Fundamental Mechanisms of Regulated Cell Death and Implications for Heart Disease

Abstract

Twelve regulated cell death programs have been described. We review in detail the basic biology of nine including death receptor-mediated apoptosis, death receptor-mediated necrosis (necroptosis), mitochondrial-mediated apoptosis, mitochondrial-mediated necrosis, autophagy-dependent cell death, ferroptosis, pyroptosis, parthanatos, and immunogenic cell death. This is followed by a dissection of the roles of these cell death programs in the major cardiac syndromes: myocardial infarction and heart failure. The most important conclusion relevant to heart disease is that regulated forms of cardiomyocyte death play important roles in both myocardial infarction with reperfusion (ischemia/reperfusion) and heart failure. While a role for apoptosis in ischemia/reperfusion cannot be excluded, regulated forms of necrosis, through both death receptor and mitochondrial pathways, are critical. Ferroptosis and parthanatos are also likely important in ischemia/reperfusion, although it is unclear if these entities are functioning as independent death programs or as amplification mechanisms for necrotic cell death. Pyroptosis may also contribute to ischemia/reperfusion injury, but potentially through effects in non-cardiomyocytes. Cardiomyocyte loss through apoptosis and necrosis is also an important component in the pathogenesis of heart failure and is mediated by both death receptor and mitochondrial signaling. Roles for immunogenic cell death in cardiac disease remain to be defined but merit study in this era of immune checkpoint cancer therapy. Biology-based approaches to inhibit cell death in the various cardiac syndromes are also discussed.

I. CELL DEATH: INTRODUCTION AND HISTORICAL PERSPECTIVE

Prior to the 1980s, cell death was generally considered passive and unregulated, the result of direct damage to cellular components by overwhelming physical, chemical, or biological insults. Even when apoptosis was described in 1972 as a “new” form of cell death that is morphologically distinct from necrosis (235), the notion that it represented a regulated process was not widely considered.

One exception to the concept of unregulated cell death, however, was the elimination of specific cells at specific times during development in invertebrates and vertebrates (153, 175, 294, 422). The term programmed cell death was initially coined to refer specifically to these developmental cell deaths, although this term is now applied more broadly. While the molecular basis for developmental cell death was not understood, the temporospatial consistency with which cells were eliminated in this context suggested an active and regulated process. For example, in the roundworm Caenorhabditis elegans, 131 cells in hermaphrodite embryos and 147 cells in male embryos (out of 1,090 and 1,178 somatic cells generated respectively) die invariantly at specific times and places. In the 1980–90s, Horvitz and colleagues identified a relatively small network of genes that orchestrate these developmental cell deaths: cell death abnormal (ced)-9 → ced-4 → ced-3 (reviewed in Ref. 302). The discovery was profound as it demonstrated unequivocally that cell death in this context was not occurring passively through unregulated damage to cellular components. Rather, these cell deaths were actively mediated through a signaling pathway shown in subsequent genetic mosaic analyses to be operating in a cell-autonomous manner (555). This work established the genetic basis of a regulated and actively mediated cell suicide process.

This same genetic pathway has been conserved over 600 million years of evolution to humans where, importantly, it mediates not only developmental cell deaths, but also cell death in postnatal life. The orthologs of ced-9, ced-4, and ced-3 in mammals encode the Bcl-2 family (187), Apaf-1 (575), and the caspase family (554), respectively, each major components of the mitochondrial apoptosis pathway. In higher organisms, apoptosis is also mediated via a second pathway involving plasma membrane death receptors (9, 14, 142). During the 1990s, intensive research expanded these genetic relationships into the biochemical events that constitute the mitochondrial and death receptor pathways (30, 263, 267, 284, 299, 338, 509).

During this important period of discovery, the prevailing assumption was that apoptosis was the sole form of actively mediated and regulated cell death. Other death processes (e.g., necrosis) were assumed to be passive and unregulated. By the beginning of the new millennium, however, it had become clear that apoptosis was not the only regulated form of cell death. Although necrosis has traditionally been viewed as unregulated, it is now recognized that a significant proportion of necrotic cell deaths, the exact proportion unclear, is actively mediated through regulated pathways. Although the signaling is distinct from apoptosis, these necrosis pathways also involve mitochondria (17) and death receptors, the latter referred to as necroptosis (194, 323, 496). In hindsight, the existence of regulated necrosis pathways in vertebrates should have come as no surprise as forms of regulated necrosis also exist in C. elegans (533).

In addition to apoptosis and regulated necrosis, a number of other death programs have emerged including ferroptosis (97, 456), pyroptosis (128, 438), parthanatos (121, 511), entotic cell death (247, 371), NETotic cell death (136, 142), lysosome-dependent cell death (433), autophagy-dependent cell death (11, 86, 289, 290), and immunogenic cell death (140, 321). While some of these programs are ubiquitous, others appear to operate only in very specialized contexts. Moreover, while each is defined by distinct morphological and/or mechanistic features, further investigation is needed to determine the true independence of some of these programs versus their functioning as “subroutines” within another program.

Herein we review current knowledge pertaining to fundamental mechanisms of regulated cell death and the roles of these death programs in the major cardiac syndromes: myocardial infarction (MI) and heart failure (TABLE 1). Emphasis will be on apoptosis and necrosis, although we will also consider autophagy-dependent cell death, ferroptosis, pyroptosis, parthanatos, and immunogenic cell death. We do not discuss entotic cell death, NETotic cell death, or lysosome-dependent cell death. Thus far, entotic cell death and lysosome-dependent cell death have not been demonstrated to mediate the demise of cardiac cells, and NETotic cell death has been implicated only indirectly through effects on neutrophils (32, 489, 519).

Table 1.

Clinician call-out box

Cell Death in Myocardial Infarction and Heart Failure
1. Cell death can occur in a passive manner or through multiple actively mediated cell suicide programs referred to collectively as regulated cell death.
2. Regulated heart muscle cell death plays important roles in the pathogenesis of reperfused myocardial infarction (ischemia/reperfusion) and heart failure with reduced ejection fraction.
3. Because regulated cell death is actively mediated through molecular pathways, the possibility exists to inhibit this signaling to reduce cardiac damage and dysfunction from these major cardiac syndromes.

II. MECHANISMS OF APOPTOSIS AND NECROSIS

In this section, we will discuss basic aspects of apoptosis and necrosis signaling in depth. Mechanistic features of autophagy-dependent cell death, ferroptosis, pyroptosis, parthanatos, and immunogenic cell death will be reviewed in section III in parallel with disease implications.

A. Morphological Definitions

Cell death was first characterized by light microscopy. Although not sensitive enough to detect the earliest evidence of irreversible cellular damage (211), this technique is useful in contrasting the morphological characteristics of apoptosis and necrosis. Apoptotic cells are characterized by three cardinal features: 1) shrinkage (apoptosis originally referred to as “shrinkage necrosis”; Refs. 234, 235); 2) fragmentation into membrane-enclosed structures containing mixtures of cell parts (apoptotic bodies); and 3) phagocytosis of apoptotic bodies by macrophages or neighboring cells. Other more variable morphological manifestations of apoptosis include plasma membrane blebbing and chromatin condensation with DNA typically plastered against the inner nuclear membrane, although chromatin condensation can also be observed in necrotic cells. The time frame for cell shrinkage, fragmentation, and phagocytosis varies greatly among organisms and tissues ranging from minutes to hours. When apoptosis proceeds efficiently, apoptotic bodies are removed without leakage of intracellular contents into the extracellular space, thereby avoiding inflammation. Moreover, apoptotic cells actively suppress an immune response through a variety of mechanisms (25, 486). Because of its rapidity and the absence of accompanying tissue inflammation, apoptosis is often a stealth process with single or small numbers of cells disappearing without a trace. In contrast, while cells stimulated to undergo apoptosis in culture usually exhibit typical apoptotic features at early time points, they often acquire necrotic features (e.g., plasma membrane leakiness) at late time points. The latter is referred to as “secondary necrosis” and reflects the absence of a phagocytic cleanup system in vitro.

While apoptosis and regulated necrosis were recognized only ~50 and ~20 yr ago, respectively, the concept of necrotic cell death as a morphological entity dates back more than 150 yr to the pioneering work of Virchow, who referred to this death process as “necrobiosis” (499). Necrotic cells exhibit cell and organelle swelling (due to pan-membrane dysfunction) and loss of defined cellular architecture (404). Necrosis in tissues in vivo typically involves swaths of contiguous (or proximate) cells, rather than single isolated cells as in apoptosis. Additionally, necrosis is accompanied by marked acute and chronic inflammation. The latter is induced by the passive and active release of mediators that recruit inflammatory/immune cells (50, 474). With time, inflammation often transitions to fibrosis including entrapment of remnants of dead cells. These features provide a permanent record that necrotic cell death has occurred.

B. Overview of Apoptosis and Necrosis Signaling

While morphology is informative, a key objective is to understand the various death programs at the molecular level. Both apoptosis and necrosis can be transduced through pathways involving cell surface death receptors (also referred to as the extrinsic pathway) or mitochondria (intrinsic pathway) (FIGURE 1). In the case of apoptosis, both pathways lead to enzymatic activation of caspases, which are a subclass of cysteine-dependent proteases that hydrolyze peptide bonds following aspartic acid residues in certain motifs (390). While the concept of “caspase-independent apoptosis” [e.g., mediated by apoptosis inducing factor (AIF) (462)] was entertained at one point in the evolution of cell death thinking, caspases are now considered obligatory for apoptosis. In contrast, AIF is now believed to have roles in necrosis and parthanatos that will be discussed below (248, 334).

FIGURE 1.

Overview of apoptosis and necrosis pathways. Both apoptosis and necrosis can be mediated through pathways that involve death receptors or mitochondria, resulting in four distinct pathways: 1) death receptor-mediated apoptosis, 2) death receptor-mediated necrosis (necroptosis), 3) mitochondrial-mediated apoptosis, and 4) mitochondrial-mediated necrosis. Regardless of upstream signaling, apoptosis in both pathways leads to activation of the cysteine-dependent proteases, called caspases. In death receptor-mediated apoptosis, binding of death ligands to their cognate receptors results in the formation of sequential protein complexes (complexes I and II). In the latter, caspase-8 (or -10, not shown) is activated. These caspases then cleave and activate downstream caspases, such as -3 and -7. In mitochondrial-mediated apoptosis, caspases are activated in a cytosolic complex (not shown) whose formation is triggered by cytochrome c. Cytochrome c gains access to the cytosol following permeabilization of the outer mitochondrial membrane, which is mediated by B cell lymphoma-2 (BCL-2)-associated X protein (BAX) and BCL-2 antagonist/killer 1 (BAK). In contrast to caspases in apoptosis, a single unifying mechanism has not yet been identified for downstream necrosis signaling that is common to both death receptor and mitochondrial pathways. In necroptosis, activation of the serine/threonine kinases receptor interacting protein kinase 1 (RIPK1) and RIPK3 is critical. RIPK3 then phosphorylates and activates a pseudokinase called mixed lineage kinase-like domain (MLKL), which oligomerizes and permeabilizes the plasma membrane to induce necroptosis. The critical event in the induction of mitochondrial necrosis is Ca²⁺-triggered opening of the mitochondrial permeability transition pore (mPTP) in the inner mitochondrial membrane. This opening causes rapid dissipation of the proton gradient across the inner membrane that drives ATP synthesis leading to energetic deficits. However, the precise mechanism leading to plasma membrane dysfunction in this pathway is unclear (see text).

Caspases exist in a hierarchy. In both the death receptor and mitochondrial pathways, the most upstream caspases, which are called initiator caspases, acquire enzymatic activity when their zymogen forms (called procaspases) are recruited into specific multiprotein complexes characteristic of each pathway. In the death receptor pathway, these complexes are assembled in response to the binding of ligand to death receptor. In the mitochondrial pathway, the critical event is permeabilization of the outer mitochondrial membrane (OMM), allowing the release of cytochrome c into the cytosol where it functions as a cofactor in assembly of the apoptosome (see below).

In contrast to the role of caspases as the final common pathway in apoptosis, no single molecule has emerged that unifies downstream necrosis signaling in the death receptor and mitochondrial pathways. In the death receptor pathway, the critical event in the induction of necroptosis is the activation of receptor interacting protein kinase 3 (RIPK3), a serine/threonine kinase. RIPK3 is often activated through phosphorylation carried out by the homologous receptor interacting protein kinase 1 (RIPK1) but, as discussed below, this is not always the case as there are alternative mechanisms. RIPK3 then phosphorylates and activates a pseudokinase termed mixed lineage kinase-like domain (MLKL), which oligomerizes and permeabilizes the plasma membrane to bring about necroptosis. RIPK3 → MLKL signaling is considered the canonical signaling module for necroptosis.

In contrast to the necroptosis pathway, the defining event in the mitochondrial necrosis pathway is calcium (Ca²⁺)-dependent opening of the mitochondrial permeability transition pore (mPTP) located on the inner mitochondrial membrane (IMM). Note that, although significantly more complex, the most critical events in apoptosis and necrosis in the mitochondrial pathway involve the OMM and IMM, respectively. Opening of the mPTP during necrosis rapidly dissipates the proton gradient across the IMM that drives mitochondrial ATP synthesis. It remains unclear, however, exactly how the resulting energetic deficit triggers manifestations of necrosis including loss of plasma membrane integrity. For example, does this result simply from dysfunction of plasma membrane ion pumps and osmotic shifts? Alternatively, are more active mechanisms involved as described above for necroptosis and as will be discussed below for pyroptosis? This issue has not been resolved.

C. Caspases

Because of the centrality of caspases to apoptosis and the complexity of their regulation, we will first describe these proteases in more detail before proceeding to signaling in the death receptor and mitochondrial pathways. Caspases are synthesized as largely inactive pro-enzymes termed procaspases. Each contains a prodomain, p20 subunit, and p10 subunit (390). The prodomain mediates important interactions of the procaspase with other proteins, the p20 subunit contains the catalytic cysteine-histidine dyad, and the p10 subunit contributes to substrate specificity. There are three subclasses of procaspases: 1) initiator (also called upstream, apical, or signaling) procaspases, including human procaspases-8, -9, -10, and perhaps -2 (whose role in cell death remains murky; Ref. 35); 2) effector (also called downstream or executioner) procaspases, including human procaspases-3, -6, and -7; and 3) inflammatory procaspases, including human procaspases-1, -4, -5, and perhaps -12 (although caspase-12 is enzymatically dead in most humans due to genetic variation raising questions about its function; Ref. 413). Most procaspases are expressed ubiquitously, exceptions being procaspase-1, which is most abundant in macrophages and monocytes, and procaspase-14, which is expressed in keratinocytes and whose function is unclear (84, 85). In addition, the substrate recognition motifs of the various caspases differ. While they each hydrolyze peptide bonds following only aspartic acid residues (P1 position), the amino acid 3 position upstream of the aspartic acid (P4 position) is also important for caspase specificity. The P4 residue tends to be a branched chain amino acid for the initiator caspases, a second aspartic acid for the effector caspases (although a valine in the case of caspase-6), and a bulky aromatic amino acid for the inflammatory caspases (355).

Apoptosis is mediated by the initiator and effector procaspases. In contrast, the inflammatory procaspases are involved in inflammasome signaling, cytokine processing, and pyroptosis, the latter a nonapoptotic form of cell death related to inflammasome activation (305, 391, 392). In addition to these roles, some caspases also function in non-cell death roles involving differentiation, proliferation, and other cellular processes (312).

Initiator procaspases exist in healthy cells as inactive monomers. Their activation during apoptosis involves a combination of dimerization and trans cleavage (28, 362), both processes induced by their recruitment into multiprotein complexes. Once activated, initiator caspases activate effector procaspases, which exist as inactive dimers in healthy cells. The activation event for effector caspases is cleavage by initiator caspases. This cutting takes place following aspartic acid residues strategically located at the junctions between prodomain, p20, and p10 subunits. The subsequent noncovalent assembly of two p20 subunits and two p10 subunits results in the active effector caspase.

Effector caspases subsequently cut multiple cellular proteins, structural and signaling, to bring about the apoptotic phenotype (129, 218). For example, caspase-3-mediated cleavage of the scramblase Xkr8 (leading to its activation) (463) and of the flippase ATP11C (leading to its inactivation) (432) are responsible for moving phosphatidylserine motifs to the outer leaflet of the plasma membrane, where they provide “eat me” signals instructing macrophages to phagocytose apoptotic bodies. Other examples include cutting of lamin A by caspase-6 (447) and of acinus (411), helicard (243), and inhibitor of caspase-activated DNase/DNA fragmentation factor 45 kD (ICAD/DFF45) (115, 285, 287, 412) by caspase-3 that bring about nuclear apoptotic changes. These examples aside, the functional significance of many instances of caspase-mediated cleavage of cellular proteins remains incompletely understood.

D. Death Receptor Pathway

1. Activating mechanisms: survival, apoptosis, and necroptosis

This pathway is regulated by the binding of specific ligands to their cognate cell surface receptors. Thus, in contrast to the mitochondrial pathway, which transduces a broad spectrum of death stressors, the death receptor pathway responds to a relatively circumscribed subset of external stimuli. Moreover, while the outcome of death receptor signaling was initially thought to be only apoptosis, it was later recognized that these receptors can also promote necroptosis, cell survival, proliferation, and inflammation (142, 190, 434, 496, 523).

Two classes of death receptors have been recognized. The traditional variety, all members of the tumor necrosis factor (TNF) superfamily, activate cell death following ligand binding (14). In contrast, so-called “dependency receptors,” which are molecularly heterogeneous, transmit death signals only when unliganded (314). Interestingly, some dependency receptors also activate survival and trophic pathways when bound by their ligands. Dependency receptors have been much less studied than the traditional variety of death receptors and remain controversial. Some potential dependency receptors [e.g., tropomyosin-related receptor A (TrkA) (44)] have been studied in the heart with respect to their trophic and survival functions, but not with respect to whether they induce cell death when unliganded (dependency mode).

Traditional death receptors include Fas (also called CD95 or Apo-1), TNF receptor 1 (TNFR1), TNF-related apoptosis inducing ligand (TRAIL) receptor 1 (also known as DR4), and TRAIL receptor 2 (also known as DR5) (277, 369, 445, 455, 500). Fas is bound and activated by Fas ligand (FasL; also known as CD95L and Apo-1L), which is an integral plasma membrane protein on an adjacent cell. Notably, Fas is found on most cells, whereas the basal expression of FasL is restricted to specific cell types (e.g., cells in immunologically privileged tissue contexts) (159, 161). But, FasL expression may be induced in response to stress, including during ischemia/reperfusion (I/R) in the heart (213). In contrast to FasL, TNF-α and TRAIL, the ligands for TNFR1 and TRAIL receptors, respectively, are soluble proteins. Death receptors generally exist as preformed trimers (135, 445). When bound by their ligand, trimers are thought to undergo a conformational change that is transmitted to their intracellular domain. This triggers assembly of one or more intracellular protein complexes at the plasma membrane. These complexes provide the platforms that determine which of the potential outcomes will result from death receptor activation.

Before describing these complexes in detail, some general principles will be noted that govern interactions between the core components of these complexes. The most important interactions are mediated in a modular fashion by death-fold motifs. While highly divergent at the primary amino acid level, death-fold motifs are structurally similar in three dimensions, each consisting of six anti-parallel α-helices (236). There are several varieties of death-fold motifs including death domains (DD), death effector domains (DED), caspase recruitment domains (CARD), and pyrin domains (PYD). Generally, death-fold motifs engage in homotypic interactions (binding to the same type of death-fold motif), although heterotypic death-fold interactions have also been reported (347).

