|Home | About | Journals | Submit | Contact Us | Français|
Mounting evidence suggests that autophagy is a more selective process than originally anticipated. The discovery and characterization of autophagic adapters, like p62 and NBR1, has provided mechanistic insight into this process. p62 and NBR1 are both selectively degraded by autophagy and able to act as cargo receptors for degradation of ubiquitinated substrates. A direct interaction between these autophagic adapters and the autophagosomal marker protein LC3, mediated by a so-called LIR (LC3-interacting region) motif, their inherent ability to polymerize or aggregate as well as their ability to specifically recognize substrates are required for efficient selective autophagy. These three required features of autophagic cargo receptors are evolutionarily conserved and also employed in the yeast cytoplasm-to-vacuole targeting (Cvt) pathway and in the degradation of P granules in C. elegans. Here, we review the mechanistic basis of selective autophagy in mammalian cells discussing the degradation of misfolded proteins, p62 bodies, aggresomes, mitochondria and invading bacteria. The emerging picture of selective autophagy affecting the regulation of cell signaling with consequences for oxidative stress responses, tumorigenesis and innate immunity is also addressed.
Macroautophagy (hereafter referred to as autophagy) is an evolutionarily conserved catabolic process that involves the sequestration and transport of organelles and macromolecules to the lysosomes for degradation.1,2 Following lysosomal degradation, recycling occurs to replenish the cell with nutrients and building blocks for anabolic processes.3 Autophagy is initiated by formation of the phagophore or isolation membrane, a crescent-shaped double membrane that expands and fuses to form a double-membrane vesicle, the autophagosome. Autophagosomes eventually fuse with lysosomes to degrade their content (Fig. 1). Autophagy has most likely evolved both as a pathway for restoring intracellular nutrient supply during starvation and as a quality control mechanism to protect the cell against damage caused by toxic macromolecules and damaged organelles. The process is also mobilized in the innate immune response against invading microbes.1,4,5
Yeast genetics has been vital for the elucidation of the molecular machinery involved in autophagy processes.2,6 So far, 34 AuTophaGy-related (ATG) genes have been reported in yeast and 15 of these are “core” ATG genes commonly required for the different autophagy pathways.6,7 Two ubiquitin (Ub)-like protein conjugation systems involving the two Ub-like modifiers Atg8 and Atg12 constitute part of the evolutionarily conserved autophagic machinery. Both systems use the E1 enzyme Atg7. Atg10 acts as an E2 for the conjugation of Atg12 to Atg5 to form a high-molecular-weight complex with Atg16, which then in turn assists the E2 Atg3 in the conjugation of phosphatidylethanolamine (PE) to Atg8. Very recently, mammalian ATG12 was also shown to conjugate to ATG3. This conjugation is not involved in starvation-induced autophagy but is important for regulation of mitochondrial homeostasis and cell death.8 In addition to the two conjugation systems, four more protein complexes make up the remainder of the molecular machinery required for autophagy (Fig. 1). In mammalian cells the serine-threonine protein kinase ULK1 (unc-51-like kinase 1, a homologue of yeast Atg1) forms a complex together with ATG13, FIP200 (focal adhesion kinase family interacting protein of 200 kD) and ATG101. The ULK1 complex regulates the initial events of autophagosome formation together with the class III phosphatidylinositol (PtdIns) 3-kinase complex containing Beclin 1, ATG14L and the PtdIns 3-kinase subunits Vps34 and Vps15. The WIPI (WD-repeat protein interacting with phosphoinositides) family proteins, notably WIPI1 and WIPI2, are Atg18 homologues acting in early steps of autophagosome formation.9,10 Finally, the ATG9 transmembrane protein is involved in membrane trafficking and is required for autophagy, but its function is still unknown.11
In yeast there is a single Atg8 protein; Drosophila and C. elegans have two, while mammals have at least seven ATG8 proteins that can be grouped into two subfamilies containing at least three MAP1 light chain 3 (LC3A, B and C) and four gamma-aminobutyrate receptor-associated protein (GABARAP) and GABARAP-like proteins (ATG8L/GEC-1/GABARAPL1, GATE-16/GABARAPL2 and GABARAPL3).12,13 ATG8 proteins are specifically cleaved at their C-termini by the ATG4 pro- teases to expose a C-terminal glycine producing the form I of the ATG8 molecule (i.e., LC3-I). This glycine residue is then conjugated to PE.14–17 The resulting PE-conjugated form II of ATG8 proteins (i.e., LC3-II) is tightly bound to the autophagosomal membranes and serves as an autophagic marker protein.14,16,18,19 The conjugation of ATG8 to PE is essential for the hemifusion of lipid membranes and is suggested to drive the expansion of the autophagosome.20 ATG8-PE is deconjugated by ATG4 protease to recycle ATG8 from the outer autophagosomal membrane.21 This step can be regulated by reactive oxygen species acting on ATG4.22 In mammals, LC3B is thought to act as the main ATG8 homologue involved in starvation-induced autophagy, but also GABARAP, GABARAPL2 and GABARAPL1 are conjugated to PE and involved in autophagy.16,23 Recent data suggest that both the LC3 and GABARAP subfamilies of ATG8s are indispensable for autophagy in mammalian cells and that there is a division of labor among them during formation of the autophagosome.24
Chaperone-mediated autophagy (CMA) is a selective form of autophagy where autophagosomes do not form, and soluble cytosolic proteins containing a degenerate short sequence motif related to the pentapeptide KFERQ are degraded by the lysosomes. Degenerate variants of this sequence motif are found in 30% of cytosolic proteins. CMA substrate proteins with a KFERQ-like motif bind to a complex of Hsc70, Hsp90 and several co-chaperones associated with the lysosomal receptor LAMP-2A and are transported directly through the membrane into the lumen of the lysosome for degradation.25,26
Another pathway involving transport of selective cargo to the lysosome with participation of part of the autophagic machinery is the biosynthetic cytoplasm-to-vacuole targeting pathway in yeast.7 In the Cvt pathway, precursors of two lysosomal enzymes, aminopeptidase I (Ape1) and α-mannosidase, are selectively delivered to the vacuole in a double-membrane vesicle after recognition of the cargo and of yeast Atg8 by the cargo receptor Atg19. Atg19 binds to an aggregate of dodecameric precursor Ape1 (prApe1) and to α-mannosidase and forms the Cvt complex. In addition, Atg11 binds to Atg19 and is required for transport of the Cvt complex to the phagophore assembly site (PAS) in yeast where the double membrane is recruited and surrounds the Cvt complex.7
In contrast to the selectivity seen in the CMA and Cvt pathways, (macro)autophagy has been considered a rather unselective process for bulk degradation of long-lived proteins and organelles that during nutrition deprivation can recycle building blocks and help restore the energy balance of the cell. However, a number of recent reports present mounting evidence of selective autophagic degradation of protein inclusions caused by aggregate-prone or misfolded proteins (aggrephagy),27–35 of organelles such as peroxisomes (pexophagy),36,37 mitochondria (mitophagy),38–43 bacteria and virus (xenophagy),44–48 surplus ER (reticulophagy),49 and ribosomes (ribophagy).50 In the following sections we will discuss recent developments in the field of selective autophagy with particular focus on the roles played by the autophagic adaptors p62/SQSTM1 (sequestosome 1) and NBR1 (neighbor of Brca1 gene). p62 and NBR1 are both selective autophagy substrates and cargo receptors for degradation of ubiquitinated substrates by autophagy.31,33,51 As discussed further below and in separate reviews in this issue of Autophagy such substrates may include protein aggregates,29,31–33,35,51 soluble proteins,37,52 midbody rings,53 damaged mitochodria,39–41 peroxisomes,37 intracellular bacteria,46–48 phagocytic membrane remnants,44 bacteriocidal precursor- and viral capsid proteins.45,54
