The Cellular Pathways of Neuronal Autophagy and Their Implication in Neurodegenerative Diseases

Abstract

Autophagy is a tightly regulated cell self-eating process. It has been shown to be associated with various neuropathological conditions and therefore, traditionally known as a stress-induced process. Recent studies, however, reveal that autophagy is constitutively active in healthy neurons. Neurons are highly specialized, post-mitotic cells that are typically composed of a soma (cell body), a dendritic tree, and an axon. Despite the vast growth of our current knowledge of autophagy, the detailed process in such a highly differentiated cell type remains elusive. Current evidence strongly suggests that autophagy is uniquely regulated in neurons and is also highly adapted to local physiology in the axons. In addition, the molecular mechanism for basal autophagy in neurons may be significantly divergent from “classical” induced autophagy. A considerable number of studies has increasingly shown an important role for autophagy in neurodegenerative diseases and has explored autophagy as a potential drug target. Thus, understanding the neuronal autophagy process will ultimately aid in drug target identification and rational design of drug screening to combat neurodegenerative diseases.

1. Introduction

Autophagy is a conserved lysosomal degradation pathway. In mammals, three types of autophagy have been described: macroautophagy, microautophagy and chaperone-mediated autophagy (CMA). These three types of autophagy differ in their mode of delivery of their substrates to lysosomes for degradation [1]. While little is known about the microautophagy process, a large body of studies has contributed to the current understanding of the macroautophagy and CMA pathways. Macroautophagy is the prototype of autophagy involving formation, delivery, and degradation of autophagic vacuoles (also called autophagosomes) through lysosomes, and will be the only form discussed in this review (hereafter referred to as autophagy). Although autophagy occurs in virtually all cell types, and likely involves highly conserved molecular machinery, emerging evidence suggests cell type/tissue-specific regulation of autophagy.

Neurons were one of the few cell types that were used in the initial identification and characterization of autophagy. With early access to electron microscopy (EM) in the last century, Alex Novikoff, Christian De Duve, and colleagues discovered and described the cell “self-eating” process of autophagy in the form of distinct vacuoles through ultrastructural analysis [2, 3]. The formation of the autophagosomes, which engulfed a portion of cytoplasm and occurred in a large number, especially following axotomy and excitotoxic insult to neurons, is associated with “chromatolysis”, a phenomenon that describes the area in neuronal cytoplasm that is devoid of organelles and filled with various types of vesicles [4]. These initial observations of neuronal autophagic activity were followed by a series of EM studies that revealed the accumulation of autophagosomes in the neurons of several human neurodegenerative diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD) and Huntington’s disease (HD). With little knowledge of autophagic process and regulation, especially in mammalian tissues, neuronal autophagy, marked by elevated levels of autophagosomes, was viewed traditionally as a cellular mechanism that was highly destructive, and therefore was suspected to be a driving force in cell death [5].

The arrival of the molecular era of autophagy study provided important evidence showing that autophagy is a primary stress response for cell survival. For example, loss-of-function studies demonstrate the essential role for autophagy-associated genes in the removal of “obsolete” proteins and organelles, thus protecting cell or neuron survival [6, 7]. Therefore, in injured neurons or neurons bearing disease-related genes, altered autophagy, associated with increased numbers of autophagosomes, can be viewed as a beneficial response of neurons in repairing or remodeling damaged cellular components necessary for sustaining normal neuronal function and survival. However, other studies also provide genetic and cellular evidence that otherwise argues for a role of autophagy in promoting neuronal death, especially in neurons with acute injury [8].

More recent studies have begun to dissect the autophagic process in neurons under various stress or pathological conditions. These studies suggest that, while accumulation of autophagosomes can arise from increased production in some cases [9], it can be caused by a mechanism that blocks fusion and degradation of autophagosomes through lysosomes in other scenarios [10]. Given the current studies exploring the potential of autophagy as a drug target for the treatment of neurodegenerative diseases, a thorough investigation of the detailed mechanism whereby autophagy participates in each disease condition, would be critical for designing therapeutic intervention and evaluating the efficacy of an “autophagy” drug [11].

Over the past several years, characterization of autophagy in mammalian cells and animal models has greatly advanced our knowledge of the autophagy process in mammals. However, despite the recent effort and exponential growth in autophagy research as a whole, the progress in understanding the basic process and regulation of neuronal autophagy remains relatively slow. Recent studies have revealed the connections between autophagy and major neurological disorders such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and Huntington’s disease (HD) [12-14]. A common theme emerging in those studies is the role of autophagy in degrading disease-related, aggregate-prone mutant proteins such as tau, huntingtin and alpha-synuclein [15]. In addition, specific pathogenic mechanisms of AD, PD, or HD may profoundly alter autophagic activity. For example, while “inappropriate” autophagic induction may contribute to the increased synthesis of β-amyloid (Aβ) [16] blocked autophagic clearance is also implicated in cytotoxicity in AD [10]. Recently, genetic animal models containing reduced autophagic activity were used to examine the role of autophagy in the pathogenesis of AD, HD, and PD. These studies provided important evidence linking dysfunctional autophagy to the specific disease process [17, 18], Friedman and Yue, unpublished). Due to numerous recent reviews on the study of autophagy in PD, AD, or HD [12-14, 19-21], this review will instead summarize recent research in understanding the basis of neuronal autophagy, especially in primary neuronal cultures and the nervous system of animal models, and autophagic activity associated with other types of neuropathological conditions. Although limited and sometimes conflicting in their current forms, these studies nonetheless begin to shed light on specific cellular pathways and the connection of the physiological function of autophagy to disease mechanism.

2. Biosynthesis of autophagosomes is conserved from yeast to human

Morphological evidence for autophagy was first reported in the 1960’s [3], but the underlying molecular mechanisms were not elucidated for another three decades. In the early 1990s, genetic screens of yeast mutants identified a number of autophagy-related (ATG) genes essential for the autophagic molecular machinery [22, 23]. Currently, 31 autophagy genes are known, many of which are required for autophagosome formation, at the nucleation, elongation, and/or fusion steps. Upon induction, an isolation membrane or phagophore forms and elongates, enveloping a portion of the cytosol, and encloses to form a double-membrane vacuole. Its outer membrane subsequently fuses with the lysosome, where its contents, together with the inner membrane, are digested by acidic hydrolases within the lysosomes (Figure 1). Up to now, at least 14 mammalian homologues of yeast ATG genes have been identified, and characterization of their functions suggests that the autophagic machinery is highly conserved in mammals [24].

Figure 1.

A schematic representation of the macroautophagy process. Following induction, a phagophore or isolation membrane is formed. This process is regulated by Atg1 (Ulk1), Atg9, and the PtdIns 3-kinase complex, which includes Beclin 1, Atg14L, and UVRAG (nucleation). Two ubiquitin-like conjugation systems, which produce Atg8-PE (LC3II) and Atg5-Atg12, mediate the elongation of the isolation membrane, closure, and the formation a double-membrane vacuole known as the autophagosome. Autophagosomes can undergo maturation and fusion with early and late endosomes and MVBs, to generate the amphisome, followed by fusion with lysosomes, to form the autolysosome. Trafficking, maturation, and fusion events are mediated by microtubules and specific SNARE and Rab proteins. ESCRT proteins are also essential for MVB fusion with autophagosomes. Alternatively, immature autophagosomes can fuse directly with lysosomes.

