Post-translational modifications of ATG8 proteins – an emerging mechanism of autophagy control Free

Autophagy is a conserved catabolic pathway that allows the degradation of cellular components via their delivery to lysosomes. These components can include surplus, damaged or potentially toxic organelles and also invading pathogens, thereby making autophagy a crucial process to maintain cell homeostasis and cope with stressful insults that can damage cell and organism integrity (Dikic and Elazar, 2018). Accordingly, autophagy plays critical roles in healthspan and longevity. Conversely, autophagy deficiencies and its malfunction has been connected with accelerated aging and multiple age-related diseases (Aman et al., 2021; Nieto-Torres and Hansen, 2021).

Differences in the molecular pathways by which cellular components or ‘cargo’ are destined for degradation in lysosomes, intracellular vesicles containing acidic hydrolases, define autophagy subtypes. During chaperone-mediated autophagy (CMA), proteins containing a KFERQ-like amino acid motif are delivered to the lysosomal lumen via specialized transmembrane transporters (Kaushik and Cuervo, 2018). In microautophagy, direct invagination of the lysosomal membrane drives the engulfment of cargo into the lysosomal lumen. Cargo degraded via microautophagy targets proteins, protein aggregates, lipids and pieces or fragments of cellular organelles (Schuck, 2020). This Review will focus on a third form of autophagy, macroautophagy. Macroautophagy (hereafter referred to as ‘autophagy’) allows for the degradation of various cargoes, for example, proteins, lipids, pathogens or whole cellular organelles, such as mitochondria. These cargoes are delivered to lysosomes via a specialized, intracellular vesicle called the autophagosome (Dikic and Elazar, 2018). Autophagosomes are de novo-formed double-membrane vesicles that act as carriers of the cellular cargo for degradation in the lysosome following autophagosome–lysosome fusion (Dikic and Elazar, 2018).

The autophagy machinery directs the initiation and expansion of the membrane needed for autophagosome formation. ATG8 family proteins (hereafter ATG8s; see below for details) are regulators of this process, and almost a third of the autophagy molecular machinery is dedicated to the incorporation of ATG8s into both the inner and outer membranes of autophagosomes (Lystad and Simonsen, 2019). The autophagy machinery is typically organized into functional complexes (Fig. 1A–C), that sequentially start with the ULK1 initiation complex, which dictates the site where autophagosomes are created (Dikic and Elazar, 2018). ULK1 is a serine/threonine kinase that, aided by regulatory proteins, including ATG13, ATG101 and FIP200 (also known as RB1CC1), phosphorylates multiple downstream targets. These events lead to the activation of the class III phosphatidylinositol-3 kinase complexes (for which PI3KC3 is the catalytic subunit) and the generation of an isolation membrane (also known as a phagophore) which, upon closure, forms the autophagosome (Fig. 1A) (Dikic and Elazar, 2018).

Accumulating reports indicate that during the selective degradation of cargo, known as selective autophagy, receptors, which are proteins that recognize and target cargo for degradation, initiate the recruitment and activation of the ULK1 complex, to trigger autophagosome formation at a given subcellular location (Lazarou et al., 2015; Turco et al., 2019). Upon the formation of phosphatidylinositol-3-phosphate (PI3P) from phosphatidylinositol in the isolation membrane, an event catalyzed by the PI3KC3 complex, the isolation membrane is recognized by PI3P-binding proteins, including the WD-repeat domain phosphoinositide­interacting proteins (WIPIs) (Fig. 1B) (Grimmel et al., 2015), and the binding protein ATG2 (which has ATG2A and ATG2B forms in mammals) (Obara et al., 2008). In parallel, ATG9 (ATG9A and ATG9B in mammals), the only transmembrane protein of the autophagy machinery, is delivered in lipid vesicles (Fig. 1B). In a coordinated fashion, ATG2, via its lipid transfer activity, delivers lipids from the endoplasmic reticulum (ER) membrane to the isolation membrane, while the scramblase activity of ATG9 accommodates these lipids among the membrane leaflets, facilitating the growth of the isolation membrane (Ghanbarpour et al., 2021; Guardia et al., 2020; Matoba et al., 2020; Osawa et al., 2019). WIPIs then help recruit the ATG8-conjugation machinery to the growing isolation membrane (Fig. 1C) (Dooley et al., 2014). Preceded by a series of ubiquitin-like conjugation events, ATG8s get incorporated via a lipid ‘anchor’ into the isolation membrane (Lystad and Simonsen, 2019). Specifically, these conjugation events start with the protease ATG4 (ATG4A–ATG4D in mammals), which cleaves ATG8s close to their C-terminus, exposing a glycine residue that allows a sequential handoff of covalent linkages to different autophagy proteins (Tanida et al., 2004a,b). These events start with the conjugation of ATG8s to ATG7 and then ATG7-bound ATG8 is transferred and linked to ATG3. Subsequently, the complex ATG5–ATG12–ATG16L1, whose formation is facilitated by the protein ATG10 promotes the transfer and covalent linkage of ATG8s to the lipid phosphatidylethanolamine (PE) in the inner and outer membranes of forming autophagosomes (Ichimura et al., 2000; Kaiser et al., 2012; Nemoto et al., 2003; Taherbhoy et al., 2011). Here, ATG8s will perform critical functions for the autophagy process including promoting isolation membrane growth, cargo recognition, autophagosome formation, and transport, as well as fusion with lysosomes (Fig. 1D) (Lystad and Simonsen, 2019). Of note, the pool of ATG8s conjugated to the inner autophagosome membrane will be degraded by lysosomal proteases at the culmination of autophagy. However, the fraction of ATG8s conjugated to the outer membrane is proteolytically cleaved by the protein ATG4 and recycled in subsequent autophagy cycles (Lystad and Simonsen, 2019; Nair et al., 2012; Yu et al., 2012). Notably, ATG8s have recently been described to also incorporate into single-membrane vesicles of endolysosomal origin to perform degradative as well as secretory functions (Nieto-Torres et al., 2021a) (see Box 1 for a brief overview of such non-canonical functions for autophagy genes).

