Piercing the armor of hepatobiliary cancer: Bcl-2 homology domain 3 (BH3) mimetics and cell death


  • Justin L. Mott,

    1. Miles and Shirley Fiterman Center for Digestive Diseases, Division of Gastroenterology and Hepatology, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • Gregory J. Gores

    Corresponding author
    1. Miles and Shirley Fiterman Center for Digestive Diseases, Division of Gastroenterology and Hepatology, Mayo Clinic College of Medicine, Rochester, MN
    • Mayo Clinic College of Medicine 200 First Street SW, Rochester, MN 55905
    Search for more papers by this author
    • Fax: 507-284-0762.

  • Potential conflict of interest: Nothing to report.

One of the cardinal features that cells acquire at malignant transformation is the ability to resist apoptosis in an overtly toxic environment. Cancerous cells survive and proliferate despite internal stresses such as gross genetic alterations and external insults including immune attack, relative hypoxia, and competition for nutrients. The evasion of cell death in this environment is as remarkable as it is necessary, and is responsible for the resistance of cancers to existing therapies. However, the mechanisms of resistance are currently being revealed. One protein family, the B cell lymphoma 2 (Bcl-2) family, has emerged as a dominant regulator of apoptosis in cancer cells. In hepatocellular carcinoma, members of this family are dysregulated, not by direct mutations, but as the downstream consequence of perturbed signaling cascades.1 For example, Bcl-xL protein expression is augmented by enhanced Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling.2, 3 Likewise, although information is more limited, the same trend of signaling-induced alterations in Bcl-2 family member expression appears to be true for cholangiocarcinoma.4–6 In this regard, hepatobiliary cancers are like most other malignancies and are addicted to Bcl-2 family proteins. This article discusses specific recent advances in overcoming resistance to apoptosis by antiapoptotic Bcl-2 family proteins, especially as it relates to targeting these proteins for treatment of hepatobiliary malignancies.


Bcl-2, B cell lymphoma 2; BH, Bcl-2 homology domain; COX2, cyclooxygenase 2; JAK/STAT, Janus kinase/signal transducer and activator of transcription.

The Armor

Bcl-2 is the founding member of the antiapoptotic multidomain proteins; its siblings include Bcl-xL, myeloid cell leukemia-1 (Mcl-1), and Bcl-2 related protein A1 (A1/Bfl). The precise mechanism by which these proteins act is still under investigation,7 but some patterns are clear. During apoptotic signaling, mitochondrial dysfunction is a final common pathway. This gateway to cell death is characterized by mitochondrial permeabilization with release of proapoptotic polypeptides into the cytosol including cytochrome c. Released cytochrome c results in activation of downstream effector caspases—cysteine proteases which dismantle the cell. Antiapoptotic Bcl-2 family proteins, which are located on mitochondria, prevent mitochondrial outer membrane permeabilization and cytochrome c release, thus preventing caspase activation and cell death.8 Bcl-2 and Bcl-xL contain four Bcl-2 homology (BH) domains with domains 1-3 forming a hydrophobic binding cleft on the surface of the protein.9 Mcl-1 folds into a similar alpha-helical structure and contains BH1-BH3 domains and a BH4-like helix.10 Antiapoptotic Bcl-2 family members bind to and inhibit proapoptotic Bcl-2 family members through interactions involving this binding groove.7

