The BH3 Mimetic ABT-737 Induces Cancer Cell Senescence

Aberrant expression of the antiapoptotic Bcl-2 family of proteins in human cancer is correlated with poor outcomes after standard chemotherapy. One approach to overcoming this blockade is the development of compounds such as ABT-737, a cell permeable small molecule Bcl-2 family antagonist that is capable of binding to Bcl-2, Bcl-x_L, and Bcl-w, but does not block the activity of Mcl-1, Bfl-1/A1 and Bcl-B (1). This agent functions by displacing BH3 proteins, such as Bim, from these proteins activating Bax and Bak to induce apoptotic cell death (2) and has the ability to kill leukemia and lymphoma (1), multiple myeloma (3), glioblastoma (4) and small cell lung cancer cell lines (5). On the basis of these results an orally active form of this compound, ABT-263, has entered clinical trials (6, 7).

Emerging results show that ABT-737, especially when used to treat solid tumors including lung, prostate, and renal cancers (1, 8), can be ineffective at inducing apoptosis. Resistance to ABT-737 is in part mediated by the elevated expression of Mcl-1 or Bfl-1/A1, which are not capable of binding this compound (9, 10). Carcinomas exhibit elevated levels of endogenous Mcl-1 and are resistant to killing by this agent (8, 11). Alternatively, we have suggested (8) that ABT-737, although not inducing cell death in tumor cells, could instead be an important anticancer agent functioning by regulating gene transcription sensitizing resistant tumor types to alternative therapies. For example, we have shown that application of ABT-737 transcriptionally induces increases in the level of tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) death receptor 5 (8) enhancing cell killing by the addition of the proapoptotic protein TRAIL. Likewise, ABT-737 resistance in glioma and pancreatic cells expressing high levels of Mcl-1 was completely reversed by cotreatment of TRAIL (4, 12).

To investigate in more depth the extent of transcriptional changes induced by ABT-737 in tumor cells resistant to apoptosis, we have treated the renal carcinoma cell line PV-10 and examined changes in mRNA levels by microarray analysis. Gene chip analysis reveals that ABT-737 treatment regulates more than 400 genes many of which are cytokines and chemokines. Interestingly, we find that these gene changes have been previously identified (13–15) as markers of cellular senescence. Senescence is a complex programmed cellular response to multiple stimuli, including DNA damage, oncogene expression, or telomere shortening associated with aging, which is characterized by irreversible growth arrest accompanied by distinct morphologic changes such as an increased number of enlarged and flattened cells (16). On the basis of this initial result, we have further examined the biochemical mechanism regulating the ability of ABT-737 to inhibit tumor cell growth and induce the senescent phenotype.

Human cancer cell lines, including 22Rv1, DU145, A549, WI-38, purchased from the ATCC (authenticated by short tandem repeats single nucleotide polymorphism, and fingerprint analyses), and PV-10 cells (8) provided by Drs. R. Gemmill and H. Drabkin (Medical University of South Carolina, Charleston, SC), were used for senescence studies. Cell lines were cultured according to the supplier's instructions, and maintained at low passage number for less than 6 months. ABT-737 (A-779024.0) and its inactive enantiomer (A-793844.0) were a gift of Abbott Laboratories. Both compounds were dissolved in dimethyl sulfoxide (DMSO; Sigma-Aldrich) at 20 mmol/L and aliquots were stored at −80°C.

Human GIPZ lentiviruses encoding short hairpin microRNAs against ATM (V2LHS_89368) and CAP3 (V2THS_15048 plus V2THS_15049), and a nonsilencing control were purchased from Open Biosystems. Arrest-In Lentiviral expression system (Open Biosystems) was used to establish PV-10 and 22Rv1 cell lines harboring small hairpin RNAs (shRNA). The following retroviral vectors were used in these experiments: pBabePuro-p53 (Addgene), expressing DN-p53 (17), and pLNCX-GFP (Clontech Laboratories Inc.) as previously described (18).

Cells were seeded in 96-well plates or culture dishes and treated with ABT-737. Cell growth was determined by a 3-(4,5-dimethylthiazoyl-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay (Sigma-Aldrich) using a microplate reader. The percentage of viable cells was evaluated by trypan blue exclusion assay as previously described (8). For clonogenic assays, cells were plated at 1,000 cells per well and cultured in medium with 0, 1, 5, 10 μmol/L of ABT-737 for 2 to 3 weeks. The cells were stained with crystal violet (19).

Cell proliferation assays was performed using a bromodeoxyuridine (BrdU) cell proliferation assay kit (CalBiochem). After 24 hours of BrdU labeling, the relative and background signal was assessed by using a fluorescent plate reader (Synergy 2, BioTek).