Death receptors contain a DD in their intracellular tail which is revealed following ligand binding to the extracellular domain (373). The DD mediates protein complex formation at the plasma membrane. Two classes of complexes have been described: death inducing signaling complex (DISC) and complex I. The DISC was first discovered and initially thought to mediate all instances of apoptosis induced by death receptors (30, 338). When it became clear that death receptor signaling could also result in other outcomes including necroptosis (194, 496), further investigation led to the discovery of complex I (323). The DISC has been best characterized in Fas signaling (431, 510), while complex I has been most intensively studied with respect to TNFR1 (91). However, the relative usage of the DISC versus complex I in mediating apoptosis has not been established and may be context-dependent.

The DISC consists of death receptor, one or two adaptor proteins, and procaspase-8 or -10 (257). Note that humans possess both procaspases-8 and -10, which are highly homologous, although procaspase-8 seems to be functionally more important (469). Mice express only procaspase-8. Following ligand binding, Fas recruits Fas-associated death domain protein (FADD) through an interaction mediated by the DDs in each protein (55). FADD also contains a DED on its opposite end and uses this to recruit procaspase-8, which contains two DEDs in its prodomain. The recruitment of procaspase-8 triggers its activation through forced proximity and trans cleavage as previously discussed. Caspase-8 goes on to cleave and activate downstream effector procaspases to induce apoptosis.

In contrast to the DISC, signaling through complex I can promote several divergent outcomes: cell survival, proliferation, inflammation, apoptosis, or necroptosis (5, 492). Assembly of complex I begins similarly to the DISC. In the case of TNFR1, however, binding of TNF-α results in the recruitment of a second adaptor protein TNFR1-associated death domain protein (TRADD), containing a DD at each end, to the cytoplasmic tail of TNFR1 via DD-DD interactions (FIGURE 2). Furthermore, rather than recruiting FADD as in the case of the DISC, TRADD recruits RIPK1 (454), also through DD-DD interactions (323). RIPK1 may also be able to bind directly to TNFR1 through DD-DD interactions without the interposition of TRADD. Complex I also includes several other proteins. The adaptor proteins TNF receptor-associated factors 2 and 5 (TRAF2 and TRAF5) and cellular inhibitor of apoptosis proteins 1 and 2 (cIAP1 and 2) are recruited to TRADD and catalyze the attachment of lysine 63 (K63)-linked ubiquitin chains onto RIPK1 (23). [Note that the name “inhibitor of apoptosis proteins” reflects a separate function of cIAP1 and -2 in the mitochondrial pathway, which will be discussed below (87).] K63 ubiquitination facilitates recruitment of the linear ubiquitin chain assembly complex (LUBAC), consisting of heme-oxidized iron responsive element binding protein 2 (IRP2) ubiquitin ligase 1 (HOIL-1; which is the catalytic subunit), HOIL-1 interacting protein (HOIP), and shank-associated RH domain interactor (SHARPIN) (138). LUBAC attaches linear [i.e., methionine 1 (M1)-linked] ubiquitin chains onto RIPK1. Also recruited into this complex are cylindromatosis (CYLD) (105, 244, 485), a deubiquitinase that removes primarily linear ubiquitin chains, and its adaptor protein spermatogenesis-associated 2 (SPATA2) that mediates CYLD recruitment into complex I (113, 249, 426, 521); and A20 (526), a deubiquitinase that removes K63-linked ubiquitin chains (526), and its binding protein A20 binding inhibitor of NF-κB-1 (ABIN-1) (109). Together these enzymes determine the ubiquitination state of RIPK1 and other proteins in complex I (502).

FIGURE 2.

Death receptor apoptosis and necroptosis pathways. Death receptor signaling is initiated by death ligand binding to its specific plasma membrane death receptor. Shown here is the binding of tumor necrosis factor-α (TNFα) to TNF receptor 1 (TNFR1). This initiates assembly of complex I on the cytoplasmic tail of the receptor. Outcomes downstream of complex I assembly include cell survival, apoptosis, or necroptosis. In the formation of complex I, TNFR1 recruits the adaptor protein TNFR1-associated death domain (TRADD) which, in turn, recruits receptor interacting protein kinase 1 (RIPK1). TRADD recruits the adaptor proteins TNF receptor-associated factor 2 and 5 (TRAF2/5) and cellular inhibitor of apoptosis proteins 1 and 2 (cIAP1/2), which attach lysine 63 (K63)-linked ubiquitin chains onto RIPK1. The K63-linked ubiquitin chains promote the recruitment of the linear ubiquitin chain assembly complex (LUBAC), which catalyzes the attachment of linear [or methionine 1 (M1)-linked] ubiquitin chains onto RIPK1. K63-linked chains on RIPK1 promote the activation of transforming growth factor-β-activated kinase 1 (TAK1) through recruitment of TAK1 binding proteins 2 and 3 (TAB2/3). TAK1 [acting through inhibitor of κB kinases (IKKs) and nuclear factor-κB (NF-κB) essential modulator (NEMO) recruited to the linear ubiquitin chains, not shown] promotes the activation of NF-κB. NF-κB is a transcription factor whose targets include multiple genes that promote cell survival and inflammation. In addition, TAK1 also activates mitogen-activated protein kinases (MAPKs), which provide additional survival signals. Cell death initiation requires the transition from complex I to cytosolic complex II. This transition is promoted by decreases in RIPK1 ubiquitination and specifically a decrease in the ratio of linear/K63-linked ubiquitin chains. Cylindromatosis (CYLD) and A20 remove primarily linear- and K63-linked ubiquitin chains, respectively. Complex IIa can take two forms. In one, the binding of TRADD-FADD-procaspase-8 signals RIPK1-independent apoptosis. In the other, binding of activated RIPK1-FADD-procaspase-8 signals RIPK1-dependent apoptosis. Furthermore, caspase-8 inhibits necroptosis by cleaving RIPK1 and RIPK3 (not shown). However, if caspase-8 is inhibited (see text), the binding of activated RIPK1 and RIPK3 in complex IIb leads to RIPK3 activation. RIPK3 then phosphorylates and activates a pseudokinase called mixed lineage kinase-like domain (MLKL), which translocates to and permeabilizes the plasma membrane to cause necroptosis.

The K63-linked ubiquitin chains on RIPK1 facilitate the activation of transforming growth factor-β-activated kinase 1 (TAK1; also called MAPKKK7) through recruitment of TAK1 binding proteins 2 and 3 (TAB2 and TAB3) (37, 223). TAK1 provides cell survival signals through the activation of mitogen-activated protein kinases (MAPKs) and nuclear factor-κB (NF-κB). TAK1-mediated activation of NF-κB is brought about through its phosphorylation of inhibitor of κB kinase α and β (IKKα and -β) which involves the recruitment of these IKKs and NF-κB essential modulator (NEMO; also called IKKγ) to linear ubiquitination chains on RIPK1 (110). Transcriptional targets of NF-κB that mediate cell survival include B cell lymphoma-2 (BCL-2), BCL-2-like-1 long form (BCL-xL), cIAP1 and -2, X-linked inhibitor of apoptosis (XIAP), and cellular-FADD-like interleukin-1β-converting enzyme-inhibitory protein (c-FLIP) (46, 193, 246, 291, 322, 494). Importantly, these pro-survival functions of RIPK1 are mediated independently of its kinase activity.

Although RIPK1 can be bypassed in the induction of apoptosis and necroptosis in the death receptor pathway (discussed below), when RIPK1 mediates either of these death outcomes, its kinase activity is required. Both ubiquitination and phosphorylation play important roles in regulating RIPK1 kinase activity. Decreases in the ratio of linear to K63-linked ubiquitin modifications (e.g., in response to changes in CYLD, SPATA2, A20, ABIN-1) favor RIPK1 activation (109, 521, 525, 526). In contrast, RIPK1 phosphorylation by TAK1, IKKs, TANK binding kinase 1 (TBK1), and MAPK-activated protein kinase 2 (MK2) is generally inhibitory, although nuances exist based on the phosphorylation site and whether phosphorylation is transient or sustained (101, 209, 252, 316, 532). These regulatory mechanisms have recently been reviewed in detail (434).

Apoptosis and necroptosis are mediated by different cytosolic complexes derived from complex I (FIGURE 2). Complex IIa signals apoptosis, while complex IIb directs the cell to undergo necroptosis. Complex IIa can induce apoptosis in either a RIPK1-independent or RIPK1-dependent manner (434). RIPK1-independent apoptosis involves a complex defined by FADD-procaspase-8, which may also contain TRADD (246, 322). In contrast, RIPK1-dependent apoptosis is directed by a complex containing RIPK1-FADD-procaspase-8 (125, 477). Activation of RIPK1 kinase activity in this setting requires the homodimerization of this protein, which is mediated through the DD (315). Moreover, RIPK1 activation is required for its interaction with FADD. The utilization of RIPK1-independent versus RIPK1-dependent apoptosis is influenced by the ubiquitination and phosphorylation states of RIPK1, which regulate kinase activation as discussed above, but the precise mechanisms directing this choice are incompletely understood (102, 150, 209, 521). The final step in the induction of apoptosis in both the RIPK1-independent and RIPK1-dependent modes is the recruitment of procaspase-8 to FADD resulting in active caspase-8 through forced proximity and trans cleavage. Moreover, in addition to signaling apoptosis, caspase-8 activation precludes necroptosis. This is brought about by caspase-8-mediated cleavage of RIPK1 and RIPK3, important mediators of necroptosis (124, 276).

In contrast to apoptosis, necroptosis is mediated by complex IIb (also called the necrosome), which includes RIPK1, RIPK3, and MLKL (FIGURE 2). The binding of TNF-α to TNFR1 induces necroptosis when apoptosis is inhibited. Inhibition of apoptosis may result through various mechanisms. For example, certain viral proteins [e.g., cowpox cytokine response modifier A (CrmA) (394)] inhibit caspase-8. The resulting blockage of apoptosis facilitates viral persistence and amplification. In these situations, the switch to cell killing by necroptosis provides a critical host response to limit viral load (61).

Another factor regulating whether necroptosis takes place is the abundance of c-FLIP (207). The gene encoding c-FLIP gives rise to alternatively spliced transcripts, one encoding c-FLIP short (c-FLIP_S) and the other c-FLIP long (c-FLIP_L) (39). c-FLIP_S consists of only two DEDs similar to those in procaspase-8, while c-FLIP_L is homologous to full-length procaspase-8 but contains mutations making it catalytically inactive. Either cFLIP isoform can heterodimerize with procaspase-8, an interaction mediated by their DEDs. Heterodimerization between c-FLIP_S and procaspase-8 markedly inhibits procaspase-8 activation (125). In contrast, the effects on caspase-8 activity resulting from heterodimerization between c-FLIP_L and procaspase-8 vary with c-FLIP_L concentration (200). High c-FLIP_L concentrations inhibit caspase-8 activity, but low to equimolar concentrations result in partial activation. At one time, these effects were thought to be mediated by interactions of c-FLIP isoforms with FADD, but recent data suggest that they actually involve direct interactions with procaspase-8 (200). When caspase-8 is inhibited by c-FLIP_S or c-FLIP_L, apoptosis is blocked and necroptosis results (361). Interestingly, the limited caspase-8 activity provided by procaspase-8-c-FLIP heterodimers inhibits necroptosis, although the mechanism is unclear as these heterodimers appear to cause only limited cleavage of RIPK1 (361). The mechanisms that inhibit apoptosis to promote necroptosis in many physiological and pathophysiological contexts, however, remain to be elucidated.

The kinase activity of RIPK3 is required for necroptosis (61, 183, 277). Although RIPK1 can be bypassed (discussed below), RIPK3 is often activated by the kinase activity of RIPK1. RIPK1 and RIPK3 interact via RIP homotypic interaction motifs (RHIM) resulting in a complex series of phosphorylation events in each protein requiring their kinase activities (61). It has been suggested that autophosphorylation, rather than trans phosphorylation, may be most important (163). In addition, it has recently been shown that the attachment of K63-linked ubiquitin chains to already active RIPK1 by Pellino 1 (PELI1) is required for the RIPK1-RIPK3 interaction and necroptosis (503). Another facilitator of this interaction is the binding of heat shock protein 90 (HSP90) and its co-chaperone CDC37 with RIPK3 (262). Conversely, A20 removes K63-linked ubiquitin chains from RIPK3, which interferes with RIPK1-RIPK3 binding (109).

A major target of RIPK3 in the induction of necroptosis is the pseudokinase MLKL (460, 564). RIPK3-mediated phosphorylation of MLKL exposes a four helical bundle in MLKL that promotes its cysteine-dependent tetramerization, progression to amyloid-like filaments, and translocation to and permeabilization of the plasma membrane (43, 51, 189, 264, 337, 398, 504). Both MLKL phosphorylation and the four cysteines are critical for these events. The small molecule necrosulfonamide, which covalently binds one of the cysteines, inhibits necroptosis downstream of MLKL phosphorylation (460). Conversely, forced polymerization of MLKL permeabilizes the plasma membrane even in the absence of RIPK3 activation (283). The mechanisms by which plasma membrane permeabilization take place and are regulated remain incompletely understood but may involve phosphatidylinositol phosphate binding (100, 104, 504), sodium (Na⁺) and Ca²⁺ influx, osmotic changes (43, 51), and activation of the endosomal sorting complexes required for transport (ESCRT)-III machinery (154, 482, 551, 558).

While RIPK1-dependent phosphorylation is critical in the activation of RIPK3 in traditional death receptor-induced necroptosis, RIPK3 and necroptosis can also be activated independently of RIPK1 in other paradigms. For example, in response to pathogen-associated molecular patterns (PAMPs), Toll-like receptors (TLRs) can induce RIPK3 activation through a mechanism that involves Toll-interleukin-1 receptor (TIR) domain-containing adapter-inducing interferon β (TRIF), which, like RIPK1, possesses a RHIM (93, 220). RIPK3 can also be activated independently of RIPK1 by DNA-dependent activator of interferon regulatory factors/Z-DNA binding protein 1 (DAI/ZBP1), another RHIM-containing protein (395, 487, 488). The fact that RIPK3 can be activated independently of RIPK1 has necessitated a change in the definition of the canonical necroptosis signaling module from RIPK1 → RIPK3 → MLKL to the present definition RIPK3 → MLKL.

2. Inhibitory mechanisms revealed by in vivo models

What has been presented thus far is a relatively linear schema in which ligand binding of death receptors promotes cell survival through complex I and induces apoptosis or necroptosis through complexes IIa or IIb, respectively. RIPK1 contributes to the survival outcome independently of its kinase activity. In contrast, while apoptosis and necroptosis can occur without RIPK1, when these forms of cell death are mediated by RIPK1, its kinase activity is required. Finally, necroptosis requires RIPK3 kinase activity and MLKL. While the preceding model describes positive regulation, it lacks some critical negative regulatory mechanisms that were revealed through the study of loss-of-function mutations in mice. Accordingly, we will next review the most critical of these mechanisms. The reader is directed to see TABLE 2 for a more complete list of the mouse models that provided these insights.

Table 2.

Genetic manipulations of death receptor pathway proteins in mice

Genotype	Phenotype	Reference Nos.
Ripk3 –/–	Normal development	354
Mlkl –/–	Normal development	530
Ripk1 D138N/D138N	Normal development	382
Ripk1 K45A/K45A	Normal development	219, 444
Ripk1 ΔG26F27/ΔG26F27	Normal development	288
Procaspase-8 –/–	Lethal E (embryonic day) 10.5	495
Procaspase-8 –/–; Ripk3 –/–	Normal development	221, 361
Procaspase-8 –/–; Mlkl –/–	Normal development	8
Fadd –/–	Lethal E11.5	548
Fadd –/–; Ripk3 –/–	Normal development	8, 92
Fadd –/–; Mlkl –/–	Normal development	8, 508
Fadd –/–; Ripk1 –/–	Lethal P (postnatal day) 1–3	559
Flip –/–	Lethal E10.5	547
Flip –/–; Ripk3 –/–	Lethal E12	92
Flip –/–; Ripk3 –/–; Fadd –/–	Normal development	92
Flip –/–; Ripk3 –/–; Procaspase-8 –/–	Normal development	221
Flip –/–; Mlkl –/–; Fadd –/–	Normal development	8
Ripk1 –/–	Lethal P1–3	233
Ripk1 –/–; Procaspase-8 –/–	Lethal P1–3	93, 402
Ripk1 –/–; Ripk3 –/–	Lethal P1–3	93
Ripk1 –/–; Mlkl –/–	Lethal P1–3	402
Ripk1 –/–; Ripk3 –/–; Fadd –/–	Normal development	93
Ripk1 –/–; Ripk3 –/–; Procaspase-8 –/–	Normal development	93, 219, 402
Ripk1 –/–; Mlkl –/–; Fadd –/–	Normal development	8
Ripk1 –/–; Procaspase-8 –/–; Dai/Zbp1 –/–; Trif –/–	Normal development	353
Ripk1 –/–; Tnfr1 –/–	Lethal P1–3	93, 402
Ripk1 –/–; Tnfr1 –/–; Trif –/–	Lethal P14–21	93
Ripk1 –/–; Tnfr1 –/–; Infar –/–	Lethal P14–21	93
Ripk1 –/–; Tnfr1 –/–; Ripk3 –/–	Normal development	93
Ripk1 mRHIM/mRHIM	Lethal P1–3	274, 353
Ripk1 mRHIM/mRHIM; Mlkl –/–	Normal development	274, 353
Ripk1 mRHIM/mRHIM; Ripk3–/–	Normal development	274, 353
Ripk1 mRHIM/mRHIM; Dai/Zbp1 –/–	Normal development	274, 353
Ripk1 mRHIM/mRHIM; Trif –/–	Lethal P1–3	353
Ripk3 D161N/D161N	Lethal E11.5	352
Ripk3 D161N/D161N; Dai/Zbp1 –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Cyld –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Trif –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Mlkl –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Tnfr1 –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Flip –/–	Lethal E11.5	352
Ripk3 D161N/D161N; Ripk1 –/–	Lethal P1–3	352
Ripk3 D161N/D161N; Procaspase-8 –/–	Normal development	352
Ripk3 K51A/K51A	Normal development	303
Ripk3 K51A/K51A; Procaspase-8 –/–	Normal development	303

a) negative regulation of apoptosis and necroptosis by ripk1.

Germline deletion of RIPK1 was observed to result in neonatal lethality accompanied by the unleashing of cell death in multiple tissue compartments (233). Subsequent conditional RIPK1 knockouts in various tissues (intestine, skin, blood, and liver) also resulted in cell death at baseline or in response to stress (74, 405, 471, 493). The type of cell death was tissue-dependent with, for example, apoptosis predominating in intestine and necroptosis in skin. Importantly, the consequences of germline RIPK1 deletion could be rescued by simultaneous deletion of 1) FADD or procaspase-8 or TNFR1 (apoptosis arm) and 2) RIPK3 or MLKL or the combination of TRIF and DAI/ZBP1 (necroptosis arm), but not by inactivation of genes in either arm alone (8, 93, 219, 353, 402, 559). In aggregate, these data indicate that the net effect of RIPK1 is to suppress apoptosis and necroptosis.

As described above, RIPK1 may promote cell survival in a kinase-independent manner or cell death (apoptosis or necroptosis) through its kinase activity. To test in vivo why RIPK1 loss was derepressing apoptosis and necroptosis, mice were created with germline knockin of kinase-dead RIPK1. These mice were normal, indicating that the non-kinase function of RIPK1 suppresses cell death in vivo (219, 288, 382, 444). Further investigation showed that knockin of a mutation that disables the RIPK1 RHIM (which mediates RIPK1 interactions with RIPK3) phenocopies the neonatal lethality of RIPK1 absence and can be rescued by deletion of RIPK3, MLKL, or DAI/ZBP1 (necroptosis arm) (274, 353). Taken together, these observations lead to the conclusion that RIPK1 suppresses both apoptosis and necroptosis in the death receptor pathway through a kinase-independent, RHIM-dependent function. Some of the kinase-independent mechanisms by which RIPK1 mediates cell survival in complex I have already been discussed above, and others may operate in complexes I and II.