The p62 cDNA was originally cloned and named p62 Lck-ligand because the encoded protein bound to the tyrosine kinase Lck.55 Shin, who originally cloned p62, noted the ability of p62 to form aggregates and coined the name sequestosome 1 (SQSTM1) based on this property.56 Later, the mouse and rat orthologues were named A170 and ZIP, respectively.57,58 The human p62 protein is 440 amino acids long and contains an N-terminal PB1 domain followed by a ZZ-type zinc finger domain, nuclear localization signals (NLS), nuclear export signal (NES), LIR (LC3-interacting region) and KIR (KEAP1 interacting region) motifs and a C-terminal UBA (Ub-associated) domain (Fig. 2A). The PB1 domain is a protein-protein interaction domain enabling p62 to interact with the protein kinases PKCζ, PKCλ/ι, MEKK3 and MEK5 as well as to NBR1.59–62 Importantly, p62 is able to homopolymerize via its PB1 domain.59,61 Interestingly, p62 harbors active nuclear import and export signals and shuttles between the nucleus and cytoplasm.63 The UBA domain of p62 binds both mono- and poly-Ub.64 p62 seems to have a preference for binding to monoUb in vitro.65 Since K63-linked Ub chains have a more open and extended conformation than K48-linked Ub chains, this is consistent with a reported in vivo preferential binding to K63-linked Ub chains.66 However, the isolated UBA domain binds with low affinity,65,67 whereas full-length p62 binds much more strongly to both mono Ub and 4x Ub when assayed in vitro using GST-pulldown assays.31 The UBA domain of p62 forms homodimers in vitro and this inhibits Ub binding.68 The importance of the UBA domain is underscored by the fact that adult onset Paget disease of the bone is frequently associated with dominant-acting mutations in one of the p62 alleles that either lead to deletion of the UBA domain or destroys its ability to bind Ub.69 Likely, p62 mutants without functional UBA increase osteoclastogenesis by potentiating receptor activator of nuclear factor-kappaB ligand (RANKL)-induced activation of nuclear factor-kappaB (NFκB).70–72 This effect may involve TRAF6 and/or atypical protein kinase Cs, and it will be interesting to see whether selective autophagy can be involved in regulating RANKL signaling. As discussed by Sumpter and Levine in another review in this issue of Autophagy, there may also be a link between mutant p62 and accumulation of paramyxovirus nucleocapsids in bone lesions of Paget disease patients.
p62 accumulates in Ub-positive inclusions in neurodegenerative diseases and proteinopathies of the liver.73–76 Immunostaining of p62 has been recommended as histochemical diagnostic marker, together with Ub and cytokeratins, for various human protein aggregation diseases.77 p62- and Ub-positive protein bodies form in response to stress including amino acid starvation, oxidative stress, accumulation of defective ribosomal products or inhibition of autophagy.28,32,33 While studying the PB1 domain-mediated interaction between p62 and atypical PKC we noticed that p62 displayed a perinuclear, punctate distribution in cells and that overexpression increases the number and size of these structures.59 We could subsequently show that p62 is a selective autophagy substrate, continuously degraded by autophagy.29,33 Our group and Komatsu's group independently showed that p62 binds directly to LC3 via the LIR motif (named LC3 recognition sequence; LRS, by Komatsu) and this motif is required for autophagic degradation of p62.33,51 The LIR motif mediates binding to all the members of the mammalian ATG8 family of proteins (both LC3- and GABARAP subfamilies).31,33,51
In a collaborative effort involving the groups of Dikic, Komatsu and Elazar we identified NBR1 as another autophagic cargo receptor that is selectively degraded by autophagy.31 The 966-amino acids long NBR1 is one of the PB1-domain containing interaction partners of p62 and has a similar domain organization as p62. They both share the arrangement of PB1, ZZ and UBA domains,78 but the twice-as-long NBR1 also has other domains not found in p62. Among these is a coiled coil domain required for dimerization of NBR1 and a novel, evolutionarily conserved domain of unknown function we denote the FW (Four W (tryptophan)) domain (Fig. 2A). This domain was recently called the NBR1 box by Kraft et al.79 We prefer the name FW domain based on the four conserved tryptophan residues since this domain is also found in another evolutionarily conserved eukaryotic protein. We found that the protein level of NBR1 is regulated by autophagy, and NBR1 is, together with p62 and Ub-proteins, located to inclusion bodies associated with human pathologies. NBR1 cannot polymerize via its PB1 domain, like p62 can. Instead it binds to p62 via this domain and can be part of, or the chain terminator of, a polymeric chain of p62 molecules. We mapped two LIR motifs in NBR1 of which LIR1, centered at Y732, is most important for autophagic degradation of NBR1. In contrast to p62, the isolated UBA domain of NBR1 binds very well to both K48- and K63-linked diUb.31 The role of NBR1 as an autophagy receptor interacting with LC3 and Ub was confirmed by a study from the group of Mathias Gautel.80
In contrast to p62, NBR1 has not been extensively studied. However, this is now changing and the protein has already been implicated in quite diverse biological contexts. NBR1 binds directly to the giant sarcomeric protein kinase titin and to p62 in the M line of the sarcomere of skeletal muscles. Mutations in titin that disrupt the binding to NBR1, cause hereditary muscle disease in humans.81 Upon loss of binding to titin, both NBR1 and p62 are dispersed from the M lines into many punctate structures probably being so-called p62 bodies (see below). It is not known if autophagy plays any role in this disease. NBR1 interacts and colocalizes with Spred2 on intracellular vesicles.82 This interaction with NBR1 is essential for the role of Spred2 as an attenuator of fibroblast growth factor signaling resulting in lysosomal degradation of activated receptors. A recent study of transgenic mice expressing a C-terminally truncated mutant of NBR1 suggests a role for NBR1 in bone remodeling. Loss of NBR1 function leads to perturbation of p62 levels and hyperactivation of p38 MAPK favoring osteoblastogenesis.83 Hence, both p62 and NBR1 are involved in bone remodeling processes and may cooperate or have antagonistic roles. More studies are needed to elucidate their relative contributions. Jorge Moscat's group has knocked out NBR1 in the T cells of mice and found that NBR1 is a critical mediator of T-cell activation and acts in the control of Th2 differentiation and allergic airway inflammation.84 Interestingly, NBR1 is involved in T cell polarity and recruited to the immunological synapse.
NBR1 displays a wider evolutionary distribution than p62. NBR1 is found in plants and fungi and in metazoans, but the gene is lost in several animal lineages including model organisms like Drosophila and C. elegans. p62 is not found in plants and fungi and seems confined to the metazoans.
