent Quikchange II XL kit according to the manufacturer's instructions. The Ramos cell line was purchased from the ATCC. The BL-41 and OPM2 cell lines were purchased from Leibniz-Institut DSMZ (Braunschweig, Germany). All human cell lines were authenticated by the suppliers using Short Tandem Repeat (STR) profiling, passaged for fewer than 6 months after resuscitation, maintained at 5% CO₂ and cultured in Gibco RPMI1640 supplemented with 10% FCS, penicillin (100 U/mL), and streptomycin (100 mg/mL). Etoposide and doxorubicin were diluted from clinical pharmacy stock (Peter MacCallum Cancer Centre). (+)-JQ1 (JQ1), IBET-151, IBET-762, Y803, and RVX-208 were provided by James Bradner (Dana Farber Cancer Institute, Boston, MA). Fedratinib was kindly provided by Sanofi-Aventis (Paris, France). For in vitro use, all small-molecule inhibitors were dissolved in DMSO to generate stock solution with a final concentration of 10 mmol/L.

Eμ-Myc lymphoma cells (1–2 × 10⁵) were incubated in the presence of JQ1, or vehicle control (DMSO), for 24 hours before flow cytometric analysis of viability [assessed by Annexin-V and propidium iodide (PI) positivity], and cell-cycle progression, assessed by nuclear DNA content (PI) staining as previously described (18). Loss of mitochondrial outer membrane potential (MOMP, % Delta Psi loss) was assessed by tetramethylrhodamine ethylester (TMRE) staining as previously described (19). Data were collected on a FACSCanto II flow cytometer (BD Biosciences) and analyzed using FlowJo Software, version 10.0.7 (Tree Star).

Whole cells lysates from cultured cells or were prepared in ice-cold immunoprecipitation lysis buffer (0.15 mol/L NaCl, 10 mmol/L Tris-Cl, pH 7.4, 5 mmol/L EDTA and 1% Triton X-100) and supplemented with protease and phosphatase inhibitors (complete EDTA-free protease and phosSTOP inhibitor cocktails, Roche Diagnostics). Lysates (20–50 μg) were separated on 10% SDS polyacrylamide gels and electroblotted onto Immobilon-P nylon membranes (Millipore). Membranes were incubated with the following primary antibodies.

Hamster anti-Bcl-2 (clone 3F11, BD Biosciences), rabbit anti–Bcl-X_L (Sana Cruz Biotechnology Inc.), rabbit anti–Mcl-1 (Rockland Immunochemicals Inc.), mouse anti-HSP90 (clone AC88, Enzo Life Sciences Inc.), mouse anti–α-tubulin (Merck KGaA), mouse anti–β-actin (clone AC-74, Sigma-Aldrich). Rabbit anti-PARP (clone 46D11), rabbit anti–P-p44/42 MAPK Thr 208/Tyr 204 (clone D13.14.4E), rabbit anti-cMyc were purchased from Cell Signaling Technology. All antibodies were incubated overnight at 4°C followed by subsequent incubation with horseradish peroxidase–conjugated secondary antibodies (DAKO). Immunoreactive bands were visualized by enhanced chemiluminescence (Amersham). All Western blots were repeated at least three times.

RNA was extracted from cell pellets using the Nucleospin RNA Extraction Kit (Macherey-Nagel) as per the manufacturer's instructions. cDNA was synthesized according to the manufacturer's instructions (Promega). Quantitative PCR analysis of samples was performed on the 7900HT Fast Real-Time PCR System (Applied Biosystems) with SYBR-green ROX mix (Agilent). GAPDH and L32 were used as the murine and human control genes, respectively. Primer sequences were: Mus musculus Mcl-1 F: GGTGCCTTTGTGGCCAAACACTTA R: ACCCATCCCAGCCTCTTTGTTTGA, Mus musculus Bcl-2 F: ATGACTGAGTACCTGAACCGGCAT R: GGGCCATATAGT TCCACAAAGGCA, Mus musculus Bcl-X_L F: AAGCGTAGACAAGGAGATGCAGGT R: GCATTGTTCCCGTAGAGATCCACA, Mus musculus total MYC F: GGACGACGAGACCTTCATCAA R: CCAGCTTCTCTGAGACGAGCTT, Mus musculus endogenous MYC F: CAGCTCCTCCTCGAGTTAG R: TGAGGAAACGACGAGAACAG, Mus musculus transgenic PhiX MYC F: TCGAACAGCTTCGAAACTCTGGTG R: TTAAATCGAAGTGGACTGCTGGCG, Mus musculus RN3 F: ATTTTGAGCGCATTGTGTTGAGC R: GGGAGCATCTGGCGACTGTTC, Mus musculus Bim F: GGATCGGAGACGAGTTCAACGAAA R:TTCAGCCTCGCGGTAATCAT, Mus musculus GAPDH F: CCTTCATTGACCTCAACTAC R: GGAAGGCCATGCCAGTGAGC, Homo sapiens MYC F: GGACGACGAGACCTTCATCAA R: CCAGCTTCTCTGAGACGAGCTT, Homo sapiens L32 F: TTCCTGGTCCACAATGTCAAG R: TTGTGAGCGATCTCGGCAC.

