BCL-2 homologues are major regulators of apoptosis and, as such, play an active role in the survival of adult neurons following injury. In recent years, these proteins have also been associated with the regulation of autophagy, a catabolic process involved in the recycling of nutrients upon starvation. Basal levels of autophagy are also required to eliminate damaged proteins and organelles. This is illustrated by the accumulation of ubiquitin-positive aggregates in cells deficient in autophagy and, in the nervous system, this is associated with progressive cell loss and signs of neurodegeneration. Given the importance of both apoptosis and autophagy for neuronal survival in adult neurons, understanding how BCL-2 homologues co-ordinately regulate these processes will allow a better understanding of the cellular processes leading to neurodegeneration. In the present review, we will discuss the roles of BCL-2 homologues in the regulation of apoptosis and autophagy, focussing on their impact on adult neurons.
Apoptosis is a form of programmed cell death that is required for the proper development and homoeostasis of multicellular organisms. In the nervous system, apoptosis is required during development to remove neurons that fail to properly connect to the growing neuronal network, as demonstrated by the phenotype of knockouts of several apoptotic regulators (reviewed in ). During this period, neurons are destined to die unless they receive proper survival signals from neighbouring cells. However, once the neuronal network is established in the mature nervous system, the remaining neurons need to survive and stay functional for the lifetime of the organism. They thus become much more resistant to apoptosis. This can be achieved through several mechanisms, one of which is the action of BCL-2 homologues, a family of proteins tightly regulating the mitochondrial pathway of apoptosis. Other processes, such as autophagy (literally self-eating) are also emerging as important regulators of neuronal survival and function. Interestingly, these can also be regulated by BCL-2 homologues. In the present review, we will discuss the roles of BCL-2 homologues in the regulation of apoptosis and autophagy, focussing on their impact on adult neurons.
The mitochondrial pathway of apoptosis is initiated when intracellular signals converge on mitochondria to cause the release of cytochrome c from the organelle (Figure 1). Cytochrome c binding to the adaptor protein APAF-1 (apoptotic protease-activating factor-1) then leads to the activation of caspase 9 and, subsequently, caspase 3/7, the proteases responsible for dismantling the cell [2,3]. Members of the BCL-2 family of proteins tightly control this process by regulating the permeabilization of the mitochondrial outer membrane [3,4]. BCL-2 homologues can be divided into three classes according their function and the presence of at least one of four BH (BCL-2 homology) domains. Pro-apoptotic BAX and BAK, containing BH1–3, but no BH4, are required for the release of cytochrome c from mitochondria, whereas anti-apoptotic BCL-2 homologues [BCL-2, BCL-XL, BCL-W, MCL-1 (myeloid cell leukaemia-1) and A1] inhibit this process. BH3-only proteins, the third class of BCL-2 homologues, contain a BH3 domain, but no other functional domain. This BH3 domain mediates the interaction between BH3-only proteins and the other two groups. Although there is still considerable debate as to whether BH3-only proteins directly or indirectly activate BAX and BAK , it is clear that, upon induction of apoptosis, BH3-only proteins are activated (through dephosphorylation, cleavage or transcriptional up-regulation), leading to the inactivation of anti-apoptotic BCL-2 homologues and activation of BAX/BAK and subsequent cytochrome c release (Figure 1) [3,4]. BH3-only proteins are thus the upstream activators of the mitochondrial apoptotic cascade.
Regulation of the mitochondrial pathway of apoptosis by BCL-2 homologues
BCL-2 HOMOLOGUES AND NEURONAL SURVIVAL
The function of BCL-2 homologues in the developing nervous system is well established [1,4]. For example, BCL-XL- and MCL-1-knockout mice have exacerbated apoptosis in the brain during development [6,7], whereas BAX-knockout mice have reduced apoptotic clearance of neurons that fail to make proper connections . In addition, knockout of several downstream mitochondrial effectors (caspase 9, caspase 3 and APAF-1) have all have increased numbers of neurons (reviewed in ). This critical role of BCL-2 homologues during development has led to the suggestion that these proteins play an equally important role in the long-term survival of adult neurons. Early work with the first identified BCL-2 homologues (BCL-2, BCL-XL, BAX and BAK) was based on the hypothesis that changes in the level of pro-apoptotic compared with anti-apoptotic BCL-2 homologues underlined the change in sensitivity to the presence of survival factors that occurs as the neurons mature. The experimental support for this hypothesis is, however, scarce, as a decrease in one anti-apoptotic BCL-2 homologue is usually compensated for by another. For example, mRNA levels of BCL-XL decrease during development in several populations of sensory neurons, but this decrease is compensated for by a concomitant increase in BCL-W mRNA . In retrospect, it is not surprising that the overall levels of anti-apoptotic compared with multi-domain pro-apoptotic BCL-2 homologues do not vary dramatically in this context, as their expression does not directly regulate apoptosis induction. A more plausible hypothesis would therefore be that, as a consequence of the changes in the survival signals that occur upon maturation, BH3-only proteins are not as prone to be activated in adult neurons as in developing neurons. This is indirectly supported by a wealth of findings demonstrating a role for BH3-only proteins in the induction of apoptosis in other systems such as during haemopoiesis (reviewed in ). This question is however still open, as most of the focus of the field has recently been on the pathological activation of apoptosis.
