In vivo cellular and molecular mechanisms of neuronal apoptosis in the mammalian CNS

Abstract

Apoptosis has been recognized to be an essential process during neural development. It is generally assumed that about half of the neurons produced during neurogenesis die before completion of the central nervous system (CNS) maturation, and this process affects nearly all classes of neurons. In this review, we discuss the experimental data in vivo on naturally occurring neuronal death in normal, transgenic and mutant animals, with special attention to the cerebellum as a study model. The emerging picture is that of a dual wave of apoptotic cell death affecting central neurons at different stages of their life. The first wave consists of an early neuronal death of proliferating precursors and young postmitotic neuroblasts, and appears to be closely linked to cell cycle regulation. The second wave affects postmitotic neurons at later stages, and is much better understood in functional terms, mainly on the basis of the neurotrophic concept in its broader definition. The molecular machinery of late apoptotic death of postmitotic neurons more commonly follows the mitochondrial pathway of intracellular signal transduction, but the death receptor pathway may also be involved.

Undoubtedly, analysis of naturally occurring neuronal death (NOND) in vivo will offer a basis for parallel and future studies aiming to elucidate the mechanisms of pathologic neuronal loss occurring as the result of conditions such as neurodegenerative disorders, trauma or ischemia.

Introduction

Programmed cell death (PCD) is a highly phylogenetically conserved mechanism by which eukaryotic cells die following a stereotyped series of molecular and cellular events commonly referred to as apoptosis (Glucksmann, 1951, Saunders, 1966). Apoptosis has been recognized to be an essential process during development where it appears to be fundamental for the control of the final numbers of neurons (and glial cells) in the central nervous system (CNS) and peripheral nervous system (PNS). Moreover, a growing body of evidence is accumulating to indicate that apoptosis is also responsible for the loss of neurons associated with physiological aging (Taglialatela et al., 1996, Kaufmann et al., 2001).

It is generally assumed that about half of the neurons produced during neurogenesis die before completion of the CNS maturation, and nearly all classes of neurons are produced in excess during development. These oversized populations of neurons are then significantly reduced during the periods of naturally occurring neuronal death (NOND), mainly upon activation of the apoptotic machinery. The neurotrophic theory has clearly established that neurons compete for limited amounts of target-derived trophic factors that have a protective role against apoptosis, and has provided a fundamental conceptual basis for a wide number of studies aiming to elucidate the mechanisms of neuron-to-target interaction eventually leading to cell survival or death. The initial concept of neurotrophic factor based on this theory has now broadened to embrace a large and molecularly heterogeneous group of surviving factors, and parallel studies have disclosed the role of electrical activity in survival during the wiring of neural networks.

Still, the general idea beyond this expanded concept is that NOND is closely related to the establishment of proper connections with targets, and under this perspective little or no attention has been dedicated to apoptosis of neural precursors and/or young neuroblasts at early developmental stages, which seems to be independent from synaptogenesis. In several areas of the brain, including the retina and the cerebellum, it appears that two subsequent periods of cell death can be observed: the first mainly occurs at the onset of neurogenesis and is not apparently related to synapse formation, while the second is linked to the wiring of young postmitotic neurons. In both periods, cell death is apoptotic (Bahr, 2000, De la Rosa and De Pablo, 2000, Lossi et al., 2002c). Why so many neurons are generated in excess and eliminated shortly thereafter remains to be understood, having also in mind that several lines of transgenic animals, in which essential cell death genes have been knocked out and/or apoptosis inhibitors have been overexpressed, display no obvious or detectable CNS defects.

In neurodegenerative disorders and traumatic neuronal injury, neuronal loss may be linked to apoptosis. However, in these conditions cell death may also be necrotic. In the latter (exogenous cell death), alterations of the cellular environment eventually result in cell swelling and disruption of the cell membrane, while the main elements of the apoptotic machinery are constitutively expressed or generated by the cell itself (cell suicide). Although the two types of cell death have initially been considered independent from each other, it is now clearly emerging that they share some cellular and molecular features, and that cells can switch from one mode of death to the other upon different conditions.

A tremendous amount of literature in the recent years has deepened our knowledge on the cellular and molecular mechanisms of apoptosis in neurons and other cells. Current knowledge about the genetic regulation of apoptosis is mainly based on studies of the nematode worm Caenorhabditis elegans. These studies have been extensively reviewed in numerous publications (Hengartner and Horvitz, 1994b, Hoffman and Liebermann, 1994, Stewart, 1994, Vaux et al., 1994, Yuan, 1995, Fraser et al., 1996, White, 1996, Meier and Evan, 1998, Liu and Hengartner, 1999). They will be only briefly mentioned here to put things under the right perspective. Also, the characterization of apoptosis in vertebrate (mammalian) neurons mainly relies on in vitro studies, and readers are again invited to consult the existing reviews on this issue (Rao and White, 1997, Selimi et al., 1997, Sastry and Rao, 2000, Denecker et al., 2001, Shastry et al., 2001).