The autophagy process and many of the steps outlined above are usually elicited following fast kinetics, to cope with the rapidly changing energetic and metabolic demands of the cells. It is therefore not surprising that autophagy is extensively regulated via post-translational modifications (PTMs), which provide rapid regulatory outputs. These PTMs include protein ubiquitylation, acetylation, lipidation, and phosphorylation, among others (Wani et al., 2015). This Review will concentrate on the recently described PTMs that have been observed on ATG8s.

ATG8 proteins are small proteins with a ubiquitin-like fold preceded by two α-helical structures (Nieto-Torres et al., 2021a). The structural resemblance between ATG8 and ubiquitin, as well as the consecutive conjugation reactions that are required to target them to lipid membranes or proteins, respectively, suggest they have similar evolutionary origins. It has recently been proposed that ATG8 conjugation to lipid membranes could be viewed as functionally similar to protein ubiquitylation, i.e. used to target ‘ATG8-tagged’ cellular membranous vesicles toward degradative pathways such as macroautophagy or, alternatively, serve as a signal for ‘ATG8-tagged’ membranes towards cellular routes such as secretion (Kumar et al., 2021) (Box 1). ATG8s are widely conserved among eukaryotes. However, whereas yeasts and other unicellular eukaryotes contain a single ATG8 homolog, during the evolution of multicellular organisms, two main subfamilies of ATG8 proteins originated, the MAP1LC3 (generally known as just ‘LC3’) and GABARAP proteins (Jatana et al., 2020). These families are represented by a diversity of members depending on the species. For example, the nematode C. elegans expresses only two different ATG8 proteins (LGG-2/LC3 and LGG-1/GABARAP), whereas humans contain seven different proteins including LC3A, LC3B, LC3B2, LC3C, GABARAP, GABARAPL1 and GABARAPL2, and some plants display 17 or more putative different members (Bu et al., 2020).

ATG8 proteins have intrinsic properties that support the autophagy process. The N-terminal α-helical structures of ATG8 proteins display membrane tethering and fusogenic properties that facilitate the expansion of the autophagosome precursor membranes (Nakatogawa et al., 2007; Zhang et al., 2023). Besides these properties, another striking feature of ATG8s is their capacity to scaffold and coordinate a number of interactions with other cellular factors that facilitate the autophagy process (Fig. 1), which was clearly appreciated in pioneering studies (Behrends et al., 2010). Via partially non-redundant protein interactomes, different ATG8 members can accomplish unique functions, which we are starting to appreciate (Johansen and Lamark, 2020). On the inner membrane of the elongating isolation membrane, research efforts have elucidated that among ATG8s, the LC3 subfamily members preferentially interact with most of the so-called autophagy receptor proteins, including p62 (also known as SQSTM1), NBR1, NDP52 (also known as CALCOCO2) and NCOA4, among others (Johansen and Lamark, 2020; Kirkin et al., 2009; Mancias et al., 2014; Pankiv et al., 2007). Autophagy receptors have the capability to bind and recruit cargo targeted for degradation, an event that is often accomplished via ubiquitin-associating (UBA) domains present in the receptors, which bind to ubiquitylated cargo. At the same time, autophagy receptors associate with ATG8s to sequester and deliver the cargo to the lumen of the growing isolation membrane. Although LC3 subfamily homologs most often support the interaction with autophagy receptors, GABARAPs can also participate in specific cases (Johansen and Lamark, 2020). The explanation for this apparent functional specialization between ATG8 subfamilies in the inner autophagosome membrane is not clear yet.