Autophagic degradation of p62 is dependent on an 11-amino acid long, linear motif (LC3-interacting region; LIR) mediating an interaction with LC3/GABARAP family proteins.33,51,85 Our group initially mapped the LIR motif to a 22 amino acid-long sequence of human p62 harboring an evolutionarily conserved motif of three acidic residues followed by a tryptophan (DDDW in p62). A triple mutation of the three aspartic acid residues to alanine or a single mutation of the W to A abolished the interaction with LC3.33 Komatsu's group subsequently showed that the 11-amino acid-long conserved motif is sufficient to confer binding to LC3.51 The groups of Komatsu and Inagaki reported x-ray and NMR structures, respectively, of the LIR peptide of p62 docked onto LC3.51,85 The p62 LIR with its core sequence DDDWTHL appears as an extended β-strand and forms an intermolecular parallel β-sheet with the β2 strand of LC3. LIR is positioned in the interface between the N-terminal arm and the C-terminal Ub-like domain of the ATG8 protein. The side chains of the W and L residues are bound in two hydrophobic pockets (the “W-site” and “L-site”) in the Ub-like domain of LC3.51,85,86 In addition to these hydrophobic interactions, electrostatic interactions involving the three D residues of the LIR motif and basic residues in the N-terminal arm and Ub-like domain of LC3 (R10, R11, K49 and K50) are also important for the p62-LC3 interaction. The importance of both the WXXL core motif and the acidic residues of the p62 LIR motif was verified by substitution of these residues with alanine.33,85,87 The crucial role of the interaction with the N-terminal arm of LC3 for binding and autophagic degradation of p62 was demonstrated by Elazar's group.88
We have compiled a sequence logo from 25 different LIR motifs from 21 different proteins that all have been tested for binding to ATG8 family proteins (Fig. 2B). The LIR motifs come from published LC3B-, GABARAP- and/or GABARAPL1-interacting proteins,52,89–92 as well as 14 unpublished LIR motifs from 12 different proteins mapped and verified by our group. As is evident from the sequence conservation revealed by the logo, the LIR motif seems to be eight amino acids long. Adopting the nomenclature of Inagaki's group (X−3X−2X−1W0X1X2 LX3),86 acidic residues (D and E) are prevalent in the X−3X−2X−1 positions and in the X1 position. The residue binding to the W-site is usually W or F although the main LIR in NBR1 harbors a Y at that position,31 as does a LIR identified in human ATG4B.92 The L-site can be occupied by either L, I or V (Fig. 2B). Based on the current knowledge, a consensus LIR motif could be written as D/E-D/E-D/E-W/F/Y-X-X-L/I/V where there is not an absolute requirement for acidic residues at all three X−3X−2X−1 positions but usually there is at least one at these positions. If there are no acidic residues at these positions there are usually acidic residues at X1 or X2, or at both, like in yeast Atg19. The requirement for aromatic residues in the W-site seems absolute as is the requirement for large, hydrophobic residues in the L-site (L/I/V). As more LIR motifs are being characterized it is likely that we will see a pattern where proteins binding preferentially to the LC3 subfamily have somewhat different residue preferences at certain positions of the eight amino acid core relative to those that prefer binding to the GABARAP subfamily of mammalian ATG8 proteins.
Following the characterization of the LIR motif in p62, similar motifs were identified in other ATG8-interactors, including yeast Atg19,85 the cargo receptor in the yeast Cvt pathway. NBR1, Nix, calreticulin and clathrin heavy chain use LIR motifs to bind to LC3 and/or GABARAP subfamily members.31,38,93–95 We recently reported that a novel protein of hitherto unknown function(s) named FYCO1 (FYVE and coiled-coil domain-containing protein 1) binds to LC3, Rab7 and PtdIns-3-phosphate (PtdIns(3)P) on autophagosomes and autolysosomes.90 FYCO1 binds to LC3 via a LIR motif located adjacent to the FYVE domain. Our results suggest that FYCO1 forms an adaptor complex on the autophagosomes that is involved in kinesin-mediated microtubule plus end-directed transport of autophagosomes and other Rab7-positive vesicles. FYCO1 is recruited to the outer membrane of autophagosomes and is as such not an autophagy substrate or cargo receptor for autophagic degradation. Lately, an important adaptor protein in the Wnt signaling pathway, dishevelled (Dvl), was shown to interact with LC3 via a LIR motif and to be degraded by autophagy.52
The groups of Inagaki and Ohsumi have collaborated on showing that both the core autophagy components yeast Atg3 and human ATG4B interact with yeast Atg8 and human LC3B, respectively, via LIR motifs.91,92 In the x-ray crystal structure of ATG4B-LC3B the sequence motif YDTL in the N-terminal arm of ATG4B is bound to the LIR docking site (LDS) of the adjacent LC3B molecule where Y occupies the W-site and L the L-site.91 The biological relevance of this interaction is not yet understood. The yeast Atg3 LIR motif (DWEDL) is required for transfer of Atg8 from the Atg8-Atg3 thioester intermediate to PE, but not for the formation of the intermediate. Interestingly, the Atg3 LIR is required for the Cvt pathway, but not for starvation-induced autophagy, and is important for lipidation of Atg8 that is bound to the Cvt pathway cargo receptor Atg19.92
The list of LIR-containing ATG8 interactors is rapidly expanding. The group of Wade Harper conducted a proteomics screen of the autophagy interaction pathway in human embryonic kidney cells (HEK293T) under conditions of basal autophagy.96 Using immunoprecipitation experiments of tagged proteins and mass spectrometry (LC-MS/MS) to identify co-precipitated proteins, Berhends and coworkers also tested the six ATG8 orthologues LC3A, B and C, GABARAP, GABARAPL1 and GABARAPL2, and found 67 so-called high-confidence candidate interaction proteins.96 Among these only p62, NBR1, FYCO1, ATG3 and ATG4B were known from before. Interestingly, more components of the conjugation machinery such as ATG7, ATG10, ATG5 and ATG16L were identified as ATG8 interactors and many proteins in the ATG8 interaction network contain domains implicated in vesicle function including a number of Rab GTPase activating proteins. Also several protein kinases feature in this network.
Using mutants in the LIR docking site (LDS) Behrends et al. found a significant fraction of ATG8 interactors to show LIR-dependent binding.96 However, the presence of many proteins that could interact in vitro with ATG8 proteins mutated in their LDS strongly suggests that ATG8 family proteins harbor another interaction surface. Roughly one third of the interactors did not show subfamily specificity, whereas a third each associated with either the LC3 or GABARAP subfamilies. When this was tested using in vitro binding assays, the specificity was more relaxed. All proteins only associated with the LC3 subfamily by the proteomics approach could bind to GABARAP proteins in vitro. Hence, in vivo there may be more specific complexes formed than revealed by the specificities observed in in vitro binding assays.
The presence of a LIR motif is not sufficient for a protein to be efficiently degraded by selective autophagy. Analogous to the oligomerization and aggregation of preApe1 in the Cvt pathway, PB1 domain-mediated polymerization of p62 is absolutely required for its degradation by autophagy.51 Clearly, in addition to LIR-dependent binding, aggregation/polymerization is required. In order to form p62 bodies that are subsequently degraded by autophagy, p62 needs, in addition to the PB1 domain and the LIR motif, also the UBA domain that binds to Ub-labeled proteins.29,33
What are the signals or labels used to target substrates to a selective autophagy pathway? As already mentioned, there is increasing evidence that Ub is used as a signal for autophagic degradation of protein aggregates and other cellular substrates.78 A simplified general model for p62-mediated selective autophagy can be proposed (Fig. 3A): First, the substrate likely needs to be recognized by molecular chaperones or other cellular systems sorting it into an autophagy degradation pathway. Second, the substrate is ubiquitinated. Third, p62 is recruited and sequesters the autophagy substrates into larger units or aggregates. Via its interaction with LC3 the phagophore is recruited and the autophagosome forms around the aggregated substrate. The ability of p62 to recognize Ub substrates was nicely illustrated in a paper from the group of Lippincott-Schwartz. They demonstrated that ubiquitination of a long-lived, cytosolic soluble protein (red fluorescent protein; mRFP) and of peroxisomes was sufficient to target these very different substrates for autophagic degradation.37 Strikingly, p62 was required for targeting both types of substrates. The authors used a strategy where they labeled peroxisomes with monoUb by means of a GFP- and monoUb-tagged peroxisomal membrane protein, pMP34. Although this strategy represents an artificial situation, the results showed convincingly that peroxisomes are targeted for autophagy when they display monoUb on their membrane facing the cytosol (Fig. 4A). Clearly, monoUb can here act as a signal for pexophagy or selective autophagy of peroxisomes. Importantly, the authors also demonstrated that siRNA-mediated knockdown of p62 strongly inhibits the autophagic degradation of peroxisomes in the absence of artificial Ub labeling.37
A functional role of the Ub chains added in various forms of selective autophagy is obviously to recruit essential adapter proteins such as p62 and HDAC6. For most Ub-labeled autophagy substrates, it is neither known which Ub ligases are involved nor which specific proteins become ubiquitinated, or the type of Ub modifications involved. Both p62 and HDAC6 seem to have a preference for K63-linked polyUb chains,66,97 and the addition of K63-linked Ub chains has been reported to correlate with the clearance of protein inclusions by autophagy.98 However, different types of Ub modifications may be added to autophagy substrates and more studies are needed to clarify the in vivo binding preferences of p62, either alone or as part of a complex with NBR1 and ALFY (autophagy linked FYVE protein).31,99 It also seems plausible to ask if recruitment of p62 creates a positive feedback loop where p62 through interaction with E3 ligases induces further ubiquitination leading to recruitment of more p62. Such a positive feedback loop would help to explain the strong correlation between p62 and Ub-proteins observed for most autophagy substrates. p62 interacts with the E3 ligase TRAF6,100 and may also via KEAP1 be responsible for recruitment of the E3 ligase cullin 3.101 Ub ligases are present in most protein inclusions and the Ub chains present in mature p62 bodies or aggresomes may differ from the Ub chains initially used to recruit p62 or HDAC6. It is experimentally very difficult to distinguish a role for Ub in recognition versus a role for Ub in the construction of autophagy competent structures.
