Endogenous and imposed determinants of apoptotic vulnerabilities in cancer

Abstract

The intrinsic apoptosis pathway is controlled by the BCL-2 family of proteins. Although the pro-survival members of this family can help cancer cells evade apoptosis, they may also produce apoptotic vulnerabilities that can potentially be exploited therapeutically. Apoptotic vulnerabilities can be driven by endogenous factors including altered genetics, signaling, metabolism, structure and lineage or differentiation state as well as imposed factors, the most prominent being exposure to anti-cancer agents. The recent development of BH3 mimetics that inhibit pro-survival BCL-2 family proteins has allowed these apoptotic vulnerabilities to be targeted with demonstrable clinical success. Here, we review the key concepts that are vital for understanding, uncovering, and exploiting apoptotic vulnerabilities in cancer for the potential improvement of patient outcomes.

Understanding mitochondrial apoptosis

Intrinsic (mitochondrial) apoptosis is an evolutionarily-conserved cell death pathway that is critical for normal development, maintenance of tissue homeostasis, and cancer prevention [1]. This pathway is controlled by the BCL-2 family of proteins (see Glossary), which contains both pro-apoptotic and pro-survival members. To trigger apoptosis, pro-apoptotic, BH3-only “activator” proteins (BIM or BID) activate the pro-apoptotic pore-forming proteins (BAX or BAK), which oligomerize to cause mitochondrial outer membrane permeabilization (MOMP) (Figure 1). MOMP releases apoptogenic factors into the cytosol such as cytochrome c, which binds APAF-1 and activates caspases to dismantle the cell and prepare it for phagocytosis [2, 3]. To balance this process, pro-survival proteins (BCL-2, BCL-X_L, MCL-1, etc.) can bind and sequester both classes of pro-apoptotic proteins to maintain cellular survival. However, when these pro-survival proteins are inhibited, they release the bound pro-apoptotic proteins, which can then go on to cause MOMP – this establishes the therapeutic index for BH3 mimetics, as described in more detail below. Curiously, MOMP has also been shown to be triggered in cells lacking known pro-apoptotic BH3-only activator proteins when all pro-survival proteins are inhibited (PMID: 27056669).

Figure 1. The mitochondrial pathway of apoptosis.

To trigger apoptosis, cellular stress or damage signals (1) produce pro-apoptotic proteins (BH3-only “activators” of apoptosis) via their upregulation (e.g., BIM or PUMA) or cleavage (BID cleaved to form active, truncated tBID) (2), which can either be sequestered by pro-survival proteins such as BCL-2, BCL-X_L or MCL1 (3) or, when these pro-survival proteins are saturated or not expressed, can activate the pore-forming proteins BAX and/or BAK (4). Activated BAX or BAK oligomerize and form pores to cause mitochondrial outer membrane permeabilization (MOMP), triggering the release of apoptogenic molecules including cytochrome c, SMAC, and OMI/HtrA2 from the mitochondrial intermembrane space. Cytochrome c complexes with APAF1 in the cytosol to form the apoptosome (5), which activates initiator caspase 9, which then goes on to activate the effector caspases 3 and 7 (6) to proteolytically dismantle cellular structures and prepare the cell for phagocytosis. Caspase activation can be blocked by XIAP (7), which in turn is inhibited by the released SMAC and OMI/HtrA2 proteins from mitochondria (7). Upstream damage or stress signaling can also activate BH3-only “sensitizer” proteins that cannot efficiently activate BAX and BAK but inhibit the activity of pro-survival BCL-2 family proteins to release any bound and sequestered BH3-only activators, which can then activate BAX or BAK to trigger MOMP (8). BH3 mimetics are small molecule drugs that mimic the activity of BH3-only sensitizers and can trigger apoptosis outright in apoptotically dependent cells or enhance sensitivity to other anti-cancer agents in unprimed cells by blocking the activity of pro-survival BCL-2 family proteins (9).

The abundance of pro-apoptotic and pro-survival proteins varies between cells, which alters their baseline apoptotic sensitivity and, consequently, their sensitivity to cellular damage and stress signaling. To illustrate, a cell that maintains its survival by expressing just enough pro-survival proteins to barely buffer endogenous pro-death signals is “primed” for apoptosis (Figure 2). In contrast, a cell that expresses a surplus of pro-survival proteins that can adequately buffer against existing and even potentially additional pro-death molecules, is “unprimed.” Finally, cells that do not express sufficient levels of BAX or BAK to initiate apoptosis or express an overwhelming abundance of pro-survival proteins are considered “apoptosis refractory.” Apoptotic priming can be assessed using a functional assay, BH3 profiling, which measures cytochrome c release in response to titrated doses of pro-apoptotic BH3 peptides [4–7]. Apoptotic priming is a critical determinant of cellular responses to chemotherapy or radiation, both in cancer cells [6, 7] as well as healthy tissues [8]. Furthermore, as described below, cells that are primed for apoptosis may be dependent on pro-survival proteins that are actively sequestering pro-death signals.

Figure 2. A model of how mitochondrial apoptosis can be targeted directly or indirectly to induce apoptosis.

