The executioner caspases are present in cells in an inactive, dimeric form, which becomes activated when cut at specific aspartate residues, resulting in a conformational change that forms the active enzyme. The active enzyme now cuts its substrates, some of which are responsible for the changes associated with apoptosis. ⁶

In general, this activating cleavage is mediated by other caspases, called initiator caspases. Unlike the executioner caspases, the inactive forms of the initiator caspases present in cells are monomeric and contain large prodomains that include sites for protein-protein interaction. Adapter molecules, which form “caspase activation platforms,” bind to the prodomains of initiator caspases, forcing them into contact, and the latter then fold into active enzymes. These can then process the executioner caspases, thereby activating them to cleave their substrates and precipitate apoptosis.

The formation of caspase activation platforms and the caspases they activate define the apoptotic pathways. Here we consider in detail only two major pathways of apoptosis, the mitochondrial pathway and the death receptor pathway. It should be noted, however, that there are others, reviewed elsewhere. ⁷ One of these involves the formation of caspase activation platforms for the engagement and activation of caspase-1, which is critically involved in inflammatory responses by processing specific cytokines and promoting their release. The activation of caspase-1 can also kill cells by a process resembling apoptosis (often called pyroptosis). The activation of caspase-1 occurs in response to many infectious agents and has been implicated in promoting forms of cancer that are associated with inflammatory conditions. Caspases can directly signal apoptosis or can use mitochondria as an intermediate, and additional point of regulation, in signaling of apoptosis.

The mitochondrial (sometimes called intrinsic) pathway of apoptosis responds to a diverse group of initiating events, including treatment with a wide variety of cytotoxic drugs, growth factor withdrawal, and oncogene activation. The mitochondrial pathway of apoptosis get its name because of the centrality of this organelle in coordinating both cell fate decision making as well as execution. Key features of apoptosis via the mitochondrial pathway include mitochondrial outer membrane permeabilization (MOMP), followed by release to the cytosol of molecules that facilitate execution of the death program, including cytochrome c, SMAC, and Omi. Cytochrome c is essential to the formation of one of the caspase activation platforms, the apoptosome, which is formed when cytochrome c activates APAF-1 to oligomerize, and the latter then binds and activates caspase-9. ¹⁰ Although caspases may perform normal physiological functions, they are best known for cleavage of proteins during apoptosis important for cell integrity, including cytoskeletal proteins and PARP-1. Among the caspase-dependent phenomena commonly observed in association with apoptosis are extracellular exposure of the phosphatidylserine that is usually found only on the inner leaflet of the plasma membrane (which results in recognition by Annexin V) and internucleosomal cleavage of chromosomal DNA (which results in laddering of DNA on electrophoresis and nuclear condensation). In vivo in a multicellular organism, an important consequence of these changes is recognition and clearance of the apoptotic cell by phagocytic cells. It may well be that commitment to cell death is already made at the point of MOMP, whereas many of the downstream morphologies we associate with apoptosis might be most important mainly in facilitating the clearance of apoptotic cells, minimizing inflammation, and optimizing cannibalization of their macromolecules. ^11–13

The key step in apoptosis commitment that is controlled by the BCL-2 family is MOMP. ^17–19 In simplest terms, there are pro- and anti-apoptotic members of the BCL-2 family. When the pro-apoptotic proteins overwhelm the anti-apoptotic ones, MOMP occurs, and the cell is irreversibly committed to apoptosis. The details of this commitment involve complicated interactions among the BCL-2 family proteins that number greater than a dozen. BAX and BAK are obligate pro-apoptotic “effectors,” proteins that homo-oligomerize to form the pores that cause MOMP. ²⁰ In order to form the oligomers required to cause MOMP, BAX and BAK must be activated. ^21,22 Proteins of the pro-apoptotic “activator” BH3-only protein subfamily, which include BIM, BID, and perhaps PUMA, can directly interact with BAX and BAK to cause an allosteric change that functionally corresponds to activation. There may also be other ways that BAX and BAK are activated, perhaps even independent of interaction with other proteins. ²³

