Active participation of cell death in development and organismal homeostasis


Author to whom all correspondence should be addressed.


The precise, regional execution of programmed cell death is required for the proper patterning and sculpting of the embryonic primordium during animal development. In addition, cell death that is not directly involved in sculpting is also widely observed. The most abundant morphological form of programmed cell death in developing animals is apoptosis, and identification of the apoptotic genetic pathways has enabled the study of apoptosis’ regulation and roles during development. Genetic and bio-imaging studies have permitted the study of the active roles of cell death in development and organismal homeostasis.


In the first decades of the 1900s, few embryologists recognized cell death as an integral part of animal development. This was probably in part because the idea of cell loss participating in embryonic development seemed to contradict the idea that development is a process in which cells proliferate, grow, and differentiate. In addition, cell death can be difficult to detect, because dying cells are rapidly removed from tissues by engulfment, shedding, or the dispersion of cell debris. However, in some developing tissues, massive cell death occurs at a specific stage and region (for example, interdigital cell death in the developing vertebrate limb, changes in the Müllerian and Wolffian ducts, and the metamorphosis of insects and amphibians). These dramatic cases sparked the first interest of researchers in naturally occurring cell death. In their paper on the naturally occurring cell death of the intersegmental muscles of silk moths, R. Lockshin and C. Williams referred to such cell death as, “Programmed cell death” (Lockshin & Williams 1964, 1965). In their paper, they stated that “as pointed out previously (Williams 1961), there can be little doubt that the death of specific cells and tissues is a part of the ‘construction manual’ for the insect as a whole. The cells that will die have been programmed to do so. Therefore, their individual deaths represent the decoding and acting-out of a fresh, albeit final, bit of genetic information”. Thus, for the first time they used “programmed cell death” for the title of a scientific manuscript.

Despite their observation, however, most cell deaths occur sporadically and asynchronously, and did not capture the interest of many researchers. Even when a few pyknotic cells appeared in actively proliferating regions, embryologists referred to such condensed small particles as, “mitotic metabolites”. Thus, researchers did not pay much attention to cell death and largely ignored its possible active roles in development until recently.

In the 1920s and 1930s, Kallius and his students Ernst and Glücksmann recognized cell death as one of the mechanisms by which cells come to form different tissues and organs (Hamburger 1992). Glücksmann (1950) extensively examined the occurrence of cell death during animal development, and classified it according to its developmental functions:

  • 1 Morphogenetic degeneration (e.g., neural tube closure).
  • 2 Histogenic degeneration (e.g., Müllerian duct, Wolffian duct, ossification, metamorphosis).
  • 3 Phylogenetic degeneration (e.g., notochord, pronephros, mesonephros).

Programmed cell death is now recognized as naturally occurring cell death that is regulated by an intracellular program. It is distinguished from necrosis, which is caused by injury, provokes an inflammatory response, and is usually detrimental to cells and tissues. The most prevalent form of programmed cell death is apoptosis, which is mediated intracellularly by a cascade of cysteine proteases called caspases (Degterev et al. 2003). Apoptosis is accompanied by characteristic cell changes, which include membrane blebbing, cell shrinkage, and chromosomal DNA fragmentation. This fragmentation is produced by CAD (caspase-activated DNAse), and is detected by TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling).

Remarkably, Glücksmann (1950) mentioned in his classic paper that the cell death in histogenic degeneration played a role in the initiation of histogenic processes by releasing certain enzymes and substrates. It was considered that the dying cells were removed by phagocytes without releasing their cell contents or eliciting an inflammatory response; however, now evidence suggests that dying cells indeed release some signaling molecules to neighboring cells to let them know their situation. In adults, apoptotic cells release molecules to induce an anti-inflammatory microenvironment and immunological tolerance, and non-apoptotic dying cells release alert or danger signals to activate inflammation (Zitvogel et al. 2010). Furthermore, in developing or regenerating tissues or wound healing, growth factors are released from dying cells that promote the proliferation of their neighboring cells (Huh et al. 2004; Perez-Garijo et al. 2004, 2005, 2009; Ryoo et al. 2004; Kondo et al. 2006; Wells et al. 2006; Fan & Bergmann 2008; Chera et al. 2009; Lee et al. 2009; Li et al. 2010).