(A) Cells that contain a low reserve of unbound anti-apoptotic proteins in their pretreatment state (prior to administration of any agents) but also lack pro-apoptotic protein expression are considered to be primed for apoptosis. These cells are typically sensitive to cytotoxic chemotherapies that cause cell stress or damage and upregulation of pro-apoptotic BH3-only proteins but are resistant to BH3 mimetics. (B-C) Cells that contain a low reserve of unbound anti-apoptotic proteins (primed for apoptosis) and (B) BCL-2 or (C) MCL-1 molecules that are actively binding and sequestering activator BH3-only proteins such as BIM, BID or PUMA, are sensitive to specific BCL-2 or MCL-1 inhibition, respectively. These cells are typically also sensitive to cytotoxic chemotherapies that cause cell stress or damage. (D) Cells that contain a high reserve of unbound anti-apoptotic proteins (unprimed) can buffer stress-induced pro-apoptotic signals and are therefore resistant to cytotoxic chemotherapies and BH3 mimetics. If additional pan inhibitor or cytotoxic chemotherapy is administered, however, MOMP could be ultimately triggered. Note that the cells in D that are experiencing cell stress or damage, perhaps due to chemotherapy treatment, would now be sensitive to additional pro-apoptotic factors as well as BH3 mimetics targeting BCL-2 or MCL-1.

Cells experience a myriad of changes while undergoing malignant transformation, including derangement of cellular growth signals, oncogenic activation, loss of microenvironmental support signaling, cell cycle checkpoint violation, ongoing DNA damage, nutrient deprivation, ROS signaling, hypoxia, and many others [9, 10]. Many of these have been previously linked directly to the increased production of pro-apoptotic signaling. For example, growth factor deficiency can cause upregulation of pro-apoptotic BIM, PUMA, and BMF [11, 12] while nutrient deprivation can upregulate pro-apoptotic Noxa [13]. This production of pro-apoptotic signaling can induce apoptosis in cells during the process of transformation and act as an anti-cancer defense mechanism. Cells that have pre-existing buffers against apoptosis, such as increased expression of pro-survival proteins BCL-2, BCL-w, BCL-X_L, BFL-1 (A1), and MCL-1, can evade transformation associated apoptosis and form tumors. However, this may come at the cost of subsequently requiring those pro-survival proteins to actively sequester their pro-apoptotic counterparts – a perilous balance that equates to an apoptotic vulnerability that can be therapeutically exploited by BH3 mimetics, a class of small molecule drugs that potently and selectively inhibit pro-survival BCL-2 family proteins by blocking their interactions with proapoptotic BCL-2 family members. Given that the transformation-associated cell stress presumably does not dissipate, the apoptotic vulnerability continues to be present in the growing tumor. The twin concepts of heightened apoptotic priming and increased dependence on pro-survival BCL-2 family proteins in cancer cells likely underlies the striking clinical success of venetoclax, a BH3 mimetic that selectively inhibits BCL-2, in the context of multiple hematological malignancies [14–16]. There are currently many agents targeting pro-survival BCL-2 family proteins in development and their specificity, selectivity and clinical progress has recently been reviewed elsewhere [17, 18].

The recent development and deployment of BH3 mimetics targeting diverse pro-survival BCL-2 family proteins with variable selectivity profiles has created major opportunities for targeting survival proteins in cancer. However, the full and timely clinical exploitation of these powerful agents requires a complete understanding of which mimetics should be used in which cancer contexts. Here, we summarize what is known about both the endogenous determinants of apoptotic dependencies, which include tumor genetic, signaling, metabolic, and structural features as well as lineage or differentiation states, as well as how these dependencies can be further shaped by drug therapies. By defining the determinants of apoptotic vulnerabilities, it will be possible to maximize the clinical impact of BH3 mimetics.

Endogenous determinants of apoptotic vulnerabilities

The endogenous apoptotic vulnerabilities that exist within a cancer cell at baseline (without further perturbations) are known to be driven by numerous factors including the cell’s lineage, differentiation state, oncogene and tumor suppressor signaling pathways, mitochondrial structure, and metabolism, which in turn regulate BCL-2 family protein expression, localization, and protein–protein interactions (Figure 3). The net integration of these and likely other less known factors produces the vulnerability states that dictate responses to BH3 mimetics.

Figure 3. Major endogenous and imposed determinants of apoptotic priming.

Key endogenous and imposed factors can affect mitochondrial priming and vulnerabilities in normal and cancer cells. These include endogenous factors such as the presence of activated oncoproteins or inactivated tumor suppressor proteins, the altered structure and metabolic functions of tumor mitochondria, and cellular differentiation states. Additionally, drugs can act as exogenous factors shaping apoptotic priming, either by altering signal transduction directly or by selecting for resistant cells with altered apoptotic dependencies. EMT, epithelial-mesenchymal transition; NK cell, natural killer cell; RTK, receptor tyrosine kinase. Image created with BioRender.com.

