Autophagy and Neuronal Cell Death in Neurological Disorders

Abstract

Autophagy is implicated in the pathogenesis of major neurodegenerative disorders although concepts about how it influences these diseases are still evolving. Once proposed to be mainly an alternative cell death pathway, autophagy is now widely viewed as both a vital homeostatic mechanism in healthy cells and as an important cytoprotective response mobilized in the face of aging- and disease-related metabolic challenges. In Alzheimer’s, Parkinson’s, Huntington’s, amyotrophic lateral sclerosis, and other diseases, impairment at different stages of autophagy leads to the buildup of pathogenic proteins and damaged organelles, while defeating autophagy’s crucial prosurvival and antiapoptotic effects on neurons. The differences in the location of defects within the autophagy pathway and their molecular basis influence the pattern and pace of neuronal cell death in the various neurological disorders. Future therapeutic strategies for these disorders will be guided in part by understanding the manifold impact of autophagy disruption on neurodegenerative diseases.

In late-onset neurodegenerative diseases (e.g., Parkinson’s), impairment of autophagy leads to the buildup of pathogenic proteins and damaged organelles, triggering neuronal apoptosis or necrosis.

Soon after the discovery of lysosomes by de Duve in the 1950s, electron microscopists recognized the presence of cytoplasmic organelles within membrane-limited vacuoles (Clark 1957) and observed what appeared to be the progressive breakdown of these contents (Ashford and Porter 1962). Proposing that “prelysosomes” containing sequestered cytoplasm matured to autolysosomes by fusion with primary lysosomes, de Duve and colleagues (de Duve 1963; de Duve and Wattiaux 1966) named this process “autophagy” (self-eating). Neurons, as cells particularly rich in acid phosphatase-positive lysosomes, were a preferred model in the initial investigations of autophagy. Early studies of pathologic states such as neuronal chromatolysis (Holtzman and Novikoff 1965; Holtzman et al. 1967) linked neurodegenerative phenomena to robust proliferation of autophagic vacuoles (AVs) and lysosomes. Although de Duve appreciated the importance of lysosomes for maintaining cell homeostasis, he was especially intrigued with their potential as “suicide bags” capable of triggering cell death by releasing proteases into the cytoplasm. Despite some support for this notion (Brunk and Brun 1972), the concept was not significantly embraced until many decades later. Instead, for many years, lysosomes and autophagy were mainly considered to perform cellular housekeeping and to scavenge and clean up debris during neurodegeneration in preparation for regenerative processes. The connection between autophagy and neuronal cell death reemerged in the 1970s from observations of Clarke and colleagues, who presented evidence that the developing brain deployed autophagy as a form of programmed neuronal cell death during which autophagy was massively up-regulated to eliminate cytoplasmic components, at once killing the neuron and reducing its cell mass for easy removal. Self-degradation was suggested as a more efficient elimination mechanism than apoptosis, which requires a large population of phagocytic cells and access of these cells to the dying region (Baehrecke 2005). Indeed, the best evidence for this process is in the context of massive cell death, as in metamorphosis and involutional states (Das et al. 2012).

Clarke proposed that autophagic cell death (ACD)—type 2 programmed cell death (PCD)—could be a relatively common alternative route to death distinct from apoptosis—type 1 PCD (Clarke 1990)—or caspase-independent cell death—type 3 PCD (Fig. 1). The distinguishing features of ACD are marked proliferation of AVs and progressive disappearance of organelles but relative preservation of cytoskeletal and nuclear integrity until late in the process (Schweichel and Merker 1973; Hornung et al. 1989). In this original concept of ACD or type 2 PCD, death is achieved by autophagic digestion of organelles and essential regulatory molecules and elimination of death inhibitory factors (Baehrecke 2005). With the advent of the molecular era of autophagy research in the 1990s, it became possible to verify the most important implication of ACD, namely, that the death could be prevented by inhibiting autophagy genetically or pharmacologically. Meanwhile, reports of prominent lysosomal/autophagic pathology in Alzheimer’s disease (AD) (Cataldo et al. 1997; Nixon et al. 2000, 2005) and other neuropathic states (Anglade et al. 1997; Rubinsztein et al. 2005) raised important questions about whether autophagy pathology signifies a prodeath program or an attempt to maintain survival—a critical question for any potential therapy based on autophagy modulation. In this article, we will examine evidence for the various neuroprotective roles of autophagy and review our current understanding of how specific stages of autophagy may become disrupted and influence the neurodegenerative pattern seen in major adult-onset neurological diseases. We will particularly focus on how neurons regulate the balance between prosurvival autophagy and well-established cell death mechanisms in making life or death decisions.

Figure 1.

Neuronal cell death: three general morphological types of dying cells in the developing nervous system, as initially classified by Schweichel and Merker (1973) and later Clarke (1990). (A,B) Type 1 (“apoptotic”) cell death: (A) A neuron, from the brain of a postnatal day 6 mouse pup, in the middle of apoptotic degeneration showing cell shrinkage, cytoplasmic condensation, ruffled plasma membrane, and a highly electron-dense nucleus. Endoplasmic reticulum (ER) is still recognizable and some are dilated. A small number of autophagic vacuoles (AVs) can be seen (arrows). (B) A late-stage apoptotic neuron displaying electron-dense chromatin balls (CB), each surrounded by a small amount of highly condensed cytoplasm. (Panel from Yang et al. 2008; reprinted, with permission, from the American Association of Pathologists and Bacteriologists.) (C) Type 2 (“autophagic”) cell death: a deafferented isthmo-optic neuron in developing chick brain after uptake of horseradish peroxidase to highlight (electron dense) endocytic and autophagic compartments. The cell death pattern features pyknosis, abundant AVs, and sometimes dilated ER and mitochondria. (Panel from Hornung et al. 1989; reproduced, with permission, from John Wiley & Sons) (D) Type 3 (“cytoplasmic, nonlysosomal”) cell death: a motoneuron displaying markedly dilated rough ER, Golgi, and nuclear envelope, late vacuolization, and increased chromatin granularity. (Panel from Chu-Wang and Oppenheim 1978; reproduced, with permission, from John Wiley & Sons) Scale bars, 1 µm (A,B); 2 µm (C,D).

