Alternative cell death mechanisms in development and beyond

Suppressing apoptotic and nonapoptotic cell death during development

Interdigital cell death during embryonic development has been used as a quintessential example to underscore the cardinal importance of apoptotic cell death as a sculpting force during morphogenesis. Surprisingly, however, none of the apoptosis-deficient mutant mice manifested severe syndactyly, the fusion among fingers or toes that results from the persistence of interdigital membranes. Among different genetic models of apoptosis deficiency, bax^−/−; bak^−/− double-knockout mice provide the best example of syndactyly (Lindsten and Thompson 2006). However, compared with that of mutants in the Wnt signaling pathway (MacDonald et al. 2004; Johnson et al. 2005; Morello et al. 2008), the syndactyly phenotype exhibited by bax^−/−; bak^−/− mice can only be considered as a partial defect. Although it is possible that the defects in morphogenesis (not just cell death) also contribute to the severe phenotype of syndactyly that is observed in certain mutants of the Wnt pathway, we should consider the possibility that the severe syndactyly phenotype involves the suppression of both apoptosis and backup cell death mechanisms. Consistent with this possibility, a detailed analysis of the interdigital cell death in apaf-1 mutant mice, which lack the key adaptor protein for mediating caspase-9 activation, led to the discovery that some cell deaths exhibit features of necrosis (Chautan et al. 1999; Cande et al. 2002). This result suggests the activation of alternative, nonapoptotic cell death mechanisms when apaf-1-dependent caspase activation is blocked. Indeed, interdigital cell death is only slightly delayed in apaf-1-deficient mice, suggesting that nonapoptotic mechanisms come into action as a backup mechanism when apoptosis is inhibited.

On the other hand, constitutive activation of the FGF (fibroblast growth factor) pathway due to loss-of-function mutations of its negative regulator (BMP [bone morphogenetic protein]) (Guha et al. 2002) or gain-of-function mutations of FGF receptors promotes syndactyly in humans (Macias et al. 1996). One possible signaling pathway by which these mutations cause syndactyly may be through suppression of the expression of Dickkopf1 (Dkk1), an inhibitor of the Wnt pathway that is highly expressed in regressing interdigital tissues. Mice deficient in Dkk1 exhibit severe syndactyly (MacDonald et al. 2004). The role of Wnt in regulating interdigital cell death is further strengthened by the syndactyly phenotype of loss-of-function mutants of other WNT inhibitors: Sfrp2 (secreted frizzled-related protein-2) (Morello et al. 2008) and Megf7/LRP4, a member of the low-density lipoprotein (LDL) receptor-related proteins (LRPs) that antagonize LRP6-mediated activation of canonical Wnt signaling (Johnson et al. 2005). Since the activation of Wnt is sufficient to induce syndactyly, Wnt signaling is probably involved in regulating both apoptotic and nonapoptotic cell death mechanisms in the interdigital tissue. This also raises the possibility for a role of nonapoptotic cell death mechanisms in tumorigenesis, because constitutive activation of the Wnt pathway, often due to loss-of-function mutations in its negative regulator, APC (adenomatous polyposis coli), is common in familial and sporadic colorectal carcinomas (Gregorieff and Clevers 2005).

An intriguing—yet unproven—working hypothesis would predict that prolonged clonogenic survival, in either interdigital tissue or cancer cells, would rely on the simultaneous blockade of both apoptotic and nonapoptotic cell death mechanisms.

Mechanisms of programmed cell death in the nematode C. elegans

The seminal work by Horvitz and colleagues (Yuan and Horvitz 2004) in the late 1980s and early 1990s provided first insights into the molecular mechanism that mediates developmental cell death in C. elegans. The gene products of ced-3 (a caspase), ced-4 (a caspase activator homologous to Apaf-1), ced-9 (a multidomain protein of the Bcl-2 family), and egl-1 (a proapoptotic BH3-only protein) constitute the prototypic apoptotic machinery that regulates the programmed death of 131 cells during development. Curiously, the development and adult lives of ced-3 and ced-4 mutant animals, which carry a significant number of extra cells that normally would have died in wild-type animals, are largely normal (Ellis and Horvitz 1991). This could be explained in a number of ways. First of all, it is possible that, in a primitive organism such as C. elegans, this cell death mechanism does not serve a major function from an organismal point of view—at least in the rather protective laboratory environment. It has been proposed that ced-3 and ced-4 are indispensable for the host defense against pathogenic bacteria (Aballay and Ausubel 2001); thus, the advantage of carrying intact ced-3 and ced-4 genes might become manifest only when the nematodes are kept in their natural habitat: soil. Alternatively, it is conceivable, yet remains to be investigated in detail, that at least a portion of “undead” cells may eventually be eliminated from ced-3 and ced-4 mutant nematodes through alternative nonapoptotic mechanisms.

