The LIR motif – crucial for selective autophagy

Macroautophagy (hereafter referred to as autophagy) is an intracellular degradation process, in which a double membrane structure called the phagophore expands and closes upon itself to sequester part of the cytoplasm to form an autophagosome (varying in diameter from 0.5 to 1.5 µm). First, the autophagosome fuses either with a late endosome forming an amphisome or directly with a lysosome forming an autolysosome. Amphisomes also fuse with lysosomes to form autolysosomes (Mizushima et al., 2011) (Fig. 1). This process can degrade all kinds of molecules and supramolecular structures in the cytoplasm, including organelles such as peroxisomes and mitochondria (Johansen and Lamark, 2011). Because of its fundamental importance in cellular homeostasis and cellular signaling, autophagy is highly relevant for a number of diseases, including cancer, neurodegenerative diseases, muscular dystrophy, lipid-storage disorders and processes such as development, ageing and innate immunity (Levine and Kroemer, 2008; Levine et al., 2011; Mizushima and Komatsu, 2011; Deretic, 2012). In addition to (macro)autophagy, microautophagy and chaperone-mediated autophagy represent distinct autophagy pathways (Arias and Cuervo, 2011; Mijaljica et al., 2011).

Pioneering genetic studies in yeast have revealed a number of AuTophaGy (ATG) genes (Nakatogawa et al., 2009; Mizushima et al., 2011). Currently, 38 Atg proteins are known in yeast. Of these, 17 (Atg1 to Atg10, Atg12 to Atg16, Atg18 and Atg22) are part of the core autophagy machinery used by all the different autophagy pathways. The components of the core autophagy machinery are well conserved from yeast to mammals and appear to act in a similar hierarchical manner. In mammals the core machinery consists of (i) the complex of uncoordinated 51-like kinase 1 and 2 (ULK1–ULK2 ); (ii) a class III phosphatidylinositol 3-kinase (PI 3-kinase) complex; (iii) ATG2A and ATG2B, and the mammalian Atg18 homologs WD-repeat protein interacting with phosphoinositides 1, 2, 3 and 4 (WIPI1, WIPI2, WIPI3 and WIPI4, respectively); (iv) ATG9; (v) a complex of the ATG12–ATG5 conjugate and Atg16L1 and; (vi) ATG8 or the microtubule-associated proteins 1A/1B light chain 3 (MAP1LC3 or LC3) proteins (Fig. 1).

Members of the ATG8 family are the only known ubiquitin-like (Ubl) proteins that are conjugated to a lipid, namely phosphatidylethanolamine (PE). ATG8-PE is present both on the outer and inner membranes of the phagophore. During autophagosome maturation, ATG8 is deconjugated from the outer membrane by ATG4. This is necessary for autophagosome biogenesis (Mizushima et al., 2011). In mammals, two subfamilies of at least seven ATG8 proteins exist: the LC3 proteins LC3A, LC3B and LC3C, with two N-terminal splice variants of LC3A, and GABARAP (γ-amino butyric acid receptor-associated protein), GABARAPL1 and GABARAPL2. In mammals, LC3B is the most prevalent and well-established autophagosome marker. Yeast has only one Atg8 homolog, Caenorhabditis elegans and Drosophila melanogaster have two, whereas Arabidopsis thaliana has nine ATG8 homologs (reviewed by Shpilka et al., 2011).

Autophagy was traditionally regarded as a non-selective, bulk degradation process mainly induced to replenish energy stores upon starvation. A distinction is usually made between basal, housekeeping autophagy that is important for quality control of proteins and organelles, and starvation- or stress-induced autophagy. During the last decade evidence has accumulated that autophagy can be highly selective (Kirkin et al., 2009a; Kraft et al., 2010; Johansen and Lamark, 2011). Selective autophagy refers to the selective degradation of, for instance, organelles (mitophagy and pexophagy), bacteria (xenophagy), ribosomes, macromolecular structures, specific proteins and protein aggregates (aggrephagy) by autophagy. The cytoplasm-to-vacuole targeting (Cvt) pathway, which selectively directs aggregated precursors of aminopeptidase 1 and α-mannosidase to the vacuole, is the only biosynthetic pathway that uses the autophagy core machinery (Lynch-Day and Klionsky, 2010). Together with the discovery of bona fide selective autophagy receptors in mammalian cells, studies of the Cvt pathway have helped elucidate some of the molecular basis for selective autophagy.