Anti-apoptotic BCL-2 family proteins include BCL-2, BCL-XL, BCL-w, MCL-1, and BFL-1 (A1). These proteins inhibit apoptosis by binding and sequestering activator proteins or monomeric “activator” BAX or BAK proteins before they can oligomerize. ^24,25 Another subfamily, the pro-apoptotic “sensitizer” BH3-only proteins, lack the ability to directly activate BAX and BAK with high efficiency, but promote apoptosis by competitively inhibiting the ability of anti-apoptotic proteins to bind activators, BAX, or BAK (Figure 14-4 ). ²¹

BH3-only proteins get their name from their possession of the BCL-2 homology 3 (BH3) region, but of no other BCL-2 homology region. ²⁶ The BH3 domain is essential for the pro-death activity of all of the pro-apoptotic BCL-2 family proteins. It is a roughly 20 amino acid amphipathic α-helix that binds into a hydrophobic BH3 binding cleft present in all of the anti-apoptotic proteins. There is a selective pattern of interaction of BH3-only proteins with anti-apoptotic proteins. ^24,27–29 For instance, the BAD BH3 domain interacts selectively with BCL-2, BCL-XL, and BCL-w, whereas the NOXA BH3 peptide interacts selectively with MCL-1. Small-molecule BH3 domain mimetics that compete for these interactions, effectively inhibiting anti-apoptotic protein function, are being tested in cancer clinical trials (see later discussion). ³⁰

However, as it is commonly used, the oft-made assertion that resistance to apoptosis is a fundamental property of all cancer begs the question, “Compared to what?” It is difficult to find experimental or clinical evidence that cancers generally are more resistant to apoptosis than the normal nonmalignant cells in the human body. The fact that cancer cells have successfully buffered prior pro-apoptotic signaling does not necessarily mean that they are well prepared to evade subsequent pro-apoptotic signaling.

In fact, many conventional chemotherapies used in the clinic kill cells via apoptosis. The specific mechanisms may vary widely. Taxanes may increase levels of BIM, alter subcellular localization of BH3-only proteins, and decrease levels of MCL-1. ^36–38 Cytotoxic agents can induce a decrease in MCL-1 levels, though whether this is a necessary and sufficient step for commitment to apoptosis in these cases is less clear. ³⁹ DNA-damaging agents, by activating p53, can activate transcription of pro-apoptotic PUMA, NOXA, and BAX (see also Chapters 3 and 15). ^40–43 In addition, p53 may promote apoptosis via direct interaction with BCL-2 family proteins. ^44,45 Space does not permit detailing all of the relationships that have been discovered between conventional chemotherapy agents and the BCL-2 family. Although there are sometimes attempts to identify the key BCL-2 family dictating response to a particular agent, it is likely that following most chemotherapies, more than one BCL-2 family protein takes part in determining the cell fate decision.

The ability of chemotherapy to selectively kill cancer cells makes it likely that in many cancer cells the apoptotic pathway is not only intact, but even more sensitive to pro-apoptotic signaling than in normal cells. This hypothesis has been directly tested using a mitochondrial assay called BH3 profiling, an assay that systematically compares mitochondrial response to a standardized panel of synthetic BH3 domain peptides. ^46,47 These studies have shown that chemosensitive cancer cells are indeed more sensitive to pro-apoptotic signals than normal cells, suggesting mitochondrial apoptotic priming as an important determinant of the therapeutic index in vivo and in vitro. Normal hematopoietic cells are the most primed of normal tissues, consistent with their perennial role as the site of dose-limiting toxicity for most cytotoxic regimens. Measuring pretreatment mitochondrial priming may even be useful as a predictive biomarker, because it correlates well with clinical response. ^46,47

From a perspective of protein biochemistry, cells that are highly primed have little anti-apoptotic reserve to buffer subsequent pro-apoptotic signaling. Very often, a highly primed cell will express abundant anti-apoptotic BCL-2 family proteins, but these will already be occupied by pro-apoptotic activators such as BIM. ^24,48,49 In these cases, the cells are dependent on the continuous function of the anti-apoptotic protein(s) for survival. Inhibition of one or more of these can result in MOMP and apoptosis in highly primed cells (Figure 14-5 ).