There is compelling evidence for the existence of nonapoptotic developmental cell death in C. elegans. One clear example is the death of linker cells in male C. elegans that occurs normally in ced-3 or ced-4 mutant males (Abraham et al. 2007). The disappearance of linker cells facilitates the connection of the male reproductive system to the exterior. Although the linker cell death was initially believed to be a “murder” by engulfment (Sulston et al. 1980), subsequent detailed analysis demonstrated that engulfment is not required for linker cell death. The death of linker cells is controlled by developmental timing because, in lin-29 mutants, which exhibit retarded development, linker cells persist beyond the point at which they would have died in wild-type animals. Dying linker cells do not manifest any signs of nuclear condensation, the morphological hallmark of apoptosis, yet extensively accumulate cytoplasmic vesicles with single membranes. These cytoplasmic vesicles are believed to be distinct from autophagosomes, as knockout of major autophagy genes—including bec-1 and unc-51, the C. elegans orthologs of Beclin 1 and Atg1/Unc1, respectively—have no effect on linker cell death (Abraham et al. 2007). While the molecular mechanism that mediates the death of linker cells is still not clear, the morphology of dying linker cells shares some resemblance with nonapoptotic cell deaths in vertebrates. One such example is the “dark” cell death of spinal motor neurons in developing caspase-3^−/− or caspase-9^−/− mouse embryos that is characterized by mitochondrial swelling and electron-translucent cytoplasmic vacuoles (Oppenheim et al. 2001) similar to that of dying linker cells. While it is clear that the nonapoptotic cell death modality that operates in male linker cells is not a “backup” mechanism that becomes manifest only when apoptosis is suppressed, it remains to be seen whether the “dark” cell death may be activated in the absence of the normal programmed cell death machinery, such as in ced-3 and ced-4 mutants.

In C. elegans expressing a mutant MEC-4 protein (which is an amiloride-sensitive Na⁺ channel from the degenerin family), mechanosensory neurons undergo necrosis (Driscoll and Chalfie 1991). The dominant gain-of–function mutant MEC-4 protein mediates an increased ion flow across the plasma membrane, eventually leading to the activation of aspartyl and calpain proteases and lysosome-mediated necrosis (Syntichaki et al. 2002). Importantly, knockdown of several autophagy genes coding for the nematode orthologs of Beclin 1, Atg8/LC3, and Atg18 prevented necrosis induced by the mutant MEC-4 protein (Samara et al. 2008), underscoring the importance of the lysosomal cell death pathway for necrotic cell death. The integrity of lysosomes may be regulated by serpins in C. elegans, as srp-6 mutant animals show enhanced sensitivity toward cellular stresses such as hypomostic shock, heat shock, oxidative stress, and hypoxia (Luke et al. 2007). However, it is not clear to what extent this lysosomal pathway is regulatable and whether it contributes to cell death during C. elegans development and tissue homeostasis.