A selective autophagy receptor needs to be able to bind specifically to cargo and to dock onto the forming phagophore enabling autophagic sequestration and degradation of the cargo. The first selective autophagy receptor to be identified was p62 [also known as sequestosome-1 (SQSTM1)] (Bjørkøy et al., 2005; Komatsu et al., 2007; Pankiv et al., 2007). p62 was well known to act as a scaffold protein in signaling pathways involving NF-κB (Moscat et al., 2007), but to also accumulate in ubiquitin-containing protein inclusions in many protein-aggregation diseases including Alzheimer disease, Pick disease, dementia with Lewy bodies, Parkinson disease and multiple system atrophy (Kuusisto et al., 2001; Zatloukal et al., 2002). We found that p62 is both a selective autophagy substrate and a cargo receptor for autophagic degradation of ubiquitylated protein aggregates (Bjørkøy et al., 2005; Pankiv et al., 2007). Consistently, knockout of autophagy in the liver of mice demonstrated that p62, which binds both ubiquitin and LC3, regulates the formation of protein aggregates and is removed by autophagy (Komatsu et al., 2007). The authors showed that blocking of autophagy resulted in a failure to degrade p62 and lead to extensive accumulation of protein aggregates, severe hepatomegaly and liver dysfunction (Komatsu et al., 2007).

p62 consists of 440 amino acids and contains an N-terminal PB1 domain, followed by a ZZ-type zinc-finger domain and a C-terminally located ubiquitin-binding UBA domain (Fig. 2). Detailed deletion mapping and point mutation analyses, together with X-ray crystallography and NMR lead to the elucidation of the LC3-interacting region (LIR) motifs of p62 and of the Cvt cargo receptor Atg19 (Pankiv et al., 2007; Ichimura et al., 2008b; Noda et al., 2008). The motif has also been called Atg8-family interacting motif (AIM) (Noda et al., 2010). The structures of p62 and Atg19 peptides bound to LC3B and Atg8, respectively, revealed a common W-x-x-L motif (x = any amino acid) (Ichimura et al., 2008b; Noda et al., 2008) (Fig. 3) (see also Box 1), and the importance of the acidic residues N-terminal to the core of the DDDWTHL LIR motif of p62 was verified by alanine substitutions (Pankiv et al., 2007; Ichimura et al., 2008a; Noda et al., 2008). The LIR motif of p62 presents as an extended β-strand that forms an intermolecular parallel β-sheet with the β2 strand of LC3B. The ATG8 family proteins have a C-terminal, ‘core’ Ubl domain that contains the conserved ‘ubiquitin fold’ and an additional N-terminal arm with two α-helices that are closed onto the core Ubl domain (Fig. 3B). The LIR-containing peptide is located in the interface of the N-terminal arm and the Ubl domain. In this LIR docking site, two hydrophobic pockets HP1 and HP2 in the Ubl domain of LC3 accommodate the side chains of the W and L residues (Ichimura et al., 2008b; Noda et al., 2008; Noda et al., 2010) (Fig. 3A). The two pockets are located on the opposite side of the hydrophobic patch (L8-I44-V70) of ubiquitin. Electrostatic interactions, which involve two of the three aspartic acid residues of the LIR motif and basic residues in the N-terminal arm and Ubl domain of LC3 (R10, R11, K49 and K50), are also important for the interaction between p62 and LC3 (Fig. 3A,B). The importance of the basic residues in the N-terminal arm of LC3B for binding and autophagic degradation of p62 has been demonstrated by domain swap experiments (Shvets et al., 2008; Shvets et al., 2011).

Different strategies have been used to identify proteins that interact with ATG8 proteins through LIR motifs, including candidate approaches (Pankiv et al., 2007; Noda et al., 2008; Sancho et al., 2012), bioinformatics searches (Kraft et al., 2012), proteomics (Behrends et al., 2010; Pankiv et al., 2010), phage display (Mohrlüder et al., 2007b) and yeast two-hybrid assays (Kirkin et al., 2009b; Novak et al., 2010; Wild et al., 2011; Popovic et al., 2012). LIR motifs have been identified by using deletion mapping and protein–protein interaction assays, and by testing deletion and point-mutated constructs. We have found that peptide array analysis is a specific and efficient method for identification of LIR motifs (Alemu et al., 2012).