Cell lineage

The most consistent apoptotic vulnerability that has been detected and exploited thus far is the BCL-2 dependence in several hematologic malignancies including chronic lymphocytic leukemia (CLL) and acute myeloid leukemia (AML) [15, 16]. These dependencies can, evidently, be quite strong given that CLL patients can remain on treatment with the BCL-2 inhibitor venetoclax for many years without developing resistance. Cells from the hematopoietic system seem to have the highest expression of BCL-2 and B cells, from which CLL arises, seem to express consistently high levels of this protein [19]. However, the positive expression of a pro-survival protein does not necessarily constitute actionable dependence since cells must be primed for apoptosis in order to have a significant degree of dependence on pro-survival proteins (Figure 2). Further, all other things being equal, higher expression of BCL-2 may decrease venetoclax sensitivity, as more drug is required to neutralize BCL-2 activity. Another key factor in the success of venetoclax in the treatment of CLL is the consistent lack of expression of other pro-survival proteins that may buffer against BCL-2 inhibition [20]. Similar to CLL, follicular lymphomas also strongly overexpress BCL-2, which is due to the t(14;18) translocation of the BCL-2 gene and the immunoglobulin heavy-chain promoter [21], yet patients with this disease have only modest sensitivity to venetoclax [22]. This is likely due to the presence and buffering capacity of pro-survival proteins BCL-X_L and MCL-1, which have been shown to be upregulated in this disease [23].

Although there is variation in expression levels of BCL-2 family members in cells of the hematopoietic system, these cells seem to express higher levels of BCL-2 specifically and are thus predisposed to being dependent on this protein. This led to hematologic cancers being explored first for therapy with BCL-2 inhibitors with demonstrated success. BCL-2 inhibitor trials in solid cancers have reported mixed results, often with only modest responses evident in subsets of patients. This is consistent with cancer cell line studies demonstrating that only subsets of cell lines from solid malignancies such as breast [24] and lung [25] carcinomas are sensitive to BCL-2 inhibition. The modest activity of BCL-2 inhibitors in solid tumors is believed to be principally due to the higher expression of other pro-survival proteins in these cancers, especially BCL-X_L [26–28]. However, BCL-X_L inhibitors and even dual-targeting BCL-2 and BCL-X_L inhibitors have not been as successful in solid tumors [29–31], perhaps suggesting they may also experience lower overall apoptotic pressure. Consistent with this concept, studies have shown that solid cancers are less primed for apoptosis than hematologic malignancies [7] and that overcoming this apoptotic resistance in solid tumors requires alternative strategies such as co-treatment with BCL-X_L inhibitors along with direct activators of pro-apoptotic BAX [32]. Beyond the hematopoietic system, BCL-2 is also expressed in certain cells within the female reproductive system as well as endocrine and thyroid glands [33]. The cancers that arise from these tissues may potentially be sensitive to BCL-2 inhibition. BCL-X_L is expressed at higher levels in neural and epithelial cells [34, 35], while MCL-1 is enriched in certain hematopoietic cells including plasma cells [36], certain types of epithelial cells, enterocytes, and microglia [37, 38]. The heightened expression of MCL-1 in plasma cells is likely linked to the reported MCL-1 dependence in the plasma cell malignancy multiple myeloma [39, 40]. The model coming into focus is that cell lineage is a critical factor driving expression of pro-survival BCL-2 family proteins and may predispose cancers arising from varying precursors to develop lineage-driven apoptotic vulnerabilities. Surprisingly, our understanding of apoptosis regulation in healthy tissues is more limited than in cancers. Thus, the degree to which BH3 mimetics beyond venetoclax will be tolerated clinically in pediatric and adult patients is not yet clear, a fact which underscores the importance of both a deep understanding of the determinants of apoptotic vulnerabilities as well as the development of therapeutic strategies that selectively sensitize tumor cells to BH3 mimetics, topics further elaborated later.

Differentiation and maturation stage

While lineage-associated regulation of apoptosis helps establish initial set points for expression of BCL-2 family proteins, the differentiation or maturation state of normal or cancer cells can modulate their expression. For example, apoptotic priming is dynamically regulated in normal cells during postnatal development, with young tissues that are growing or remodeling being more primed for apoptosis compared with adult tissues [8, 41–43], making young tissues consistently more sensitive to ionizing radiation or chemotherapy than their mature counterparts [8, 44–46]. This is consistent with differentiation state being a powerful arbiter of apoptotic sensitivity in both healthy and cancerous cells. It is likely that cancers deriving from more mature and apoptosis-resistant progenitors may exhibit reduced levels of pro-survival protein dependence. Indeed, pediatric malignancies seem to be more sensitive to anti-cancer therapies than their counterparts in adults, which may be linked to differences in apoptotic sensitivity of the cancers that arise from differently primed cells of varying maturation stages. It has also been shown that the differentiation state of AML cells affects venetoclax sensitivity, with phenotypically primitive AML being sensitive to BCL-2 inhibition while monocytic AML exhibits reduced expression and dependence on BCL-2 [47]. Similarly, T-cell acute lymphoblastic leukemias of the early T cell progenitor subgroup, which are at high risk for relapse, are dependent on BCL-2 while those exhibiting features of more mature T cells are instead dependent on BCL-X_L [48].