Studies in yeast revealed that, unlike endosomes and secretory vesicles, autophagosome formation does not require budding from existing organelles as its membrane source. Rather, autophagosomes may form from de novo membrane cisternae in the pre-autophagosomal structure (PAS), which contains several Atg protein complexes, and resides adjacent to the yeast vacuole [25-27]. It was shown that Atg9, the only known integral membrane protein associated with autophagic membranes, shuttles between a peripheral site in the cytoplasm and the PAS, therefore regulating the delivery of membrane to the PAS for expansion [27, 28]. The Atg1 kinase complex, which contains regulatory subunit, Atg13, is involved in the retrieval of Atg9 from the PAS and is required for autophagosome formation [28, 29]. Interestingly, the mammalian homologue of Atg9 (mAtg9) was shown to cycle between the trans-Golgi network (TGN) and late endosomes, which may serve as sources for membrane elongation. Starvation or rapamycin treatment induced redistribution of mAtg9 from TGN to peripheral late endosome membranes. Knock-down of Atg1 human homolog, ULK1, prevented this starvation-induced redistribution, suggesting that mAtg9 trafficking is ULK1-dependent [30].

In addition to the recruitment of membranes, Atg9 may also play a role in assembling protein complexes at the PAS [31]. Vps34, a class III phosphatidylinositol (PtdIns) 3-kinase mediates vesicular trafficking through its interactions with Vps15 and Atg6, and forms two distinct complexes: one with Atg14, which regulates autophagy-specific function, and the other with Vps38, for endosome-to-Golgi trafficking [32]. The PtdIns 3-kinase complex produces phosphatidylinositol 3-phosphate (PtdIns(3)P) which may recruit effector proteins, such as the Atg18-Atg2 complex, and together, both complexes are essential for nucleation [25]. The mammalian homologues of Vps34, Vps15 and Atg6 are hVps34, p150 and Beclin 1, respectively. Recent studies have found additional proteins that interact with the hVps34-p150-Beclin 1 complex: UVRAG [33], Atg14L (putative yeast Atg14 homologue) [34] and Rubicon [35]. These studies suggest the existence of multiple hVps34-Beclin 1 kinase complexes, which are involved in specific membrane trafficking mechanisms, including autophagy [35].

Two ubiquitin-like conjugation systems mediate autophagic membrane elongation; one involving the conjugation of Atg5 with Atg12 [36] and the other involving the covalent linkage of Atg8 with phosphatidylethanolamine (PE) [37]. Both Atg12 and Atg8 modifications share a single E1-like activating enzyme, Atg7, but are processed by two separate E2-like conjugating enzymes; Atg10 and Atg3 respectively [38]. The two ubiquitin-like conjugation systems are highly conserved from yeast to mammals [36, 39]. The majority of Atg5 and Atg12 exists in the conjugated form and interacts noncovalently with multimeric protein, Atg16, in yeast and its functional counterpart, Atg16L, in mammals [40]. In one of the first studies linking the small GTPase Rab family to Atg proteins, Atg16L was shown to directly interact with Golgi-resident Rab33 and modulates autophagosome formation [41]. The Atg5-Atg12·Atg16 complex is essential for autophagosome formation and facilitates Atg8 conjugation with PE [40] through E3-like activity of Atg5-Atg12 for Atg8 conjugation [42].

Microtubule-associated protein 1 light chain 3 (LC3) is a mammalian homolog of yeast Atg8 and is cleaved at its C-terminal region by cysteine protease Atg4. This processed form (LC3-I) resides in the cytoplasm until it undergoes two ubiquitin-like modifications to become covalently linked to PE. Both the lipidated form LC3 (termed LC3-II) and Atg12-Atg5·Atg16 are recruited to the isolation membrane [38]. Whereas Atg12-Atg5·Atg16 dissociates upon autophagosome completion, LC3-II remains coupled to the autophagic membrane until fusion with the lysosome [39]. Prior to this fusion event, autophagosomes undergo a maturation process involving fusion with early and late endosomes, including multivesicular bodies (MVB), to form amphisomes [43]. Subsequently, these vacuolar bodies can then fuse with lysosomes to form the autolysosome (Figure 1). These fusion events are mediated by SNARE proteins and Rab proteins, specifically Rab7, which is essential for maturation [44, 45].

3. Neuronal autophagy: process and function in health and disease

3.1 Basal level autophagy

3.1.1 Unique feature of neuronal autophagy: scarcity of autophagosomes

Previous studies using mice expressing an autophagosome marker, green fluorescent protein-tagged LC3 (GFP-LC3), suggested that autophagy is distinctly regulated in different tissues [46]. For example, food limitation triggers a rapid upregulation of autophagy in liver and heart as indicated by the formation of a large number of GFP-LC3-associated autophagosomes, whereas it fails to induce GFP-LC3-labeled autophagosomes in the CNS, despite the strong expression of GFP-LC3 in many types of neurons. In addition, LC3 exists predominantly in a soluble form (LC3-I) in healthy neurons, and levels of the lipidated form (LC3-II) do not change after starvation [46]. Recently, we have investigated the localization of GFP-Atg5, a green fluorescent Atg5 fusion protein used for monitoring isolation membrane, in CNS neurons of transgenic mice [47], and the results show that like GFP-LC3, GFP-Atg5 is largely diffuse regardless of whether food is available or withdrawn (Yue, unpublished data). Indeed, ultrastructural analysis also confirmed scarcity of double-membrane vacuoles that are characteristic of autophagosomes in healthy neurons [48]. In contrast, many tissues, including liver and heart, have many GFP-LC3 puncta even when food is not limited, thus providing direct evidence for constitutively active autophagy under normal conditions.

The above evidence suggests that neurons normally maintain low levels of autophagosomes, and perhaps a low rate of autophagosome biosynthesis, even with the fluctuation of available nutrients. The current hypothesis is that neurons are prohibited from large-scale autophagy induction in response to starvation due to their ability to utilize multiple energy sources to maintain normal neuronal function. For example, glial cells provide nutrient and neurotrophin support to neurons and peripheral organs supply necessary nutrients to the CNS under hypothalamic regulation [49]. Thus, it is possible that even after 48 hours of fasting in mice, it is insufficient to cause nutrient or trophic factor depletion in the CNS neurons due to compensatory mechanisms of nutrient supply. Interestingly, without showing the availability of nutrient or trophic factors, a recent study reported that mTOR activity (an important regulator of autophagy) was significantly reduced, at least in hypothalamic neurons, after 48 hours of starvation in mice [50]. Despite the reduced mTOR activity, there was no report of the formation of GFP-LC3 or GFP-Atg5 puncta in hypothalamic neurons (Yue unpublished data), [46]. While the mTOR activity in hypothalamic neurons of GFP-LC3 or GFP-Atg5 transgenic mice after 48 hour-starvation remains to be determined, other studies observed mTOR-independent regulation of autophagy in neurons [51, 52]. Importantly, Young et al., showed that insulin plays a critical role in suppressing the induction of autophagy in primary neuron cultures. Through screening culture medium components, they found that insulin (but not other components, including amino acids) was the key factor and that the absence of insulin induced autophagy in primary neurons, in an mTOR-dependent manner. Surprisingly, a highly potent Akt inhibitor also efficiently induces autophagy, however, via an mTOR-independent pathway [51] This result suggests that the active insulin signaling is responsible for the low level of autophagosome formation in healthy neurons, and therefore, the insulin pathway may provide a critical mechanism for controlling basal autophagy in neurons. Although the autophagy control mediated by insulin signaling is not restricted to neurons [53], it can be adapted to the neuronal physiology that is related to the high consumption of energy via distinct sources. In addition, multiple parallel signaling pathways, including the mTOR pathway, exist in neurons to regulate autophagy.