Our present work summarizes the literature on apoptosis of mammalian central neurons in vivo, with a particular attention for NOND in the postnatal cerebellum, since this is our major field of interest. In doing so we will also consider with attention the relationship of apoptosis and cell proliferation in early developmental death of neural precursors and/or young neuroblasts, and the data available on transgenic and mutant animals.

The issue of morphological diversity in developmental cell death has been reviewed in detail (Clarke, 1990). In an ultrastructural study on several embryonic tissues it was proposed that there are three main types of cell death during normal development, on the basis of the role of lysosomes in cell disruption (Schweichel and Merker, 1973). In the first type, cell death occurs without any detectable activation of endogenous lysosomes, but cells are eventually destroyed by phagocytosis and secondary lysosome activation by tissue macrophages. This process has also been referred to as heterophagocytosis. In the second type of cell death (autophagocytosis) cells are eliminated after activation of their own lysosomal enzymes. In the third type, there is no obvious lysosome intervention. The first two types are by far more common and have been described by various authors starting from the 1960s. The ultrastructural features of type 1 cell death in Schweichel and Merker classification correspond to the current definition of apoptosis, while at least a variant of type 3 shares several features with necrosis (Clarke, 1990).

Apoptosis was originally defined as a distinct mode of cell death on the basis of a series of characteristic ultrastructural features (Fig. 1E and F) according to the following sequence of events: nuclear and cytoplasmic condensation, cell fragmentation and phagocytosis (Kerr et al., 1972). Initially Kerr et al. used the term “shrinkage necrosis” to describe this form of cell death. Subsequently they coined the term “apoptosis” (from the Greek=falling of the leaves), which indicates the dropping of leaves from trees or petals from flowers, to emphasize the role of this type of cell death in normal tissue turnover.

Apoptosis involves a series of stereotyped, morphologically well defined phases, that are most clearly evident at the electron microscope level.¹

Changes in the nucleus represent the first unequivocal evidence of apoptosis (Fig. 1E). Chromatin condensation and segregation into sharply delineated masses that abut on the nuclear envelope are typically observed at the onset of apoptosis. Chromatin masses are very electrondense and often show a characteristic crescent-like appearance. High magnification electron micrographs reveal that these masses are made up of closely packed, fine granular material.

This initial condensation eventually leads to true nuclear pyknosis. In parallel with nuclear changes, cytoplasm condensation also occurs, and the cell membrane becomes convoluted with the onset of protuberances of various sizes that may give the cell a star-like appearance. As the cytoplasm density increases, some vacuoles may become evident, but the cell organelles remain unaffected, although they become abnormally closely packed, likely as a consequence of the loss of cytosol. However, ribosomes can be detached from the rough endoplasmic reticulum and from polysomes, and these latter eventually disappear. As the process continues, the cell and its nucleus assume a more irregular shape and nuclear budding occurs to produce discrete fragments, still surrounded by an intact nuclear envelope. Eventually, the cell is fragmented into membrane-bounded apoptotic bodies which still display a sharp segregation of condensed chromatin in nuclear fragments and well preserved organelles. Apoptotic bodies are rapidly cleared out in tissues by macrophages or neighboring cells, and are degraded within heterophagosomes (Fig. 1F).

We have recently described the clearance of apoptotic cells during NOND in the postnatal rabbit cerebellum in vivo (Lossi et al., 2002c) after labeling proliferating cells with a modified “window-labeling” technique (Belecky-Adams et al., 1996) to follow their fate. We have thus observed that the whole apoptotic cerebellar granule cells (CGCs) are phagocytosed by dark (medium to highly electrondense) glial elements, before being fragmented into apoptotic bodies. These glial elements likely corresponded to microglia. At early stages of apoptosis, CGCs in the external granular layer (EGL) were often seen to be contacted by glia, that bent around them and became engulfed with the entire apoptotic cell. Cells with the typical features of late apoptosis were more easily observed in the internal granular layer (IGL) and commonly fragmented into several apoptotic bodies inside the heterophagosomes. It was rather easy to spot some of these phagocytic cells engulfed with apoptotic material in close apposition to blood capillaries, and indeed we have provided ultrastructural evidence for the presence of intraluminal blood monocytes engulfed with heterophagosomes. These latter contained highly condensed nuclear DNA pre-labeled in vivo with a S-phase marker 24 h before. Since proliferation of the CGCs only occurs in the EGL, our data demonstrated that in the limited span of time between tracer administration and sacrifice some CGCs completed their division, entered the apoptotic program and were cleared by the glia. This observation gives an unequivocal demonstration of the remarkable speed by which apoptotic cells are removed from the intact nervous tissue.