The plasticity of differentiation states can also modulate apoptotic vulnerabilities. For example, the epithelialmesenchymal transition (EMT), a cellular program through which cell–cell and cell–substrate interactions are remodeled, leading to epithelial cell detachment and adoption of a mesenchymal cell fate, is critical for the progression of certain solid tumors [49]. Among other properties, the EMT confers on cancer cells a broad ability to resist cell death following treatment with cytotoxic and targeted chemotherapies [49]. Thus, it is reasonable to assume that EMT induction impacts the regulation of the apoptotic signaling network in cancer cells. Indeed, recent studies have begun to demonstrate how this may occur. For example, the EMT transcription factor ZEB1 can directly bind to the BIM promoter to suppress its transcription, inhibiting the apoptotic response to EGFR inhibitors in mesenchymal EGFR mutant lung cancer models [50]. Additional studies have highlighted a particularly critical role for the anti-apoptotic BCL-2 family protein, BCL-X_L. Cells that have undergone an EMT not only appear to exploit BCL-X_L for chemoresistance, but also rely upon it for their survival. For example, one recent study demonstrated that EMT renders cells more dependent upon BCL-X_L through a PERK–NOXA-dependent mechanism [51]. Another recent study corroborated this concept by demonstrating that mesenchymal kidney cancers are dependent upon BCL-X_L, speculating that this dependence may derive from the fact that epithelial shedding triggers cell death via anoikis, and that cancer cells that undergo an EMT may be positively selected due to enhanced BCL-X_L activity in order to avoid anoikis [52], a concept further supported by the observation that TGF-β1 treatment of epithelial cells can promote both apoptosis and EMT [53]. Thus, changes in differentiation state that promote tumor progression may sensitize cancer cells to BCL-X_L inhibition.

Oncogenes and tumor suppressors

BCL-2 family pro-survival proteins are frequently altered in cancer genomes in a manner analogous to classical oncogenes, a fact which provided early support for the idea that evasion of apoptosis is a hallmark property of cancers that is required in order for neoplastic cells to survive in the face of hostile intracellular and microenvironmental stresses characteristic of the disease [54]. For example, the classic discovery that follicular lymphoma cells frequently harbor the aforementioned t(14;18) translocation established that resistance to apoptosis through increased expression of a pro-survival BCL-2 family protein can promote cancer development [2]. More recently, cancer genome characterization projects have extended this concept, for example through the discovery that MCL-1 amplifications occur recurrently in diverse malignancies, including breast, lung, bladder, ovarian, and prostate cancers [55].

It is now also clear that oncogenes and tumor suppressor genes (TSGs) promote malignant transformation at least in part through their regulation of apoptosis. For example, TP53, the most commonly altered TSG in human cancers, encodes a positive regulator of the pro-apoptotic proteins PUMA, NOXA, BID, and BAX [56–59]. Indeed, p53-mediated, transcriptional PUMA induction has been demonstrated to be the major driver of apoptosis in response to certain instances of DNA damage [60]. Likewise, oncogenic pathways, when hyperactivated, can suppress pro-apoptotic proteins and/or upregulate pro-survival proteins. For example, ERK phosphorylates BIM, leading to its βTRCP-dependent proteasomal degradation; thus, hyperactive ERK pathway activity leads to suppression of BIM levels [61, 62]. BAD is phosphorylated at S112 and S136 by ERK and AKT, respectively, leading to 14-3-3 binding and inactivation [63, 64]. Finally, PI3K pathway hyperactivation can both suppress PUMA expression via FOXO3A and promote MCL-1 translation through mTORC1 [65, 66].

Complicating the situation is the fact that, contrary to the themes earlier, oncogenes can also promote apoptosis. This concept, most thoroughly established for the MYC oncogene, is based on the fact that MYC positively regulates the expression of pro-apoptotic genes, including BAX, BID, and BIM [8]. Thus, both cancers and actively proliferating, MYC-high non-malignant tissues exhibit increased apoptotic priming, a fact which may both explain some of the toxicities associated with chemo- and radiotherapy treatments observed in children [8] while offering a novel treatment strategy for aggressive, MYC-high, drug resistant disease [67, 68], topics elaborated elsewhere in this review.

Cytokine, growth factor and hormone signaling

Signaling pathways regulated by cytokines, extracellular ligands, growth factors and hormones have been shown to modulate BCL-2 family protein expression and activity and can therefore modulate apoptotic vulnerabilities in cancer cells. For example, estrogen receptor (ER) signaling is known to induce levels of BCL-2 [24] in breast cancer cells, a finding which motivated ongoing trials examining venetoclax in ER+ breast cancer. Signaling pathways activated by cytokines including interleukin 6, interleukin 7 and others have been shown to modulate levels of BCL-2 family proteins and either promote survival, promote apoptosis, or modulate sensitivity to BH3 mimetics. Similarly, CD40 ligation from follicular helper T cells in the lymph node microenvironment can shift CLL cells toward survival and provoke resistance to BCL-2 inhibition. By extension, blocking CD40 signaling may potentially restore sensitivity, even in this so-called “sanctuary” site.