Interestingly, in contrast to mice lacking RIPK1, four humans were recently reported with homozygous loss-of-function mutations in both RIPK1 alleles (protein absent in 3 of 4 individuals) (68). While these people survived neonatal life, all four manifest lymphopenia, susceptibility to infections, arthritis, or inflammatory bowel disease. In addition, cells derived from these individuals exhibited increased susceptibility to necroptosis.

b) negative regulation of necroptosis by catalytically active c-flip_L-procaspase-8 heterodimers.

This mechanism was discussed earlier and originates from the curious observation that germline deletion of FADD or procaspase-8 causes embryonic lethality (495, 548). Why loss of apoptosis signaling was killing the embryo went unexplained for multiple years and led to the incorrect hypothesis that accumulation of excessive cells was responsible. A mechanism was not elucidated until it was observed that simultaneous deletion of RIPK3 or knockin of either of two kinase-dead RIPK3 mutants rescued the lethality of either FADD or procaspase-8 deletion (8, 92, 221, 303, 352, 361). This indicates that necroptosis is activated in the absence of FADD or procaspase-8. Further work, described above, revealed that c-FLIP_L-procaspse-8 heterodimers possessing low, but detectable, caspase-8 activity suppressed necroptosis, although the precise substrates remain unclear.

c) as yet unresolved: possible negative regulation of apoptosis by ripk3.

A somewhat surprising finding regarding the regulation of apoptosis emerged from the study of kinase-dead RIPK3 knockin mice. One of these mutants (D161N) exhibited embryonic lethality due to excessive apoptosis in the yolk sac vasculature. This was rescued by simultaneous deletion of procaspase-8 (352). In contrast, no rescue resulted from co-deletion of TNFR1, CYLD, FLIP, TRIF, DAI/ZBP1, or MLKL. The mechanism of this effect is thought to involve complex formation between RIPK3 (D161N), RIPK1, FADD, and procaspase-8 resulting in activation of procaspase-8. In contrast, wild-type RIPK3 does not engage in these interactions. However, a second RIPK3 kinase-dead (K51A) knockin mouse did not exhibit embryonic lethality or activation of apoptosis (303), despite the fact that both RIPK3 kinase-dead mutants were able to rescue embryonic lethality in mice lacking procaspase-8. The differential abilities of the two RIPK3 kinase-dead mutants to activate apoptosis is hypothesized to result from differences in the conformations of RIPK3 resulting from the precise mutations employed rather than from the absence of kinase activity. Further work will be needed to test this hypothesis and its physiological significance.

The conclusions from these in vivo models can be summarized as follows: 1) RIPK1 suppresses early postnatal lethality in mice by inhibiting both apoptosis and necroptosis in the death receptor pathway, a function that is independent of its kinase activity but depends on its RHIM. 2) Caspase activity resulting from the c-FLIP_L-procaspase-8 heterodimer suppresses embryonic lethality from FADD or procaspase-8 deletion by inhibiting RIPK3-dependent necroptosis.

E. Mitochondrial Pathway

The mitochondrial pathway, which can signal cell death through apoptosis or necrosis, is activated by diverse stimuli including loss of nutritional/growth/survival factors, hypoxia, ischemia, hypoxia/reoxygenation, ischemia/reperfusion, oxidative stress, nitrosative stress, proteotoxic stress, DNA damage, increases in concentrations of cytoplasmic or mitochondrial Ca²⁺, and multiple toxins and drugs (e.g., chemotherapeutics and targeted cancer therapies) (73). These stimuli are transduced to mitochondria directly or through other organelles [e.g., endo/sarcoplasmic reticulum (ER/SR)] to bring about cell killing.

The central event in apoptosis in the mitochondrial pathway is permeabilization of the OMM [often referred to as mitochondrial outer membrane permeabilization (MOMP)]. This allows various mitochondrial apoptogens to access the cytosol where they promote procaspase activation. In contrast, the major event in necrosis in the mitochondrial pathway is opening of the mPTP in the IMM (341).

1. Mitochondrial apoptosis pathway

MOMP is regulated by the BCL-2 family proteins (57) (FIGURE 3). There are three subfamilies, which are defined by whether they promote or inhibit cell death and which BCL-2 homology domains (BH) they contain. The first subfamily comprises the pro-survival BCL-2 proteins, which contain BH1–4. These include BCL-2 itself, BCL-xL, myeloid cell leukemia-1 (MCL-1), BCL-2-like protein 2 (BCL-W), and others. The second subfamily is made up of the multi-domain pro-cell death BCL-2 proteins, whose structures are classically defined as containing BH1–3 but, in some cases, have been interpreted to also include BH4. These include BCL-2-associated X protein (BAX) and BCL-2 antagonist/killer 1 (BAK). A third member, BCL-2 related ovarian killer (BOK), is important in embryonic development, but its role in cell death is not clear (232, 293). The third subfamily comprises BH3-only proteins, each of which promotes cell death and contains only the BH3 domain. These include BCL-2-like 11 (BIM), BH3 interacting domain death agonist (BID), p53-upregulated modulator of apoptosis (PUMA), phorbol-12-myristate-13-acetate-induced protein 1 (NOXA), BCL-2-associated agonist of cell death (BAD), BCL-2 interacting killer (BIK), Harakiri (HRK)/BCL-2 interacting protein death protein 5 (HRK/DP5), BCL-2 modifying factor (BMF), and others. These three subfamilies of BCL-2 proteins engage in complex interactions to regulate MOMP. The overview is that BH3-only proteins transduce death signals from the periphery to activate BAX and BAK and inactivate the pro-survival BCL-2 proteins. Conversely, the pro-survival BCL-2 proteins seek to inactivate BAX and BAK and to neutralize the BH3-only proteins.

FIGURE 3.

Mitochondrial apoptosis pathway. The mitochondrial apoptosis pathway can be triggered by diverse extracellular and intracellular stress stimuli. The central event in the execution of this pathway is permeabilization of the outer mitochondrial membrane (OMM), which is tightly regulated by the B cell lymphoma-2 (BCL-2) family proteins. These are divided into three subfamilies according to their function and BCL-2 homology (BH) domains: 1) pro-survival proteins, including BCL-2, BCL-2-like-1 long form (BCL-xL), and myeloid cell leukemia-1 (MCL-1); 2) multidomain pro-cell death proteins, including BCL-2-associated X protein (BAX) and BCL-2 antagonist/killer 1 (BAK); and 3) pro-cell death BH3-only proteins, which are further subdivided into “activators” and “sensitizers.” Activator BH3-only proteins [BCL-2-interacting domain death agonist (BID), BCL-2-interacting mediator of cell death (BIM), p53 upregulated modulator of apoptosis (PUMA), and perhaps phorbol-12-myristate-13-acetate-induced protein 1 (NOXA)] directly bind and conformationally activate BAX and BAK. Conversely, the pro-survival BCL-2 proteins oppose these events both through sequestering the activator BH3-only proteins and through interacting with and inhibiting BAX and BAK. Sensitizer BH3-only proteins [BCL-2 antagonist of cell death (BAD), BCL-2-interacting killer (BIK), BCL-2-modifying factor (BMF), and Harakiri (HRK)] indirectly activate BAX and BAK through binding the pro-survival BCL-2 proteins, which both displaces the activator BH3-only proteins as well as prevents the pro-survival proteins from binding and neutralizing BAX and BAK. Activation of BAX and BAK drives their homo- and hetero-oligomerization within the OMM. This results in membrane permeabilization and release of apoptogenic factors from mitochondria, including cytochrome c, second mitochondria-derived activator of caspase (SMAC)/direct inhibitor of apoptosis (IAP) binding protein with low PI (DIABLO), and OMI/high temperature requirement protein A2 (OMI/HtrA2). The binding of cytochrome c and dATP to Apaf-1 promotes formation of the apoptosome, in which procaspase-9 is activated. Caspase-9 subsequently activates downstream procaspases including -3 and -7. IAPs bind and inhibit procaspase-9 and active caspases-3 and -7. SMAC/DIABLO and OMI/HtrA2 bind and inhibit IAPs resulting in activation of caspases. Caspase-8 can also cleave and activate BID (tBID), allowing tBID to bind and activate BAX, thereby linking the death receptor and mitochondrial apoptosis pathways.

Here we consider this regulation in more detail. A specific subset of BH3-only proteins termed “activators” (which include BID, BIM, PUMA, and perhaps NOXA) bind directly to and induce conformational changes in BAX and BAK (FIGURE 3). The regulation of BAX has been studied in most detail and will be described. While BAX is also regulated transcriptionally (e.g., by p53) (325) and through posttranslational modifications such as phosphorylation (292), its conformational activation through interactions with BH3-only activators is critical. This involves binding of the BH3 helix of a BH3-only activator to the BAX trigger site located in its NH₂ terminus (148). This results in displacement of the linker between α-helices 1 and 2 (α1 and α2), rearrangement of internal helices, exposure of BAX’s own BH3 helix, and ultimately exposure of a transmembrane domain in α9, the most COOH-terminal helix (147). This transmembrane domain inserts tightly into the OMM, shifting the location of BAX from the cytoplasm, where it normally resides in healthy cells in an inactive state, to mitochondria. In addition, BAX mitochondrial translocation may also be facilitated by other active mechanisms that are not completely understood (383, 531). While BAX activation begins in the cytoplasm, lipid and protein interactions at the mitochondria also contribute (56). BAK resides constitutively at the OMM, and it also undergoes conformational activation. Following activation, BAX and BAK engage in homo- and hetero-oligomerization within the OMM to induce MOMP. The precise mechanism of MOMP is not clear, but may involve pore formation as suggested by liposome studies. MOMP, in turn, allows release of mitochondrial apoptogens to the cytosol where they facilitate procaspase activation.

The pro-survival BCL-2 proteins inhibit MOMP through several mechanisms. First, they bind and sequester the BH3-only activators, an interaction involving the BH3 domain of the activator and a hydrophobic groove formed by the convergence of BH1–3 in the COOH terminus of the pro-survival BCL-2 protein (referred to hereafter as the “groove”). An additional more controversial mechanism is direct binding of the pro-survival BCL-2 proteins to BAX and BAK, an interaction mediated by the BH3 domain of BAX or BAK and the groove in the pro-survival BCL-2 protein. This inhibitory mechanism is likely applicable only to partially activated BAX and BAK because the BH3 domain needs to be exposed. A third mechanism by which the pro-survival BCL-2 protein BCL-xL inhibits BAX is through promoting BAX retrotranslocation from mitochondria (111).

The BH3-only activators not only activate BAX and BAK through direct binding, but also indirectly by inhibiting the pro-survival BCL-2 proteins. The latter involves binding of the BH3 helix of the activator to the groove in the pro-survival BCL-2 proteins. The remaining BH3-only proteins that are not direct activators of BAX and BAK are called “sensitizers.” Using their BH3 helices, they also can bind the groove in the pro-survival BCL-2 proteins, which will displace the activator BH3-only proteins, releasing them to activate BAX and BAK. The result of this competition reflects the relative stoichiometry and binding constants of the proteins. But, under conditions of apoptosis induction, the sensitizers win out freeing the activators from sequestration.

BH3-only proteins (both activators and sensitizers) become activated through distinct biochemical mechanisms providing a means for them to relay a diversity of death stimuli to BAX and BAK and/or the pro-survival BCL-2 proteins. For example, PUMA and NOXA are transcriptionally activated by p53 (345, 364). BIM is activated by ER stress through mechanisms that increase its abundance. These include BIM dephosphorylation which prevents its ubiquitination and proteasomal degradation, and C/EBP homologous protein (CHOP)-mediated increases in Bim transcription (384). Caspase-8, which is activated during death receptor-mediated apoptosis, cleaves BID to produce truncated BID (tBID), in which the BH3 domain is exposed (164, 263), allowing tBID to then bind and activate BAX and BAK. This mechanism serves to link the death receptor and mitochondrial apoptosis pathways, thereby providing a critical amplification loop needed for death receptor ligands to maximally activate caspases in many cell types. BAD is activated when released from sequestration by 14-3-3 proteins. This results from BAD dephosphorylation at a critical serine residue, which can be triggered by loss of survival factors, such as insulin-like growth factor-1 (IGF-1) and interleukin-3 (IL-3), that reduce AKT activity, or by calcineurin-mediated dephosphorylation (76, 83, 506).

Induction of MOMP by activated BAX and BAK oligomers leads to the release of mitochondrial proteins including cytochrome c, second mitochondria-derived activator of caspase (SMAC)/direct inhibitor of apoptosis (IAP) binding protein with low PI (DIABLO) (106, 497), and OMI/high temperature requirement protein A2 (HtrA2) (118, 464) (FIGURE 3). In the cytosol, these proteins facilitate caspase activation and apoptosis as described below. Also released are AIF (462), which, as previously mentioned, likely plays roles in necrosis and/or parthanatos rather than apoptosis (248, 334), and endonuclease G (EndoG) (266), which was once thought to partner with AIF to cut DNA, but whose role in cell death is currently unclear (112). AIF will be discussed in later sections of this review.

In the cytosol, cytochrome c induces caspase activation by triggering the assembly of a multiprotein complex called the apoptosome (267, 284, 575). This consists of a circular array of seven molecules of the adaptor protein Apaf-1, each bound by one cytochrome c molecule (derived from the mitochondria) and one dATP molecule (from the cytosol), and seven molecules of procaspase-9 (2). The binding of cytochrome c presumably induces a conformational change in Apaf-1 that allows it to interact with procaspase-9 through CARDs in each protein. This induces procaspase-9 activation through forced proximity and trans cleavage, and caspase-9 subsequently cleaves and activates effector caspases.

In the cytosol, various IAP proteins (cIAP1, cIAP2, and XIAP) function as caspase inhibitors (87, 89, 108, 407) (FIGURE 3). These are the same proteins discussed previously in the context of their RIPK1 ubiquitination in complex I. In the mitochondrial pathway, these molecules function to inhibit caspases through several mechanisms, as delineated for XIAP. These include binding to procaspase-9 to prevent its dimerization and activation (442) as well as inhibition of already activated effector caspases-3 and -7 through binding that blocks their substrate sites (47, 199, 403, 465). In addition, XIAP promotes the degradation of effector caspases in the proteasome through its E3-ligase function (466). While these mechanisms inhibit apoptosis, XIAP also promotes its own degradation, presumably to provide a negative-feedback control mechanism (424, 542).

When released in the cytosol during apoptosis, SMAC/DIABLO and OMI/HtrA2 relieve inhibition of caspases by IAPs to allow apoptosis to proceed (464) (FIGURE 3). The mechanism involves binding of SMAC/DIABLO or OMI/HtrA2 to IAPs at or near their caspase binding sites, thereby freeing the caspases (185, 453). In addition, OMI/HtrA2 possesses a serine protease activity that permanently inactivates XIAP through cleavage (118, 464, 537).

Our discussion of the mitochondrial pathway thus far has focused on apoptosis and, more specifically, how BCL-2 proteins link death signals with MOMP and how mitochondrial apoptogens released during MOMP induce caspase activation and cell death. However, mitochondria have long been known to play critical roles in necrotic cell death, dating back to before it was even realized that necrosis could be a regulated process. Accordingly, we will next consider how regulated necrosis is mediated in the mitochondrial pathway and how necrosis and apoptosis signaling at this organelle may interconnect.

2. Mitochondrial necrosis pathway

The details of mitochondrial-mediated necrosis are less well defined than those of apoptosis or necroptosis. Nevertheless, some key principles have emerged. First, the triggering event for necrosis in the mitochondrial pathway is Ca²⁺-induced opening of the mPTP in the IMM. This occurs quickly, usually within minutes following elevation in Ca²⁺ concentration within the mitochondrial matrix (for example, see Ref. 528). Importantly, mPTP opening takes place in the absence of, or hours before, MOMP, indicating that mPTP-mediated cell death is not occurring secondarily to mitochondrial-mediated apoptosis. (The mechanism by which cytochrome c release may sometimes occur at late time points following mPTP opening is discussed below.) While a definitive knowledge of the components of mPTP would be needed to test directly the hypothesis that opening of this pore triggers mitochondrial-mediated necrosis, strong evidence is provided by studies in which ppif, the gene encoding cyclophilin D, was deleted in the mouse (17, 341, 425). Cyclophilin D is a peptidyl prolyl cis-trans isomerase that is located in the mitochondrial matrix (66, 172). While not a core component of mPTP, cyclophilin D is a strong positive regulator of Ca²⁺-induced pore opening (17). Moreover, promotion of mPTP opening by cyclophilin D is dependent on its prolyl isomerase activity, although the relevant substrates have not been identified. Importantly, cells lacking cyclophilin D from various tissues of ppif −/− mice are resistant to Ca²⁺-induced mPTP opening and necrotic cell death. These data suggest that Ca²⁺-induced mPTP opening is critical for necrosis in the mitochondrial pathway.

While observations that high Ca²⁺ concentrations induce mitochondrial swelling date back >60 yr (204, 389), the original definition of mPTP was based on pharmacological criteria (173). mPTP is a nonselective pore permeable to molecules <1.5 kDa in mass, whose opening is promoted by high Ca²⁺ concentrations in the mitochondrial matrix. Ca²⁺-induced mPTP opening is potentiated by oxidative stress, increases in phosphate, and depletion of adenine nucleotides (ATP and ADP) and inhibited by high concentrations of protons (H⁺). Some of these conditions occur during the reperfusion phase of I/R, in particular the oxidative stress and reversal of acidosis. Additional operational specificity was imparted to this pharmacological definition by the observation that cyclosporine A (CsA), an inhibitor of cyclophilin D, inhibits Ca²⁺-induced mPTP opening (67). As will be discussed in section III, CsA blocks I/R injury in some preclinical models of I/R (273, 367), although it appears to have failed in clinical trials (69, 278, 381).

As recently reviewed (226), most models of mPTP between ~1990–2007 depicted it as complex involving the adenine translocase (ANT) and sometimes the mitochondrial phosphate carrier (PiC) in the IMM, the voltage-dependent anion channel (VDAC) in the OMM, and cyclophilin D in the matrix. ANT, which was often hypothesized to be the pore-forming unit in the IMM, is in fact capable of forming pores in artificial membranes (409). Additionally, bongkrekic acid and atractyloside, which bind and regulate ANT in cells, inhibit or promote mPTP opening, respectively (179, 203). These data notwithstanding, subsequent knockout/knockdown studies have ruled out each of these components as essential to the pore function of mPTP (17, 171, 251, 341). It must be emphasized, however, that despite being excluded as components of the pore, ANT, PiC, and cyclophilin D are each critical sensitizers of Ca²⁺-induced mPTP opening (17, 239, 251, 341).