Following discovery of p62 as a selective autophagy receptor, the related neighbor of BRCA1 gene 1 (NBR1) was found to act as an aggrephagy receptor (Kirkin et al., 2009b). Subsequently, nuclear dot protein 52 kDa (NDP52) was found to be an important xenophagy receptor (Thurston et al., 2009) together with optineurin (Wild et al., 2011). These and the other autophagy receptors discussed below use LIR-motif-dependent interactions to target their cargos for autophagic degradation.

In addition to the role of p62, NDP52 and optineurin in selective autophagy, these proteins have also recently been shown to regulate innate immunity signaling pathways and, thus, were suggested to represent a new class of pattern recognition receptors, the sequestosome-1-like receptors (SLRs) (Deretic, 2012). The SLRs currently consists of p62, NBR1, NDP52, optineurin and Tax1-binding protein 1 (TAX1BP1) (Fig. 2). They all contain a dimerization or multimerization domain, a LIR domain (an atypical LIR motif in the case of NDP52 and TAX1BP1) and an ubiquitin-binding domain. These three features of SLRs are required for the efficient execution of their role as autophagic cargo receptors (Pankiv et al., 2007; Ichimura et al., 2008b; Itakura and Mizushima, 2011; Deosaran et al., 2013). Studies of selective autophagy in mammalian cells and of the Cvt pathway in yeast revealed that the cargo must either be aggregated or represent a reasonably large structure that enables the binding of many receptor molecules; alternatively, the autophagy receptors themselves need to be able to multimerize the cargo (Lynch-Day and Klionsky, 2010; Johansen and Lamark, 2011).

The dual nature of SLRs as autophagy receptors and scaffolding proteins that act in signaling pathways is intriguing. Since the levels of SLRs are regulated by autophagy, the rate of autophagy obviously impacts on signaling that involves SLRs. To discuss these signaling pathways is beyond the scope of this Commentary, but it is worth noting that accumulation of SLRs occurs during cellular stresses, including infection and inflammation, oxidative stress, ER-stress and metabolic stress (Johansen and Lamark, 2011; Deretic, 2012). This accumulation may have dramatic effects on stress-related signaling pathways, but the exact role of autophagy in the control of signal transduction – beyond its effect on receptor levels – is poorly understood. One exception is the regulation of the KEAP1–NRF2 oxidative-stress-response pathway, in which p62 binds to and sequesters KEAP1, leading to its autophagic degradation and the concomitant induction of NRF2 (Komatsu et al., 2010; Jain et al., 2010; Taguchi et al., 2012). A positive feedback loop is established in that increased p62 levels activate NRF2, which, in turn, further increases p62 levels (Jain et al., 2010). Recently, it has been found that the liver toxicity of accumulated p62 is due to constitutive upregulation of the NRF2 oxidative-stress-response pathway (Inami et al., 2011).

One of the striking features of SLRs is their ability to mediate the selective autophagy of substrates that apparently have no structural similarities. Substrates targeted by p62 include ubiquitylated protein aggregates and membrane-embedded structures, such as intracellular bacteria and peroxisomes. The single feature that unites the various structures appears to be that they become ubiquitylated before they are degraded. The LIR motif is absolutely required for targeting of SLRs and bound cargoes into the lumen of autophagosomes (Johansen and Lamark, 2011). Selective autophagy depends on a direct interaction between the LIR motif and ATG8 homologs that are conjugated to the inner, concave membrane of the phagophore. However, the LIR motif by itself does not bring a protein to the inner surface of a phagophore, and the majority of LIR-motif-containing proteins are not substrates for selective autophagy. Selective autophagy of p62 depends on its PB1-domain-driven polymerization, but for the delivery of p62-associated cargos, ubiquitin binding and interactions with other proteins are also important. For the selective autophagy of protein aggregates, p62 collaborates with autophagy-linked FYVE protein (ALFY), a nuclear scaffolding protein that is recruited to cytosolic protein aggregates in a p62-dependent manner. ALFY interacts directly with ATG5 and phosphatidylinositol (3)-phosphate [PtdIns(3)P], and may act as a scaffold protein that induces the assembly of an autophagy-compatible structure (Clausen et al., 2010; Filimonenko et al., 2010). In flies that lack the ALFY ortholog Blue cheese (Bchs), accumulation of the p62 ortholog Ref(2)P in ubiquitin-positive protein aggregates has been observed (Clausen et al., 2010) suggesting a conserved role for ALFY.

