Deciphering the Molecular Events Necessary for Synergistic Tumor Cell Apoptosis Mediated by the Histone Deacetylase Inhibitor Vorinostat and the BH3 Mimetic ABT-737

Histone deacetylase inhibitors (HDACi) can induce tumor cell apoptosis, inhibit cell proliferation by blocking progression through the G1/S or G2/M cell cycle checkpoints, induce cellular differentiation, suppress angiogenesis, and modulate antitumor immunity (1). Using genetic mouse models of cancer, we and others have showed that direct tumor cell killing by these agents is essential for their antitumor responses in vivo (2, 3). Some studies have indicated that apoptosis mediated by HDACi proceeds through the extrinsic (death receptor) pathway following transcriptional upregulation of tumor necrosis factor superfamily ligands and/or receptors (2, 4–7). However, in the majority of cases including all studies that we have performed, activation of the intrinsic (mitochondrial) apoptosis pathway plays a fundamental role in mediating HDACi-induced cell death (3, 8–12).

The intrinsic apoptosis pathway is regulated by pro- and antiapoptotic Bcl-2 family proteins that consists of 3 major subgroups: (i) multidomain prosurvival proteins (Bcl-2, Bcl-x_L, Bcl-w, Mcl-1, A1); (ii) BH3-only proapoptotic proteins (Bid, Bim, Bik, Bmf, Noxa, Puma, Hrk, Bad); and (iii) multidomain proapoptotic proteins (Bax, Bak, Bok; ref. 13). ABT-737 is a BH3 mimetic compound developed to specifically inhibit the activity of prosurvival Bcl-2 family proteins (14). Using competitive fluorescence polarization assays and recombinant Bcl-2 family proteins, ABT-737 was shown to bind with relatively high affinity to Bcl-2, Bcl-x_L, and Bcl-w and to have much lower capacity to bind to and inhibit Mcl-1 and A1 (14). Our subsequent cell-based assays confirmed that ABT-737 was a potent inhibitor of Bcl-2 and Bcl-x_L, and a poor inhibiter of Mcl-1 and A1, but surprisingly was only a weak inhibitor of overexpressed Bcl-w in Eu-myc lymphoma cells (15).

Although HDACi show great promise as cancer therapeutics, given their diverse molecular and biological antitumor activities and their manageable clinical side effects, these agents may be best utilized in the clinic in combination with existing chemotherapeutics or with novel small molecule or biological agents in a targeted manner (16). We previously showed that overexpression of prosurvival Bcl-2 proteins inhibits the apoptotic and therapeutic activities of HDACi (3, 17) and combination treatment with ABT-737 restored the efficacy of HDACi against tumors overexpressing Bcl-2 and Bcl-x_L (15). Although these studies clearly established that combining HDACi and ABT-737 resulted in more potent antitumor activities in vitro and in vivo, the molecular events necessary for synergistic tumor cell apoptosis mediated by the combination treatment had not been elucidated.

Herein we show that the HDACi, vorinostat, “primes” tumor cells overexpressing Bcl-2 for rapid apoptosis following treatment with ABT-737 by inducing expression of certain BH3-only genes. We determined that upregulation of bmf, but not bim, by vorinostat was functionally critical to prime cells for apoptosis mediated by ABT-737 in vitro and in vivo. Highly expressed Bmf was bound by Bcl-2 at baseline, however, the addition of ABT-737, which avidly binds to Bcl-2, resulted in release of Bmf from Bcl-2, allowing subsequent interaction between Bmf and Mcl-1. Synergistic apoptosis induced by vorinostat and ABT-737 was suppressed in cells with mutated p53, although the induction of BH3-only genes by vorinostat in the absence of a functional p53 pathway was unaffected. The functional importance of Bmf and p53 in mediating a robust response to vorinostat and ABT-737 was shown in cells with knockout or knockdown of bmf and mutated p53 that displayed a severely attenuated response to the vorinostat/ABT-737 combination. We showed that basal expression of the p53 target gene, noxa, and to a lesser extent, puma, was reduced in cells with mutant p53. Finally, knockout of noxa in p53-competent, Bcl-2–overexpressing cells also suppressed the apoptotic response to the combination of vorinostat and ABT-737.

These studies provide important mechanistic insight into the molecular events that underpin synergistic tumor cell death mediated by HDACi combined with ABT-737. Bmf was a common transcriptional target of structurally diverse HDACi and induction of this gene was critical for the combined effects of vorinostat and ABT-737. This defines a novel functional relationship between Bmf and ABT-737 in killing cells overexpressing Bcl-2 and highlights the concept of “apoptotic priming” of tumor cells by an anticancer agent for death mediated by a second agent. In this instance, vorinostat-induced induction of bmf “primed” tumor cells overexpressing Bcl-2, for subsequent death mediated by ABT-737. However, the cells were additionally sensitized to ABT-737–mediated apoptosis through the expression of wild-type p53, likely through the maintenance of adequate basal levels of noxa and puma.

