Initiation of autophagosome formation

Autophagosome formation appears to start at phagophore (autophagosome precursor; also known as pre-autophagosomal structures)-assembly sites (PAS). The source of the membrane is unclear, although recent data support a contribution from the endoplasmic reticulum (ER) (Axe et al., 2008). The formation of the phagophore requires the activity of the class-III phosphoinositide 3-kinase (PI3K) Vps34, which forms phosphatidylinositol 3-phosphate [PtdIns(3)P]. Chemical inhibition of Vps34 by 3-methyladenine or wortmannin blocks the formation of new autophagosomes. Vps34 is part of a large macromolecular complex specific for autophagy that contains the proteins Atg6 (also called beclin 1), Atg14 and Vps15 (p150) (Itakura et al., 2008). Other proteins that are involved in the initiation stage of autophagosome formation include Atg5, Atg12, Atg16, the newly identified autophagy-related protein focal adhesion kinase (FAK)-family interacting protein of 200 kD (FIP200), which interacts with Atg1 (also called ULK1) (Hara et al., 2008), and the mammalian orthologue of Atg13 (Chan et al., 2009; Hosokawa et al., 2009; Jung et al., 2009).

Autophagosome elongation

Atg5 and Atg12 are involved in the first of two ubiquitylation-like reactions that control autophagy. Atg12 is conjugated to Atg5 in a reaction that requires Atg7 [ubiquitin-activating-enzyme (E1)-like] and Atg10 [ubiquitin-conjugating-enzyme (E2)-like]. Atg5-Atg12 conjugates are localised onto the PAS and dissociate upon completion of autophagosome formation. The process of Atg5-Atg12 conjugation depends on Vps34 function, and Vps34 activity, along with autophagy, is positively regulated by the small GTPase Rab5 (Ravikumar et al., 2008). Atg5-Atg12 interacts non-covalently with Atg16L (Atg16-like) to form a complex of approximately 800 kDa (Ohsumi and Mizushima, 2004).

The second ubiquitylation-like reaction involves the conjugation of microtubule-associated protein 1 light chain 3 (MAP1-LC3; also known as Atg8 and LC3) to the lipid phosphatidylethanolamine (PE). LC3 is cleaved at its C-terminus by Atg4 to form cytosolic LC3-I, which is covalently conjugated to PE to form membrane-associated LC3-II, a process that requires the activities of Atg7 (E1-like) and Atg3 (E2-like) (Ohsumi and Mizushima, 2004). The Atg5-Atg12 conjugate seems to modulate LC3-I conjugation to PE by acting in an E3-like fashion (Hanada et al., 2007) and the Atg5-Atg12-Atg16L complex also determines the sites of LC3 lipidation (Fujita et al., 2008). In this way, LC3 is specifically targeted to Atg5-Atg12-associated membranes, which are expanded phagophores.

Crosstalk between the two ubiquitylation-like systems has also been suggested. Atg10 can interact with LC3-I and facilitate LC3-I conjugation to PE (Nemoto et al., 2003). Similarly, Atg3 co-immunoprecipitates with Atg12, and overexpression of Atg3 increases Atg5-Atg12 conjugation (Tanida et al., 2002). Although Atg5-Atg12 is lost upon completion of autophagosome formation, LC3-II remains associated with autophagosomes even after fusion with lysosomes, after which LC3-II on the cytosolic face of autolysosomes can be delipidated and recycled (to LC3-I). Conjugated LC3 is a substrate for Atg4, and Atg4 delipidates LC3-II and removes it from the membrane. Furthermore, the production of reactive oxygen species (ROS; specifically H₂O₂) during starvation results in oxidation and inactivation of Atg4. This inhibits LC3-II delipidation and might promote autophagosome formation (Scherz-Shouval et al., 2007).

LC3-II is the only known protein that specifically associates with autophago-somes and not with other vesicular structures. Thus, LC3-II levels correlate with autophagic vacuole numbers, which can also be assessed by scoring LC3-positive vesicle numbers (Kabeya et al., 2000). LC3-II can mediate membrane tethering and hemifusion, and these functions might be crucial for the expansion of autophagosome membranes. These data also raise the possibility that LC3 assists the final fusion of the pre-autophagosomal double-membrane `cups' into fused vesicles (Nakatogawa et al., 2007).