The secretion of morphogens, a group of extracellular signaling molecules, is one of the most important regulatory mechanisms of animal development. A morphogen is secreted from a localized source and forms a concentration gradient. The responses of tissues and cells vary with the morphogen concentration. Signaling centers such as the ZPA (zone of polarizing activity) (Sanz-Ezquerro & Tickle 2000), floor plate (Homma et al. 1994), and notochord (Offner et al. 2005) secrete the morphogen Sonic Hedgehog, and all of these areas also undergo massive cell death. It is speculated that the apoptosis at signaling centers is involved in the secretion of morphogens. Another function of apoptosis in this region would be as a buffering system that regulates the number of morphogen-expressing cells (Sanz-Ezquerro & Tickle 2000). Similar function of apoptosis that regulates the size of organizing centers was studied in Xenopus neuroectoderm (Offner et al. 2005).

Among other morphogenic roles, apoptosis is required for Xenopus tadpole tail regeneration (Tseng et al. 2007). In epithelial cells (Madin-Darby canine kidney [MDCK] cells), apoptosis affects the reorganization of neighboring cells’ actin cytoskeleton (Rosenblatt et al. 2001) or actively generates a force that promotes morphogenesis in Drosophila embryo (Toyama et al. 2008). The apoptosis of larval epidermal cells is required for the proliferation of histoblasts in Drosophila pupae (Ninov et al. 2007). Thus, developmental cell death is not simply a process for getting rid of unwanted cells, but is an active participant in the communication of the multi-cellular community (Fig. 1).

Figure 1.

 Roles of apoptosis in a multi-cellular community. AER, apical ectodermal ridge; FGF, fibroblast growth factor; ZPA, zone of polarizing activity; (A) Cell death in signaling centers. (B) Secretion of factors from dying cells. ATP, adenosine tri-phosphate; TGFβ, transforming growth factor β. (C) Epithelial cell death regulates the actin organization of neighboring cells. (D) Mutual interactions between cell death and proliferation.

Quantitative detection of cell death

Critical discoveries about cell death were achieved by studying it quantitatively. Levi-Montalcini and Hamburger carefully tracked the number of postmitotic neurons in the chick spinal motor column, and quantitatively examined the effects of neuronal target tissue removal or target tissue graft on the number of neurons. Based on this series of experiments, they concluded that the neuronal cell loss in developing neural tissue is due to cell death and not to migration away or to transformation into other cell types (Hamburger 1992). This quantitative neuronal-survival assay was also used to identify nerve growth factor by Cohen & Levi-Montalcini (1957).

Cell-lineage tracing over the entire development of Caenorhabditis elegans has proved as useful for investigating programmed cell deaths as for identifying specific cell fates (Sulston & Horvitz 1977). One hundred and thirty-one cell deaths occur during C. elegans development, and all the deaths are genetically programmed. Studies conducted by Horvitz and colleagues successfully identified the genes required for all 131 cell deaths (Ellis & Horvitz 1986; Ellis et al. 1991). The identification of the cell death genes in C. elegans and the discovery of the conserved caspase-mediated apoptotic pathway enabled the genetic and pharmacological manipulation of apoptotic cell death in vivo. Thus, researchers have been able to explore the in vivo roles of apoptosis experimentally.

Detection of cell death in studies of development

Although the occurrence of massive cell death during development is now widely accepted, quantitative studies on cell death in vivo are still limited and difficult to conduct. Even in C. elegans, the direct observation of cell death is not easy, because it occurs at various parts of the body at different developmental times, and the dead cells are engulfed and disappear within about 15 min (Sulston & Horvitz 1977). In C. elegans, the quantitative detection of cell death was made easier by the isolation of engulfment mutants, such as ced-1 and ced-2 (Hedgecock et al. 1983). Although, no such engulfment mutants are available in mammals, DNase II-deficient mice, in which the macrophages carry engulfed CAD-cleaved fragmented DNAs, facilitate the detection of apoptotic areas by TUNEL staining (Nagasaka et al. 2010). However, the ability to detect the precise sites of apoptosis is still limited, because the apoptotic cell-engulfed macrophages can move from their original location.