The Peter MacCallum Cancer Centre Animal Ethics Committee approved all in vivo procedures in this study. C57BL/6 mice were purchased from the Walter and Eliza Hall Institute (Melbourne, VIC). For transplantation of Eμ-Myc lymphomas in vivo, cohorts of 6- to 8-week-old syngeneic C57BL/6 mice were inoculated via tail vein injection with 1 to 4 × 10⁵ Eμ-Myc lymphoma cells. For in vivo use, JQ1 was reconstituted in 1 part DMSO to 9 parts 10% (w/v) Hydroxypropyl-β-cyclodextrin (HPBCD; Cyclodextrin Technologies Development Inc.) in sterile water, or DMSO vehicle control. JQ1 was dosed at 50 mg/kg 5 days per week (5d/2d) via intraperitoneal injection, commencing 3 days after lymphoma inoculation, for a total of 5-weeks therapy or until treatment failure. For detection of GFP-positive cells, 10 μL whole blood was incubated in 200 μL red cell lysis buffer (150 mmol/L NH₄Cl, 10 mmol/L KHCO₃, 0.1 mol/L EDTA) and washed twice in ice-cold flow cytometry buffer (2% FCS and 0.02% NaN₃ in PBS). Data were collected on a FACSCanto II flow cytometer (BD Biosciences) and analyzed using FlowJo Software, version 10.0.7 (Tree Star). Full blood count analyses were performed on the CELL-DYN Sapphire Blood Analysis Instrument (Abbott Laboratories).

DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen). Fragment-sequencing libraries were prepared and then processed for exome capture using SureSelect Mouse All Exon chemistry (Agilent Technologies). Captured libraries were sequenced on Illumina HiSeq 2000 using a paired-end sequencing strategy achieving approximately 100x average read depth over captured bases. Short-read sequence alignment and somatic variant calling was done as previously described (20). Variant validation was performed by Sanger sequencing directly from gDNA using BigDye terminator chemistry (Life Technologies, Thermo Fisher Scientific) and a custom oligonucleotide primer designed 3′-prime of the target base. Primer sequence used was Mus musculus Nras R: TGGCAAATACACAGAGGAACC.

Statistical analysis was performed using GraphPad Prism Software, Version 6.0c.

JQ1 is a potent and selective BET bromodomain inhibitor with a k_d of 49 to 190 nmol/L for BET bromodomain family members (14). To determine whether Eμ-Myc lymphomas were sensitive to BET bromodomain inhibition in vitro, we cultured a series of independently-derived lymphomas in increasing concentrations of JQ1 for 24 hours before assessment of cell viability using Annexin-V/PI positivity (Fig. 1A–C). Dose-dependent induction of cell death was evident in the majority (4/5) of Eμ-Myc lymphomas. Importantly, we also identified a primary Eμ-Myc lymphoma (^#6066) with relative de novo resistance to JQ1 at concentrations up to 1.5 μmol/L (Fig. 1A).

Defective p53 signaling mediates resistance to conventional chemotherapeutics and is associated with poorer outcomes in the lymphoma clinic (21). To elucidate the requirement for p53-competence in JQ1-mediated apoptosis we also tested the sensitivity of Eμ-Myc lymphomas with defective p53 signaling. An Eμ-Myc lymphoma derived on a p53-null background (^#3391) and a second lymphoma bearing a previously characterized p53 mutation (^#106; refs. 17, 22) were both as sensitive to JQ1 as a p53 wild-type lymphoma (^#4242; Fig. 1B), despite relative etoposide and doxorubicin resistance (Fig. 1D). Thus, JQ1 potently induces cell death in Eμ-Myc lymphomas that are chemoresistant due to loss or dysfunction of p53.