On the outer membrane, ATG8s also establish critical interactions with other proteins involved in autophagosome formation, transport and fusion. ATG8 proteins, and more specifically GABARAPs, help to engage and retain the ULK1 complex machinery and the PI3KC3 complex components via direct interactions, hence sustaining isolation membrane expansion and facilitating autophagosome formation, as determined in mammalian cells (Birgisdottir et al., 2019; Joachim et al., 2015). In addition, the binding of GABARAPs to ATG2 appears to sustain autophagosome closure (Bozic et al., 2020). ATG8s, and especially LC3 subfamily homologs, also mediate the attachment of molecular motors to the outer surface of autophagosomes via adaptor proteins. Specifically, LC3B can recruit the adaptor JIP1 (also known as MAPK8IP1), which connects autophagosomes with dynactin and dynein to facilitate their retrograde transport toward the nuclear periphery and their ultimate autophagosome–lysosome fusion, an essential step for the degradation of cargo (Fu et al., 2014). The proteins LC3A and LC3B can also bind the adaptor protein FYCO1, which recruits kinesin motors that dictate anterograde transport of vesicles, toward the plasma membrane of the cell (Pankiv et al., 2010), a step thought to counteract dynein-mediated transport and premature fusion of autophagosomes with lysosomes, during autophagosome biogenesis and maturation (Nieto-Torres et al., 2021c). Finally, ATG8 can facilitate the recruitment of tethering and fusion-related proteins to promote autophagosome and lysosome fusion, which requires the coordinated action of the SNARE complex, tethers, including the HOPS complex and RAB GTPases. The tethers, recruited by the active form of RAB GTPases, facilitate proximity between the autophagosome and lysosomes, and activate the SNARE complex, which ultimately triggers autophagosome–lysosome fusion (Barz et al., 2021; Wang et al., 2016). The GABARAP subfamily of proteins recruits the scaffolding protein PLEKHM1, which in turn engages the HOPS complex and the GTPases RAB7 and ARL8B (Marwaha et al., 2017; McEwan et al., 2015). In parallel, ATG8s facilitate the recruitment of membrane-attached SNARE proteins, including STX17, which associates with the outside membrane of autophagosomes and facilitates their closure and subsequent fusion with lysosomes (Kumar et al., 2018) (Fig. 1). Overall, ATG8 members display key roles on the autophagosome surface that ultimately lead to autophagosome–lysosome meeting and fusion.

The characterization of the structure of ATG8s, as well as the identification of ATG8-binding partners, allowed the discovery of various binding interfaces and motifs that sustain the functional processes described above. The identification of the first autophagy receptor in mammals, p62 (Pankiv et al., 2007), also illuminated the structural requirements for the ATG8–p62 interaction, which applies to many of the characterized direct interactors of ATG8. p62 and other ATG8-binding proteins display a short stretch of amino acids of variable length (10–20) whose core is defined by the sequence W/F/Y-X-X-L/I/V and constitutes the so-called LC3-interacting region (LIR). Note that this is sometimes named the ATG8-interacting motif (AIM) (Birgisdottir et al., 2013). This sequence of amino acids, which usually displays an extended β-sheet structure, assembles with high affinity into two well-defined hydrophobic pockets in ATG8 proteins that constitute the LIR-docking site (LDS) (Fig. 2). Both the sequence of the LIR and those of the adjacent N- and C-terminal amino acid stretches, which might constitute an extended LIR, modulate protein binding to ATG8s (Sakurai et al., 2017). Usually, acidic residues present immediately preceding the LIR, or phosphorylation of residues in that area, strengthen the interaction with ATG8s via electrostatic interactions with basic residues that surround the ATG8 LDS (Rogov et al., 2023). The specific sequence of the LIR and its flanking amino acids, as well as non-conserved residues in the LDS, might confer the specificity of ATG8 proteins binding to specific interactors (Wirth et al., 2019). For example, FYCO1 preferentially binds to LC3A and LC3B versus other human ATG8s, whereas ULK1 shows a preferential binding toward GABARAP (Alemu et al., 2012; Olsvik et al., 2015). Researchers have developed powerful bioinformatic tools that allow the prediction of LIR within protein sequences (Jacomin et al., 2016). Experimental validation is generally needed to confirm the functionality of these predictions.