Since active caspase-3 or TUNEL-positive cells are rapidly engulfed by macrophages and displaced rapidly from their original site, researchers have developed tools for the detection of the early phase of apoptosis, such as the activation of the initiator caspase-9 in tissue sections. Such data can provide information about the spatial and temporal pattern of apoptosis initiation. Since more than 1000 cellular substrates for caspases have been reported so far (Dix et al. 2008; Mahrus et al. 2008), an antibody that specifically recognizes a caspase-9-cleaved substrate was sought as a powerful tool for studying the initiation of apoptosis in fixed tissues. Other than the proform of executioner caspases such as caspase-3, the intermediate filament vimentin was the first discovered caspase-9 substrate. Vimentin is widely expressed in developing tissues, including neural tissue, in early development. The neoepitope of caspase-cleaved vimentin is conserved in human and mouse, and an antibody specific for this neoepitope (anti-V1 antibody) was generated and used to examine caspase-9 activation in the developing embryo (Nakanishi et al. 2001). Immunostaining of E10.5 mouse embryos with the anti-V1 antibody revealed clusters of V1-positive signals in the lamina terminalis of the forebrain. Some apoptotic cells in the lamina terminalis were stained by both the V1 antibody and TUNEL; however, most apoptotic cells were either V1-positive or TUNEL-positive (Nakanishi et al. 2001). Thus, the anti-V1 antibody seems to recognize vimentin cleavage at an earlier stage of apoptosis than the degradation of chromosomal DNA that is detected by TUNEL staining.

At a later stage of neural development, vimentin expression is reduced and another intermediate filament, nestin, takes its place. Semaphorin 7A (Sema7A), a membrane-anchored member of the semaphorin family of axon-guidance proteins, was identified as another substrate for caspase-9 (Pasterkamp et al. 2003). Intriguingly, an antibody against the N-terminal portion of mouse Sema7A preferentially immunolabels the form of Sema7A cleaved by caspase-9, enabling caspase-9 activity to be monitored in vitro cell culture and in vivo. Immunostaining with this antibody revealed that Sema7A is cleaved by caspase-9 in the olfactory sensory neurons (OSNs) projecting to the anterior-medial and posterior ventral regions of the olfactory bulb, where caspase-3 is strongly activated, around the perinatal stage (Fig. 2, Shizue Ohsawa and Masayuki Miura, unpubl. data, 2006) (Ohsawa et al. 2010). Sema7A is a direct substrate for caspase-9, and a knockout of the caspase-9 gene in mice abolishes the Sema7A immunoreactivity. Caspase-9 or apaf-1 mutant mice exhibit misrouted axons, impaired synaptic formation, and defects in OSN maturation. In the aged olfactory bulb, the cleavage of Sema7A by caspase-9 and the caspase-9-like LEHD-cleaving activity is increased, compared with the young olfactory bulb. The YVAD (caspase-1-like)-, DEVD (caspase-3-like)-, and IETD (caspase-8-like)-cleaving activities did not differ between the young and aged olfactory bulb, suggesting that caspase-9 is specifically activated in the olfactory bulb with aging (Ohsawa et al. 2009). In both the developing and aged olfactory system, caspase-9 is activated and appears to execute non-apoptotic functions. The non-apoptotic functions of caspases, which include roles in cell proliferation, migration, differentiation, and immunity are currently being studied intensely (Kuranaga & Miura 2007).

Figure 2.

 Non-apoptotic caspase activation during olfactory development. (A) Coronal sections (5 μm thick) of embryonic day 14.5 (E14.5), E16.5, E18.5, and P4 wild-type mouse olfactory bulb were stained with anti-active-caspase-3 (green), anti-GAP-43 (blue; a marker for immature neurons including olfactory sensory neurons), and anti-OMP (magenta; a marker for mature olfactory sensory neurons (OSNs)) antibodies. Caspase-3 is strongly activated in the olfactory sensory axons projecting to the medial regions of the anterior olfactory bulb during late embryonic development (E16.5 and E18.5, arrowheads). Scale bar, 100 μm. (B) Coronal sections (5 μm thick) of E18.5 apaf-1 heterozygous (+/−) and homozygous (−/−) mutant olfactory bulb were stained with anti-Sema7A (green; a marker for the cleaved form of Sema7A by caspase-9), anti-NCAM (blue; a marker for OSN axons), and anti-OMP (magenta) antibodies. Arrowheads show OSN axons labeled with the anti-Sema7A antibody in the apaf-1 (+/−) mice. Scale bar, 100 μm. (C) Non-apoptotic functions of caspases during olfactory development. Mice mutant for apaf-1 or caspase-9 exhibit misrouted axons and impaired OSN maturation including synaptic formation, indicating that Apaf-1/caspase-9-mediated caspase signaling is required for the proper axonal outgrowth and maturation of olfactory sensory neurons. These non-apoptotic functions might be mediated by modulating the amount of Sema7A.