Necrosis (from the Greek=degeneration of the corpse) indicates the death of a cell or group of cells usually as a result of injury, disease or pathological state. Therefore, necrosis is traditionally associated with one or more exogenous factors leading to non physiological cell death. This latter may be divided into two main stages: the death of the cell that can be defined as the irreversible alteration of the cellular mechanisms that enable the cell to maintain its homeostasis, and the subsequent degeneration of the dead cell (Clarke, 1990). In general terms, necrosis involves large numbers of cells which are often grouped together and is associated with an inflammatory reaction. In the irreversibly injured cells, the morphological changes characteristic of necrosis consist in clumping of the chromatin without marked changes in its distribution: usually chromatin condenses in multiple small clumps with irregular outlines, and is poorly demarcated from the surrounding nucleoplasm. Sometimes densities in the matrix of abnormally swollen mitochondria, and local membrane disruption become evident. At later stages, there is a more or less pronounced disintegration of cell organelles and membranes, although the cell somehow maintains an overall identity. The chromatin disappears at the end of the process, leaving only “ghost-like” cell residues. Usually, groups of several cells undergo necrosis in tissues, with a marked inflammatory response that eventually leads to the removal of the necrotic debris by cells of the mononuclear phagocytic system.

Starting from pathology, but nowadays also in the broader field of cellular and developmental neurobiology, the term apoptosis rapidly caught on, in particular to emphasize the normal occurrence of this type of cell death in contrast to the association of necrosis with pathological insults (Migheli et al., 1994, Stewart, 1994, White, 1996, Marks and Berg, 1999). This dichotomy of apoptosis versus necrosis has been originally based upon the morphological differences between the two modalities of cell death, although some authors started to question such a sharp morphological distinction already at the beginning of the nineties (Clarke, 1990). In more recent times, it has become clear that, in mammalian cells, the gap between classical apoptosis and necrosis is filled by many intermediate morphological types, in which blebbing may be more or less evident and varying degree of chromatin condensation and margination may be apparent (Leist and Jaattela, 2001). In parallel, with the more and more in depth dissection of the cellular and molecular mechanisms of apoptosis (see Section 1.2) it became clear that specific cellular pathways are activated in apoptosis leading to an enforcement of the concept that, in this type of death, cells are responsible of their own demise, a reason for which apoptosis is commonly referred to as a “cell suicide”. However, in more recent times this axiomatic association of apoptosis and physiological cell death turned out to be an oversimplification for several reasons. First, the intrinsically necessary elimination of specific cell populations during development² of multicellular organisms is often, but not always characterized by an apoptotic morphology (Schwartz et al., 1993, Leist and Jaattela, 2001). Second, apoptosis besides being relevant to an array of physiological functions (that in addition to development, comprise the differentiation and maturation of various types of cells, and several functions of the immune system), is involved with cell injury induced by a spectrum of physical and chemical agents (Boobis et al., 1990, Stewart, 1994, Ortiz et al., 2001, Yakovlev and Faden, 2001, Dainiak, 2002). Third, apoptosis is concerned with oncogenesis, tumor homeostasis, and the action of cytotoxic drugs employed in chemotherapy (Hoffman and Liebermann, 1994, Stewart, 1994, Mimeault, 2002, Singh et al., 2002). Fourth, more and more evidence is accumulating to show a role of apoptosis in several neurodegenerative disease, including Alzheimer disease (AD) and aging (Marks and Berg, 1999, Sastry and Rao, 2000), although it has been questioned whether during aging of human brain there is cell loss at all (Morrison and Hof, 1997).

In addition, a more accurate analysis of the cellular mechanisms of apoptosis suggested that at least some executioners of apoptotic and non apoptotic cell deaths may be identical (Benchoua et al., 2001, Moroni et al., 2001, Smith, 2001, Cole and Perez-Polo, 2002, Fujikawa et al., 2002, Hou and MacManus, 2002, Schwab et al., 2002).

Therefore, as the investigation of cell death in different systems proceeds, we find more and more variations to the classical concept of apoptosis versus necrosis (Lockshin and Zakeri, 2002).

As already mentioned, apoptosis and necrosis are not the sole modes of cell death (Clarke, 1990). Since the discovery of caspases (see below), the most widely studied caspase-independent cell deaths were those of the autophagic type (Lockshin and Zakeri, 2002). However, long before caspases were discovered, the role of lysosomes in cell death has been extensively investigated (Clarke, 1990, Zakeri et al., 1995, Clarke et al., 1998).

The term autophagy (from the Greek=self-eating) refers to a type of cell death in which the cytoplasm is actively destroyed by lysosomal enzymes well in advance before nuclear changes become visible. The most characteristic features are the appearance of large autophagic vacuoles of lysosomal origin in the cytoplasm (Fig. 1G). In these cells, although many of the characteristic changes of apoptosis eventually become evident, they are notably delayed, and substantial cellular degradation is evident, before the typical nuclear alterations of apoptosis occur. Finally, when about three quarters of the cytoplasm has been destroyed, it begins to condense, and chromatin coalescence and margination become apparent. In parallel, agarose gel electrophoresis reveals the appearance of DNA oligomer fragmentation, and the remnants of the cell are phagocytosed as in classical apoptosis (Zakeri and Lockshin, 2002).