THE STAGES OF AUTOPHAGY

Since its initial description, autophagy has referred to the process by which unwarranted constituents of the cell are sequestered within vacuoles and degraded by lysosomes (Klionsky et al. 2010). Lysosomal digestion of the cell’s own cytoplasmic material is the common cardinal feature of all forms of autophagy. Different mechanisms for delivery of substrates to the lysosomal compartment, however, distinguish among the three major subtypes of autophagy in mammalian cells (Fig. 2). The least well understood of these, microautophagy, is a constitutive and sometimes selective process by which proteins or organelles may be engulfed by invagination of the lysosomal or endosomal membrane and scission of the cargo-containing vesicle inside the lumen. Cargo selection and delivery to endosomes is facilitated by the chaperone Hsc70 and internalization relies on the same endosomal sorting complex required for transport (ESCRT) I and III complexes involved in multivesicular body (MVB) formation (Sahu et al. 2011). In chaperone-mediated autophagy (CMA), cytosolic proteins containing a KFERQ motif are delivered to a LAMP2A-containing complex on the lysosomal membrane via a complex of chaperones, including Hsc70 and HSP70, and cochaperones (EGHSP40, HSP90, HIP, HOP, and BAG1). The substrate is subsequently unfolded and degraded with the help of lysosomal Hsc70 and a complex of proteins that disassembles the transmembrane translocation complex (Bandyopadhyay et al. 2008, 2010). CMA activity is constitutive in most cells but up-regulated under conditions of stress or nutrient deprivation.

Figure 2.

Schematic for the three major subtypes of autophagy—macroautophagy, chaperone-mediated autophagy, and microautophagy. The mammalian target of rapamycin (mTOR) kinase, the principal regulator of macroautophagy, is activated by inhibiting the tuberous sclerosis complex (TSC1) and TSC2, thereby increasing the function of the GTP-binding protein Rheb (A). Insulin or growth factors suppress autophagy by activating the class I phosphatidylinositol 3 kinase (PI3K)–Akt/protein kinase B (PKB) pathway. Abundant intracellular stores of amino acids and ATP suppress autophagy by inhibiting AMP-activated protein kinase (AMPK), an activator of the TSC complex. Macroautophagy is orchestrated by complexes composed of autophagy gene (Atg)-related proteins, which coordinate specific steps in autophagy induction and sequestration as described in the text. The process is initiated when an “isolation” membrane is created from a preautophagosomal structure (PAS) under the direction of the class III PI3K complex and Atg proteins, including Beclin 1 (Atg6) (B). Two ubiquitin-like protein conjugation pathways (C) direct the expansion of the isolation membrane as it sequesters a region of cytoplasm and organelles into a double-membrane-limited autophagosome (AP). Microtubule-associated protein light chain 3-II (LC3-II), formed by phosphoethanolamine conjugation of LC3-I, translocates to the autophagosome membrane (C). Digestion of the sequestered cytoplasmic cargo is initiated when a lysosome (Ly) fuses with the outer membrane of the autophagosome to form an autophagolysosome (APL) and lysosomal hydrolases are introduced. The completion of substrate digestion within autolysosomes (AL) ultimately results in lysosome retrieval. In chaperone-mediated autophagy, cytosolic proteins containing a KFERQ motif are selectively chaperoned (Chap) and delivered to the lysosomal lumen for degradation through the binding to LAMP2 located on the lysosomal membrane. In microautophagy, small quantities of proteins or organelles directly enter lysosomes by invagination of the lysosomal or endosomal membrane.

A third form of autophagy, macroautophagy, mediates either bulk or selective degradation of cytoplasmic constituents and is arguably the mechanism with the highest degradative capacity. In this process, a double-membrane structure, termed the “phagophore,” originates from several possible sources of cellular membrane and becomes specialized as a preautophagosomal structure (PAS) (Fig. 2B). Under the coordination of two ubiquitin-like protein conjugation pathways (Fig. 2C) (Ohsumi 2001), the PAS elongates into an “isolation” membrane that envelops a region of cytoplasm ultimately forming a double-membrane-limited vesicle, the autophagosome. The outer membrane of the autophagosome fuses with a lysosome or late endosome, creating an autolysosome or an amphisome, respectively. Amphisome formation predominates in axons in which autophagosomes are actively formed and fuse rapidly with late endosomes before being delivered by retrograde transport to the lysosomes concentrated mainly at proximal axon levels and in the perikaryon (Lee et al. 2011). The digestion of sequestered material by dozens of acidic hydrolases (Gordon and Seglen 1988; Liou et al. 1997; Eskelinen 2005; Fader and Colombo 2009; Noda et al. 2009) is promoted when the autolysosome is acidified by vacuolar ATPase (vATPase), a proton pump on the lysosomal membrane, and cathepsins become fully activated (Yoshimori et al. 1991). Autophagy induction is regulated by activity of the mammalian target of rapamycin (mTOR) kinase, which is controlled by growth factor supply and signaling via the PI3 kinase-AKT pathway and by specific amino acids and ATP through AMP kinase (Fig. 2A) (He and Klionsky 2009; Ravikumar et al. 2010). Lysosomes are reformed from autolysosomes through a process also regulated in part by mTOR (Yu et al. 2010).

NEUROPROTECTIVE ACTIONS OF AUTOPHAGY

Macroautophagy is constitutive in neurons and nonselective under nutrient deprivation conditions but can selectively target an organelle that is damaged and exposes a molecular signal that promotes sequestration (Dikic et al. 2010; Weidberg et al. 2011; Youle and Narendra 2011). Recognition of this selective form of macroautophagy has highlighted the working relationship between autophagy and the ubiquitin-proteasome system (UPS) in which a decline in UPS activity may up-regulate autophagy (Fortun et al. 2003; Korolchuk et al. 2009). Certain proteins play regulatory roles in both processes, such as p62, an adaptor protein for autophagy, VCP/P97, which acts through p62 and ubiquitin (Tresse et al. 2010), and Parkin, an E3 ubiquitin ligase implicated in Parkinson’s disease (PD) (Yoshii et al. 2011).