An important role for both p62 and HDAC6 is to sequester the autophagy substrates into larger units or aggregates before they become degraded by autophagy. Aggregation is very likely itself a signal for degradation by autophagy with no need for other labels as long as the aggregate is recognized directly by a cargo receptor or adapter that also can bind to ATG8 family proteins.
Structures similar to p62 bodies were first described and characterized in dendritic cells by the group of Philippe Pierre and named DALIS (dendritic cell aggresome-like inducible structures).102 Using puromycin to induce the formation of defective ribosomal products (DRiPs), they were able to show that DALIS are used for storage and ubiquitination of newly synthesized defective proteins.103 Their data also indicated that DALIS have an ordered structure. Puromycin-labeled DRiPs were recruited to the center of DALIS as early as 5 min after their formation. Their data also indicated that DRiPs are ubiquitinated after their entry into DALIS, and Ub conjugation enzymes such as E1, E2-UBC4/UBC5 and CHIP (E3 ligase) are found in the center of DALIS.103 Later studies revealed that structures similar to the DALIS can be formed in many cell types and the term ALIS was therefore introduced as a more general name for these structures.28 We subsequently showed that p62 is a major protein in ALIS and concluded that p62 bodies and ALIS are indistinguishable structures.33 The formation of p62 bodies is normally a transient event and their contents are degraded by autophagy or by the proteasome.28,33,63 These structures are highly ubiquitinated and are distinct from aggresomes and have a characteristic rounded appearance when visualized by light microscopy. They can adopt variable sizes (mostly 0.5–1 µm, but some can be several µm in diameter) and they are formed in response to various stressors.28,33 It should be noted that LC3 is efficiently recruited to p62 bodies.28,29,104,105 GFP-LC3-positive dots, commonly used as a marker for autophagosomes, may in addition to autophagosomes also represent some nonmembrane-bound p62 bodies. However, most of the p62 bodies visualized by fluorescence microscopy in the size range of 0.5–1 µm are sequestered into autophagosomes.29
The transient accumulation of p62 bodies is associated with stress conditions such as the accumulation of misfolded proteins or induction of autophagy.28 However, using inhibitors of autophagosomal maturation, we have observed that Ub-proteins are continuously delivered to autophagic vesicles also under normal growth conditions. What is interesting is that no accumulation of Ub-proteins in vesicular structures was observed in cells depleted for p62, indicating that p62 is essential for constitutive autophagy of misfolded proteins.33 We recently reported that two other p62 interaction partners, NBR1 and ALFY, are essential components of puromycin-induced p62 bodies, and both proteins are also needed for constitutive autophagy of Ub-proteins.31,99 In conclusion, we believe that p62, NBR1 and ALFY constitute a complex responsible for degradation of Ub-proteins via constitutive autophagy and for the stress-induced formation of p62 bodies (Fig. 3A).99
ALFY is a 400 kDa scaffold protein that was initially shown to be redistributed from the nucleus or nuclear envelope to autophagic structures in response to amino acid starvation or proteasomal inhibition.106 The protein was later found to interact with p62, and redistribution of ALFY in HeLa cells depends on the ability of p62 to shuttle between the cytoplasm and nucleus.99 Unlike p62 and NBR1, ALFY is not a selective autophagy substrate, and ALFY is probably degraded by autophagy only when associated with p62-bodies or other types of aggregates.35,99 Support for a role of ALFY in constitutive autophagy of Ub-labeled proteins also comes for studies of Drosophila lacking the ALFY homologue Blue Cheese (bchs). These flies have a reduced life span and accumulate Ub-positive inclusions and display neurodegeneration.107 The C-terminal part of ALFY contains a p62-interacting BEACH domain,99 an ATG5-interacting WD40 repeat region,35 and a PtdIns(3)P-binding FYVE domain.106 The large size of ALFY also makes it a candidate scaffold involved in the assembly of p62 bodies. However, more studies are needed to clarify the individual roles of p62, NBR1 and ALFY in the formation and autophagic degradation of p62 bodies.
The formation of p62 bodies under stress conditions is in part due to the increased expression of the p62 gene. In addition, the decision to make p62 bodies depends on the expression of specific molecular chaperones (Fig. 3A). Increased expression of the Hsp/Hsc70 co-chaperone Bag3 was recently shown to induce the formation of p62 bodies.30 The amount of misfolded proteins gradually increases during cell aging, and a shift from Bag1 to Bag3 expression during aging correlates with a shift from proteasomal degradation of misfolded proteins into autophagic degradation. Co-precipitation of Bag3 with p62 indicates that these proteins are in the same complex.30 An important interaction partner of Bag3 in selective autophagy of misfolded proteins is the small heat shock protein HspB8.108 HspB8 expression is induced by proteasomal inhibition.109 A complex consisting of Bag3 and HspB8 was initially implicated in autophagic clearance of mutant Huntingtin.108 Later a similar complex including CHIP was shown to be important for selective autophagy of SOD in amyotrophic lateral sclerosis (ALS).109 Furthermore, a complex of Bag3, HspB8, Hsc70 and CHIP, or a similar Drosophila complex containing the Bag3 homologue Starvin, were shown to be needed for constitutive autophagy in muscles.110 The complex is required for maintenance of Z disks and facilitates selective autophagy of damaged components such as filamin.110
Another group of proteins involved in the sorting and degradation of misfolded proteins are the ubiquilins (ubiquilin 1–4). Their domain architecture with an N-terminal ubiquilin (UbL) domain and a C-terminal UBA domain have clear similarities with that of p62 (Fig. 2A). The PB1 domain of p62 and the UbL domain of ubiquilin both interact with the S5A/Rpn10 subunit of the 26S proteasome, indicating a role in the shuttling of ubiquitinated substrates to the proteasome.66,111 Ubiquilin is involved in ER-associated degradation (ERAD) as part of a complex also containing Erasin and the AAA+ ATPase chaperone p97/VCP (valosin-containing protein).112 p97/VCP has an important role in ERAD in the delivery of retranslocated proteins to the UPS. ERAD substrates may then be ubiquitinated and shuttled to the 26S proteasome by ubiquilin. p97/VCP is a multifunctional protein that also binds to HDAC6 to regulate delivery of Ub aggregates to aggresomes.113 Ubiquilin was recently implicated in both CMA and autophagy and it is degraded by both pathways.114 For the latter pathway, ubiquilin is recruited to an LC3-containing complex and stimulates formation of autophagosomes. Although the interaction between LC3 and ubiquilin appears to be indirect, this suggests that ubiquilin may act as an autophagy receptor. It is not known whether p62 and ubiquilin have redundant roles or are associated with different subfractions of misfolded proteins.