Our current knowledge on the gene regulation of apoptosis in nerve cells (and other cell types) is mainly derived from studies on the nematode worm C. elegans (Yuan, 1995, Meier and Evan, 1998, Liu and Hengartner, 1999, Hengartner, 2001). In C. elegans, 131 out of 1090 somatic cells undergo developmental apoptosis. Four genes named ced-3, ced-4, ced-9 and egl-1 form the “death machinery” which is responsible for the execution of the cells undergoing PCD (Hengartner, 1997). Genetic studies in mutant animals have demonstrated that ced-3 and ced-4 are both required for cell execution. Also they have shown that ced-9 is a survival factor that negatively regulates the activity of ced-3 and ced-4 (Hengartner et al., 1992). Ectopic expression of ced-4 can induce cell death. However, the CED-4 protein is not directly responsible for the killing of cells. Rather, CED-4 acts as an adaptor molecule that regulates the killing activity of CED-3. Genetic studies in which ced-3, ced-4 and ced-9 were ectopically expressed in the six touch cells of C. elegans led to the demonstration that ced-4 acts upstream of ced-3 and may be regulated by ced-9 (Shaham and Horvitz, 1996). In keeping, the CED-9 protein acts as a survival factor that protects cells from being killed by interacting with CED-4. A recently discovered member of the death effector family in C. elegans is EGL-1 (Liu and Hengartner, 1999). The loss of egl-1 function results in survival of all cells that normally undergo apoptosis. EGL-1 might activate PCD by binding to CED-9 and inhibiting its activity. Thereby CED-4 is released from the CED-9/CED-4 containing complex (Conradt and Horvitz, 1998).

After the mammalian homologues of the death machinery genes of C. elegans were cloned, it became clear that these genes have similar functions in the regulation of apoptosis in all species examined so far.

The mammalian homologues of CED-3 are a growing family of related proteases, which are collectively referred to as caspases (caspase=cysteine aspartate protease; Thornberry and Lazebnik, 1998). The first mammalian caspase to be identified was the interleukin-1β-converting enzyme (ICE or caspase 1; Yuan et al., 1993, Schwartz and Milligan, 1996). Later it was demonstrated that caspase 3 shows the highest sequence homology to CED-3 (Fernandez-Alnemri et al., 1994, Nicholson et al., 1995). To date the 14 members of the caspase family can be divided into three subfamilies on the basis of the peptide-sequence preferences of their substrates: (i) the ICE-protease subfamily (caspases 1, 4, 5, 13 and 14, and murine caspases 11 and 12); (ii) the CED-3 subfamily (caspases 3, 6, 7, 8, 9 and 10); and (iii) the caspase 2 subfamily (Table 1).

Among these, caspases 3, 6, and 7 have a short pro-domain and degrade vital cellular proteins (see also Table 2); the others have long pro-domains that mediate protein–protein interactions, and only in certain circumstances (perhaps with the exception of caspases 1 and 11) may trigger apoptosis (Thornberry and Lazebnik, 1998).

Each caspase is initially synthesized as a zymogen and requires processing at specific cleavage sites to generate the active enzyme (Stennicke and Salvesen, 1999). The caspases that are the first to be activated, in turn trigger other downstream caspases giving rise to a proteolytic cascade that culminates with the execution of apoptosis.

Evidence for the in vivo relevance of individual caspases in mammalian apoptosis within the CNS (and PNS) is mainly based upon studies on knockout mice (see Section 3).

Different subsets of caspases are activated depending upon the pro-apoptotic stimulus. For example, caspases 3, 6, and 8 are part of the Fas/TNF-mediated death pathway, while caspases 9 and 3, together with apoptosis protease activated factor 1 (Apaf1) and cytochrome c participate in mitochondria-associated cell death (Zimmermann et al., 2001). These two pathways do not seem to be completely independent since a link was reported through BID, a protein that mediates cytochrome c release from mitochondria in response to activation of cell surface death receptors (Li et al., 1998, Luo et al., 1998).

The substrate specificity of CED-3 is more similar to caspase 3 than to caspase 1 or 2 (Xue et al., 1996). Therefore, the CED-3/CED-4/CED-9 complex of C. elegans resembles the caspase 9/caspase 3/Apaf1/cytochome c complex of mammals.

The differences in morphology of apoptotic cells in C. elegans and mammals may be related to differences in downstream events of apoptosis. In C. elegans, apoptotic cells show chromatin shrinkage and cytoplasm condensation. However, there are no evident apoptotic bodies as in mammals. Each of the 14 different mammalian caspases may be responsible for cleavage of several different molecules, reflecting the more complex structural morphology of apoptosis (Table 2).