If ANT, PiC, VDAC, and cyclophilin D do not form the pore, what does? Unexpectedly, recent work has suggested that components of the F₁-F₀ ATP synthase in mitochondrial complex V constitute the pore-forming unit of mPTP (22). However, different groups have implicated different sets of subunits including the oligomycin sensitivity-conferring protein (OSCP) in the peripheral stalk of the F₁ subunit (152) and the c-subunit in the F₀ portion (6, 33). Furthermore, based on reconstitution studies, the group supporting the OSCP model proposed that the pore is composed of dimers of the F₁-F₀ ATP synthase. Both the OSCP and c-subunit models are supported by RNAi-mediated depletion of OSCP and ATP5G1 and ATP5G3 (the latter two proteins isoforms of the c-subunit). However, other research subsequently challenged these models through loss-of-function experiments using CRISPR/Cas9-mediated gene editing of Atp5O (encoding OSCP), Atp5g1, Atp5g2, and Atp5g3 (all encoding the c-subunit), as well as Atp5f1 (encoding the membrane domain of the b-subunit, which imparts structure to the F₁ stalk). Those data report that decreases in these subunits do not result in loss of mPTP function (181, 182). In addition, in silico modeling of the c-subunit has raised doubts as to whether its predicted structural and biophysical properties are consistent with mPTP (569). It should be noted that manipulations of some of these F₁-F₀ ATP synthase components perturb other aspects of mitochondrial function thereby introducing potential confounding factors. Furthermore, it has been counter-argued that careful examination of the data claiming that deletion of the genes encoding OSCP and the b-subunit does not abolish mPTP function (181) actually demonstrates partial inhibition, supporting the notion that these components of the F₁-F₀ ATP synthase do contribute to the pore (21). Thus this controversy remains unresolved (16), and further work will be needed to settle this critical issue. In addition, if components of the F₁-F₀ ATP synthase, in fact, do constitute the pore, a second important issue will be to find mutations that allow a separation of ATP synthetic function from necrosis function (if possible at all), a challenge because both processes involve dissipation of the proton gradient across the IMM.

Irrespective of the identity of mPTP, high Ca²⁺ concentrations in the mitochondrial matrix are the major stimulus for mPTP opening. Accordingly, an important challenge is to understand how pathophysiological stimuli relevant to various disease scenarios bring about these increases. A plausible pathway to explain this has been suggested for I/R (336) (FIGURE 4). The production of lactic acid in the context of anaerobic metabolism during ischemia creates an intracellular acidosis (64). This is partially corrected by activation of the plasma membrane Na⁺/H⁺ exchanger (NHE), which operates in reverse mode to transport H⁺ out of the cell in exchange for Na⁺ into the cell (335). The subsequent overload of intracellular Na⁺ activates the plasma membrane Na⁺/Ca²⁺ exchanger (NCX) to eject Na⁺ at the expense of importing Ca²⁺ (162, 205). The subsequent rise in cytoplasmic Ca²⁺ concentration triggers the release of large amounts of Ca²⁺ from the ER/SR through ryanodine and inositol trisphosphate (IP₃) receptors. This Ca²⁺ reaches the mitochondrial matrix, its transport across the OMM and IMM regulated by VDACs (443) and the mitochondrial Ca²⁺ uniporter (MCU) (375), respectively. As previously noted, mPTP opening occurs mainly during reperfusion when reactive oxygen species (ROS) and correction of intracellular acidosis sensitize mPTP opening. This schema was the conceptual basis for the use of Na⁺/H⁺ exchange inhibitors (e.g., cariporide) in I/R, which showed promise in animal studies but failed in clinical trials for reasons that are not completely clear, perhaps circumvention by redundant mechanisms (31, 228, 335, 478).

FIGURE 4.

Mitochondrial necrosis pathway. The mitochondrial necrosis pathway can be triggered by diverse stimuli including loss of nutrient/survival factors, increase in intracellular calcium (Ca²⁺), reactive oxygen species (ROS), ischemia, and ischemia/reperfusion. The central event in the execution of this pathway is opening of the mitochondrial permeability transition pore (mPTP) in the inner mitochondrial membrane (IMM). Ischemia deprives the myocardium of oxygen (O₂), resulting in anaerobic metabolism and intracellular acidosis. In response, the cell pumps out the excess proton (H⁺) and accumulates sodium (Na⁺) by reverse operation of the Na⁺/H⁺ ion exchanger (NHE). The excess Na⁺ is, in turn, pumped out by the Na⁺/Ca²⁺ exchanger (NCX), which leads to an increase in intracellular Ca²⁺. Further increases in Ca²⁺ levels result from Ca²⁺-induced Ca²⁺ release from the endo/sarcoplasmic reticulum (ER/SR), resulting in intracellular Ca²⁺ overload. Increased concentrations of Ca²⁺ in the mitochondrial matrix promote mPTP opening during reperfusion which neutralizes intracellular acidosis and stimulates ROS production. The pro-cell death protein BCL-2-associated X protein (BAX) also sensitizes Ca²⁺-induced mPTP opening and necrosis. The consequences of mPTP opening are 1) loss of the H⁺ gradient across the IMM (mitochondrial membrane potential, Δψ_m), which leads to cessation of ATP synthesis, and 2) influx of water into the hyperosmolar mitochondrial matrix leading to mitochondrial swelling and eventual rupture of the outer mitochondrial membrane (OMM).

Multiple signaling molecules have been reported to impact Ca²⁺-induced mPTP opening (reviewed in Ref. 248), but mechanistic insights are unavailable in most cases. Accordingly, we will comment only on some recent advances. Unexpectedly, the multidomain pro-cell death proteins BAX and BAK have been demonstrated to facilitate Ca²⁺-induced mPTP opening and necrotic cell death (225, 528). While the details pertaining to heart disease will be discussed in more detail in section III, the basic observation is that mice with combined cardiomyocyte-specific deletion of Bax and germline deletion of Bak exhibit rescue of I/R-induced necrotic morphology (528). Furthermore, absence of BAX and BAK was found to render mouse embryonic fibroblasts (MEFs) resistant to very early Ca²⁺-induced mPTP opening that precedes mitochondrial cytochrome c release and caspase activation by hours (225, 528). Absence of BAX alone is sufficient for this effect as reintroduction of physiological levels of BAX into Bax/Bak double knockout cells restored sensitivity to Ca²⁺-induced mPTP opening. The sufficiency of BAK alone has not yet been tested. BAX oligomerization is known to be required for MOMP during apoptosis (151, 238). In contrast, BAX does not undergo oligomerization during necrosis (528). Moreover, oligomerization-defective BAX mutants, which are unable to support MOMP and apoptosis (151, 238), were sufficient to restore sensitization of Ca²⁺-induced mPTP opening in Bax/Bak double knockout cells (225, 528), providing strong support for the notion that BAX mediates necrosis and apoptosis through fundamentally different mechanisms. The underlying molecular mechanism, however, remains incompletely understood. One group suggests that BAX functions as an OMM component of mPTP by facilitating OMM permeability and conductance (225). Another group proposes that BAX-mediated mitochondrial fusion, a recognized non-cell death function of BAX (196, 224), is important in the sensitization to Ca²⁺-induced mPTP opening (528). These hypotheses are not mutually exclusive.

In general, the role of mitochondrial dynamics (fusion and fission) in cell death remains poorly defined (306). Multiple studies demonstrate that mitochondria undergo fission during apoptosis, and fission has been hypothesized to facilitate MOMP. MOMP per se, however, can be dissociated from mitochondrial fission by the overexpression of pro-survival BCL-2 proteins, which inhibit MOMP but not fission (436). Nevertheless, mitochondrial fission appears to contribute to apoptosis and MOMP as evidenced by experiments showing that these processes are inhibited by pharmacological or genetic inhibition of dynamin related protein 1 (DRP1), which mediates fission (45, 132). Given these observations, the data showing that increased mitochondrial “connectivity,” whether through BAX-mediated mitochondrial fusion or inhibition of DRP1-mediated fission, sensitizes cells to Ca²⁺-induced mPTP opening are intriguing (528). In keeping with these observations, genetic deletion of mitofusins 1 and 2 (MFN 1 and 2), which mediate mitochondrial fusion, lessen infarct size during myocardial I/R, which is primarily a necrotic process (174). Somewhat confusing, however, is that pharmacological inhibition of DRP1 has also been reported to reduce infarct size (368). How do we reconcile these data? Do these observations reflect other unrelated functions of the molecules that were manipulated which happen to track with changes in mitochondrial connectivity? For example, DRP1 also mediates BAX oligomerization (329), and MFN2 also functions as a Parkin receptor in mitophagy (52) and in promoting mitochondrial-ER tethering (79, 349). Or do these findings reflect the influence of changes in mitochondrial size and shape on other aspects of mitochondrial anatomy (e.g., cristae structure; Refs. 63, 429) and function? Or, perhaps they reflect fundamental differences in how apoptosis and necrosis are signaled in the mitochondrial pathway? Taken together, these data suggest that mitochondrial dynamics and mitochondrial-mediated cell death (apoptosis and necrosis) are functionally and mechanistically interconnected, but additional work will be required to understand the precise interrelationships.

In thinking about the effects of BAX and BAK even more broadly, it should be noted that, in addition to their actions at the mitochondria, these proteins exert effects at other organelles that could impact mitochondrial-mediated necrosis indirectly. For example, BAX and BAK also reside at the ER (574), where they promote accumulation of luminal Ca²⁺ through a mechanism that may involve antagonism of a Ca²⁺ leak induced by the interaction of pro-survival BCL-2 proteins with IP₃ receptor 1 (IP₃R1) (359, 360). This ensures that adequate reserves of ER Ca²⁺ are available to respond to a subsequent death stimulus. Moreover, ER-localized BAX may be particularly important in mediating cell death in response to oxidative stress (430). BAX and BAK also induce permeabilization of the lysosomal membrane (227), although the relevance of this to necrotic cell death is not known.

The mechanisms by which mPTP opening brings about necrotic cell death are not clear. Opening of this pore has two major consequences: 1) immediate dissipation of the proton gradient across the IMM leading to cessation of ATP synthesis, and 2) the rapid ingress of water down its osmotic gradient into the solute-rich mitochondrial matrix resulting in mitochondrial swelling. Apoptotic cells shut down many energy-requiring cellular processes, including DNA repair, protein translation, and proteasomal function, often through caspase-mediated cleavage of the relevant effectors. In contrast, necrotic cells not only stop making ATP, but continue to spend it in an unrestrained manner (260, 410, 449, 461). Thus cellular ATP levels plummet during necrosis, a hallmark of this cell death process. This “energetic crisis” is often invoked as the explanation for the loss of plasma membrane integrity, another defining feature of necrotic cells. While deficits in ATP may provoke osmotic stresses through dysfunction of plasma membrane pumps (117), it remains unclear whether this mechanism accounts for plasma membrane leakiness during mitochondrial-mediated necrosis. Other possibilities meriting further exploration include 1) the extent to which loss of ATP disrupts plasma membrane repair processes (27, 498) and 2) the involvement of yet to be identified mechanisms that actively disrupt the plasma membrane, analogous to the roles of MLKL in necroptosis (43, 51, 189, 264, 337, 398, 504) and gasdermins in pyroptosis (184, 230, 406, 439, 512) (discussed in sect. IIIC3).

Some signaling loops appear to operate during necrosis to amplify energy deficits. For example, poly(ADP-ribose) (PAR) polymerase-1 (PARP-1) participates in DNA repair and transcription (245), processes that continue during necrosis and may be further induced (e.g., by oxidative damage to DNA during reperfusion). The synthesis of PAR and its transfer onto target proteins by PARP-1 requires NAD⁺. Since NAD⁺ is an essential cofactor for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in glycolysis, this process is compromised. Moreover, ATP is consumed by compensatory mechanisms attempting to restore NAD⁺ levels (29). Thus PARP hyperactivation, which is the central process in the cell death program called parthanatos (121) (discussed in sect. IIIC4), may also contribute to necrotic cell death.

Another potential amplification loop in necrosis is mediated by AIF, which functions as a mitochondrial oxidoreductase in healthy cells (77). We have previously noted that this protein has been inadvertently misnamed because its main cell death roles appear to be in necrosis and parthanatos rather than apoptosis (248, 334). PAR polymers from PARP-1 hyperactivation, calpains, and BAX can promote the release of AIF to the cytosol (334, 462, 552, 553). Following this, AIF translocates to the nucleus and promotes large-scale DNA cleavage (200 kb to 50 kb) as opposed to internucleosomal DNA cleavage mediated by CAD/DFF40 in apoptosis (115, 285, 287, 412). This DNA cleavage, which is carried out in partnership with macrophage migration inhibitory factor (MIF) (511), further activates PARP-1 resulting in more NAD⁺ consumption and energetic deficits in a positive feed-forward manner.

Interconnections exist between apoptosis and necrosis signaling. We have already considered how the differential usage of complexes IIa and IIb adjudicate the decision between apoptosis and necroptosis in the death receptor pathway. Although less defined, it is highly likely that apoptosis and necrosis also interconnect in the mitochondrial pathway (10). As previously mentioned, mPTP opening may lead to release of cytochrome c and caspase activation at late time points (17). This most likely results from the aforementioned osmotic effects of mPTP opening which promote OMM rupture rather than BAX/BAK-dependent MOMP. The extent to which this caspase activation contributes to the amplification of necrotic cell death, however, remains undetermined. Connections in the other direction (apoptosis → necrosis) may also exist. For example, during apoptosis, NADH dehydrogenase (ubiquinone oxidoreductase core subunit S1) Fe-S protein 1 (NDUFS1) undergoes caspase-mediated cleavage resulting in oxidative stress, mitochondrial depolarization, and swelling (401), events that may reflect necrosis. This raises the possibility that other mitochondrial proteins may also be caspase substrates. Further investigation is needed to delineate bidirectional cross-talk between apoptotic and necrotic signaling within the mitochondrial pathway.

III. REGULATED CELL DEATH IN HEART DISEASE

A. Overview of Myocardial Infarction and Heart Failure

Regulated and unregulated forms of cell death have been implicated in the pathogenesis of multiple forms of heart disease ranging from MI without reperfusion, MI with reperfusion (I/R), heart failure of diverse etiologies, myocarditis, congenital heart disease, and others (95, 167, 240, 327, 356, 366, 419, 435, 527). MI and heart failure have been studied most intensively, and we will focus on these syndromes in light of their high clinical relevance and because the available information allows them to be discussed mechanistically (TABLE 3).

Table 3.

Effects of death pathway manipulations in cardiac disease models and humans

Pathways	Cell Type	Cardiac Disease Model	Outcomes	Reference Nos.
Death receptor pathway in MI
Fas –/– (lpr mutant)	Global	I/R (in vivo)	Reduced infarct size	258
Tnfr1 –/–	Global	MI without reperfusion (in vivo)	No change in infarct size	250
Tnfr2 –/–	Global	MI without reperfusion (in vivo)	No change in infarct size	250
Tnfr1 –/–; Tnfr2 –/–	Global	MI without reperfusion (in vivo)	Increased infarct size	250
Necrostatin-1 (WT mice)		I/R (in vivo)	Reduced infarct size	448
Necrostatin-1 (global ppif –/– mice)		I/R (in vivo)	No additional reduction in infarct size	272
Ripk3 –/–	Global	I/R (in vivo)	Reduced infarct size	351, 563
miR-103/107 antagomir		I/R (in vivo)	Reduced infarct size and necrosis	507
miR-873		I/R (in vivo)	Reduced infarct size and necrosis	508
NRF-siRNA		I/R (in vivo)	Reduced infarct size and necrosis	508
miR-223 tg overexpression	Cardiomyocyte	I/R (in vivo)	Reduced infarct size and necrosis	388
miR-223 –/–	Global	I/R (in vivo)	Increased infarct size and necrosis	388
Mitochondrial pathway in MI
Bcl-2 tg overexpression	Cardiomyocyte	I/R (in vivo)	Reduced infarct size	38, 54
Bax –/–	Global	I/R (isolated perfused)	Reduced infarct size, apoptosis, and necrosis	192
Bax –/–	Global	MI without reperfusion (in vivo)	Reduced infarct size and post MI dysfunction	191
Bax –/–; Bak –/–	Bax (cardiomyocyte)	I/R (in vivo)	Reduced infarct size, apoptosis, and necrosis	528
	Bak (global)
Ppif –/–	Global	I/R (in vivo)	Reduced infarct size and necrosis	17, 341
Ppif tg overexpression	Cardiomyocyte	Baseline (in vivo)	Necrosis	17
Ppif –/– + Bax –/–; Bak –/–	Bax (cardiomyocyte)	I/R (in vivo)	No additional reduction in infarct size	528
	Bak (global)
	Ppif (global)
Puma –/–	Global	I/R (isolated perfused)	Reduced infarct size, apoptosis, and necrosis	483
Cyclosporine A		I/R (various models)	Reduced infarct size (with some variability)	367, 273
Cyclosporine A (humans)		I/R (in vivo)	Reduced infarct size	381
Cyclosporine A (humans)		I/R (in vivo)	No reduction in infarct size or post MI dysfunction	69
Death receptor and mitochondrial pathways in MI
cIAP2 tg overexpression	Cardiomyocyte	I/R (in vivo)	Reduced infarct size, apoptosis, and necrosis	62
UCF-101		I/R (in vivo)	Reduced infarct size	24, 282
ARC tg overexpression	Cardiomyocyte	I/R (in vivo)	Reduced infarct size	386
Caspase inhibitors		I/R (in vivo)	Reduced infarct size	195, 198, 538, 544
Procaspase-3 –/–	Procaspase-3 (cardiomyocyte)	I/R (in vivo)	No reduction in infarct size or post MI dysfunction	206
Procaspase-7 –/–	Procaspase-7 (global)
Death receptor pathway in HF
Procaspase-8 tg overexpression	Cardiomyocyte	Baseline (in vivo)	0.023% TUNEL+ cardiomyocytes (control 0.002%)	524
(limited activation)			Lethal heart failure
Procaspase-8 tg overexpression	Cardiomyocyte	Baseline (in vivo)	Rescue of TUNEL positivity and heart failure	524
(limited activation + delayed caspase inhibition)
Traf2 –/–	Cardiomyocyte	Baseline (in vivo)	Dilated cardiomyopathy and heart failure	169
Traf2 +/–	Cardiomyocyte	MI without reperfusion (in vivo)	Dilated cardiomyopathy and heart failure	169
Traf2 –/–; Tnfr1 –/–	Traf2 (cardiomyocyte)	Baseline (in vivo)	Rescued pathological phenotype of Traf2 –/–	169
	Tnfr1 (global)
Traf2 –/–; Ripk3 –/–	Traf2 (cardiomyocyte)	Baseline (in vivo)	Rescued pathological phenotype of Traf2 –/–	169
	Ripk3 (global)
Tak1 –/–	Cardiomyocyte	Baseline (in vivo)	Dilated cardiomyopathy and heart failure	265
Tak1 –/–; Tnfr1 –/–	Tak1 (cardiomyocyte)	Baseline (in vivo)	Rescued pathological phenotype of Tak1 –/–	265
	Tnfr1 (global)
Ripk3 –/–	Global	MI without reperfusion (in vivo)	Reduced cardiac remodeling and heart failure	298
Ripk3 –/–	Global	Doxorubicin	Reduced cardiac dysfunction	563
Mitochondrial pathway in HF
L-type Ca2+channel (LTCC) tg overexpression	Cardiomyocyte	Baseline (in vivo)	Ca2+ overload, necrosis, heart failure, premature death.	346
LTCC tg overexpression	LTCC (cardiomyocyte)	Baseline (in vivo)	Rescued heart failure	346
Ppif –/–	Ppif (global)
Ppif –/–	Global	Doxorubicin	Decreased cardiac dysfunction and necrosis	346
Ppif -/	Global	Pressure overload	Dilated cardiomyopathy and heart failure	114
			Shift from fatty acid to glucose use
Ppif –/–	Global	Swimming	Mortality in 5 days	114
			Increased hypertrophy and pulmonary edema
Bnip3 –/–	Global	I/R (in vivo)	Reduced apoptosis, cardiac remodeling, and heart failure	96
Bnip3 –/–	Global	Doxorubicin	Reduced mitochondrial abnormalities and mortality	90
Bnip3 tg overexpression	Cardiomyocyte	Baseline (in vivo)	Pathological remodeling and heart failure	96
Gαq tg overexpression	Cardiomyocyte	Baseline (in vivo)	Increased apoptosis, hypertrophy, and heart failure	3, 71
Gαq tg overexpression	Cardiomyocyte	Pregnancy	Heart failure	3
Gαq tg overexpression + caspase inhibition	Cardiomyocyte	Pregnancy	Rescued heart failure	180
NIX/BNIP3L tg overexpression	Cardiomyocyte	Baseline (in vivo)	Increased apoptosis and heart failure	557
Gαq tg overexpression	Cardiomyocyte	Baseline (in vivo)	Rescued apoptosis and heart failure	557
Autophagy-dependent cell death
Beclin-1 +/–	Global	Pressure overload	Reduced autophagy and cardiac dysfunction	570
Beclin-1 tg overexpression	Cardiomyocyte	Pressure overload	Increased autophagy and cardiac dysfunction	570
Beclin-1 +/–	Global	I/R (in vivo)	Reduced autophagy and infarct size	310
Beclin-1 +/–	Beclin (global)	Baseline (in vivo)	Accelerated heart failure and early mortality	475
Desmin tg overexpression	Desmin (cardiomyocyte)
Atg5 –/– (postnatal)	Cardiomyocyte	Baseline (in vivo)	Heart failure and TUNEL+ cardiomyocytes	344
Atg5 –/– (embryonic)	Cardiomyocyte	Pressure overload	Heart failure	344
Ferroptosis
Deferoxamine (iron chelator)	Perfusion	I/R (isolated perfused)	Reduced infarct size	144
Glutaminolysis inhibitor	Perfusion	I/R (isolated perfused)	Reduced infarct size	144
GPX4 tg overexpression (mitochondrial targeted)	Global	I/R (isolated perfused)	Reduced release of creatine kinase	72
Ferrostatin-1		I/R (in vivo)	Reduced infarct size and myocardial enzyme release	120, 268
			Cardiac dysfunction and pathological remodeling
Ferrostatin-1		Doxorubicin	Reduced cardiac dysfunction	120
Ferrostatin-1		Heart transplantation	Reduced cardiac cell death	268
Dexrazoxane		I/R (in vivo)	Reduced infarct size and myocardial enzyme release	120
Dexrazoxane		Doxorubicin	Reduced cardiac dysfunction	120
Pyroptosis
Caspase-1 –/–	Global	I/R (in vivo)	Reduced infarct size	229
Caspase-1 –/–	Global	MI without reperfusion (in vivo)	Reduced cardiac dysfunction	317
Caspase-1 tg overexpression	Cardiomyocyte	Baseline (in vivo)	Heart failure	317
Caspase-1 tg overexpression	Cardiomyocyte	I/R (in vivo)	Increased infarct size	467
ASC –/–	Global	I/R (in vivo)	Reduced infarct size	229
ASC –/– bone marrow	ASC (global)	I/R (in vivo)	Reduced infarct size compared with WT	229
Into WT mice			Same infarct size compared with ASC –/–
WT bone marrow	ASC (global)	I/R (in vivo)	Reduced infarct size compared with WT	229
into ASC –/– mice			Same infarct size compared with ASC –/–
Parthanatos
Parp-1 –/–	Global	I/R (isolated perfused)	Reduced cardiac contractile dysfunction	165, 378, 567
Parp-1 –/–	Global	I/R (in vivo)	Reduced infarct size and myocardial enzyme release	543, 573
PARP-1 inhibitor		I/R (in vivo)	Reduced infarct size	269, 378, 517, 572
PARP-1 inhibitor	Perfusion	I/R (isolated perfused)	Reduced cardiac contractile dysfunction	98, 567
Parp-1 –/–	Global	Pressure overload	Reduced pathological remodeling	379
Mif –/–	Global	I/R (in vivo)	Increased infarct size	324, 387
Mif –/–	Global	I/R (isolated perfused)	Increased infarct size	387