Apart from the insulin pathway, specific proteins or protein modifications expressed in neurons may also contribute to additional regulation of basal autophagy in neurons. Previously, we showed that expression of a microtubule associated protein 1B (MAP1B), which is enriched in neurons and binds LC3 with high affinity, remarkably affects the formation of GFP-LC3 puncta (autophagosomes), and this effect is dependent on the status of certain modifications of MAP1B [54], see section 10 of this review).

Taken together, we propose that the regulation of basal autophagy in CNS neurons at least involves the following mechanisms: (1) a non-cell-autonomous mechanism whereby nutrients, hormones (insulin), or neurotrophic factors are supplied by peripheral organs or glial cells, (2) a cell-autonomous control mechanism by intrinsic nutrient-mediated signaling or specific factors expressed in neurons. Furthermore, current evidence also raises the question as to what context or extent does mTOR-mediated signaling regulates autophagy in neurons.

While the hypothesis for the low level of basal autophagy (in the form of “classical” autophagosomes) and its regulation remains to be tested further, a recent study provides evidence suggesting an alternative hypothesis that autophagosomes, once formed, are rapidly cleared by fusion with lysosomes in primary neuronal culture. It further proposes that the clearance of autophagosomes by basal autophagy is highly efficient in neurons and accounts for the low levels of autophagosomes [10]. Apparently, future experiments focused on directly measuring or imaging the dynamics of autophagy flux will be needed to solve this issue.

3.1.2 Essential role of basal autophagy in cellular homeostasis and neuroprotection

Current evidence suggests that neuronal autophagy is tightly regulated and autophagosomes are maintained at minimal level. However, recent studies in mutant mice with targeted deletion of Atg5 or Atg7 specifically in brain, unequivocally demonstrate the importance of basal autophagy in CNS neurons [6, 7]. Unlike Atg5- or Atg7-deficient yeast or mammalian MEF cells that survive and grow normally with regular culture conditions, CNS neurons lacking Atg5 or Atg7 in animal models undergo progressive degeneration, even when mice are housed under normal conditions. Importantly, Atg5 or Atg7 deficient neurons develop a massive number of inclusion bodies that are labeled by ubiquitin, suggesting that continuous autophagy is required to prevent the build-up of intracytoplasmic protein aggregation. These studies provide undeniable proof for the existence and critical function of basal autophagy in neurons. Paradoxically, the evidence that autophagy is in continuous action through autophagosome-lysosome degradation is at odds with the scarcity of autophagosomes or low level of basal autophagy in the healthy neurons. Although it appears to be consistent with the hypothesis of highly efficient basal autophagy[10], an alternative hypothesis is that basal autophagy in neurons proceeds via a distinctive mechanism (e.g. in the absence of “classical” autophagosomes formation). Even though autophagy genes (ATG system) are required, they participate in the lysosomal degradation pathway that differs from canonical autophagy (in the presence of well-documented autophagosome formation) [55]. One well-known example for such an autophagy-related cellular process, is the yeast Cvt (cytoplasm-to-vacuole targeting) pathway, which shares most of its protein components with autophagy, but is biosynthetic and functions to specifically deliver certain enzymes to vacuoles [56]. Future studies should investigate the possibility of non-canonical or adapted autophagy in neurons.

Although CNS neurons are generally vulnerable to the loss of the autophagy pathway, the degree of vulnerability and the formation of intracellular inclusions vary significantly among different neuron types as observed in the mutant brains deficient in Atg5 or Atg7. The ablation of the Atg5 or Atg7 gene may trigger a cell-type specific cellular response to autophagy deficiency and/or a cell type-dependent mechanism contributing to the neurotoxicity. For example, Purkinje cells deficient in Atg5 or Atg7 display very few ubiquitin-associated inclusions, whereas these cells are among the most vulnerable neuron types [57, 58]. In contrast, a large number of ubiquitin-associated inclusions are seen in the brain region where neuronal loss was hardly detected when autophagy was genetically inhibited (Waguri S, Komatsu M, unpublished data). While it is possible that the compensatory pathways to the loss of autophagy vary in different neuronal types and account for the difference in their vulnerability, it may also reflect the disparity in their intrinsic demands for autophagy, as well as relative levels of basal autophagy. Neurons exposed to high levels of various stresses may contain high levels of basal autophagy and are therefore, far more vulnerable, whereas the neurons that require low levels of basal autophagy through their lives are relatively resistant to autophagy deficiency.

3.2 Induced autophagy in neurons: signaling and distinctive stress response

The primary function of autophagy is to provide cellular response to nutrient limitation by mobilization of autophagosome-lysosome degradation pathway. A large body of evidence shows that nutrient-related signaling pathways regulate autophagic activity in mammals. These pathways include insulin and amino acid pathways, mTOR kinase complex [59-61], AMP-activated protein kinase (AMPK) [62] and Beclin 1/Vps34 (class III PI-3 kinase, PItdIns3K) lipid kinase complex [63]. Despite the growing evidence linking these signaling pathways to autophagy regulation, the details for how they control the complex process of autophagy is lacking. Furthermore, little is known about tissue/cell type-dependent regulation. A recent study in primary cortical neuronal culture shows that insulin plays a critical role in controlling the induction of autophagy [51]. A unique feature of neurons, with respect to their fuel supply, is that they use glucose (or ketones) almost exclusively as a blood-borne energy substrate to provide energy and carbon chains for protein synthesis. Therefore, compared to other tissues/cell types, CNS neurons may depend less on autophagy to provide free amino acids or energy under physiological condition. It can be further speculated that the primary function of neuronal autophagy at the basal level is different than as a primary nutrient-starvation response [49]. Accordingly, neurons are likely to use a distinctive mechanism for autophagy regulation. Adding to the complexity of autophagy regulation in neurons, a recent study shows that cultured neurons from male rats more readily undergo autophagy in response to 24 hours of nutrient deprivation compared to female neurons [64], indicating gender differences in autophagic capacity of neurons.