While we have shown that p62 is required for the delivery of Ub-proteins in autophagic vesicles,33 Ub-labeled autophagosomes accumulate in cells either lacking or expressing disease-associated mutants of p97/VCP.115 Hence, loss of p62 or p97/VCP seems to block different steps in the same autophagy pathway. This is particularly interesting since mutations in p62 or p97/VCP can both lead to Paget disease of the bone.
The importance of constitutive autophagy of misfolded proteins has been strongly supported by autophagy knockout studies. Loss of ATG5 or ATG7 in mice results in the accumulation of Ub aggregates in neurons and hepatocytes although there is no evidence that proteasomal degradation is affected.116,117 Knockdown of autophagy in flies similarly results in the accumulation of Ub aggregates.118,119 The most likely explanation is that a significant amount of Ub-labeled proteins are degraded by constitutive autophagy under normal conditions. Constitutive autophagy may in particular be important in differentiated, non-proliferating cells. Tissue-specific knockout of ATG genes causes accumulation of p62-positive protein aggregates in neurons,116,117,120 hepatocytes,32 skeletal muscle,121,122 cardiac muscles,123 pancreatic β-cells,124,125 and podocytes in the kidneys.126
It has been demonstrated, both in cell culture and in vivo, that mammalian p62 and the Drosophila homologue Ref(2)P are required for formation of ubiquitinated protein aggregates.29,32,33,127 Strikingly, Komatsu and coworkers found that the observed accumulation of Ub-positive protein aggregates in hepatocytes upon liver-specific knockout of the Atg7 gene in mice is dependent on p62. The hepatomegaly and liver dysfunction is strongly ameliorated by simultaneous knockout of p62.32 The high level of Ub aggregates observed in hepatocytes of autophagy-deficient mice indicates that this organ may be particularly sensitive to p62 accumulation.
p62 and Ub are also major constituents of two types of aggregates found in chronic liver diseases, i.e., intracellular hyaline bodies (IHBs) found in hepatocellular carcinoma, and Mallory-Denk bodies (MBs) characteristic for alcoholic and nonalcoholic steatohepatitis.128 p62 is needed for the formation of both these structures.74 In particular, IHBs seem to resemble p62 bodies, whereas MBs are distinguished from IHBs in that they contain abnormal keratins in addition to p62 and Ub.128
One discrepancy that has been noted is the much lower level of Ub proteins in Atg7/p62 double-knockout tissues as compared to Atg7-deficient tissues.32 One likely scenario is that Ub-proteins are less aggregated in p62-deficient cells and therefore can be more easily degraded by CMA or the proteasome. There is crosstalk between different degradation pathways for misfolded proteins, and the loss of one degradation system may result in the activation of other systems.129 However, a high constitutive level of p62 caused by autophagy inhibition may itself contribute to an increased formation or decreased degradation of Ub-proteins. It is suggested that p62 accumulating due to autophagy inhibition delays the delivery of ubiquitinated proteins to the proteasome.130 Also, at least part of the pathophysiological effects seen in the autophagy-deficient liver can be attributed to chronic activation of Nrf2 caused by high levels of p62.131 In the brain of p62 knockout mice, K63-linked Ub-proteins increase and the mice display an Alzheimer-like phenotype.132,133 Distinct from the liver, the presence of p62 does not exaggerate neurodegeneration in autophagy-deficient mice.32
The term aggresome denotes a general type of protein aggregate that is induced by the transport of smaller Ub-aggregates along microtubule tracks into the MTOC region where the structure is formed usually encaged by intermediate filaments.134 This is believed to be a protective mechanism since small and diffuse aggregates are more toxic than aggresomes.135 HDAC6 is required for the formation of aggresomes (Fig. 3B). It binds to Ub and dynein molecular motors, and this enables HDAC6 to shuttle soluble Ub-aggregates to aggresomes.136,137 HDAC6 binds Ub via a C-terminal BUZ domain and has preference for K63-linked chains.97,113 Lewy bodies in Parkinson disease are characterized by the presence of aggregated DJ-1 or α-synuclein and are among the best studied aggresomal structures. Mutations in the E3 ligase Parkin are common in recessive forms of Parkinson. Parkin was, in a recent study, reported to cooperate with the heterodimeric E2 enzyme UbcH13/Uev1a to mediate K63-linked polyubiquitination of misfolded DJ-1, and this is essential both for HDAC6-mediated transport of DJ-1 aggregates and formation of Lewy bodies.97 Another essential role of HDAC6 is to promote the creation of an actin network enhancing fusions between aggresome-containing autophagosomes and lysosomes.138 HDAC6 may also be important for the transport of the autophagy machinery to the aggresomes.136 Both in mammals and flies, HDAC6 promotes autophagic degradation of polyglutamine inclusions.136,139 Autophagy is important for degradation of aggresomes since proteins must be unfolded before they can be degraded by the proteasome or CMA.27 Several studies present evidence for an autophagic degradation of polyglutamine inclusions.136,140–142 However, dependent on the specific aggregating protein involved, not all types of aggresomes can be degraded by autophagy,143 and it is often unclear whether it is the large aggregates or their precursors that are most efficiently degraded.
The presence of p62 is a common feature of most cytoplasmic and nuclear protein aggregates found in human diseases.74 Most pathological aggregates are induced by a specific protein other than p62 and the recruitment of p62 is a later event induced by the ubiquitination of the aggregates (Fig. 3B). Hence, p62 is for example not needed for the formation of polyglutamine inclusions of Huntington disease,144 and the presence of p62 in aggregates is believed to reflect a role of p62 in autophagic degradation. In support of this, LC3B mutants defective in p62 binding are defective in autophagy-mediated clearance of polyglutamine inclusions.145 What should be noted is that NBR1 and ALFY are likely to be present together with p62 in protein inclusions.31,35,99 Do these three proteins collaborate to enhance the autophagic degradation of aggresomes, analogous to their role in the formation and degradation of the transient p62 bodies? ALFY is recruited to polyglutamine inclusions as part of a complex containing p62, NBR1, LC3, ATG5, ATG12 and ATG16L.35 ALFY is required for efficient autophagic clearance of polyglutamine or α-synuclein inclusions in mammalian cells dependent on a direct interaction between ALFY and ATG5. Furthermore, overexpression of ALFY promotes degradation of polyglutamine inclusions in a neuronal lentiviral model, and overexpression of the Drosophila homologue Bchs diminishes neurotoxicity in a Drosophila eye model of polyglutamine toxicity.35 For further discussions on ALFY, see the article by Yamamoto and Simonsen in this issue of Autophagy.