BCL-2 HOMOLOGUES AND NEURODEGENERATIVE DISEASES
Considerable effort has been made towards the understanding of the role of apoptosis and BCL-2 homologues in neurodegenerative diseases. Evidence of apoptosis (caspase 3 activity and BAX up-regulation) has been found in patient samples, as well as in classic animal models of neurodegeneration such as stroke, AD (Alzheimer's disease), HD (Huntington's disease) and PD (Parkinson's disease) (reviewed in [10,11]). Importantly, overexpression of BCL-2 or BCL-XL, or genetic deletion of BAX, resulted in increased neuronal survival in in vivo models of stroke [12,13], AD , PD [15,16] and HD , further suggesting an important role for BCL-2 homologues in the regulation of neurodegenerative diseases. On the other hand, the precise role of BH3-only proteins in the aetiology of neurodegenerative diseases remains poorly understood, although recent studies have suggested an important role for the p53 tumour suppressor. Work in the context of tumorigenesis has delineated the mechanisms through which p53 regulates apoptosis in response to genotoxic stress [18,19]. This occurs at least in part through the up-regulation of several BH3-only proteins [PUMA (p53 up-regulated modulator of apoptosis), Noxa and human BIK] [19,20]. The existence of similar pathways has been uncovered in neurons where p53 plays a role in several types of injury . For example, Puma has been shown to kill damaged neurons downstream of p53 [22,23]; however, despite these recent advances, the extent to which apoptosis is responsible for neuronal cell death following injury and the exact contribution of BCL-2 homologues is not clear. One reason for this is that genetic deletion of most apoptosis-related genes involved in neuronal cell death are embryonic lethal. For example, knockouts for BCL-XL and MCL-1, the two main anti-apoptotic BCL-2 homologues in the cortex, do not survive past E13 and E17 (where E is embryonic day) respectively [6,7]. This has rendered the study of these genes in adult brain more challenging. In addition, neuronal cell death occurring in neurodegenerative diseases can often present characteristics of several forms of cell death, including apoptosis, necrosis and autophagy (although the latter is also a survival mechanism; see below).
The presence of such a variety of morphologically distinct types of cell death underscores the range of cellular processes, including apoptosis, contributing to both survival and death of neurons. It could therefore be postulated that preventing cell death itself is probably not sufficient to ensure that neurons remain both alive and functional over the lifetime of the organism. This is of particular importance given the limited regeneration capacity of the brain and lifetime of neurons. Neurons thus need highly efficient ways to eliminate damaged proteins and organelles, properly sort protein, and sustain their metabolic needs. Part of this is achieved through the activity of molecular chaperones and the UPS (ubiquitin–proteasome system). Autophagy is a third quality control mechanism playing an important role in the turnover of protein aggregates and dysfunctional organelles in the brain.
Autophagy is a catabolic process in which proteins, portions of the cytoplasm or whole organelles are delivered to lysosomes for degradation [3,24,25]. Autophagy is induced under nutrient starvation conditions, where it allows recycling of nutrients necessary for cell survival [3,25]. In addition, a basal constitutive level of autophagy controls normal turnover of organelles and cytosolic components . Beside its well-characterized starvation-induced role in energy homoeostasis and intracellular recycling, autophagy has been linked to several cellular processes promoting cellular repair and survival . This includes cellular remodelling associated with differentiation during development, organelle turnover and lipid metabolism [26,27]. The quality control function of autophagy, including removal of protein aggregates and damaged organelles, as well as the maintenance of genomic integrity, has received a lot of attention as it is thought to play a major role in cancer progression and neurodegenerative diseases [3,24]. Of note, autophagy has also been associated with cell death under some circumstances .