While nutrient-related pathways are likely the conserved mechanism for the control of neuronal autophagy [10, 51], the direct in vivo evidence that these pathways contribute to the induction of autophagy (e.g. synthesis of a large number of autophagosomes) in CNS neurons remains to be shown. Despite the lack of in vivo evidence of neuronal autophagy mediated by nutrient signaling, recent studies in animal models indicate that, in contrast to nutrient starvation which rarely induces autophagy, a variety of stress-related signals, neuron injuries, and neuropathogenic pathways cause rapid formation and accumulation of autophagosomes in neurons. As a result, the intracytoplasmic area occupied by autophagic activity is sharply increased and the capacity of autophagic degradation could be maximally expanded. During the process of autophagy induction, neurons may undergo a significant change in autophagy regulation, involving a deregulation process that allows neurons to transition from basal level (neuron-specific process) to the activated state (well-conserved and known as induced autophagy) involving large-scale biosynthesis of autophagosomes. Here we summarize some examples of induced autophagy in neurons.

3.2.1 Hypoxic-ischemia

A combined hypoxic-ischemia (H/I) procedure in rodents has been widely used as a model for studying human ischemia or stroke. H/I causes extensive CNS neuron damage or injury and consequent neuronal death. Two recent studies of mouse models, one in neonatal brains [8] and the other in adult brains [65], showed that H/I induced robust formation of autophagosomes within a very short period of time (hours), concomitant with increased production of LC3-II levels. This induced autophagy occurred in different types of neurons in the hippocampus and striatum, suggesting that it is not cell type-specific. The rapid appearance of autophagic hallmarks in these two in vivo models demonstrated that H/I is a powerful stimulus for autophagy induction in CNS neurons. At the cellular level, H/I produces pro-inflammatory cytokines and simultaneous activation of both pro-survival (e.g., upregulation of Hsp70, phosphorylation of ERK and AKT) and pro-apoptotic signaling pathways (e.g., release of cytochrome c and AIF from mitochondria, cleavage of caspase-9 and -8) [65]. Therefore, at present, it is unknown which signals of autophagy induction are triggered by the H/I procedure. Interestingly, Koike et al further demonstrates that mice deficient in essential autophagy gene, Atg7, show nearly complete protection from H/I-induced neuron death, arguing strongly for the involvement of autophagy in a pro-death pathway in this specific setting [8]. Although the significant role of autophagy in triggering cell death pathways has yet to be clarified, this study suggests that H/I-induced autophagy, rather than an epiphenomenon of neuronal death, is an active process that controls cell fate. More importantly, inhibition of autophagy represents an attractive strategy for drug design in alleviating neuronal damage associated with H/I injury.

3.2.2 Excitotoxic Stimuli

Many excitotoxic stimuli provoke neurodegeneration through poorly defined cell mechanisms. The Clarke group has previously used NMDA (glutamate mimetic)-treated organotypic hippocampal slices to investigate this mechanism [66]. They found that, within two hours following the drug administration, numerous autophagosomes appeared in CA1 and CA3 pyramidal neurons, concomitant with signs of neurodegeneration. Moreover, this induction of autophagosome formation was inhibited by treatment of a JNK-signal blocking peptide, which also prevented neuronal death [66]. In another excitotoxicity-related model with focal injection of kainic acid, hippocampal extracts contain a significant increase in LC3-II levels as early as 4 hours following the injection [67]. Consistent with autophagic activity elicited in hippocampus, local administration of kainic acid caused a rapid increase in autophagosomes and autolysosome formation in the cytoplasm of striatal cells. In addition, it also augmented the ratio of LC3-II/LC3-I, LAMP2, cathepsin B, release of cytochrome c, and activation of caspase-3, suggesting stimulation of both apoptosis and autophagy [68]. Pre-treatment of autophagy inhibitor 3-methyladenine (3-MA) reversed the change in LC3-II levels, autophagosome formation, and loss of striatal cells [68]. In Lurcher mice, a genetic mouse model of excitotoxicity associated with Purkinje cell-specific degeneration, constitutive activation of glutamate δ2 receptor results in autophagy induction, preceding Purkinje cell degeneration. Moreover, the mutant glutamate δ2 receptor-mediated cell toxicity in transfected cells was blocked with autophagy inhibitor 3-MA [69].

Current knowledge indicates that excitotoxicity occurs when excitatory glutamate receptors, such as those for NMDA or kainic acid, are overexcited by high dosage of these ligands. The activated receptors allow the receptor channels to open, resulting in Ca²⁺ influx and consequent activation of a number of enzymes, including phospholipases, endonucleases, and proteases, such as calpain. Recent studies show that Ca²⁺ triggers strong stimulation of autophagic activity [70]. Therefore, it is possible that excitotoxic-related signals induce elevated levels of autophagy by a common Ca²⁺-mediated pathway. Furthermore, the above studies suggest that manipulation of autophagy, by blocking its activation, may be beneficial in the treatment of excitotoxic-associated neuronal death.

3.2.3 Methamphetamine and MPP⁺

Administration of methamphetamine (METH) in animals generally produces selective degeneration of dopamine (DA) neuron terminals without significant cell body loss [71]. This model is currently used to study specific aspects of neuron terminal dystrophy and degeneration in PD. To investigate the mechanism underlying the METH-mediated toxicity, the Sulzer group found that METH treatment promoted the rapid formation of autophagosomes, particularly in neuronal varicosities and, ultimately, within cell bodies of DA neurons from midbrain neuronal cultures [72]. A recent study in PC12 cells showed a similar response to METH administration, and provided additional evidence that suppression of autophagy, by blocking PI-3K class III, activity precipitates neuronal death. This study implicates a beneficial role of METH-induced autophagy in neural protection, perhaps through repairing or remodeling injured neurons [73].

A related PD cell model is the treatment of the neurotoxin 1-methyl-4-phenylpyridinium (MPP⁺), which produces mitochondria-targeted injury and contributes to parkinsonism induced by 1-methyl-4-phenyl-1,2,3,6-tetrahydro-pyridine in mammals. Administration of MPP⁺ in primary DA neuronal culture elicited an increased profile of intracellular autophagosomes, which was suppressed by extracellular signal-regulated protein kinase (ERK) kinase, a MEK inhibitor [74]. However, it is unclear why 3-MA, an inhibitor for autophagy, did not impede autophagosome formation induced by MPP⁺ in this setting. Interestingly, RNAi knock-down of autophagy gene LC3 or block of MEK with a specific inhibitor in MPP⁺-treated human neuroblastoma SH-SY5Y cells reduced formation of autophagosomes and cell death. One plausible explanation for MPP⁺-induced autophagy is through injured mitochondria, which elicited signals involving MEK for increased autophagic activity. Although the result shows that reduced autophagy is correlated with delay of cell death, whether or not autophagy acts as an executor of cell death is unclear [74].