The CED-4 homologue in mammals is Apaf1 (Zou et al., 1997). Apaf1 is part of a large protein complex called the apoptosome (Adams and Cory, 2002). As already mentioned, cellular demolition in apoptosis is carried out by caspases. Their activation, however, requires a number of scaffolding proteins that act in the apoptosome. In particular, biochemical studies have revealed that caspase 3 processing required not only caspase 9, but also Apaf1 and cytochrome c (Liu et al., 1996, Li et al., 1997). Upon binding to cytochrome c (also referred to as Apaf2) which is released from mitochondria at the onset of apoptosis, a series of conformational changes lead to Apaf1 multimerization and association with procaspase 9 (also referred to as Apaf3) with the generation of the about 1 MDa molecular weight apoptosome (Adams and Cory, 2002).

The biological role of Apaf1 in the nervous system is still under debate, mainly as a consequence of a series of unexpected findings derived from studies in transgenic animals (see Section 3).

Since the first observation that the ced-9 gene encodes a protein that is similar to mammalian B-cell lymphoma-2 (BCL-2) protein (Hengartner and Horvitz, 1994a), at least 15 proteins structurally related to CED-9 and BCL-2 have been identified to date, and grouped in the so-called BCL-2 protein family (Adams and Cory, 1998, Chao and Korsmeyer, 1998, Newton and Strasser, 1998, Sadoul, 1998, Sanchez and Yuan, 2001). All members of the family posses at least one of the four conserved BCL-homology (BH) domains (BH1, BH2, BH3, and BH4). Besides to BCL-2, some of these proteins (such as BCL-X_L and BCL-W) act as survival factors, whereas others (such as BAX, BAK, BAD, BID) are pro-apoptotic. CED-9 and the anti-apoptotic members of the BCL-2 family protect cells from death by at least two different mechanisms. First BCL-2 and BCL-X_L can heterodimerize and form pores, similar to those of the membrane-inserting domains of diphtheria toxin and colicins (Liang and Fesik, 1997), that act as channels for ions or even small proteins to cross the outer mitochondrial membrane. Thus, BCL-2 and BCL-X_L (located at the cytoplasmic side of the outer mitochondrial membrane) can prevent the release of cytochrome c from mitochondria, avoiding assembling of the apoptosome, and thereby protecting cells from being killed (Li et al., 1998, Luo et al., 1998). Second, in parallel to what happens in C. elegans (where CED-9 protects cells from death by interacting with CED-4, which, in turn, regulates, the activity of CED-3), BCL-X_L in mammals interacts with Apaf1. This interaction occurs via the BH4 domain and results in inhibition of the association of Apaf1 with procaspase 9, again with blockage of apoptosome formation and caspase 9 activation (Adams and Cory, 2002).

Work on distribution of the BCL-2 family members in the nervous tissue (see 2 In vivo analysis of NOND in the mammalian CNS, 3 Transgenic models and mutant animals) led to the conclusion that, in individual neuronal populations, there is an intricate network of interactions, with delicate balance and interchangeable functions, that modulate the pro- and anti-apoptotic function of these proteins. Noteworthy, function interchangeability of individual members of the family, such as, for example, the conversion of BCL-2 to a BAX-like death effector by caspase 2 (Cheng et al., 1997), generates further complication in our understanding of the molecular mechanisms mediated by the BCL-2 family proteins.

Several different stimuli can initiate the apoptotic death of neurons. However, the finding that common morphological and biochemical alterations are observed independently upon the event that triggers apoptosis, suggests that most apoptotic pathways converge on a restricted number of common effector pathways (Sastry and Rao, 2000).

Basically two major pathways can be differentiated by the relative timing of caspase activation and mitochondrial release of cytochrome c. In the first, which is exemplified by activation of death receptors, an effector caspase is activated prior to mitochondrial alterations. In the other, cytochrome c is released from the mitochondrial intermembrane space prior to caspase activation.

Death receptors are cell surface receptors that trigger apoptosis. There are several types of death receptors in different tissues, but two members of the tumor necrosis factor receptor (TNFR) family were recently demonstrated to be involved in neuronal death (Raoul et al., 2000): Fas (CD95/Apo-1) and the p75 neurotrophin receptor (p75^NTR; Fig. 2A and B).

Fas has been extensively studied as a death receptor in lymphocytes. Upon exposure to cell death-triggering stimuli, lymphocytes express at the surface of their membrane the Fas ligand (Fas-L) which binds Fas through an autocrine or paracrine mode. Upon formation of the Fas/Fas-L complex, the Fas-associated death domain (FADD) adaptor protein activates the signaling procaspase 8, which eventually activates the effector caspase 3 (Blatt and Glick, 2001).

In the classical paradigm of trophic deprivation in cultured phaeochromocytoma (PC12) cells, CGCs, and spinal motorneurons, blockage of Fas/Fas-L interaction leads to reduction of cell death, indicating that Fas activation is an obligatory key step in trophic factor deprivation induced-apoptosis (Brunet et al., 1999, Le Niculescu et al., 1999, Raoul et al., 1999). Other studies have further shown that up-regulation of Fas-L expression in neurons consistently occurs as an early step in apoptosis induced by injury or stress (Raoul et al., 2000).