Vorinostat (Merck), panobinostat (Novartis), romidepsin (Gloucester), VPA (Sigma-Aldrich), etoposide, taxol, vincristine (Peter MacCallum Cancer Centre), ABT-737 (Abbott Park), were dissolved in dimethyl sulfoxide (DMSO). For in vivo experiments vorinostat and ABT-737 were prepared as previously described (15).

Development of Eμ-myc, Eμ-myc/bcl-2, Eμ-myc/mcl-1, Eμ-myc/bim^−/−, Eμ-myc/bmf^−/−, Eμ-myc/noxa^−/−, Eμ-myc/apaf-1^−/−, and Eμ-myc/p53^−/− lymphomas has been previously described (3, 15, 18–20). MSCV retroviral plasmids for expression of Bcl-2 and A1 were described previously (15). For gene knockdown studies, shRNAmirs in the pLMP vector directed against p53 (kindly provided by Dr Ross Dickens, Walter and Eliza Hall Institute), bmfshRNA1 (sense—AGCAAGCCAGGGTGAAACTTAA), bmfshRNA2 (sense—CGCCCAGAGTAAGGAATGTCTT), and IFI16 (sense—CTCATATCAGATTATTTGGAAT) were used. Retroviral transduction of Eμ-myc lymphoma cells and Western blotting was performed as previously described (3). Blots were probed with α-Bcl-2 (BD Sydney), α-Bim (Stressgen), α-Bmf (Alexis), α-p53 (Novocastra), α-acetylated-Histone H3 (Upstate), and α-ß-actin (Sigma) used as a control. Lymphomas were also analyzed for p53 pathway status by treatment with 20 nmol/L etoposide Western blotting with α-p53 (Novocastra) and α-p19^ARF Invitrogen and cell death as measured by propidium iodide (Sigma-Aldrich) uptake and annexin V-APC (DB Biosciences) binding.

Eμ-myc/bcl-2 (1 × 10⁶ cells) were incubated for 8, 16 hours with 2.5 μmol/L vorinostat then washed with media and incubated and for a further 1, 2, 4, and 8 hours with 0.1 μmol/L ABT-737 or 0.1 μmol/L ABT-737e (an enantiomer of ABT-737) or 4 and 8 hours with ABT-737 or ABT-737e treatment alone. Eμ-myc lymphoma cells (1 × 10⁶ cells/mL) were also incubated in the presence of the indicated compounds for 24 hours. Cell death was analyzed by propidium iodide (Sigma-Aldrich) uptake and annexin V-APC (BD Biosciences) as described (3, 15). Colony assays were performed as previously described (17). Synergy was calculated using the Biosoft CalcuSyn program based on the Chou-Talalay method (21).

For immunoprecipitation, Bcl-2 and Mcl-1 antibodies were bound to Protein G Sepharose 4 Fast Flow-beads (Amersham) and cross-linked using 40 mmol/L dimethyl pimedilate in 0.1 M borate (pH 9). Eμ-myc/bcl-2 cells were treated with 0.5 μmol/L ABT-737 for 8 hours. Lysates were prepared on ice in lysis buffer (20 mmol/L Tris-HCl pH 7.4, 135 mmol/L NaCl, 1.5 mmol/L MgCl₂, 1 mmol/L EGTA, 1 mmol/L EDTA, 1% Triton X-100) supplemented with Complete protease inhibitors (Roche). Equilibrated antibody-bound beads were incubated with 500 μg protein at 4°C over night. After 4 washes with lysis buffer, the beads were boiled in SDS-sample buffer and analyzed by Western blotting.

1 × 10⁶ cells were harvested in 250 μL of TRIzol reagent (Invitrogen), RNA was isolated using an Invitrogen isolation method and cDNA synthesis was performed according to manufacturers instructions (Promega). Quantitative PCR analysis of samples was performed on ABI7900 light cycler with SYBER-green ROX mix (Thermo Scientific-Abgene) with primers spanning exons of the BH3-only family of proteins utilizing ribosomal housekeeping gene RPL32. Primer sequences for the complete BH3 family set can be requested from the author.

Approximately 5 × 10⁴ Eμ-myc/bcl-2 cells were injected intravenously (i.v.) into the tail veins of 5-week-old male C57Bl/6 mice. Mice were monitored and treated when an average total white cell blood count of over 13 × 10⁶/mL as assessed by Bayer ADVIA 120 hematology analyzer (GMI Ramsey) showed leukemia. Therapy consisted of administration of vorinostat (200 mg/kg) for 12 hours and ABT-737 (100 mg/kg) for 4 hours later or each drug independently along with a 16-hour vehicle control.

Vorinostat and ABT-737 combination therapy was performed on 7 × 10⁴ 4T1 breast carcinoma cells were harvested and injected subcutaneously (s.c.) into the flanks of 4-week-old female Balb/c mice. Once the tumor reached 9 mm², daily simultaneous intraperitoneal (i.p.) administration of 200 mg/kg vorinostat and 150 mg/kg ABT-737 began. The mice were treated and monitored for 14 days. Eμ-myc/bcl-2 tumors from the lymph nodes of treated mice were subjected to immunohistological staining with hematoxylin/eosin and TUNEL analysis as previously described (17).