‘Apoptosis’ without the canonical regulators of apoptosis

Caspases are phylogenetically restricted to animals only. Although the genomes of plants and single-cell eukaryotes do not encode canonical caspases, they nevertheless undergo regulated cell death with morphological features similar to that of apoptosis. For example, yeast cells can undergo cell death associated with chromatin condensation, DNA fragmentation, and phosphatidylserine exposure; when yeast is made to express a CDC48 mutant (Madeo et al. 1997); when yeast is treated with external stressors, including hydrogen peroxide or acetic acid; or when yeast undergoes chronological aging (Madeo et al. 2002). This has given rise to the notion that yeast can also undergo apoptosis, which is at odds with the idea that apoptosis must be mediated by canonical regulators including proapoptotic multidomain Bcl-2 family proteins and caspases. Many investigators in the cell death field argue that, since yeast does not encode any canonical caspase or (known) Bax/Bak analogs, yeast cell death should not be termed “apoptosis.” However, it is intriguing that dying yeast cells can acquire the morphological features of apoptosis (Carmona-Gutierrez et al. 2010), suggesting that several hallmarks of apoptosis—including nuclear pyknosis, chromatin condensation, and phosphatidylserine exposure—are not truly specific biomarkers for caspase activation and thus must be interpreted with caution. For example, cell death with morphological features of apoptosis during the development of vertebrates has traditionally been regarded as evidence of the activation of one uniform, caspase-driven biochemical cell death pathway. However, we must consider the possibility that at least some of those cell deaths with morphological features of apoptosis that occur during the development of vertebrates arise without the obligatory contribution of the canonical apoptosis pathway.

Nonetheless, yeast possess some elements that resemble those involved in the canonical pathway of animal cell apoptosis. Mitochondria are involved in mediating yeast cell death (Cheng et al. 2008), and the yeast genomes do encode homologs of mammalian cell death effectors, including a metacaspase termed YCA1 (Madeo et al. 2002), apoptosis-inducing factor (Aifp) (Wissing et al. 2004), and HtrA2/Omi (Nma111p) (Fahrenkrog et al. 2004). Since yeast Aifp is a mitochondrial DNase that translocates to the nucleus upon lethal stress with H₂O₂ or acetic acid (Wissing et al. 2004), while Nma111p is a serine protease that is normally localized in the nucleus (Fahrenkrog et al. 2004), activation of Aifp and Nma111p might contribute to the acquisition of the apoptotic phenotype. Moreover, YCA1 overexpression can induce yeast cell death with morphological features of apoptosis, and its deficiency reduces cell death induced by oxidative stress, acetate, or chronological aging (Madeo et al. 2002). Intriguingly, plant metacaspases, which are distant phylogenetic relatives of animal caspases, as is YCA1, can cleave some nuclear substrates that are also cleaved by animal caspases (Sundstrom et al. 2009). Thus, although metacaspases have a different cleavage specificity than caspases—they hydrolyze proteins after arginine or lysine (basic residues), not after aspartate (an acidic residue)—they can share substrates over large evolutionary distances. Nonetheless, alternative interpretations are possible, and YCA1 might regulate autophagy in an indirect fashion. For instance, YCA1 has been shown to be a stress response protein associated with chaperones. Loss of YCA1 in Δyca1 cells induces an increase in stress response protein levels and the formation of autophagosomes (Lee et al. 2010). It remains to be seen if the increased resistance of Δyca1 yeast cells to H₂O₂ or acetic acid is due exclusively to the lack of proteolytic execution, or whether it might involve a constitutively activated stress response that mediates an elevated level of cytoprotection.

The fact that yeast apparently lacks a canonical apoptosis pathway, but nonetheless can acquire an “apoptotic” morphology in response to lethal stimuli, points to the possibility that mammalian cells may similarly succumb to an “apoptotic” cell death without the canonical apoptosis effectors coming into action. For example, the caspase-independent, phylogenetically conserved death effector termed AIF has been shown to mediate chromatin condensation and induce phosphatidylserine exposure as a backup mechanism in both human cells (Susin et al. 2000) and C. elegans (Wang et al. 2007) when caspase activation is inhibited. In some paradigms of yeast cell death (Wissing et al. 2004) as well as in mammalian neurons (Cregan et al. 2004), AIF is important for cell death induction, suggesting that caspase-independent apoptosis may be mediated, at least in some instances, by AIF. However, systematic studies are needed to explore the question regarding in which cases cells dying with an apoptotic morphology succumb to caspase-independent programs.

Autophagic cell death—elimination of cells in toto

Morphological studies of developing embryos have led to the detection of certain dying cells with extensive accumulation of vesicles, a phenomenon that has been called “autophagic cell death” (Schweichel and Merker 1973). However, the activation of autophagy serves in most conditions to promote survival (Levine and Yuan 2005); the existence of autophagic cell death and the exact contribution of the canonic autophagic pathway to “autophagic cell death” are not yet established.