Redundancy and/or collaboration clearly exist between different SLRs, although this is not very well studied. NBR1 and p62 bind directly to each other through their PB1 domains, and collaborate in selective autophagy of misfolded proteins and probably also midbody rings (Kirkin et al., 2009b; Pohl and Jentsch, 2009; Kuo et al., 2011). These two proteins also collaborate in pexophagy. Here, binding and clustering of peroxisomes is mediated by NBR1 in a process that depends on the coincident membrane binding of its amphipathic J domain and the adjacent UBA domain (Deosaran et al., 2013). Specialized intracellular pathogens have often developed strategies to avoid or use autophagy for their own purposes, but other pathogens are efficiently degraded by selective autophagy if they are released into the cytoplasm or upon membrane rupture (Mostowy and Cossart, 2012). Microbes that are released into the cytosol are ubiquitylated and then recognized by SLRs (Dupont et al., 2009; Thurston et al., 2009; Zheng et al., 2009). Membrane remnants associated with exposed microbes can also be polyubiquitylated and targeted for autophagic degradation by p62 (Dupont et al., 2009). Furthermore, p62 promotes autophagic killing of intracellular microbes. Cytoplasmic precursors of antimicrobial peptides (ubiquitin or ribosomal precursor proteins) are transported by p62 into autolysosomes or microbe-containing autolysosomes. Here, the precursors are converted into peptides that have been shown to kill Mycobacterium tuberculosis (Ponpuak et al., 2010); and these peptides might also be potent against other microbes. Viruses can also act as substrates and p62 has been implicated in xenophagic elimination of Sindbis virus (Orvedahl et al., 2010). Efficient xenophagy of Salmonella enterica serotype Typhimurium (S. typhimurium) is mediated by p62, NDP52 and optineurin, with the p62-containing microdomains on ubiquitin-coated bacteria appearing to be physically separated from areas that are occupied by NDP52 and optineurin (Cemma et al., 2011; Mostowy et al., 2011; Wild et al., 2011). However, although NDP52 is recruited to ubiquitin-coated bacteria through its C-terminal Zinc-finger (ZF) domain (Fig. 2), it is initially targeted to damaged Salmonella-containing vacuoles that are marked by galectin-8, which binds exposed β-galactoside-containing glycans. In this way, cytosolic galectin-8 functions as an ubiquitin-independent ‘danger’ receptor and ‘eat-me’ signal (Thurston et al., 2012). Galectin 8 also detects non-bacteria induced damage to endosomes or lysosomes, suggesting that membrane rupture is the initial common event detected during invasion by microbes (Thurston et al., 2012). The atypical LIR motif in NDP52, termed CLIR, comprises the tripeptide Leu-Val-Val and binds specifically to LC3C (Fig. 3A). Efficient recruitment of the other ATG8 family members to bacteria-degrading autophagosomes depends on both NDP52 and LC3C (von Muhlinen et al., 2012). TAX1BP1 (T6BP) is a cargo receptor with homology to NDP52. TAX1BP1 binds ubiquitin and contains the same atypical LIR motif as NDP52 (Newman et al., 2012) but its role in xenophagy is unknown.

NDP52 is required for degradation of the micro RNA (miRNA)-processing enzyme DICER, and the main miRNA effector AGO2 by selective autophagy (Gibbings et al., 2012). An ubiquitin-independent role of optineurin in aggrephagy has recently been reported (Korac et al., 2013). It should also be noted that p62 has a role in mitophagy (Johansen and Lamark, 2011), although this process is primarily mediated by specific mitochondrial membrane receptors, as discussed below.