Although the LDS–LIR interaction is the most frequently observed between ATG8 and its binding partners, several direct ATG8 interactors lack canonical LIRs. Pioneering studies in Arabidopsis thaliana describe the interaction of the proteasome subunit RPN10 with ATG8, which involves a new conserved interacting surface known as ubiquitin-interacting motif (UIM) for its resemblance to some already known ubiquitin-binding properties (Marshall et al., 2019). This finding reinforces the similarity between ATG8s and ubiquitin. UIMs, however, appear to be specific for ATG8 proteins and be of high affinity. They contain the consensus sequence Ψ-F-Ψ-Ω/T (where Ψ and Ω are small hydrophobic and aromatic residues, respectively). These amino acids fit into a pocket on ATG8 named the UIM-docking site (UDS), and as happens with the LIR, adjacent residues can affect UIM–UDS binding (Marshall et al., 2019). Interestingly, this protein-binding platform is located far from the LDS on the ATG8 structure and therefore might facilitate the simultaneous loading of protein interactors (Fig. 2). Currently, 17 proteins, conserved throughout species, have been implicated in UIM-mediated binding to ATG8. Many of them are involved in vesicle dynamics, such as SNAREs, RAB GTPases and clathrin adaptors (Marshall et al., 2019). Interestingly, UIM-mediated autophagy seems to be crucial to maintaining proteostasis via regulating the levels of the protein segregase CDC48 (also known as P97 or VCP), including its neurodegeneration-related pathogenic variants (Marshall et al., 2019). Although some functional implications of the UIM-based interactions have been tested in plants and yeast, such as the aforementioned degradation of CDC48 (Marshall et al., 2019), the relevance of this novel interacting platform needs to be formally tested in mammalian and human cells. Future studies might also uncover potential additional conserved interactors utilizing the UIM–UDS binding interface.

Other well-characterized protein interactions between ATG8s and their binding partners are mediated by alternative regions. One example is lamin B1 whose interaction with LC3B is sustained by the N-terminal arm of LC3B, specifically residues R10 and R11 (Dou et al., 2015). This process is crucial for the formation of cytoplasmic chromatin fragments and their subsequent autophagic degradation, events enriched in cells undergoing senescence, a cellular state that involves a permanent proliferation arrest with implications for cancer and aging (Dou et al., 2015). Similarly, LC3B R10 and R11 facilitate the recruitment of damaged mitochondria to autophagosomes (Chu et al., 2013). Mitochondria are degraded via a selective type of autophagy named mitophagy, for which an array of different receptors participate (Palikaras et al., 2018). Interestingly, in this case, targeting to mitophagy is facilitated by a protein–lipid interaction between LC3B and the lipid cardiolipin, when the latter becomes exposed to the outer membrane of damaged mitochondria (Chu et al., 2013). The molecular implications of LC3B R10 and R11 allowing both interactions with lamin B1 and cardiolipin need further investigation. Finally, recent studies have demonstrated the relevance of an arginine-rich motif in LC3B that forms an RNA-binding domain (RBD). This domain is required for a specific type of mRNA degradation now called LC3B-mediated mRNA decay (Hwang et al., 2022). The relevance of this newly described RBD and its physiological function needs to be further investigated.

Overall, ATG8s highly concentrate on autophagosome membranes and via alternative binding interfaces support the recruitment of cellular cargoes and other factors to orchestrate the autophagy process.

Accumulating reports are starting to elucidate how the vast array of protein–protein interactions in which ATG8 proteins participate, are regulated. This is overall an understudied topic, but some aspects of it are beginning to be understood, including the temporal regulation of protein–protein interactions. Post-translational modification (PTM) of both ATG8 proteins and/or their interactors is likely important to control their associations and in turn regulate key aspects of the autophagy process (Table 1). PTMs potentially modify diverse properties of ATG8 proteins and/or their interactors. These include changes in protein levels, subcellular localization, conjugation to lipids and accessibility to ATG8-binding interfaces. All these processes might have profound effects on the interactome of ATG8s, and consequently, this could constitute major regulatory mechanisms, as discussed below (see also Figs 2 and 3).

Ubiquitylation and proteosome-mediated degradation seem to also operate directly on ATG8s to control their levels and the extent of autophagy function. The ubiquitin-activating enzyme UBA6 together with the ubiquitin ligase BIRC6 monoubiquitylate LC3B at K51 and trigger its proteosome-mediated degradation in mammalian cells (Jia and Bonifacino, 2019) (Table 1 and Fig. 3). The ubiquitylation process mediated by UBA6 and BIRC6 might also operate on LC3A and LC3C, as these proteins similarly become ubiquitylated even though it is unclear at what amino acid ubiquitylation happens. This regulatory pathway appears to be unique to LC3 members as UBA6- and BIRC6-mediated ubiquitylation does not happen in GABARAP members (Jia and Bonifacino, 2019). Consequently, depletion of UBA6 or BIRC6 increases autophagy flux and degradation of p62 and NBR1 in non-neuronal mammalian cells and neurons (Jia and Bonifacino, 2019).