Cells (1 × 10⁵) were plated and after 24 hours, fresh medium was added. Conditioned medium was collected following treatment and used at a 2:1 dilution in the human interleukin-6 (IL-6) or IL-8 ELISA Kit (R&D Systems) according to the manufacturer's instructions. SA-β-gal activity was determined using a SA-β-gal staining kit from Cell Signaling Technology according to the manufacturer's instruction. Senescent cells were identified as blue-stained cells by standard light microscopy, and a total of 500 cells were counted in 3 random fields on a slide to determine the percentage of SA-β-gal–positive cells.

Western blot analysis was done as described previously (8) and is described in detail in the Supplementary Material and Methods.

The electrophoretic mobility shift assay (EMSA) is described previously (20) and in detail in the Supplementary material.

Real-time PCR, primers and microarray gene expression profiling are performed as described in detail in the Supplementary Experimental Procedures. Dual luciferase assay was performed (8) and is described in detail in the Supplementary materials.

We have identified a group of cancer cell lines (8) that are resistant to the proapoptotic activity of ABT-737. To examine changes in gene transcription after ABT-737 treatment of PV-10 cells, which are a kidney cancer cell line resistant to apoptosis induced by this agent, we carried out microarray analysis using an Affymetrix human gene chip (U133 Plus 2.0 Array) with mRNA from cells either treated with vehicle or ABT-737 for 24 hours. Our analysis of these results revealed that more than 430 genes were significantly changed (>2-fold, P < 0.05) either up or down after exposure to ABT-737 (Fig. 1A). Genes that are regulated more than 4-fold by ABT-737 are shown in (Supplementary Table S1). IL-6, IL-8, and IL-11 and chemokines CXCL2 and CXCL5 and other secreted cytokines/chemokines are induced more than 2-fold (Fig. 1B). Quantitative real-time PCR (qT-PCR) analysis of IL-6 and IL-8 showed that ABT-737, but not its inactive enantiomer, was able to induce changes in mRNA levels confirming the microarray analysis (Fig. 1C). Significant increases in other mRNA transcripts including IL-11, CXCL2, CXCL5, IRF1, and GADD45A were also confirmed by qT-PCR analysis (Supplementary Fig. S1). To further validate gene changes shown in the microarray analysis, we also included genes that are downregulated by ABT-737 treatment. QT-PCR analysis confirmed that ABT-737 inhibited expression of inhibitor of differentiation 2 (ID2) and ID3 transcripts (Supplementary Fig. S1). To determine whether the change in IL-6 and IL-8 mRNA levels is mirrored with a change in the protein produced, we performed an ELISA assays. Secretion of IL-6 and IL-8 protein was both increased in PV-10 cells in response to ABT-737 treatment (Fig. 1D). To examine whether these gene changes were seen in other cell lines after ABT-737 treatment, we studied the effect of ABT-737 on 6 additional cancer lines, 4 renal cancer cell lines, KRC/Y, KV-6, CAKI, and 786-0, and 2 lung cancer cell lines, A549 and HOP-62. In all of these cell lines with only minor exceptions, ABT-737 was able to induce IL-6 and IL-8 and downregulate ID2 and ID3 transcripts (Supplementary Fig. S2). Pathway analysis shows marked changes in genes associated with protein binding, signal transduction, apoptosis in addition to changes in cytokine activity, and inflammatory response mediators.

Gene microarray analysis revealed that ABT-737 induced a significant increases in the transcription factor CCAAT/enhancer-binding protein beta (C/EBPβ; Fig. 1B). Consistent with these results, qT-PCR and Western blot analysis showed significant induction of C/EBPβ mRNA and protein levels in PV-10 cells treated with ABT-737 (Fig. 2A and B). Because the IL-6 promoter contains both C/EBPβ and NF-κB binding sites, we examined the possibility that activation of these 2 transcription factors played an important role in regulatingIL-6 mRNA levels. Previously, we had reported that NF-κB promoter activity is increased by ABT-737 (8) and EMSA show that ABT-737 treatment of PV-10 cells activates NF-κB activity (Fig. 2C). We find that ABT-737 treatment, but not the inactive enantiomer, of PV-10 cells was able to activate both the IL-6 and IL-8 promoters (Supplementary Fig. S3). Using a plasmid in which the IL-6 promoter was cloned in front of a luciferase reporter, mutation of either the C/EBPβ or NF-κB sites (Fig. 2D), blocked the transcriptional activation of this gene by ABT-737. These DNA microarray results show that in a cell line that does not undergo apoptosis after treatment ABT-737 is capable of inducing changes in gene transcription.