3.2.4 Proteasome inhibition: a compensatory induction

Emerging evidence implicates inhibition of the ubquitin-proteasome system (UPS) in neurodegenerative diseases, such as AD and PD. Autophagy-lysosomal degradation and UPS are the two major proteolytic pathways, but they differ in various aspects including their substrates, capacity, and molecular and cellular machinery. Previous study of the effects of chronic low-level proteasome inhibition on neural homeostasis indicated that inhibition of UPS causes increased autophagic activity in clonal SH-SY5Y cells [75]. The view that increased autophagic activity is a compensatory response of impaired UPS system is supported by a few other studies. Pandey et al used the fly genetic system to provide in vivo evidence that significantly strengthened and expanded this view. They showed that histone deacetylase 6 (HDAC6), a microtubule-associated deacetylase that interacts with polyubiquitinated proteins, plays a critical role in this compensated induction of autophagy for impaired UPS function. Moreover, they found that forced expression of HDAC6 is able to rescue degeneration associated with UPS dysfunction in an autophagy-dependent fashion [76]. This study raises an important question regarding the cross-talk between the two main cellular degradation routes – autophagy and UPS – and whether or not HDAC6 is positioned at the intersection of these two pathways. Interestingly, a more recent report shows that genetic deletion of 26S proteasomes in mouse brain causes neurodegeneration and Lewy-like inclusions in midbrain DA neurons, which is accompanied by increased expression of autophagy genes and formation of numerous vacuoles resembling autophagosomes [77]. Although the mechanism of interaction between the two pathways is unclear, the compensatory up-regulation of autophagy for degradation of protein aggregates, which normally exceed the capacity of UPS degradation, should be explored as drug target for the treatment of neurodegenerative diseases.

3.2.5 Lysosomal enzyme/lipid storage or suppression: Autophagy “vicious circle”

A number of studies exploring autophagy in lysosome-related neurological diseases provide a new understanding of the pathogenic mechanisms of “old” diseases. A subset of these diseases, also known as lysosomal storage disease, is caused by deficiencies of lysosomal hydrolases, resulting in accumulation of undigested materials in the lysosomal or related compartment. These disorders include Niemann–Pick type C (NPC) [78, 79], mucolipidosis type IV with mutations in Mucolipin 1 (MCOLN1), multiple sulfatase deficiency and mucopolysaccharidosis type IIIA [80]. A common cellular pathologic feature in these diseases and their animal models is the manifestation of aberrant autophagic activity as evidenced by the accumulation of autophagosomes and increased levels of LC3-II, accompanied by neurodegeneration. Direct deletion of the genes encoding lysosomal enzymes such as cathepsin D (CD-/-) or combination of cathepsin B and L (CB-/- and CL-/-) in mouse brain, resulted in a similar phenotype [81]. The current view is that lipid trafficking molecules or lysosomal enzymes are required for the delivery, fusion or clearance of autophagosomes; deficiencies in these molecules causes disruption of this process, resulting in build-up of autophagosomes/autolysosomes unable to complete the digestion. This view supports a hypothesis that impaired autophagic activity is part of the disease mechanism for lysosomal storage diseases due to the inability of neurons to clear autophagosomes. However, these studies also provide evidence that otherwise amends this hypothesis. Pacheco et al showed that, in addition to impaired autophagy flux, there is also induction of autophagy in NPC by signaling through a complex of the class III phosphoinositide 3-kinase/beclin-1 [79]. A significant number of nascent autophagosomes (immature or before fusion with lysosomes) are found in CD-/- or CB-/- and CL-/- brain, suggesting that they are newly synthesized autophagosomes [81]. In a study of fibroblasts obtained from patients carrying mucolipidosis type IV mutation, Vergarajauregui et al suggests that the autophagosome accumulation is due to increased de novo autophagosome formation and delayed fusion of autophagosomes with late endosomes/lysosomes [80]. Therefore, the available evidence indicates that there is induced synthesis of autophagosomes in addition to the compromised autophagosome fusion or clearance in lysosomal storage diseases.

The deficiency in lysosomal degradation is expected to affect the net outcome of all forms of autophagy, including macroautophagy, microautophagy and chaperon-mediated autophagy. In line with the concept that blocking one type of degradation (e.g., proteasome or CMA) stimulates compensatory pathways (e.g. autophagy), intracellular accumulation of non-degraded materials resulting from lysosomal enzyme deficiency is expected to feed back to the upstream signaling which triggers the biosynthesis of autophagosomes. Since autophagosome degradation through lysosomes is impaired, it creates a “dilemma” for neurons to dispose autophagosomes, amassing a large number of autophagosomes of various stages. Thus, we propose a hypothesis that explains the sequence of the events in the disease, considering lack of detectable autophagosomes in healthy neuron. The accumulation of undigested materials due to lysosomal deficiency triggers de novo autophagosome biosynthesis in lysosomal storage disease. Impaired lysosomal degradation will trap these autophagosomes, causing build-up of the induced autophagosomes, and more undigested materials. This can continue in circles, and as a result of this rampant autophagy “vicious circle”, neurons may exhaust their energy and generate various toxic species that are harmful.

3.3 ESCRT: Autophagy meets endocytosis

Previous studies from the Seglen group showed that autophagosomes can fuse with endosomes and form a hybrid product, termed amphisomes, providing early evidence of interaction between autophagy and endocytosis [82]. Although this observation suggests the convergence of autophagosomes and endosomes, perhaps under specific settings, the significance of this process has not been fully appreciated. Recently, several studies showed that mutations in CHMP2B, a subunit of the endosome sorting complex required for transport of (ESCRT)-III, are linked to frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) and prevents autophagic degradation [83]. Along with these studies, genetic and ultrastructural analysis in Drosophila melanogaster showed that subunits of the endosomal sorting complex required for transport (ESCRT)- I, -II and -III, as well as their regulatory ATPase Vps4 and the endosomal PtdIns(3)P 5-kinase Fab1, are all required for normal autophagy function [84, 85]. These studies are important in several aspects. First, they reveal a novel mechanism involving dysfunctional autophagy in the pathogenesis of the specific type of FTD and ALS. Second, since the ESCRT proteins are known to function in the sorting of transmembrane proteins into the inner vesicles of the multivesicular body (MVB) during endocytosis [86], these studies suggest that MVB or endocytic pathways are critical routes for the trafficking of autophagosomes to lysosomes, and defective ESCRT function prevents fusion or maturation of autophagosomes. Third, they suggest that neurons are particularly vulnerable to abnormal function in the autophagosome-MVB (endosomes) pathway, and this may be associated with dendritic maintenance [83].

Several outstanding questions, however, arise from these studies. The ESCRT machinery is evolutionarily conserved [86], and its function in MVB biogenesis has been shown in many species including yeast and human. Curiously, although ESCRT/MVB, as well as autophagy, are highly conserved and well-characterized in yeast, the interaction between ESCRT/MVB and autophagy has not been demonstrated. Therefore, it raises a question whether the ESCRT-MVB pathway is only important for autophagy in higher eukaryotes, including fly and mammals. Furthermore, amphisomes are not detected in yeast, and autophagosomes fuse directly with the vacuole, which is mediated by SNARE proteins [87]. A related question is whether ESCRT-MVB represents the primary, if not the only, route for delivery of autophagosomes to lysosomal degradation, and whether this route occurs only in specific cell/tissue context. Previous studies also suggest that a considerable proportion of endocytosed cargo merges with autophagic pathway prior to being degraded by lysosomes in neurons [20]

