Autophagy is a conserved cellular degradative process important for cellular homoeostasis and survival. An early committal step during the initiation of autophagy requires the actions of a protein kinase called ATG1 (autophagy gene 1). In mammalian cells, ATG1 is represented by ULK1 (uncoordinated-51-like kinase 1), which relies on its essential regulatory cofactors mATG13, FIP200 (focal adhesion kinase family-interacting protein 200 kDa) and ATG101. Much evidence indicates that mTORC1 [mechanistic (also known as mammalian) target of rapamycin complex 1] signals downstream to the ULK1 complex to negatively regulate autophagy. In this chapter, we discuss our understanding on how the mTORC1–ULK1 signalling axis drives the initial steps of autophagy induction. We conclude with a summary of our growing appreciation of the additional cellular pathways that interconnect with the core mTORC1–ULK1 signalling module.
- amino-acid starvation
Autophagy is a conserved cellular process whereby damaged organelles and cytosolic proteins are degraded using double membrane-enclosed vesicles, known as autophagosomes . The term ‘autophagy’ was coined in 1963 by Christian de Duve when he described the presence of lysosomal-like vesicles containing degraded cellular proteins and organelles. Since this discovery, three main types of autophagy have been described: chaperone-mediated autophagy, micro-autophagy and macro-autophagy. The most well understood of these is macro-autophagy, which will be the focus of this chapter and hereby simply referred to as autophagy. Autophagic degradation maintains cellular homoeostasis and promotes survival in almost every mammalian cell and this fundamental mechanism is integral to normal physiology and multiple disease states as summarized elsewhere [2–4]. In the present chapter, we discuss our understanding of the signalling events that are engaged when a cell activates autophagy.
The overall degradative flux of autophagy can be divided into three phases (Figure 1): first, sequestration, in which cellular components are captured by double-bilayer membranes to form the autophagosome; secondly, transport of the autophagosome to the lysosome; and thirdly, degradation or maturation, which involves vesicle fusion and mixing of autophagosomal contents with lysosomal hydrolases to eventually release degraded by-products back into the cytosol through membrane permeases. The production of autophagosomes is coordinated through the function of at least 35 ATG (autophagy) gene products, which were first described in yeast, with following characterization of the mammalian orthologues (see  for detailed review).
Autophagosome initiation, elongation and closure
The initial sequestration step of autophagy can be differentiated further into three stages: initiation, elongation and closure. Initiation relies on the recruitment of multiple ATG protein complexes on to a cup-shaped membrane assembly site termed the phagophore (or isolation membrane). In mammalian cells, one of the earliest events during autophagosome initiation is the recruitment of the ULK1 (uncoordinated-51-like kinase 1) complex, which comprises ULK1, mATG13, FIP200 (focal adhesion kinase family-interacting protein 200 kDa) and ATG101 . The homologous yeast complex, constructed around the kinase ATG1, is described in further detail below. A second complex recruited during autophagosome initiation contains the class III PI3K (phosphoinositide 3-kinase), VPS34, along with regulatory subunits Beclin1/ATG6, ATG14 L and p150/VPS15. Activation of the Beclin1–PI3K complex results in production of PI3P (phosphatidylinositol 3-phosphate) to provide a lipid signal that recruits other autophagy effectors such as DFCP1 (double FYVE domain-containing protein 1) and members of the WIPI (WD-repeat protein interacting with phosphoinositides) family to the forming phagophore. The role of PI3K signalling for autophagy is discussed in a dedicated chapter of this volume (Chapter 2).
Initial phagophores elongate to enwrap cargo and this stage is dependent on recruitment and orchestrated action of the ATG5–ATG12 and ATG8–PE (phosphatidylethanolamine) ubiquitination-like conjugation systems. These two pathways are inter-related as the ATG12–ATG5 conjugate is required for progression of the second conjugation system, which ultimately converts ATG8 family proteins into their PE-conjugated (active) forms. Generation and enrichment of ATG8–PE on the phagophore has been proposed to drive a number of essential processes during autophagosome formation including binding of autophagy cargo receptors, membrane elongation and, finally, membrane closure.