The chromatin immunoprecipitation (ChIP) assay was performed as per the manufacturer's instructions using a cell signaling EZ ChIP kit (Upstate). Cross-linked protein/DNA was incubated with either 1 μg of isotype control antibody or 5 μg of antiacetylated histone H3 (Upstate) for 16 to 24 hours at 4°C. The immuno-precipitated DNA was analyzed by PCR with primers targeting the promoter regions of specific BH3-only genes (available from author).

Previously we showed that coincubation of HDACi and ABT-737 induced apoptosis in Bcl-2–overexpressing lymphomas derived from Eμ-myc transgenic mice (Eμ-myc/bcl-2; ref. 15). We hypothesized that HDACi may sensitize or “prime” cells for ABT-737–induced apoptosis. Treatment with vorinostat (2.5 μmol/L) alone for 24 hours had little or no effect on the viability of Eμ-Myc/bcl-2 lymphomas whereas treatment with ABT-737 (0.5 μmol/L) caused less than 20% of the lymphoma cells to undergo apoptosis (Supplementary Fig. S1A–D). However, pretreatment of Eμ-Myc/bcl-2 lymphomas with vorinostat (2.5 μmol/L) potently sensitized the tumor cells to ABT-737–induced apoptosis (Supplementary Fig. S1A–D). The synergistic effects of vorinostat and ABT-737 required effective inhibition of Bcl-2, as an enantiomer of ABT-737 with significantly less Bcl-2-inhibitory activity (14) did not synergize with vorinostat (Fig. 1A). We confirmed that a second independently derived Eμ-Myc/bcl-2 lymphoma was also sensitized to ABT-737–induced apoptosis following pretreatment with vorinostat (2.5 μmol/L) for 16 hours (Supplementary Fig. S1E).

We hypothesized that vorinostat primed Eμ-Myc/bcl-2 lymphomas for ABT-737–mediated apoptosis by increasing the expression of proapoptotic BH3-only genes. Treatment of Eμ-myc/bcl-2 lymphomas with vorinostat alone or vorinostat and ABT-737 resulted in induction of 3 BH3-only genes, bmf, bim, and noxa, more than 2-fold above baseline (Fig. 1B and C). Interestingly, bmf (∼12-fold), noxa, and bim (both ∼3.5-fold) were induced to maximal levels following single agent treatment with vorinostat (Fig. 1B). In contrast, hrk was maximally induced (∼100-fold) after 16 hours pretreatment with vorinostat followed by 8 hours incubation with ABT-737 (Fig. 1C). Only moderate induction of hrk (6-fold) was observed when Eμ-myc/bcl-2 lymphoma cells were treated with ABT-737 alone for 8 hours and none of the other Bcl-2 family genes showed consistent changes in expression following single agent ABT-737 treatment (Fig. 1B and C). Changes in the expression of Bmf, Bim, and Hrk protein mimicked the magnitude and kinetics of their transcriptional profiles (Fig 1D). All detectable Bmf isoforms showed increased expression following treatment with vorinostat (Fig. 1D, lanes 1–3), although there was a 1 hour delay in maximal protein expression compared to the kinetic change in bmf mRNA (compare with Fig. 1B). In contrast, there was no observable change in Bmf levels following treatment with ABT-737 alone (Fig. 1D). Similarly, the expression of Bim_EL, Bim_L, and Bim_S protein isoforms increased over time in response to vorinostat in the presence and absence of ABT-737 (Fig. 1D). Densitometric analysis revealed that Bmf protein levels were induced maximally 11-fold above baseline, whereas BimL was induced 8-fold above baseline. Maximal levels of Hrk were observed following pretreatment with vorinostat for 16 hours and an additional incubation with ABT-737 for 4 to 8 hours (Fig. 1D, lanes 6 and 7). Changes in Noxa protein could not be evaluated by Western blotting, as a suitable antibody specific for endogenous mouse Noxa is not currently available. Expression of the BH3-only family member Bad and multidomain proapoptotic member Bax did not change during treatment with vorinostat and/or ABT-737 (Fig 1D). These results indicate that bmf, bim, and noxa are genes that are up-regulated during pretreatment with vorinostat and hence the protein products of these genes may “prime” Eμ-Myc/bcl-2 lymphoma cells for rapid apoptosis mediated by ABT-737.

To determine if bmf, bim, noxa, and hrk levels were altered following treatment with vorinostat in vivo, tumor and normal cells were harvested from the same lymph nodes of vorinostat-treated C57BL/6 mice bearing GFP^+ve Eμ-Myc lymphomas. Vorinostat induced a robust and early induction of bmf in Eμ-Myc lymphomas whereas little or no change was observed in normal cells (Supplementary Fig. S2A). More modest tumor cell-specific induction of bim and noxa was also observed following treatment of mice bearing Eμ-Myc lymphomas (Supplementary Fig. S2B and C). Maximal induction of hrk was observed at later time points (Supplementary Fig. S2D), consistent with the notion that transcriptional activation of this gene occurs as a consequence of apoptosis. Western blot analysis confirmed expression of Bmf in Eμ-myc/bcl-2 lymphomas isolated from the lymph nodes of vorinostat-treated tumor-bearing mice (Supplementary Fig. S2E).