I/R, ischemia/reperfusion; MI, myocardial infarction; tg, transgenic.

1. Myocardial infarction

MI results from acute and prolonged deficits in the supply of oxygen, nutrients, and survival factors to the myocardium relative to the demands of this tissue. Although this imbalance can be created through various means, it is most commonly caused by sudden and prolonged ischemia. The latter often results from thrombotic occlusion of a coronary artery following rupture/erosion of an atherosclerotic plaque (270, 271). Total occlusion of a coronary artery manifests clinically as ST-segment elevation MI (STEMI), which is usually transmural at the pathological level. Infarction is accompanied by multiple structural and functional consequences, the most irreversible of which is myocardial cell death. This involves both cardiomyocytes and non-myocytes, although the latter is less well documented (423, 472, 473). Moreover, in contrast to other cardiac syndromes (e.g., heart failure), where multiple processes contribute to pathogenesis, cell death is the initiating and central cardiac event in MI (TABLE 3).

Work in the 1980s and 1990s established reperfusion therapy (initially shown with thrombolytic agents; later using angioplasty/stenting) as beneficial in reducing infarct size (340). Infarct size is critical because it is a major determinant of subsequent cardiac dysfunction/heart failure and mortality. Although the benefit of reperfusion in preventing these serious outcomes of MI is unquestionable, it is clear that the reperfusion process itself begets myocardial cell death through a combination of oxidative stress, Ca²⁺ overload, and inflammation (collectively referred to as reperfusion injury) (178, 549). While it is extremely difficult to estimate the relative magnitudes of cardiac damage resulting from reperfusion per se as opposed to the prolonged ischemia that precedes it, some investigators have proposed that reperfusion accounts for up to ~50% of the total myocardial damage (549). In animal models, reperfusion generally limits the transmural extent of infarction to the subendocardial/mid-myocardial portion of the area at risk.

Non-ST-segment MI (non-STEMI) results from acute obstruction of a coronary artery that is often subtotal or “flickering” between subtotal and total. This dynamicity results in some degree of reperfusion and the resulting pathological injury is classically subendocardial. Interestingly, the incidence of STEMI is decreasing, while that of non-STEMI has remained relatively constant (546), the causes of this shift being unclear but possibly related to healthier lifestyles, reduction of cardiac risk factors, or use of statins. Most MI studies in animals, however, involve coronary artery ligation without or with subsequent reperfusion and, thus, model non-reperfused or reperfused STEMI rather than non-STEMI.

In this section, we review genetic and pharmacological manipulations of various cell death pathways and their effects on infarct size and post-MI cardiac function. Some general points will be noted here. First, in most cases, genetic and pharmacological inhibition of cell death signaling reduces cardiomyocyte death and infarct size only in the context of reperfused MI (I/R). Such reductions are not usually seen in permanent ischemia models and, when reductions are observed in those contexts, the possibility of inadvertent reperfusion needs to be considered. Why are reductions in cardiomyocyte death and infarct size not achieved without reperfusion? Several possible explanations can be envisioned: 1) initial regulated cell death transitions to unregulated cell death if the death stimulus persists. 2) Cellular repair processes are initiated with reperfusion so that, if the initial ischemic killing can be reduced, the net effect is tissue salvage. 3) Inhibition of cell death signaling is actually mitigating reperfusion injury rather than ischemic damage. While suggestive information support each of these possibilities, definitive data are lacking. Moreover, these possibilities are not mutually exclusive.

Second, despite extensive research, it remains unclear which cell death programs are most important during MI. Studies predating the regulated cell death era have traditionally considered unregulated necrosis to be the primary mode of cardiomyocyte death in this syndrome based primarily on morphological criteria (211, 212). When apoptosis became recognized as the first regulated cell death process, multiple studies demonstrated that genetic manipulations that inhibit apoptosis signaling in the death receptor and mitochondrial pathways resulted in reductions in apoptosis markers (e.g., TUNEL) in cardiomyocytes and in infarct size in mice subjected to myocardial I/R (38, 54, 137, 156, 213, 258, 309, 320). These experiments were interpreted as showing that cardiomyocyte apoptosis is critical in the pathogenesis of MI with reperfusion. More recent studies have raised doubts about the relevance of apoptosis in cardiomyocyte loss during MI (206). Moreover, genetic inhibition of pathways that mediate mitochondrial-dependent necrosis (17, 341), necroptosis (258, 351, 563), ferroptosis (15, 72, 120, 144), pyroptosis (229), parthanatos (543), and autophagy-dependent cell death (290, 310) have also demonstrated reductions in cardiomyocytes death and infarct size in the context of I/R (TABLE 3). This raises the question as to why multiple death pathways are being activated during reperfused MI. Possible explanations include 1) multiple cell death processes are, in fact, operating during MI with some cells dying through one modality and other cells through another. 2) Overlap/connections among mechanisms and markers have confounded analyses that assign cell death to one or another program. 3) Redundancy in signaling mechanisms has resulted in another cell death program taking over when a given program is inhibited. These issues need to be sorted out possibly through combined manipulations of cell death pathways coupled with a more comprehensive analysis of pathway-specific markers.

2. Heart failure

Heart failure, which is a highly heterogeneous syndrome in terms of etiologies, is generally divided at the pathophysiological level into heart failure with reduced ejection fraction (HFrEF) and heart failure with preserved ejection fraction (HFpEF) (34). HFrEF is characterized by defects in cardiac contraction, whereas HFpEF is defined at the cardiac level primarily by impaired filling. The role of cardiomyocyte death has been explored almost exclusively in HFrEF, where the attrition of cardiomyocytes (referred to as “dropout”) has been noted.

Measurement of the magnitude of decreases in numbers of cardiomyocytes is highly controversial and fraught with methodological issues, which will not be reviewed here. Because of these limitations in quantifying numbers of cardiomyocytes, markers of cell death programs have been employed. The rate of TUNEL-positive cardiomyocytes in failing human hearts has been estimated to be 0.08–0.25% as compared with 0.001–0.002% in nonfailing hearts (167, 366, 418–420). While these numbers reflect only a single cell death marker, they are instructive in two respects. First, they show that rates of cardiomyocytes death are higher in failing versus nonfailing human hearts. Second, they show that the overall magnitude of cardiomyocyte death is quite low, raising the question as to whether cardiomyocyte death is important in the pathogenesis of heart failure. As will be discussed below, genetic manipulations of cell death pathways demonstrate that cardiomyocyte death is an important component in the progression to heart failure (TABLE 3). However, in contrast to MI where cardiomyocyte death is the central event, it is one component process among many in the pathogenesis of heart failure.

B. Apoptosis and Necrosis

1. Death receptor pathway in MI

A number of studies have examined the role of death receptor signaling in I/R (TABLE 3). Some pre-dated and some post-dated the recognition that this pathway stimulates necroptosis as well as apoptosis. Accordingly, the cell death markers analyzed and accompanying mechanistic experiments often do not definitively differentiate between these death programs.

The death ligand FasL is upregulated in the rodent myocardium during global I/R in isolated, perfused hearts, although the cell type of origin was not identified (213). To test the role of Fas signaling in the pathogenesis of I/R, mice deficient in Fas (lpr mutant with global absence of Fas) were studied at an age before they manifest baseline immunological abnormalities. Reductions in infarct size accompanied by decreases in the percentage of TUNEL-positive cardiomyocytes were observed following I/R in vivo (258). Reductions in TUNEL were interpreted as indicating modulation of apoptosis. Necroptosis readouts were not assessed. These data suggest that Fas signaling contributes to myocardial damage during I/R.

Similar experiments were performed using global TNFR knockout mice in non-reperfused MI. Ablation of either TNFR1 or TNFR2 did not alter infarct size compared with controls 24 h following permanent coronary occlusion (250). In contrast, combined deletion of both receptors increased infarct size. While it is possible that this result reflects loss of the pro-survival effects of TNFR1 at complex I, it is curious that simultaneous deletion of TNFR2 was needed to reveal this effect. TNFR2, which lacks a DD, is not thought to directly signal through complex I, although it may influence TRAFs, IAPs, and NF-κB through other mechanisms (122). In addition, this receptor is mainly present in immune cells where it activates cell proliferation (514), although there is some evidence that its expression is upregulated in the heart at least late after infarction (328). This study suggests that the overall effect of TNF-α signaling in non-reperfused MI is to limit infarct size, and either TNFR1 or TNFR2 suffices for this effect.

The small molecule inhibitor necrostatin-1 (and its optimized analog necrostatin-1s) allosterically inhibits RIPK1 kinase activity (81, 82) to block necroptosis. Necrostatin-1, given at the time of reperfusion, reduced infarct size in mice subjected to I/R in vivo (448). These findings were extended to an in vivo pig model that showed that necrostatin-1, administered 10 min before reperfusion, reduced infarct size and ameliorated cardiac dysfunction in a dose-dependent manner (242). Interestingly, necrostatin-1 did not provide further reduction in infarct size in ppif −/− mice (lacking cyclophilin D), which were already partially cardioprotected (272) as previously shown (17). These data raise the possibility that RIPK1 and cyclophilin D, a key regulator of mPTP, signal in the same pathway. While there is no biochemical evidence linking RIPK1 and cyclophilin D/mPTP, a connection has been found between RIPK3 (which can be activated by phosphorylated RIPK1) and mPTP-mediated necrosis (563), which will be discussed below. Another caveat that may be relevant to this study is that necrostain-1 has been reported to have off-target effects (26, 60, 134, 470), one of which involves inhibition of ferroptosis (134), a death program that also mediates I/R (72, 120, 144). Markers of ferroptosis were not assessed in this study. Nonetheless, these data suggest that necroptosis may mediate I/R.

Global deletion of RIPK3 limited infarct size following I/R (351, 563). Interestingly, Ca²⁺/calmodulin-dependent protein kinase IIδ (CaMKIIδ), a facilitator of mPTP-mediated necrosis (216, 571), was identified as a novel target of RIPK3 (563). RIPK3 activates CaMKIIδ through phosphorylation and indirectly through oxidation. Furthermore, pharmacological inhibition of CaMKII, a known meditator of mPTP opening and I/R injury, blocks mPTP opening and cardiomyocyte killing specifically in response to RIPK3 overexpression (563). While mitochondria do not appear essential for TNF-α-induced necroptosis in some systems (468), it remains possible that mPTP opening contributes to RIPK3-induced cardiomyocyte killing. Whether the RIPK3-CaMKIIδ connection obviates the need for MLKL in bringing about the necrosis phenotype was also explored with MLKL knockdown, but was inconclusive because the resulting depletion of MLKL was incomplete (563). These data suggest, however, that RIPK3 provides a connection between necroptosis and mPTP-mediated necrosis.

Several studies have tested the involvement of microRNAs (miRs) and long noncoding RNAs (lncRNAs) in regulating death receptor signaling during I/R. For example, miR-103/107 can be upregulated by oxidative stress in cell systems to decrease FADD abundance and unleash necroptosis (507). Expression of miR-103/107 is upregulated in the ischemic region during I/R in vivo, and administration of specific anti-miRs significantly attenuates necrosis, infarction, and improves cardiac function (507). miR-103/107 is functionally inhibited by the lncRNA H19. This may serve to mitigate the FADD lowering and necroptosis-promoting effects of miR-103/107, although the involvement of this lncRNA was not tested in vivo. In contrast to miR-103/107, which promotes necroptosis, miR-873 inhibits this cell death program by decreasing the abundance of RIPK1 and RIPK3 in cardiomyocytes (508). miR-873 is downregulated during I/R, and intracoronary delivery of miR-873 limits I/R-induced necrosis and infarct size. miR-873 abundance is regulated by a lncRNA named necrosis-related factor (NRF) that functions as a sponge. NRF abundance is upregulated during I/R, which is functionally important as its silencing decreases infarct size. Interestingly, NRF appears to be a transcriptional target of p53 (508). A third study identified miR-223 as a potential regulator of multiple components of the death receptor pathway, including TNFR1, Death Receptor 6 (DR6), and IKKα, as well as NOD-like receptor family pyrin domain containing 3 (NLRP3), an inflammasome component involved in pyroptosis (388). miR-223 abundance increased in the heart in response to I/R. Moreover, cardiac-specific transgenic overexpression of miR-223 reduced markers of necrosis (but not apoptosis) and infarct size, while global knockout did the opposite.

2. Mitochondrial pathway in MI

A number of studies have examined the effects of various BCL-2 proteins on I/R injury, including the pro-survival subfamily, the multidomain pro-cell death subfamily, and the BH3-only proteins. Again, many of these studies were performed before the realization that BAX and likely BAK can mediate necrosis, in addition to apoptosis, in the mitochondrial pathway (TABLE 3).

Two groups independently studied the effect of cardiac-restricted transgenic BCL-2 overexpression on I/R injury in vivo and found reductions in apoptotic markers and infarct size accompanied by amelioration of cardiac dysfunction (38, 54).

Despite the redundancy between BAX and BAK signaling in some cell systems (e.g., MEFs; Ref. 520), global deletion of BAX alone reduced infarct size, apoptotic and necrotic markers, and cardiac dysfunction following I/R in isolated, perfused hearts (192). The same group found that BAX deletion resulted in modest reductions in infarct size in non-reperfused MI in vivo accompanied by amelioration of cardiac dysfunction 28 days later (191). In a separate cohort of mice also subjected to non-reperfused MI, cardiac caspase-3 activity at 2 h (apoptosis marker) and release of cardiac enzymes into the blood at 24 h (necrosis markers) were also reduced by Bax deletion. These data show that the absence of BAX reduces infarct size and cardiac remodeling during MI.

As discussed previously, more recent work has demonstrated that, in addition to their traditional roles in promoting MOMP and apoptosis, BAX and perhaps BAK mediate primary necrosis in the mitochondrial pathway by sensitizing cells to Ca²⁺-induced mPTP opening (225, 528). Mice with cardiomyocyte-restricted deletion of Bax and global deletion of Bak exhibited reduced infarct size following I/R compared with wild-type mice, accompanied by reductions in morphological features of necrosis, including mitochondrial abnormalities and derangement of sarcomere structure, and infarct size (528). Additionally, global deletion of ppif (encoding cyclophilin D) in the context of the cardiomyocyte-restricted Bax and global Bak knockout (triple knockouts) did not result in additional reduction in infarct size (528). These data provide genetic evidence that BAX/BAK signaling resides in the same (or an overlapping) pathway with cyclophilin D, a positive regulator of mPTP opening. This result further supports the notion that BAX/BAK signal necrosis in this context.

PUMA, which is transcriptionally induced by p53, is a BH3-only protein that functions as a direct activator of BAX and BAK (57, 397). Global deletion of PUMA reduced infarct size in isolated, perfused hearts subjected to I/R, which was accompanied by decreases in apoptosis and necrosis markers and lessening of cardiac dysfunction (483).

The defining feature of mitochondrial-mediated necrosis, Ca²⁺-induced opening of the mPTP, is facilitated by cyclophilin D. Two groups independently created global knockout mice lacking cyclophilin D and observed inhibition of Ca²⁺-induced mPTP opening and necrotic cell death (17, 341). Apoptotic cell death was unaffected. Moreover, absence of cyclophilin D markedly reduced infarct size in response to I/R. Conversely, cardiac-specific overexpression of cyclophilin D caused mitochondrial swelling and cell death (17). These studies provided the initial genetic evidence that necrosis mediated through the mitochondrial pathway is important in MI.

The cyclophilin D inhibitor CsA reduces Ca²⁺-induced mPTP opening and anoxic injury in cardiomyocytes (350). Multiple studies have shown that CsA, and analogs, ameliorates myocardial damage during I/R in various animal models (367), albeit with some variability in results (273). This drug was also suggested to reduce infarct size during reperfused MI in humans, although the study involved a small number of subjects (381). Ultimately, however, the most definitive human study to date failed to show reductions in cardiac remodeling, heart failure, and mortality, the most clinically relevant end points (69). This lack of efficacy probably reflects the fact that reduction of infarct size was not replicated in this cohort (69, 278). As will be discussed below, these data do not negate the possibility that cyclophilin D or mPTP is an important therapeutic target in I/R. Rather, they raise the possibilities that CsA may not be the optimal drug and/or that other death pathways need to be simultaneously blocked.