It is widely believed that autophagy exerts mainly prosurvival influences on normal cells and its increased activity has been linked to enhanced longevity in animal models (Cavallini et al. 2008; Morselli et al. 2010; Rubinsztein et al. 2011). Autophagy protects cells under stress from nutrient deprivation (Levine and Yuan 2005), loss of energy, or states of protein aggregation by breaking down less essential cellular constituents to recover amino acids, lipids, and other metabolites needed to maintain bioenergetics and synthesize more adaptive proteins, thereby forestalling apoptosis. Another important protective strategy in neurodegenerative disease states is the accelerated elimination by autophagy of damaged and potentially toxic proteins as well as proapoptotic molecules. For example, activated caspase 3 appears in affected neurons in mouse models of AD but is mainly sequestered within autophagosomes (Yang et al. 2008). Similarly, mitophagy, the selective removal of damaged mitochondria by autophagy, is a crucial line of defense against apoptosis believed to fail in some neurodegenerative diseases. In mitophagy, the E3-ubiquitin ligase Parkin, or a related protein such as Nix (Bnip3L) (Sandoval et al. 2008), is recruited to the compromised mitochondrion by certain mitochondrial proteins exposed by the injury. The ubiquitinated protein(s) serve as a binding site for p62 and light chain 3 (LC3) to induce sequestration (Narendra et al. 2008).

AUTOPHAGY AND CELL DEATH PATHWAYS

The cross-regulation of autophagy and apoptosis provides insight into how autophagy protects against apoptotic cell death and conversely how the apoptotic program subverts autophagy to achieve its goal. Central to this cross talk is the interaction between the key apoptosis regulators, Bcl-2 and Bcl-xL, and beclin 1, a component of two separate protein complexes regulating autophagosome formation and endosome maturation, respectively. As a Bcl-2 homology (BH)–3 domain-only protein (Oberstein et al. 2007), beclin 1 may interact with Bcl-2 and Bcl-xL (Pattingre et al. 2005), which diminishes its interactions with these complexes and deprives cells of autophagy’s antiapoptotic actions. Mutating the BH3 domain of beclin or the BH2 receptor domain of Bcl-2/Bcl-xL abolishes this capacity to inhibit beclin-dependent autophagy (Maiuri et al. 2007). Conversely, dissociation of beclin 1 and Bcl-2/Bcl-xL, resulting in increased autophagy and antiapoptotic effects, can be achieved in several ways. These include competitive displacement of either partner by interacting proteins (Bad, Bnip3, or Tbid), which include components of the beclin 1 complexes that support autophagy (e.g., UVRAG, Atg14L/Barkor, and Hmgb1). The interaction is also blocked by beclin phosphorylation via death-associated protein kinase (DAPK) (Zalckvar et al. 2009), MAP kinase (ERK1 or JNK1)-mediated phosphorylation of Bcl-2 (Wei et al. 2008; Tang et al. 2010), or Traf6-mediated ubiquitination of beclin 1 (Shi and Kehrl 2010).

Pathological activation of calpains, a family of calcium-activated neutral proteases, can also inhibit autophagy by truncating Atg5, which then translocates from the cytosol to mitochondria where it associates with Bcl-xL and can trigger cytochrome c release (Yousefi 2006). Calpains can also inactivate caspases and potentially convert apoptotic death to necrotic death (Lankiewicz et al. 2000; Syntichaki and Tavernarakis 2002). Whether apoptosis or necrosis ensues depends on various factors, including the developmental status of the neuron, extracellular glutamate levels, cytosolic and mitochondrial calcium levels, mitochondrial membrane potential, ATP levels, and calpain activity (Pang and Geddes 1997; Nasr et al. 2003; Pang et al. 2003).

AUTOPHAGY IN NEURONS AND NEURODEGENERATION

The paucity of AVs in healthy neurons (Mizushima et al. 2004; Nixon, et al. 2005) initially led investigators to suggest that autophagy activity in neurons is relatively low unless induced by some form of stress (Holtzman et al. 1967). Recent evidence has shown, however, that neuronal autophagy is constitutive and, in fact, may be quite active, although AVs remain scarce because lysosomal clearance of these intermediates is very efficient (Boland et al. 2008). One important implication is that, although neurons seem able to accommodate high autophagy activity with minimal ill effects, they are unusually vulnerable to any impairment of autolysosomal clearance. The special reliance of neurons on autophagy has long been suspected from observations that many human primary lysosomal diseases, e.g., lysosomal storage disorders, preferentially affect the brain. This vulnerability of neurons to autophagic/lysosomal dysfunction is not surprising when it is considered that neurons must survive throughout the life of the organism and lose the capability for mitosis, which helps mitotic cells cope with accumulating waste materials by diluting this burden through cell division. Also, neurons must maintain especially large volumes of membrane and cytoplasm associated with long axons and dendrites and must continually traffic autophagy-related compartments long distances back to the cell body where substrate clearance by lysosomes is most active (Lee et al. 2011). Even briefly inhibiting lysosomal proteolytic activity disrupts transport of AVs and causes them to selectively accumulate in dystrophic swelling (Lee et al. 2011) and in perikarya (Ivy et al. 1984; Bednarski et al. 1997).

Investigating cell death pathways in neurons and autophagy is complicated by the slow evolution of neurodegenerative diseases in which prodeath disease and prosurvival factors within a given neuron may battle over many months or years and proteolytic systems may cross talk extensively, yielding in the end a complex picture that defies classification as a distinct pattern of cell death. In AD, for example, neurons with dystrophic axons containing activated caspases, activated calpains, and massive AV accumulations may persist for months or years (Coleman 2005; Rao et al. 2008; Adalbert et al. 2009). Moreover, in most chronic neurodegenerative diseases, only a tiny percentage of the neurons may be degenerating at any given time and may be at different stages of the degenerative process when analyzed. Moreover, cell death patterns may vary across neuronal cell types even in the same disease model, whereas nonneuronal cells in brain may be mounting quite different biochemical responses. Analysis at the single cell level is, therefore, the optimal approach to cell death investigations in brain.