To confirm that vorinostat-mediated upregulation of BH3-only genes was not a response to induction of apoptosis, Eμ-myc/bcl-2 cells were treated with vorinostat and ABT-737 in the presence of the pan-caspase inhibitor zVAD-fmk. Apoptosis mediated by vorinostat/ABT-737 was inhibited by zVAD-fmk (Supplementary Fig. S3A), however bmf, bim, and noxa mRNA still increased following treatment with vorinostat (Supplementary Fig. S3B). Of note, the induction of hrk following treatment with vorinostat and ABT-737 was suppressed compared to the level of gene induction seen in cells that underwent apoptosis following exposure to both agents (compare Figs. 1C and S3B). Similar results were obtained using Eμ-myc lymphomas overexpressing the prosurvival family member Bcl-2A1 (A1) that we had previously shown to be resistant to vorinostat/ABT-737 combination treatment (15), as ABT-737 does not inhibit A1 (refs. 22, 23; Supplementary Fig. S3C and D). Taken together these results suggest that vorinostat primes cells for ABT-737–mediated death by specifically inducing expression of certain BH3-only genes, namely bmf and to a lesser extent bim and noxa. In contrast, the upregulation of hrk following treatment with vorinostat and ABT-737 appears to be a secondary response triggered by apoptotic signaling and hrk is therefore unlikely to be important for vorinostat-mediated priming for death by ABT-737.

To determine if the transcriptional activation of selective BH3-only genes by vorinostat was due to drug-induced hyperacetylation of their promoter regions, ChIP assays were performed. Eμ-myc/bcl-2 cells were treated with vorinostat for 16 hours and ChIP assays were performed using an anti-acetylated-histone H3 antibody. Treatment with vorinostat resulted in very strong induction of histone H3 acetylation at the bmf promoter and weaker acetylation of histone H3 at the bim promoter (Supplementary Fig. S4A and B). This suggests that up-regulation of bmf and bim following treatment with vorinostat occurs as a direct consequence of changes in the levels of acetylation at the promoters of these target genes. Additional studies showed that vorinostat-mediated induction of bmf, bim, and noxa mRNA was not inhibited by continuous cotreatment of cells with cyclohexamide to inhibit de novo protein synthesis (data not shown). Together, these data indicate that vorinostat induces direct hyperacetylation of histone H3 within the promoter regions of genes such as bmf and bim resulting in their transcriptional upregulation.

To determine the functional importance of Bmf and Bim for the synergistic killing of cells overexpressing Bcl-2 by vorinostat and ABT-737, we generated Eμ-myc/bim^−/− (3) and Eμ-myc/bmf^−/− (19) lymphoma cells with overexpression of Bcl-2. Exposure of Eμ-myc/bcl-2, Eμ-myc/bim^−/−/bcl-2, and Eμ-myc/bmf^−/−/bcl-2 lymphomas to the vorinostat/ABT-737 combination revealed that Eμ-myc/bim^−/−/bcl-2 cells were as sensitive as Eμ-myc/bcl-2 cells whereas Eμ-myc/bmf^−/−/bcl-2 lymphomas displayed significantly less apoptosis (Fig. 2A). To confirm the important role of bmf in mediating the synergistic apoptotic effects of vorinostat and ABT-737, Eμ-myc/bcl-2 lymphomas were transduced with retroviral vectors expressing 2 different short hairpin RNAs (shRNA) directed towards bmf or an shRNA against the human Ifi16 gene as a control. Both shRNAs targeting bmf robustly inhibited the expression of the proapoptotic Bmf I protein isoform (Supplementary Fig. S5). Exposure of Eμ-myc/bcl-2 lymphomas expressing the IfI16 shRNA (Eμ-myc/bcl-2/Ifi16shRNA) to vorinostat and ABT-737 resulted in apoptosis in these cells that was very similar in magnitude and kinetics to that observed in Eμ-myc/bcl-2 cells (Fig. 2B). In contrast, knockdown of Bmf by 2 different shRNAs in Eμ-myc/bcl-2/bmfshRNA1 and Eμ-myc/bcl-2/bmfshRNA2 cells resulted in resistance to the combination treatment (Fig. 2B). These results confirm that Bmf plays an important role in mediating robust apoptosis of Eμ-myc/bcl-2 lymphomas upon treatment with vorinostat and ABT-737.