The induction of IL-6, IL-8, CXCL2, and CXCL5 and other secreted proteins has been identified (13, 15) as associated with the induction of cellular senescence. Senescent cells secrete multiple factors that alter tissue microenvironments, including the inflammatory cytokines such as IL-6 and IL-8 (13, 15, 21). These cytokines and chemokines function to reinforce the senescent growth arrest through autocrine and paracrine mechanisms (21, 22). The transcription factor C/EBPβ cooperates with IL-6 to amplify the activation of inflammatory mediators, including IL-8 (15).

To test senescence induction, we first examined the effect of ABT-737 on PV-10 cell growth. Long-term exposure of PV-10 cells to ABT-737 induced significant growth inhibition and a G₁/S cell cycle blockade (Fig. 3A; Supplementary Fig. S4). Direct detection of apoptosis with Annexin V staining showed that apoptotic cells were not significantly increased (Supplementary Fig. S4). However, a dose-dependent inhibition of colony formation and an inhibition of BrdU incorporation was seen after the addition of ABT-737 to both PV-10 and 22Rv1 cells (Fig. 3B and C). To test the possibility that cell growth arrest in response to ABT-737 was caused by the induction of cellular senescence, 5 human carcinoma cell lines including PV-10 cells were exposed to ABT-737 followed by staining with SA-β-gal, a marker of senescent cells. PV-10, A549, and 22Rv1 cells stained positively for this senescent marker whereas W1-38 cells, normal human fibroblasts, and DU145 human prostate cancer cells showed no response (Fig. 3D and E). Cells that were positive for SA-β-gal showed a flattened and enlarged morphology that was consistent with senescence. Although the WI-38 cell line was resistant to ABT-737 induced senescence, γ-irradiation (IR) caused senescence-like morphologic changes in WI-38 human fibroblast cells, suggesting that ABT-737 did not induce senescence in all cell lines.

Overwhelming DNA damage leads to apoptosis, but lower levels of damage have been suggested to be as a causal factor in the induction of both cellular senescence and growth arrest (23). To investigate whether ABT-737 induced lower levels of DNA damage, we investigated the ability of this agent to induce DNA double-strand breaks (DSB), followed by the stimulation of ataxia telangiectasia mutated (ATM) kinase-dependent phosphorylation of the histone H2A variant H2AX, γ-H2AX (7). Western blot analysis revealed a time and dose-dependent induction of both ATM phosphorylation and γ-H2AX in PV-10 cells, 22Rv1, and A549 cells (Fig. 4A and B). Similar results were revealed by immunofluorescent staining of PV-10 cells showing that ABT-737 and DNA damaging agents IR or doxorubicin (DXR) caused the induction of γ-H2AX (Supplementary Fig. S5). Recently, chromatin structures associated with modified histones, such as dimethyl-histone H3 at Lys9 (H3K9m), were identified as a critical feature of cellular senescence. During senescence H3K9m increases and contributes to the transcriptional repression of growth promoting genes found in senescence (24, 25). Using an anti-H3K9m antibody to stain treated and untreated cells, immunofluorescence microscopy shows that ABT-737, as well as DXR and IR, induced this histone modification (Supplementary Fig. S5).

To test the role of the ATM protein kinase as a mediator of the induction of senescence by ABT-737, we infected PV-10 and 22Rv1 cells with lentiviruses encoding short hairpin microRNA targeting ATM (shRNAmir-ATM). The initiation and maintenance of senescence-associated cytokine secretion requires the DNA damage response proteins, including ATM kinase, nijmegen breakage syndrome 1, and check point kinase 2 (26). In a pool of PV-10 cells ATM levels were decreased by shRNA, we found that a marked reduction in ATM levels prevented ABT-737 from inducing γ-H2AX (Fig. 4C). Moreover, ABT-737 treatment of cells with lower levels of ATM protein kinase caused a markedly reduced level of SA-β-Gal and senescence-like morphologic changes (Fig. 4D). To examine whether the induction of DSBs followed by ATM activation was the driving force behind the ABT-737–mediated senescence induction of IL-6 and IL-8, we measured levels of IL-6 and IL-8 mRNA in PV-10-shRNAmir-ATM and PV-10-shRNAmir-control cells. QT-PCR analysis showed that ATM knockdown prevented the ABT-737–mediated induction of IL-6 and IL-8 mRNA (Supplementary Fig. S6). Similar results were observed in 22Rv1 prostate cancer cells depleted of ATM. These findings support the concept that activation of ATM by ABT-737-induced DNA damage signals the induction of senescence leading to an increase in IL-6 and IL-8 mRNA.