The Arrows

BH3-only proteins are proapoptotic members of the Bcl-2 family which display sequence homology only with the BH3 domain, a 10-20 amino acid sequence that forms an amphipathic alpha helix. BH3-only proteins are cytoplasmic, but undergo translocation to the mitochondria when activated. Activation of the BH3-only proteins occurs through several diverse pathways, including proteolytic cleavage (Bid [BH3 interacting domain death agonist]),11 phosphorylation (Bim [Bcl-2 interacting mediator of cell death]),12 dephosphorylation (Bad),13 and increased protein synthesis (Noxa14 and Puma [p53 upregulated modulator of apoptosis]15). The BH3-only proteins act as messengers of the death signal, and mobilization of their active form results in activation of the killer Bcl-2 family proteins Bax (Bcl-2 associated x protein) and Bak (Bcl-2 antagonistic killer).16 It is the multidomain (BH1-BH3) proteins Bax and Bak that are responsible for the permeabilization of the outer mitochondrial membrane during cell death. Whether the BH3-only proteins directly engage and activate Bax and Bak17, 18 or whether they act indirectly by neutralizing antiapoptotic Bcl-2 proteins to disinhibit Bax and Bak19, 20 is actively being pursued. Their binding to antiapoptotic proteins is well studied, and occurs within the hydrophobic binding cleft; for instance, Fig. 1A depicts the Bak BH3 peptide bound to Bcl-xL.21 Some BH3-only proteins demonstrate selective binding to antiapoptotic Bcl-2 proteins (Fig. 2A).

Figure 1.

Hydrophobic binding pocket of Bcl-xL. (A) Space-filling projection of Bcl-xL bound to a BH3 peptide (black) derived from the Bak protein generated from data by Sattler et al.21 (1bxl) available at the Protein Data Bank, generated using Protein Explorer (http://proteinexplorer.org74). Bcl-2 homology domains are yellow (BH1), red (BH2), and green (BH3). (B) ABT-737 derivative (black) bound to the hydrophobic binding cleft of Bcl-xL oriented similar to that in (A). Generated from data by Bruncko et al.48 (2O2N, protein data bank I.D. number).

Figure 2.

BH3-only protein and BH3 mimetic binding partners. (A) The hydrophobic binding cleft of Bcl-2 and Bcl-xL are highly similar, and not surprisingly, BH3-only proteins that bind Bcl-2 generally bind Bcl-xL.22 In contrast, Mcl-1 and A1 bind some (Bim, Bid, and Puma) but not all (Bad) of the binding partners for Bcl-2/Bcl-xL. Noxa appears to be selective for Mcl-1 and A1.50 Similarly, BH3 mimetics have been described that indiscriminately bind Bcl-2 family members (Gx15-070, TW-37), whereas ABT-737 is a Bcl-2/Bcl-xL-selective BH3 mimetic. To date, a Noxa-like BH3 mimetic that preferentially binds Mcl-1/A1 has not been described. (B) Schematic illustration of the mitochondrial apoptosis pathway regulated by the antiapoptotic proteins Bcl-2, Bcl-xL, and Mcl-1. The proteins function to prevent mitochondrial membrane permeabilization and subsequent cell death. Combination chemotherapy that targets both Bcl-2/Bcl-xL and Mcl-1 may have value in treating hepatocellular and cholangiocarcinoma. Bcl-xL (1r2d) and Mcl-1 (1wsx) proteins are illustrated as ribbon diagrams.

The Battle

Cancer cells inappropriately resist cell death in the face of toxic stimuli. This resistance commonly is dependent on the Bcl-2 family of antiapoptotic proteins.22 For instance, Bcl-2 is up-regulated in B cell non-Hodgkin's lymphoma as well as the most common and lethal human cancers: non–small cell lung cancer,23 breast, prostate,24 and colorectal cancer (reviewed in Shangary and Johnson25). Bcl-2 overexpression results in development of high-grade lymphoma in the mouse.26 Other antiapoptotic Bcl-2 family members can similarly provide inappropriate apoptosis resistance in cancer, including Bcl-xL24, 27 and Mcl-1.5, 24 Similar to Bcl-2, overexpression of Mcl-128 or Bcl-xL29 induces malignancy in animal models. Indeed, targeting multiple antiapoptotic Bcl-2 family members may be a very efficient means to induce cancer cell death or to sensitize tumors to chemotherapeutic agents.20, 30, 31 Consistently, the proapoptotic BH3-only protein Bim32, 33 and possibly others34, 35 act as tumor suppressors. Alterations in protein expression of antiapoptotic and proapoptotic Bcl-2 family members have been described in hepatobiliary cancers. As summarized in Table 1, Bcl-xL36 and Mcl-137 play significant cytoprotective roles in hepatocellular carcinoma, whereas Mcl-138 plays a dominant role in protecting cholangiocarcinoma cells.