Insect metamorphosis provides a convenient and highly synchronized model of developmental cell death, as larval organs are destroyed in toto during a short period of time. Metamorphic cell death of the salivary glands from developing Drosophila melanogaster larvae is initiated by a rise in ecdysone, which promotes the up-regulation of several caspases (dronc, drice, and strica) and ark, along with several Atg genes (Berry and Baehrecke 2007). However, inhibition of caspases is not sufficient to block the degeneration of the salivary gland, suggesting the involvement of additional mechanisms. When combined with caspase inhibition, mutations in Atg5 or loss-of-function mutations in Atg2, Atg3, or Atg18 lead to the persistence of salivary glands for ∼24 h beyond the time when they normally would have died in control larvae. Thus, autophagy, or at least certain autophagic genes, are involved in the death of salivary gland cells during metamorphosis. However, it is also possible that only a selected few among all of the genes involved in mediating canonical autophagy are involved in mediating “autophagic cell death.” This could explain why deficiencies in some autophagy genes (e.g., atg1, atg2, and atg18) have stronger effects on metamorphic cell death than those in others (e.g., atg7).

Interestingly, drpr, the fly homolog of the C. elegans engulfment gene ced-1, as well as several factors involved in the engulfment of dying cells from programmed cell death—such as six microns under (simu), croquemort (crq), ced-6, src oncogene at 42 A (Src42A), ced-12, crk, and myoblast city (mbc)—were found recently to be involved in the activation of autophagy during metamorphosis of salivary gland. Furthermore, Drpr was found to be required for induction of both autophagy and cell death in metamorphic salivary glands (McPhee et al. 2010). Thus, the mechanism of engulfment may provide a direct link to connect cell death to autophagy.

Although the published data (McPhee et al. 2010) is consistent with a cell-autonomous mode of action for Drpr to mediate autophagy activation during degeneration of salivary gland, an alternative, yet-to-be-excluded possibility is that Drpr may activate autophagy in cells other than the dying ones, and that this distant autophagy may be required for the coordination of the cell death process that culminates in the attrition of an entire tissue. In this regard, it might be interesting to examine if dying cells in metamorphic salivary glands also express Pretaporter, a ligand that binds to the extracellular region of Drpr and participates in the clearance of apoptotic cells (Kuraishi et al. 2009), and whether Pretaporter exposure is regulated by atg genes.

During the developmental programmed cell death in C. elegans, two pathways control the engulfment of dying cell corpses on the side of engulfing cells. The first pathway—which includes ced-1, ced-6 (GULP), ced-7 (ABC transporter), and dyn-1 (Dynamin, a large GTPase)—functions to recognize dying cells and transduce engulfment signals. The second pathway—including ced-2 (CrkII), ced-5 (Dock 180), ced-10 (Rac guanosine triphosphatase), ced-12 (ELMO), and psr-1 (phosphatidylserine receptor)—activates cytoskeleton rearrangement in the engulfing cells (Reddien and Horvitz 2004). During phagocytosis, the receptor CED-1 clusters in the phagocytic cup before quickly encircling the cell corpse. Mutations in engulfment genes, such as ced-1, can enhance the persistence of undead cells caused by partial loss-of-function mutations in killer genes, such as ced-3 (Reddien et al. 2001). Thus, the engulfment process may have a role in ensuring the completion of the cell death process. On the other hand, degradation of apoptotic cells in the engulfing cells involves the sequential enrichment of CED-1, DYN-1, phosphatidylinositol 3-phosphate (PI3P), and the small GTPase RAB-7 on the phagocytic membranes of the engulfing cells, followed by the fusion of endosomes and lysosomes to phagosomes (Yu et al. 2008). Thus, an alternative possibility is that Drpr performs a similar function in dying cells as that in the engulfing cells by inducing the formation of (auto)phagosomes during the metamorphic death of salivary gland cells .