There are three types of autophagy: microautophagy, CMA (chaperone-mediated autophagy) and macroautophagy, which differ in the mechanism used to deliver cargo to the lysosomes for degradation. Microautophagy will not be addressed here. In CMA, protein substrates are recognized by the HSC70 (heat-shock cognate 70 stress protein) chaperone and directly delivered to lysosomes by interacting with LAMP2a (lysosome-associated membrane protein 2a) . In the case of macroautophagy (hereafter referred to as autophagy and the main form of autophagy discussed), organelles and portions of the cytoplasm are engulfed in a double membrane organelle called the autophagosome, which then fuses with a lysosome for bulk degradation (Figure 2) [3,25]. Autophagosome formation is regulated by a set of ATG (autophagy-related) genes in a pathway that is conserved from yeast to mammals. The nucleation of the autophagosome is initiated by the activation of the class III PI3K (phosphoinositide 3-kinase) Vps34 in complex with Beclin1 (ATG6 in yeast) and other regulatory proteins. Subsequent elongation of the autophagic membrane and incorporation of the cytoplasmic content requires two ubiquitin-like conjugation systems. One involves the covalent conjugation of ATG12 to ATG5, whereas the other requires the addition of a PE (phosphatidylethanolamine) group to LC3 (one of the mammalian homologues of yeast ATG8), leading to its association with the autophagosome [3,25].
Autophagy pathway in mammalian cells
AUTOPHAGY AND NEURODEGENERATIVE DISEASES
Of the ATG genes, several are also tumour suppressors (for example, Beclin1  and Bif-1 ), despite the fact that autophagy can provide a survival advantage to cells in metabolically stressed hypoxic regions of the tumour . These opposite effects underscore the function of autophagy in both the generation of nutrients and as a quality control mechanism. This latter activity can explain, at least in part, the tumour-suppressor role of autophagy as it promotes decreased oxidative stress and lower occurrence of tumour-promoting genetic mutations . As mentioned previously, proper quality control mechanisms are also crucial for long-term survival of neurons. Indeed, one of the hallmarks of neurodegenerative diseases is the accumulation of protein aggregates. Although each disease has been characterized by the presence of a specific protein within these aggregates (α-synuclein in PD, β-amyloid plaques and Tau in AD, and mutant Huntingtin in HD), the common underlying mechanism probably involves changes in the UPS . Interestingly, a functional involvement of autophagy in the control of neurodegeneration was highlighted by studies in mice lacking an autophagic response, owing to the genetic deletion of important ATG genes (ATG5 and ATG7) specifically in the brain [33,34]. These mice had a progressive accumulation of ubiquitin-containing aggregates, developed progressive deficits in motor function accompanied by neurodegeneration and died within a few months, in a process reminiscent of neurodegenerative disease. Further support for a link between protein aggregates and autophagy comes from studies examining the clearance of disease-related protein aggregates. Degradation of α-synuclein has been shown to be dependent on both CMA and macroautophagy, whereas clearance of mutant Huntingtin and β-amyloid plaques is reduced when Beclin1 protein levels are decreased [35–37]. Interestingly, the same studies found lower Beclin1 levels in affected human samples compared with control brains [35,36]. Furthermore, induction of autophagy through Beclin1 overexpression or rapamycin treatment reduced aggregate formation, supporting further a role for autophagy in the regulation of protein aggregate formation [35,36].
Several pathways could lead to activation of autophagy following accumulation of misfolded or aggregated proteins. Accumulation of misfolded proteins can overcome the clearance capacity of the proteasome and lead to inhibition of the ERAD [ER (endoplasmic reticulum)-associated degradation] pathway, an ER quality control mechanism that relies on the proteasome to degrade ER-associated misfolded proteins (reviewed in [38,39]). ERAD inhibition causes ER stress and activation of the UPR (unfolded protein response), leading to decreased protein synthesis, up-regulation of ER chaperones and activation of autophagy. Sustained UPR activation also induces apoptosis. The relevance to neurodegenerative diseases is highlighted by the observation that both α-synuclein and mutant huntingtin activate ER-associated autophagy. A second pathway activating autophagy in response to misfolded proteins could involve the p53 family member p73. Deletion of one copy of p73 in mice has been shown to induce phospho-tau accumulation and to lead to neurodegeneration during aging in an AD mouse model . Interestingly, p73 can activate autophagy , suggesting that these two functions could be related. Additional work is required to directly address the relevance of p73 in the induction of autophagy and its relationship with neurodegenerative diseases.