AUTOPHAGY OVERACTIVATION—AN UNCOMMON CELL DEATH PATHWAY IN NEURONS

Although autophagy is generally neuroprotective, can too much of a good thing at the wrong time be deleterious? ACD in the context of involution and metamorphosis during invertebrate development may be among the rare instances in which ACD, as originally defined by Clarke, takes place, and the phenomenon in this situation seems to serve the physiological purpose of eliminating entire cell populations. In mammalian cells, however, ACD strictly defined as death executed by the autocannibalism of constituents essential for cell survival and dependent on autophagy genes is uncommon. An autophagic (type 2 PCD) pattern of death blocked by Atg-gene deletion has been observed in cells in which apoptosis is prevented by deleting Bcl-2 family members, Bax, and Bak or by inhibiting caspases (Shimizu et al. 2004; Yu et al. 2004). Even in this specialized situation, however, it is not entirely clear that autophagy is sufficient to execute death without help from proteases that mediate necrosis. In cells that have competent caspases, additional cell death pathways are invariably activated during cell death showing the type 2 PCD morphological pattern, although autophagy may serve as the upstream trigger (Pyo et al. 2005; Nixon 2006). In some cases, AV proliferation occurs in the context of cell death executed by caspases and may facilitate execution but is not essential for death (Nixon 2006).

Autophagy inhibition by 3-methyl adenine (3MA) has been used to implicate autophagy in cell death execution by showing blocked or delayed cell death after this treatment, although the interpretation of protection via autophagy inhibition should be qualified because this compound inhibits not only the class III PI3 kinase regulating autophagy but also class I PI3 kinase involved in endocytosis. Also, in most of these cases, cytoprotection is not absolute and death ultimately ensues via cytochrome c release and caspase 3 activation (Uchiyama 2001; Canu et al. 2005; Kaasik et al. 2005), indicating that an apoptotic pathway may be operating in parallel. Furthermore, the inhibition of a particular cathepsin (Uchiyama 2001; Canu et al. 2005; Kaasik et al. 2005) also delays or blocks cell death in many of the same models, further supporting the idea that lysosomal destabilization and cathepsin release ultimately triggers apoptosis.

Inhibiting autophagy has been reported to be beneficial in several neurodegenerative states, the clearest example being hypoxic ischemic (HI) brain injury. In the most compelling of these studies, Atg deletion reduced autophagy-related pathology and nearly completely prevented severe hippocampal damage after HI brain injury in neonatal or adult mice (Koike et al. 2008). In both mouse models, however, effects of autophagy modulation were upstream of the neuronal cell death executioner pathways that were also activated in these models—both caspase 3-dependent and -independent pathways, in the case of the neonatal model, and the caspase 3-independent pathway only in the adult model. Similar observations on this injury model have been made using 3MA to assess the involvement of autophagy in neuronal death. (Wen et al. 2008; Puyal et al. 2009; Piras et al. 2011; Wang et al. 2011; Xin et al. 2011; Zhang 2012). Beneficial effects of autophagic inhibition have also been seen in several models of chronic neurodegenerative diseases. In 6-hydroxydopamine (6-OHDA) injured rat substantia nigra neurons, up-regulated autophagy (as evidenced by AV accumulation) and cell death were prevented by 3MA pretreatment (Li et al. 2011). Neuronal death induced by dysfunctional ESCRT-III, which is associated with frontotemporal dementia linked to chromosome 3, could be delayed by 3MA administration or by knocking down Atg5 and Atg7, suggesting that autophagic stress by excess accumulation of autophagosomes is detrimental to neuronal survival (Lee et al. 2007; Lee and Gao 2009).

When autophagy inhibition appears to be cytoprotective, the particular step in autophagy involved is not straightforward. Autophagy induction is often accompanied by increased lysosome biogenesis and up-regulated hydrolase expression (Dehay et al. 2010; Palmieri et al. 2011; Settembre et al. 2011). In situations in which lysosomal function is also compromised or lysosomes are destabilized, it is possible that autophagy inhibitors relieve autophagic stress on compromised lysosomes by slowing delivery of autophagic substrates rather than by attenuating an overaggressive autocannibalistic process. Indeed, healthy neurons in culture seem to tolerate strong autophagy induction (Lee et al. 2011) unless lysosomal function is also impaired. In situations in which blocking autophagy is protective, fending off lysosome destabilization could be more germane to the mechanism of cytoprotection than is the blockade of authentic ACD.

NEURONAL CELL DEATH ASSOCIATED WITH FAILURE OF AUTOPHAGY

In contrast to autophagy overactivation, autophagy failure is commonly linked to a lysosome-dependent form of cell death (Boya 2011), which is relevant to the loss of neurons in various neurodegenerative diseases (Nixon et al. 2008; Settembre et al. 2008a,b; Dehay et al. 2010; Rubinsztein et al. 2011). A neuronal cell death with type 2 morphology is seen in lysosomal storage disorders owing to defects at autolysosomal stages of autophagy—loss of function of lysosomal hydrolases or structural proteins (Hartmann et al. 2000; Koike et al. 2000; Ko et al. 2005; Willenborg et al. 2005; Nixon et al. 2008; Settembre et al. 2008a; Wolfe and Nixon 2012). For example, cathepsin D is ubiquitously expressed, yet mutations that markedly lower cathepsin D activity affect the brain disproportionately causing profuse AV accumulation in neurites and progressive neuronal loss in the neocortex and hippocampus (Tyynela et al. 2000). Lysosomal dysfunction leading to the destabilization of lysosomal membranes and death of nonneuronal cells is well established (Brunk and Svensson 1999; Boya and Kroemer 2008; Boya 2011) and likely relevant to neurons in neurological disorders (Nixon et al. 2008; Dehay et al. 2010), although the evidence is mostly in vitro at present. Autophagy failure at early stages of this pathway may also induce cell death although possibly with less florid autophagic pathology. Deleting factors critical for autophagy induction or autophagosome formation such as FIP200, Atg5, or Atg7 has been shown to induce delayed neuronal cell death and cytoplasmic accumulation of ubiquitinated proteins or organelles (Hara et al. 2006; Komatsu et al. 2006, 2007; Liang et al. 2010).