To examine whether the DNA damage response caused by ABT-737 occurs as a result of activation of the caspase cascade or was secondary to an off-target direct effect of this drug, we treated cells with z-VAD-FMK, a pan-caspase inhibitor. As a positive control, in these experiments we added TRAIL to ABT-737, a combination we have previously showed is highly synergistic and induces caspase activation and apoptosis (8). When PV-10 cells were treated with ABT-737, caspase-3 (p32 precursor) is processed into 1 inactive (p24) and 2 active (p20 and p17) cleavage fragments. The cleaved subunit of caspase-3 generated in response to ABT-737 is insufficient to further process poly(ADP-ribose) polymerase (PARP), a substrate commonly cleaved during apoptosis. However, the γ-H2AX induction by ABT-737 correlated with caspase-3 cleavage, suggesting the possibility that caspase-3 cleavage induced by ABT-737 might be the cause of the DNA damage response (Fig. 5A). Because z-VAD-Fmk is not a specific caspase-3 inhibitor, we depleted caspase-3 levels in PV-10 and 22Rv1 cells with lentiviral infection of shRNAmir against caspase-3, and examined the ability of ABT-737 to induce senescence. Western blot analysis confirmed in both PV-10 and 22Rv1 cells that partial knockdown of caspase-3 expression inhibited the ABT-737 induction of γ-H2AX (Fig. 5B). To initiate DSBs, caspase-3 results in the activation of caspase-activated DNase (CAD) by cleaving its inhibitor ICAD (27). We show that ABT-737 treatment of PV-10 cells induced ICAD cleavage, which was absent in caspase-3 knockdown cells (Fig. 5C). Exposure of ABT-737 to cells with decreased caspase-3 expression induced a reduced percentage of SA-β-gal–positive cells compared with those cells treated with shRNAmir-control (Fig. 5D). QT-PCR analysis also showed that caspase-3 knockdown prevented the ABT-737–mediated induction of IL-6 and IL-8 mRNA (Fig. 5E). These results support the hypothesis that ABT-737 produced insufficient levels of cleaved caspase-3 to drive apoptosis, but sufficient levels to cause the DNA damage-mediated induction of senescence and induction IL-6 and IL-8 (13, 15, 26).

Because reactive oxygen species (ROS) is known to be an inducer of senescence, we evaluated whether treatment of PV-10 and 22Rv1 cells with ABT-737 might increase oxidative stress. ROS levels, as assessed by dichlorofluorescein, after ABT-737 treatment of both PV-10 and 22Rv1 cells were increased (Fig. 6A), and N-acetyl cysteine (NAC), a precursor of intracellular glutathione, prevented this ROS increase. NAC also partially blocked ABT-737–mediated cleavage of caspase-3 and ICAD suggesting that activation of caspase-3 and DNA damage is in part due to the elevation of ROS (Fig. 6B). Consistent with the role of DNA damage in the induction of senescence, NAC treatment also reduced percentages of SA-β-gal–positive cells (Fig. 6C).

Both DNA damage and the induction of senescence are associated with the activation of p53 and the transcription of genes, such as p21 (28). Increases in cell cycle inhibitor p21 and p16 gene transcription are thought to play a critical role in the mechanism by which cells undergo senescence (29, 30). Both PV-10 and 22Rv1 cells when treated with ABT-737 show increased levels of both wild-type p53 and p21 protein levels (Fig. 7A), but no change in p16 mRNA levels (Supplementary Fig. S1A). As predicted from the changes in p53, the upregulation of p21 protein levels occurred at the transcriptional level and the enantiomer had no activity in regulating p21 (Fig. 7B). To examine whether ABT-737—induced increases in p21 levels could account for the growth arrest seen after treatment, we measured the effect of ABT-737 on the activity of the Cdk2 protein, a kinase that plays an important role both in controlling the G₁/S transition, and has a clear impact on the induction of oncogene-induced senescence (31). The Cdk2 kinase was immunoprecipitated from treated cells and incubated with histone H1 as a substrate. We find that ABT-737 treatment inhibits Cdk2 activity in a dose-dependent fashion in both PV-10 and 22Rv1 cells (Supplementary Fig. S7).