A change in the immunoreactivity of linker Histone H1 is another marker for early apoptosis. In both mammals and Drosophila, dying cells fail to be immunolabeled with an anti-H1 monoclonal antibody, AE-4. Real-time imaging of caspase activation and H1 dynamics in cultured mammalian neural cells revealed that H1 changed its location in the nucleus after caspase-3 activation (Ohsawa et al. 2008). From the staining patterns of AE-4 and anti-active caspase-3 antibodies, cells undergoing the transition from caspase activation to the apoptotic H1 change could be identified as H1-positive caspase-activated cells, providing a novel criterion for early apoptosis. Notably, according to these staining patterns, many OSNs in the developing mouse olfactory epithelium showed sustained caspase activity without the H1 change. As suggested by the anti-Sema 7A antibody staining, caspases may have unique functions in OSNs.

Genetically encoded probes for studying apoptosis in vivo

The live-imaging analysis of caspase-activated cells is useful for tracing cell-death signaling in vivo. A genetically encoded indicator for caspase activation called “Sensor for activated caspase based on FRET” (SCAT) has been developed (Takemoto et al. 2003). This method enables the quantitative monitoring of caspase-3 activity by fluorescence resonance energy transfer (FRET) between two types of fluorescent proteins (ECFP and the YFP variant Venus) that are linked by a peptide containing the caspase-3 cleavage sequence, DEVD. In the SCAT3 probe, ECFP is the FRET donor and Venus is the FRET acceptor, and the caspase activity is determined from the FRET from ECFP to Venus, because cleavage of the linker DEVD sequence by activated caspases abolishes the FRET. The change in FRET is represented as the emission ratio of Venus/ECFP. If the caspase activity is low, FRET occurs, resulting in a “high” emission ratio, but when caspase is activated, the emission ratio is reduced by the abolishment of FRET. In this way, the caspase activity can be represented semi-quantitatively, and this method can distinguish not only the strong activation of caspases that is required to execute apoptosis but also the weak caspase activation that is sometimes used for cell signaling (Fig. 3) (Kanuka et al. 2005; Kuranaga et al. 2006; Ribeiro et al. 2007).

Figure 3.

 SCAT3, a genetically encoded sensor for caspase activation. (A) Fluorescence resonance energy transfer (FRET) between two types of fluorescent proteins (ECFP and Venus) that are linked by a peptide containing the caspase-3 cleavage sequence, DEVD. Arrow indicates the caspase-3 cleavage site. (B) SCAT3-ratio imaging of HeLa cells treated with tumor necrosis factor-α (TNF-α)/cycloheximide. When the caspase activity is low or non-existent, FRET occurs resulting in a “High emission ratio”. When caspase is activated, the emission ratio is reduced by FRET abolishment. This enables the monitoring of caspase-3 like activity semi-quantitatively in vivo.

“Apoliner” is another caspase-activity sensor. This probe consists of, from the N-terminus, the transmembrane domain of mouse CD8, monomeric red fluorescent protein (mRFP), the caspase cleavage site of Drosophila inhibitor of apoptosis 1 (DIAP1) (DQVD), and enhanced green fluorescent protein (EGFP) with a nuclear localization signal (NLS-EGFP). The caspase cleavage of Apoliner generates mCD8-mRFP and NLS-EGFP, so caspase activation can be monitored by changes in the fluorescent protein’s localization (Bardet et al. 2008). The cleavage site of Apoliner contains the BIR1 domain of DIAP1, which enhances recognition by caspase, is cleaved specifically and efficiently by caspase. Although the FRET-based SCAT probe is more quantitative, it requires equipment for FRET imaging, while Apoliner does not.