One intriguing observation is that nuclear p62 and ALFY also collaborate in the formation of nuclear aggregates that in several ways seem to be similar to the p62 bodies.99 These structures are associated with nuclear PML bodies.35,63,99 PML bodies have been proposed to function as sites for proteasomal degradation of misfolded proteins.146 The recruitment of p62 and ALFY to PML bodies is needed for accumulation of Ub-proteins in PML bodies.63,99 p62 also contributes to the recruitment of PML bodies and proteasomes to, or close to, nuclear inclusions formed by polyglutamine-expanded ataxin1Q84, and our data suggest a role for p62 in degradation of these inclusions.63
The idea of selective removal of damaged or surplus mitochodria by autophagy (mitophagy) has recently received support from studies lending some mechanistic insight into this complex process. Using yeast genome-wide screening, the groups of Ohsumi and Klionsky independently discovered the mitophagy receptor, Atg32.40,41 The Atg32 protein is a single-spanning mitochondrial outer membrane protein with a free N-terminal region facing the cytosol. Atg32 mutant yeast cells are defective in mitophagy. Atg32 binds both to the selective autophagy factor Atg11 and to Atg8, and functions specifically in mitophagy. Atg32 has a LIR motif important for the interaction with Atg8.40 In yeast there seems to be no involvement of Ub in mitophagy.
In mammalian cells, loss of membrane potential of damaged mitochondria was proposed to lead to autophagic degradation, as suggested from studies of hepatocytes from GFP-LC3 transgenic mice.147 Mitophagy is important during the developmental elimination of mitochondria in reticulocytes. Nix/BNIP3L is essential for this clearance of mitochondria occurring during erythrocyte maturation.148,149 The Dikic group recently showed that Nix/BNIP3L recruits the ATG8 family Ub-like modifier GABARAPL1 to damaged mitochondria via a LIR motif in its N-terminal region.38 Ablation of the Nix:LC3/GABARAP interaction slows mitochondrial degradation in murine reticulocytes. Also, Nix recruits GABARAPL1 to depolarized mitochondria. Hence, Nix acts as a mitophagy receptor for selective clearance of mitochondria during erythrocyte differentiation (Fig. 4B). Using a phage display screening, the Wilbold group also identified Nix as binding to GABARAP via a LIR motif.95 The role of Nix/BNIP3L in mitophagy is covered in detail in the article by Novak and Dikic in this issue of Autophagy.
Several recent reports have described a specific mitophagy pathway that is Ub-dependent and has been referred to as Parkin-mediated mitophagy (Fig. 4C). Richard Youle's group provided the first evidence for a role of Parkin in mitophagy of damaged mitochondria showing that Parkin is recruited to and important for the degradation of damaged mitochondria in response to the treatment of cells with ROS or the uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP).42 They also subsequently showed that PINK1 rapidly accumulates on damaged mitochondria and that PINK1 stabilization on the mitochondria is both necessary and sufficient for recruitment of Parkin.150 The Springer group reported that PINK1, Ub, p62 and the voltage dependent anion channel 1 (VDAC1) are required for Parkin-mediated mitophagy in both neuronal and non-neuronal cells.39 PINK1 interacts with Parkin and is needed for the translocation of Parkin to the membrane of damaged mitochondria. Parkin-mediated ubiquitination then results in accumulation of p62, clustering of mitochondria and autophagic degradation.39 See also the article by Springer and Kahle in this issue of Autophagy. The group of Tso-Pang Yao demonstrated that Parkin-mediated ubiquitination of damaged mitochondria also serves to recruit HDAC6, and that HDAC6 is needed for efficient degradation. They compared the role of HDAC6 in mitophagy with the role of HDAC6 in aggresome formation, and showed that HDAC6-mediated transport of damaged mitochondria results in the formation of perinuclear mito-aggregates resembling aggresomes.151 Other Parkin substrates relevant for mitophagy are the mitofusins which become ubiquitinated in a PINK1- and parkin-dependent manner following mitochondrial damage.152,153 The mitofusins are transmembrane GTPases located in the mitrochondrial outer membrane involved in regulating fusion of mitochondria. Ubiquitination of mitofusins may be involved in segregating damaged mitochondria for mitophagy, via an inhibition of mitochondrial fusion events, and/or the ubiquitinated mitofusins may act as tags for autophagic cargo receptors.152,153
Nix is also important for Parkin-mediated mitophagy as reported in a study on CCCP-induced mitophagy, concluding that Nix is required for translocation of Parkin to the damaged mitochondria.154 Recent papers from the groups of Komatsu and Youle disagree with the Springer group on the exact role of p62 in mitophagy.155,156 Both the Komatsu and Youle groups demonstrate that p62 is required for Parkin-induced mitochondrial clustering, but using p62 knockout murine embryo fibroblasts they claim p62 is not required for mitophagy. The Youle group shows this also using siRNA knockdown in HeLa cells and finds that VDAC1 is dispensable for both the clustering and autophagic degradation of mitochondria following CCCP-treatment of cells.156 However, it seems clear that Parkin (but not a pathogenic mutant) promotes K63-linked polyubiquitination of mitochondrial substrate(s) leading to recruitment of p62. Thereafter, p62 aggregates dysfunctional mitochondria dependent on its PB1-domain-mediated homopolymerization similar to its aggregation of polyubiquitinated proteins. In the study by Komatsu and coworkers, a delay in mitochondrial degradation was observed in p62 KO MEFs complemented with a LIR mutant of p62 (W340A) suggestive of involvement of p62 in Parkin-mediated mitophagy. A possible explanation for the different results on the role of p62 in Parkin-mediated mitophagy can be functional redundancy.155 NBR1 is a likely candidate that may compensate for the absence of p62. More studies are needed to resolve this conundrum. The essential substrates ubiquitinated by Parkin in Parkin-mediated autophagy remain to be identified. It is also important to clarify the differences between mitophagy pathways utilized in degradation of damaged versus surplus mitochondria.
Autophagy used as a defense system against infection has been termed “xenophagy”.4,157 Several articles in this thematic issue of Autophagy discuss the selective autophagy of bacteria (See Shahnazari et al., Cemma et al., Ogawa et al., Randow, and Ponpuak and Deretic). Autophagic degradation of invading bacteria is important in order to protect cells from bacterial colonization. Invading Salmonella enterica serovar Typhimurium (S. typhimurium) reside and replicate within intracellular compartments termed Salmonella-containing vacuoles (SCVs), but the bacteria can also be released into the cytosol and are then targeted by the host Ub conjugation system. Two different autophagy receptors were recently shown to be involved in autophagic degradation of Ub-labeled S. typhimurium.46,48 The group of John Brumell implicated p62 in this process (Fig. 4D). They showed that p62 and LC3 are recruited to the surface of Ub-labeled bacteria, and the recruitment of p62 is needed for efficient autophagy of the bacteria. The UBA domain is critical for the recruitment of p62 to Ub-positive S. typhimurium, and both the UBA and LIR domains are required for autophagy of the bacteria.48 In this study siRNA knockdown of endogenous NBR1 suggested that NBR1 does not play any major role in the autophagy targeting of the bacteria. In a separate study, the group of Felix Randow characterized another autophagy receptor, NDP52 (nuclear dot protein 52 kDa) (Fig. 2A), which interacts directly with LC3 and is needed for efficient autophagy of S. typhimurium (Fig. 4D). NDP52 uses a C-terminal zinc finger to bind to Ub chains located on cytosolic bacteria.46 Both p62 and NDP52 restrict intracellular replication of S. typhimurium. Interestingly, p62 and NDP52 are recruited independently to distinct microdomains surrounding the bacteria and cooperate for efficient autophagy of the intracellular bacteria (See article by Cemma et al. in this issue of Autophagy). It remains to be clarified whether a putative LIR motif in the N-terminal region of NDP52 mediates the binding to LC3.79
Other intracellular pathogens like Shigella flexneri and Listeria monocytogenes are adapted to a cytosolic lifestyle and they break the vacuolar membrane after entry into host cells (Fig. 4E). Whereas S. enterica become ubiquitinated and degraded by autophagy, pathogens that replicate in the cytosol seem to avoid host ubiquitination.158 An interesting study on L. monocytogenes illustrates how this bacterium depends on the bacterial protein ActA to escape from autophagic degradation.47 ActA-mediated recruitment of the Arp2/3 complex, Ena/VASP and actin prevents ubiquitination of the bacteria and thereby protects the bacteria against autophagic recognition. In contrast, ActA deletion mutants are ubiquitinated, recognized by p62, and degraded by autophagy.47 Another study on S. flexneri showed that vacuolar membrane remnants are polyubiquitinated, recognized by p62 and targeted for autophagic degradation (Fig. 4E).44 These membrane remnants contain the p62-interacting Ub ligase TRAF6. TRAF6 activates NEMO and both NEMO and p62 act in the NFκB signaling pathway promoting inflammatory and prosurvival pathways.44 It is not known if NBR1 is acting together with p62 in xenophagy of ActA-deleted Listeria and Shigella membrane remnants.