3. Molecules that operate in both death receptor and mitochondrial pathways in MI

As previously discussed, cIAP1, cIAP2, and XIAP play roles in both the death receptor and mitochondrial pathways. In the death receptor pathway, they activate survival pathways through ubiquitination events in complex I that suppress apoptosis and necroptosis (434). In the mitochondrial pathway, IAPs limit apoptosis by inhibiting caspases (87, 89, 108, 407). cIAP2 is upregulated by ischemia in the heart, and transgenic mice with cardiac-specific overexpression of cIAP2 exhibited modest reductions in infarct size following I/R (62) (TABLE 3). This was accompanied by reductions in both blood levels of cardiac enzymes (a marker of necrosis) and fewer TUNEL-positive cells (apoptosis). The study did not explore the relative extents to which these effects were mediated through inhibition of death receptor or mitochondrial pathways.

As previously discussed, the mitochondrial apoptogens SMAC/DIABLO and OMI/HtrA2 relieve IAP-mediated inhibition of caspases by directly binding IAPs and displacing caspases (185, 453, 464). In addition, OMI/HtrA2 also cleaves and permanently inactivates XIAP through its serine protease function (118, 464, 537). UCF-101, a small molecule inhibitor of the OMI/HtrA2 serine protease activity, was shown to reduce XIAP loss, caspase activation, and infarct size following I/R when administered 10 min before reperfusion (24, 282). The efficacy of UCF-101 in limiting infarct size is notable because its presumed site of action is postmitochondrial, meaning after potential mitochondrial dysfunction has been induced during apoptosis. This raises the possibility that survival may be promoted through additional mechanisms such as IAP-mediated effects at complex I.

Apoptosis repressor with caspase recruitment domain (ARC) (149, 241), a CARD-containing protein that is highly expressed in cardiomyocytes, is a potent inhibitor of several cell death programs (131, 313, 347). ARC blocks death receptor-mediated apoptosis through interactions with FADD that interfere with DISC formation (347). ARC suppresses mitochondrial-mediated apoptosis by inhibiting BAX conformational activation through direct binding (347). An additional mechanism by which ARC inhibits mitochondrial death signaling is through its interaction with p53 in the nucleus (131). This inhibits p53 tetramerization, which both interferes with p53 function as a transcription factor and promotes export from the nucleus. Finally, ARC attenuates the ER stress response in some cell types through mechanisms involving CHOP (313). Although ARC undergoes proteasomal degradation during oxidative stress and I/R (130, 348), cardiomyocyte-specific transgenic overexpression reduces infarct size (386). Global ARC deletion has also been reported to exacerbate infarct size following I/R (99). But this study was performed in a small number of mice of mixed genetic background and has not been replicated using large numbers of genetically homogeneous mice (R. N. Kitsis, unpublished data). The lack of an effect of ARC deletion on increasing infarct size may reflect that endogenous ARC is also degraded during I/R in wild-type mice.

Various caspase inhibitors (typically poly- or pan-caspase inhibitors) have been shown to variably reduce infarct size in mice subjected to I/R accompanied by amelioration of cardiac dysfunction (195, 198, 538, 544). Although decreases in apoptosis markers were observed in most studies, the precise caspases that were antagonized and death programs that were blocked were not defined in most cases. Of particular interest is whether caspase inhibition attenuated mitochondrial dysfunction during I/R and, if so, what is the mechanism. Additionally, among the myriad substrates that are cleaved by caspase-3 during apoptosis, some are contractile proteins such as troponin T and α-actinin (65), the preservation of which by caspase inhibition could potentially lessen cardiac dysfunction.

While the above experiments showed some efficacy of caspase inhibitors to decrease infarct size following I/R, one study using combined deletion of procaspases-3 and -7 failed to show any effect on acute infarct size, postinfarct heart failure, or cardiomyocyte death, also using an I/R model (206). This study is important because caspases-3 and -7 are the major effector caspases. A limitation of this study, however, is that the third effector caspase, caspase-6, was not eliminated. Although its repertoire of substrates is more limited than that of caspases-3 and -7, it is difficult to know whether some compensation occurred.

4. Death receptor pathway in heart failure

As previously noted, 0.08–0.25% of cardiomyocytes in failing human hearts are TUNEL-positive compared with 0.001–0.002% in controls (167, 366, 418–420). A gain-of-function approach was used to assess the significance of low, but abnormal, rates of cardiomyocyte death in the pathogenesis of heart failure. Transgenic mice were created with cardiomyocyte-specific overexpression of an inducible procaspase-8 mutant (524) (TABLE 3). This mutant encoded a fusion protein consisting of a myristoylation signal (for plasma membrane targeting), three FK506 binding protein modules (FKBP), and the p20 and p10 subunits of procaspase-8. It had previously been shown in cell culture that addition of a small molecule that binds the FKBP modules could activate procaspase-8 by forced proximity (339). When this small molecule was administered to mice, they died acutely due to massive cardiomyocyte death, a result that is not informative about the pathogenesis of human heart failure. However, when left untreated for several months, spontaneous heart failure and mortality were noted in these transgenic mice, but not in wild-type mice or mice expressing a catalytically inactive version of the same transgene. Importantly, rates of TUNEL-positive cardiomyocytes in the mice dying of heart failure were 0.023% compared with 0.002% in wild-type mice and those with the catalytically dead mutant. Moreover, delayed administration of a pan-caspase inhibitor to the transgenic mice destined for heart failure rescued both cell death and heart failure. This study demonstrated that rates of cardiomyocyte death that are even 4- to 10-fold lower than those observed in failing human hearts are adequate to cause a lethal dilated cardiomyopathy. It should be noted that this study used caspase-8 activation as an experimental tool to address this question, but did not explore the role of caspase-8 and the death receptor pathway per se in mediating heart failure in response to various pathological stimuli.

TRAF2 is usually thought to be an adaptor protein in complex I, but it also contains a RING domain with E3-ligase activity (7, 259). While cIAP1/2 and LUBAC are generally considered the major enzymes in complex I responsible for adding K63-linked and linear ubiquitin chains, respectively, to RIPK1, it is possible that TRAF2 also plays a role. Cardiomyocyte-specific transgenic expression of TRAF2 activates NF-κB (41). Furthermore, cardiomyocyte-specific deletion of TRAF2 in mice, or expression of a TRAF2 mutant lacking the RING domain (which appears to function as a dominant negative) in neonatal cardiomyocytes, induces apoptosis and necrosis (169). The apoptosis component is sensitive to pharmacological inhibition of RIPK1 kinase activity by necrostatin-1. The necrosis component is mitigated by RIPK3 absence or MLKL knockdown. A constitutively active TAK1 mutant protects against both apoptosis and necrosis resulting from the TRAF2 dominant negative mutant expression, similar to the baseline suppression of both of these forms of cell death in vivo by endogenous TAK1 (265). While the effects of the dominant negative TRAF2 mutant suggest that its E3-ligase function is important in suppressing cell death at least in the heart, the precise mechanism remains to be defined. Potentially relevant TRAF2 substrates include RIPK1 and TAK1 (119, 562), but their roles remain to be tested. The preceding mechanism may be important in the pathogenesis of heart failure because TRAF2 abundance increases in response to pressure overload and post-MI heart failure, and cardiomyocyte-specific TRAF2 deletion induces dilated cardiomyopathy and heart failure (169). Thus TRAF2 appears important in the heart in pathological, as well as baseline, conditions.

The role of RIPK3 in heart failure was tested using a non-reperfused MI model in which infarct size was unchanged (298). Global RIPK3 knockout reduced adverse cardiac remodeling and dysfunction at 30 days. Markers of necroptosis signaling were not assessed in this study. The anthracycline doxorubicin, a chemotherapeutic agent used to treat multiple solid tumors and leukemias in adults and children, causes a dose-dependent cardiomyopathy in ~10% of patients. While the etiology of doxorubicin-induced cardiomyopathy is complex, apoptotic and necrotic cell death is believed to be involved. Global absence of RIPK3 also attenuated doxorubicin-induced cardiac dysfunction (563).

5. Mitochondrial pathway in heart failure

In light of the previously discussed role of cyclophilin D in mPTP opening and necrosis during I/R (341), the role of this pathway was also tested in heart failure models. Given the importance of Ca²⁺ as a trigger for mPTP opening, this was first done in mice with cardiac Ca²⁺ overload. This was induced by cardiomyocyte-specific overexpression of the β2a subunit of the plasma membrane L-type Ca²⁺ channel (LTCC) (346) (TABLE 3). These mice exhibited cardiomyocyte Ca²⁺ overload, cardiomyocyte necrosis, heart failure, and premature death. To test the importance of necrosis in causing heart failure, the transgenic mice overexpressing the β2a subunit of the LTCC were crossed with global ppif −/− mice. Interestingly, heart failure was rescued by the absence of cyclophilin D, but not by overexpression of BCL-2, implicating necrotic cell death in pathogenesis. In the same study, cyclophilin D loss was also shown to rescue fibrosis (as a surrogate for necrosis) and cardiac dysfunction resulting from administration of doxorubicin (346). This experiment suggests that the role of cyclophilin D and necrotic cell death in heart failure may extend beyond just Ca²⁺ overload and implicates mitochondrial-mediated cardiomyocyte necrosis in heart failure.

To further extend the role of cyclophilin D/mPTP and cardiomyocyte necrosis to other clinically relevant heart failure models, ppif −/− mice were subjected to pressure overload, a stress common to heart failure of multiple etiologies (114). Unexpectedly, pressure overload in ppif −/− mice resulted in marked systolic dysfunction accompanied by chamber dilation within 7 days, while wild-type controls manifest only relatively mild cardiac dysfunction at 6 wk. Pressure overload did not significantly increase the frequency of TUNEL-positive cardiomyocytes in either ppif −/− or wild-type mice. Thus the heart failure phenotype exhibited by ppif −/− mice was not cell death related. In contrast, transcript profiling in the baseline state suggested changes in substrate utilization away from fatty acids and toward glucose in ppif −/− mice, a metabolic change often observed in failing hearts (127). This effect was hypothesized to result from impaired Ca²⁺ efflux from the mitochondrial matrix in ppif −/− mice resulting in activation of Ca²⁺-dependent dehydrogenases. Interestingly, a swimming regimen, a stimulus that induces physiological cardiac hypertrophy with maintained or augmented systolic function, induced mortality in ppif −/− mice within 5 days. An exaggerated hypertrophic response accompanied by pulmonary edema was observed in ppif −/−, but not wild-type, mice. This study demonstrates that complex metabolic changes leading to heart failure are induced by pathological and physiological hypertrophic stimuli in the absence of cyclophilin D.

BCL-2/adenovirus E1B 19-kDa interacting protein 3 (BNIP3) and 19-kDa interacting protein 3-like protein X/BCL-2/adenovirus E1B 19-kDa interacting protein 3-like (NIX/BNIP3L) are two related BH3-only-like proteins that mediate a complex mixture of apoptotic and necrotic cell death (36, 49, 311). Each of these proteins also serves as adaptors between ubiquitinated OMM proteins and microtubule-associated protein 1 light chain 3 (LC3) in some mitophagy paradigms (561). The precise mechanisms by which BNIP3 and NIX/BNIP3L induce cell death are not understood, nor is the relationship between the cell death and mitophagy functions. Unlike classic BH3-only proteins, the BH3 domain of BNIP3 and NIX/BNIP3L is not required for cell death. Rather, mitochondrial targeting, which is mediated by the transmembrane domain in each protein, is obligatory for this function. BNIP3 and NIX/BNIP3L are each expressed as two isoforms generated by alternative splicing. The long isoform contains the transmembrane domain and induces cell death. The short isoform lacks the transmembrane domain and heterodimerizes with the long isoform to antagonize cell death. The relevance of these proteins to cardiac disease resides in their regulation by pathological stimuli. BNIP3 abundance increases at the transcriptional level in response to hypoxia (168, 396), while NIX/BNIP3L is transcriptionally upregulated by signaling mediated by Gα_q, a protein coupled to certain cell surface receptors (angiotensin II type 1 receptor, endothelin receptor, and α₁-adrenergic receptor) that mediate cardiac hypertrophy (557).

BNIP3 appears to induce primarily apoptotic cardiomyocyte death during cardiac remodeling post-I/R (96) and necrotic cardiomyocyte death following doxorubicin treatment (90). With regard to the I/R studies (96), cardiomyocyte-specific deletion of Bnip3 does not influence infarct size, perhaps because it is not yet fully induced by hypoxia in the time frame of cardiomyocyte killing in this model. However, cardiomyocyte-restricted absence of this protein decreases post-I/R apoptotic cell death in the peri-infarct zone and remote myocardium, thereby attenuating adverse cardiac remodeling and heart failure. Conversely, inducible, cardiomyocyte-specific transgenic overexpression of BNIP3 stimulates pathological modeling and heart failure even in the absence of infarction. Doxorubicin also induces BNIP3 expression in cardiomyocytes, and deletion of Bnip3 in these cells in vivo protects against abnormalities in mitochondrial morphology and respiration and necrotic cell death induced by doxorubicin (90). One relevant mechanism appears to be that the absence of BNIP3 prevents doxorubicin-induced disruption of an interaction between cytochrome c oxidase subunit 1 (COX1) and uncoupling protein 3 (UCP3), two proteins that reside in the IMM.

NIX/BNIP3L is important in mediating cardiomyocyte apoptosis mediated through Gα_q (557). Transgenic mice with cardiac-restricted overexpression of Gα_q develop cardiac hypertrophy, cardiomyocyte apoptosis, and heart failure (3, 71). In addition, a proportion of mice go into heart failure during pregnancy (3), likely induced by the hemodynamic stresses of pregnancy, rather than the vascular pathology that operates in human peripartum cardiomyopathy (376). Pregnancy-related heart failure in Gα_q transgenic mice can be rescued by caspase inhibition (180). As previously noted, NIX/BNIP3L is upregulated by Gα_q. Cardiomyocyte-restricted transgenic overexpression of NIX/BNIP3L phenocopies the cardiomyocyte apoptosis and heart failure exhibited by Gα_q transgenic mice (557). Moreover, transgenic overexpression of the short inhibitory isoform of NIX/BNIP3L attenuates cardiomyocyte apoptosis and heart failure in Gα_q transgenic mice. A proportion of NIX/BNIP3L may reside at the ER/SR, as well as at the mitochondria. Selective targeting of NIX/BNIP3L to mitochondria appears to induce primarily BAX/BAK and caspase-dependent apoptotic death, while targeting to ER/SR seems to induce primarily mPTP-dependent necrosis (53).

C. Other Death Programs

1. Autophagy-dependent cell death

Autophagy is an evolutionarily conserved process that degrades cellular components in the lysosome. It plays important roles in maintaining quality control at baseline and is augmented in response to stress to restore cellular homeostasis (326). Given its role in attenuating cellular stress, autophagy generally functions as a survival process (261). However, some evidence suggests that, under certain circumstances, autophagy may beget cell death. Autophagy-dependent cell death remains a controversial concept. It is defined as a regulated form of cell death that is caused by autophagy, as opposed to occurring in parallel with, or as a compensation for, autophagy (142). It remains unclear if autophagy-dependent cell death results from excessive rates of autophagy (possibly reflecting excessive catabolism of cellular components) or whether it is attributable to a qualitative change in the nature of the autophagy process itself (FIGURE 5). It should be noted that autophagy-dependent cell death is not synonymous with another regulated cell death program termed lysosome-dependent cell death (142, 433), which is caused by the release of lysosomal enzymes into the cytosol resulting from lysosomal membrane permeabilization. In addition, it is not clear if leakiness of the lysosomal membrane is a feature of autophagy-dependent cell death.

FIGURE 5.

Autophagy-dependent cell death. The “conventional” view of autophagy-dependent cell death is that increases in rates of macroautophagy induce cell death through excessive catabolism of cellular components. Specific death machinery and specific autophagy-dependent cell death markers have not been identified. In contrast, autosis is a form of autophagy-dependent cell death that is characterized by a specific cellular morphology (see text) and mediated by the plasma membrane Na⁺-K⁺-ATPase.

Before considering autophagy-dependent cell death, it should be noted that several autophagy programs have been described including macroautophagy, chaperone-mediated autophagy, microautophagy, and endosomal microautophagy (139). Studies pertaining to autophagy-dependent cell death have focused on macroautophagy, which will be referred to hereafter simply as autophagy. This process degrades substrates including macromolecules (e.g., proteins, lipids, and nucleic acids) and organelles by transporting them to lysosomes in double membrane vesicles termed autophagosomes. Autophagosomes subsequently fuse with the lysosomal membrane delivering their cargo to the lysosomal lumen where acidic hydrolases reside. Detailed descriptions of this process can be found elsewhere (139, 326). The selection of substrates in autophagy may be nonselective (in bulk) or selective, an example of the latter being the mitophagic removal of damaged mitochondria, which are marked by ubiquitination of OMM proteins. Autophagy is activated in response to diverse cellular stresses including starvation, ischemia, oxidative stress, genotoxic stress, proteotoxic stress, and infection.

Autophagy-dependent cell death was historically defined as “evidence of cell death” accompanied by cytoplasmic autophagic vacuoles (e.g., autophagosomes), these features most easily identified by electron microscopy (80). To reemphasize, however, this morphological definition does not discriminate among the possibilities that cell death is 1) being caused by autophagy, 2) accompanying autophagy but unrelated, or 3) inducing autophagy as a compensatory mechanism. The deficiencies of this morphological definition underscore the importance of 1) establishing cause-and-effect between autophagy and cell death and 2) having specific death readouts that reflect autophagy-dependent cell death. As described below in the heart, the cause-and-effect issue has been addressed by the studies in which autophagy is perturbed genetically following which cell death is assessed. However, death readouts specific for autophagy-dependent cell death still have not been identified, with the exception of a subset of autophagy-dependent cell death termed autosis (290) that will be discussed below.

In the heart, autophagy can be activated by prolonged ischemia, I/R, and during heart failure of various etiologies (475, 570). To test whether autophagy is responsible for cell death and the pathogenesis of these syndromes, the general strategy has been to genetically modulate autophagy in these disease contexts (TABLE 3). Beclin 1 is critical for early autophagosome formation. Global deletion of a single Beclin 1 allele (as biallelic deletion in the germ line is lethal; Ref. 556) ameliorates cardiac dysfunction resulting from severe pressure overload (570). Conversely, cardiomyocyte-specific Beclin 1 overexpression does the opposite. The expected effects of these Beclin 1 manipulations on rates of autophagy were confirmed using various autophagy markers. The only cell death readout in these studies was TUNEL, a marker of apoptotic cell death. However, TUNEL was not informative in this particular context because, as is frequently observed, pressure overload did not significantly increase the frequency of TUNEL-positive cardiomyocytes even in wild-type mice. These data show that increases in Beclin 1, resulting in augmentation of rates of autophagy, exacerbate cardiac dysfunction in this severe pressure overload model. Although this effect may be occurring through increases in cell death, this was not shown explicitly. That said, this latter criticism is addressed to some extent in a second study that used the same global Beclin 1 heterozygote knockout mouse in an I/R model (310). Loss of one Beclin 1 allele reduced infarct size and frequency of TUNEL-positive cardiomyocytes, thus providing more direct evidence that autophagy may exacerbate cardiomyocyte death under some pathological conditions. It should be noted, however, that in the setting of ischemia without reperfusion, stimulation of autophagy appears to be protective (310, 427, 428).

To add further confusion, however, a third study inhibited autophagy under basal and pressure overload conditions using a different genetic approach and obtained the opposite result (344). Autophagy related 5 (Atg5), which is involved in autophagosome maturation, was deleted specifically in adult cardiomyocytes in vivo. This spontaneously resulted in severe systolic dysfunction and heart failure within 1 wk accompanied by increased frequency of TUNEL-positive cardiomyocytes. Cardiomyocyte-specific deletion of Atg5 in embryonic life caused no baseline phenotype, but severe heart failure developed within 1 wk of pressure overload. These data suggest that endogenous levels of autophagy are required for normal systolic function under basal and pressure overload conditions, and reductions in autophagy under these conditions are accompanied by increased cardiomyocyte death.