GABARAP is subjected to ubiquitylation and proteasomal degradation via a different molecular pathway in mammalian cells. During starvation, GABARAP migrates from the pericentriolar material to nascent autophagosomes. This process is regulated by the protein PCM1, which directly interacts with GABARAP via a LIR (Joachim et al., 2017). In the absence of PCM1, the E3 ubiquitin ligase Mib1 binds and ubiquitylates (mono-, di- and poly-ubiquitylation) GABARAP, which leads to its proteasomal degradation (Joachim et al., 2017). Interestingly, Mib1-mediated ubiquitylation of GABARAP likely occurs at K13 and K23, which happen to be poorly conserved amino acids among other ATG8 proteins. This suggests a specific PTM regulatory pathway could be operating on GABARAP (Joachim et al., 2017).

Overall, the impact of these ubiquitylation-mediated regulatory axes on autophagy control warrants further investigation to understand their physiological relevance, and it opens opportunities for therapeutic interventions.

Acetyl-CoA levels are a direct indicator of nutrient cellular status and a strong regulator of autophagy (Mariño et al., 2014). Importantly, the function of autophagy-related proteins can be controlled via acetylation in response to the nutritional status of the cell. The acetylation step is catalyzed by lysine acetyltransferases such as the p300/CBP family proteins, while the deacetylation step is catalyzed by lysine deacetylases (KDACs) such as the sirtuins (Sadoul et al., 2011).

LC3B is directly acetylated for nucleocytoplasmic translocation and activation (LC3B isoform was verified via personal communication, see Acknowledgements section). In multiple human cells including HEK293 and HepG2 and mouse cells such as MEFs and C2C12 cells, LC3B as well as GABARAP and GABARAPL1 can be observed in both the nucleus and the cytoplasm under nutrient-rich conditions. Upon starvation, the pool sizes of these ATG8s change, and they predominantly localize to the cytoplasm, where they can be conjugated to autophagosomes. When starvation ends, LC3B redistributes again to both the nucleus and the cytoplasm. Such trafficking is dependent on acetylation at K49 and K51 on LC3B. Nuclear p300 (also known as EP300) and CBP (also known as CREBBP) acetylate LC3B at these sites and prevent the cytoplasmic redistribution of LC3B (Table 1 and Fig. 3). Because p300 is activated by phosphorylation from mammalian target of rapamycin (mTOR), mTOR inactivation during starvation results in reduced LC3B acetylation from p300 and CBP. Additionally, SIRT1, but not other KDACs with nuclear localization (of which SIRT6, SIRT7, and class I and II KDACs were tested), deacetylates LC3B at these sites and promotes their starvation-induced cytoplasmic localization (Huang et al., 2015; Xu and Wan, 2023) (Table 1 and Fig. 3). Such acetylation-dependent nucleocytoplasmic translocation of LC3B for starvation-induced autophagy seems to be conserved, as a similar regulatory mechanism was observed for nucleocytoplasmic translocation of the Atg8 protein in a moss (Wu et al., 2021). Studies on the downstream signaling pathway in human HEK293 and HeLa cells, as well as mouse NIH3T3 cells, revealed that the DOR (also known as TP53INP2) autophagy regulator protein binds to the deacetylated LC3B and exports it out of the nucleus to the cytoplasm in human cells (Huang et al., 2015; Nowak et al., 2009), where it binds ATG7 for the conjugation to the autophagosomes. Similarly, deacetylation of LC3B has been shown to be required for binding to the autophagy receptor p62 in the cytoplasm of human HEK293 cells (Song et al., 2019); however, it is unclear whether this is also due to cytoplasmic translocation of LC3B or due to another mechanism. Additionally, acetylation might also contribute to stabilizing non-lipidated inactive LC3B by inhibiting mono-ubiquitylation at K51 (Song et al., 2019), which otherwise leads to proteasome-dependent degradation of LC3B (Jia and Bonifacino, 2019). Additional studies showing the direct requirement of K49 and/or K51 acetylation to inhibit ubiquitylation are required to test this crosstalk in human cells. Interestingly, a recent report has described that multiple ATG8 proteins, when localized in the nucleus, act as coactivators of the transcriptional factor LMX1B, which regulates the expression of autophagy genes in dopaminergic neurons (Jiménez-Moreno et al., 2023). Importantly, at least as tested for LC3B in HEK293 cells, the nuclear localization of LC3B due to the K49 and K51 acetylation is required for the LC3B–LMX1B interaction (Jiménez-Moreno et al., 2023). Similar to what is seen in mammalian cells, in Drosophila, the acetylation status of the ATG8 family member Atg8a correlates with nutrient availability and dictates the association of Atg8a with the transcriptional factor Sequoia, which controls the expression of autophagy genes (Jacomin et al., 2020).