In fixed tissue, CD8::PARP::Venus works well to detect caspase activation. Human poly-ADP-ribose polymerase-1 (PARP) contains the typical caspase-3 cleavage site DEVD, and is efficiently cleaved by caspase-3-like effector caspases. A highly specific antibody that recognizes the caspase-cleaved neoepitope of PARP is commercially available. Overexpression of the caspase substrate CD8::PARP::Venus enhances the sensitivity of the immunological detection of the caspase-cleaved product. Using this probe, caspase activity localized to the dendrites of sensory neurons in Drosophila was detected in the pupal stage (Williams et al. 2006).

The endogenous activity of caspase is regulated by inhibitor of apoptosis proteins (IAP), which was originally found in baculoviruses (Gyrd-Hansen & Meier 2010). In Drosophila, loss-of-function mutations in the thread (th) gene, which encodes DIAP1, result in early embryonic death through the inappropriate activation of caspases (Wang et al. 1999). DIAP1 contains a carboxy-terminal RING finger domain and functions as an E3 ubiquitin ligase (Gyrd-Hansen & Meier 2010). It suppresses caspase activation by binding directly to caspases and promoting their degradation through the ubiquitin-proteasome pathway. DIAP1 itself is a short-lived protein, with a half-life of approximately 30 min, and is degraded by the ubiquitin-proteasome pathway. During periods of programmed cell death, DIAP1 degradation is promoted by its binding to the proapoptotic proteins Reaper (Rpr), Head involution defective (Hid), and Grim. At the same time, caspases are released from DIAP1 and promote programmed cell death. Thus, the balance between the DIAP1 protein level and caspase activation determines whether cells will survive or die by apoptosis. DIAP1 turnover is therefore an indicator of the early caspase activation pathway. To monitor the DIAP1 protein turnover in living cells, a mutant DIAP1 that cannot bind to caspases was fused with the fluorescent protein Venus. This probe, PRAP (PRe-Apoptosis signal-detecting probe, based on DIAP1 degradation), was shown to reflect the endogenous turnover of DIAP1 well (Fig. 4) (Koto et al. 2009).

Figure 4.

 Development of external sensory organ. External sensory organ of Drosophila notum is formed by two external cells, the shaft (sf) and the socket (so) cells, and two inner cells, the internal neuron (n) and the sheath (sh) cell. The progenitor cells, pI (SOP), pIIa, pIIb, and pIIIb divide to form the five cells of the external sensory organ. The socket and shaft cell arise from the pIIa cell. The neuron, sheath, and glial (g) cell arise from the pIIb lineage. The glial cell soon dies by apoptosis. Caspase activation was monitored by SCAT3 and was detected only in the glial cell before it died.

During apoptosis, phosphatidylserine (PS) is exposed on the outer leaflet of the plasma membrane. Annexin V (A5) specifically binds to PS in calcium-dependent manner, and fluorescently labeled A5 is widely used to detect apoptosis in vitro. To achieve secretion of the A5, a secretion signal peptide was added to N-terminus of the human A5 protein. The secreted form of A5 fused with yellow fluorescent protein was constructed and expressed in zebrafish, and successfully labeled apoptotic cells in the living zebrafish (Van Ham et al. 2010). PS exposure is also observed in non-apoptotic cell death, caused by cathepsin B activation, perturbed Ca2+ homeostasis, or protein kinase C (PKC) activation (Hirt et al. 2000; Foghsgaard et al. 2001). Thus, this method can be used to detect both apoptotic and non-apoptotic cell death in vivo.

What does the live imaging of caspase-activation pathways tell us?