A very different role of p62 in xenophagy has been reported by the group of Vojo Deretic. In a recent study, they found that in addition to the delivery of Mycobacterium tuberculosis into autolysosomes, efficient killing of the bacterium depends on a p62-mediated uptake of specific ribosomal and bulk ubiquitinated cytosolic proteins into the same autolysosomes.54 Interestingly, p62 is in this case not needed for the autophagic delivery of the bacterium, but only for the delivery of the bactericidal precursors that are processed to active compounds within the autolysosomes (Fig. 4E).
Evidence for the use of p62 as an adaptor for selective autophagy also in antiviral defense comes from the group of Beth Levine.45 They studied the role of autophagy in mice where the central nervous system was infected with a lethal Sindbis (SIN) virus and showed that SIN capsid proteins are cleared by selective autophagy involving p62. The clearance of surplus capsid proteins somehow protects against cell death although viral replication is not affected. p62 and SIN proteins accumulate in mouse neurons lacking ATG5 function, and they could co-immunoprecipitate p62 with SIN capsid protein in virus-infected HeLa cells. Knockdown experiments show that p62-mediated selective autophagic degradation of the capsid protein reduces virally-induced cell death. It is presently not known if there is a direct interaction between p62 and SIN capsid protein (Fig. 4F). See the review on the role of p62 in the degradation of the Sindbis virus capsid in the article by Sumpter and Levine in this issue of Autophagy.
In yeast, Atg11 acts as an adaptor protein for Ub-independent selective autophagy. Atg11 is responsible for the recruitment of autophagy substrates to the PAS where phagophores are made (Fig. 5). In the Cvt pathway, the delivery of the Ape1 complex to Atg11 is mediated by the cargo receptor Atg19 that interacts with the Ape1 complex, Atg11 and Atg8.7 The cytosolic cysteine protease Lap3 is selectively degraded by autophagy utilizing the Cvt pathway components because it binds to the prApe1 oligomer and is transported to the vacuole under nitrogen starvation in a manner dependent on both Atg19 and Atg11.159 Similar to Lap3, yeast Ald6 is degraded by selective autophagy upon induction of bulk autophagy by nutrient limitation.160 However, in contrast to Lap3, the cytosolic acetaldehyde dehydrogenase Ald6 is transported by autophagosomes to the yeast vacuole by an unknown mechanism. Because overexpression of Ald6 is toxic during nutrient starvation, its activity may need to be regulated by specific autophagic degradation to ensure cell survival upon nutrient limitation.160
Mitophagy in yeast can be induced by a reduction in membrane potential, nitrogen starvation or entry into stationary phase. There seems to be a single mitophagy pathway in yeast that depends on the two outer membrane proteins Atg32 and Atg33. Similar to Atg19, Atg32 contains a LIR motif that binds to Atg8 and it also interacts with Atg11.40,41 In contrast, Pichia pastoris Atg30, the cargo receptor for pexophagy in this methylotrophic yeast, has no LIR motif, but is located on peroxisomal membranes and interacts with PpAtg11.161 A detailed discussion of mitophagy in yeast is found in the article by Wang and Klionsky in this issue of Autophagy.
In mammals, there seems to be both Ub-dependent and Ub-independent mitophagy pathways, and the current belief is that Nix/Bnip3L, but not p62, is essential for Ub-independent mitophagy (see above). There is also a recent report indicating that p62 can be involved in Ub-independent selective autophagy of misfolded proteins. A mutant of superoxide dismutase 1 (SOD1) causing amyotrophic lateral sclerosis (ALS) is degraded by autophagy in a Ub-independent way due to a direct interaction with p62.162 It is also likely that ubiquitination is not involved in p62-mediated selective autophagy of Sindbis virus capsid proteins (see above). Hence, this may be another example of Ub-independent selective autophagy like the one reported for the mutant SOD1 protein. It remains to be determined how this protein is recognized by p62.
Similar to p62 and NBR1, the SEPA-1 protein in C. elegans oligomerizes and forms cytoplasmic aggregates that are selectively removed by autophagy during embryogenesis. Furthermore, SEPA-1 acts as a cargo receptor or adaptor for autophagic degradation of a specialized type of protein aggregates called germ P granules by binding directly to the P granule component PGL-3 and to the ATG8 homologue LGG-1.163 This way the P granules become sequestered in autophagosomes and degraded by autophagy. Although Ub is not involved, this is very similar to the selective autophagy mediated by p62.
Both p62 and NBR1 are involved in cellular signaling processes that seem unrelated to autophagy. In these cases selective autophagy can be an important mechanism to regulate the level and lifetime of the signaling complexes they participate in. The groups of Jorge Moscat and Sylvia Stabel found p62 to bind to atypical protein kinase C,58,60 and it was subsequently demonstrated that p62 acts as a scaffold or adapter protein in NFκB signaling pathways.164–166 In addition to the atypical protein kinase Cs, PKCζ and PKCλ/ι, p62 also binds to the kinases MEKK3, MEK5 and ERK1.59–62,167 Clearly, many of the phenotypes observed in p62 knockout mice can be attributed to defects or perturbations in cellular signaling pathways.168 Studies involving knockout, knockin and transgenic mice show that p62 plays important roles in bone remodelling, obesity, oxidative stress response and cancer development.70,131,169–172
An antagonistic relationship between p62 and ERK1 activation has been established,167,170 and increased autophagy of p62 under hypoxia is responsible for activation of the ERK1/2 pathway.173 This shows that removal of p62 by selective autophagy can have an impact on signaling (Fig. 6A). A recent study performed with both fly and mammalian macrophages show that constitutive, p62-mediated selective autophagy is required for cell spreading and Rho1-induced cell protrusions. It is suggested that p62 may mediate selective autophagic degradation of a regulator of the Rho pathway.174 In the Wnt signaling pathway the adaptor protein Dvl is degraded by selective autophagy to downregulate the pathway.52 The E3 ligase VHL (Von Hippel-Lindau) ubiquitinates Dvl2 which is then bound by p62, which in turn facilitates aggregation and LC3-mediated autophagosome recruitment of Dvl2 under starvation conditions. The details and biological roles are further discussed in the article by Gao and Chen in this issue of Autophagy.