ATG1, ATG6/Beclin1 and ATG8 represent the three core regulatory complexes essential for canonical autophagy, which describes the most widely observed forms of autophagy in mammalian cells . The present chapter focuses on the regulation of canonical autophagy, which, characteristically, is rapidly activated in cells sensing starvation of nutrients such as amino acids (Figure 2). By contrast, other specialized non-canonical forms of autophagy can be activated in particular cell contexts. Non-canonical types of autophagy are notable for not requiring all of the major core regulatory modules, thus defining, for example, ATG1- or ATG8-independent subtypes of autophagy.
Regulation of yeast autophagy by TOR–ATG1
The best understood activation signal for autophagy is amino-acid starvation and this is transmitted using the TOR (target of rapamycin)–ATG1 pathway . Cells use TOR–ATG1 signalling as a mechanism to coordinately regulate synthesis of proteins via translation and degradation via autophagy  (Figure 3). The key early studies showed how the yeast Saccharomyces cerevisiae starved of amino acids increased ATG1 kinase activity to promote autophagy activation and this stimulatory effect was mimicked by the TOR inhibitor rapamycin . Yeast lacking ATG1 could not undergo autophagy and could not endure starvation conditions showing the essential survival function for nutrient-dependent autophagy .
Maximal activity of yeast ATG1 requires interaction with its regulatory cofactors ATG13, ATG17, ATG29 and ATG31 (Figure 4). In the current model, following autophagy activation, ATG1 and ATG13 assemble on to a stable ATG17–ATG29–ATG31 subcomplex . Another key feature is that the function of ATG13 is controlled by phosphorylation. Under amino-acid-rich conditions, ATG13 is phosphorylated by TOR on at least eight serine residues . Following amino-acid starvation and loss of TOR activity, ATG13 becomes hypo-phosphorylated, allowing full activation of ATG1 by stabilizing the overall complex or altering the conformation of ATG1–ATG13 binding. The importance of ATG1–ATG13–ATG17 complex formation has been demonstrated by mutagenesis approaches. The ATG17-C24R point mutation disrupts binding to ATG13 and this mutant was inefficient at supporting ATG1 kinase activity and autophagy in an ATG17 deletion strain . Point mutations in the C-terminal region of ATG1 (Y878A/R885A and R885E/K892E) that prevent binding to ATG13 accordingly prevent activation of autophagy . Mutations in ATG13 that disrupt binding are similarly defective for autophagy regulation .
ATG1 functional domains
The initial membrane assembly site in yeast is specifically termed the PAS (pre-autophagosome structure). TOR inhibition following amino-acid starvation stimulates autophagy by promoting assembly of an active ATG1–ATG13–ATG17–ATG29–ATG31 complex on the PAS. Once localized on the PAS, multiple functional domains of ATG1 control separate immediate-early and late steps for autophagosome formation and maturation (Figure 5). Crystal structures of the ATG17–ATG29–ATG31 subcomplex have shown that dimerization of ATG17 may play a key role in bridging and coordinating the traffic of multiple lipid vesicle precursors at the PAS . Part of this mechanism involves interaction with ATG1, which has as one of its key roles, direct binding of membrane vesicles via a C-terminal EAT (early autophagy targeting/tethering) domain. Within the C-terminal region of ATG1, the EAT domain appears overlapping or juxtaposed to the sequence that binds ATG13, leading speculatively to potential modes of co-operation or competition [15,18].
The ATG1 complex at the PAS is the pivotal director that recruits further autophagy factors. Consistent with this role, a defined sequence within the internal regulatory region of ATG1 binds ATG8, which has been termed the AIM (ATG8 family-interacting motif) [16,19]. Interestingly, mutation of the AIM did not prevent ATG1 localization and initial formation of the PAS, although ATG1–ATG8 interaction was critical for complete autophagosome formation and proper delivery to the yeast lysosomal-like vacuole compartment, highlighting multiple steps controlled by ATG1.
In the current model, the fully assembled ATG1 complex is a fundamental component that provides a structural scaffold for early autophagosome formation. Other mutagenesis data further illustrate distinct ATG1 kinase-dependent and -independent roles during autophagosome formation . Kinase-inactive ATG1 provided some function as the PAS was formed, but full progression of autophagy was impaired, for example, since factors like ATG8 abnormally accumulated. Consistent with a block in membrane flow, kinase-inactive ATG1 did not drive the generation of properly sized autophagosomes and functional autophagy. In summary, the first essential function of the ATG1 autophagy complex is to direct membranes and protein factors on to autophagosome assembly sites. A subsequent phase that requires ATG1 kinase function and ATG8 involves disassembly and remodelling of membrane and potentially other protein components to allow progression on to elongation, closure and transport phases of autophagy.