Both yeast and mammalian cells can selectively eliminate damaged or superfluous mitochondria by mitophagy (reviewed by Ashrafi and Schwarz, 2013). In yeast, mitophagy is orchestrated by Atg32, an integral protein of the outer mitochondrial membrane (OMM) with a N-terminus that faces the cytosol and C-terminus located in the intermembrane space (Kanki et al., 2009; Okamoto et al., 2009). Atg32 can interact with Atg8 indirectly through Atg11 and directly through its LIR motif in the N-terminal cytosolic domain. Atg32 recruits Atg8 and Atg11 to the mitochondria surface to form an initiator complex essential for mitophagy (Kondo-Okamoto et al., 2012). In mammalian cells, three integral OMM proteins that all have a LIR motif in their cytosolic N-terminal domain are implicated in mitophagy (Fig. 2). Two homologous BCL2 homology 3 (BH3)-only proteins, Bnip3 and Nix (also known as Bnip3L), are able to induce mitophagy and can also activate cell death (reviewed by Zhang and Ney, 2009). Bnip3 induces the removal of both mitochondria and endoplasmic reticulum (Hanna et al., 2012). Homodimerization of Bnip3 through the transmembrane domain facilitates the interaction between the LIR motif of Bnip3 and LC3B. Nix also facilitates LIR-dependent mitophagy (Novak et al., 2010) (supplementary material Table S1). During erythroid cell maturation, Nix mediates the complete removal of mitochondria (Schweers et al., 2007; Sandoval et al., 2008). Additionally, Nix is involved in depolarization-induced mitophagy. Nix has a core LIR motif identical to Bnip3. However, in contrast to Bnip3, Nix does not interact with LC3B but with GABARAP-L1 during mitochondrial stress (Schwarten et al., 2009; Novak et al., 2010). Hence, residues flanking the core LIR motif might be involved in determining specificity. Bnip3 and Nix are both involved in hypoxia-induced mitophagy (Zhang et al., 2008; Bellot et al., 2009). The third mitophagy receptor in the OMM is FUNDC1; it acts in hypoxia-induced mitophagy, but with a different mechanism that involves the dephosphorylation of its LIR motif, which enhances its binding to LC3B (see below) (Liu et al., 2012).

So far few autophagy receptors are known to only interact with one substrate under certain circumstances. Starch-binding-domain-containing protein 1 (Stbd1) and the E3-ubiquitin ligase Cbl represent such specialized receptors (Fig. 2). Stbd1 binds glycogen in vitro and is associated with glycogen in cells; it binds more tightly to abnormal glycogen that is poorly branched (Jiang et al., 2011). Stbd1 binds to GABARAP-L1 through a LIR motif and has been proposed to act as an autophagy receptor for glycogen in a process termed glycophagy (Jiang et al., 2011).

Kinase activity can also be regulated by selective autophagy that involves interaction with the LIR motif. For instance, Cbl has been identified as an autophagy receptor for the active, non-receptor, membrane-associated tyrosine kinase Src (Sandilands et al., 2012). Increased Src activity promotes tumorigenesis but excessive Src signaling can be cytotoxic (Yeatman, 2004). When integrin signaling through the focal adhesion kinase (FAK)–Src pathway is disrupted in cancer this can lead to excessive and cytotoxic Src activity. By using its LIR motif to bind LC3B, Cbl is able to switch the targeting of Src from the proteasome to autophagic degradation, thereby promoting cancer cell survival (Sandilands et al., 2012).

The presence of functional LIR motifs in components of the core autophagy machinery demonstrates that LIR-motif-mediated interactions do not only help targeting cargo receptors to autophagosomes but are also involved in regulating autophagosome formation and maturation. In addition to the core autophagy machinery, several other LIR-motif-containing proteins are involved in autophagosome formation, transport and maturation (fusion to lysosomes) (see Fig. 4) (supplementary material Table S1).

The ATG proteins of the core autophagy machinery are involved in all steps of autophagosome formation (Fig. 1) (Mizushima et al., 2011). The yeast serine/threonine kinase Atg1 (ULK1 in mammals) forms a large complex with Atg13 and the Atg17–Atg31–Atg29 ternary complex. Recently, two independent studies reported a LIR-motif-dependent interaction between Atg1 and Atg8 (Kraft et al., 2012; Nakatogawa et al., 2012). Atg1 is present on autophagosomes in an Atg8-dependent manner before it is transported to the vacuole for its degradation. Kraft et al. (Kraft et al., 2012) also showed that Atg13, in complex with Atg1, is degraded by autophagy. Similarly, Atg1 and Atg13 are degraded by autophagy during nutrient starvation in Arabidopsis (Suttangkakul et al., 2011), but the role of the interaction between their LIR motifs and ATG8 is currently unknown. Mutations in the LIR motif of Atg1 result in reduced autophagy but do not influence its functions during initiation of autophagosome formation (Nakatogawa et al., 2012). This indicates that Atg1 is also involved in late events of autophagy. Kraft et al. showed that the Atg1–Atg8 interaction is conserved and maintained in mammals, by demonstrating that ULK1 associates with autophagosomes in a LIR-motif-dependent manner (Kraft et al., 2012). We identified that the same LIR motif in ULK1 is required for its starvation-induced association with autophagosomes (Alemu et al., 2012). In contrast to that in Atg1, the LIR motif of ULK1 does not significantly mediate its degradation. We also mapped LIR motifs in ULK2, and the ULK complex proteins ATG13 and FIP200, and demonstrated their binding to ATG8 proteins with a preference for the GABARAP-subfamily (Alemu et al., 2012). It is possible that LIR-ATG8 interactions of ULK complex proteins facilitate and/or stabilize tethering of the ULK complex to the phagophore.