Given the important functional role of Bmf in priming vorinostat-treated cells for ABT-737–induced apoptosis, we wished to determine if other HDACi also specifically up-regulated bmf. Treatment of Eμ-myc/bcl-2 cells with the HDACi panobinostat (LBH589), valproic acid (VPA), romidepsin (depsipeptide), and vorinostat resulted in a robust increase in bmf mRNA that was substantially greater than the response of any other BH3-only gene assessed (Fig. 2C). In contrast, other apoptotic stimuli did not significantly induce bmf (Fig. 2D). Consistent with previous reports (24, 25), treatment with the DNA damaging agent etoposide, exposure to UV irradiation, and deprivation of serum resulted in enhanced expression of the p53-response gene puma (Fig. 2D) whereas noxa was induced following etoposide treatment (Fig. 2D) as previously showed (26). Interestingly, treatment of Eμ-myc/bcl-2 lymphomas with concentrations of taxol or vincristine capable of mediating ∼50% death of Eμ-myc lymphomas (data not shown) did not cause any substantial increase in any of the BH3-only genes studied.

We next determined if the requirement for vorinostat-mediated bmf induction for synergistic apoptosis of Eμ-myc/bcl-2 lymphomas in vitro was maintained in the in vivo setting. C57BL/6 mice bearing Eμ-myc/bcl-2, Eμ-myc/bcl-2/bmfshRNA1, Eμ-myc/bmf^−/−/bcl-2, and Eμ-myc/bcl-2/IFI-16shRNA tumors were treated with vehicle, vorinostat, ABT-737, or a combination of vorinostat and ABT-737. In mice bearing Eμ-myc/bcl-2 or Eμ-myc/bcl-2/IFI-16shRNA lymphomas, treatment with vorinostat or ABT-737 alone caused a slight decrease in WBC numbers and spleen weight (Fig. 3A). The single agent effect on WBC numbers and spleen weights was significantly enhanced following combination treatment with vorinostat and ABT-737 (Fig. 3A). Enhanced induction of apoptosis in mice bearing Eμ-myc/bcl-2 lymphomas treated with the vorinostat/ABT-737 combination compared to single agent treatment was confirmed by TUNEL staining on lymph node tissue (Fig. 3B). The synergistic effect of vorinostat/ABT-737 was almost completely abrogated in mice bearing Eμ-myc/bcl-2/bmfshRNA1 and Eμ-myc/bmf^−/−/bcl-2 lymphomas (Fig. 3A).

Having showed that in tumor cells engineered to overexpress Bcl-2, treatment with vorinostat increased the levels of Bmf and this was important to “prime” cells for subsequent killing by ABT-737, we wanted to confirm that vorinostat could act in the same manner in a tumor where endogenous Bcl-2 was highly expressed. We previously determined that the mouse breast carcinoma cell line 4T1.2 was relatively resistant to apoptosis induced by single agent vorinostat treatment in vitro and in vivo (27). Western blot analysis revealed that 4T1.2 cells expressed levels of Bcl-2 comparable to the expression level seen in Eμ-myc/bcl-2 lymphomas (Fig. 4A). Treatment of 4T1.2 cells with vorinostat or ABT-737 alone for 24 hours induced minimal apoptosis however synergistic tumor cell death was apparent following combination treatment with ABT-737 and vorinostat (Fig. 4B). Moreover, a time course assay showed that pretreatment of 4T1.2 cells for 16 hours with vorinostat sensitized the cells to ABT-737–induced apoptosis with similar kinetics and magnitude of cell death to those observed in Eμ-myc/bcl-2 lymphomas (Fig. 4C). Bmf mRNA and protein was strongly induced following treatment of 4T1.2 cells with vorinostat with kinetics of gene induction that mirrored those observed in Eμ-myc cells (Fig. 4D and E). Finally, we showed that the combination of vorinostat and ABT-737 was superior in vivo for the treatment of 4T1.2 tumors compared to single agent treatment (Fig. 4F). These data are consistent with the notion that high levels of endogenous Bcl-2 may confer a level of resistance to vorinostat-mediated apoptosis, which can be overcome with prior incubation of cells in vorinostat, followed by the addition of a BH3-mimetic. Effects of tumor microenvironment and/or differences in local drug concentrations in subcutaneous versus lymphoid may account for the combination effect of vorinostat and ABT-737 being less robust in the 4T1.2 system compared to that seen against Eμ-myc/bcl-2 lymphomas.

Bmf was originally identified in a yeast 2-hybrid screen using Mcl-1 as bait and co-immunoprecipitated with overexpressed Bcl-2, Bcl-x_L, Mcl-1, and Bcl-w in HEK-293 cells (28). BIOCORE analysis using recombinant proteins and Bmf-BH3 peptides indicated that the binding affinity of Bmf was higher for Bcl-2 and Bcl-x_L than for Mcl-1 (29). Consistent with these data, we showed that Bmf co-immunoprecipitated with Bcl-2 whereas little or no Bmf was detected following immunoprecipitation of Mcl-1 (Supplementary Fig. S6). Interestingly, treatment of cells with ABT-737 resulted in a decrease in the amount of Bmf that co-immunoprecipitated with Bcl-2 and detectable Bmf that co-immunoprecipitated with Mcl-1 (Supplementary Fig. S6). This shows that Bmf preferentially binds Bcl-2 over Mcl-1. However, ABT-737, which can functionally interact with Bcl-2 but not Mcl-1 (14, 15, 22, 23) appears to effectively compete with Bmf for binding to Bcl-2, resulting in release of Bmf that is subsequently free to interact with Mcl-1.

