AIM/LIR-based fluorescent sensors—new tools to monitor mAtg8 functions

ABSTRACT

Macroautophagy/autophagy, a catabolic process by which cytoplasmic materials are degraded and recycled in lysosomes/vacuoles, remains a rapidly expanding research topic with the need for constantly improved methodologies to study each step of this pathway. Recently Lee and colleagues, as well as Stolz et al., independently reported the development of new AIM/LIR-based fluorescent sensors, which mark individual endogenous mammalian Atg8-family (mAtg8) proteins without affecting the autophagic flux. When expressed in cells, each sensor selectively recognizes individual mAtg8 isoforms and distinguishes mammalian MAP1LC3/LC3 proteins from the related GABARAPs. Such selectivity was achieved by using various LC3-interacting regions with high binding affinity to either a subgroup, or a specific, mAtg8 isoform as part of the sensor. Here we discuss the utility of these sensors in autophagy research and highlight their strengths, weaknesses and future directions.

KEYWORDS:

• Atg8
• Atg8-interacting motif
• autophagy assay
• fluorescent probes
• GABARAP
• LC3
• LC3-interacting region
• sensor

Abbreviations

AIM	=	Atg8-interacting motif
Atg8	=	Autophagy-related 8
GABARAP	=	GABA type A receptor-associated protein
GFP	=	green fluorescent protein
GIM	=	GABARAP interaction motif
HyD	=	hydrophobic domain
LIR	=	LC3-interacting region
MAP1LC3/LC3	=	microtubule associated protein 1 light chain 3
RFP	=	red fluorescent protein

Ever since the discovery of autophagy, considerable effort has been devoted to developing a wide range of methods to monitor, analyze and understand this fascinating process. All the autophagy assays, thoroughly described in a previous publication [1], have greatly contributed to broadening our knowledge about the autophagy process and its complexity, while also helping us understand the mechanisms of its induction and regulation. Among these methods, Atg8 (autophagy-related 8) is the most widely used protein marker employed to monitor autophagosome formation and localization, as well as to measure autophagic flux [1]. The advantage of Atg8 proteins lies in the fact that the core autophagy mechanism related to autophagosome formation and degradation is common for the majority of selective and nonselective autophagy pathways described to date. MAP1LC3A/LC3A (microtubule associated protein 1 light chain 3 alpha), LC3B, LC3B2, LC3C, GABARAP (GABA type A receptor-associated protein) and GABARAPL1/2 (GABA type A receptor-associated protein like 1/2) are related to the Atg8 protein, described initially in yeast. Members of the Atg8 protein family (Atg8/LC3/GABARAP) are central autophagy components involved in autophagosome formation in most model organisms [2–4].

During autophagy, Atg8/LC3/GABARAP proteins are cleaved by a cysteine protease belonging to the Atg4/ATG4 family [5,6] and subsequently lipidated [2,5] leading to the association of Atg8-family proteins initially with the phagophore and later with mature autophagosomes [7,8]. Because lipidated Atg8-family proteins mostly and specifically decorate autophagic structures, fluorescently-labeled Atg8-family proteins are routinely used to study autophagosome dynamics. First, labeled Atg8-family proteins are mainly used to analyze autophagosome formation, which is achieved by counting the number of GFP-Atg8/LC3/GABARAP puncta [1] or by analysis of protein levels by western blots, after normalization to the levels of a housekeeping protein unaffected by autophagy. Second, the subcellular distribution of autophagosomes, and their further degradation in vacuoles or autolysosomes, can be monitored in real time by microscopy observation when fluorescently-labeled Atg8-family proteins are used. Finally, fluorescently-labeled Atg8-family proteins are routinely used to measure autophagic flux, for instance by western blot analysis of the levels of unlipidated and lipidated forms of Atg8-family proteins after treatment with lysosomal/vacuolar inhibitors [1], by use of a tandem RFP-GFP-Atg8 fluorescence microscopy assay [7,9–11] or by a recently developed, fluorescent probe, GFP-LC3-RFP-LC3ΔG, which does not need to be combined with lysosomal inhibitors [12]. However, none of these methods circumvents concerns associated with the possibility of artifacts caused either by overexpression of the reporter mammalian Atg8-family (mAtg8) proteins, or the inability to distinguish and monitor multiple, endogenous mAtg8 family members. It is known that different mAtg8s have varying expression levels in different cell types [13–16]. Also, diverse functions for the LC3 and GAPARAP subfamilies are known. For example, during the early and late stages of autophagosome formation, LC3 proteins are mainly important for recruitment of autophagic receptors, while GABARAPs facilitate transport and membrane fusion [17–24].