Degeneration of midgut during metamorphosis of Drosophila provides another example of autophagic cell death. Like in salivary glands, programmed cell deaths triggered by steroids lead to the destruction of larval midgut structures. Despite high levels of caspase activity in metamorphic midgut in wild-type Drosophila, caspases are not required for midgut degradation. Dying midgut cells manifest the accumulation of pGFP-Atg8a in cytoplasmic puncta, which is a sign of autophagy. Furthermore, knockdown of several Atg genes that are transcriptionally up-regulated at the time of midgut cell death severely delays midgut degradation (Denton et al. 2009). These data indicate again an important role for Atg genes in midgut destruction, yet do not clarify whether these genes act in a cell-autonomous fashion.

Autophagic cell death has also been noted in Dictyostelium discoideum, a simple eukaryote that lives in the soil and feeds on bacteria (Giusti et al. 2009b). In the vegetative state, when food is abundant, Dictyostelium cells divide by binary fission; however, when food is exhausted, starvation triggers the formation of multicellular aggregates in which coordinated morphogenesis and cellular differentiation facilitates the formation of a fungus-like structure that carries spore-forming cells on the top of a stalk composed of cells that have undergone programmed death and lignification. The Dictyostelium genome does not encode caspases or Bcl-2 family proteins, yet does contain a set of atg genes similar to those found in yeast and mammals. One particular Dictyostelium cell line can be driven into programmed death by starvation and simultaneous addition of the differentiation factor DIF. This programmed death is autophagic in that it is accompanied by the accumulation of vacuoles, which are suppressed by mutations in atg1. However, atg1 mutant Dictyostelium cells still die in response to starvation plus DIF, while manifesting a necrotic phenotype characterized by mitochondrial uncoupling, the clustering of organelles, lysosomal permeabilization, and eventual disruption of the cells into debris (Giusti et al. 2009a). Although Atg1 is clearly required for the vacuolization of Dictyostelium, the manner in which Atg1 is involved in cell death is still debatable. It is possible that Atg1 is indeed required for cell death of Dictyostelium induced under starvation condition with the presence of DIF. In the absence of Atg1, however, a “backup” cell death mechanism would be activated to mediate cell death. Alternatively, Atg1 may be required only for vacuolization, which would constitute an epiphenomen that accompanies cell death but is not required for the lethal process. In the Atg1 mutant, the cell death exhibits an alternative morphology, but the fundamental mechanism that mediates cell death may not have changed (Kosta et al. 2004).

Transgene-enforced overexpression of Atg1 in Dictyostelium is sufficient to induce ectopic autophagy and cell death in several Drosophila tissues, including the imaginal discs, fat body, and salivary glands (Scott et al. 2007). Cell death induced by overexpression of Atg1 shows typical signs of apoptosis, including caspase activation and DNA fragmentation, and is inhibited by the caspase inhibitor p35. Thus, in contrast to metamorphic cell death of the salivary gland, where caspase activation occurs upstream of or in parallel with autophagy activation, high levels of autophagy induced by Atg1 overexpression may lead to cell destruction in a caspase-dependent manner. Since Atg1 clearly has functions other than mediating autophagy—e.g., regulating axonal transport (Toda et al. 2008)—it remains to be seen whether Atg1-induced caspase activation is truly due to excessive autophagy.

Necroptosis—receptor-interacting protein 1 (RIP1) kinase activity-dependent necrotic cell death

Necrosis had traditionally been considered an accidental cell death that is not subject to cellular regulations. Necrotic cell death is morphologically characterized by an increase in cell volume, swelling of organelles, and plasma membrane rupture, followed by leakage of intracellular contents. However, it has now been established that at least a part of necrotic cell death may be executed through a mechanism termed “necroptosis” (Degterev et al. 2005). Recent advances have begun to elucidate the molecular mechanisms that ignite and execute this necrotic cell death program. Necroptosis may be activated upon stimulation by tumor necrosis factor α (TNFα), FasL, and TRAIL, the very same ligands that can activate apoptosis in the death receptor pathway. Thus, cell death induced by the activation of the death receptor may be executed through alternative cell death pathways, apoptosis, or necroptosis.

