BH3-only proteins in apoptosis at a glance

In mammalian cells, there are two pathways to apoptosis (Hotchkiss et al., 2009; Strasser et al., 2011) (see Poster): the ‘extrinsic’ pathway, which is triggered by engagement of cell surface ‘death receptors’ of the tumour necrosis factor receptor family with their ligands, and the ‘intrinsic’ (or mitochondrial) pathway, which is triggered by diverse cellular stresses, such as cytokine deprivation, DNA damage or oncogene activation. These pathways are largely independent (Strasser et al., 1995), each activating different initiator caspases, but they converge at the level of effector caspases.

The intrinsic pathway is controlled by interactions between proteins related to BCL2, the oncoprotein activated by chromosome translocation in follicular lymphoma (Tsujimoto et al., 1985). BCL2 inhibits apoptosis (Vaux et al., 1988), as do four close homologues – BCL-X_L, BCL-W, MCL1 and A1 – which share four BCL2 homology (BH) domains (see Poster). Other relatives that are similar in sequence and structure – BAX, BAK (and possibly BOK) – instead promote apoptosis, as do distant relatives (BIM, PUMA, NOXA, BID, BMF, BAD, BIK and HRK), which have only the BH3 domain in common with BCL2 or each other. Induction of apoptosis requires the activation of members of both of these death-promoting families (Cheng et al., 2001; Zong et al., 2001). Exposure to stress results in the induction of BH3-only proteins, which neutralise the pro-survival proteins. Subsequent activation of BAX and BAK, which involves their conformational change and homo-oligomerisation on the mitochondrial outer membrane, leads to its permeabilisation. The consequent release of cytochrome c into the cytoplasm promotes the formation of the apoptosome complex, which activates initiator caspase-9.

The ∼26-residue BH3 domain is an amphipathic α-helix that binds with high affinity to a hydrophobic groove on the surface of pro-survival BCL2 family members (Sattler et al., 1997). Individual BH3-only proteins have varying affinities for different pro-survival proteins (Opferman et al., 2003; Chen et al., 2005; Kuwana et al., 2005; Certo et al., 2006; Ku et al., 2011). Whereas BIM, tBID (the cleaved active form of BID; see below) and PUMA bind with high affinity to all pro-survival proteins, others are more selective; for example, BAD binds with high affinity only to BCL2, BCL-X_L and BCL-W, and NOXA only to MCL1 and A1. Importantly, the ‘promiscuous’ binders are more potent killers than the more selective binders, but co-expression of ‘selective binders’ that have complementary specificity (e.g. BAD plus NOXA), kills cells as potently as the ‘promiscuous’ BIM (Chen et al., 2005). This suggests that efficient apoptosis requires neutralisation of all pro-survival BCL2 family members expressed within a given cell.

However, there is still vigorous debate on how the BH3-only proteins trigger activation of BAX and BAK (Chipuk et al., 2010; Strasser et al., 2011). The ‘direct activation’ model (Letai et al., 2002; Kuwana et al., 2005; Chipuk et al., 2010) (see Poster) proposes that certain BH3-only proteins termed ‘activators’ – for example BIM, tBID and perhaps PUMA (Cartron et al., 2004) – bind transiently with low affinity to BAX and BAK (‘hit and run’), thereby triggering their conformational change and subsequent oligomerisation on the mitochondrial outer membrane. According to this model, the other BH3-only proteins, termed ‘sensitisers’ (for example, BAD, BIK, HRK), bind only to pro-survival BCL2 family members, but in doing so they liberate any complexed ‘activator’ BH3-only proteins to activate BAX or BAK. Recently, the distinction between ‘activator’ and ‘sensitiser’ BH3-only proteins has become more blurred, with claims that NOXA (Dai et al., 2011) and BMF (Du et al., 2011) can act as direct activators.