Table 1. Dysregulated Protein Expression of Bcl-2 Family Members in Hepatocellular Carcinoma and Cholangiocarcinoma
 Bcl-2 Family ProteinHepatocellular CarcinomaCholangiocarcinoma
  1. Protein expression of antiapoptotic Bcl-xL and Mcl-1 is often increased in hepatocellular carcinoma, whereas expression of proapoptotic Bid protein is decreased. Decreased Bax protein has also been described in p53-positive hepatocellular carcinomas.71 Note, however, that the majority of hepatocellular carcinomas are p53-negative, and others have found increased Bax protein expression independent of p53 status.70 In cholangiocarcinoma, Bcl-2 and Bcl-xL protein expression is increased less frequently than Mcl-1, which seems to be the dominant regulator of cholangiocarcinoma resistance to cell death. No data are available for Bid protein expression in cholangiocarcinoma and limited data are available on Bax protein expression; both increased and decreased Bax expression has been described.

AntiapoptoticBcl-2not expressed6464, 65
 Bcl-xL↑↑↑36, 6667
 Mcl-1↑↑↑37, 68↑↑↑5, 67
ProapoptoticBid↓↓69no data 
 Bax↑/↓70, 71↑/↓72, 73

Conversely, cancer cells may be under “BH3 stress”, with increased active BH3-only proteins.22 This activation may be the result of genetic insults accumulated during oncogenesis,39 or from oncogene activation40 such as c-myc.41 This BH3 stress may drive the need for increased levels of antiapoptotic Bcl-2 family members to allow the tumor cells to survive and proliferate. A treatment designed to tip the balance in favor of BH3-only proteins and death signaling may then prove to be of value in treating some malignancies.42 Not surprisingly, there is much interest in antagonizing the Bcl-2 multidomain antiapoptotic proteins for potential therapeutic application in cancer, representing a new drug class: BH3 mimetics.


Several molecules with the capacity to bind the hydrophobic binding cleft of antiapoptotic Bcl-2 family members thus, mimicking BH3 peptide binding, have been described. For instance, the molecule HA14-1 was shown to displace a Bak-derived BH3 peptide from Bcl-2 at micromolar concentrations and caused caspase activation and cell death in HL-60 cells at similar concentrations.43 Antimycin A, a mitochondrial respiratory chain inhibitor, was found by screening for compounds that selectively killed Bcl-xL–overexpressing murine hepatocyte cell lines compared to isogenic cells with lower Bcl-xL expression. Antimycin A was predicted to bind within the hydrophobic binding cleft of Bcl-xL and indeed competes for the same binding site as a Bak BH3-peptide. A derivative that does not inhibit mitochondrial respiration, 2-methoxy-antimycin A3, similarly bound Bcl-2 and induced cell death.44 Recently, (−)-gossypol has been shown to bind with high affinity to Bcl-2, Bcl-xL, and Mcl-1.45 A limitation of HA14-1, antimycin A, and gossypol is their toxicity in the absence of Bax or Bak.31 If we take the mechanism of action of these drugs to be BH3 mimetic, then, like BH3-only proteins,46 they should be dependent on Bax and Bak for killing. Thus, the compounds described above likely have “off-target” effects, and high potency BH3 mimetics that specifically act through the Bax/Bak pathway are desirable.