Autophagy failure, depending on where the defect is along the pathway, can trigger neuronal cell death in several ways. When proteolytic clearance steps are compromised, autolysosomes/lysosomes accumulate mutant and oxidized proteins, protein oligomers and aggregates, damaged organelles, and other incompletely digested products, certain of which increase the permeability of lysosomal membranes causing hydrolases to be released into the cytoplasm, in some cases even from otherwise intact lysosomes (Kroemer and Jaattela 2005). A long list of exogenous agents, including many anticancer drugs (Erdal et al. 2005), are able to disrupt lysosomal membrane integrity directly and induce rapid lysosome-dependent cell death. Similarly, the list of endogenous factors able to induce lysosomal membrane permeabilization (LMP) is growing and includes ceramide, sphingosine, oxidized lipids or lipoproteins, reactive oxygen species (ROS), calpains, certain caspases, and a few proteins/peptides implicated in AD, such as Aβ and ApoΕ (Brunk and Svensson 1999; Yuan and Yankner 2000; Boya and Kroemer 2008; Johansson et al. 2010; Repnik et al. 2012). Agents that cause cataclysmic disruption of lysosomal membranes are likely to induce rapid necrosis during which released hydrolases participate as both a trigger and as executioners along with caspases that are activated by cathepsin-mediated cleavage (Hayashida et al. 1993; Werneburg et al. 2007). Slower release of cathepsins from lysosomes may first activate apoptotic cascades via the caspase-mediated cleavage of BID releasing of mitochondrial cytochrome c (Guicciardi et al. 2000), degradation of antiapoptotic Bcl-2 homologs, and activation of Bax that releases mitochondrial apoptosis-initiating factor (AIF) (Bidere et al. 2003) and can also induce LMP (Werneburg et al. 2004; Kilinc et al. 2010). In a pathological situation in which either autophagy induction, substrate recognition, or sequestration are impaired, the resultant increase in numbers of damaged mitochondria can trigger apoptosis through the intrinsic pathway and via ROS generation that oxidizes membrane lipids and destabilizes the lysosome membrane. Reduced autophagic elimination of other proapoptotic factors, such as activated caspases, may also accelerate apoptosis under these conditions (Yang et al. 2008).

AUTOPHAGY IN MAJOR NEURODEGENERATIVE DISEASES

Autophagy-related pathology has been noted in late-onset neurodegenerative diseases including AD (Cataldo et al. 1997; Nixon et al. 2005; Nixon and Cataldo 2006), PD (Anglade et al. 1997), amyotrophic lateral sclerosis (ALS) (Hart et al. 1977; Nakano et al. 1993; Sasaki 2011), Huntington’s disease (HD) (Roos et al. 1985; Rudnicki et al. 2008), and several others (Liberski et al. 1995; Yue et al. 2002; Rudnicki et al. 2008). The pathology in different disease states has so far been rarely assessed quantitatively. The extent of autophagy waste “storage” in perikarya and neuronal processes and the nature of the AV subtypes that accumulate may differ significantly across these diseases, which in some cases provide useful clues to the underlying disturbance of autophagy. Combined with biochemical analyses of cell and animal models, a picture is emerging from studies of different disorders that autophagy may be altered at early stages (e.g., induction and autophagosome formation) or at late stages (e.g., proteolytic clearance of autophagosomes by lysosomes), or both, which has a different impact on the pattern of neurodegeneration.

PD

In PD, the death mainly of dopaminergic neurons in the substantia nigra is associated with accumulation of the cytosolic protein α-synuclein within inclusions called Lewy bodies. The PD cell death pattern shows features of apoptosis and necrosis as well as increased numbers of autophagosomelike structures (Fig. 3) (Stefanis 2005). Autophagy defects at early stages of autophagy as well as at later lysosomal clearance stages contribute to this complex pattern. Mutations of Parkin (Park2) and PINK1 account for the majority of autosomal recessive cases of parkinsonism (Gasser 2009). Both proteins operate in a pathway that controls mitochondrial fusion-fission events (Yao and Wood 2009) but also play roles in mitophagy that are critical to PD pathogenesis. PINK1, a serine/threonine kinase, localizes to the outer mitochondrial membrane and serves as a sensor of mitochondrial membrane polarization. On normal mitochondria, PINK1 is proteolytically cleaved constitutively (Abeliovich 2010), but it becomes selectively stabilized on the outer mitochondrial membrane when mitochondrial function is compromised. This enables Parkin (a cytoplasmic E3 ubiquitin ligase) to bind and ubiquitinate certain exposed membrane proteins, which recruits p62 and LC3, that initiate mitophagy of the damaged organelle (Geisler et al. 2010; Narendra et al. 2010; Wild and Dikic 2010). It is believed that PD-causing mutations of PINK and Parkin impede mitophagy, causing damaged mitochondria to accumulate and potentially initiate apoptotic events. Additional influences of defective PINK on mitophagy and other mitochondrial dynamics, such as fission–fusion events and ROS production (Wood-Kaczmar et al. 2008; Dagda et al. 2009; Michiorri et al. 2010), may add complexity to the net role mitophagy failure plays in PD. Mutations of α-synuclein in familial PD also disrupt substrate access during CMA by binding abnormally tightly to LAMP2A, thereby blocking not only its own uptake into lysosomes but that of other CMA substrates (Cuervo et al. 2004). Wild-type α-synuclein modified by dopamine, seems to cause similar CMA dysfunction suggesting that this form of α-synuclein toxicity applies to sporadic as well as familial forms of PD (Xilouri et al. 2009).

Figure 3.

Neuropathology in PD. A neuron of the substantia nigra from a case of juvenile parkinsonism. The pale central area of the perikaryon, next to a Lewy body (asterisk), shows disappearance of rough ER and mitochondria accumulation, which may imply impaired mitophagy. (Panel from Hayashida et al. 1993; reproduced, with permission, from Springer Science + Business Media.) (Inset) A dopaminergic neuron from a PD case, filled with neuromelanin-containing vesicles, shows a shrunken nucleus with highly condensed, marginalized chromatin (arrows). Scale bar, 1 µm. (Panel from Hartmann et al. 2000; reproduced, with permission, from the National Academy of Sciences.)