The ‘indirect activation’ model (see Poster) posits that the primary role of all BH3-only proteins is to bind to pro-survival BCL2 family members, thereby preventing them from binding to and neutralising any activated BAX or BAK molecules (Chen et al., 2005; Willis et al., 2005; Willis et al., 2007; Uren et al., 2007). This model envisages that activation of a BAX or BAK protein is the result of a conformational change that exposes its BH3 domain, which, if not bound by a pro-survival BCL2-like protein, would catalyse self homo-oligomerisation (see Dewson and Kluck, 2009). The two models are not mutually exclusive; indeed, they might act in parallel during developmentally programmed death of haematopoietic cells (Merino et al., 2009). Variations of these models have also been proposed (Chipuk et al., 2010; Lovell et al., 2008; Strasser et al., 2011).

BH3-only proteins are regulated by diverse transcriptional and post-translational mechanisms (see Poster). Loss of individual BH3-only proteins, except BIM, results in a relatively normal phenotype in the absence of stress conditions (see Table in Poster), indicative of functional overlap. Different cytotoxic conditions induce distinctive, albeit overlapping, expression patterns that can vary with cell type.

There are three major splice variants of BIM (BIM_S,BIM_L and BIM_EL), the most potent being BIM_S, which is extremely hard to detect in vivo (O’Connor et al., 1998). Transcription of Bim is reportedly induced by the forkhead transcription factor FOXO3a, in response to cytokine withdrawal (Dijkers et al., 2000); by CHOP, a member of the C/EBP transcription factor family, during endoplasmic reticulum (ER) stress (Puthalakath et al., 2007); and by MYC, possibly indirectly (Egle et al., 2004). Bim transcription is repressed by the miR-17-92 microRNA cluster (Ventura et al., 2008), which in turnis repressedbyglucocorticoids (Molitoris et al., 2011). BIM_EL and BIM_L are also thought to be regulated post-translationally by sequestration to the microtubule-associated dynein motor complex (Puthalakath and Strasser, 2002). Phosphorylation of BIM can either be activating or inactivating, depending on the site. Phosphorylation by extracellular signal-regulated kinase (ERK) lowers the affinity of BIM for MCL1 and BCL-X_L, and targets it for ubiquitylation and proteasomal degradation (Ewings et al., 2007). By contrast, phosphorylation by Jun N-terminal kinase (JNK) increases the pro-apoptotic activity of BIM (Lei and Davis, 2003; Putcha et al., 2003). Dephosphorylation by protein phosphatase 2A (PP2A) following ER stress also enhances its activity (Puthalakath et al., 2007).

BIM is prominent in lymphoid cells, but is also readily detectable in many other cell types (O’Reilly et al., 2000). Bim^–/– mice exhibit prominent lymphoid hyperplasia, indicating that BIM is a major regulator of lymphoid homeostasis (Bouillet et al., 1999). BIM is crucial for the deletion of autoreactive thymocytes (Bouillet et al., 2002) and B cells (Enders et al., 2003), and for the termination of acute (Pellegrini et al., 2003) and chronic (Hughes et al., 2008) immune responses. Bim^–/– mice develop excessive antibody levels which, depending on the genetic background, can lead to autoimmune kidney disease (Bouillet et al., 1999). Remarkably, loss of even only a single allele of Bim can prevent the fatal polycystic kidney disease that afflicts Bcl2^–/– mice (Bouillet et al., 2001), indicating that BCL2 and BIM are the principal regulators of apoptosis in developing nephrons.

Loss of BIM increases the resistance of many haematopoietic cell types to diverse cytotoxic conditions, including cytokine deprivation and exposure to abnormal calcium flux, taxol and glucocorticoids (Bouillet et al., 1999). Although BIM is not a direct transcriptional target of the DNA damage response protein P53, loss of BIM renders normal (Bouillet et al., 1999; Erlacher et al., 2005) and transformed (Happo et al., 2010) lymphoid cells more resistant to DNA-damage-inducing agents, including chemotherapeutics. BIM co-operates with PUMA to mediate apoptosis triggered by cytokine deprivation and exposure to γ-irradiation, dexamethasone and ER stressors (Ekoff et al., 2007; Erlacher et al., 2006).