Shedding—physical elimination of cells

The intestinal epithelium is a self-renewing monolayer of epithelial cells arising from stem cells localized at the base of the crypts. In the mouse, it takes 2–3 d for the epithelial cells to migrate from the crypts to the tips of villi, where they are shed at a rate of 1400 cells per villus per day. Thus, the intestinal epithelium has one of the highest turnover rates in the body (Watson and Pritchard 2000). Shedding of cells leads to their separation from the extracellular matrix and hence is expected to lead to anoikis, a specific form of apoptosis that is induced by detachment. However, apoptosis-deficient (e.g., Casp-3^−/−, Apaf-1^−/−, and Bax^−/−; Bak^−/−) mice fail to exhibit a defect in the intestinal epithelium. Although Bcl-2^−/− mice might have supernumerary apoptotic cells in the intestine, no defect in the villus dimension of mice overexpressing Bcl-2 in the intestine, as well as in Bax^−/−; Bak^−/− mice, has been reported, suggesting that the turnover of intestinal epithelial cells may involve a nonapoptotic cell death mechanism.

Electron microscopic studies revealed a fascinating diversity of cell death morphologies during the shedding of mammalian small intestine cells (Fig. 2; Mayhew et al. 1999). A large percentage of cell deaths occurs upon the formation of a tight junction between the cell that is going to be extruded and neighboring epithelial cells, which may “push” the cell that is destined to die toward the lumen. In this case, live cells are extruded into the lumen and the shed cells acquire an apoptotic morphology only when they have lost contact with the epithelium. However, some shed cells with an anuclear phenotype may die before being shed. Alternatively, certain epithelial cells die in situ and are engulfed by macrophages. Finally, certain epithelial cells die with a morphological resemblance to necrosis, including cell swelling, a decrease in cell electron density, and complete degradation of organelles and membranes. The mechanism that regulates necrosis in the mammalian intestine is unknown. Nonetheless, these morphological studies suggest that multiple mechanisms may be involved in the turnover of the small intestine, consistent with the normal intestinal phenotype of apoptosis-deficient mice.

LKB1, a serine–threonine kinase, is a tumor suppressor that is inactivated in Peutz-Jeghers syndrome, an inherited condition characterized by the development of multiple benign hamartomatous polyps in the gastrointestinal tract. A reduction of apoptosis (determined by the TUNEL method) was found in small intestines of patients with Peutz-Jeghers syndrome, and the expression of LKB1 can stimulate p53-dependent apoptosis (Karuman et al. 2001). Thus, the loss of LKB1 may be sufficient to reduce the rate of intestinal epithelial turnover. How this is achieved is not entirely clear. LKB1 catalyzes the activating phosphorylation of kinases of the AMPK family, which regulates cell cycle and cell growth. Moreover, LKB1 can activate SIK1 (salt-inducible kinase 1), a key modulator of anoikis (Cheng et al. 2009). However, since the turnover of intestine epithelial cells may involve multiple cell death mechanisms beyond anoikis, it appears plausible that LKB1 must affect other mechanisms of cell loss in order to promote the formation of hamartomatous polyps.

Cornification—the specialized death of keratinocytes in skin

Our skin is covered by dead cells, so-called corneocytes, which perform an essential function as a physical and chemical barrier. Keratinocytes, the major cell type in the epidermis, undergo a specialized form of programmed cell death, termed cornification, that is different from canonical apoptosis. During cornification, the keratinocytes lose their organelles, including the nucleus, to become dead corneocytes. In the final stage, corneocytes are shed from the skin by a process termed desquamation (Lippens et al. 2009). Cornification does not require the canonical apoptosis mediators, as none of the apoptosis-deficient mouse models manifests a defect in cornification. Skin development does require caspase-14, a nonapoptotic caspase expressed exclusively in the cornifying epithelial cells and activated during cornification (Denecker et al. 2007). Caspase-14^−/− mice show defects in stratum corneum formation and reduced skin barrier. In caspase-14^−/− mice, profilaggrin, a caspase-14 substrate, is not properly processed into mature filaggrin, which is required for the proper moisturizing and formation of skin barrier (Denecker et al. 2007). However, it is clear that the cornification and desquamation process is largely normal in caspase-14^−/− mice; thus, the mechanism that leads to the death of keratinocytes during cornification that removes all of the organelles is still not clear.