Mitochondrial structure and shape

Mitochondria contain inner and outer mitochondrial membranes, abbreviated as the IMM and OMM, respectively. The smooth OMM serves as a physical interface with the cytosol and other cellular compartments, while the IMM delineates the mitochondrial matrix and is comprised of two sub-compartments: an inner boundary membrane that runs parallel to the OMM, and the cristae – convoluted invaginations that contain the core proteins required for cellular respiration. The mitochondrial network in living cells is constantly being reshaped through a series of processes known as mitochondrial membrane dynamics, which involve both organelle fusion and fission as well as ultrastructural remodeling [69]. These dynamic changes, employed to meet the changing signaling and metabolic needs of the cell, in turn regulate apoptotic priming. For example, during cell death, cristae undergo significant structural changes that are dependent upon the mitochondrial fusion regulator OPA1 [70, 71]. These changes support the redistribution of cytochrome c from the cristae to the intermembrane space, followed by its complete release into the cytosol through apoptotic pores in the outer mitochondrial membrane. Similarly, during apoptosis, BAX and BAK initiate mitochondrial fragmentation through stabilization of the fission regulator DRP1 [69]. Even mitochondrial shape has been shown to affect the ability of the pore-forming proteins to induce MOMP: hyperfragmented mitochondria that have a more spherical shape are less permissive to stabilizing interactions between BAXα9 and the mitochondrial outer membrane [72]. This may protect cells from apoptosis and invalidate any apoptotic vulnerabilities that may exist.

Given the relationship between mitochondrial membrane dynamics and apoptotic cell death, it is intuitive that tumor cells may alter mitochondrial dynamics homeostasis as a means of promoting their growth and survival [73]. Indeed, recent evidence suggests that tumors exhibit recurrent amplification of mitochondrial dynamics-regulating genes [74]. Further, oncogenic signaling pathways exert direct control over mitochondrial membrane dynamics, as exemplified by studies demonstrating that the ERK/MAPK pathway, frequently hyperactivated in cancer, directly phosphorylates DRP1 to promote mitochondrial fission which is essential for tumor growth [75, 76]. Evidence suggests that mitochondrial membrane dynamics also powerfully influence tumor responses to apoptosis-inducing therapies. For example, small molecule inhibitors of driver oncogenes sensitize EGFR, BRAF, KRAS, and PIK3CA mutant tumors to SMAC mimetic therapies via their effects on mitochondrial membrane dynamics [74]. Further, tumor cells with altered mitochondrial membrane dynamics can be positively selected for by apoptosis-inducing therapy because of their increased resistance to apoptosis. This has been observed in leukemia cells, in which the mitochondrial chaperonin CLPB is upregulated upon the acquisition of resistance to venetoclax. CLPB’s resistance-conferring function is explained by its interaction with OPA1, which promotes the stabilization of mitochondrial cristae to prevent cytochrome c release following venetoclax treatment, highlighting the possibility that in certain cases, cells may recover from BAX/BAK mediated MOMP provided that cytoplasmic cytochrome c release is prevented [77]. Together, these findings provide critical evidence that mitochondrial structure and dynamics modulate tumor sensitivity to apoptosis inducing therapy, suggesting that mitochondrial structural regulation may be a point of convergence for diverse apoptosis resistance mechanisms.

Cellular metabolism

The intrinsic apoptosis pathway is largely regulated at the mitochondria, an organelle that is also the principal metabolic site of the cell. It is thus understandable that bidirectional crosstalk exists between metabolic and apoptotic regulation. For example, changes in mitochondrial membrane dynamics can simultaneously affect both cellular metabolism and apoptotic commitment [69, 73]. Further, discrete metabolic pathways regulate apoptotic potential. For example, changes in pentose phosphate pathway flux can alter the caspase activation state through NADPH-regulated redox inactivation of cytochrome c [78], and BAX activation and consequent MOMP are regulated by ceramide and sphingosine metabolism [79, 80]. Reciprocally, phosphorylation of NOXA and BAD can drive glucose consumption through the pentose phosphate pathway and glycolysis, respectively, while BCL-X_L has been linked to oxidative phosphorylation via ATP synthase [81–84].

Excitement surrounding venetoclax and other direct activators of cancer cell apoptosis inspired a series of recent studies to more fully articulate the mechanisms of metabolic-apoptotic crosstalk. This included one recent study that used CRISPR-based loss-of-function screens to map cellular metabolic pathways that influence the sensitivity of AML cells to venetoclax-induced apoptosis. Among other findings, this work revealed that heme biosynthesis is a powerful modulator of apoptotic priming via its effects on the stability and function of the electron transport chain, and that the altered regulation of heme biosynthesis may explain, in part, both the therapeutic window that exists for venetoclax between AML cells and normal hematopoietic stem cells as well as the refractoriness of certain leukemias with upregulated heme biosynthesis to induction chemotherapy [85]. In a similar manner, other studies have used unbiased functional genomic or metabolomic approaches to define metabolic pathways that influence apoptosis regulation and clinical sensitivity to venetoclax. Examples include studies identifying metabolic pathways leading to increased oxidative phosphorylation [86], fatty acid oxidation [87], and nicotinamide metabolism [88] as important modulators of apoptotic priming and consequent venetoclax sensitivity in patients with leukemias. These and other studies also suggest that therapeutic targeting of metabolic pathways whose activation statuses are altered in cancer, and which function to suppress apoptotic commitment, may be a particularly effective strategy to reverse venetoclax and chemotherapy resistance [89–92].