Noxa (Oda et al., 2000) and Puma (Nakano and Vousden, 2001; Yu et al., 2001) are both direct transcriptional targets of p53. However, although lymphocytes and fibroblasts lacking PUMA are highly resistant to γ-irradiation and DNA-damage-inducing chemotherapeutics (Jeffers et al., 2003; Villunger et al., 2003a), those lacking NOXA remain relatively sensitive (Shibue et al., 2003; Villunger et al., 2003b). Conversely, loss of NOXA renders fibroblasts and keratinocytes more resistant to UV radiation than does loss of PUMA (Naik et al., 2007). Thus, different DNA lesions might modulate the target gene preferences of P53.

E2F1 also targets both Noxa and Puma for transcription (Hershko and Ginsberg, 2004) and the hypoxia-inducible factor HIF1α targets Noxa (Kim et al., 2004). Puma transcription is induced when cells are deprived of cytokines or treated with glucocorticoids, staurosporine or phorbol esters (Ekoff et al., 2007; Erlacher et al., 2005; Jeffers et al., 2003; Kieran et al., 2007; Michalak et al., 2008; Villunger et al., 2003b). NOXA is reportedly inhibited by cytosolic sequestration following phosphorylation by cyclin-dependent kinase 5 (CDK5) (Lowman et al., 2010).

Unlike most BH3-only proteins, which are relatively unstructured until they are bound to pro-survival proteins (Hinds et al., 2007), BID is structurally similar to BAX and BAK (Chou et al., 1999; McDonnell et al., 1999). Full-length BID is expressed in most tissues, but is inactive until cleaved proteolytically. Because BID can be cleaved by caspase-8, its active form, tBID, can link the extrinsic and intrinsic apoptosis pathways (see Poster). Hepatocytes from Bid^–/– mice are resistant to killing through the death receptor FAS, but their lymphocytes remain sensitive (Kaufmann et al., 2007; Yin et al., 1999), implying that apoptosis signal amplification through the mitochondrial pathway is essential for killing in certain cell types, but not in others.

There are two major isoforms of BMF, produced from different initiation sites on a common transcript (Grespi et al., 2010). BMF can be negatively regulated by sequestration to the cytoskeleton through binding to the myosin V motor complex (Puthalakath et al., 2001). Although studies in cell lines implicated BMF in detachment-mediated apoptosis (anoikis) (Puthalakath et al., 2001), endothelial cells and fibroblasts from Bmf^–/– mice remain sensitive (Labi et al., 2008), suggesting that other BH3-only proteins are also triggered by anoikis. BMF activity might be regulated by JNK phosphorylation (Lei and Davis, 2003). Bmf^–/– mice develop progressive B cell hyperplasia (Labi et al., 2008), indicating that BMF regulates lymphoid homeostasis. Lymphocytes lacking BMF have increased resistance to dexamethasone and to inhibitors of histone deacetylases (HDACs) (Labi et al., 2008), which prevent transcription by promoting chromatin compaction.

Phosphorylation of BAD by the protein kinase AKT during cytokine signalling reportedly results in its binding and sequestration by 14-3-3 scaffold proteins (del Peso et al., 1997). Bad^–/– mice appear normal (Ranger et al., 2003), except for a mild thrombocytosis (Kelly et al., 2010) that is probably caused by reduced inhibition of BCL-X_L in the absence of BAD (Mason et al., 2007). Mouse embryonic fibroblasts (MEFs) lacking BAD remain sensitive to many apoptotic stimuli (Ranger et al., 2003). Combined loss of BIM and BAD enhances the survival of activated B cells that are deprived of cytokines and accelerates γ-radiation-induced thymic lymphomagenesis, which is indicative of functional overlap (Kelly et al., 2010).