Collectively, these findings indicate that acetylation controls the subcellular localization and potentially also stability of the LC3B protein. These results also imply that LC3B deacetylation is a required step for global activation of LC3B-mediated autophagy during starvation. Preliminary experiments suggest that GABARAPs could also undergo regulation via acetylation (Huang et al., 2015). Further analysis to prove this observation and to test whether other LC3s are similarly regulated by acetylation, needs to be performed. Similarly, whether selective autophagy of specific cargoes is reliant on acetylation remains to be elucidated. Finally, the acetylation status of LC3B might have further roles in autophagy via association to transcription factors and subsequent regulation of gene expression of autophagy genes. Given the key regulatory functions controlled by ATG8, defects associated with this pathway have been shown to have profound physiological consequences. For example, impaired autophagy accompanied by deficiency of SIRT1, the deacetylase of LC3B, is also associated with several diseases in the kidney, brain and heart of humans and mice (Donmez and Outeiro, 2013; Kitada et al., 2016; Kong et al., 2015; Nadtochiy et al., 2011; Paraíso et al., 2013; Polak-Jonkisz et al., 2013; Tang, 2017; Zhang et al., 2021). Also, the autophagic-mediated degradation of nuclear progerin, which normally occurs in a manner dependent on SIRT1-mediated deacetylation of LC3B, does not occur in in human fibrotic liver tissue, leading to its abnormal accumulation (Bai et al., 2022). Future studies should elucidate the causal relations between LC3B acetylation and diseases and utilize them for therapeutic applications of the diseases.

Lipidation involves covalent modifications of proteins by the addition of specific lipids, typically targeting these proteins to particular membranes (Xie et al., 2015). During autophagy, lipidation of ATG8 proteins refers to a cascade of ubiquitylation-like enzymatic reactions that attach a C-terminal glycine residue of these proteins to the amino headgroup of the membrane lipid phosphatidylethanolamine (PE) (Hanada et al., 2007; Oh-oka et al., 2008; Wesch et al., 2020) (Table 1 and Fig. 3). The formation of ATG8–PE conjugates is hypothesized to mainly occur at the growing edges of the isolation membrane due to the curvature sensitivity of ATG3, which facilitates ATG8 conjugation to PE (Ichimura et al., 2000; Metlagel et al., 2013). It has been suggested that lipidated ATG8s further interact with membranes through aromatic residues (F77 and F79 in the ubiquitin-like domain of yeast Atg8), and this interaction exposes binding sites on ATG8s, facilitating recruitment of the cargo or interactions with adaptor proteins, as well as contributing to creating curvature of the membrane and ultimately facilitating the formation of a sphere-like structure (Maruyama et al., 2021). However, an alternative model of the association of ATG8 with autophagic lipid membranes has been proposed by another research group (Zhang et al., 2023). This study demonstrated that the N-terminal regions of membrane-conjugated ATG8 are associated with lipid bilayer in cis (Zhang et al., 2023) and are inserted into the membrane instead of being exposed to the cytoplasm. This apparent discrepancy between studies can potentially be explained by a potential influence of the membrane composition and curvature on distinct conformational states of ATG8s. The ATG8-induced membrane curvature is favored by some degree of topological selectivity and specificity, i.e. enrichment of ATG8s in the convex versus concave face of the growing isolation membrane. Even though the exact process and the underlying signaling events of such selectivity are not well understood, it has been demonstrated that the ATG12–ATG5–ATG16L1 complex can be detected exclusively on the convex face of the isolation membrane, creating a mesh-like network of proteins that has been proposed to stabilize the shape of the forming vesicle (Kaufmann et al., 2014; Mizushima et al., 2001).

The final step of autophagosome maturation involves the de-conjugation of ATG8 from the outer membrane. In vitro assays as well as in vivo experiments in yeast have shown that ATG4 proteases cleave the bond between the ATG8 C-terminal glycine and PE leading to ATG8 deconjugation and recycling (Abreu et al., 2017; Kauffman et al., 2018). There are four orthologs of ATG4 protease in mammalian cells, ATG4A–ATG4D. (Fernández and López-Otín, 2015). However, it has been suggested that ATG4D might play a leading role in the delipidation of ATG8 proteins (Betin and Lane, 2009). Further work using the single and quadruple knockouts of ATG4 proteases in mammalian cells lines as well as mice models has demonstrated some redundancy in the function of different orthologs of ATG4 (Kauffman et al., 2018; Tamargo-Gómez et al., 2021).

Overall, lipidation is an important post-translational modification of ATG8 that regulates the conjugation of these proteins to autophagic membranes allowing them to perform a variety of functions, from binding different cargo to interacting with adaptors and transport proteins. Besides this, it has been demonstrated in in vitro work and mammalian cell lines that conjugation of ATG8 proteins can also occur to phosphatidylserine (PS) on single-membrane structures, to sustain alternative processes, termed non-canonical functions of the autophagy machinery (see Box 1 for more details) (Durgan et al., 2021; Lystad and Simonsen, 2019).