Regulation of the mammalian ULK1 complex
Mammalian genomes contain potentially five ATG1 orthologues, but of these, only ULK1 and ULK2 show conservation along their entire length . To date, ULK1 is the best-characterized mammalian ATG1. As in yeast, the ULK1 complex transmits signals from the amino-acid-dependent mTORC1 (mechanistic TOR complex 1) to regulate autophagy. To avoid confusion, it is worth noting that mTOR was originally termed mammalian TOR, but has since been renamed mechanistic TOR, which is the terminology now recognized by the HUGO Gene Nomenclature Committee.
The internal regulatory region of ULK1 contains a conserved motif [microtubule-associated protein LIR (LC3-interacting region)] that binds to ATG8 family proteins [16,20] (Figure 5). The ULK1 C-terminus contains EAT- and mATG13-binding domains . Interestingly, the mammalian mTORC1–ULK1 signalling module acquired additional novel mechanisms through evolution. In contrast with yeast, the core mammalian ULK1 complex (with mATG13, FIP200 and ATG101) remains stable regardless of the cellular nutrient status. Although there is no sequence similarity, FIP200 has been proposed to be the functional equivalent of ATG17 in the complex, whereas a yeast equivalent of ATG101 has yet to be characterized.
A critical feature of the mammalian signalling model is direct binding between activated mTORC1 and the ULK1 complex that is stimulated under amino-acid-rich conditions [6,8] (Figure 4). mTORC1 inhibits autophagy by phosphorylating ULK1 on Ser757 within the internal regulatory region and mTORC1-mediated phosphorylation of mATG13 could be providing a further inhibitory brake on autophagy [21–25]. Following amino-acid starvation, mTORC1 is inactivated and binding to the ULK1 complex is lost, thereby releasing the ULK1 complex to become catalytically active and translocation on to early autophagosome assembly sites near ER (endoplasmic reticulum) membranes. As in yeast, translocation of the ULK1 complex to phagophore initiation sites is an early event following amino-acid starvation in mammalian cells [25–27]. In the current model, ULK1 drives phagophore assembly and elongation by directing protein and membrane components. Without proper ULK function, other early steps of autophagosome assembly such as PI3P production from the Beclin1 complex are blocked . ULK1 may play both structural and dynamic roles as kinase-inactive forms can be observed to localize to phagophore sites, but block overall autophagy flux [18,27]. Overall, the conserved TOR–ATG1 signalling module generally involves translocation of a multimeric ATG1 complex on to autophagosome assembly sites to orchestrate downstream protein and membrane traffic following nutrient starvation.
The autophagy field is now in the process of further understanding how the basic TOR–ATG1 signalling axis is integrated into the larger network organization of the cell. Below, we summarize a number of exciting areas that require additional characterization to provide a more complete picture (Figure 6).
AMPK regulation of autophagy
Nutrient-dependent autophagy is best characterized in terms of amino-acid starvation, but autophagy is also activated in response to energy deprivation, for example when mammalian cells are starved of glucose [22,29,30]. The link of cellular energetics to autophagy highlights a complex signalling network that interconnects AMPK (AMP-activated protein kinase), mTORC1 and ULK1. When energy is limiting, cellular ATP levels decrease while AMP levels increase, leading to an activation of AMPK and phosphorylation of a wide range of downstream targets that control cell energy balance. AMPK plays a key role in the activation of autophagy and this mode of regulation involves phosphorylation of ULK1 on at least six serine and threonine residues within the internal regulatory region (reviewed in ). Details on how these AMPK phosphorylation events regulate autophagy need clarification, but there seems to be a mechanism of cross-talk as mTORC1 phosphorylation of ULK1 disrupts the AMPK–ULK1 interaction . These recent insights are consistent with a tight interconnection between mTORC1/AMPK/ULK1 signalling that integrates multiple cell pathways in response to nutrients. As with TOR regulation of autophagy, AMPK-dependent signalling to ATG1 arose early in evolution as supported by evidence from yeast .