During the course of our experiments we identified an Eμ-myc/bcl-2 lymphoma (designated *Eμ-myc/bcl-2) that displayed considerable resistance to apoptosis induced by the vorinostat/ABT-737 combination (Fig. 5A). Bmf was up-regulated in vorinostat-treated *Eμ-myc/bcl-2 cells indicating that a lack of bmf induction was not responsible for the observed resistance to vorinostat/ABT-737 (Fig. 5B). The parental *Eμ-myc lymphoma retained sensitivity to vorinostat-induced apoptosis (Supplementary Fig. S7A), however, these cells were relatively resistant to apoptosis mediated by etoposide (Supplementary Fig. S7B). Eμ-myc lymphomas have a propensity to inactivate the p53 pathway through alterations of the ARF, mdm2, and p53 loci resulting in chemoresistance in these tumors (30, 31). In comparison to etoposide-sensitive Eμ-myc cells, *Eμ-myc cells expressed significant levels of a higher molecular weight form of p53 that was not induced by etoposide (Supplementary Fig. S7C). Moreover, the *Eμ-myc lymphoma expressed constitutively high levels of p19^ARF, a phenotype consistent with deletion or functional mutation of p53 (31, 32). We subsequently sequenced p53 from *Eμ-myc cells and identified an insertion of 33 nucleotides resulting in an 11 amino acid insertion within the C-terminal, oligomerization domain of the protein.

To definitively show that wild-type p53 was required for maximal apoptosis mediated by the ABT-737/vorinostat combination we developed Eμ-myc/p53^−/−/bcl-2 and Eμ-myc/bcl-2/p53shRNA lymphomas (Supplementary Fig. S7C). Apoptosis mediated by the combination of vorinostat and ABT-737 was significantly compromised in Eμ-myc/p53^−/−/bcl-2 and Eμ-myc/bcl-2/p53shRNA lymphomas compared to control Eμ-myc/bcl-2 and Eμ-myc/bcl-2/IFI16shRNA cells (Fig. 5C). However, robust induction of bmf by vorinostat was retained in Eμ-myc/p53^−/−/bcl-2 (Fig. 5D) and Eμ-myc/bcl-2/p53shRNA cells (Fig. 5E) confirming that an intact p53 pathway was not necessary for vorinostat to induce expression of this functionally important BH3-only gene.

To determine if the combination of loss of bmf and p53 had a more pronounced effect on apoptosis induced by the vorinostat/ABT-737 combination than loss of either gene alone, we identified a rare Eμ-myc/bmf^−/− lymphoma (termed *Eμ-myc/bmf^−/−) with a defect in the p53 pathway as evidenced by loss of detectable p53 protein (in the presence or absence of exposure to etoposide) and increased expression of p19^ARF protein (Fig. 6B and C). These cells were relatively resistant to etoposide-induced apoptosis (Fig. 6A) consistent with these cells being defective in p53 function. We subsequently expressed Bcl-2 in *Eμ-myc/bmf^−/− cells and determined the effect of vorinostat combined with ABT-737. As expected, Eμ-myc/bcl-2 lymphomas were highly sensitive to the combination of vorinostat and ABT-737 while *Eμ-myc/bmf^−/−/bcl-2 lymphomas with knockout of bmf and loss of function of p53 were highly resistant to combination treatment (Fig. 6D).

We confirmed that optimum apoptosis induced by vorinostat/ABT-737 required both bmf and p53 using sensitive clonogenic assays. As expected, treatment of Eμ-myc/bcl-2 or Eμ-myc/bcl-2/IFI16shRNA lymphomas with vorinostat and ABT-737 resulted in almost complete loss of clonogenic potential (Fig. 6E). In contrast, the clonogenic potential was partially rescued in Eμ-myc lymphomas with knockdown of bmf (Eμ-myc/bcl-2/bcl-2shRNA) or mutation of p53 (*Eμ-myc/bcl-2) and this rescue was further enhanced in *Eμ-myc/bmf^−/−/bcl-2 lymphomas with knockout of bmf and functional inactivation of p53 (Fig. 6E). These data indicate that both bmf and p53 play important functional roles in mediating robust tumor cell apoptosis following combination treatment with vorinostat and ABT-737.