Despite the development of genetically encoded probes, the accessible tissues or animals that can be used for confocal microscopic observation are limited. The Drosophila pupa is a useful system for tracing apoptosis and its signals during development (Takemoto et al. 2007). Microchaetes are typical structures of the Drosophila peripheral nervous system. They are formed by two outer support cells (the socket cell and the shaft cell) and two inner cells (the neuron and the sheath cell), which all arise from asymmetric divisions of the sensory organ precursor (SOP) or precursor I (pI) cell (Fig. 4). The development of the entire sensory organ lineage, from the pI cell to bristle elongation, was traced by live imaging, and the dynamics of the cell-death signaling were monitored using PRAP (Koto et al. 2009). PRAP was detected in the pI cell and persisted in the pIIa and pIIb daughter cells; however, the PRAP signal disappeared from them before the pIIb cell divided. During the next cell division of pIIb, no PRAP was detected in the pIIa, pIIIb, or glial cell. The next division of pIIa produces the socket and shaft cells. During this step, no PRAP was observed in the pIIIb, pIIa, socket, or shaft cells. The final division, of pIIIb, produces the neuron and sheath cell, and these cells were also negative for PRAP. Following this final cell division and the death of the glial cell, the PRAP fluorescence re-emerged, in the socket cell and the shaft cell. Just before the beginning of bristle elongation, PRAP rapidly vanished from the shaft cell. Thus, the degradation of the endogenous caspase inhibitor DIAP1 was regulated in a cell-lineage- and stage-specific manner (Fig. 5). SCAT3 live imaging indicated that executioner caspase was activated only in the dying glial cells (Fig. 4). Since DIAP1 inhibits both initiator and executioner caspases, DIAP1 degradation might only activate the initiator caspase and regulate non-apoptotic caspases functions in the shaft cell. In fact, initiator but not executioner caspase activity is required for bristle elongation. Thus, the live imaging of cell-death signaling revealed that time-limited and reversible changes in cell death signaling occur during organ development (Fig. 5) (Koto et al. 2009). The DIAP1 degradation during SOP asymmetric cell division was the unique in vivo evidence that the generation of a new cell type during development is a stressful process associated with the activation of cell-death signaling pathway.

Figure 5.

 Dynamics of Drosophila inhibitor of apoptosis 1 (DIAP1) degradation during the asymmetric cell division of a sensory organ precursor cell. (A) Schematic representation of DIAP1 degradation and caspase activation. (B) Schematic representation of PRAP (PRe-Apoptosis signal-detecting probe). In vivo live imaging analysis of PRAP (green) was begun about 16 h after puparium formation. Each nucleus is marked by Histone2B-ECFP (magenta).

Programmed cell death evolved to recognize differences in cell populations and to participate in organismal maintenance

In metazoan development, different cell populations are generated within the same body (Fig. 6). This process must be intrinsically stressful for both the differentiating cells and their surrounding cells. For example, neural crest cells disperse on distinct pathways and produce different derivatives in a specific place in vertebrate embryo. During the migration and differentiation process, melanocytes but not neuronal cells migrate on the lateral migration pathway. A few neuronal cells that migrate in the lateral migration pathway are removed by apoptosis, suggesting that localized environmental factors control the selection of cells to establish the precise pattern of differentiated cell population (Wakamatsu et al. 1998). In early development, at the immunologically immature stage, the dramatic expansion of cell types can be achieved without interference by the immune system. Instead of immunological reactions, cell stress-response molecules such as JNK play major roles in inducing apoptosis (Igaki 2009), and the removal of apoptotic cells is executed by neighboring cells. Later in development, organogenesis begins, and each organ independently generates its own unique structure and function. As in the early developmental stage, stress responses are thought to play major roles in eliminating unwanted cells during each organogenesis event, but immune cells actively participate in surveying and eliminating the unwanted cells at the whole-body level.

Figure 6.

 Cell stress is generated during the establishment of a cellular community. Differentiation of cell in cellular communities during development is intrinsically stressful for both the differentiating cell and their surrounding cells.

In holometabolous insects, most larval tissues are scrapped during metamorphosis, and the cells or tissue fragments are engulfed by macrophage-like cells, called hemocytes. At the same time, imaginal cells proliferate rapidly and start to build the adult tissues. Thirty years ago, S. Natori and colleagues pointed out the similarity of the clearance of dying cells during development and the host defense mechanism, as quoted below.

In immune systems in higher organisms, macrophages do not normally phagocytize their own cells or tissue fragments, because they discriminate self and non-self strictly. Thus, insect phagocytes may also recognize larval tissue fragments as non-self at this stage, although these tissues are self in the larval stage (Komano et al. 1981).

We assume that some defence mechanism is needed in the ontogenic process of insects to eliminate unnecessary tissues or cells, and that the same mechanism is activated not only during ontogenesis but also in emergency such as on invasion of foreign substances, since insects have no alternative defence system such as the immune network in vertebrates (Takahashi et al. 1986).