Other members of the core autophagy apparatus, Atg3 in yeast and ATG4B in mammals, undergo LIR-motif-dependent interactions with ATG8 that potentially serve a regulatory role (Satoo et al., 2009; Yamaguchi et al., 2010). Among the four ATG4 homologs (ATG4A, ATG4B, ATG4C, ATG4D), ATG4B is the main human ATG4 homolog that efficiently processes ATG8 precursors and ATG8-PE (Li et al., 2011). The crystal structure of the human ATG4B–LC3B complex indicates conformational changes in ATG4B upon binding of the LC3 substrate that facilitate access of LC3 to the catalytic site of ATG4B (Satoo et al., 2009). Interestingly, in the crystal structure, the N-terminal LIR motif of ATG4B interacts with a LIR-binding site on an adjacent (non-substrate) LC3. This interaction with the LIR motif stabilizes an open conformation of the N-terminal tail of ATG4B, which presumably favors membrane targeting. Since ATG4 also mediates deconjugation of ATG8 proteins, a process that requires membrane targeting, the conformation of the N-terminal tail might, therefore, regulate the deconjugation activity of ATG4 (Satoo et al., 2009). The yeast E2-like enzyme Atg3 contains a canonical LIR motif (WEDL) that is essential for the efficient transfer of Atg8 from Atg3 to PE. The Atg3 LIR motif is required for the Cvt pathway but not for starvation-induced autophagy. The interaction between the LIR motif of Atg3 and Atg8 liberates Atg8 from being bound by the LIR motif of Atg19, thus allowing Atg8–PE conjugation (Yamaguchi et al., 2010).

Although some steps of autophagosome formation are well understood, the membrane origins of autophagosomes are still debated. Multiple membrane sources were found to be involved, such as endoplasmic reticulum (ER), mitochondria, ER-mitochondria contact sites and plasma membrane (Hamasaki et al., 2013; Weidberg et al., 2011). The plasma membrane can contribute directly to the formation of ATG16L1-positive autophagosome precursors that depend on interactions between ATG16L1 and the clathrin heavy chain (Ravikumar et al., 2010). Hence, clathrin-mediated endocytosis might be involved in regulating the initial stages of autophagosome formation (Ravikumar et al., 2010). Interestingly, the clathrin heavy chain also interacts with GABARAP through a LIR motif on a surface-exposed α-helix in the flexible linker region (Mohrlüder et al., 2007a). Structural studies show that LIR motifs adopt a β-conformation when bound to ATG8-proteins and form an intermolecular parallel β-sheet (Fig. 3B) (Noda et al., 2010). It will, therefore, be interesting to learn which conformation the clathrin LIR has upon binding to GABARAP. Clathrin and GABARAP are both involved in trafficking of the GABA_A receptor, suggesting that the LIR-mediated interaction has a physiological relevance. However, it has not been studied whether this interaction impacts on autophagosome formation. Calreticulin, which competes with clathrin for binding to GABARAP (Mohrlüder et al., 2007a), also has a LIR motif very similar to that of clathrin (supplementary material Table S1) (Mohrlüder et al., 2007b). Calreticulin is a luminal Ca²⁺-dependent chaperone of the ER, but is also involved in variety of cytosolic functions as a regulator of intracellular Ca²⁺ homeostasis (Wang et al., 2012). It is presently not known whether these LIR interactions are relevant for both autophagosome formation and trafficking of the GABA_A receptor, or for only the latter.

The autophagosome precursor that is generated by clathrin-dependent endocytosis might represent phagophore precursors. During and after phagophore formation, proteins are recruited to the forming autophagosome in a ‘retrieve–recycle’ manner. The tumor protein 53-induced nuclear protein 2 (TP53INP2; also known as and, hereafter, referred to as DOR) exits the nucleus in response to cellular stress or the activation of autophagy (Nowak et al., 2009; Mauvezin et al., 2010). Cytoplasmic DOR then localizes to autophagosomes where it interacts with the transmembrane protein VMP1 (Nowak et al., 2009). On autophagosomes, DOR interacts with LC3B through its LIR motif (Sancho et al., 2012), but it does not colocalize with the autolysosome-associated protein LAMP1, indicating that DOR localizes only to early autophagosomes (Mauvezin et al., 2010). Through its interaction with VMP1, DOR presumably acts as a scaffold protein that recruits ATG8 proteins to the autophagosome (Nowak et al., 2009). The LIR motif in DOR overlaps with its nuclear export signal (Sancho et al., 2012). Hence, mutation of the core LIR residues of DOR blocks its nuclear exit in response to autophagy activation. Interestingly, DOR and its homolog TP53INP1 share two highly conserved regions, including the LIR motif (Sancho et al., 2012). LIR-mediated localization of TP53INP1 to autophagosomes induces autophagy- and caspase-dependent cell death, and it has been suggested that TP53INP1 displaces p62 from LC3B, which then promotes cell death (Seillier et al., 2012).