The BH3-only genes puma and noxa are direct transcriptional targets of p53 (33–35) and both genes are elevated in premalignant B cells from Eμ-myc mice (20). As shown in Figure 7A, the expression of noxa mRNA was significantly diminished in p53-defective *Eμ-myc/bcl-2 lymphomas whereas expression of puma was decreased to a lesser extent. In contrast, expression of bmf and bad was not affected in *Eμ-myc/bcl-2 lymphomas (Fig. 7A). As we had observed that noxa was induced following vorinostat treatment (Figs. 1–3) and given that noxa expression was suppressed in *Eμ-myc lymphomas, we sought to determine the effect of knockout of noxa on apoptosis mediated by the vorinostat/ABT-737 combination. Eμ-myc/noxa^−/− lymphomas (20) transduced to overexpress Bcl-2 displayed biochemical features of wild-type p53 activity (Fig. 7B). Treatment of Eμ-myc/bcl-2 and Eμ-myc/noxa^−/−/bcl-2 lymphomas with vorinostat and ABT-737 in a time-course assay showed that the Eμ-myc/noxa^−/−/bcl-2 displayed a significant level of resistance to the combination treatment comparable to that observed in Eμ-myc/p53^−/−/bcl-2 lymphoma cells (Fig. 7C). This indicates that like bmf, noxa plays an important role in mediating tumor cell apoptosis following combined treatment with vorinostat and ABT-737.

HDACi are a promising class of new anticancer drugs that show single agent clinical activity against a range of hematological malignancies, most notably cutaneous T-cell lymphoma (36). Using syngeneic mouse models of cancer we have shown that overexpression of prosurvival Bcl-2 family proteins inhibits the apoptotic and therapeutic activities of diverse HDACi, including vorinostat, panobinostat, romidepsin, and valproic acid (3, 15, 37, 38). Moreover, we showed that the resistance to vorinostat-induced apoptosis mediated by Bcl-2 or Bcl-x_L could be reversed using ABT-737, a small molecule inhibitor of Bcl-2 proteins (15). Herein, we have used mouse genetics to decipher the apoptotic proteins and pathways necessary for vorinostat and ABT-737 to kill tumor cells overexpressing Bcl-2.

ABT-737 often has weak single agent activity against a range of tumor cell lines even though these cells frequently express high levels of Bcl-2/Bcl-x_L (39). Such cells may be insensitive to ABT-737 due to concomitantly high levels of the prosurvival protein Mcl-1, which is very poorly inhibited by ABT-737, or due to other events such as phosphorylation of Bcl-2 (22, 23). Moreover, it has been recognized that the mere presence of high levels of Bcl-2 or Bcl-x_L does not necessarily mean that such cells are dependent or “addicted” to the activity of these proteins. Indeed, the concept of “apoptotic priming” has been used to explain why certain cells may be either inherently sensitive or resistant to ABT-737 treatment (40). As activation of the intrinsic apoptotic pathway is regulated by the functional interaction between prosurvival Bcl-2 and proapoptotic BH3-only proteins, cells in which prosurvival Bcl-2 family proteins are largely bound by BH3-only proteins are considered to be “primed” for apoptosis mediated by ABT-737 (40). Cells that express prosurvival Bcl-2 proteins at levels that exceed the capacity of endogenous BH3-only proteins to bind them may be considered to be “unprimed” and therefore less sensitive to single agent ABT-737 treatment (40). It therefore follows that an agent capable of increasing the expression or activity of BH3-only proteins, and/or decreasing the expression/activity of prosurvival Bcl-2 family members, may sensitize “unprimed” tumor cells to ABT-737–mediated apoptosis (Supplementary Fig. S8A). Consistent with this model, combining ABT-737 with proapoptotic stimuli such as conventional chemotherapeutic drugs, γ-irradiation and novel “targeted” small molecule and biological agents can induce synergistic apoptosis in tumor cells that are relatively resistant to single agent treatment (39).

HDACi and ABT-737 used in combination can induce synergistic tumor cell apoptosis (15, 41). We have showed that vorinostat-mediated induction of bim may be important for the single agent apoptotic and therapeutic activity of vorinostat (3) and bim was important for the combined activities of suberoyl bis-hydroxamic acid (SBHA) and ABT-737 in human tumor cell lines (41). However, as we clearly showed herein, knockout of bim had no effect on the ability of the vorinostat/ABT-737 combination to kill Eμ-myc/bcl-2 lymphomas even though bim was induced by vorinostat at both the mRNA and protein level. In contrast, knockout or knockdown of bmf impaired synergistic tumor cell apoptosis induced by vorinostat and ABT-737. In Eμ-myc/bcl-2 lymphomas, Bmf was robustly sequestered by Bcl-2 however, upon addition of ABT-737, Bmf was rapidly released from Bcl-2, presumably as a result of avid binding of ABT-737 to Bcl-2. Bmf was then free to interact with Mcl-1 and possibly other prosurvival Bcl-2 proteins (see Supplementary Fig. S8A). As ABT-737 is relatively ineffective against tumors expressing high levels of Mcl-1 (15), the ability of free Bmf to interact with this prosurvival Bcl-2 family member likely plays an important and unexpected role in mediating the rapid death of vorinostat-primed Eμ-myc/bcl-2 lymphomas following addition of ABT-737.