It is worth considering the distinction between expression of an anti-apoptotic protein, and dependence on that anti-apoptotic protein. If one enforces BCL-2 overexpression via transfection of a plasmid vector in a healthy cell line already well established in culture, one will not necessarily obtain a BCL-2–dependent cell line, because there is likely no dependence on the additional BCL-2. The cell line was doing fine without it before transfection, and inhibition of BCL-2 would just restore the status quo ante. In oncogenesis, however, increased expression of BCL-2 is selected for, not extrinsically induced. This implies that the increased BCL-2 that was selected for is required for the continuous buffering of pro-apoptotic signals. In biochemical terms, that means that BCL-2 is largely already in complex with pro-apoptotic BCL-2 family members such as BIM or BAX. In the instance of the transfected cell line, however, BCL-2 is mainly empty and ready to buffer subsequent pro-apoptotic signaling. In the case where BCL-2 expression has been positively selected in the face of extant pro-apoptotic signaling, the BCL-2 is mainly full. Rather than providing anti-apoptotic reserve, the BCL-2 in the latter case is actually storing pro-apoptotic proteins at the mitochondrion, priming it for sensitivity to apoptosis. ²⁴ The cancer cell just discussed is thus dependent on continuous BCL-2 function for survival. In other cancer cells, similar to the transfected cell line, BCL-2 expression may be incidental, rather than selected for, and the BCL-2 not required tonically. Protein and transcript levels of multiple BCL-2 family members, as well as functional assays such as BH3 profiling, have been used to identify cells that are dependent on BCL-2 and hence likely to be sensitive to antagonists of BCL-2 (see later discussion). ^48,50,51

A general property of kinase inhibitors in cancer, whether tyrosine kinase or serine/threonine kinase inhibitors, is their utilization of the mitochondrial apoptotic pathway to kill cancer cells. Interestingly, although doubtless other BCL-2 family members also contribute, BIM, a pro-apoptotic activator BH3-only protein, is regulated by many different kinase inhibitors. In turn, lethality of kinase inhibition depends on the upregulation of BIM. BIM upregulation is required for killing by imatinib whether in CML or in gastrointestinal stromal tumors (GIST) where the constitutively active c-KIT kinase is the relevant target. ^56,57 Similar observations have been made for inhibitors of EGFR, MEK, ALK, and b-RAF. ^37,58–61 A fascinating human genetic study found that a germline polymorphism of BIM that removes the BH3 domain that is required for pro-apoptotic function is common in an Asian population. ⁶² Intriguingly, this polymorphism conferred inferior clinical response to inhibitors of both BCR-ABL in CML and of EGFR in non–small-cell lung cancer, apparently confirming its importance in cancer cell–fate decision making in clinical use of kinase inhibitors.

Proteasome inhibitors such as bortezomib have become a mainstay of multiple myeloma therapy and have found application in other diseases such as mantle-cell lymphoma. These also apparently kill via the mitochondrial apoptotic pathway. In this case, the most often cited mechanism of response is the pro-apoptotic BH3-only family protein NOXA, which can selectively bind and inactivate MCL-1, an anti-apoptotic protein also in the BCL-2 family. ⁶³ Of course, proteasome inhibitors alter the levels of thousands of proteins, so that there may be several proteins that are affected that are important for cell-fate decision after proteasome inhibitors.

These efforts are in their infancy, and important questions need to be resolved. For example, should the agent have a narrow or broad spectrum of inhibition of anti-apoptotic BCL-2 family proteins? Broad-spectrum agents will likely be more toxic to more kinds of cancers; however, there is the possibility of a narrowed therapeutic index that could limit clinical utility. Will these agents be best used as single agents or in combination with others? Although some diseases may demonstrate clinical sensitivity to single agents, it may well be that combination with the powerful, albeit less selective, pro-death signaling induced by conventional cytotoxic agents will be necessary to provide the significant benefit in long-term clinical outcomes that is so badly needed in many cancers. In addition, combinations with targeted agents such as kinase inhibitors have proved promising in vitro.

When a death receptor is engaged by its ligand, this can result in apoptosis. In this death receptor pathway of apoptosis, the initiator caspase is caspase-8, and its activation platform involves the adapter protein, FADD, which binds to the prodomain of the caspase, allowing it to form its active dimer.

The simplest form of this process is seen in the ligation of CD95 (this may apply to the TRAIL receptors as well). Ligation of CD95 causes the exposure of a protein interaction region in the intracellular part of the receptor, which rapidly binds to FADD. Several receptors and the associated FADD form the caspase activation platform, and caspase-8 is recruited to FADD for its activation. The proteolytic activity of the bound caspase-8 now cuts the caspase, removing the prodomain and releasing the now-stabilized dimer (Figure 14-6 , A).