The direct binding of blood cells to “damaged-self” tissue has been hypothesized to be an ancestral function of the immune system (Seong & Matzinger 2004). Similarly, the tissue cell removal system might be widely used for development, tissue remodeling, repair, and regeneration. After the epidermal wounding of a Drosophila larva, circulating blood cells are recruited to the site of damage and captured there. After capture, the blood cells spread across the wound surface, become phagocytically active, and clear the wound site of debris (Babcock et al. 2008). Recently, Nagata et al. (2010) demonstrated the importance of dead-cell clearance, by showing that the inefficient engulfment of dead cells activates the immune system and leads to autoimmune diseases in mammals. Thus, programmed cell death and the rapid clearance of dead cells are essential for organismal homeostasis. The programmed cell death machinery may have evolved to recognize the status of individual cells and to maintain organismal homeostasis in both developing and adult animals.

Origin of the homeostatic function of caspase-mediated cell death in eukaryotes

Caspase proteins are conserved in metazoans. A distant family of caspases, called metacaspases, is found in protozoans, fungi, and plants. Another family, called paracaspases, is found in Dictyostelium and metazoans (Uren et al. 2000). The yeast metacaspase YCA1 regulates the cell death induced by stresses such as oxidative stress, acetate, or chronological aging in Saccharomcyes cerevisiae. Yeast cells with a mutation in YCA1 initially show resistance to the apoptosis due to chronological aging. However, the survivors lose their ability to regrow, indicating that damaged cells accumulate within the population of apoptosis-deficient cells. Moreover, wild-type cells outlast the yca1-null mutants in direct competition assays during long-term aging (Herker et al. 2004). Thus, the apoptosis in yeast might have a selective advantage for this unicellular organism in a population. Furthermore, old yeast culture medium contains substances that stimulate the survival of aged cells, suggesting some beneficial effects of cell death in old culture (Herker et al. 2004). Thus, in monocellular eukaryotes, programmed cell death appears to be indispensable to maintain the homeostasis of a cell population and/or cell community.

Caspases and the innate immune system

The conserved connection between programmed cell death mechanisms and defense mechanisms is supported by the study of ced-3 and ced-4 mutants in C. elegans. These mutant animals develop normally, except that all the programmed cell death is abolished under standard laboratory culture conditions. When C. elegans is fed Salmonella typhimurium, the bacteria colonize the intestine, leading to an increased level of cell death in the worm gonad. This S. typhimurium-mediated germ-line cell death is prevented in the ced-3 and ced-4 mutants. However, these mutants are hypersusceptible to S. typhimurium-mediated killing (Aballay & Ausubel 2001). These results suggest that caspase is involved in the C. elegans defense response to pathogen attack. However, whether germ cell death is required for this susceptibility has not been clarified.

The Drosophila innate immune system is composed of two major pathways, Toll and IMD (immune deficiency) (Buchon et al. 2009). The Toll pathway recognizes Gram-positive pathogens (mostly fungi), and the IMD pathway responds to Gram-negative ones (mostly bacteria). These two independent pathways regulate distinct classes of nuclear factor-κB (NF-κB) proteins. The Toll pathway activates Dorsal and Dif, while the IMD pathway leads to the activation of Relish. dredd, a Drosophila orthologue of caspase-8 is an essential component of the IMD pathway.

Caspase-1 (Interleukin-1β-Converting Enzyme, ICE), the first reported mammalian homologue of Ced-3 (Miura et al. 1993; Yuan et al. 1993), was identified as a unique cysteine protease that processes pro-interleukin-1β (IL-1β) to mature IL-1β. Caspase-1 also processes IL-18 and IL-33, and thus is responsible for both inflammatory responses and cell death. The Caspase-1-activating complex is called the “inflammasome” (Schroder & Tschopp 2010). It is composed of Caspase-1, ASC (apoptosis-associated speck-like protein containing a caspase-recruitment domain), and Apaf-1-like molecules called NLRs (NOD-like receptors), which can sense different bacteria, toxins, or endogenous “danger signals” released from damaged cells. Thus, Caspase-1 activation in macrophages might be an important cellular response occurring through the Apaf-1-like components’ sensing of infection or cellular damage in the body.

In mammals, caspase-1 regulates not only inflammatory cytokine release, but also the unconventional protein secretion of leaderless proteins such as fibroblast growth factor-2 (FGF-2) and thioredoxin-1 (Keller et al. 2008). Since many of the caspase-1-dependent secreted proteins are involved in inflammation, cytoprotection, and tissue repair, caspase-1 might play a role in the reconstruction of tissues under stressed or damaged conditions.