BIK was the first BH3-only protein to be identified (Boyd et al., 1995; Chittenden et al., 1995). Bik transcription can be activated by E2F1 and P53 (see Chinnadurai et al., 2008) and Bik activity is postulated to be inhibited by phosphorylation mediated by casein kinase II (CSNK2) (Verma et al., 2000). Although BIK is widely expressed, Bik^–/– mice display no overt abnormalities and Bik^–/– haematopoietic cells and fibroblasts remain sensitive to diverse apoptotic stimuli (Coultas et al., 2004). Defects in spermatogenesis in Bik^–/–Bim^–/– mice suggest that BIK and BIM share the role of eliminating supernumerary germ cells (Coultas et al., 2005).

HRK activity is enhanced by JNK-mediated phosphorylation (Ma et al., 2007). HRK expression is restricted to the central and peripheral nervous systems and contributes to apoptosis of certain neuronal cells induced by cytokine deprivation (Coultas et al., 2007; Imaizumi et al., 1997). Although HRK expression in human hematopoietic cells has also been reported (Inohara et al., 1997), loss of HRK has no impact on mouse haematopoietic cells (Coultas et al., 2007).

BH3-only proteins as tumour suppressors Inhibition of apoptosis is a fundamental step in tumour development (Hanahan and Weinberg, 2000) and, conversely, induction of apoptosis might be important for tumour suppression. Mutations crippling genes encoding BH3-only proteins have been detected in a range of human cancers (see Table in Poster) and direct tests using mice in which certain genes have been knocked out have provided experimental evidence that certain BH3-only proteins can indeed serve as tumour suppressors.

BIM was the first BH3-only protein shown to be a tumour suppressor. Using Eμ-Myc transgenic mice, in which Myc expression is driven by the immunoglobulin heavy chain (IgH) enhancer, modelling the chromosome translocation in Burkitt lymphoma (Adams et al., 1985), loss of only a single allele of Bim was found to accelerate MYC-driven B lymphomagenesis (Egle et al., 2004). Loss of BIM also accelerates renal cancers induced in mice by mutant Ras (Tan et al., 2005). Homozygous deletion of BIM is evident in many mantle cell lymphomas (Tagawa et al., 2005) and silencing of BIM by promoter methylation and mutation is common in B cell lymphomas (Mestre-Escorihuela et al., 2007). Certain Burkitt lymphomas carry point mutations in MYC that prevent MYC from activating BIM (Hemann et al., 2005). The miR-17-92 cluster, which represses BIM expression, is often overexpressed or amplified in human cancers (Xiao et al., 2008).

As a result of P53 mutations, over 50% of human tumours exhibit impaired ability to induce PUMA or NOXA in response to genotoxic damage or oncogenic stress (for examples see Table in Poster). Loss or knock-down of PUMA accelerated B lymphomagenesis in Eμ-Myc mice (Garrison et al., 2008; Hemann et al., 2004; Michalak et al., 2009) and loss of NOXA accelerated γ-radiation-induced thymic lymphomagenesis (Michalak et al., 2010). However, Noxa^–/–Puma^–/– mice do not develop tumours (Michalak et al., 2008) and, although the combined loss of PUMA and NOXA accelerated MYC-driven lymphomagenesis, the effect is not as great as that provoked by loss of P53 (Michalak et al., 2009). Therefore, induction of apoptosis through PUMA and NOXA constitutes only one mechanism by which P53 suppresses tumour development; others include cell cycle arrest, DNA repair and cellular senescence. Loss of BMF or BAD also accelerates lymphomagenesis in mouse model systems (Labi et al., 2008; Frenzel et al., 2010; Kelly et al., 2010).