The mechanisms that regulate membrane trafficking in autophagy are poorly understood. The Rab GTPases (a large family of monomeric, small GTPases) in their active form are spatially organized into distinct membrane regions, where they recruit effectors to regulate intracellular vesicle trafficking events (Stenmark, 2009). Rab GTPase-activating proteins (GAPs) negatively regulate the activity of Rab GTPases. The Rab GAP TBC1D25 is recruited to phagophores and autophagosomes through direct interactions between its LIR motif and ATG8 homologs, and its GAP activity regulates the fusion between autophagosomes and lysosomes (Itoh et al., 2011). TBC1D25 inhibits Rab33B, a Golgi-resident Rab (Itoh et al., 2011). Active Rab33B binds to ATG16L1 and is involved in recruitment of the ATG12-ATG5-ATG16L complex to preautophagosomal structures (Itoh et al., 2008). The authors suggest a model whereby TBC1D25 uses ATG8 proteins as scaffolds to regulate autophagosomal maturation (Itoh et al., 2011).

Recently, 14 of 36 human TBC (Tre2, Bub2, Cdc16)-domain-containing Rab GAPs were shown to interact with ATG8 proteins in yeast two-hybrid screens (Popovic et al., 2012). One of these, TBC1D5 contains two LIR motifs, which are both required for ATG8 binding and its co-localization with ATG8 to autophagosomes upon starvation-induced autophagy (Popovic et al., 2012). TBC1D5 is involved in retrograde traffic from endosomes to the Golgi. Interestingly, the N-terminal LIR of TBC1D5 interacts with the Vps29 subunit of the retromer complex, a recycling endosome sorting complex responsible for vesicle delivery from early endosomes to the Golgi. The binding of TBC1D5 to Vps29 can be titrated out by LC3 (Popovic et al., 2012), indicating that TBC1D5 acts as a molecular switch between endosomes and autophagy. Furthermore, the C-terminal LIR of TBC1D5 can tether the endosome and autophagosome, thereby mediating autophagosome maturation (Popovic et al., 2012).

The examples above demonstrate crosstalk between endocytosis and autophagy, which then converge in lysosomal degradation. Rab7 is involved in maturation of both autophagosomes and endosomes, as well as in the transport of autophagosomes and endosomes towards lysosomes for degradation. The Rab7 effector FYVE and coiled-coil-domain-containing protein 1 (FYCO1) is localized on phagophores, autophagosomes and late endosomes and, in addition to Rab7, interacts with LC3 and PtdIns(3)P (Pankiv et al., 2010). FYCO1 binds to LC3 through a LIR motif in the middle of the connecting loop between its FYVE and GOLD domains. This flexible loop is predicted to be folded in a way that blocks the interaction between the FYVE domain and PtdIns(3)P. Binding to LC3B on autophagic structures releases this inhibition and targets FYCO1 exclusively to PtdIns(3)P-containing membranes that contain LC3B (Pankiv et al., 2010). FYCO1 couples autophagosomes and other Rab7-positive vesicles to molecular motors. Depending on the direction of vesicle movement, FYCO1 is coupled to kinesin molecular motors. This is further supported by the identification of a potential kinesin-binding site in FYCO1 (Pankiv et al., 2010).

Mitogen-activated protein kinase 15 (MAPK15) is another protein that localizes to autophagosomes through an interaction between its LIR motif and ATG8 proteins (supplementary material Table S1) (Colecchia et al., 2012). The kinase activity of MAPK15 is known to affect the rate of both basal and starvation-induced autophagy (Colecchia et al., 2012) but the substrates of MAPK15 that are involved in autophagy are unknown.