Self/non-self recognition activates cell death during development

The immune system is activated when cells recognize other cells as non-self. The recognition of self/non-self is mediated by immune cells in adults, but during development, neighboring cells can also sense differences in cells and remove damaged or slowly growing cells by inducing cell death (cell competition) (Moreno et al. 2002; Li & Baker 2007).

Sponges (phylum Porifera) are the phylogenetically oldest metazoan phylum. Marine sponges have the machinery for apoptosis, which they execute during the formation of the asexual reproduction bodies called gemmules and to get rid of unwanted cells accumulating during development and homeostasis. Interestingly, sponges may also have a system for recognizing self and non-self. When two sponge species grow towards each other, they first adhere and then they separate. In allografting experiments, the induction of apoptosis was observed during allograft rejection. DEVDase activity increased during the rejection period, and an inhibitor for caspase-3 prevented the apoptosis of the rejected graft. Two cDNAs encoding caspase-3-related protein were cloned (CAS3l and CAS3s) from the marine sponge G. cydonium. CAS3s and CAS3l appear to be generated by the alternative splicing of the same gene. CAS3s resembles human caspase-3, and CAS3l contains a CARD domain at its N-terminus. This CAS3-encoding gene is induced in the rejected apoptotic graft (Wiens et al. 2003, 2004). Although whether a direct cellular interaction between self and non-self activates the caspase-mediated apoptotic signaling or whether immune cells are involved has not been tested yet, the sponge rejection system might represent a prototype of the activation of cell death signaling by the recognition of self/non-self.

Amphibian metamorphosis is regulated by thyroid hormone (TH), and for many years it was thought that TH plays a crucial role in tail regression by inducing programmed cell death in a variety of tissues in a cell-autonomous manner. However, an immune-mediated rejection system has been shown to play crucial roles in tail resorption during amphibian metamorphosis. The keratin-related genes ouro1 and ouro2 are specifically expressed in the regressing tail skin at the climax of metamorphosis. Ouro proteins function as tail antigens whose expression allows tail skin cells to be recognized as non-self by immune cells, thereby mediating an immune-based mechanism of tail degeneration (Mukaigasa et al. 2009). This newly identified immune cell-mediated self-tissue or cell rejection system may also be used in other vertebrate development.

Concluding remarks

Cell death is widely observed in embryos, developing tissues, and self-renewing tissues. If cell death is intrinsically included in developmental and homeostatic processes, the elimination of stressed, damaged-self, or non-self-like cells in the body must be executed in many different ways. In deficient mice for caspase (knockout of caspase-3, caspase-7, caspase-9 or double knockout of caspase-3/7) and its activating pathway genes (knockout of apaf-1 and double knockout of bax/bak), effects of gene knockout on the early development, except for brain, are not so drastic. In C57BL/6 genetic background, knockout animals of caspase-3, caspase-9, or apaf-1 sometimes survive to adulthood. Whereas it has not been tested in detail whether apoptosis is severely prevented or not, the redundant pathway could play roles in activation of caspase cascade in these knockout mice (Kumar 2007; Giam et al. 2008). Another possibility is the activation of alternative cell death pathway to remove unwanted cells during development. Caspase-3 activation was prevented in most tissues in apaf-1 knockout embryo; however, many non-apoptotic dead cells were observed (Nagasaka et al. 2010). In adult mice, the intestinal epithelium, a self-renewing monolayer of epithelial cells that is generated from the intestinal stem cells of crypts, turns over in 2–3 days. A caspase or apaf-1 deficiency does not affect this turnover of epithelial cells. A morphological study of the small intestinal epithelium indicated that multiple pathways are used to execute these cell deaths (Mayhew et al. 1999). Thus, organisms expend much effort to execute cell death, because cell death is critical for building up and maintaining cellular communities during development as well as in adults.


I thank all of the lab members of Department of Genetics, The University of Tokyo, especially E. Kuranaga, Y. Yamaguchi, T. Chihara, A. Koto and S. Ohsawa for stimulating discussions. I also thank N. Funayama (Kyoto University) for discussions about cell death in sponges. I thank E. Kuranaga, A. Koto, S. Ohsawa, K. Takemoto and K. Nonomura for unpublished pictures and figures. This work was supported in part by grants from the Japanese Ministry of Education, Science, Sports, Culture and Technology and a RIKEN Bioarchitect Research Grant.