At the molecular level, the fact that the protein aggregates found in both autophagy-deficient mice and patient samples contain ubiquitin suggests the presence of some level of cross-talk between the UPS and autophagy. Indeed, the ubiquitin-binding proteins p62/SQSTM1 (sequestosome 1) and NBR1 (neighbour of BRCA1 gene 1) also interact with LC3, leading to their inclusion in autophagosomes along with their ubiquitinated cargo [42,43]. Knockout of p62 can prevent the accumulation of protein aggregates in autophagy-deficient mice, suggesting a functional role for p62 in mediating the lysosomal degradation of protein aggregates . However, the relationship between the UPS and autophagy might not be that simple. A recent study has suggested that inhibition of autophagy caused the accumulation of protein aggregates indirectly, through the stabilization of p62 . The accumulation of p62 did not affect proteosomal activity, but impaired delivery of ubiquitinated substrates to the 26S proteasome, leading to the formation of p62-containing aggregates. This is similar to the negative effect of overexpressed or mutant α-synuclein on the degradation of MEF2D (myocyte enhancer factor 2D), a transcription factor required for neuronal survival . Nuclear MEF2D is normally exported to the cytosol where it is degraded in a CMA-dependent fashion. Overexpression of α-synuclein blocks this degradation, leading to the accumulation of non-functional cytosolic MEF2D and decreased cell survival. One caveat of this observation, however, is that it is still not clear whether protein aggregate accumulation is a cause or a consequence of neurodegenerative disease.
Irrespective of the actual consequences of autophagy modulation on the appearance of protein aggregates, other roles of autophagy probably play a key role in the maintenance of cellular homoeostasis. One example is the recycling of damaged organelles, especially mitochondria. This is most evident in the context of PD, where mitochondrial damage lies at the heart of the cellular dysfunction leading to the loss of dopaminergic neurons . Indeed, exposure to mitochondrial complex I inhibitors such as rotenone or MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) are linked to PD, whereas several PD-associated genes code for proteins affecting mitochondrial function. For example, mitochondrial DJ-1 (PARK7) plays an antioxidant role, whereas Pink1 (PARK6) is a kinase anchored on the mitochondrial outer membrane that has been proposed to modulate the activity of Parkin (PARK2), another PD-related protein partially localized to mitochondria . Interestingly, both DJ-1 and Parkin play a role in autophagy [46,47]. In the case of Parkin, it has been shown to mediate the autophagic degradation of damaged mitochondria, a process commonly referred to as mitophagy . A picture is thus emerging, at least in the case of PD, in which failure to properly dispose of damaged mitochondria through autophagy could lead to oxidative stress, cellular damage and subsequent cell death.
BCL-2 HOMOLOGUES AND AUTOPHAGY
In light of the preceding discussion, it could be postulated that defects in mitochondrial function, UPS or autophagy will negatively affect each of these processes, leading to the accumulation of damaged organelles and proteins, impairing cellular functions further. One twist to the story was the discovery that, in addition to their well-characterized roles in apoptosis, BCL-2 homologues regulate autophagy [48,49]. BCL-2 and BCL-XL bind directly to a BH3 domain in Beclin1, inhibiting Beclin1 and consequently autophagy . BH3-only proteins or phosphorylation of BCL-2 by JNK1 (c-Jun N-terminal kinase 1) can disrupt this interaction, thereby inducing autophagy [48,50]. On the basis of the fact that BCL-2 opposes the activity of BAX and BAK, it could be hypothesized that BAX and BAK promote autophagy. This is supported by the fact that BAX interacts with Bif-1/Endophilin B1, a protein that has been implicated in BAX activation, but also mitochondrial fission and autophagy . However, an increase in autophagy was reported in the absence of BAX and BAK following pro-apoptotic treatments . These studies indicated that cells which fail to die by apoptosis, owing to the absence of BAX and BAK, survive by inducing autophagy and thus are not formally required for autophagy to occur. On the other hand, whether BAX and BAK play a role in the activation of autophagy upon nutrient starvation remains to be determined. Similarly, very little is known about the involvement of BH3-only proteins in this process. In theory, BH3-only proteins would be expected to promote autophagy since they can displace binding of Beclin1 to BCL-2. This could prove difficult to demonstrate, however, because of their potent apoptosis-inducing properties. Nevertheless, a few so-called sensitizing BH3-only proteins (BAD and BIK) have been shown to promote autophagy in some circumstances [48,52]. A similar caveat probably applies to the study of autophagy in knockouts of anti-apoptotic BCL-2 family members. As haemopoietic cells and the developing brain, two systems which exhibit high rates of apoptosis, are the most studied organs in these knockout animals, it is likely that any effect on autophagy would be masked by the apoptotic phenotype. We were recently able to circumvent this limitation by knocking out anti-apoptotic MCL-1 in post-mitotic neurons using the CaMKIIα (Ca2+/calmodulin-dependent protein kinase IIα) promoter to drive Cre expression (M. Germain and R.S. Slack, unpublished work). These animals had a massive increase in autophagy, but only limited apoptosis, in cortical neurons within 1 week of Cre expression, demonstrating in vivo that MCL-1 regulates autophagy.