In the case of TNFR1, the process is much more complex. Binding of ligand to TNFR1 induces exposure of the intracellular region of the receptor, but this does not bind FADD. Instead, another adapter, called TRADD, is bound, as well as a number of other signaling proteins, some of which modify the associated proteins with ubiquitin. Among these is RIPK1, a protein kinase, discussed further later. Also among these are proteins involved in the activation of a transcription factor, nuclear factor-κB (NFκB). The initial complex (complex I) is then released from the receptor, and its subsequent signaling outcome depends on additional protein interactions. Among these is a protein called c-FLIP_L (also called CFLAR).

c-FLIP_L closely resembles caspase-8 but lacks a catalytic cysteine. When complex I is released, TRADD binds to FADD in the cytosol, and this recruits caspase-8. However, if c-FLIP_L is present, the complex that forms (complex IIa) contains heterodimers of caspase-8 and c-FLIP_L, and this does not promote apoptosis (discussed later). However, if the activity of NFκB is blocked or disrupted, or if the expression of c-FLIP_L is otherwise blocked, complex IIa activates caspase-8 (see Figure 14-6, B). Another complex that forms in cells exposed to TNF involves the kinase RIPK1 (complex IIb). Like TRADD, RIPK1 can bind to FADD and form a caspase-activation platform. Again, if c-FLIP_L is present, caspase-8–c-FLIP_L heterodimers form, and apoptosis does not proceed. However, if c-FLIP_L is absent, this can also result in caspase-8 activation (see Figure 14-6, C).

The formation of the RIPK1-FADD complex is not restricted to the death receptor pathway. It can form as a consequence of DNA damage, although the precise signaling mechanisms are not fully elucidated. ⁶⁴ Another way in which this forms is as a result of ligation of some Toll-like receptors (TLRs), molecules that recognize components of infectious organisms, such as bacteria. Some TLRs recruit a signaling molecule, called TRIF, and this can engage RIPK1 to form a complex with FADD. ⁶⁵ As with the other cases, the subsequent activation of caspase-8 to promote apoptosis depends on the presence of c-FLIP_L. Efforts to therapeutically activate death receptor pathways for cancer therapy are in development. ⁶⁶

Although MOMP and the release of cytochrome c can then lead to activation of caspase-9, as we have noted previously, this event is not necessary in this case. Instead, the release of other proteins, such as Smac and Omi, is required, as these bind to and neutralize XIAP. As a result, the executioner caspases cleaved by caspase-8 can now precipitate apoptosis in the cell. This is illustrated in Figure 14-6, D. This explains the otherwise apparently paradoxical observation that anti-apoptotic Bcl-2 proteins can often inhibit apoptosis engaged by the death receptor pathway.

Understanding necrosis is important in cancer biology for several reasons. Most importantly, cancers that contain high levels of necrosis often have a poor prognosis. This may be because such cancers grow so quickly that they outstrip their blood supply. Another important aspect of necrosis is that it is inflammatory: That is, the contents of the dying cells release mediators (DAMPs, see earlier discussion) that trigger inflammatory responses. Inflammation itself damages tissues and can cause cycles of repair and proliferation involving growth factors that can promote oncogenesis and tumor expansion. In many adult cancers, inflammation may be a major component of the process that promotes tumorigenesis.

As we noted earlier, it is useful to distinguish between cell death that is passive versus active, and this applies to necrosis. Passive necrosis occurs when cells are irreparably damaged by external forces, or when processes that are essential for sustaining cellular homeostasis are blocked or disrupted. This can occur as a consequence of mechanical stress or toxic chemicals or as a result of specialized pore-forming peptides or proteins (e.g., some venoms and bacterial toxins, complement, perforin). Because such processes are intuitively fairly obvious, our focus here is on forms of necrosis that are active, that is, depend on cellular processes and events that, if inhibited, preserve the survival of the cell. For the most part, these all correspond to forms of cellular “sabotage,” although as we will see, there is one form of necrosis that might be “suicide.”