The control of induction of Fas-L expression is achieved by at least two pathways: the first involves Jun amino-terminal kinase (JNK; Le Niculescu et al., 1999, Martin-Villalba et al., 1999), the second protein kinase B (PKB or Akt; Francois and Grimes, 1999, Brunet et al., 1999, Brunet et al., 2001). It seems also possible that Fas activation itself may trigger neuronal death, in the absence of Fas-L up-regulation (Raoul et al., 1999).

Although p75^NTR was the first neurotrophin receptor to be discovered, its biological role and mode of action still are not completely understood. Initially p75^NTR was the only known receptor for nerve growth factor (NFG), but its low affinity for NGF and lack of signaling motifs in the cytoplasmic domain appeared to be major shortcomings to assign it a prominent role in the transduction of NGF biological effects. After the discovery of the high affinity tyrosine kinase neurotrophin receptors (Trks), p75^NTR was hypothesized to act as a Trk-co-receptor (Kaplan and Stephens, 1994, Wolf et al., 1995). More recently, attention has been focused on its putative role in neuronal apoptosis (Barker, 1998, Barrett, 2000, Raoul et al., 2000, Miller and Kaplan, 2001).

This was first suggested when it became clear that p75^NTR was a member of the same family of transmembrane receptor as TNFR and Fas. The first reports on the apoptotic effect of p75^NTR came from in vitro studies on neuronal cell lines (Alles et al., 1991, Barrett and Bartlett, 1994, Rabizadeh and Bredesen, 1994, Bredesen and Rabizadeh, 1997). Examples of a role for p75^NTR as an inducer of apoptosis in vivo are still relatively few and somehow contradicting. After antisense oligonucleotide administration to rat pups it was demonstrated that p75^NTR is necessary for post-axotomy death of sensory neurons (Cheema et al., 1996). To study the potential pro-apoptotic effects of p75^NTR, a transgenic mouse expressing the p75^NTR intracellular domain under the control of a Tα1 α-tubulin promoter was generated (Majdan et al., 1997). Interestingly this animal model showed that the ability of p75^NTR to cause apoptosis is restricted not only to certain neuronal types, but also to certain temporal windows during the life of these neurons. However, p75^NTR knockout mice show reduced neuronal death in the retina, certain spinal cord interneurons, and sympathetic neurons (Bamji et al., 1998, Frade and Barde, 1999). In the retina and spinal cord, observations in ngf^−/− mutants, indicate that NGF binding to p75^NTR is the triggering event responsible for apoptosis (Frade and Barde, 1999). Sympathetic neuron cell death is instead triggered by p75^NTR binding to brain-derived neurotrophic factor (BDNF), as suggested by analysis of bdnf^−/− mice (Majdan et al., 1997). Finally, it should be mentioned that additional putative p75^NTR-associated signal-transduction elements, such as the transcription factor nuclear factor kappa B, may modulate neuronal apoptosis (Taglialatela et al., 1997, Taglialatela et al., 1998).

Despite of the above, still it remains difficult to assign a clear role in apoptosis to p75^NTR as a consequence of our lack of knowledge of its intracellular signal transduction pathway (Barrett, 2000). This latter is unlikely to be the same as for Fas (or TNFR) since p75^NTR has a different death domain (Liepinsh et al., 1997). Interestingly, a novel caspase-dependent pathway that does not involve the death domain, but BCL-2 has recently been described (Coulson et al., 1999).

While in the death receptor pathway apoptosis is triggered by a relatively small number of structurally-related ligands, mitochondrial apoptosis in neurons can be triggered by a variety of structurally-unrelated agents (Sastry and Rao, 2000). This implies that mitochondrial apoptosis may be induced by more than one single mechanism.

A key event in the mitochondrial pathway is the release of cytochrome c into the cytosol. Experiments in cell-free systems led to hypothesize that cytochrome c release in mitochondrial apoptosis is either caused by a rupture of the outer mitochondrial membrane and/or by the so-called mitochondrial permeability transition (MPT), that is controlled by a voltage- and Ca²⁺-sensitive pore, referred to as the permeability transition (PT) pore (Blatt and Glick, 2001).

Subcellular localization studies have shown that the anti-apoptotic members of the BCL-2 family (BCL-2, BCL-X_L) reside on the mitochondrial outer membrane, while the pro-apoptotic family members (BAD, BAX, BID) may be either cytosolic or present on the cytoplasmic surface of the outer mitochondrial membrane (Zimmermann et al., 2001). During apoptosis these pro-apoptotic molecules are activated and translocate to the mitochondria where they induce the release of cytochrome c (and other proteins) from the intermembrane space. On the other hand, the anti-apoptotic proteins BCL-2 and BCL-X_L work to prevent cytochrome c release from mitochondria, and thereby protect cells from death (Kluck et al., 1997, Yang et al., 1997). Subsequent events involve formation of the apoptosome and caspase activation.