New Ammunition

Significant progress has been made in this area recently as several publications describe novel BH3 mimetics with potential clinical use. The rationally designed ABT-737 is a small molecule that binds within the hydrophobic cleft of Bcl-2, Bcl-xL, and Bcl-w30, 31, 47, 48 and kills in a fashion dependent on Bax and Bak in the nanomolar range.30, 31 ABT-737 induces apoptosis in cancer cells and cell lines as a single agent in some cases, and sensitizes to chemotherapeutic agents paclitaxel,47 cytosine arabinoside (Ara-C), doxorubicin,30 and etoposide.31 Figure 1B shows an acyl-sulfonamide–based ligand similar to ABT-737 bound to the hydrophobic binding cleft of Bcl-2. Interestingly, this BH3 mimetic does not bind and inhibit Mcl-1 or A1/Bfl, thus opening the door to resistance mediated by up-regulation of one of these antiapoptotic proteins;30, 31, 47, 49 failure to bind Mcl-1 is a potential limitation for the use of this drug in hepatocellular and cholangiocarcinoma, both of which are characterized by up-regulation of this Bcl-2 family protein (Table 1). However, perhaps combination therapy may overcome this potential problem (Fig. 2B). For instance, protein levels of the BH3-only protein Noxa are increased by proteasome inhibition, and Noxa specifically binds Mcl-1 and A1.50 Additive effects were seen combining the proteasome inhibitor bortezomib (currently approved by the U.S. Food and Drug Administration as Velcade for the treatment of multiple myeloma) with ABT-737 in multiple myeloma cells.51 Perhaps ABT-737 treatment targeting Bcl-xL along with bortezomib would be synergistic in the treatment of hepatocellular carcinoma. Likewise, treatment with sorafenib (BAY 43-9006), a kinase inhibitor that is approved for the treatment of renal cell carcinoma, results in the loss of cellular Mcl-1 expression in several cell lines, including cholangiocarcinoma cells.52 Sorafenib has a beneficial effect in the treatment of human hepatocellular carcinoma,53 likely in part by down-regulating Mcl-1. A combination of sorafenib plus ABT-737 is effective against non–small cell lung cancer cells49 and similarly could be synergistic in the treatment of hepatocellular carcinoma. Similarly, cyclin-dependent kinase inhibition decreases Mcl-1 protein levels in cholangiocarcinoma cells,38 and inhibition of cyclin-dependent kinases with seliciclib also cooperates with ABT-737 to increase apoptosis.31, 49 Finally, cyclooxygenase-2 (COX-2) acts in cholangiocarcinoma cells54 and hepatocellular carcinoma cells55 to increase Mcl-1 protein expression, and COX-2 inhibition decreases Mcl-1 protein expression54, 55 and sensitizes tumor cells to apoptosis. Combined treatment with a COX-2 inhibitor plus ABT-737 may be of therapeutic interest.

Activity against Bcl-2, Bcl-xL, Bcl-w, and Mcl-1 has been reported for another small molecule, Gx15-070, in the nanomolar range.40 Given its broad range of binding partners, resistance should not be as common. In culture, Gx15-070 has shown synergy with bortezomib,56 human epidermal growth factor receptor-2 (HER2) kinase inhibitors (lapatinib and GW2974),57 and the death ligand TRAIL (tumor necrosis factor–related apoptosis-inducing ligand) in cholangiocarcinoma cells (S.F. Bronk, unpublished observations). Further, Gx15-070 has single-agent cytotoxic activity, as well as additive activity with fludarabine or chlorambucil, against chronic lymphocytic leukemia.58 Gx15-070 killing is reportedly Bax/Bak-dependent, because baby mouse kidney epithelial cells expressing adenovirus E1A and dominant-negative p53 demonstrated caspase activation upon treatment with Gx15-070 when derived from wild-type mice, but not when derived from Bax/Bak double knockout mice.40 This drug has exceptional promise for the treatment of hepatobiliary malignancies.