Multiple mTORC1–AMPK–ULK1 connections
It had already been recognized that AMPK inhibits mTORC1. One route involves AMPK-dependent stimulation of the tuberous sclerosis protein complex TSC2–TSC1, which is a GTPase-activating factor that negatively regulates the GTP-binding protein Rheb. As Rheb is required for mTORC1 activity, low energy and activated AMPK thereby cause TSC2–TSC1 to inactivate Rheb-mTORC1. In the second inhibitory pathway, AMPK directly phosphorylates the mTORC1 subunit raptor (regulatory-associated protein of mTOR). This phosphorylation stimulates binding between raptor and 14-3-3 proteins which results in mTORC1 inhibition. Although AMPK and mTORC1 coordinate to signal downstream of ULK1, there are additional connections to signal upstream. ULK1 can phosphorylate all three subunits of AMPK and repress their activity , which may represent a form of negative feedback. Another feedback loop links ATG1 to the inhibition of mTOR as first demonstrated in Drosophila . Mammalian cells have a homologous pathway that includes ULK1-dependent phosphorylation of raptor leading to decreased mTORC1 activity [34,35]. As such, AMPK- and ULK1-mediated phosphorylation of raptor work together to inhibit mTORC1 function, but these mechanisms must coordinate with additional raptor phosphorylation events that positively regulate mTORC1. As mTORC1 primarily serves to repress ULK, feedback from ULK1/2 that inhibits mTORC1 might form a positive reinforcement circuit. These multiple mechanisms are likely to enable the cell to finely tune amino acid and energy signals to tightly coordinate autophagy with other homoeostatic pathways.
Post-translational and transcriptional control of ULK1
mTORC1–AMPK–ULK1 have been established as core components of the nutrient-dependent autophagy module. ULK1 regulation has yet other layers of complexity to be discovered. Proteomic MS studies have so far revealed 29 phosphorylation sites in yeast ATG1 and similar approaches have highlighted over 16 phosphorylation sites in ULK1 . Phosphorylation events are observed throughout the different functional regions of ATG1/ULK1 (Figure 5). These modifications on ATG1/ULK1 reflect combined action of autophosphorylation and other kinases. Autophosphorylation of ATG1 and ULK1 in the kinase domain is critical for catalytic function of the protein and autophagy regulation [36–38]. Future experiments need to characterize more precise roles for the AMPK and mTORC1 sites found in the internal regulatory domain of ULK1. In addition, this region may serve to integrate signals from a range of kinases. There is evidence that protein kinase B (also known as Akt) can also phosphorylate ULK1 within its internal regulatory region . Furthermore, several sites within the ULK1 C-terminal EAT domain are phosphorylated. For these, it is attractive to speculate additional regulatory mechanisms to control membrane binding and protein interactions.
ULK1 serves a central role during early autophagy regulation and so it is reasonable that this component receives multiple upstream signals. Emerging evidence highlights that ULK1 is regulated by protein acetylation on lysine residues in the kinase and internal regulatory regions . Protein acetylation is a form of post-translational modification more widely appreciated for its roles in regulating histones and the p53 tumour suppressor. Mammalian cells also have a pathway where prolonged growth factor starvation causes activation of GSK3 (glycogen synthase kinase 3) which, in turn, stimulates the acetyltransferase Tip60 [HIV-1 Tat (transactivator of transcription)-interactive protein 60 kDa)] to acetylate ULK1 and promote autophagy. Interestingly, acetylation is being recognized for its fundamental role in coordinating cellular metabolic pathways including autophagy [40–42]. Other types of prolonged stress can activate ULK1 function by increasing gene expression. For example, ULK1 expression is increased in a number of cancer cells through the action of p53 and ATF4 (activating transcription factor 4) following DNA damage, hypoxia or disruption of ER homoeostasis (also termed the unfolded protein response) [43,44]. These acetylation and transcriptional mechanisms provide a complementary level of control on ULK1 that function on relatively slower timescales than the amino-acid-sensitive mTORC1/ULK1 pathway.
Downstream of ATG1/ULK1
Fundamental questions still remain on how ATG1/ULK1 truly controls autophagy. The activated ATG1/ULK1 complex translocates on to early autophagy initiation to direct assembly, but the underlying biochemical basis is unclear. Phosphorylation or autophosphorylation on ATG1/ULK1 might drive the assembly by promoting a conformational change and interactions with membranes and additional autophagy factors. Activated ULK1 also phosphorylates the mATG13 and FIP200 subunits of the core complex and although details need to be resolved, these modifications could be serving as autophagy recruitment signals [18,23–25,45]. Besides driving the formation of early autophagy membranes, ULK1 phosphorylates several other substrates and these events may link ATG1/ULK1 with other major autophagy regulatory pathways.