How can anti-apoptotic BCL-2 homologues negatively regulate both a survival (autophagy) and a pro-death (apoptosis) mechanism? In the case of tumour cells, overexpression of BCL-2 can promote both survival (anti-apoptotic function) and increased genomic instability (by blocking autophagy), stimulating tumour growth . It is, however, harder to reconcile these two functions in terminally differentiated cells such as neurons, although there are precedents for such a dichotomy. For example, the p53 tumour suppressor can positively regulate both apoptosis and survival mechanisms, including the activation of several metabolic pathways and autophagy, the final outcome depending on both the cellular context and the extent of the damage [18,19]. In the case of BCL-2 family proteins, the range of mechanisms regulating their function in response to stress (transcriptional activation, phosphorylation and UPS-mediated degradation [53,54]) and the fact that they can participate in several survival/death functions (apoptosis, autophagy and calcium signalling) suggest some similarities to p53.
Two scenarios could thus be envisioned. First, different autophagic signals could be regulated differently by BCL-2 homologues. For example, although overexpression of BCL-2 can inhibit autophagy induced by acute starvation [48,49], it does not prevent autophagy induced by growth factor removal . Secondly, the effect of BCL-2 homologues on autophagy could be proportional to the intensity of the signal. In this scenario, overactivation or sustained activation of autophagy that could potentially result in cell death would be blocked or dampened by BCL-2 expression, whereas basal or transient autophagy would not be affected (Figure 3). One variation on this idea is that BCL-2 homologues could serve as a rheostat to set a baseline level of autophagy, thereby allowing the quality control roles of autophagy to be carried out while preventing its potentially deleterious effects. Several recent findings support this idea. First, autophagy has been suggested to be pro-apoptotic in conditions where it is sustained . Secondly, decreasing BCL-2 or MCL-1 levels increases the amount of autophagy in cell lines in the presence of nutrients (, and M. Germain and R.S. Slack, unpublished work). A similar increase in autophagy was observed in the brain of MCL-1-conditional-knockout mice (M. Germain and R.S. Slack, unpublished work). We thus propose that the endogenous levels of anti-apoptotic BCL-2 homologues promote survival of long-lived differentiated cells such as adult neurons by blocking apoptosis, while allowing proper quality control processes to occur. Disruption of this balance by overexpressing BCL-2 would, however, have dire consequences, as it would promote both survival and increased genomic instability through the inhibition of autophagy, two steps required for cancer progression.
Model for the regulation of autophagy by BCL-2 homologues
Although the exact contribution of BCL-2 homologues to autophagy regulation and its relationship to neurodegenerative diseases remain elusive, the realization that these proteins regulate various processes is a step towards the understanding of apoptosis as integrated within a more general cellular context in terminally differentiated cells. In this context, it might be worthwhile revisiting the protection afforded by BCL-2 overexpression or BAX deletion on models of neurodegenerative diseases to assess whether pathways other than apoptosis are also affected.
M.G. is the recipient of a research fellowship from the Heart and Stroke Foundation of Canada. The authors' work is supported by grants from the Heart and Stroke Foundation of Canada.
We would like to thank the Life Side for providing the title, as well as the Death Side (especially Melissa Kelly) for comments and helpful suggestions in the preparation of this manuscript.
apoptotic protease-activating factor-1
- BH domain
BCL-2 homologue domain
myeloid cell leukaemia-1
myocyte enhancer factor 2D
p53 up-regulated modulator of apoptosis
unfolded protein response