As we mentioned, when a cell dies by apoptosis, it is rapidly engulfed by other cells and removed from the system before any loss of plasma membrane integrity. However, if such engulfment does not occur, the dying cells will take on some features of necrosis as plasma membrane integrity is lost. This is called secondary necrosis and can often be distinguished from other forms of necrosis by some features of apoptosis, such as chromatin condensation. Whether or not secondary necrosis is inflammatory, and thus may contribute to some disease states, remains controversial, but there is accumulating evidence that defects in the engulfment and clearance of apoptotic cells can cause pathology arising from the effects of secondary necrosis.

A second form of necrosis that is related to apoptosis is caspase-independent cell death (CICD). ⁶⁷ This occurs when a cell undergoes MOMP in the mitochondrial pathway of apoptosis, but caspase activation is blocked or disrupted. Although several mechanisms have been suggested for why CICD occurs, an emerging consensus suggests that this is mainly due to energetic consequences of the “mitochondrial catastrophe” produced by MOMP. ⁶⁸ Notably, regulatory events that control MOMP, such as expression of anti-apoptotic BCL-2 proteins, effectively block CICD and thus distinguish this form of necrosis. Although cells that undergo MOMP but do not engage caspases often die by CICD, cells can survive this event, ^69,70 and it is possible that such survival contributes to cancer and/or chemotherapeutic resistance in some cases.

As mentioned earlier, passive necrosis occurs when essential cellular processes are disrupted, but repair mechanisms can insulate a cell from such insults. One major repair mechanism is autophagy (discussed in much more detail later), and therefore a cell that dies as a consequence of such damage may show extensive evidence of autophagy. As a result, we may categorize cells dying in this way as so-called autophagic (type II) cell death. In such cases, however, inhibition or disruption of autophagy will render the cell more susceptible to necrosis, and therefore it may be more appropriate to regard this process as one of necrosis.

Ischemia/reperfusion injury is a complex process we can regard as a form of cell sabotage and thus active cell death, although it is not fully understood. There are three components of ischemia/reperfusion injury that are generally agreed on: a rise in intracellular calcium, the production of reactive oxygen species, and disruption of the mitochondria.

On reperfusion, potassium channels in the plasma membrane open, releasing potassium from the cytosol, and this in turn causes the opening of plasma membrane calcium channels and a rise in intracellular calcium (Ca²⁺). The increase in Ca²⁺ has several effects. First, it can activate the protease calpain, which can inflict proteolytic damage to the cell. Pharmacologic inhibitors of calpain have some protective effects in ischemia/reperfusion injury. Second, Ca²⁺ activates NADPH-oxidases, causing the production of ROS, which can also damage the cell by targeting lipids and DNA. Damage to the DNA elicits activation of polyADP-ribose polymerase (PARP), and extensive activation of PARP can deplete stores of NADH, accelerating death. Scavenging ROS or inhibition of PARP can also produce some protection against this form of injury.

Both Ca²⁺ and ROS may precipitate necrosis in large part by affecting the mitochondria through a process called the mitochondrial permeability transition. The inner mitochondrial membrane is critical for mitochondrial function, as the control of the distribution of protons and other ions across this membrane drives ATP production and other activities of the organelle. High levels of Ca²⁺ and/or ROS cause the opening of a “pore” in the membrane that dissipates these critical gradients, and the matrix swells, ultimately destroying the organelle. The components of the permeability transition pore are largely unknown, with one exception: a protein of the mitochondrial matrix, cyclophilin D, has a major role in the process. Animals lacking cyclophilin D display no developmental abnormalities, but are resistant to ischemia/reperfusion injury. Whatever the function of the mitochondrial permeability transition might be, it may be safe to conclude that necrosis induced by ischemia/reperfusion that engages this mechanism represents a form of cellular sabotage. Whether this process can be used to promote cancer cell death therapeutically is not known.

Necroptosis involves the activation of another kinase, RIPK3, which binds to RIPK1 but can also be engaged by other proteins (a putative sensor of viral DNA, called DAI-1, activates RIPK3 and causes necroptosis independently of RIPK1). The activation of RIPK3 rapidly induces cell death with the features of necrosis.

How cell death occurs on activation of RIPK3 is largely unknown and is an area of intense research. One other player in the process is another protein, MLKL, which may itself be a kinase or may act as an adapter to bring specific substrates to RIPK3. Although a number of effector mechanisms have been proposed, including mitochondrial ROS and other mediators, at this point we simply do not know how this form of necrosis occurs.