Imposed determinants of apoptotic vulnerabilities

Although endogenous features of tumors described earlier clearly template their apoptotic priming states, these states are not immutable. Instead, a significant amount of work in recent years has focused on rational strategies to shape tumor apoptotic priming using pharmacological approaches. Most notably, researchers have leveraged the idea that just as driver oncogenes often suppress apoptotic priming through biochemical modulation of the BCL-2 network, therapies targeting oncogenic drivers can reverse these effects, driving apoptosis or sensitizing tumor cells to BH3 mimetics. Many examples of this concept exist, including studies demonstrating that BIM induction is central to the apoptotic response of EGFR mutant [93–96] and ALK rearranged [97] lung cancers, HER2 amplified and PIK3CA mutant breast cancers [98], and BRAF mutant melanoma [99] and colorectal cancer [100]. Further, additional studies have revealed that even when oncopathway targeted therapies fail to induce significant levels of apoptosis on their own, they can modulate the regulation of the BCL-2 network in a manner that primes its sensitivity to blockade of one or more anti-apoptotic BCL-2 family members. This concept is demonstrated, for example, in studies identifying strong synergy between MEK and either BCL-X_L [27] or MCL-1 [101] inhibitors in KRAS mutant tumors, between mTORC1 and BCL-X_L inhibitors in PIK3CA mutant breast cancers [66], and between BRAF and MCL-1 inhibitors in BRAF mutant tumors [102], among many other examples.

The concept of therapy-induced apoptotic priming extends well beyond oncopathway targeted therapies. In fact, the therapeutic indices of diverse anticancer agents, including those targeting cell cycle checkpoints (e.g., CDK4/6 inhibitors, aurora kinase inhibitors), metabolism (e.g., glycolysis inhibitors), and DNA integrity (e.g., DNA damaging agents, ionizing radiation, topoisomerase inhibitors), are linked to the fact that already-stressed cancer cells exhibit increased apoptotic sensitivity [103, 104]. A wealth of information is available on the specific BCL-2 family protein changes induced by these anti-cancer agents, changes which can ultimately sensitize cells to BH3 mimetics. For example, topoisomerase inhibitors lead to cleavage and activation of the activator BID [105, 106], an effect which, like BIM induction following targeted therapy treatments [6, 94, 95, 107], would be expected to cause an overall increase in the number and strength of apoptotic vulnerabilities depending on which pro-survival proteins are expressed. Still other agents can cause changes in BCL-2 family protein interactions that very specifically alter apoptotic vulnerabilities. For example, proteasome inhibitors such as bortezomib cause ER stress, an unfolded protein response, and subsequent upregulation of Noxa [108], which binds and inhibits only MCL-1. This essentially mimics the activity of an MCL-1 inhibitor and potently triggers apoptosis in MCL-1 dependent cells. In cells that express both MCL-1 and BCL-2, this Noxa upregulation results in sensitivity to BCL-2 inhibitors [109].

The timing of therapeutic strategies that alter apoptotic vulnerabilities is an important and emerging consideration. For example, administering a DNA damaging agent may increase BCL-X_L dependence only after the damage is first sensed by DNA damage response elements such as p53 to activate downstream transcriptional programs, a process which may not manifest as increases in apoptotic vulnerabilities until 48 to 72 hours later in some cases [110]. Thus, the timing of administration of BH3 mimetics designed to synergize with these agents should be dictated by this mechanistic consideration, particularly for those mimetics that are expected to cause stronger on-target toxicities than the well-tolerated venetoclax. More recently, studies have also suggested that drug treatments can not only lead to the modulation of apoptotic priming on short timescales (hours to days) but can also select on longer timescales (weeks to months) for resistant cancer cells that exhibit hypersensitivity to BH3 mimetics. A recently reported example includes the oncogene MYC, which can prime cancer cells for apoptosis through the mechanisms described earlier. MYC is commonly upregulated in AML cells with acquired resistance to BRD4 inhibitors and in BRAF mutant melanoma cells with acquired resistance to BRAF/MEK inhibitors, causing these resistant cells to exhibit hypersensitivity to BH3 mimetics and laying the groundwork for serial treatment strategies that act as anti-cancer “evolutionary traps” [67, 68].