We next sought to confirm whether JQ1-induced cell death occurred via the intrinsic apoptotic pathway. In addition to phosphatidylserine exposure (assessed by Annexin-V staining; Fig. 1A–C), JQ1-treatment induced accumulation of sub-diploid DNA on propidium iodide nuclear staining (Fig. 1E) and mitochondrial outer-membrane permeabilization (Fig. 1F). Furthermore, JQ1 treatment potently induced proteolytic cleavage of PARP as a downstream marker of caspase activation (Fig. 1G). To confirm the phenotype is not limited to the JQ1 pharmacophore, we assessed apoptosis induction by a panel of structurally distinct bromodomain inhibitors. Specifically, we showed proapoptotic activity at on-target concentrations by JQ1, IBET-151, IBET-762, RVX-208, and Y803 (OTX015/MK-8628) against Eμ-Myc lymphoma cells in vitro (Supplementary Fig. S1A).

As these results implicated the intrinsic pathway of apoptosis induction, we retrovirally expressed Bcl-2 as a means to prevent mitochondrial permeabilization in lymphomas with primary sensitivity to JQ1. These Eμ-Myc MSCV-Bcl-2 lymphomas were rendered highly resistant to JQ1-induced mitochondrial permeabilization (Fig. 1F) and apoptosis (Fig. 1H) compared with their isogenic controls. In the absence of apoptosis induction, Eμ-Myc MSCV-Bcl-2 lymphomas underwent G₁ cell-cycle arrest (Fig. 1I). Further, we show ectopic retroviral expression of Bcl-2 in the human IG-cMYC translocated multiple myeloma cell line, OPM2, abrogates JQ1-induced apoptosis (Supplementary Fig. S1B–S1D). Thus, BET-inhibition mediates apoptosis in human and mouse multiple myeloma and lymphoma cells that can be inhibited by forced expression of Bcl-2.

The Eμ-Myc model of aggressive B-cell lymphoma/leukemia was generated with a transgene that juxtaposes murine cMyc to the immunoglobulin Eμ-enhancer element to generate a B-cell–restricted tumor (15). Analogous switch translocations of the cMYC locus occur in Burkitt lymphoma, subgroups of diffuse large B-cell lymphoma and multiple myeloma. Initial descriptions of BET-inhibition in IG-cMYC–translocated lymphoid malignancies reported specific downregulation of cMYC and cMYC-associated transcriptional programs (4). We therefore anticipated that JQ1's activity in the Eμ-Myc model would be mediated by transcriptional downregulation of cMyc. Unexpectedly, no reductions in cMyc transcript were detected following treatment with pro-apoptotic concentration of JQ1 (1 μmol/L) over a time course corresponding to apoptosis induction in a sensitive lymphoma (Fig. 2A). Furthermore, JQ1 did not downregulate the Myc-target gene, Rn3 (Fig. 2B; ref. 23). In contrast, the human IG-cMYC-translocated Burkitt lymphoma cell lines, BL-41 and Ramos, or multiple myeloma cell line, OPM2, showed robust suppression of MYC mRNA levels within 2 hours of drug treatment (Fig. 2C). In these human lines, BET-inhibition displaces BRD4 from endogenous immunoglobulin-associated super-enhancers brought into play by the chromosomal translocation (3). The orchestrated expression of Myc from the Eμ-Myc transgene may not accurately reflect this pathophysiology, providing an explanation for these divergent observations. To investigate whether the epigenetic regulation of transgenic Myc expression in Eμ-Myc lymphomas was differentially regulated compared with endogenous expression programs, we designed distinct primer sets specific for the Eμ-Myc transgene (by flanking the PhiX bacteriophage region adjacent to transgenic Myc), and the endogenous Myc loci, respectively. As expected, significantly higher expression of the Eμ-Myc transgene–derived Myc transcript compared with endogenous Myc transcript was observed, with an average ratio of 340:1 (Fig. 2D). Endogenous, but not transgenic Myc transcript was suppressed by JQ1-treatment in Eμ-Myc lymphoma cells (Fig. 2E and F). Thus, although JQ1 treatment completely suppressed the expression of endogenous Myc (as is also seen with human MYC in the context of immunoglobulin switch translocations), there was no suppression of the Eμ-Myc transgene. Accordingly, because transgenic Myc expression dominates in Eμ-Myc cells, MYC protein expression was not downregulated before apoptosis induction (Fig. 2G). As previously demonstrated, we confirmed that JQ1 treatment is associated with loss of MYC protein expression in a human MYC-translocated multiple myeloma cell line, OPM2 (Supplementary Fig. S2A). However, BRD4-independent retroviral expression of MYC in OPM2 cells did not completely suppress the induction of apoptosis by JQ1 (Supplementary Fig. S2B). Consistent with our data in Eμ-Myc cells, these data suggest that apoptosis induction by JQ1 is not dependent on loss of MYC expression. Therefore, the robust induction of apoptosis in Eμ-Myc lymphoma differs mechanistically from other experimental systems where decreased MYC expression appears to be important to mediate the biological effects of BET inhibitors (4, 8, 24).