As discussed in the consideration of death receptor signaling, RIPK1 can recruit the adapter, FADD, into a caspase activation platform that engages caspase-8 and the caspase-like molecule, c-FLIP_L. Although the caspase-8–c-FLIP_L heterodimer is an enzymatically active protease, it does not promote apoptosis (for reasons that we still do not understand). However, it has now become clear that this activity serves an important function in cells: It blocks the activation of RIPK3 (Figure 14-7 ).

Ligation of TNFR1, or other ways in which RIPK1 is engaged and activated, generally does not cause necroptosis, unless the formation of the FADD–caspase-8–c-FLIP_L complex is blocked or disrupted. ^71,72 Mice in which the genes for FADD or caspase-8 are deleted die early in embryogenesis, but development is fully rescued by ablation of RIPK3. The effect of ablating c-FLIP is the same, but its rescue is more complex: In this case death occurs only if both RIPK3 and FADD are deleted, most likely because activity of FADD to activate caspase-8 promotes death in cells lacking c-FLIP.

How, exactly, the caspase-8–c-FLIP_L heterodimer blocks necroptosis is not fully known. One simple possibility is that it cuts RIPK1 (and perhaps RIPK3), rendering it inactive. However, other possibilities exist, either upstream of RIPK1 or downstream of RIPK3. For example, an enzyme, CYLD, that removes ubiquitin from RIPK1 (required for its necroptotic function) is targeted by caspase-8 ⁷³ and therefore may also play a role in the protective function of caspase-8–c-FLIP_L.

Why is the system “built” this way? One possibility is that many viruses have evolved mechanisms to inhibit the activation of caspase-8 as a way to avoid immune attack, and therefore, linking RIPK3-induced necroptosis to the pathway ensures that cells infected by such viruses still die. Another, similar argument relates to c-FLIP_L, which is rapidly turned over in cells: When intracellular parasites (including viruses) disrupt the translation of cellular proteins, levels of c-FLIP_L decline, thereby sensitizing the cell to necroptosis. The destruction of the cell not only limits the replication of the parasite, but also engages the inflammatory response elicited by the necrotic cells.

Because necroptosis appears to be a bona fide form of cellular suicide, it is possible that this cell death process can function in tumor suppression. As more is learned about this phenomenon, cancers may be found in which oncogenesis is offset by engagement of necroptosis. Alternatively, there may be ways to exploit the process for tumor cell destruction in some forms of cancers.

The c-IAP molecules are E3 ligases that have, among their targets, the kinases RIPK1 and RIPK3. In healthy cells, the c-IAPs restrict the accumulation of these kinases and therefore prevent their activation. When cells are exposed to the Smac mimetics, the kinases accumulate, but if c-FLIP_L is present, the action of FADD-caspase-8–c-FLIP_L prevents death. However, if c-FLIP_L is absent, the cells will die by caspase-8–mediated apoptosis or via necroptosis.

In normal conditions, autophagy functions at low basal levels to ensure the occasional, selective, and necessary turnover of proteins and organelles due to aging and normal wear and tear. Damaged proteins and organelles display the “eat me” signal, the addition of polyubiquitin chains to proteins, which are recognized by autophagy receptors that recruit autophagosomes. In response to stress and starvation, autophagy is robustly induced. This ensures the elimination of damaged proteins and organelles that are induced by stressful conditions and also provides an alternate source of bulk metabolic substrates through intracellular recycling when external nutrient sources are limiting.

There are many growth regulatory and stress pathways that control the switch from basal to induced autophagy. Many of these pathways converge on the major nutrient-sensing and cell growth–promoting mammalian target of rapamycin (mTOR). Nutrients activate mTOR, which shuts off initiation of autophagosome formation by directly phosphorylating components of the atg1/unc-51–like kinase 1 (ULK1) complex (see Figure 14-8). ⁹⁹ Starvation suppresses mTOR activity, de-repressing autophagy and providing a direct link between nutrient availability, cell growth, and catabolism (Figure 14-8 ). Simply put, cells have no need to eat themselves when there is plenty of food, and regulation of autophagy by mTOR integrates that important relationship.