Expression of wild-type p53 was important for tumor cell death mediated by the vorinostat/ABT-737 combination. Loss of function of p53 did not affect vorinostat-induced expression of BH3-only genes, however, the basal level expression of p53-target genes puma, and in particular noxa, was significantly reduced. Noxa has exquisite binding specificity for Mcl-1 whereas Puma has been showed to more broadly bind numerous prosurvival Bcl-2 proteins and additionally interact with, and directly activate, Bax and Bak (40). Accordingly, we posit that decreased expression of puma and noxa through loss of p53 function would likely raise the “apoptotic threshold” that is mediated primarily through the intrinsic pathway (Supplementary Fig. S8B). Clearly p53 has tumor suppressor and apoptotic functions that are independent of puma and/or noxa (42), however, in the context of synergistic apoptosis mediated by vorinostat and ABT-737, loss of noxa alone was sufficient to blunt the apoptotic effects of the combination. The ability of Noxa to selectively interact with Mcl-1 with high affinity (29) likely underpins the important role of noxa in mediating optimal tumor cell death by vorinostat and ABT-737. Indeed, it appears that noxa plays an important role in synergistic apoptosis mediated by ABT-737 and a range of other anticancer agents. For example, NOXA but not BIM was important for synergistic death of human melanoma cells mediated by ABT-737 combined with antimelanoma drug imiquimod even though both BH3-only genes were induced by imiquimod (43).

The requirement for both bmf and p53 to mediate the most potent apoptotic response to the vorinostat/ABT-737 combination was showed using Eμ-myc/Bcl-2 lymphomas with knockout or knockdown of bmf and a concomitant loss of p53 function. Our data are consistent with a model whereby Bmf induction is important for vorinostat to “prime” cells for apoptosis mediated by ABT-737, whereas p53 activity lowers the “apoptotic threshold” by maintaining elevated levels of noxa and possibly puma. These data provide important mechanistic insight into the molecular events necessary for vorinostat to function synergistically with ABT-737. Moreover, our finding that Bmf was more important than was Bim, for priming of Eμ-myc/Bcl-2 lymphoma cells by vorinostat for ABT-737–induced apoptosis was unexpected, given that Bim has broader binding specificity for prosurvival Bcl-2 proteins than does Bmf and, unlike Bmf, Bim has been reported to directly activate Bax and Bak (40). This indicates that it is not simply induction of any BH3-only gene that “primes” cells for apoptosis mediated by ABT-737, but rather specific BH3-only proteins, or combinations of BH3-only proteins may be required for this process. This hypothesis is in agreement with recent studies by Letai and colleagues that described “priming” as specific competitive interactions between 2 subclasses of BH3-only proteins based on their ability to directly bind Bak and Bax (activators) or compete for binding of prosurvival Bcl-2 family members (sensitizers). Letai and colleagues suggested that primed cells undergo apoptosis via specific release of activator BH3-only members allowing oligomerization of Bax and/or Bak (40).

Different ABT-737–sensitizing agents likely engage specific BH3-only proteins to synergize with ABT-737, presumably in a cell-type specific manner. Thus, in different cell types across and within species there are particular requirements for expression of activator BH3-only members such as Bim or sensitizer BH3-only members such as Bmf, with the relative number and levels of expression of BH3-only members from each subclass determining the state of cellular priming. This was apparent when comparing our results with those of Chen and colleagues who showed a requirement for Bim in mediating synergistic apoptosis by ABT-737 and SBHA (41), whereas we showed that Bim was dispensable for apoptosis induced by the vorinostst/ABT-737 combination. In addition, knockdown of NOXA attenuated the combined apoptotic effects of CPT-11 and ABT-737 in HCT16 cells but not in HT-29 cells whereas knockdown of NOXA equivalently suppressed the synergistic effects of bortezomib and ABT-737 in both cell lines (44).

These preclinical studies provide the most comprehensive functional link between tumor genotype and the apoptotic and therapeutic effects of HDACi combined with ABT-737. Our work provides the basis for the rational use of this combination in the clinic and identifies potential biomarkers such as HDACi-mediated induction of bmf and p53 status to putatively predict likely therapeutic efficacy.

R.W. Johnstone: commercial research grant, Merck and Co. The other authors declared no potential conflicts of interest.

We thank Drs David Huang, Andreas Strasser, Jerry Adams, Philippe Bouillet and Suzanne Cory from the Walter and Eliza Hall Institute for helpful advice and Eu-myc mice and cells. We thank Drs Scott Lowe, Ross Dickins, Saul Rosenberg, Steve Elmore, Alex Shoemaker and Victoria Richon for helpful advice and providing reagents. Vorinostat was kindly provided by Merck, ABT-737 was kindly provided by Abbott Laboratories. We would like to thank Dr Jessica Bolden, Ralph Rossi and Claudia Soratroi for discussions and technical help.

R.W. Johnstone is a Principal Research Fellow of the National Health and Medical Research Council of Australia (NHMRC) and supported by NHMRC Program and Project Grants, the Susan G. Komen Breast Cancer Foundation, the Prostate Cancer Foundation of Australia, Cancer Council Victoria, Victorian Breast Cancer Research Consortium, and the Australian Rotary Health Foundation. A. Villunger is supported by the AICR (grant no. 06-440) and the Austrian Science Fund (FWF). R.W. Johnstone received a collaborative research grant from Merck and Co. for research involving vorinostat.

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