One signaling pathway regulated by autophagy is Wnt signaling; autophagy enhances the degradation of Dishevelled2 (Dvl2), a transducer of the Wnt pathway and, thus, negatively affects Wnt signaling (Gao et al., 2010). The C-terminal DEP domain of Dvl2 contains a LIR motif that binds to ATG8 proteins (Gao et al., 2010; Zhang et al., 2011). The N-terminal DIX domain mediates the self-oligomerization of Dvl2 that is necessary for its ubiquitylation, and facilitates its binding to LC3B and GABARAP (Gao et al., 2010). Ubiquitylation of Dvl2 is enhanced during starvation, and is essential for its interaction with p62 and subsequent targeting to autophagosomes. Thus, p62 mediates the indirect association of Dvl2 with LC3B and GABARAP (Gao et al., 2010). This is also likely to be the case for the interaction of Dvl2 with GABARAP-L1 (Zhang et al., 2011). Consequently, Dvl2 is degraded by autophagy through its LIR-dependent interaction with ATG8 proteins and by means of the autophagy receptor p62. Very recently, it was shown that β-catenin is selectively degraded by autophagy during nutrient deprivation via the formation of a β-catenin-LC3 complex depending on a LIR motif in β-catenin. A regulatory feedback mechanism is at work, in which active Wnt/β-catenin signalling represses autophagy and p62 expression, while β-catenin is itself targeted for autophagic clearance in autolysosomes upon autophagy induction (Petherick et al., 2013).

Since 25% of the known LIR motifs harbour an S or T residue as the ‘any amino acid residue’ at position –1 immediately N-terminal to the aromatic residue of the LIR motif (supplementary material Table S1), it is conceivable that binding affinity of LIR motifs is regulated through phosphorylation. Indeed, NDP52 recruits TANK-binding kinase 1 (TBK1) to the bacterial surface (Thurston et al., 2009). Optineurin also recruits TBK1 to ubiquitylated Salmonella, resulting in the subsequent phosphorylation of optineurin at S177 located at the ‘any amino acid residue’ at position –1, which strongly enhances the binding to LC3B (Wild et al., 2011). Optineurin and NDP52 occupy the same microdomains on the bacteria. Thus, NDP52-bound TBK1 can also phosphorylate the LIR of optineurin, thereby enhancing the response.

The phosphorylation state of the LIR motif of Bnip3 has been shown to determine whether it executes pro-survival mitophagy or apoptosis. Phosphorylation of S17 and S24 that flank the LIR increases binding of Bnip3 to LC3B and GABARAP-L2, and induces mitophagy. When its LIR motif is unphosphorylated, Bnip3 functions as a BH3-only protein and promotes apoptosis (Zhu et al., 2013). Furthermore, the interaction of FUNDC1 with LC3B is enhanced during hypoxia through the dephosphorylation of the tyrosine residue (Y18 binding to HP1) in its LIR, which facilitates mitophagy (Liu et al., 2012).

Not only the LIR motifs but also the binding surface of ATG8 proteins may be phosphorylated to regulate binding of LIR-motif-containing proteins. For instance, phosphorylation of S12 in the N-terminal arm of rat LC3B by protein kinase A (PKA) negatively affects autophagy (Cherra et al., 2010). This residue is adjacent to the two Arg residues (R10 and R11) that bind to two aspartic acid residues (bold) in the DDDWTHL LIR motif of p62, but it has not been studied whether phosphorylation of S12 affects the docking of p62. Further studies are required to thoroughly address the question to which extent LIR motifs and LIR docking sites are regulated by posttranslational modifications.

The interaction between LIR motifs and ATG8 proteins is crucial for the recruitment of cargo to the inner surface of the phagophore, and for the recruitment of effector proteins to the outer autophagosomal membrane where these effectors mediate transport and maturation of autophagosomes (Fig. 4). The characterization of LIR-motif-containing proteins and the elucidation of their roles in autophagy are still at an early stage. Additional examples on how LIR–ATG8 interactions are regulated through phosphorylation or other post-translational modifications are clearly anticipated in future studies. It will be interesting to see whether an interaction of LIR motifs with ATG8 proteins is also involved in processes other than autophagy. The binding of TBC1D5 to Vps29 through its N-terminal LIR motif suggests that there are binding partners for LIR-motifs other than ATG8 proteins. Finally, it will also be interesting to investigate whether these LIR–ATG8 interactions can be explored as druggable targets.

We thank members of our group for critical reading of the manuscript, and Steingrim Svenning for help with Fig. 3.