It is also conceivable that the ESCRT-MVB pathway is actually part of the yet-to-be-defined basal autophagy process in neurons. Formation of ubiquitin- or p62/SQSTM1-labeled inclusions and autophagosomes accumulation are two important features associated with cells or neurons deficient in ESCRT components. Interestingly, time-course studies show that the formation of ubiquitinated inclusions occurs prior to the appearance of autophagosomes in cortical neuronal culture expressing mSnf7-2 siRNA or CHMP-2B^intron5. The early formation of ubiquitinated protein inclusions in these neurons defective in ESCRT function phenocopies the results from Atg5 or Atg7 deficient neurons [6, 7], suggesting that the ESCRT-MVB pathway is likely related to the Atg5 or Atg7-regulated function, which is required for the maintenance of protein homeostasis or basal autophagy function. Since impaired ESCRT pathway does not affect biosynthesis of autophagosomes (unlike Atg5 or Atg7), autophagosome biosynthesis is induced as a response to the increasing levels of ubiquitinated proteins. Thus, an alternative hypothesis is that formation of ubiquitin- or p62-labeled protein inclusions is the early response of neurons to the deficiency in basal autophagy caused by dysfunctional ESCRT, whereas the accumulation of autophagosomes is the secondary response to the increased levels of undigested protein and/or a result of impaired ESCRT function, which may be related to autophagosome fusion or maturation. The relationship of ESCRT with basal autophagy should be explored in the future.

4. Specialized Neuronal Autophagy in the Axons

4.1 Role of neuronal autophagy in axonal homeostasis

The axon is a highly specialized neuronal compartment that performs many functions independently from the soma. Accumulating evidence has revealed that, not only is autophagy uniquely regulated in the neuron, but it may also have a distinct function and regulation within the axonal compartment, independent from the dendrites and soma. Early morphological evidence showed that, following axotomy [88] or excitotoxic injury[4], double-membrane vacuoles resembling autophagosomes started accumulating and localized in dilated axon terminals. In later studies, Hollenbeck described the sequestering of autophagic substrates in autophagosomes at the most distal region of the axon, which were retrogradely transported to the soma [89]. Mammalian autophagosomes form in various regions throughout the cytoplasm. Upon maturation, they move toward lysosomes, which are primarily located in the juxtanuclear cytoplasm of the cell body [90, 91]. More recently, autophagosome formation in response to an excitotoxic insult was monitored in Lurcher mice expressing GFP-LC3. The constitutive activation of the GluRδ2 receptor resulted in an excitotoxic injury that triggered a rapid and robust accumulation of GFP-LC3-labeled autophagosomes in axonal dystrophic swellings [54]. The link between increased autophagy activity and axonal dystrophy would suggest local biosynthesis of autophagosomes at the distal axon terminals.

To elucidate the physiological role of neuronal autophagy, mutant mice containing a neural cell type-specific deletion of Atg7 were generated. Establishment of these mutant mice allowed the study of cell-autonomous events in cerebellar Purkinje cells deficient in autophagy. Characterization of the mutant Purkinje cells revealed the accumulation of aberrant organelles and membrane structures in dystrophic axon terminals. This result suggests a specific role for neuronal autophagy in the maintenance of membrane homeostasis at the axon terminal. We hypothesize that this highly specialized neuronal autophagy is required for keeping the balance of the membrane network, which normally involves cycling of membranous structures or vesicles, at the axon terminals to support synaptic activity. This particular function of autophagy is no surprise considering that the typical autophagic process involves dynamic membrane rearrangement and turnover. Future study should investigate in detail how this “self-eating” process participates in axonal membrane turnover and what membrane substrates axonal autophagy removes under physiological conditions. The answers to these questions are expected to advance our current understanding of autophagy in the neuron, as well as the disease processes that are associated with dysfunctional autophagy in the neuron and the axon.

4.2 MAP1B-LC3 interaction regulates autophagosome formation and possibly axonal transport

LC3 is best known as an autophagosome-associated protein. Despite the abundance of LC3 in neurons, the vast majority of LC3 is in soluble form and is not associated with autophagosomes. To understand neuron-specific autophagosome formation, Wang et al identified and analyzed the interaction between LC3 and microtubule-associated protein 1B (MAP1B), which is an abundant protein in the axons and plays an important role in regulating microtubule stability. They found that MAP1B binds to LC3 with high affinity, and over-expression of MAP1B in non-neuronal cell culture significantly reduces the number of LC3-associated autophagosomes, presumably through the MAP1B-LC3 interaction [54]. Since MAP1B (or MAP1A) is highly expressed in neurons, this result may potentially explain the scarcity of the autophagosomes in healthy neurons and the lack of GFP-LC3 puncta in GFP-LC3 transgenic brain. In contrast to MAP1B, phosphorylated MAP1B (MAP1B-P) was found to be associated with LC3-autophagosomes. It was further shown that LC3 was associated with increased levels of MAP1B-P in dystrophic axon terminals, coincident with the presence of a large number of autophagosomes [54]. Moreover, during development, MAP1B-P is highly expressed and is most concentrated in the distal axons of growing neurons [92]. The conserved role of MAP1B-P in axonal growth or repair during development or injury implicates autophagy in the remodeling of axon terminal structures for regeneration. Thus, we propose that interactions of LC3 with MAP1B or MAP1B-P provide a mechanism that regulates autophagosome formation in neurons or the axons.

The hypothesis that neuronal autophagy maintains axonal homeostasis is also in line with the idea that autophagy substrates are removed from the axons through axonal transport back to the soma where lysosomes perform degradation [89, 93]. Interestingly, recent studies have revealed a role of MAP1B in retrograde axonal transport [54]. It is possible that the MAP1B-P-LC3 interaction may regulate axonal transport of “autophagic cargo”. In support of this idea, microtubules and the dynein motor complex were previously shown to be required for autophagy process [90]. Dynein, the minus-end directed motor protein, mediates retrograde transport and interacts with the dynactin protein complex, which is implicated in cargo binding [94]. Inhibition of dynein impeded rapid movement of autophagosomes in mouse embryonic fibroblasts (MEF) cells and inhibited clearance of mutated alpha-synuclein or huntingtin by preventing fusion with the Lysosome [90, 95]. Furthermore, microtubule networks may facilitate the transport of LC3-positive autophagosomes,[96] but are not required for fusion with lysosomes.[54, 97, 98]

Although limited, recent studies involving live imaging showed that GFP-LC3 “dots” undergo retrograde transport in primary cerebellar granule cells [93]. In addition, both anterograde and retrograde transport of GFP-LC3-labeled autophagosomes were detected in neurites of differentiated PC12 and normal rat kidney cells.[91, 99] Although bidirectional movement was observed, Jahreiss et al. propose an overall bias for microtubule-dependent movement toward the nucleus, which is mediated by dynein [91]. In summary, we anticipate that future study, using live-imaging, should elucidate the dynamic process of autophagy in neuron or axon.