BCL-2 family members have previously been identified as bona fide BET protein–regulated genes and important mediators of the bromodomain inhibitor–induced apoptotic response in models of MLL-rearranged AML (9). Having demonstrated activation of the intrinsic apoptotic pathway by JQ1 despite sustained Myc expression, we next assessed the capacity for direct modulation of BCL-2 family members. Short exposures to proapoptotic JQ1 concentrations significantly reduced Bcl-2 and Bcl-X_L mRNA with no effect on Mcl-1 mRNA levels (Fig. 3A–C). Prosurvival BCL-2 proteins sequester proapoptotic BH3-only proteins, including BIM, thereby preventing BAX/BAK activation and induction of MOMP (25). In the context of malignant peripheral nerve-sheath tumors, genetic knockdown of BRD4 and treatment with JQ1 both potently induce the expression of proapoptotic BIM (26). Moreover, BIM is regulated independently of p53 to augment apoptosis in Myc-driven lymphomagenesis (27). Concurrent to Bcl-2 and Bcl-X_L suppression, JQ1 treatment rapidly increased Bim transcription (Fig. 3D). To elucidate the functional importance of Bim in the apoptotic response to JQ1, we treated two Eμ-Myc lymphomas derived on a Bim-null background (^#24 and ^#20). These demonstrated high-level JQ1 resistance (up to 2 μmol/L) compared with a representative Bim wild-type control (^#4242; Fig. 3E). These Bim-deficient lymphomas do not have a general antiapoptotic phenotype as they remain equally sensitive to etoposide, doxorubicin and non–BET-targeted therapeutics [including PI3K inhibitor (ref. 17) and cyclin-dependent kinase (CDK) inhibitors (ref. 28); Fig. 3F]. Taken together with the results presented in Fig. 1, these data indicate that JQ1-mediated apoptosis induction occurs due to a shift in the balance of BCL-2 family proteins, favoring mitochondrial permeabilization via a relative excess of Bim to Bcl-2 and Bcl-X_Ltranscription.

Having demonstrated a capacity to induce p53-independent apoptosis in vitro, we next assessed the efficacy of JQ1 in vivo. In mice bearing established lymphoma with palpable lymphadenopathy, a single 50 mg/kg dose of JQ1 induced a statistically significant reduction in spleen weight and increased cell death in explanted lymph nodes (Fig. 4A and B). Prolonged JQ1 treatment delayed the outgrowth of lymphomas, as evidenced by reduced numbers of cells in leukemic phase (Fig. 4C) and improved median survival (18 d vs. 34 d, P < 0.0001; 26 d vs. 52 d, P < 0.0001 for ^#4242 and ^#299, respectively, Fig. 4D and E). Similar results were achieved in mice bearing etoposide-resistant lymphoma derived on a p53-null background (Fig. 4F, 20 d vs. 33 d, P < 0.001). To substantiate our in vitro observations that JQ1 mediates its activity via engaging the intrinsic apoptotic pathway, we next assessed its efficacy in the context of Bcl-2 overexpression in vivo. Retroviral expression of Bcl-2 significantly abrogated the survival advantage conveyed by JQ1 (Fig. 4G; prolongation of median survival 4d ^#4242 MSCV-Bcl-2, P < 0.001, vs. 16d for ^#4242 MSCV). Prolonged JQ1 treatment was well-tolerated in vivo and no significant changes in body weight were observed (Supplementary Fig. S3A). Moreover, there was minimal hematological toxicity with no significant changes in hemoglobin or neutrophils, although modest thrombocytopenia was observed (Supplementary Fig. S3B–S3D).