To examine whether the p53 pathway is essential for the induction of senescence, we generated a pool of 22Rv1 cells expressing dominant-negative p53 (DN p53; rev. 17). Western blot analysis confirmed that after ABT-737 treatment overexpression of DN p53 caused a substantial reduction in p21 (Fig. 7C) and inhibited the ability of ABT-737 to induce SA-β-Gal (Fig. 7D). As a result of DN p53 expression, ABT-737 induction of increases in the IL-6 and IL-8 mRNA were also significantly reduced (Fig. 7E). To show whether changes in IL-6 and IL-8 are necessary for ABT-737 induction of senescence, C/EBPβ protein was ablated by siRNA-treatment (Supplementary Fig. S8). QT-PCR analysis of these cells with C/EBPβ knockdown showed that the ABT-737–mediated induction of IL-6 and IL-8 mRNA levels was significantly reduced (Supplementary Fig. S8). Similarly, senescence as measured by SA-β-gal staining induced by ABT-737 was also reduced by decreasing C/EBPβ levels (Supplementary Fig. S8). Together, these results indicate a model whereby ABT-737 induces a signaling cascade including changes in ROS/Caspase-3/DDR/p53/p21/IL-6/IL-8 that leads to cancer cell senescence (Fig. 7F).

We show that the addition of ABT-737 to cancer cell lines that do not undergo apoptosis (8) inhibits cancer cell proliferation associated with the induction of the senescent phenotype. The induction of senescence in both normal and cancerous cell types occurs with marked changes in gene transcription and protein secretion (13) including significant secretion of cytokines and chemokines normally associated with the inflammatory response. Likewise, we have found that the application of ABT-737 to PV-10 kidney cancer cells regulated the transcription of more than 400 genes, many of which are associated with an “inflammatory” phenotype (15, 21). We have identified marked increases in IL-6 and IL-8 driven by changes in the levels of the C/EBPβ transcription factor, mimicking defined events delineated in oncogene-induced and other types of senescence (13, 32). The senescent phenotype was further shown by the induction of SA-β-gal staining, a flattened morphology, a marked decrease in BrdU incorporation, and a G₁/S cell cycle blockade. Together these results show that in a solid tumor cell lines cells that do not undergo apoptosis after ABT-737 treatment, this compound is capable of inducing senescence.

We propose that ABT-737, although not inducing apoptotic levels of DNA damage, is capable of inducing low levels of DNA damage as measured by increases in γ-H2AX and ATM phosphorylation. Both the pan-caspase inhibitor z-VAD-FMK and expression of a shRNA specific to caspase-3 blocked the ability of ABT-737 to induce γ−H2AX, showing that activation of this caspase by ABT-737 was essential to the induction of DNA damage. Activation of caspase-3 has been implicated in specific gene regulation, for example, in the repression of CD28 expression in aging lymphocytes (33). More recently, activation of both CAD and caspase-3 have been shown to be necessary for myoblast differentiation and to induce p21 mRNA transcription (34). Caspase activation was sufficient to induce biologic changes, but insufficient to induce apoptosis. Activation of ROS production has been associated both with the regulation of caspase-3 activity and the induction of senescence (35, 36). As suggested by the ability of NAC to partially block both caspase-3 activation and senescence, the addition of ABT-737 to these tumor cells ROS plays a role in low level caspase-3 activation. Similarly, it has been reported that peroxynitrite leads to senescence of red blood cells through caspase-3 activation (37). Very low doses of γ-radiation induce only transient DNA damage that is quickly repaired. Higher doses of radiation are needed for prolonged DNA-damage that then leads to the induction of senescence-associated markers (26), and still higher doses are needed to induce cell death. Finally, the amount and extent of DNA damage has been linked to the ability of cells to increase the secretion of the immune mediator cytokines/chemokines associated with senescence (38), similar to what is seen here.

Increases in p21 is an essential control point in senescence (30) not only because modulation of this protein blocks cyclin-dependent protein kinases, but also because p21 regulates the transcriptional activation of multiple genes necessary for the induction of the senescent phenotype (39). In multiple tumor types, activation of the p53 (40) has been shown to be essential for the induction of senescence. Using overexpression of dominant-negative p53, we show that the p53 pathway is also necessary for ABT-737 senescence in the 22Rv1 prostate tumor cells. Other DNA damaging agents require p53 as an essential mediator of senescence (41). On the basis of our experiments, we hypothesize that treatment of p53-positive solid tumor patients with ABT-737, or its oral counterpart ABT-263, may yield significant stable disease based on the induction of cellular senescence. Future human clinical trials will be needed to evaluate this important possibility.

No potential conflicts of interests were disclosed.

We thank Abbott Laboratories for providing ABT-737. Dr. Myung Lee (University of North Texas Health Science Center) kindly provided pGL3-IL-6 firefly luciferase reporter plasmids including wild-type, deletion mutants of C/EBPβ or NF-κB, and IL-8 firefly luciferase reporter plasmids.

This work received partial funding from DOD W8IXWH-08 (ASK) and 1P30-CA138313 supported the flow cytometry shared resource.

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.