α-Synuclein is also a substrate for macroautophagy (Webb et al. 2003), which may be recruited as a compensatory mechanism when CMA is impaired. Macroautophagy stimulation by beclin 1 gene transfer, ATG overexpression, or rapamycin significantly ameliorates pathology in α-synuclein PD models (Spencer et al. 2009; Yu et al. 2009). Certain PD-related disease factors, however, may also disrupt lysosomal function and the clearance of substrates by macroautophagy. One example is 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), a well-characterized neurotoxin model of PD induced by mitochondrial dysfunction and oxidative stress (Yokoyama et al. 2008; Bezard and Przedborski 2011). Lysosome number drops in dopaminergic neurons very early after MPTP exposure, related in part to LMP and cytosolic release of cathepsins—events that are partially prevented by cathepsin inhibitors. Autophagosome accumulation accompanies lysosomal depletion. All of these events and neurodegeneration are attenuated by genetic or pharmacological activation of transcription factor EB (TFEB) or rapamycin, both of which increase lysosomal biogenesis (Dehay et al. 2010) in addition to inducing autophagy. Other factors contributing to general lysosomal dysfunction in some forms of PD include loss-of-function mutations of the lysosomal ATPase, ATP 13A2, a rare cause of PD (Ramirez et al. 2006; Di Fonzo et al. 2007), and mutations in the lysosomal hydrolase glucocerebrosidase that cause the lysosomal storage disorder Gaucher disease and also increase the risk of PD (Sidransky et al. 2009). The neuroprotective effects of enhancing autophagy in most PD models (Dagda et al. 2009; Dadakhujaev et al. 2010), but not all (Zhu et al. 2007), supports the pathogenic importance of autophagy defects observed in PD and PD mouse models.

Polyglutamine Expansion Diseases

Polyglutamine disorders are adult-onset progressive neurodegenerative diseases caused by expansion of a cytosine-adenine-guanine (CAG) repeat motif within the coding region of various genes. The respective proteins containing the long polyglutamine tracts typically form aggregates in the cytoplasm. Generally the age at disease onset correlates with the increased number of CAG repeats. These aggregate-prone proteins tend to be good macroautophagy substrates, and blocking autophagy leads to their increased aggregation and neurotoxicity when the protein is overexpressed (Ravikumar et al. 2004). Besides HD, the first repeat expansion disease discovered to be associated with defective autophagy as discussed below, AVs are increased in animal models of spinocerebellar ataxia 1 and 7, and spinal and bulbar muscular atrophy and neuroprotective effects of autophagic enhancement indirectly imply a role of autophagy in the disease pathogenesis (La Spada and Taylor 2010).

HD, the best known of the polyglutamine disorders, is caused by mutations of the Huntingtin (Htt) gene, encoding a 350-kDa protein that is ubiquitously expressed even though HD principally affects the brain and particularly the striatum. Although the relative neurotoxicity of aggregated versus nonaggregated Htt remains controversial, soluble forms of mutant Htt are increasingly implicated as the toxic species (Dunah et al. 2002; Arrasate et al. 2004; Cui et al. 2006). Several caspases and calpains, known to be activated in HD brain, generate toxic amino-terminal fragments of Htt most closely linked to HD pathogenesis (Ona et al. 1999; Sanchez et al. 1999; Gafni and Ellerby 2002; Graham et al. 2006). Caspase 6 cleavage of mutant Htt seems particularly important in mediating the sensitivity of HD striatal neurons to excitotoxicity (Graham et al. 2006), although polyglutamine expansion also impedes various antiapoptotic effects of Htt including inhibitory effects of caspase 3 activation (Gervais et al. 2002; Zhang et al. 2006). Although activation of apoptotic pathways is likely, the typical morphologic pattern of apoptosis is incomplete (Hickey and Chesselet 2003).

Autophagy was initially implicated in HD from the presence of abnormal endosomes and AV accumulations in HD neurons (Tellez-Nagel et al. 1974) and by the similar morphologies seen in cell models after mutant Htt is overexpressed (Kegel et al. 2000). Htt aggregates, which are poor UPS substrates (Jana et al. 2001), can be cleared by autophagy. Autophagy inducers accelerate their clearance while improving toxicity phenotypes in cell and animal models of HD, whereas inhibiting autophagosome formation or fusion with lysosomes promotes inclusion formation and cell death (Ravikumar et al. 2004). Aggregate-prone Htt seems to be a selective macroautophagy target because autophagy modulation does not affect wild-type Htt levels. Interestingly, acetylation of mutant Htt targets it to autophagosomes (Jeong et al. 2009) and selective clearance of mutant Htt can be achieved by hyperacetylation of the mutant protein with HDAC (histone deacetylase) inhibitors, which are neuroprotective in various HD models (Kazantsev and Thompson 2008).

The efficacy of autophagy inducers in clearing overexpressed mutant Htt from autophagosomes suggests that lysosomal clearance mechanisms are not the primary site of autophagy disruption in HD. Evidence from HD tissue and disease models points instead to defective recognition of cargoes during early sequestration steps. In several HD models, autophagosomes form properly and are cleared even though protein turnover by macroautophagy is paradoxically slow (Martinez-Vicente et al. 2010). The autophagosomes in these cells, however, appear relatively devoid of cargo, suggesting a possible failure of autophagosome membranes to properly engage the substrates during sequestration. In this regard, mutant Htt aggregates sequester beclin 1, which can interfere with autophagosome membrane nucleation (Shibata et al. 2006) and perhaps promote derangement of the sequestration process. Deficient autophagy would be expected to compound effects of Htt-related alterations of mitochondrial function, including loss of antiapoptotic functions of Htt when mutated in HD (Gil and Rego 2008).

ALS and Related Disorders

ALS, the prototypical degenerative disease of motor neurons, is sporadic in approximately 90% of cases but the remaining cases are familial, caused by mutations of at least three different genes, namely, the Cu/Zn superoxide dismutase-1 gene (SOD-1), the TAR DNA-binding protein 43 gene (TARDBP or TDP-43), and the fused in sarcoma/translation in liposarcoma gene (FUS/TLS) (Liscic and Breljak 2011). The recent finding that ALS also arises from mutations of TDP-43, a major cause of frontotemporal dementia (FTD), has highlighted the overlapping clinicopathological features of ALS, FTD-MND (FTD with motor neuron disease), or FTD-U (FTD with ubiquitin-immunoreactive inclusions). In the latter disorders, ubiquitinated inclusions appear in the cytoplasm replacing the normal nuclear staining of TDP-43 (Hirano 1996; Nakano 2000; Mackenzie 2007; Strong 2008; Strong et al. 2009; Geser et al. 2010).