How can we reconcile these two opposing data? One obvious difference is that different genes were manipulated, with Beclin 1 loss of function ameliorating cardiac pathology (pressure overload and I/R) versus Atg5 loss of function causing pathology at baseline and with pressure overload. Arguing against this explanation, however, is a fourth study showing that heterozygous global deletion of Beclin 1 exacerbates desmin-related cardiomyopathy (475). Nevertheless, the notion that differences in Beclin 1 and Atg5 may underlie these discrepant results is not resolved. In addition to its distinct role in autophagosome formation, Beclin 1 is also a BH3-only protein that is bound and negatively regulated by BCL-2 (363, 377). While it is generally believed that the Beclin 1-BCL-2 interaction does not interfere with the anti-cell death properties of BCL-2, some data suggest that enhancement of Beclin 1-BCL-2 or Beclin 1-BCL-xL binding resulting from phosphorylation of the BH3 domain of Beclin 1 may activate BAX (300). This raises speculation as to whether loss of Beclin 1, but not Atg5, somehow mitigates cell death. A second difference between the Beclin 1 and Atg5 loss-of-function studies is that the Beclin 1 gene was deleted in a global manner and Atg5 in a cardiomyocyte-specific manner. Insufficient information exists to assess the significance of this difference. A third difference is that the severity of pressure overload may have been greater in the Beclin +/− versus Atg5 −/− study, again a consideration that cannot yet be resolved. In summary, we are unable to reconcile these discrepant results. Moreover, these studies connect changes in macroautophagy most convincingly with changes in cardiac function rather than in autophagy-dependent cell death, making it difficult to draw conclusions specifically about effects on cell death.

a) autosis.

Recently, a new program of autophagy-dependent cell death termed autosis was described (290). Autosis is characterized by morphological changes distinct from those of apoptosis or necrosis that include early nuclear convolution followed by rapid ballooning of the perinuclear space. Autosis induction tracks with increased rates of autophagy and is prevented by pharmacological or genetic inhibition of autophagy, but not by inhibition of apoptosis, necrosis, or even lysosomal enzymes. Autosis can be induced by nutrient deprivation in cell culture and by hypoxia/ischemia in the brain. Most interestingly, autosis appears to be mediated by the plasma membrane Na⁺-K⁺-ATPase, and neuroprotection results from pharmacological inhibition of this molecule (FIGURE 5). These are early days for autosis, and clearly more work will be needed to understand its mechanism and potential significance as a cell death process in various contexts including heart disease.

2. Ferroptosis

Ferroptosis is an iron-dependent form of cell death characterized by oxidative damage to cellular membranes (97, 142) (FIGURE 6). Fenton reaction-generated ferric iron (Fe³⁺) increases ROS and activates lipoxygenases that damage cellular membranes [especially phosphatidyl ethanolamine-containing polyunsaturated fatty acids (PUFAs)] (539). Lipid peroxidation can be prevented, and ferroptosis blocked, by molecules that either chelate iron (e.g., deferoxamine, dexrazoxane) or oppose lipid peroxidation (e.g., ferrostatin-1, liproxstatin-1, vitamin E) (97, 134). The major endogenous mechanism to combat lipid peroxidation is glutathione peroxidase 4 (GPX4) (51, 308). GPX4 function requires the cofactor glutathione (GSH) and is therefore dependent on the plasma membrane cystine/glutamate antiporter system x_c⁻ because cystine is needed to generate GSH (97). Molecules that inhibit system x_c⁻ (e.g., erastin, sulfasalazine, sorafenib, glutamate) deplete the cellular pool of GSH and inhibit GPX4 function leading to ferroptosis (540). The morphological manifestations of ferroptosis are distinct from those of apoptosis and necrosis and include dense, compact mitochondria with loss of cristae (97). Some studies have reported OMM rupture (134, 535), but this is not a consistent or characteristic feature (B. Stockwell, personal communication). Loss of plasma membrane integrity has been used as a readout for ferroptosis (1), but it is presently unclear whether this is part of the primary process or simply reflects late downstream effects of oxidative damage. In fact, one of the most important unanswered questions is how ferroptosis kills cells, specifically it is not clear as to which cellular membranes need to be damaged to bring about this form of cell death (123).

FIGURE 6.

Overview of ferroptosis. Ferroptosis is an iron-dependent form of regulated cell death mediated by lipid peroxidation of cellular membranes, although which damaged membranes induce cell death is not yet clear. Ferroptosis can be initiated by iron overload in which Fenton reaction-produced Fe³⁺ activates lipoxygenases or by mechanisms that inactivate glutathione peroxidase 4 (GPX4), which opposes lipid peroxidation. GPX4 can be inactivated by extracellular glutamate overload, which inhibits antiporter system x_c⁻ . System x_c⁻ normally imports cystine, which is converted to cysteine that subsequently generates glutathione (GSH), a cofactor for GPX4. Fe³⁺ imported through the transferrin receptor (TR) is converted to Fe²⁺ in endosomes by the metalloreductase six-transmembrane epithelial antigen of prostate 3 (Steap3) and released from endosomes by divalent metal transporter 1 (DMT1). Erastin, sulfasalazine, and sorafenib are small molecules that inhibit system x_c⁻ to induce ferroptosis. Ras synthetic lethal 3 (RSL3) and FIN56 are small molecules that inhibit GPX4 also inducing ferroptosis. Ferrostatin-1, liproxstatin-1, vitamin E, and CoQ10 are lipid antioxidants that inhibit ferroptosis. In addition, FIN56 also inhibits CoQ10 through the mevalonate pathway. Deferoxamine (DFO) binds Fe²⁺ mainly extracellularly, while dexrazoxane and ciclopirox (CPX) bind Fe²⁺ intracellularly.

Ferroptosis has been implicated in cell death in a variety of disease settings including treatment with cancer drugs (540), glutamate-induced neuronal cell death and neurodegenerative disease (97, 446), drug-induced liver damage (295), acute hypoxia-induced kidney injury (134, 279, 446), and I/R in the liver (134) and heart (72, 120, 144). A strong rationale for investigating ferroptotic death in the heart is the known cardiotoxicity of iron and the fact that reperfused MI in humans is associated with iron overload in the peri-infarct zone (40).

Five studies involving ferroptosis are directly relevant to the heart (15, 72, 120, 144) (TABLE 3). The first confirmed that iron overload takes place in cardiomyocytes and nonmyocytes during I/R in vivo (15). It then went on to demonstrate that ferroptosis can be induced in primary cultures of adult mouse cardiomyocytes using either Fe³⁺ or either of two small molecule activators of ferroptosis: the system x_c⁻ inhibitor erastin or the GPX4 inhibitor Ras synthetic lethal 3 (RSL3). The read-out employed for cell death was loss of plasma membrane integrity, which could be rescued by ferrostatin-1, a lipophilic antioxidant. Importantly, cell death occurred independently of caspase-3 activation or alterations in autophagy. While, as discussed above, loss of plasma membrane integrity is not specific for ferroptosis and may not even be part of the direct mechanism, the fact that cell death could be induced by erastin and RSL3, which activate specific steps in ferroptotic cell death, demonstrates that ferroptosis can be induced in cardiomyocytes. Since mammalian target of rapamycin (mTOR) has previously been shown to decrease infarct size in myocardial I/R in vivo (13) and because mTOR also impacts iron handling (166), this study also investigated the effects of mTOR overexpression and knockout on cell death induced by Fe³⁺ and ferroptosis agonists in cultured adult cardiomyocytes. mTOR overexpression reduced, and knockout increased, cell death induced by these stimuli, and the mechanism may involve mTOR modulation of ROS production.

A second study showed that the iron chelator deferoxamine reduced infarct size following global I/R in isolated, perfused mouse hearts (144). This same study, which implicated glutaminolysis in the pathogenesis of ferroptosis, showed that a small molecule inhibitor of glutaminases also reduced infarct size in the same system.

A third study, which was published before ferroptosis was discovered, provides complementary information (72). Again, using global I/R in isolated, perfused hearts, the data show that global transgenic overexpression of a mitochondrial-targeted mutant of GPX4 is cardioprotective as measured by reduced release of creatine kinase, an indicator of loss of cell membrane integrity. Mitochondrial-targeted GPX4 also reduced mitochondria lipid peroxidation, and lessened mitochondrial and contractile dysfunction post-I/R.

A fourth study demonstrated that treating mice with either ferrostatin-1, or the iron chelator dexrazoxane, reduced infarct size as well as serum markers of myocardial injury during I/R in vivo (120). These compounds also improved survival and prevented cardiac dysfunction in mice treated with doxorubicin, which induces mitochondrial iron load and lipid peroxidation.

A fifth study showed that ferrostatin-1 blocked cardiomyocyte death both in heart transplantation and the traditional coronary artery ligation I/R models in vivo (268). In the transplantation model, ferrostatin-1 reduced levels of hydroperoxy-arachidonoyl-phosphatidylethanolamine, a mediator of ferroptosis, and deaths of cardiomyocytes and cardiac fibroblasts, but not endothelial cells. Furthermore, ferroptosis was shown to promote neutrophil recruitment in the transplantation model through a TLR4/TRIF/type 1 interferon (IFN) signaling pathway specifically in coronary vascular endothelial cells. Since ferrostatin-1 inhibited neutrophil recruitment both in this model and the I/R model, inhibition of cardiomyocyte death may have occurred, in part, through antagonizing neutrophil-dependent mechanisms. However, ferrostatin-1 also reduced infarct size in isolated buffer-perfused hearts, indicating that a component of the rescue involved inhibition of ferroptosis in cardiac cells.

Taken together, these studies show that ferroptosis can take place in adult cardiomyocytes. In addition, they implicate ferroptosis in the pathogenesis of myocardial I/R. An important caveat, however, is that the conclusion that ferroptosis is responsible for cell death in this syndrome is based primarily on the observation that inhibition of myocardial damage results from molecules known to block ferroptosis. Data that would strengthen the involvement of ferroptosis in this paradigm would include markers of the specific molecular events that mediate ferroptosis. These might include increases in pathogenic modified fatty acids such as hydroperoxy-arachidonoyl-phosphatidylethanolamine (268) and inhibition of GPX4 or system x_c⁻ (97). The role of ferroptosis in other cardiac syndromes, such as heart failure, also merits exploration.

3. Pyroptosis

Pyroptosis is a regulated form of cell death that is closely tied to the innate immune response and characterized by permeabilization of the plasma membrane and extracellular release of inflammatory cytokines (FIGURE 7). Plasma membrane rupture during pyroptosis is mediated by gasdermin D (GSDMD) (184, 230, 439). In healthy cells, GSDMD is held functionally inactive through an intramolecular interaction between its COOH-terminal (inhibitory) and NH₂-terminal (pro-death) domain. Proteolysis of GSDMD by inflammatory caspases (human caspases-1, -4, and -5 and mouse caspases-1 and -11) generates the functionally competent NH₂-terminal cleavage product (GSDMD-NT) (94, 286). GSDMD-NT preferentially binds to components primarily found in the plasma membrane inner leaflet of mammalian cells (phosphatidylinositol phosphates and phosphatidylserine), and the inner and outer leaflets of bacterial cells (cardiolipin), but does not bind to components present in the outer leaflet of mammalian cells (phosphatidylethanolamine and phosphatidylcholine) (286). Through still largely undefined mechanisms, GSDMD-NT monomers oligomerize to form pores in the plasma membrane and cause the release of intracellular material inducing cell death (94, 286). Interestingly, GSDMD itself is also released from cells undergoing pyroptosis, but does not damage neighboring cells likely due to the aforementioned membrane component binding specificity that prohibits association with mammalian outer leaflets (286). Gasdermin E [GSDME; also known as Deafness, autosomal dominant 5 (DFNA5)] is a second member of the conserved gasdermin family that displays a similar membrane-perforating function resulting in pyroptotic-like cell death. Importantly, however, GSDME is processed into its active NH₂-terminal form by caspase-3 rather than by inflammatory caspases as is the case for GSDMD (406, 512).

FIGURE 7.

Overview of pyroptosis. The sentinel event in pyroptosis is the creation of pores in the plasma membrane by gasdermin D (GSDMD). GSDMD is activated through cleavage by inflammatory caspases (caspases-1, -4, and -5 in humans). Procaspase-1 is activated in response to diverse stresses including infection, toxic insults impinging on the cell, and multiple intracellular stresses [e.g., reactive oxygen species (ROS), cytosolic DNA, and others]. Pathogen-associated molecular patterns (PAMPs) and danger-associated molecular patterns (DAMPs) resulting from these stresses promote assembly of inflammasomes, which are subcellular multiprotein complexes in which procaspase-1 activation takes place. This involves the sensing of PAMPs and DAMPs by pattern recognition receptors such as NOD-like receptor (NLR) pyrin domain (PYD) containing receptor 3 (NLRP3). NLRP3 subsequently recruits ASC, which recruits and activates procaspase-1 (see text for details). Other proteins besides NLRP3 can also function as the “apical” protein in inflammasome assembly. Inflammasome formation is “primed” through transcriptional and nontranscriptional mechanisms (see text for latter). Transcriptional priming occurs downstream of Toll-like receptors (TLRs) and TNFR1 through NF-κB-mediated increases in NLRP3, ASC, and procaspase-1 expression. In addition, the expression of downstream effectors in this pathway including the cytokines pro-interleukin (IL)-1β and pro-IL-18 are also transcriptionally upregulated. Noncanonical activation of procaspase-4 and -5 can also take place through their binding to lipopolysaccharide (LPS) in the cytoplasm. Once activated, inflammatory caspases cleave and activate GSDMD as noted above and process pro-IL-1β and pro-IL-18 to their mature forms, which escape through GSDMD-generated pores. Interaction of NLRP3 with mitochondrial antiviral signaling protein (MAVS) and cardiolipin facilitate inflammasome assembly and activation.

As noted, inflammatory caspases cleave GSDMD to GSDMD-NT (230, 439), although caspase-8 may also contribute in certain contexts (370). Activation of procaspase-1 occurs in multiprotein complexes known as inflammasomes. Canonical inflammasomes respond to diverse pathogen and host-derived danger signals, and typically consist of a pattern recognition receptor (PRR), which detects PAMPs and danger-associated molecular patterns (DAMPs), an adapter protein, and procaspase-1. Classic PRRs include NOD-like receptor (NLR) family members, the most well studied of which is NLR pyrin domain (PYD) containing receptor 3 (NLRP3). In the NLRP3 inflammasome, NLRP3 recruits the adapter protein apoptosis-associated specklike protein containing a CARD (ASC) through interactions involving the PYD in each protein, thereby providing a nucleation platform for ASC filament formation (296). Filamentous ASC “specks,” in turn, facilitate the recruitment and activation of caspase-1 through CARD-CARD interactions (256, 491). Noncanonical caspase activation may take place through the binding of gram-negative bacteria-derived lipopolysaccharide (LPS) to procaspases-4, -5, or -11 in the cytosol, which then triggers cleavage and activation (231, 440). Active caspase-1 processes pro-IL-1β and pro-IL-18 into mature IL-1β and IL-18 and promotes plasma membrane rupture via GSDMD culminating in lytic cell death and a pro-inflammatory milieu. Caspase-11 is sufficient to cleave GSDMD and induce cell death but does not directly process pro-IL-1β or pro-IL-18.

Inflammasome assembly depends on a two-step process that involves priming and activation. Priming occurs in response to receptor activation including engagement of TLR2, -3, or -4; TNFR1; or IL-1 receptor. Priming can be transcriptionally independent or dependent. Nontranscriptional, or early, priming occurs rapidly and transiently (10 min to 1 h after stimulation) as a result of posttranslational modifications of NLRP3. These include phosphorylation (451, 452, 457), ubiquitination (176, 385, 450, 536), and nitrosylation (188), events that are regulated proximally by the adapter proteins myeloid differentiation primary response 88 (MyD88), IL-1 receptor-associated kinase 1 (IRAK-1), and TRIF (126, 202, 217, 275, 393). Conversely, transcriptional priming takes longer (several hours) and is mediated by NF-κB, which elicits increased de novo expression of certain cytokines (pro-IL-1β, pro-IL-18) and inflammasome components (NLRP3, ASC, procaspase-1) that normally have low baseline expression (19, 170).

The intracellular location and precise mechanisms leading to inflammasome activation are still largely undefined. Studies in macrophages have demonstrated that mitochondria serve as nodes for canonical NLRP3 inflammasome assembly and signaling. In response to stimuli that trigger inflammasome activation, NLRP3 is recruited to mitochondria through an interaction with an OMM adapter referred to as mitochondrial anti-viral signaling protein (MAVS) (374, 458). The mitochondrial membrane component cardiolipin also binds to NLRP3 and may facilitate its subcellular translocation (208). Increases in mitochondrial-derived ROS and the oxidation of liberated mitochondrial DNA (mtDNA) from dysfunctional or damaged mitochondria drive assembly and activation of the NLRP3 inflammasome (343, 441, 565, 568). Oxidized mtDNA can also directly bind to NLRP3 to promote inflammasome function and subsequent caspase-1 activation (441). Interestingly, mitophagy is also stimulated during this process and regulates caspase-1 activity and inflammatory cytokine processing through the clearance of damaged mitochondria, thereby providing an important negative feedback mechanism (343, 566). Inflammasomes themselves can also be targeted for autophagic degradation, thereby promoting resolution of the inflammatory response (437). Perhaps, not surprisingly, inhibition of autophagy has been shown to promote inflammasome activation (343). While the role of these mitochondrial factors in modulating NLRP3 inflammasome activity has been demonstrated primarily in immune cells, their necessity in cardiac cell types, including cardiomyocytes, requires further investigation.

During I/R or non-reperfused MI, the expression of canonical inflammasome components (including ASC and NLRP3) is upregulated, as are the activation of caspase-1 and the secretion of pro-inflammatory cytokines IL-1β and IL-18 (229, 319, 417). However, the cell types in which inflammasome assembly and resulting cytokine processing take place remain unclear. In fact, it appears that inflammasome assembly is significantly more robust in non-cardiomyocytes (perhaps fibroblasts) and bone marrow-derived cells that are recruited into the heart during infarction than in cardiomyocytes themselves (229), although there is evidence that inflammasomes also assemble in cardiomyocytes to a more limited extent (319, 459, 529). Adoptive transfer experiments using ASC −/− mice as donors or recipients demonstrated that loss of inflammasome function in either the bone marrow compartment or in non-cardiomyocyte cardiac resident cells is sufficient to protect the heart from I/R. This suggests that inflammasome function is needed in both compartments for full infarction to take place (229) (TABLE 3). Cardiomyocyte-specific transgenic overexpression of procaspase-1 demonstrated that high levels of this inflammatory caspase are sufficient to activate caspase-3 (possibly an overexpression effect) and exacerbate apoptosis and infarct size in response to I/R (467), as well as under basal conditions, resulting in heart failure (317). Interestingly, however, this occurs in the absence of IL-1β and IL-18 maturation, despite the high levels of caspase-1 transgene expression in cardiomyocytes (317). This perhaps reflects sluggish inflammasome assembly in this cell type. Conversely, global deletion of procaspase-1 or ASC reduces infarct size following I/R (229), and global deletion of procaspase-1 ameliorates post-MI heart failure (317).

Since global knockout mice were used in all of these studies, however, it cannot be concluded that these effects are mediated by loss of caspase-1 in cardiomyocytes, especially in light of the results described above demonstrating that inflammasome assembly is more robust in non-cardiomyocytes (229). Targeting components of the inflammasome machinery using cell type-specific deletion approaches would help to clarify this issue. An important complementary approach would be to assess the effects of GSDMD ablation in various cardiac cell types.