Another protein that is normally located in the intermembrane space of mitochondria is the apoptosis-inducing factor (AIF). AIF is a flavoprotein which shares homology with the bacterial oxidoreductase and, that, similarly to cytochrome c, is a phylogenetically old, bifunctional protein (Susin et al., 1999). Upon death signaling, AIF translocates to the nucleus, binds to DNA and provokes chromatin condensation and large scale (approximately 50k bp) DNA fragmentation, apparently in a caspase-independent manner (Daugas et al., 2000). In keeping, overexpression of BCL-2 prevents the release of AIF from mitochondrion, but not its apoptogenic activity (Susin et al., 1999). Recent data show that AIF is released from mitochondria by a mechanism distinct from that of cytochrome c, but probably mediated by PARP-1 (Yu et al., 2002). Interestingly, in embryonic morphogenesis, genetic inactivation of AIF appears to abolish early neuronal death of proliferating precursor cells and young postmitotic neuroblasts (Joza et al., 2001). In addition, it has been shown that the phenotype of harlequin (hq) mutant mice, which display progressive degeneration of terminally differentiated cerebellar and retinal neurons, is due to a proviral insertion in the aif gene, causing about an 80% reduction in AIF expression (Klein et al., 2002). Mutant CGCs are susceptible to exogenous and endogenous peroxide-mediated apoptosis, but can be rescued by AIF expression. Overexpression of AIF in wild-type neurons further decreases peroxide-mediated cell death, suggesting that AIF serves as a free radical scavenger.

Recently, an additional protein with the dual name Smac/DIABLO, released together with cytochrome c during apoptosis, has been identified (Du et al., 2000, Verhagen et al., 2000). Smac/DIABLO promotes caspase activation by associating with the apoptosome and inhibiting a family of proteins that function as inhibitors of apoptosis (IAPs). In some cellular systems, cytochrome c is necessary but not sufficient for cell death. In these systems Smac/DIABLO may be the second factor required for the so-called competence to die (Deshmukh and Johnson Jr., 1998).

Differentiated neurons are postmitotic cells completely devoid of replicative capability. Most mammalian CNS neurons enter the postmitotic state during embryonic life. In doing so some dividing cells exit from the cell cycle and enter a phase of mitotic quiescence commonly referred to as the G₀ phase. Indeed postmitotic neurons are believed to enter an “extended G₀ phase” which, unlike other cell types, is irreversible.³ Investigation of the mechanisms of neuronal apoptosis has led to the unexpected discovery that, in many instances, the quiescent and dormant cell cycle machinery is “resuscitated”. In this section, we present an overview of recent data suggesting that uncoordinated expression of cell cycle-related molecules is one fundamental mechanism of apoptosis in certain neurons. Readers are invited to consult a number of reviews specifically devoted to this issue for further information (White, 1996, Yoshikawa, 2000, Liu and Greene, 2001).

The cellular and molecular mechanisms of cell cycle regulation and checkpoint in metazoans have been extensively reviewed in the recent past (Rhind and Russell, 2000, Walworth, 2000, Walworth, 2001, Bartek and Lukas, 2001, Bulavin et al., 2002, Melo and Toczyski, 2002).

Three families of proteins are primarily responsible of cell cycle regulation: the cyclins, the cyclin-dependent kinases (CDKs), and the cyclin-dependent kinase inhibitors (CKIs). Cyclins are a group of proteins whose abundance varies substantially during the cell cycle. They associate with their cognate CDKs acting as activating subunits, and eventually leading to active CDK complexes. CDKs allow progression through the different phases of the cell cycle by phosphorylating their target substrates.

For protection from a variety of different types of insults or stress resulting in DNA damage, eukaryotic cells have developed a system of checkpoints that delay progression to the next phase of the cell cycle and activates DNA repair. When DNA damage is irreparable, checkpoints eliminate potentially hazardous cells by permanent cell cycle arrest or cell death (Bartek and Lukas, 2001). During normal cell cycle progression, initiation of mitosis is triggered by a complex process of activation of the cyclin-dependent protein kinase cell division cycle 2 (Cdc2) kinase. Cdc2 kinase is the major “engine” that drives the G₂→M transition. Prior to mitosis the Cdc2–cyclin B1 complex is held in the cytoplasm in an inactive state by Cdc2 phosphorylation at Thr14 and Tyr15. Dephosphorylation of these sites in mammals is regulated by two cell division cycle 25 (Cdc25) phosphatases, Cdc25B and Cdc25C. Cdc25B removes inhibitory phosphates in cytoplasmic Cdc2; with G₂ progression, dephosphorylated Cdc2 translocates into the nucleus where it is a target for Cdc25C, which maintains Cdc2 dephosphorylation at inhibitory sites (Bulavin et al., 2002). In mammals, the checkpoint kinase 1 (Chk1) is the most important molecule that acts upstream Cdc25C in DNA replication and damage checkpoints (Walworth, 2001). Chk1 phosphorylates Cdc25C in vitro at Ser216. Strikingly, dephosphorylation of Cdc25C coincides with mitotic entry.