BH3-only proteins play a crucial role in the killing of tumour cells by anti-cancer agents (see Strasser et al., 2011). Efficient killing of MYC-driven lymphomas in mice by DNA-damaging agents requires the P53 targets PUMA and NOXA, and also, unexpectedly, BIM (Happo et al., 2010). BIM (in concert with PUMA) is essential for inducing apoptosis of tumour cells following glucocorticoid treatment (Molitoris et al., 2011), MEK inhibition (Abrams et al., 2004; Cragg et al., 2008) or microtubule stabilisation by paclitaxel (Puthalakath et al., 1999; Sunters et al., 2003). Inhibitors of oncogenic kinases such as imatinib, which targets BCR–ABL in chronic myeloid leukaemia (CML) cells, require BIM activity and in certain settings also that of BAD and BMF (Costa et al., 2007; Cragg et al., 2008; Cragg et al., 2007; Gong et al., 2007; Kuroda et al., 2006). BIM and BMF activity mediate the sensitivity of lymphoid malignancies to glucocorticoids (Ploner et al., 2008) and HDAC inhibitors (Lindemann et al., 2007). BMF enhances the apoptotic response of BCL2-overexpressing lymphomas to combined treatment with the HDAC inhibitor Vorinostat and the BH3 mimetic ABT-737 (see below) (Wiegmans et al., 2011). BID contributes to responsiveness to HDAC inhibitors (Lucas et al., 2004) and proteasome inhibitors (e.g. bortezomib) (Duechler et al., 2005), and bortezomib induces NOXA, BIM and BIK in a variety of human cancer cells (Fennell et al., 2008).

Small-molecule mimics of BH3-only proteins are being developed for cancer therapy. ABT-737 (Oltersdorf et al., 2005) and its orally active derivative ABT-263 (Tse et al., 2008) are both bona fide BH3 mimetics because they kill cells through a BAX- and BAK-dependent mechanism (reviewed by Lessene et al., 2008). Both compounds bind with high affinity to BCL2, BCL-X_L and BCL-W, but do not interact with MCL1 or A1 (Oltersdorf et al., 2005). They are relatively well tolerated, despite the transient thrombocytopaenia caused primarily by inhibition of BCL-X_L, which is the limiting factor for platelet survival (Mason et al., 2007; Zhang et al., 2007). In preclinical studies, ABT-737 was efficacious as a single agent for treating certain leukaemias and lymphomas (Del Gaizo Moore et al., 2007; Del Gaizo Moore et al., 2008; Konopleva et al., 2006; Oltersdorf et al., 2005; Shoemaker et al., 2008; Tse et al., 2008; Vogler et al., 2008), multiple myelomas (Chauhan et al., 2007; Kline et al., 2007) and small cell lung cancers (Oltersdorf et al., 2005), but not for epithelial cancers (Huang and Sinicrope, 2008; Witham et al., 2007). High levels of MCL1 confer resistance to ABT-737, but this could be overcome by reducing MCL1 levels (reviewed by Lessene et al., 2008). ABT-737 can synergise with conventional anti-cancer agents, as well as with targeted inhibitors of oncogenic kinases (reviewed by Cragg et al., 2009). ABT-263, currently in early Phase I–II clinical trials, has shown promising efficacy in chronic lymphocytic leukaemia (CLL) patients (Gandhi et al., 2011; Tse et al., 2008; Wilson et al., 2010).

BH3-only proteins are essential for initiating apoptosis through the mitochondrial pathway. Elucidating which of the pro-apoptotic BH3-only proteins are essential for the killing of malignant cells by conventional and targeted anti-cancer agents can provide valuable insights into why certain therapies succeed in some patients but fail in others. In the era of BH3 mimetics, this knowledge will help tailor the use of rational combination therapy and increase therapeutic efficacy whilst limiting collateral damage to normal tissues.

A.S. and S.C. are supported by the National Health and Medical Research Council [grant number 461221, Australia Fellowship (AS)]; the National Institutes of Health [grant number CA43540]; the Leukemia and Lymphoma Society [grant number SCOR 7413]; and operational infrastructure grants through NHMRC IRISS and the Victorian State Government OIS. L.H. has a Leukaemia Foundation Australia scholarship. Deposited in PMC for release after 12 months.