A third small-molecule inhibitor was designed by starting from (−)-gossypol and making rational modifications based on a comparison of structural docking of (−)-gossypol and a Bim BH3 peptide to Bcl-2;45, 59 this inhibitor is termed TW-37. This agent displaces a Bid BH3 peptide from Bcl-2, Bcl-xL, and Mcl-1 at low micromolar to nanomolar concentrations (Ki values of 320 nM, 480 nM, and 180 nM, respectively45). Importantly, TW-37–induced death appears to be dependent on either Bak or Bax based on short hairpin RNA–mediated silencing of these proteins. Verhaegen et al. achieved compelling silencing of Bax and (separately) Bak in melanoma cell lines and reduced cell death by 35%-50% when TW-37 was combined with the mitogen-activated protein kinase/extracellular signal-regulated kinase inhibitor U0126.59 TW-37 exhibits single-agent killing of PC-3 prostate cancer cells45 and synergizes with U0126 in melanoma cell lines.59 Again, given its ability to target Bcl-xL and Mcl-1, this drug could prove important in treating hepatobiliary malignancies. TW-37 alone induces increased p53 protein levels in melanoma cell lines, and the combination of U0126 and TW-37 is a more potent inducer of p53. Because p53 is an activator of transcription for the BH3-only proteins Noxa and Puma,14, 15 it will be interesting to learn the role of these BH3-only proteins in this model. However, in terms to hepatocellular carcinoma, we note that p53 is frequently disabled in this neoplasm.

Gold Standard

The ideal drug candidate could be expected to exhibit several features: (1) preferential killing of cancer cells compared to noncancer cells; (2) Bax/Bak-dependent apoptosis activity; and (3) binding to and inhibition of multiple antiapoptotic Bcl-2 family members. Selective induction of apoptosis of cancer cells will be necessary to avoid untoward effects in patients. In this respect, ABT-737, Gx15-070, and TW-37 have all shown preferential killing of tumor versus normal cells (peripheral blood mononuclear cells for ABT-737 and Gx15-070, and normal melanocytes for TW-37), with ABT-737 causing a reduction in platelets and lymphocytes.47 Clinical trials with Gx15-070 so far have not revealed unexpected toxicity.60 The other BH3 mimetics are still in preclinical development. From first principles of apoptosis signaling, it seems unlikely that clinically significant hepatic toxicity will be uncovered, because the normal liver is not under BH3 stress. Hematopoietic cells, on the other hand, may require a survival signal for maintenance, possibly explaining the thrombocytopenia observed with ABT-737.47, 61 The requirement for Bax/Bak-dependent killing seems to distinguish the 2 groups of BH3 mimetics described above, because HA14-1, gossypol, and antimycin A all can kill equally well mouse embryonic fibroblasts deficient for both Bax and Bak. Interestingly, the induction of apoptosis by antimycin A may involve an independent, novel mechanism. Antimycin A killing seems to depend on antiapoptotic Bcl-xL rather than proapoptotic Bax/Bak, because Bcl-xL–overexpressing cells are preferentially sensitive to antimycin A compared to Bcl-xL(low) cells,44 and selected mutagenesis of Bcl-xL decreased antimycin A toxicity, but Bcl-xL mutants still protected against staurosporine-induced death.62 In this respect, such a compound could be very relevant to hepatocellular carcinoma, which is characterized by Bcl-xL overexpression. Gx15-070 and TW-37 both compete for binding to Bcl-2, Bcl-xL, and Mcl-1, and ABT-737 has activity against Bcl-2 and Bcl-xL. However, it may not be necessary to target all antiapoptotic members of the Bcl-2 family with a single agent. Combination chemotherapy could be employed in treating hepatobiliary cancers with ABT-737 plus an agent that disables Mcl-1 (Fig. 2B). Also, Mcl-1 protein expression can be silenced by the microRNA mir-29b in cholangiocarcinoma cells.63 Agents which up-regulate mir-29b may be useful in conjunction with ABT-737 in hepatobiliary cancer therapy. As these or other BH3 mimetics enter clinical trials, their therapeutic potential may be realized.