ATG8 protein phosphorylation is an important regulatory mechanism that can impact different relevant steps of the autophagy pathway (Table 1 and Fig. 3). The first reports described PKA-mediated phosphorylation of LC3A at S12 as having an inhibitory effect on autophagy, as determined by autophagy flux assays in mammalian cells (Cherra et al., 2010). Although the impact of this PTM on regulating the LC3A interactome is unknown, the authors speculated that it could interfere with the membrane-tethering properties of LC3A (Cherra et al., 2010). Subsequent studies revealed autophagy inhibition correlating with LC3A S12 phosphorylation during so-called activation-induced cell death, a specific form of cell death in T cells mediated by T cell receptor signaling (Corrado et al., 2016), suggesting an important regulatory role for LC3A S12 phosphorylation. Equivalent serine residues appear to be conserved in LC3C and GABARAPL2, but their potential phosphorylation remains unknown.

Pioneering studies have described the physical interaction between LC3 members, specifically the protein LC3B, and the serine/threonine kinases STK3 and STK4 in mammalian cells (Behrends et al., 2010). These kinases function in the Hippo pathway, which controls cell proliferation (Yu and Guan, 2013). STK4 phosphorylates LC3B at T50 (Shrestha et al., 2020; Wilkinson et al., 2015), an event suggested to facilitate autophagosome–lysosome fusion (Wilkinson et al., 2015). Follow-up studies determined that this PTM, which interestingly happens in very close vicinity to the LDS interface (Fig. 2), modulates the binding of several ATG8 interactors and regulates the association of the transport-related protein FYCO1 in human cells (Nieto-Torres et al., 2021b,c; Shrestha et al., 2020). In the absence of LC3B T50 phosphorylation, FYCO1 associates preferentially with LC3B, and autophagosomes display an aberrantly enhanced anterograde transport toward the cell periphery, which might, in turn, compromise autophagosome–lysosome fusion. LC3B T50 phosphorylation decreases the binding of FYCO1 to LC3B. Phosphorylation of LC3B, subsequently, facilitates the retrograde transport of autophagosomes toward the nuclear periphery, via dynein-mediated transport, and the subsequent autophagosome–lysosome fusion (Nieto-Torres et al., 2021c). FYCO1 binding to LC3B might be relevant in the initial steps of autophagy to prevent premature retrograde transport of autophagosomes and fusion with lysosomes. However, it could be detrimental later on once autophagosomes are mature and ready to deliver cargo for degradation to lysosomes (Nieto-Torres et al., 2021c). Therefore, LC3B T50 phosphorylation could be crucial to controlling the timing of these events. What controls the ability of STK4 to phosphorylate LC3B remains to be investigated. Interestingly, STK4 and other homologous kinases can assemble into a macromolecular complex called striatin-interacting phosphatase and kinase (STRIPAK) (Kück et al., 2019; Tang et al., 2019). In flies, ATG8 interacts with Striatin, the scaffolding protein of STRIPAK, and a regulatory subunit of protein phosphatase PP2A. This event is required for the regulation of autophagosome transport (Neisch et al., 2017). Interestingly, recent reports in human cells indicate that striatin is a direct target of ULK1, which activates PP2A and autophagy upon starvation induction (Hu et al., 2021). These findings suggest that the STK4-containing STRIPAK complex might be a node for regulation by multiple autophagy-regulating kinases.

Besides STK4-mediated regulation, strikingly, LC3B T50 can potentially be phosphorylated by other kinases, including STK3, protein kinase C (PKC) and NEK9 (Shrestha et al., 2020; Wilkinson et al., 2015). Accordingly, NEK9-mediated LC3B phosphorylation regulates p62 degradation via autophagy (Shrestha et al., 2020). An LC3B T50 phospho-mimetic mutant, which resembles a permanent phosphorylation state of LC3B, was used in the studies described above. Interestingly, this mutant confers aberrant phenotypes in terms of autophagosome transport and protein–protein interactions, which argues that this PTM might need to be temporally regulated (Nieto-Torres et al., 2021c; Shrestha et al., 2020). PKC has been reported to phosphorylate LC3B at amino acids T6 and T23, yet the impact of these events on autophagy has not been identified (Jiang et al., 2010). Finally, recent studies have revealed the role of ATG8 phosphorylation in controlling autophagosome–lysosome fusion. The kinase TBK1 accumulates at the site of autophagosome formation and is known to play crucial roles in the degradation of depolarized mitochondria and intracellular pathogens via phosphorylation of autophagy receptors (Richter et al., 2016). Besides this, TBK1 directly phosphorylates membrane-bound LC3C (at S93 and S96) and GABARAPL2 (at S87 and S88), stabilizing these proteins in the outer autophagosome membrane (Herhaus et al., 2020). Specifically, these PTMs impede ATG4 binding and premature ATG4-mediated deconjugation of LC3C and GABARAPL2 from the outer membrane of autophagosomes. This ensures the participation of LC3C and GABARAPL2 in the early stages of isolation membrane formation and facilitates a steady autophagy flux, as defined by in vitro experiments and in mammalian cells (Herhaus et al., 2020). Presumably, once this is achieved, either the kinase activity of TBK1 will cease and/or dephosphorylation of LC3C and GABARAPL2 via phosphatases will allow their recycling; these are interesting possibilities for future investigation.