Entosis—cell eating cell

Apoptotic cells are commonly engulfed by professional phagocytes (such as macrophages) or nonprofessional neighboring cells. However, live cells may also be engulfed by another cell, a process termed entosis, providing an interesting case of “cannibalism” in cell biology (Fig. 3; Overholtzer et al. 2007). Entosis was first discovered in human tumors, and may be activated when cells lose their attachment to the extracellular matrix. It is still unclear what might be the “eat-me” signal for entosis, as the internalized live cells do not expose phosphatidylserine residues on the outer leaflet of their plasma membranes, and hence lack (one of) the most important engulfment signals of apoptotic ells. The entotic cell–cell interaction requires cadherin to establish cell contact. The entotic internalization process then requires actin polymerization and myosin II, Rho, and Rho kinase ROCK activity in the internalized cells. Once internalized, the engulfing cell appears to use the lysosome to mediate the degradation of the intruded cell without the involvement of caspases (Overholtzer et al. 2007). Although entosis has been described to occur only in tumors, it is conceivable that this mechanism may also contribute to cell death during development and tissue homeostasis.

Physiological significance of alternative cell death mechanisms

Although apoptosis has been initially viewed as the primary regulated mechanism of cell death in higher animals, the importance of alternative cell death mechanisms is underscored by the largely normal development of mutant mice with mutations in the canonical apoptosis mediators. Morphological observations suggest that additional nonapoptotic cell death pathways may operate in the normal development of mammalian embryos (Schweichel and Merker 1973). Although nonapoptotic mechanisms are most clearly illustrated under apoptotic-deficient conditions, it is also possible that nonapoptotic cell death mechanisms might function at the front line as the primary cell death mechanism in a cell type-dependent and stimulus-dependent manner. For example, although necroptosis may be used largely as a backup mechanism under physiological conditions when stimulation of death receptors fails to activate apoptosis, necroptosis may also operate as one of the primary mechanisms that mediates pathological cell death. Moreover, the mechanisms that mediate the turnover of epithelial cells, such as intestinal epithelial cells and keratinocytes, appear primarily nonapoptotic. Finally, since yeast cells, which lack a canonical apoptosis pathway, can acquire a typical apoptotic morphology, we must consider the possibility that some of the “apoptotic” cell death that occurs during development may be executed through noncanonical pathways.

How may cells choose to die through apoptotic versus nonapoptotic pathways? Apoptotic cell death is highly regulatable, and thus may operate in individual cells to lead to selective elimination. In contrast, autophagic cell death such as that which occurs during metamorphosis may function to eliminate a tissue in toto. For epithelial cells that are exteriorly exposed, physical removal of dying or dead cells through shedding may be the most efficient mechanism of elimination.

Apoptosis is executed through a highly stereotyped series of biochemical events that guarantee a swift and noninflammatory removal of corpses. Although the translocation of phosphatidylserine to the outer leaflet has been known to be a caspase-dependent event in apoptosis, the exposure of phosphatidylserine can also occur in nonapoptotic cell death (Hirt et al. 2000) as a result of cathepsin B activation (Foghsgaard et al. 2001), perturbed Ca²⁺ homeostasis, or activation of PKC. Thus, the same mechanism that is employed to engulf apoptotic cells may also be used to clean up nonapoptotic cell death.

While apoptosis has been studied extensively for more than two decades, nonapoptotic cell death mechanisms are only beginning to be elucidated. In spite of the initial consensus that apoptosis would be the mechanism of programmed cell death, it turned out that apoptosis-deficient mice have largely normal phenotypes, indicating that nonapoptotic cell death mechanisms must play a major role in development and tissue homeostasis. Given that the vast plant kingdom and primitive eukaryotes such as yeast and Dictyostelium lack the major player(s) of the canonical, caspase-driven apoptotic pathways, yet effectively control and execute their own death, nonapoptotic cell death pathways may be evolutionarily more ancient than that of the apoptotic pathway. The mechanistic elucidation of these alternative cell death mechanisms is likely to lead to exciting new insights into the regulation of programmed and accidental cell death, with major implications for the fields of biology and medicine.