4. Parthanatos

Parthanatos is initiated by, and dependent on, the hyperactivation of PARP-1, the subsequent accumulation of PAR polymers in the cytosol, and AIF-dependent DNA fragmentation (121). In addition to these biochemical signatures, parthanatos is distinguished from apoptosis by its lack of inter-nucleosomal DNA fragmentation and apoptotic bodies and from necrosis by the absence of cellular swelling (12, 552, 553). The prototypical stimulus for PARP-1 activation is DNA damage. However, oxidative stress, hypoxia, inflammatory stimuli, and excitotoxcity (in neurons) can also promote PARP-1 activation (70, 78, 121, 560). During parthanatos, elevated levels of PAR polymers precipitate the cleavage and release of AIF from mitochondria through mechanisms that remain incompletely understood but require calpains, BID, and BAX (42). The direct binding of AIF and PAR polymers may also facilitate this process (334, 513, 553). AIF, which lacks nuclease function, mediates large-scale DNA fragmentation and chromatin condensation independently of caspases or EndoG (307, 505, 513, 534). Instead, AIF binds and recruits MIF, a cytokine recently demonstrated to have nuclease activity, to the nucleus to execute DNA fragmentation (511). In addition, the excessive generation of PAR polymers by PARP-1 simultaneously depletes the cell of NAD⁺ and ATP, as discussed in section II, resulting in energetic collapse that is also thought to contribute to cytotoxicity and may provide an amplification loop for necrotic cell death (121).

Reperfusion of the infarcted myocardium generates a burst of reactive oxygen and nitrogen species that can promote the activity of PARP-1 (160, 297). Indeed, elevated PARP-1 expression and activation have been observed in rodent hearts after MI and failing human myocardium (269, 378, 379, 484). By utilizing global Parp-1 knockout mice, several independent groups have demonstrated that loss of PARP-1 confers cardioprotection against I/R injury. Isolated, perfused Parp-1 −/− hearts had more viable myocardium, improved contractility, and higher NAD⁺ content compared with wild-type counterparts following I/R (165, 378, 567) (TABLE 3). Consistent with these findings, Parp-1 −/− mice exhibited reduced infarct size and lower serum creatine kinase levels compared with wild-type mice following I/R in vivo (543, 573). With the use of an isolated, perfused rat heart model, treatment with a pharmacological inhibitor of PARP-1 afforded improved functional recovery of the reperfused myocardium (98, 567). Administration of a PARP-1 inhibitor in vivo also conferred a cardioprotective benefit in response to I/R in the rat. Inhibitor treatment reduced infarct size and attenuated systolic dysfunction, largely mirroring results obtained using Parp-1 −/− mice (269, 517, 572). Parp-1 −/− mice were also protected from pressure overload-induced cardiac dysfunction and pathological remodeling, although the underlying mechanism was not defined (379). Importantly, myocardial apoptosis (as determined by TUNEL and DNA laddering) was reduced in Parp-1 −/− mice following I/R (573). However, established features of parthanatos were not rigorously examined in these studies, as most of this work preceded the definition of parthanatos as a regulated cell death program. Therefore, whether parthanatos per se occurs in the injured and/or failing heart and whether specific inhibition of this cell death program provides benefit remain unknown. To this point, two studies using Mif −/− mice demonstrated that loss of endogenous MIF paradoxically exacerbated infarct size and further suppressed functional recovery following I/R (324, 387). Moreover, whether parthanatos occurs within, and kills, cardiomyocytes has not been investigated, although forced expression of PARP-1 in neonatal rat cardiomyocytes was shown to reduce cell viability (the type of death and underlying mechanism were not explored) (379). Addressing these points remains critical to our basic understanding of parthanatos in heart disease. Nevertheless, based on these promising preclinical results, a phase I clinical trial (safety and dosing) of the PARP-1 inhibitor INO-1001 was performed in 40 patients with STEMI undergoing percutaneous coronary intervention (PCI) (330). INO-1001 reduced serum PARP-1 activity and inflammatory markers, although also resulted in elevation of some liver transaminases. For unclear reasons, clinical testing never progressed to phase 2.

5. Immunogenic cell death

Immunogenic cell death (ICD) is a regulated process in which dying cells are able to initiate an immune response that kills other cells expressing the same antigens (140, 142, 143). This process is initiated when the dying cells expose or release DAMPs. DAMPs include endogenous cellular proteins, nucleic acids, and small molecules such as the ER protein calreticulin, the chromatin binding protein high-mobility group box 1 (HMGB1), type I IFNs, annexin A1, nucleic acids, and ATP. DAMPs bind to PRRs on or within antigen presenting cells (APCs), such as dendritic cells. The APCs subsequently internalize the dead cell corpses providing antigens that processed and displayed on the surface of APCs as peptides bound to major histocompatibility complex (MHC) molecules. Binding of this complex to an appropriate T cell receptor, along with a costimulus provided by an interaction of B7-1 (CD80) or B7-2 (CD86) on the APC to CD28 on the T cell, stimulates T cell proliferation and differentiation. This process is referred to as T cell activation (253). Activated T cells are then recruited and kill target cells (e.g., cancer cells) expressing the same antigens as those expressed by the dying cells that initiated this process. This specificity is provided by an interaction between peptide MHC on the target cells with the T cell receptor.

Conversely, T cell activation is opposed by a variety of mechanisms (20). These include inhibition mediated by cytotoxic T lymphocyte-associated protein 4 (CTLA-4; CD152) and programmed cell death protein 1 (PD-1; CD279), both proteins resident on the surface of T cells (400). [It should be noted that PD-1, despite its name, is not primarily involved in cell death (4).] While some functional overlap exists, CTLA-4 restrains T cells primarily in secondary lymphoid organs, and PD-1 exerts its effects mainly in peripheral tissues (522). Stimuli that activate T cells induce CTLA-4 trafficking to the cell surface allowing this protein to steal B7-1 and B7-2 away from binding to CD28, thereby depriving T cells of this activating costimulus (280, 281, 491). T cell-activating stimuli also induce the expression of PD-1. Ligands for PD-1 include the cell surface proteins programmed cell death ligand 1 (PD-L1; CD274), which is expressed on APCs and cells of multiple tissues, and PD-L2 (CD273), whose expression is more limited (103, 116, 133, 254, 255). PD-L1 or PD-L2 binding to PD-1 activates Src homology 2-containing phosphotyrosine phosphatase (SHP2), which inhibits signaling from the T cell receptor and CD28 (201, 550). In addition, the expression of PD-L1 and PD-L2 are induced by IFN-γ secreted by activated T cells, which provides a negative feedback loop that restrains T cell activation to protect peripheral tissues from damage (133, 145, 254). In addition to these canonical mechanisms, further complexity continues to emerge in this intensively studied field (400, 522).

Elucidation of the CTLA-4 and PD-1 pathways provided the cornerstone for the development of immune checkpoint inhibitor (ICI) therapies for cancer, arguably the most dramatic advance in cancer treatment in the past decade (400). Although cancer cells use the mechanisms described above to activate immunogenic cell death, these cells often evade immune surveillance through a number of mechanisms (20, 141, 372, 399). For example, they can overexpress PD-L1 to prevent T cell activation. ICIs are monoclonal antibodies that bind CTLA-4 (ipilimumab), PD-1 (nivolumab, pembrolizumab, cemiplimab), or PD-L1 (atezolizumab, avelumab, durvalumab) and, in so doing, block these proteins from engaging in the interactions described above (400). The result is an unleashing of T cell activation that promotes cancer regression leading to an unprecedented extension of life in some of the most refractory and lethal malignancies.

As might be anticipated, ICIs can also induce immune-related damage to multiple tissues (215, 358, 518). In particular, myocarditis has been reported (197, 214, 301, 333, 414). The median time of onset following initiation of therapy has been reported to be 27–30 days, but with variability (333, 414). Clinical manifestations have included heart block, refractory ventricular tachycardia, and heart failure. A retrospective analysis of 101 cases of severe ICI-associated myocarditis indicated an overall mortality of 46%, rising to 67% when two ICIs were employed (333). ICI-associated myocarditis is characterized by lymphocytic and macrophage infiltration involving both the conduction system and cardiac muscle (214). Elevations in serum troponin levels have been noted consistent with cardiomyocyte necrosis (214, 301). Deep sequencing of the T cell receptor of two patients suggested that infiltrates in heart, skeletal muscle, and the cancer showed clonal expansion of similar T cell populations (214). More recently pericarditis and vasculitis have also been associated with ICIs (414).

The occurrence of ICI-associated myocarditis is not unexpected based on studies in mice. For example, germline deletion of Pdcd1, encoding PD-1, in BALB/c mice caused myocarditis and dilated cardiomyopathy (357). Pathogenesis in this case appears to have been driven by an autoantibody against cardiac troponin I (365). Pdcd1 knockout has also been reported to exacerbate myosin peptide-induced myocarditis, and this was mediated by CD4+ T cells (476). In addition, adoptive transfer experiments showed that deletion of Pdcd1 in CD8+ T cells exacerbated myocarditis induced by ovalbumin (476). Second, germline combined knockout of CD274 (encoding PD-L1) and CD273 (encoding PD-L2) also exacerbated the severity of ovalbumin-induced myocarditis (157). This outcome was phenocopied by administration of a blocking antibody against mouse PD-L1 (157). Finally, germline knockout of Ctla4 caused a lethal lymphoproliferative syndrome, which included myocarditis (48, 237, 481, 516).

The frequency of ICI-associated myocarditis is quite low, estimated to be only ~1% (301). While one explanation for this low occurrence rate is likely under-reporting of this new syndrome, another is the basal inhibition of immune mechanisms in the myocardium (158). One inhibitory mechanism appears to involve induction of PD-L1 expression in cardiomyocytes and microvascular endothelial cells in the heart in response to several stress stimuli. Despite the currently reported low frequency of ICI-associated myocarditis, we suspect that this syndrome will become more common as ICIs become even more widely employed and used in combination with other cancer therapies that promote DAMP release from both cancer cells as intended and from cardiac cells as an untoward effect (e.g., doxorubicin and radiation; Ref. 107).

The discussion thus far has focused on how ICIs unleash T cell activation, which can promote cancer regression and induce autoimmune tissue damage. However, we have not yet considered the mechanisms by which cytotoxic lymphocytes kill cells. In cancer contexts, two mechanisms have been demonstrated (304). In the first, the binding of death ligands on the plasma membrane of (FasL) or secreted from (TRAIL, TNF-α) cytotoxic lymphocytes can induce cancer cell death through binding to their cognate death receptors on the cancer cell, as we have previously discussed. The second killing mechanism involves perforin-1 and granzyme B. Cytoplasmic granules containing perforin-1 and granzyme B are released from cytotoxic lymphocytes in a directional manner such that they enter the cleft between the T cell and the cancer cell (immunological synapse) (501). Perforin-1 creates pores in the cancer cell plasma membrane allowing entry of granzyme B. Granzyme B is a serine protease that hydrolyzes peptide linkages following aspartic acid residues in the preferred motif IEXD↓XG. This bears some similarity to preferred motif of caspase-8 [(I/L/V)EXD↓] (177, 480). Consistent with this, granzyme B is able to cleave and activate caspase-8 substrates BID (18, 515) and procaspase-3 (75, 155, 222, 318). While these events are suggestive, the mechanisms by which granzyme B actually kills cells is more complex and differs between humans and rodents (222). Human granzyme B can induce cell death by cleaving BID leading to BAX/BAK-dependent MOMP, cytochrome c release, and procaspase-3 activation (186, 380). Mouse granzyme B, on the other hand, can directly cleave and activate procaspase-3 (222). In addition, granzyme B can also kill cells lacking BID or BAX/BAK, suggesting the existence of other cell death pathways (479). One involves granzyme B-mediated cleavage of mitochondrial complex I components NDUFS1, NDUFS2, and NDUFV1 resulting in a ROS-dependent form of cell death, which could be necrotic as well as apoptotic (59, 210). It is possible that different mechanisms may be activated by granzyme B depending on which death options remain available in the context of the genetic mutations in a given cancer cell.

In contrast to cancer cells, the extent to which ICI-activated T cells kill cardiomyocytes and other heart cells is unclear. While ICIs can result in inflammatory infiltrates involving both cardiomyocytes and the conduction system and while there is evidence of cardiomyocyte necrosis (troponin leak), several key issues remain unclear: 1) To what extent does cardiac cell death contribute to the pathogenesis of ICI-associated myocarditis? 2) Which cell death programs are involved? 3) Are the mechanisms by which T cells kill cardiac cells the same or different from those utilized for cancer cells? As ICIs become even more widely deployed, answers to these questions will be critical.

Finally, it is worth considering whether immunogenic cell death, in the absence of ICIs, may contribute to cardiomyocyte loss during MI or heart failure. This may seem unlikely given the baseline inhibition of immune mechanisms in the myocardium and the fact that this inhibition appears to intensify under stress conditions (158). In addition, involvement of immunogenic cell death may appear “unnecessary” given the plethora of cardiomyocyte-intrinsic death programs described in this review. Nevertheless, further investigation will be needed to test this possibility.

IV. CONCLUSIONS AND THERAPEUTIC CONSIDERATIONS

A. Summary of the Roles of Regulated Cell Death in MI and Heart Failure

The most important conclusion from the above studies is that regulated forms of cell death contribute significantly to the pathogenesis of both MI with reperfusion (I/R) and heart failure (specifically HFrEF). Some caveats need to be emphasized, however. First, while heart samples from patients with these syndromes show various markers confirming the occurrence of regulated cell death programs, the genetic and pharmacological studies establishing cause and effect were performed in animal models. Second, the primacy of cell death in pathogenesis differs greatly between MI and heart failure. During acute MI, cardiomyocyte death is the central event in pathogenesis. Acute loss of these cells in the infarct zone drives subsequent adverse remodeling in the remaining myocardium leading to heart failure. Importantly, the studies presented above show that much of this acute cell death occurs through regulated death programs and, therefore, may be amenable to therapeutic manipulation. In contrast, while levels of cardiomyocyte death are substantially elevated in heart failure, their absolute magnitude is quite low compared with acute MI. Genetic studies in mice, however, again show that this cell death occurs through regulated programs and, in aggregate, contributes to heart failure progression. Nevertheless, cell death is only one component in the pathogenesis of heart failure as dysfunction of viable cardiomyocytes is also important.

What cell death programs are most important in I/R and heart failure? Recent data prompt reconsideration of the relative contributions of apoptosis and necrosis in I/R. Early studies suggested that apoptosis was the primary form of cell death in this syndrome. In some cases, this conclusion was based on genetic manipulations that were thought at the time to impact only apoptotic cell death but were later recognized to also affect necrosis. Raising further questions concerning a role for apoptosis in I/R are data suggesting that caspase-dependent cell death is less robust in adult as compared with fetal/neonatal cardiomyocytes (415, 416, 421). What do we know now? While current data clearly establish a role for non-apoptotic cell death programs in I/R, we remain uncertain whether cardiomyocyte apoptosis plays a role. Given the temporal and spatial complexity of cell death during I/R, we feel that further studies to examine this issue are needed. In the interim, however, we suggest that authors employ the term cardiomyocyte death in the context of I/R rather than the term cardiomyocyte apoptosis.

What non-apoptotic cell death programs mediate cardiomyocyte loss during I/R? The studies presented above demonstrate a role for both mitochondria-mediated necrosis and necroptosis. In addition, emerging data implicate ferroptosis, although it remains unclear whether this program is functioning as an independent cell death modality or an amplification loop for the necrotic cell death. Regardless, the molecules that mediate ferroptosis may provide new therapeutic targets. PARP-1-dependent processes also appear important in I/R. Although we discussed these in the context of the cell death program parthanatos, the significance of PARP-1 activation may be more related to its depletion of NAD⁺ and disruption of cellular energetics. The data regarding autosis in ischemic neuronal injury are intriguing, and encourage further study of this form of autophagy-dependent cell death in myocardial I/R. Pyroptosis may also contribute to I/R, but the major challenge is in understanding the cell types in which it is taking place. Current data suggest that pyroptosis operates primarily in non-cardiomyocytes, bone marrow-derived cells recruited to the heart during MI and perhaps cardiac fibroblasts, but further study is needed.

What cell death programs contribute to the pathogenesis of heart failure? Both apoptotic and necrotic cardiomyocyte death have been implicated, but their relative contributions remain unclear. Very limited information is available regarding the roles of other cell death programs in heart failure.

B. Therapeutic Considerations

The first issue is which cardiac syndromes provide the optimal targets for anti-cell death therapies. We would argue that those requiring short-term inhibition of cell death may be the most logical starting points because they limit the possibility of undesirable outcomes including the initiation/progression of cancer and interference with antiviral defenses. The most common and important acute indication is I/R in the setting of MI or possibly cardiac surgery including transplantation. A second important short-term indication may be in the prevention of cardiotoxicities related to cancer treatments. These often involve a cell death component and can result from conventional chemotherapy, targeted therapies, and radiation (331, 332). An additional advantage of targeting cell death in the contexts of cardiac surgery and cancer treatments is that, unlike MI, these are scheduled events allowing a therapy to be initiated at or before death stimuli. What about targeting cell death in nonacute conditions such as heart failure? Given the importance of this syndrome and its residual mortality despite current treatments, heart failure may well be a future candidate for therapies that inhibit cardiomyocyte death. While the necessity for ongoing cell death inhibition raises the concerns noted above, the lethality of heart failure may justify such an approach. Risk-benefit analyses are an important consideration for all medical treatments.

Although an effective therapy for I/R would be impactful, this has traditionally been a very difficult problem. How can cell death be inhibited in I/R? Based on the data presented, targeting necrosis in the death receptor and mitochondrial pathways would likely be the highest priorities, although inhibition of ferroptosis, PARP-1 activation, and autosis might also be beneficial. The most promising approaches in the death receptor pathway involve inhibition of RIPK1 and RIPK3, and the latter may have the advantage of possessing a wider spectrum of relevant downstream targets (351, 563). Several possible approaches exist to inhibiting necrosis in the mitochondrial pathway. The most obvious is to inhibit mPTP opening. As discussed, CsA showed initial promise in a small clinical trial, but later failed in further testing (69, 381). While there may be multiple explanations for this result (278), one possibility is that CsA is not the optimal drug to antagonize cyclophilin D and mPTP opening. Another possibility is that additional positive regulators of mPTP may also require inactivation. An understanding of the proteins that constitute mPTP could provide insights into this question as well as reveal novel drug targets. Another potential target in the mitochondrial pathway is BAX. In fact, highly selective small molecule inhibitors of BAX have recently been reported (146), which have the theoretical advantage of inhibiting both necrotic and apoptotic cell death.

GRANTS

This work was supported by National Institutes of Health Grants R01HL127339 and R15HL135726 (to D. P. Del Re), R01HL116507 (to Q. Liu), as well as R01HL128071, R01 HL130861, and R01HL138475 (to R. N. Kitsis); Department of Defense Grant PR151134P1 (to R. N. Kitsis); American Heart Association Grants 15PRE25080032 (to D. Amgalan) as well as 15CSA26240000 and 18SRG34280018 (to R. N. Kitsis); Fondation Leducq (to R. N. Kitsis); Deutsche Forschungsgemeinschaft Grant EXC 306 (to A. Linkermann); and research grants from the Else Kröner-Fresenius Stiftung (to A. Linkermann).

DISCLOSURES

A. Linkermann has received research grants from Pfizer, Novartis, Fresenius Medical Care, and the Else Kröner-Fresenius Stiftung; has material transfer agreements with Genentech, Glaxo Smith Kline, and Apogenix; and has received honoraria and/or travel grants from Astellas, Otsuka, Genentech, Alexion, and Tekmira. R.N. Kitsis is co-founder of Aspida Therapeutics Inc. and a consultant for Amaron Bio. No conflicts of interest, financial or otherwise, are declared by the other authors.