There has been much recent interest into mechanisms of acquired resistance to BET inhibition, which has proven difficult to engineer in vitro (29, 30). We observed universal treatment-failure despite ongoing dosing in mice bearing ^#4242 lymphoma that was initially sensitive to JQ1 (Fig. 4D). Re-challenge of previously treated Eμ-Myc^#4242 lymphomas (annotated JQ1R) with JQ1 demonstrated sustained drug resistance ex vivo compared with matched JQ1 naïve control (JQ1S; Fig. 5A). Re-transplantation of JQ1R and re-treatment with JQ1 in vivo according to the same schedule used for parental lymphoma resulted in far inferior survival outcomes (Fig. 5B, prolongation of median survival 6d ^#4242 JQ1R, P < 0.0001, vs. 16d for ^#4242 MSCV).

We used global RNA sequencing (RNA-seq) to investigate compensatory transcriptional pathways being activated in JQ1-resistant cells (Supplementary Methods). These analyses identified 108 genes that we were significantly up- or downregulated in JQ1R cells, including Bcl-2 (Fig. 5C–E). As we hypothesized aberrant regulation of the intrinsic apoptotic pathway–mediated resistance to JQ1, we compared differentially regulated genes (JQ1R vs. JQ1S, FDR < 0.05 and logFC > 1) to genes known to regulate the intrinsic mitochondrial apoptotic pathway (Fig. 4F). This analysis demonstrated Bcl-2 as the only apoptosis-related gene differentially expressed that we have subsequently validated using RT-PCR and immunoblot (Fig. 5G and H). Bim induction was observed in JQ1S and JQ1R cell lines following 16 hours of exposure to JQ1 at both the mRNA and protein level (Supplementary Fig. S4A–S4B); however, excess Bcl-2 likely buffers the increased Bim in JQ1R cells to prevent BAX/BAK activation and apoptosis induction. Gene set enrichment analyses (GSEA) was used to identify transcriptional programs differentially expressed in JQ1-resistant cells where a KRAS expression signature was strongly correlated with JQ1R cells (Fig. 5I). Therefore, despite initial significant (including p53 independent) disease responsiveness, sustained JQ1 exposure leads to the emergence of resistant clones associated with alterations in BCL-2 protein expression. This correlates with abrogation of in vivo activity by forced retroviral expression of BCL-2.

Our initial experiments identified a lymphoma (^#6066) with de novo resistance to JQ1-induced apoptosis (^#6066, Fig. 1A). As seen with ^#4242 MSCV-Bcl2 lymphoma (Fig. 1I), a secondary cytostatic response was observed in the ^#6066 lymphoma that was resistant to apoptosis (Fig. 6A). However, the reduced apoptosis in the ^#6066 lymphoma was not explained by blunting of the transcriptional upregulation of Bim following JQ1 treatment (Fig. 6B). Overall survival in JQ1-treated mice bearing this lymphoma was poor (median survival 10d vs. 17d, P = 0.0004, Fig. 6C) compared with prior experiments with lymphomas demonstrating in vitro sensitivity (Fig. 4B and C). Consistent with p53-dependent mechanisms of apoptosis induction by chemotherapy, the ^#6066 lymphoma has wild-type p53 expression and is sensitive to etoposide and doxorubicin in vitro (Fig. 1D). Interestingly, we also observed cross-resistance to the dual JAK-2/BET inhibitor (31), Fedratinib (Supplementary Fig. S5A). To elucidate this mechanism of selective resistance to JQ1, we performed whole-exome sequencing of ^#6066 and identified somatic variants by comparison with Eμ-Myc germline DNA. Notably, ^#6066 bears a hot-spot mutation of Nras (Q61K), evoking constitutive RAS pathway activation (Fig. 6D). As RAS pathway activation conveys resistance to apoptosis mediated via BCL-2 (32), we hypothesized that Nras activation mediated resistance to JQ1. To test this in an isogenic setting, the JQ1-sensitive, Nras wild-type lymphoma (^#4242) was transduced with empty vector (MSCV-GFP) or MSCV-NrasQ61K resulting in constitutive RAS pathway activation as evidenced by phosphorylation of MAPK (Fig. 6E). Like ^#6066 (bearing a spontaneous Nras mutation), 4242 MSCV-NrasQ61K showed relative resistance to the apoptotic effects of JQ1 (Fig. 6F). Similarly, OPM2 cells transduced with MSCV-NrasQ61K display increased phosphorylation of MAPK and are more resistant to JQ1-induced apoptosis compared with the empty vector control (Supplementary Fig. S5B–S5C). Having demonstrated enrichment of a KRAS gene signature and upregulation of Bcl-2 in JQ1R cells, we hypothesized that RAS activation in lymphoma cells may regulate the expression of BCL-2 to convey resistance to JQ1. We show RAS-activation correlated with BCL-2 upregulation as evidenced by Western blot (Fig. 6G) and reduced sensitivity to BH3-mimetic treatment (Fig. 6H). Thus, constitutive RAS pathway activation mediates resistance to the apoptotic effects of BET-inhibition concurrent with upregulation of BCL-2.