Outstanding questions and key challenges

Before targetable apoptotic vulnerabilities can be fully exploited for cancer therapy, a series of challenges and outstanding questions must be addressed:

• What cancers have endogenous or induced apoptotic vulnerabilities that can be targeted therapeutically? A growing body of mechanistic studies, combined with cell line profiling efforts, have begun to clarify how apoptotic priming states and BCL-2 family dependencies vary across human cancers [51]. Further, the deployment of clinical grade BH3 profiling assays is allowing for the more highly resolved mapping of both the overall apoptotic priming state and specific BCL-2 family dependencies in clinical samples in real time, work which is being complemented by emerging ex vivo BH3 mimetic profiling efforts [111, 112]. These methods can be further complemented by cytogenetic or expression biomarkers of BH3 mimetic sensitivity. For example, the t(11;14) translocation in multiple myeloma, which puts control of cell cycle-promoting cyclin D1 (CCND1) under the immunoglobulin heavy chain (IGH) promoter, is associated with heightened dependence on BCL-2 for reasons that are unclear. Likewise, increased BCL-2 dependence is also observed in estrogen receptor (ER)-positive metastatic breast cancers with confirmed BCL-2 expression, which is upregulated in a subset of these tumors. Finally, the baseline expression of a BCL-2 family member targeted by a BH3 mimetic, universally across the tumor cells present within a patient, should be a minimal criterion for enrollment of patients.

• What are the most effective strategies for combining BH3 mimetics with other anti-cancer agents? While in many cases, the most effective chemotherapeutic agents for a given cancer will pair well with an appropriate BH3 mimetic, this will not always be the case. For example, chemotherapeutics that drive non-apoptotic forms of cell death like senescence or ferroptosis may fail to synergize with BH3 mimetics. More broadly, the most synergistic, and potentially best tolerated, BH3 mimetic-containing combination therapies are expected to sometimes comprise agents which, on their own, are not particularly effective because one agent creates a tumor-specific apoptotic vulnerability that is exploited by the second agent. For example, KRAS mutant tumor models respond impressively to combinations of MEK inhibitors with BCL-X_L or MCL-1 inhibitors, and PIK3CA mutant tumor models similarly respond well to combinations of mTORC1 and BCL-X_L inhibitors, despite the fact that in each case, the constituent drugs are relatively ineffective on their own [27, 66, 101]. Indeed, BCL-X_L and MCL-1 appear to cooperatively regulate survival in a wide swath of tumor models, where in many cases selective inhibition of either target alone is ineffective [51].

• What regional variation or temporal dynamics of apoptotic vulnerabilities exist in tumors? Apoptotic vulnerabilities exploitable with BH3 mimetics may oscillate in a stochastic or non-stochastic manner over time and space, potentially impacting clinical sensitivity to these agents. The basis for these effects likely includes changing apoptotic competence during the cell cycle and fluctuating or spatially varying nutrient and growth factor availability, though much work is needed to understand these factors and their clinical impact more fully.

• What are the short- and long-term toxicities associated with therapies that include BH3 mimetics and how do they vary across heterogenous patient populations? Defining the toxicities associated with BH3 mimetic agents, particularly when they are combined with targeted or cytotoxic chemotherapies, has to date been a largely empirical exercise. Toxicities observed with single agent BH3 mimetics to date, including the thrombocytopenia that results from the exquisite sensitivity of platelets to direct BCL-X_L inhibition [113] and the cardiovascular toxicity observed with MCL-1 inhibitors, are consistent with the predictions of models systems [114, 115]. Similarly, for the case of BH3 mimetic-based combination therapies, the increased rate of infections seen in multiple myeloma patients treated with bortezomib plus venetoclax is easily understood [116], whereas the mechanistic basis for arrhythmias associated with combined BTK inhibition and venetoclax is less clear [111]. Finally, toxicities experienced in pediatric patients treated with BH3 mimetics may vary from those evident in adults due to the increased apoptotic sensitivity of cells within growing tissues [8].

• How can BH3 mimetics be refined to improve efficacy while reducing toxicity? The development of the first BH3 mimetic, ABT-737, was a stunning achievement of fragment-based drug discovery that opened the door to what are now a family of BH3 mimetic drugs with impressive potency, selectivity, and bioavailability [25]. However, it is clear that drugs targeting additional players in the mitochondrial apoptotic network, including inhibitors of less-studied BCL-2 family pro-survival proteins like BFL-1 and BCL-w and direct activators of BAX and BAK, could find utility in circumstances where existing BH3 mimetic drugs fail [51, 117]. Further, drugs which circumvent the on-target toxicities of existing BH3 mimetics, such as recently described BCL-X_L degraders that spare platelets, could dramatically improve the depth and breadth of activity of these agents [118]. Finally, methods to achieve tumor-selective MCL-1 modulation, for example by targeting of the protein’s upstream regulators, may be a useful strategy for overcoming the on-target toxicities observed with direct MCL-1 inhibitors to date [66].