4.3 Ulk1 and axonal autophagy

Additional evidence supporting the hypothesis that autophagy is important in axonal transport and function comes from the study of yeast Atg1 and its homologue, Unc-51 (C. elegans uncoordinated-51). Atg1/Unc-51 is a highly conserved serine/threonine kinase. In initial C. elegans screenings, unc-51 mutants displayed disruptions in axon ultrastructure, including the presence of abnormal membranous structures in the axon and large swellings in the terminal [100]. The murine homologue, Unc51.1, is required for neurite extension in primary cerebellar granule neurons and governs vesicular membrane organization during outgrowth [101]. Thus, Unc-51 may provide a potential link between local membrane dynamics and autophagy at the axon terminal. Interestingly, the human homolog, Unc-51-like kinase 1 (ULK1), interacts with two MAP1B-LC3 related proteins: Golgi-associated ATPase enhancer (GATE-16) and GABA_A receptor associated protein (GABARAP), and have recently been shown to associate with p62/SQSTM1 [102, 103]. Nerve growth factor (NGF) binding with TrkA receptors facilitates K-63 polyubiquitination of ULK1, promoting its association with the UBA domain of p62/SQSTM1 and recruiting ULK1 into the active TrkA complex. ULK1 interacts with syntenin and SynGAP (which regulate endocytosis) and results in the trafficking of NGF-bound TrkA receptors into endocytic vesicles [104]. These results provide a possible mechanism for the cross-talk between two important membrane trafficking pathways – autophagy and endocytosis – and their role in axonal physiology. One possibility is that, by fusion with autophagosomes, some types of membrane compartments, including endosomes, are removed from the axons and delivered to lysosomes for degradation. This process may be essential to maintain the homeostasis of the axonal membrane network [58].

4.4 Axonal autophagy and axonopathy

Axonal dystrophy, a hallmark of axonopathy, can be triggered by neuronal injuries, excitotoxicity, and various neurodegenerative conditions. Despite the prevalence of this pathology, the molecular mechanisms underlying axonopathy, as well as the connection between axonopathy and neurodegeneration, remain poorly understood. Dysfunctional autophagy has recently been implicated in axonal dystrophy. Severe axonal swellings were observed when autophagy was suppressed in the mouse brain [6, 7]. Genetic ablation of Atg7, specifically in Purkinje cells of mutant mice, however, resulted in cell-autonomous axonal dystrophy and degeneration, implying an essential role for autophagy in membrane trafficking and turnover in axons. In contrast to Lurcher mice, axon terminals of Atg7 deficient mice lacked autophagosome-like vesicles, but amassed abnormal organelles and membranous structures [58]. Moreover, mice with Purkinje-cell specific deletion of either Atg5 or Atg7, exhibited axonal dystrophy much earlier than dendritic tree atrophy and cell death [57, 58]. Thus, overactive or insufficient autophagy may contribute to axonopathy, which is a prominent feature of human neuropathology.

Accumulating evidence indicates that axonal degeneration precedes neuronal cell body death and undergoes a self-destruct mechanism that is distinct from apoptotic death in the soma [105, 106]. A classic example of axonal degeneration is Wallerian degeneration, in which transected axons or neurites undergo complete fragmentation distal to the site of injury [106]. In mice carrying the spontaneous-occurring slow wallerian degeneration (Wld^s) mutation, Wallerian degeneration is markedly delayed following transection. However, the “dying back” model, a progressive retrograde degeneration of the distal axon, may better reflect the chronic injury associated with neurodegenerative diseases [106]. NGF deprivation induced accumulation of autophagosomes in the distal tips of PC12 neurites supporting the hypothesis that autophagy induction contributes to dying back degeneration in which distal neurites are more fragile than proximal segments. Indeed, knocking down levels of Atg7 or Beclin1 caused a significant delay of neurite degeneration after NGF deprivation in sympathetic neurons [107]. These findings suggest that autophagy activation contributes to neurite degeneration, thus providing a valuable clue about the possible involvement of autophagy in the mechanism of axonal degeneration.

In summary, based on current evidence, we propose that basal autophagy plays an important role in the maintenance of axonal homeostasis by the removal of “autophagic cargo” from distal axons or terminals through retrograde axonal transport. Although the “autophagic cargo” can be associated with LC3, it is largely unclear whether they are in the form of typical autophagosomes [55]. Future study should investigate the nature of “autophagic cargo” in healthy neurons or axons. In contrast, pathological conditions can induce autophagy involving stimulated local synthesis of autophagosomes in axons. Degradation by the autophagic process may occur in the axons, which may require the accessibility of lysosomes or late endosomes in the axons. Or degradation may take place in the soma, which then requires retrograde transport of autophagosomes. Nonetheless, the inability of axons to either degrade or transport a large number of autophagosomes produced locally would have deleterious consequence to the axons, causing axonal dystrophy and degeneration (Figure 2).

Figure 2.

Proposed model for neuronal autophagy in the axons. Under normal conditions, basal autophagy maintains axonal homeostasis by removal of “autophagic cargo” (top). The “autophagic cargo” undergoes retrograde axonal transport to the soma and fuses with lysosomes for degradation. Stress or injury can induce local biosynthesis of autophagosomes, resulting in their accumulation in axons and axon terminals. It has been suggested that, under pathological conditions, precursors of degradative vesicles or lysosomes may be anterogradely transported to the axon terminal and contribute to the degradation of autophagosomes (bottom left). Neurons deficient in autophagy amass proteins, organelles, and aberrant membrane structures at axon terminals, resulting in gross axonal swellings or dystrophy (bottom right).

5. Specific function of p62/SQSTM1 in autophagy-mediated protein degradation

5.1 Role of neuronal autophagy in protein homeostasis

One of the critical findings in the study of Atg5 or Atg7-deficient neurons is the recognition that autophagy is required to suppress spontaneous protein aggregation. Loss of these autophagy genes causes accumulation of polyubiquitinated proteins. These appear as inclusion bodies, which increase in size and number with aging. However, proteasome function, which is generally known to cause abnormal ubiquitin-mediated proteolysis when impaired, is not compromised. Thus, blocking neuronal autophagy results in delayed global turnover of cytoplasmic components, accumulation of misfolded and/or unfolded proteins, followed by ubiquitination and the formation of inclusion bodies. Furthermore, these findings reveal an important role for basal autophagy in the protein clearance or protein homeostasis in neurons. While autophagy is absolutely required to degrade proteins for the maintenance of homeostasis in healthy neurons, its function becomes even more prominent in neurons expressing disease-related proteins. Recent studies show that autophagy contributes to the degradation of aggregate-prone proteins such as polyglutamine containing proteins and mutant alpha-synuclein [15]. Although it remains largely unclear how autophagy may achieve selective degradation of these disease-related proteins, expression of disease-related proteins in cells or neurons stimulates autophagy activation, and elevated autophagy is correlated with reduced levels of protein aggregates, concomitant with a decrease in neurotoxicity [61, 108].