A significant autophagy response occurs very early in ALS. Autophagosomes are frequently seen in otherwise normal-appearing perikarya of motor neurons in sporadic ALS patients (Sasaki 2011) and significantly are elevated by the time clinical symptoms appear (Venkatachalam et al. 2008; Pasquali et al. 2009). Autophagosomes and autolysosomes appear in close association with p62-positive inclusions and both become more frequent in degenerating neurons (Sasaki 2011). The high number of autophagosomes may imply elevated autophagy induction given the decreased mTOR phosphorylation seen in several genetic ALS models (Morimoto et al. 2007; Li et al. 2008; Hetz et al. 2009). Although damaged mitochondria increase in motor neurons in SOD1 mouse ALS models (Liu et al. 2004; Martin et al. 2007; Vande Velde et al. 2008), these are frequently in AVs, suggesting that mitophagy is competent (Wong et al. 1995). Investigations to date have not identified a significant impairment of cargo recognition, sequestration, or autophagosome formation.

Alternatively, or in addition to the possibility of increased autophagy induction, defective autophagosome clearance is also consistent with this morphological pattern and with other data, especially given that neurons usually clear autophagosomes efficiently (Boland et al. 2008). Indeed, growing evidence points to a defect in autophagosome clearance owing to impaired fusion of endosomes, MVP, or lysosomes in some forms of ALS. Defects in the ESCRT machinery that regulates fusion of endosomes/MVB with autophagosomes to produce amphisomes (Filimonenko et al. 2007) have been implicated in frontotemporal dementia linked to chromosome 3 (FTD3) (Skibinski et al. 2005) and ALS (Momeni et al. 2006; Parkinson et al. 2006). Cell models of these diseases have shown that loss of function of various ESCRT genes disrupts autophagosome maturation by impeding autophagosome fusion with lysosomes or endosomes. For example, mutations in charged multivesicular body protein-2B (CHMP2B) in familial ALS disrupt ESCRT pathway function causing aggregates of ubiquitinated proteins and p62 to accumulate (Filimonenko et al. 2007; Lee et al. 2007). Loss-of-function mutations of dynactin, a component of the dynein complex regulating retrograde axonal transport, impair autophagy by disrupting autophagosome retrograde transport and fusion with lysosomes (Ravikumar et al. 2005; Laird et al. 2008).

AD

AD is defined by the presence of “senile or neuritic plaques” consisting of β-amyloid deposits surrounding foci of degenerating or dystrophic axons and dendrites and intraneuronal filamentous aggregates of the microtubule-associated protein tau. AD leads to widespread death of many subtypes of neurons beginning in the hippocampus and spreading progressively to the cortex in a circuit-selective pattern. Neuronal death in AD does not conform to a conventional cell death pattern and may vary in its features among neuron subtypes, showing in various proportions the features of apoptosis (e.g., activation of multiple caspases), necrosis (e.g., calpain activation and cathepsin up-regulation), and florid autophagic/lysosomal pathology.

Abnormalities of the lysosomal system in AD (Fig. 4) are a continuum that includes endosome anomalies associated with endocytic pathway up-regulation and increased lysosome biogenesis—the earliest appearing disease-specific cellular pathologies in the disease (Nixon et al. 2006). AVs progressively accumulate profusely in affected neurons and become the predominant organelles within enormously swollen dystrophic neurites, a hallmark of AD neuropathology. The selective accumulation of AVs of all types within the dystrophic swellings likely reflects selective and specific impairment of the axonal transport of autophagy/lysosomal-related compartments (Lee et al. 2011). Neuritic dystrophy in AD is much more extensive than that seen in other aging-related neurodegenerative diseases, and the huge mass of accumulated waste protein within AVs is reminiscent of that seen in some primary lysosomal storage disorders (Nixon et al. 2008).

Figure 4.

Autophagic pathology and neurodegeneration in AD brains. (A) Cathepsin D (Cat D) immunocytochemistry of AD neocortex reveals that many amyloid plaques are cathepsin D positive (arrows). (B) A plaque of higher magnification depicts numerous Cat D-positive dystrophic neurites. (C) Electron microscopy shows that dystrophic neurites are filled mainly with autophagosomes and autolysosomes containing undigested materials. (D) A tangle-bearing neuron showing scattered bundles of paired helical filaments (arrow and inset) and a peripherally displaced but otherwise normal nucleus. (E) The boxed area of D is shown at higher magnification. The cytoplasm contains numerous AVs including double-membrane dense structures and multilamellar bodies (arrows) as well as many small dense bodies or lysosomes (arrowheads). (Panels C–E from Nixon et al. 2005; reproduced, with permission, from the author.) (F,G) Neurons in AD neocortex immunolabeled with Cat D antibodies depicting a later stage of degeneration with altered nuclear morphology and accumulation of hydrolase-positive lipofuscin or lipofuscin-containing autolysosomes (arrows). (F) The same neocortical pyramidal neuron showing atrophic changes by Nissl staining (F, inset; scale bar, 10 µm) contains AVs and many lipofuscin granules when examined ultrastructurally. The nucleus is highly electron dense and the morphology of the whole cell suggests dark-neuron degeneration, whereas the nucleus of a neuron in (G) shows chromatin clumping and marginalization, suggesting ongoing apoptotic changes. (G, inset) shows hydrolase-positive aggregates of lipofuscin granules. (Panels F and G from Cataldo et al. 1994; reprinted, with permission, from Elsevier.)

Current evidence indicates that autophagy is principally defective at the stage of autolysosomal proteolysis in AD. The presence of abundant mature autophagosomes containing partially digested substrates and cathepsins (Nixon et al. 2005) contrasts with the exceptional efficiency of AV clearance in normal neurons (Boland et al. 2008). Neuritic dystrophy and selective AV accumulation can be reproduced by blocking lysosomal proteolysis pharmacologically or genetically in in vitro and in vivo models (reviewed in Nixon and Yang 2011), whereas they are not seen when autophagy is strongly induced in otherwise healthy neurons (Boland et al. 2008; Lee et al. 2011). Lysosomal proteolysis failure is further supported by the recently identified role of presenilin 1 (PS1) as a chaperone essential for the delivery of the proton pump, vacuolar vATPase, to lysosomes, which is essential for lysosome acidification and protease activation. Mutations of PS1, the most common cause of early-onset familial AD (Sherrington et al. 1995), lead to markedly defective lysosomal acidification and autolysosomal maturation (Lee et al. 2010), explaining why PS1 mutations potentiate autophagic/lysosomal, amyloid, and tau pathologies as well as accelerate neuronal cell death in patients with PS-FAD or mouse models (Cataldo et al. 2004).