Consistent with a role in controlling intracellular membranes, ATG1/ULK1 signalling directs trafficking of the ATG9 autophagy transmembrane proteins towards autophagosome assembly or elongation and this process is conserved from yeast [46,47]. Experiments that combined Drosophila and mammalian systems identified ATG1/ULK1-dependent phosphorylation of a myosin light-chain kinase protein (ZIP kinase in mammals, spaghetti-squash activator in Drosophila) . Once activated by ULK1, ZIP kinase phosphorylates myosin II regulatory light-chain to control starvation-induced trafficking through direct interaction with ATG9, providing an additional mechanism for ULK1 to modulate downstream autophagy.
AMBRA1 (activating molecule in BECN1-regulated autophagy 1) is an interesting ULK1 substrate that bridges to Beclin1/VPS34 signalling . ULK1-mediated phosphorylation was required for disassembly of a dynein-light-chain–AMBRA1 complex from a dynein microtubule-dependent motor. Thus it has been proposed that ULK1 might regulate autophagy by allowing release and translocation of an active AMBRA1–Beclin1–VPS34 complex to autophagosome assembly sites. One should note that AMBRA1 has further levels of complexity with mechanisms for regulating autophagy by promoting ubiquitination, protein stability and activity of ULK1 . Additional mechanisms also exist to link the ULK and Beclin1–VPS34 pathways. Components of the exocyst complex (better understood for secretory membrane transport) were shown to directly interact with ULK1 and Beclin1–VPS34 . The model from this work has ULK1 directing formation and localization of different subcomplexes of Beclin1–VPS34 with multiple exocyst subunits. Thus ULK1 has several substrates and binding partners that may allow mechanistic coordination of ULK1, Beclin1 and ATG9 autophagy regulatory pathways.
ATG1/ULK1 for neurobiology
A discussion on ATG1/ULK1 function needs to include roles in neuronal vesicular transport that occur in parallel with autophagy. Caenorhabditis elegans with mutation of their ATG1 homologue (unc-51) were originally characterized for uncoordinated behaviour with an underlying axonal defect. Roles for ATG1/ULK1 in neuronal development are conserved in flies and mammals and a number of molecular pathways have been described. The ULK1 C-terminal domain binds syntenin and SynGAP which are proteins implicated in neuronal vesicular trafficking [52,53]. Drosophila ATG1 directly phosphorylates factors involved in vesicular transport such as unc-14, unc-76 and VAB-8 L . Other data support a neuronal signalling complex containing ULK1/2, TrkA NGF (nerve growth factor) receptor, TRAF6 (tumour necrosis factor receptor-associated factor 6) and p62 . Interestingly, despite the progress made in understanding these molecular details, it still remains unclear how the ATG1/ULK1 complex may be coordinating autophagy and more general vesicular trafficking pathways in neurons and perhaps wider cell types.
Resting cells carry out low-level autophagy as a basal homoeostatic mechanism. Following nutrient starvation, autophagy is rapidly activated to degrade proteins and recycle basic cellular building blocks. The autophagy response on amino-acid depletion is controlled by the conserved TOR–ATG1 signalling pathway. We now understand that activated ATG1 family proteins drive autophagy by localizing to sites of autophagosome formation and assembling complexes of additional autophagy protein factors and membrane precursors. As autophagy plays a universal role in maintaining normal biology of cells and is implicated in multiple disease contexts, it has become important to understand all facets of autophagy regulation in order to devise strategies of modulating the process. ATG1 is an attractive signalling pathway for future study due to its central role in coordinating autophagy and other cellular processes.
• When amino-acid levels are limiting, the cell responds by inhibiting protein synthesis while increasing protein degradation through autophagy.
• Activation of canonical autophagy following amino-acid starvation is controlled by a conserved pathway centred by TOR and the autophagy kinase ATG1.
• An activated ATG1 protein complex is assembled early during the regulatory process at sites of autophagosome formation.
• Regions within ATG1 have dedicated functions such as binding protein factors, membrane localization and kinase-mediated remodelling of autophagosome assembly.
• Mammalian ULK1 is a signalling nexus that integrates multiple post-translational and gene expression signals to control downstream autophagy and membrane trafficking.
- © The Authors Journal compilation © 2013 Biochemical Society