In sum, phosphorylation of ATG8s is a key step to regulate the protein–protein interactions that govern the autophagy process (Table 1 and Fig. 3).

ATG8 proteins are crucial players in autophagy and most of their functions are accomplished through the coordination of multiple protein–protein interactions that initiate and sustain vital stages of autophagy, including autophagosome formation, cargo recruitment, autophagosome transport and autophagosome–lysosome fusion. Although recent evidence indicates that different ATG8 members might specialize in certain stages of the process and/or establish preferential protein–protein interactions, it is not fully understood how individual members are able to multitask and sustain different functions in both a space- and time-regulated manner.

Supported by recent reports, PTMs of ATG8s play a crucial role in controlling autophagy given their transient yet strong regulatory nature (Table 1). Based on this literature, ATG8 PTMs can be synthesized into a potential ‘multilayered post-translational regulatory network’ (Fig. 3), as follows: (1) regulation of levels of ATG8s via ubiquitylation and proteosomal degradation, which might constitute a rate-limiting step for the downstream stages of the autophagy process; (2) regulation of the nucleo-cytoplasmic levels of ATG8s via acetylation, which dictates the availability of ATG8s in the nucleus or cytoplasm to perform autophagy-related functions; (3) vesicle targeting of ATG8s to autophagosomes or alternatively to vesicles of endolysosomal origin via lipidation controls the involvement of these proteins in canonical autophagic functions or in non-canonical degradative or secretory processes (see Box 1); and (4) regulation of the ATG8 protein interactome via PTMs (thus far mostly identified for protein phosphorylation), which either operate directly on ATG8 or on their binding partners. This PTM-mediated regulatory step controls the functions dictated by ATG8s at a given time and location, including autophagosome transport, receptor recruitment and autophagosome–lysosome fusion. This simplified and integrative regulatory model based on the findings described here warrants further investigation. It will be important to test whether specific types of PTMs could apply to all ATG8 members, as well as to analyze the conservation of these PTM-mediated regulatory layers among different species and organisms. Also, it is intriguing to speculate whether different PTMs might be activated via alternative signaling pathways. For example, phosphorylation could be mostly guided via extracellular or local signals subsequently executing regulatory changes at specific subcellular locations, whereas acetylation might be controlled essentially by the metabolic status of the cell.

Future efforts should focus on the unbiased and targeted identification of ATG8 PTMs. Public proteomic databases (e.g. https://www.phosphosite.org) are compiling information on PTM data which many times comes from global proteomic efforts, and have available information on ATG8s. Mechanistic ATG8-specific studies will resolve the function and interactions of these modifications. Besides this, the topological resolution of PTMs, and specifically whether they occur on proteins attached to the inner or outer autophagosome membrane needs further investigation. Along these lines, pioneer proteomic studies have started to identify ATG8-mediated protein interactions that exclusively occur inside ATG8-positive vesicles using proximity ligation assays coupled with protease-mediated degradation (le Guerroué et al., 2017; Zellner et al., 2021). Finally, given the increasing roles found for ATG8s in non-canonical autophagy, it will be desirable to be able to differentiate ATG8s located in autophagosomes versus other types of vesicles to distinguish between canonical and non-canonical ATG8 functions. Additional research on the upstream regulators of both canonical and non-canonical functions of ATG8s is also necessary. Overall, dissecting the regulatory axes operating on ATG8s to favor canonical and non-canonical functions of the autophagy machinery via PTMs will facilitate identifying the interplay between these regulatory pathways in normal physiology. Besides this, these research efforts will help to understand the potential implications of ATG8 PTM regulatory pathways in diseases (see Box 2).

In conclusion, post-translational regulatory mechanisms play a key role in the rapid activation and/or inhibition of autophagy, and it is well known how the early steps of the process are regulated via PTMs of key autophagy regulators. From recent research summarized here, ATG8s are emerging as a PTM-related regulatory hub, owing to the impact PTMs have on controlling multiple crucial functions of ATG8s in autophagy. Similarly, and although not discussed in this Review due to space constraints, PTMs on interactors of ATG8 might also play key roles in controlling autophagy, yet we are only starting to understand the impact of these regulatory pathways (Rogov et al., 2023). Further efforts in this exciting new research direction will grant a better understanding of the molecular regulation of autophagy and its potential underappreciated impact on health and disease.

We thank Dr Sara Landeras-Bueno for preparing Fig. 2. We thank Dr Wei Liu (Zhejiang University School of Medicine, China) for providing information on the acetylatable ATG8 isoform investigated in Huang et al. (2015).