5.2 Link of p62/SQSTM1 to autophagic degradation

Emerging evidence suggests that p62/SQSTM1, originally identified as an ubiquitin-associated protein, provides a link between autophagy and selective protein degradation. Recent studies show that p62/SQSTM1 is an LC3-interacting protein [103, 109]. Because p62/SQSTM1 can bind a large number of proteins through its multiple protein-protein interaction motifs, it may mediate diverse signaling pathways including cell stress, survival, and inflammation. Structural analysis reveals that p62/SQSTM1 N-terminal Phox and Bem1 (PB1) domain exhibits self-oligomerization, and the C-terminal ubiquitin-associated (UBA) domain can bind ubiquitinated proteins, suggesting a link between p62/SQSTM1 and disease-related protein inclusion formation. Consistent with the notion that p62/SQSTM1 is likely more related to autophagy-lysosome degradation, lysosomal inhibition, but not proteasomal inhibition, resulted in marked accumulation of p62 as well as LC3-II. Moreover, ablation of autophagy in neurons leads to rapid and robust increase in p62 protein levels [54, 109]. These observations suggest that p62 is selectively degraded by autophagy via the LC3-interacton. In addition, since autophagosome formation and long-lived protein degradation are intact in p62-knockout mice, p62 is considered a specific substrate of autophagy rather than a molecule involved in autophagosome formation.

Recent studies have identified the LC3 recognition sequence (LRS) in uncharacterized linker regions between the zinc finger and UBA domains in murine p62. The LRS is comprised of 11 amino acids (Ser334-Ser344), which include an acidic cluster and hydrophobic residues (DDD or DEE and WXXL or WXXV). Interestingly, this sequence is almost identical to a previously reported LC3-interacting region (LIR) of human p62. The crystal structure of the LC3-LRS complex, solved at 1.56Å resolution, reveals that the acidic cluster of Asp337-Asp339 in LRS interacts with basic residues in the N terminus of LC3, and that the Trp-340 and Leu-343 residues are inserted into two hydrophobic pockets, exposed on the ubiquitin domain of LC3. LC3 has basic residues at its N-terminal α-helix surface, and these residues are involved in the interaction with the acidic cluster of LRS, whereas the other two Atg8 mammalian homologues, GATE-16 and GABARAP, have acidic residues in their respective N-terminal α-helical surfaces [110]. It is thus conceivable that p62 is a more favorable target for LC3 than GATE-16 or GABARAP.

5. 3 Role of p62 in inclusion body formation

Characterization of protein inclusions in autophagy-deficient cells or neurons reveals that nearly all inclusions are positive for ubiquitin and p62 [109]. This raises an important question regarding the role of p62 in the formation of protein inclusions. To answer this, Komatsu et al crossed p62-/- mice to mutant mice with autophagy deficiency in specific tissues. They found that loss of p62 greatly reduces the formation of ubiquitinated protein inclusions resulting from impaired autophagy in mice. Similar studies were also performed in fruit flies and again the results indicate that the presence of p62 is necessary for the occurrence of protein inclusions in neurons [111]. To understand the molecular basis of this function of p62, it was shown that p62 proteins harboring mutations in LRS, escape efficient degradation by autophagy, leading to inclusion formation despite normal autophagy in cells. In addition, degradation of the PB1 mutant of p62, which is defective in oligomerization, is markedly attenuated, but no protein inclusions occur. These results suggest that increased levels and oligomerization of p62 are necessary for the formation of inclusion bodies, and furthermore, oligomerization of p62 via PB1 is a critical event for facilitating their degradation by autophagy.

Considerable evidence has shown that p62 is a component of protein inclusion bodies found in human disorders such as in liver injuries (e.g., alcoholic hepatitis, steatohepatitis, and α1-antitrypsin deficiency) and neurodegenerative diseases (e.g., AD, PD, and ALS). Although autophagy is beneficial in the clearance of toxic protein aggregates and prevents cellular toxicity, whether or not p62 is indeed essential for the formation of disease-related inclusions and the exact role of p62 in the disease process remains unknown at present. Intriguingly, the formation of ubiquitin-positive aggregates induced by proteasome inhibition is also suppressed in p62- deficient cells, suggesting that p62 is a general mediator of inclusion formation. Thus, we hypothesize that p62 normally functions as an adaptor that links certain proteins to autophagy machinery for degradation. When cellular degradation systems (proteasomes or autophagy-lysosomes) are compromised or overwhelmed, p62 levels are sharply increased (especially with impaired autophagy) and it simply behaves as “glue” that bonds many autophagic protein substrates together, manifesting as protein aggregates and eventually large protein inclusions.

5. 4 Potential mechanism of neurodegeneration caused by autophagy-deficiency

Emerging evidence indicates that soluble proto-fibrils or small oligomers of disease-related proteins are the cytotoxic species, whereas large inclusion bodies derived from the sequestration of toxic protein oligomers are protective. Therefore, we can hypothesize that the presence of p62 is protective in suppressing the toxicity of proto-fibrils or protein oligomers, given the role of p62 in facilitating the formation of relatively less toxic inclusion bodies. Accordingly, loss of p62 in autophagy-deficient neurons would exacerbate the neurodegenerative process because toxic proteins would accumulate intracellularly and no inclusion bodies would form. Surprisingly, in Atg7/p62 double knockout mice, loss of p62 was found to be associated with reduced liver dysfunction or little change in the neurodegenerative process caused by autophagy-deficiency, despite the absence of any inclusion bodies [109]. Apart from the specific function of p62 in liver, these results seem to argue against a critical role of protein inclusions in the context of cytotoxicity, especially in the absence of disease-related proteins. Alternatively, they may suggest that the proteins accumulated upon impairment of autophagy are largely neutral or at least neurons are tolerant to the build-up of those proteins. In this regard, the degradative function of autophagy in protein homeostasis control is likely more significant towards disease-related, aggregate-prone proteins than normal long-lived protein substrates of autophagy.

So what could be the primary cause that leads to neuronal death when neuronal autophagy is impaired? As discussed above, analyses of knock-out mice with Purkinje cell-specific Atg7 deletion (Atg7^flox/flox;pcp2) demonstrated that loss of Atg7 initially causes cell-autonomous, progressive dystrophy (manifested by axonal swelling) and degeneration of the axon terminals followed by cell-autonomous Purkinje cell death and mouse behavioural deficits. Furthermore, the mutant Purkinje cells developed aberrant organelles in the swelling axons, suggesting the important role of autophagy in the regulation of local axonal membrane trafficking and turnover [58]. The study also implicates impairment of axonal autophagy as a mechanism for axonopathy associated with neurodegeneration. Interestingly, such axonopathy associated with Atg7-deficient Purkinje cells and hypothalamic neurons is still observed in Atg7/p62-double knockout neurons, indicating that the development of axonopathy in Atg7-deficient neurons is p62-independent [109]. Therefore, we hypothesize that axonal dystrophy and degeneration provide an important mechanism for the neurodegeneration caused by autophagy-deficiency.

6. Conclusion and perspectives

The past several years have witnessed a rapid growth of autophagy research. With emerging evidence that links the aberrant autophagic activity to various neurodegenerative diseases, it is of upmost importance to understand the basic process of neuronal autophagy at molecular and cellular levels. This would aid in revealing the exact role of autophagy in the neuropathogenesis associated with different diseases. Currently, the particular challenge facing us is to establish robust and physiological cellular and animal models that allow us to dissect neuronal autophagy, and to identify specific factors that contribute to the regulation of neuronal autophagy process. We anticipate that, with the development of new experimental approaches, particularly in the area of cell biology, mouse genetics, live-imaging and proteomics, we will make significant progress in this field and provide valuable knowledge to the design of drugs targeted at autophagy in the treatment of neurodegenerative diseases.