Autophagosome maturation and fusion

Autophagosomes form randomly in the cytoplasm. They are then trafficked along microtubules in a dynein-dependent manner to lysosomes, which are clustered around the microtubule-organising centre (MTOC; located near the nucleus). There, autophagosomes tether, dock and fuse with lysosomes, and the contents of the two vesicles are mixed (Jahreiss et al., 2008). The details of this aspect of mammalian autophagy are still unclear, although the SNARE proteins that are involved in yeast autophagosome-vacuole fusion have been characterised (Ishihara et al., 2001). En route to fusion with the lysosomes, autophagosomes can fuse with endosomes to form amphisomes. It is currently not clear whether amphisome formation is a prerequisite for autophagosome-lysosome fusion. Reduced activity of some ESCRT (endosomal sorting complex required for transport) proteins disrupts autophagosome-lysosome fusion, resulting in the accumulation of autophagosomes (Skibinski et al., 2005). Similarly, LAMP2 deficiency has been associated with the accumulation of autophagosomes, which is consistent with a defect in fusion (Eskelinen et al., 2002).

mTOR-dependent pathway

One of the classical pathways that regulate autophagy is controlled by the mammalian target of rapamycin (mTOR). mTOR activity inhibits mammalian autophagy, but its activity is inhibited under starvation conditions, which contributes to starvation-induced autophagy. Autophagy is also induced by inhibiting mTOR with drugs such as rapamycin (Rubinsztein et al., 2007). Mammalian Atg13, ULK1 and ULK2 have recently been identified as direct targets of mTOR (Chan et al., 2009; Hosokawa et al., 2009; Jung et al., 2009). Atg13 binds to ULK1 and ULK2 (ULK1/2) and mediates their interaction with FIP200. Under nutrient-rich conditions, mTOR is associated with this complex. Inhibition of mTOR leads to its dissociation from the complex and results in the partial dephosphorylation of Atg13 and ULK1/2; this activates ULK1/2 to phosphorylate FIP200 and thereby induce autophagy (Chan et al., 2009; Hosokawa et al., 2009; Jung et al., 2009).

Several signalling molecules regulate autophagy via mTOR. Insulin or insulin-like growth factor (IGF) signalling activates mTOR via PI3K and Akt. Activation of adenosine-monophosphate-activated protein kinase (AMPK) inhibits mTOR activity via tuberous sclerosis complex 1 (Tsc1), Tsc2 and Ras homology enriched in brain (Rheb) (Hall et al., 2008). p53, a commonly mutated gene in human cancers, can positively and negatively regulate autophagy via the mTOR pathway. Oncogenic or genotoxic stress stabilises and activates p53. This can stimulate autophagy by activating AMPK or by upregulating phosphatase and tensin homologue (PTEN) and Tsc1. Genetic or chemical inhibition of p53, however, also activates autophagy (Levine and Abrams, 2008).

Recently, further insights have been provided into the mechanisms behind starvation-induced autophagy. Autophagy can be inhibited by the binding of the apoptosis-related proteins B-cell lymphoma 2 (Bcl2) or basal-cell lymphoma-extra large (Bcl-XL) to beclin 1 (Atg6). Starvation induces Jun N-terminal kinase 1 (Jnk1) activity, which phosphorylates Bcl2, thereby disrupting the interaction between beclin 1 and Bcl-2 and inducing autophagy (Wei et al., 2008). This mechanism might also account for the upregulation of autophagy after proteasome inhibition or ER stress, as one study showed that ER-stress-induced autophagy is Jnk1-dependent and that proteasome inhibition can induce ER stress (Ding et al., 2007). Several other beclin-1 binding partners such as activating molecule in beclin-1-regulated autophagy (Ambra), Rab5, ultraviolet-radiation resistance-associated gene (UVRAG) and beclin-1-associated autophagy-related key regulator (Barkor) can activate beclin 1.