Inheritance of the ε4 allele of the apolipoprotein E gene (ApoE4) is the strongest genetic risk factor for late-onset AD. ApoE4 has numerous biological activities but prominent among them are allele-specific effects on cholesterol regulation that promote the specific endocytic dysfunction arising in early AD (Cataldo et al. 1996, 2000). Moreover, a unique proteolytically processing pattern of ApoE4 in lysosomes yields a “molten globule” structure that induces reactive intermediates, which destabilize lysosomal membranes leading to lysosomal leakage and apoptosis (Ji et al. 2002, 2006; Mahley and Huang 2006). The expression of ApoE4, but not the ApoE3 allele, increases levels of intracellular Aβ peptide (Aβ), enlarges lysosomes and alters their morphology in a mouse AD model, and causes neurodegeneration of neurons typically vulnerable in AD (Belinson et al. 2008). Similarly, overexpression of human Aβ 42, but not Aβ 40, in Drosophila neurons induces age-related autophagic/lysosomal dysfunction and neurotoxicity (Ling et al. 2009) believed to arise from lysosomal membrane destabilization mediated directly by Aβ (Yang et al. 1998; Glabe 2001) or by oxidized, incompletely degraded autophagic substrates (Terman and Brunk 2006; Kurz et al. 2008).

The prominent defects in clearance of autophagic substrates by lysosomes may be compounded by an increased induction of autophagy in AD, which is known to accelerate autophagy pathology (Boland et al. 2008) and increase neurotoxicity. Lysosome biogenesis in AD brain and AD mouse models (Cataldo et al. 1995, 1996, 2004; Ginsberg et al. 2010) is up-regulated as is the transcription of a wide range of autophagy-related genes in the AD cortex (Lipinski et al. 2010) and in CA1 hippocampal neurons specifically (S Ginsberg and R Nixon, unpubl.). Some analyses of mTOR signaling, although not all, are consistent with increased autophagy induction (Bhaskar et al. 2009; Caccamo et al. 2010; Nixon and Yang 2011). Although an initial report of lowered beclin 1 levels in brain suggested that early steps of autophagy may be impaired (Pickford et al. 2008; Jaeger et al. 2010), this finding has not been confirmed in a later more detailed analysis (R Nixon, P Mohan, and E Masliah, unpubl.).

The significance of autophagy failure to the pathogenesis of AD is underscored by the efficacy of autophagy enhancement in ameliorating AD-related pathologies (e.g., Aβ deposition and tau pathology) and synaptic and cognitive deficits in AD models (Karuppagounder et al. 2009; Ahmed et al. 2010; Greco et al. 2010; Spilman et al. 2010; Majumder et al. 2011; Tian et al. 2011; Yang et al. 2011). Interestingly, in one recent study, rapamycin administration before development of AD pathology and presumably before lysosomal failure delayed the onset and diminished the severity of the AD phenotype, whereas administration after the appearance of pathology had little beneficial effect, highlighting the importance of lysosomal failure (Majumder et al. 2011). The therapeutic efficacy of genetic manipulations selectively targeting the lysosome to increase its proteolytic efficiency (Sun et al. 2008; Yang et al. 2011) has established the particular importance of lysosomal proteolysis defects in promoting disease progression.

CONCLUDING REMARKS

Investigations of autophagy in neurodegenerative disease have greatly intensified since the first reviews appeared on this subject a decade ago (Nixon et al. 1995). Autophagy is now firmly established as a vital homeostatic and quality control mechanism in healthy neurons and as a cytoprotective response when further induced in chronic neurodegenerative diseases. Autophagy protects neurons against metabolic challenges by generating supplies for energy production and cell repair, eliminating toxic damaged proteins and organelles, and suppressing apoptotic signaling. With increasing cellular age, autophagy becomes less efficient despite a greater need to eliminate mounting levels of damaged proteins, organelles, and other apoptosis-inducing stimuli. In AD, PD, and likely other disorders, progressive autophagy dysfunction, including impairment driven by causative gene mutations or other disease risk factors, compounds the adverse effects of aging on autophagy in promoting neuronal cell death. Defects in autophagy that impair the early stages of this process—induction or cargo recognition and sequestration—cause autophagic substrates, including mitochondria, to accumulate in the cytoplasm and trigger apoptosis, as in PD. Failure at later stages of autophagy related to the efficient digestion of sequestered substrates leads to the buildup of partially digested toxic products in autolysosomes and lysosomes, as seen in AD. In this situation, consequent injury and destabilization of lysosomal membranes and leakage of cathepsins can trigger either apoptotic or necrotic cascades (Fig. 5). Extensive cross talk between regulators of autophagy and apoptosis has provided insight into how autophagy protects against apoptosis and how the balance between these opposing forces contributes to the complex patterns of neuronal cell death seen in nervous system diseases. Although much more work is needed to clarify these mechanisms, recent preclinical investigations in neurodegenerative disease models have provided glimpses that autophagy modulation may be a fruitful therapeutic strategy.

Figure 5.

Possible states of impairment in the autophagy pathway in different neurodegenerative diseases and possible links to neuronal cell death. The autophagy defect may conceivably promote neuronal cell death via two possible mechanisms: (1) impaired mitophagy resulting in accumulation of damaged mitochondria and mitochondrial membrane permeability (MMP) leading to cytochrome c release and apoptotic cell death; and (2) impaired lysosomal clearance of autophagy substrate leading to lysosomal membrane permeabilization (LMP) and cathepsin release into cytosol, thereby inducing either apoptotic or necrotic cell death. Increased autophagy induction in the face of a downstream block in the pathway may be a counterproductive response. AP, autophagosome; CMA, chaperone-mediated autophagy; Cyto c, cytochrome c; Ly, lysosome; MA, macroautophagy.