The p62 bodies have been suggested to act as signaling hubs where p62 interacts with TRAF6 and caspase-8, and may affect both prosurvival and pro-apoptotic signaling pathways.175,176 Along the same lines, the polyubiquitinated vacuolar membrane remnants resulting from Shigella entry that are recognized by p62 and targeted for autophagic degradation, also act as signaling hubs promoting inflammatory and prosurvival pathways.44 Analogous to p62, another autophagic adaptor or cargo receptor recognizing Ub-labeled invading bacteria, NDP52, is part of a TBK1 signaling complex containing the kinases TBK1 and IKKε as well as the scaffold proteins Nap1 and Sintbad.46 Hence, both p62 and NDP52, which both may help clear Ub-labeled intracellular bacteria by autophagy,46,48 act as true innate immune receptors involved in cytosolic surveillance. Interestingly, p62 is induced by lipopolysaccharide (LPS) and is also involved in regulation of aggregation of the MyD88 adaptor in Toll receptor signaling, suggesting that p62 may be regulated by, and act as a regulator of, innate immunity signaling pathways.177,178
p62 is a proteotoxic stress response protein and is induced at the mRNA and protein level by oxidants, sodium arsenite, cadmium, ionophores, proteasomal inhibitors or overexpression of polyglutamine-expanded proteins.144,179 The transcription factor Nrf2 induces a battery of genes upon oxidative stress as part of a protective response against oxidative damage to organelles and macromolecules. Under conditions without oxidant or electrophile stress, Nrf2 is bound to KEAP1, an adaptor for the Cul3 E3 ligase complex, and rapidly degraded by the proteasome. But upon oxidative stress KEAP1 is modified on specific cysteine residues, and Nrf2 is stabilized and turns on a set of antioxidant response genes.180–182 p62 was recently shown to be able to induce Nrf2 when overexpressed, and the induction of p62 is itself strongly inhibited in Nrf2 knockout mice.183,184 Earlier this year, five different groups reported that p62 binds to KEAP1 and thereby mediates the induction of Nrf2.101,131,185–187 Three groups mapped a so-called KIR motif in p62 that binds to the Kelch repeat domain of KEAP1.101,131,187 The LIR and KIR motifs are located directly adjacent to each other, and the binding to LC3/GABARAP family proteins can be competed by KEAP1 in vitro and slows the degradation of GFP-p62, in a reporter cell system to monitor autophagic degradation of p62,188 suggesting that there is a mutually exclusive binding to either the LIR or the KIR in a single p62 molecule.187 However, since p62 is a polymer, both KEAP1 and LC3 may be bound by a polymeric p62.
A pertinent question to ask is if KEAP1 is degraded by autophagy. We found that overexpressed KEAP1 is degraded by autophagy in a p62-dependent manner. However, due to lack of good antibodies for immunofluorescence and EM studies of endogenous KEAP1 we have been unable to determine whether endogenous KEAP1 is a selective autophagy substrate or not. Copple et al. found that knockdown of p62 leads to increased protein levels of endogenous KEAP1 caused by a decrease in its degradation rate.185 This could be due to reduced degradation of KEAP1 via autophagy. On the other hand, Fan et al. suggest that KEAP1 is not itself an autophagy substrate but forms a complex with p62 and LC3 under oxidative stress conditions to facilitate efficient removal of protein aggregates via autophagy.186 Clearly more studies are needed to determine if endogenous KEAP1 is degraded by selective autophagy or by the proteasome, or both. We also demonstrated that Nrf2 binds together with a small Maf protein to an antioxidant response element in the p62 upstream region and activates transcription of the gene. This way p62 sets up a positive feedback loop in the KEAP1-Nrf2 pathway (Fig. 6B).187
These findings connect the earlier reports about p62 as a stress response protein directly to an oxidative stress response signaling pathway. Related to this, Eileen White's group showed that under conditions of metabolic stress and autophagy deficiency the failure to properly regulate p62 levels leads to oxidative stress and increased tumorigenicity.189 High levels of p62 due to autophagy defects is sufficient to alter NFκB regulation and gene expression and to promote tumorigenesis. Previously, it had been shown that p62 is required for efficient transformation of murine embryo fibroblasts by the activated Ha-Ras oncogene, and for tumor formation in a murine model mimicking human lung adenocarcinoma.169 The groups of Moscat and White report opposite effects on the regulation of NFκB signaling. The experiments performed are not directly comparable, so more studies are needed to elucidate the reasons for this discrepancy. Depending on the level of p62 there may clearly be different outcomes. On one hand, p62-mediated selective autophagy is part of the antioxidative response that protects cells against ROS-induced damage of organelles and genome integrity. On the other hand, blockade of autophagy leads to p62-mediated accumulation of Ub aggregates that increases ROS and leads to DNA damage, thereby augmenting tumorigenesis.190 It is also highly likely that selective autophagy may help tumor cells survive under conditions of metabolic stress and that efficient transformation by many oncogenes may depend on selective autophagy.
An important focus for future studies will be to identify and characterize more cargo receptors for selective autophagy and identify the nature of their substrates. Such proteins must be able to interact with ATG8 family proteins and with their substrate as well as being able to multimerize to form an aggregated structure. However, a self-association property is not needed for cargo receptors sitting in the membranes of organelles, like Nix and Atg32. It will be interesting to see whether all cargo receptors depend on a LIR motif for their interaction with ATG8 family proteins or if there are other motifs mediating binding to another surface(s) of the ATG8 family proteins. An obvious first place to look for novel cargo receptors will be among the recently characterized ATG8-interacting proteins. In addition, other screens, like yeast two-hybrid screens, are likely to yield novel cargo receptors. Surprisingly few human cargo receptors in addition to p62 have so far been characterized. These are NBR1, Nix and NDP52, and in yeast there is Atg19 and Atg32. This is in stark contrast to the large number of LC3/GABARAP-interacting proteins that have now been identified. Many of these interacting proteins can be assumed to have roles in vesicle transport, fusion processes or other processes that may be distinct from autophagy. However, analyses of these proteins will very likely result in the identification of more proteins important for autophagosome formation, autophagosome maturation and/or selective autophagy. To improve our understanding of selective autophagy, it is clearly also very important to elucidate the different roles played by the different members of the mammalian LC3 and GABARAP subfamilies.
It is important to understand how autophagy substrates are selected and labeled for degradation. For Ub-labeled substrates, are there distinct Ub linkages responsible for sorting substrates for degradation by the proteasome or by selective autophagy? For misfolded proteins that can be degraded by several systems, it will be important to further analyze the roles of molecular chaperones, and how they respond to signaling events regulating the different degradation pathways. There is likely to be a fine-tuned regulation between different degradation pathways. Aggregation versus degradation is another interesting layer of regulation that is particularly relevant for autophagy substrates. The detection limit of the light microscope has restricted our ability to investigate the fate of small aggregates. However, new developments in live cell imaging and high resolution fluorescence microscopy will shed new light on this problem highly relevant for protein aggregation diseases.
It will be particularly interesting and likely challenging to elucidate roles of selective autophagy in different signaling pathways. The roles played by p62 in tumorigenesis are clearly linked to the autophagy status of the tumor cells as well as to the environmental stress that they may experience. The role played by p62 in mitophagy may be linked to tumorigenesis because of the tumorigenic effect of ROS produced by uncleared damaged mitochondria. However, as mentioned above there are unresolved issues that need to be clarified about the role of p62 in Parkin-mediated mitophagy. Is it a cargo receptor here or only an “aggregator”?
The contribution of autophagic cargo receptors, like p62 and NDP52, in innate immunity and signaling during xenophagy is another field where future research is expected to give important insight. In general, for several of the complicated selective autophagy processes the level of redundancy of autophagic cargo receptors is important to decipher. Related to this, we know that p62 interacts with NBR1 and ALFY, but we do not know much about under what conditions these three proteins collaborate and when/where they have distinct roles.
In conclusion, research on selective autophagy has the potential to contribute strongly to increased understanding of, and identification of, drug targets for important diseases like cancer, neurodegenerative diseases, diabetes and infections. Although initiated in the early 1960s, the research on different aspects of autophagic processes, including all variants of autophagy, has grown almost exponentially since the start of this century and it will be very interesting to see the future development of this fascinating field of research.
This work was supported by grants from the FUGE program of the Norwegian Research Council, the Norwegian Cancer Society, the Aakre Foundation and the Blix Foundation to T.J.