mTOR-independent pathway

An mTOR-independent pathway for the induction of autophagy has recently been discovered (Sarkar et al., 2005). In this pathway, the inhibition of inositol monophosphatase (IMPase) reduces free inositol and myoinositol-1,4,5-triphosphate (IP3) levels, which leads to an upregulation of autophagy. Lithium, carbemazepine and valproate – drugs that are used to treat a range of neurological and psychiatric conditions – induce autophagy via this pathway (Sarkar et al., 2005). Further studies have shown that this pathway is regulated by intracellular Ca²⁺ levels and cyclic AMP (cAMP), and have revealed additional drugs that might induce autophagy, such as verapamil (an L-type Ca²⁺-channel antagonist) and clonidine (an imidazoline-receptor agonist that reduces cAMP levels) (Williams et al., 2008). In this pathway, autophagy is inhibited when intracellular cAMP levels are increased by adenylyl cyclase (AC), which influences autophagy by activating the guanine nucleotide exchange factor Epac. Epac then activates Rap2B, which inhibits autophagy by activating phospholipase Cϵ (PLCϵ), resulting in the production of IP3, which mediates the release of Ca²⁺ from ER stores. Increased intracytosolic Ca²⁺ blocks autophagy by activating calpains (a family of Ca²⁺-dependent cysteine proteases), which mediate their effects on autophagy through Gsα, which is activated after calpain cleavage. This, in turn, increases AC activity to increase cAMP levels, thus forming a loop. Other compounds that induce autophagy include ceramide, sphingosine, trehalose, fluspirilene, trifluoperazine, pimozide, nicardipine, penitrem A, niguldipine and amiodarone (Zhang et al., 2007), but their mechanisms of action are not well understood.

Starvation

One of the ancestral purposes of autophagy has been as a means of recycling macromolecules to provide new nutrients at times of starvation. This is a key role for autophagy in single-cell organisms such as yeast, and is a crucial function in mammals as well. For instance, in the immediate newborn period there is relative starvation before breastfeeding is established. Autophagy-deficient mice are apparently normal at birth but die soon thereafter, owing to an inability to recycle nutrients (Kuma et al., 2004).

Autophagy appears to be important for mammalian development as well. Fertilisation induces autophagy, and autophagy-defective oocytes derived from oocyte-specific Atg5-knockout mice failed to develop beyond the four- and eight-cell stages if they were fertilised by Atg5-null sperm (Tsukamoto et al., 2008). Autophagy also appears to regulate the clearance of apoptotic corpses during development [for instance, during embryonic cavitation (Qu et al., 2007)].

Aggregation-prone proteins

The key proteolytic systems in mammalian cells are autophagy and the ubiquitin-proteasome system. The proteasome can only process unfolded proteins, as it has a narrow opening; however, oligomers and organelles, which cannot enter the proteasome, are accessible to autophagy. Indeed, autophagy appears to be a crucial route for the degradation of aggregation-prone proteins. These include neuro-degenerative-disease-associated proteins such as mutant huntingtin (which causes Huntington's disease), and mutant forms of α-synuclein (which causes forms of familial Parkinson's disease). The clearance of such proteins is impaired when autophagy is compromised. However, autophagy upregulation is a possible therapeutic strategy for such diseases, as drugs that induce autophagy by mTOR-dependent or mTOR-independent routes enhance the clearance of such mutant proteins and alleviate their toxicity in a range of animal models (Sarkar et al., 2008).

This function of autophagy might also be relevant physiologically, because protein aggregates form in the brain and some other tissues (from endogenous `normal' proteins) in Atg5or Atg7-knockout mice, which have compromised autophagy (Hara et al., 2006; Komatsu et al., 2006). Such a mechanism might also contribute to other neurodegenerative diseases. For instance, recent studies suggest that autophagosome-lysosome fusion is impaired in a form of frontotemporal dementia that is caused by mutations in a gene in the ESCRT complex, and by mutations in the dynein-complex component dynactin that cause a form of motor neuron disease (Munch et al., 2004; Skibinski et al., 2005). This will impair the clearance of autophagic substrates and predispose cells to aggregate formation.