• Can BH3 mimetics drive genomic instability and mutagenesis? A key consideration when targeting endogenous or imposed apoptotic vulnerabilities with BH3 mimetics is the strength of the dependence that is present in a cancer cell. While strong endogenous or imposed apoptotic vulnerability can support strong sensitivity to BH3 mimetics, mild dependence due to expression of multiple pro-survival proteins can buffer against inhibition of a single protein in a manner that may ultimately make BH3 mimetics in these settings counterproductive. This is based on the recent discoveries showing that inducing MOMP with BH3 mimetics in only a subset of mitochondria within a cell – a phenomenon referred to as “minority MOMP” – can lead to sub-maximal caspase and DNAase activation, which can be insufficient to carry out all the execution stages of apoptosis and instead cause DNA mutations and other cellular adaptations that promote tumor progression [119, 120].

• How do BH3 mimetics modulate anti-tumor immune responses? Recent preclinical studies have suggested the BH3 mimetics, including venetoclax, can improve the activity of both immune checkpoint inhibitors and NK cell therapies [121, 122], exciting developments that nevertheless should be considered with care given the well-established vulnerability of certain immune populations to these drugs. Further, it is also now clear that therapies that induce MOMP generally represent a “point of no return”, after which cell death is inevitable. However, combining MOMP-inducing agents like BH3 mimetics with deletion of downstream executioner caspases leads to the profound activation of Type I interferon (IFN) signaling in dying cells [123, 124]. This is important, as tumor-selective Type I IFN signaling can augment multiple steps in the tumor-immunity cycle, amplifying the therapeutic activity of diverse immunotherapies [125, 126]. Thus, decoupling mitochondrial cell death from caspase-driven immune suppression may be a promising strategy for unlocking the immune stimulatory potential of BH3 mimetic therapies.

Concluding remarks and future perspectives

The convergence of three recent events - the development of selective, potent, and in vivo bioavailable BH3 mimetic drugs, our increased understanding of the determinants of apoptotic priming and vulnerabilities, and evidence of striking clinical successes with venetoclax – have together set the stage for the large-scale integration of direct apoptosis activating therapies into the cancer clinic. In fact, it is not implausible to suggest that patients with a very broad range of cancers could potentially benefit from the simple combination of the most effective chemotherapy for their cancer with an appropriate BH3 mimetic agent for their tumor. Nevertheless, fully achieving the therapeutic potential of BH3 mimetic drugs in the clinic requires that we develop a more complete understanding of the determinants of sensitivity and toxicity along with an improved set of pharmacological tools (see Outstanding questions).

Outstanding Questions.

• What is the complete landscape of apoptotic priming states across defined tumor types and subtypes?

• What are the most effective strategies for building BH3 mimetic-based combination therapies containing cytotoxic or targeted chemotherapies?

• What are the key toxicity challenges associated with BH3 mimetic-based therapies?

• Can improved pharmacological strategies make BH3 mimetics safer and more effective?

• How do apoptotic vulnerabilities in tumors vary in space and time?

• What is the nature and risk of tumor-promoting effects of BH3 mimetics?

• How can BH3 mimetics modulate the immune system, and how does this impact therapeutic outcomes?

Highlights.

• During cancer development, cells with pre-existing buffers against apoptosis, such as increased expression of pro-survival BCL-2 proteins, can evade transformation associated apoptosis and form tumors

• The requirement for pro-survival BCL-2 family proteins in cancer cells endows them with vulnerabilities that are exploitable using BH3 mimetic drugs

• Diverse endogenous features of cancer cells modulate BCL-2 family protein dependencies, as do imposed features such as the cellular response to drug treatments

• The parallel development of selective, potent BH3 mimetic drugs, combined with our evolving understanding of the landscape of BCL-2 family vulnerabilities in cancer, has set the stage for the large-scale integration of direct apoptosis activating therapies into the clinic

Glossary

Anoikis

a form of cell death triggered by detachment from a surrounding extracellular matrix

BCL-2 family of proteins

a family of pro-survival and pro-apoptotic proteins whose interactions dictate a cell’s apoptotic equipoise. Pro-survival members include BCL-2, BCL-X_L, MCL-1, BCL-w, and BFL-1, while pro-apoptotic members include the pore-forming proteins BAX and BAK as well as BH3-only activators and sensitizers including BIM, BID, BAD, PUMA, NOXA, HRK, and BMF

BH3 profiling

a live cell-based functional assay to measure apoptotic priming and dependencies

BH3 mimetic

a class of drugs targeting the pro-survival BCL-2 family proteins, most notably BCL-2, BCL-X_L, and MCL-1

BRD4 inhibitor

an inhibitor of BRD4, a bromodomain-containing protein that binds to acetylated histones to regulate gene expression and is important for the viability of certain cancers, including AML

EMT

epithelial-mesenchymal transition, a process whereby cancer cells acquire characteristics associated with increased aggressiveness and drug resistance

MOMP

mitochondrial outer membrane permeabilization, the BAX/BAK-dependent process that leads to cytoplasmic cytochrome c release and downstream caspase activation during apoptosis

PERK

an endoplasmic reticulum (ER)-resident protein that signals ER stress

Sanctuary site

a physiological site where cells are protected, for example from drug-induced cell death

SMAC

second mitochondrial activator of caspases, a protein which facilitates activation of caspases downstream of MOMP