G₁ phase is a period when cells make critical decisions about their fate, including the option to replicate DNA (in other words entering the S phase) and complete cell division. The decision to enter the S phase is made at the so-called restriction point in mid-to-late G₁ (Bartek et al., 1996). The genes critical for G₁/S transition and coordination of S–G₂–M progression are regulated by the parallel retinoblastoma protein (Rb) and Myc pathways (Walworth, 2000). According to the current two-wave model of the G₁ checkpoint response in mammalian cells, an initial rapid transient p53-independent response is followed by the delayed yet more sustained G₁ arrest imposed by p53 (Bartek and Lukas, 2001).

The retinoblastoma tumor suppressor gene encodes a nuclear protein, Rb, which plays a central role in cell cycle control. Rb and the Rb-related proteins p107 and p130 are among the best-characterized substrates of G₁ CDKs (Weinberg, 1995). Rb is a member of a family of proteins that interact with many transcription factors. The early gene 2 factor (E2F) is a transcription factor that appears to be the major physiological target of Rb.⁴ The Rb family proteins bind to the transactivation domain of E2F, and strongly activate transcription of E2F-responsive genes (Stevaux and Dyson, 2002). Through activation of these genes, E2F is believed to positively regulate cell cycle progression.

Phosphorylation and dephosphorylation of Rb regulates its E2F–protein binding activity. The unphosphorylated or hypophosphorylated active form of Rb predominates in the G₁ phase and binds to E2F, to repress its transcriptional activity.

Transcriptional repression is relieved in late G₁, when Rb family proteins become highly phosphorylated by one or more CDKs (Lee et al., 1997, Zarkowska and Mittnacht, 1997, Lundberg and Weinberg, 1998). Activated E2F triggers transcription of many genes involved in DNA replication and cell growth control.

In more recent years, the view of E2F-dependent transcription has broadened. E2F-regulated genes have a role in DNA synthesis and repair, mitosis, and, directly relevant to the purpose of the present discussion, apoptosis (Stevaux and Dyson, 2002). In particular, E2F1 has a physiological role in DNA-damage responses, perhaps through expression of DNA-repair and pro-apoptotic genes, including Apaf1, casp 3 and casp 7 (Müller et al., 2001).

In the areas of persistent neurogenesis during adulthood, Rb immunoreactivity is high in proliferating neuronal precursors, but reduced during terminal differentiation (Yoshikawa, 2000). A transient increase in the Rb protein level appears to be an important step in the initiation of terminal mitosis of neuronal progenitors, and is then followed by a drastic reduction during terminal differentiation and maintained at low levels in postmitotic neurons (Slack et al., 1998, Callaghan et al., 1999). In keeping, an increasing body of evidence is leading to the notion that postmitotic neurons generally undergo apoptosis when the cell cycle regulators that promote Rb phosphorylation are activated. Moreover, E2F family members can trigger apoptosis, and E2F1-induced apoptosis can be specifically inhibited by Rb (Liu and Greene, 2001).

The Rb-related proteins p107 and p130 appear to be able not only to substitute for many of the Rb functions in growth regulation, but also provide other fundamental functions that may extend beyond those of Rb. The abundance of p130 in differentiated neurons indicates that the growth arrest dependent upon E2F4/p130 interaction might be a key event in the maintenance of the neuronal G₀ state (Persengiev et al., 1999).

Although there are probably multiple sensors that record and transduce DNA damage within mammalian cells, one of the most important is the p53 tumor suppressor protein. p53 plays a critical role as a transducer of damage to genomic integrity into growth arrest and/or apoptosis. DNA damage triggers stabilization and accumulation of p53, which then initiates either a G₁ cell cycle arrest or apoptosis (Fraser et al., 1996). In addition to DNA damage, several other stimuli lead to p53 activation (Blatt and Glick, 2001). In the normal state, cells contain fairly low levels of p53, since the inactive form of the protein is highly unstable. For activation, p53 must be phosphorylated. Although a large number of kinases is capable to phosphorylate p53, response to DNA damage is most likely mediated by DNA-dependent protein kinase (DNA-PK) which is a substrate of certain caspases, the product of the ataxia-teleangectasia gene (ATM), and Chk2 (Lee and McKinnon, 2000, Blatt and Glick, 2001).

When neurons mature in vitro the subcellular localization of p53 changes from the nucleus, in immature cells, to the cytoplasm, in fully differentiated neurons (Eizenberg et al., 1996). It seems that, in response to appropriate stimuli, p53 translocates to the nucleus and plays a regulatory role in directing primary neurons toward differentiation or apoptosis.

Several lines of evidence converge to indicate that neurons undergo apoptosis by p53-dependent or -independent pathways, according to the stimulus responsible for genotoxic stress. However, it is clear that p53 has a significant role in apoptosis that follows DNA damage in vivo (Wood and Youle, 1995). p53 is also a fundamental component of the p75^NTR apoptotic signal cascade (Aloyz et al., 1998).