Initial descriptions of targeted BET inhibition evoked much interest, due in part to the opportunity of indirect targeting of the previously “undruggable” oncogene, cMYC (4). MYC activation is relevant across the full spectrum of human cancers, and is of particular relevance to aggressive B-cell malignancy. In that context, the presence of an isolated MYC translocation is necessary for the diagnosis of Burkitt lymphoma, with excellent outcomes expected in patients fit for intensive chemoimmunotherapy (33). Conversely, MYC translocations convey poor risk in multiple myeloma (34). Despite intense clinical research, the prognostic significance of both MYC translocation and/or MYC overexpression at the protein level remains controversial in DLBCL. It is increasingly evident that it is not MYC overexpression per se that determines DLBCL outcomes, rather its co-occurrence with co-operating lesions (35). Of note, MYC plus BCL2 protein expression in the context of activated B-cell (ABC) subtype disease (“double expressors”) or germinal center B-cell DLBCL with switch translocations of both MYC and BCL2 (“double hit” lymphomas; DHL), correlates with much worse outcomes (36, 37).

Our mechanistic data using genetically defined Eμ-Myc lymphomas confirms other reports of a preclinical efficacy signal for BET inhibitors in the context of MYC dysregulation (4–6). We also now show that engagement of the intrinsic apoptotic pathway by p53-independent mechanisms is a critical determinant of BET inhibitor activity in the lymphoma setting. Our results highlight an important caveat in the extrapolation of mechanistic data using epigenetically active drugs from transgenic animals to human tumors. Specifically, the epigenetic regulation of cMyc (and retrovirally expressed Bcl-2) is divergent between Eμ-Myc lymphomas and human IG-cMYC translocated disease. Of particular interest is that despite complete insensitivity of transgenic cMyc to BET-inhibitor-mediated transcriptional downregulation, the majority of lymphomas remain exquisitely sensitive to JQ1-evoked apoptosis. These data are consistent with the reported co-activator function of BRD4 to MYC target gene expression (3), although our initial analysis of Myc-target genes showed little or no change in expression following JQ1 treatment. We also show that the epigenetic regulation of endogenously expressed BCL-2 family proteins mediates the apoptotic response to BET-inhibition, principally due to a shift in the ratio of Bim to Bcl-2 and Bcl-_XL transcription. A previous study in MLL-rearranged AML demonstrated that BRD4 bound to a locus containing the transcriptional start site of the Bcl-2 gene and BET inhibitor treatment resulted in decreased binding of BRD4, CDK9, and Pol II within this region and loss of phospho-Pol II and H3K4me3 transcriptional activation marks, concomitant with decreased Bcl-2 expression (9). Our data support the notion that Bcl-2 is a critical target gene of BRD4 in tumors where Myc is the primer driver oncoprotein or in situations such as MLL-AF9-driven leukemia where MYC activity is an important secondary event (9). The functional and clinical significance of our observations are further supported by the finding that the antiapoptotic phenotype of constitutive RAS expression mediates primary resistance to JQ1, and that lymphomas progressing on JQ1 therapy variably upregulate BCL-2 family proteins.

Understanding mechanisms of BET-inhibitor activity will assist in the translation of clinical compounds to lymphoma trials. Through this study and others (3), it is apparent that transcriptional downregulation of MYC is not a pre-requisite to BET-inhibitor efficacy in all tumor contexts. Though perturbations of the mitochondrial apoptotic pathway mediate sensitivity to BET-inhibition, it does not follow that the presence of BCL-2 overexpression will always be a necessary predictor of BET-inhibitor responses. Where BCL-2 lies downstream of NFκB activation, for example, in ABC-subtype DLBCL, BRD4 antagonism may downregulate BCL2 transcription downstream of RELA (38). Likewise, translocated BCL-2 expressed on BRD4-loaded superenhancers may be directly targeted by BET-inhibitors in DHL (39). Conversely, MCL-1–dependent lymphomas, or those where BCL-2 is stabilized by RAS pathway activation are likely to resist BET-inhibitor induced apoptosis. Moreover, it is apparent that there are other mechanisms of BET-inhibitor resistance in different tumor contexts. Two recent studies reported concordant data supporting a role of activated Wnt signaling in mediating resistance to BET inhibitor in MLL-rearranged AML (29, 30), whereas another study demonstrated that in triple-negative breast cancer, resistance to JQ1 was conferred by constitutive hyperphosphorylation of BRD4 through decreased activity of PP2A (40).

Despite initial disease control, Eμ-Myc lymphomas progress within a short time-period on single-agent JQ1 therapy. Treatment failure is due at least in part to cell-autonomous mechanisms of resistance to BET inhibitor–evoked apoptosis, as explanted lymphomas exhibited in vitro resistance to JQ1. Moreover, re-transplant of these lymphomas to JQ1-naïve mice resulted in attenuated therapeutic responses in vivo. We therefore predict that single-agent BET-inhibitor treatment outcomes will be limited by rapid emergence of resistant disease in patients. The risk of BET-inhibitor resistance could be mitigated by treatment in low disease-burden states (e.g., post-remission induction), or more likely by the rational incorporation of drug combination therapy. Preliminary preclinical data suggest BET inhibitors may be effective in combination with BH3-mimetics directly engaging the mitochondrial pathway, shown in the context of DHL, or drugs mitigating prosurvival signaling downstream of RAS (e.g., PI3K or MEK), shown in the context of T-cell acute lymphocytic leukemia (41, 42).

Taken together, this study details the apoptotic proteins and pathways engaged by JQ1 to induce death of lymphoma cells, and provides clear evidence that induction of apoptosis is critical for the therapeutic effects of JQ1. Importantly, the molecular and biological responses to JQ1 occurred independently of Myc downregulation and were not reliant on an intact p53 pathway. We demonstrate that increased expression of BCL-2 through activation of Nras or following initial exposure to JQ1 is sufficient to induce drug resistance. These studies provide important information regarding the mechanisms of action of, and mechanisms of resistance to JQ1 that are important for further clinical development of BET protein inhibitors.

J.E. Bradner is President, NIBR at Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.J. Hogg, A. Newbold, B.P. Martin, J. Shortt, R.W. Johnstone

Development of methodology: S.J. Hogg, A. Newbold, B.P. Martin, M. Lefebure, J.E. Bradner, J. Shortt

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.J. Hogg, A. Newbold, S.J. Vervoort, L.A. Cluse, G.P. Gregory, M. Lefebure, E. Vidacs, R.W. Tothill

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.J. Hogg, A. Newbold, S.J. Vervoort, B.P. Martin, G.P. Gregory, M. Lefebure, R.W. Tothill, J.E. Bradner, J. Shortt

Writing, review, and/or revision of the manuscript: S.J. Hogg, A. Newbold, B.P. Martin, G.P. Gregory, R.W. Tothill, J. Shortt, R.W. Johnstone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Newbold, B.P. Martin, E. Vidacs

Study supervision: J. Shortt, R.W. Johnstone

We thank members of the Gene Regulation laboratory and Prof Mark Dawson (Peter MacCallum Cancer Centre) for helpful comments, advice, and suggestions.

This work was supported by funding from the Leukemia Foundation of Australia (to S.J. Hogg, A. Newbold, G.P. Gregory, and M. Lefebure), the Cancer Therapeutics CRC (to S.J. Hogg and G.P. Gregory), the Royal Australasian College of Physicians (to G.P. Gregory), and the Arrow Bone Marrow Transplant Foundation (to M. Lefebure). J. Shortt is supported by funding from the Eva and Les Erdi/ Snowdome Foundation Victorian Cancer Agency Fellowship, and the Cancer Council Victoria. R.W. Johnstone is a senior principal research fellow (APP1077867) of the National Health and Medical Research Council of Australia (NHMRC) and is supported by an NHMRC program grant (APP454569), the Cancer Council Victoria and the Victorian Cancer Agency.