Although mAtg8s are among the most crucial components of the autophagy machinery, little is known about overlapping and specific functions of each isoform and none of the available methods described above provide the proper tool to analyze their individual features. Despite the observation that mAtg8 proteins are highly structurally conserved, individual isoforms diverge around the Atg8-interacting motif (AIM)/LC3-interacting region (LIR)-binding site. The AIM/LIR-binding site contains 2, highly evolutionarily conserved, hydrophobic pockets (the HP1 pocket, also named the W-site, and the HP2 pocket, alternatively called the L-site) formed by the ubiquitin-like fold which, because aromatic and hydrophobic residues from an AIM/LIR motif dock into these cavities, are absolutely crucial for AIM/LIR binding (for more information see review [25]). Structural studies of Atg8/LC3/GABARAPs have revealed that individual mAtg8s diverge from each other in the amino acid composition and electrical charge of the surface located close to the HP1 site, which is responsible for the ionic interaction between the AIM/LIR motif and the AIM/LIR-binding site. This diversity is reflected in selectivity towards specific binding partners due to varying affinities of the different types of AIM/LIR motifs found in interacting proteins [13]. Therefore, an important part of understanding autophagy regulation by the different mAtg8 isoforms, as well as the non-autophagic role(s) of each mAtg8 protein, have eluded a thorough study. To meet researchers’ needs, novel AIM/LIR-based fluorescent sensors were recently designed independently by Lee and colleagues [26] and by Stolz et al. [27] to monitor mAtg8 puncta (Figure 1).

Figure 1. The AIM/LIR-based fluorescent sensors bind selectively to mAtg8 proteins in an AIM/LIR-dependent manner. Summary of AIM/LIR-based fluorescent sensors constructed by Stolz et al. [27] and Lee et al. [26] is presented in the attached table. Columns present each LC3/GABARAP protein and rows indicate the AIM/LIR-based fluorescent sensor used for detection. Description: efficient sensor binding “+”; most efficient binding and preferential colocalization of a pan-specific sensor “+*”; binding detected in affinity isolation assay “(+)”; improved sensor containing PB1 domain (mCh-PB1-AS3-67) is available “[+]”.

The foundation of both approaches was to engineer peptides based on AIM/LIR sequences that detect endogenous mAtg8 proteins. The AIM/LIR sequence is a short linear motif that was first identified in the human SQSTM1/p62 protein and in yeast Atg19 [28,29]. Subsequently, X-ray crystallography and NMR studies elucidated the structure of the SQSTM1 LIR bound to LC3B and the Atg19 AIM bound to Atg8, showing conservation of the mechanism of interaction between AIM/LIR motifs and Atg8/LC3/GABARAP proteins across eukaryotes [29,30]. To date, many AIM/LIR motifs have been identified in various Atg8-interacting partner proteins, including cargo receptors or specific proteins selectively degraded by autophagy, components of the core autophagic machinery, and proteins associated with vesicles and their transport. Most of them are categorized as canonical AIM/LIR motifs that fit the following consensus listed in the iLIR database (http://repeat.biol.ucy.ac.cy/iLIR/) and are referred to as the xLIR motif: [A/D/E/F/G/L/P/R/S/K)(D/E/G/M/S/T/V-W/F/Y-D/E/I/L/Q/T/V-A/D/E/F/H/I/K/L/M/P/S/T/V-I/L/V], where the amino acids marked in bold correspond to the residues that are most essential for the interaction with Atg8-family proteins as they dock into the HP1 and HP2 pockets present on the surface of Atg8/LC3/GABARAP proteins [31,32]. Not surprisingly, this significant diversity in amino acid composition within the AIM/LIR motif affects the selection of specific interactors by different mAtg8 isoforms in mammals. For instance, there is growing evidence that LC3- or GABARAP-subfamily members recognize diverse autophagy adaptors and receptors [13]. Rogov and colleagues even defined a GABARAP interaction motif (GIM), which can be presented as a sequence ([W/F-V/I]-X2-V) that is distinct from the LIR sequence [33].

To develop AIM/LIR-based fluorescent sensors, Stolz et al. [27]. searched the library of peptides that possess bona fide LIR properties and are selective for individual mAtg8 isoforms, whereas Lee and colleagues [26] created sensors by testing 34 known LIR motifs and identified 2 where one specifically detects LC3A/B-positive puncta and another that detects all mAtg8-positive puncta, but preferentially associates with GABARAP/GABARAPL1-positive structures (Figure 1). In both cases, the stability and sensitivity of the sensors had to be optimized by using multiple copies of the LIR motif itself, or by the addition of another motif/domain already proven to increase the protein's association with the phagophore membrane or with membranes in general.

Membrane-associated proteins targeted to specific intracellular organelles often possess multivalent domains and/or motifs such as a dimerization/oligomerization domain, a hydrophobic motif, or a specific lipid/protein-binding motif for their proper subcellular localization. Such multivalency (interactions in which 2 or more molecular recognition events occur simultaneously) provides a strategy for increasing protein avidity to specific cellular components by accumulation of several affinities from domains and motifs existing in proteins [34]. For instance, Atg19 contains multiple Atg8-binding sites that enable it to bend the phagophore membrane [29,35,36]. The multidomain autophagic receptor SQSTM1 contains, among other domains and motifs, a LIR motif mediating the interaction with mAtg8s and an N-terminal PB1 domain for oligomerization, which stabilizes its binding to LC3B-coated surfaces [37]. Also NBR1, an autophagic receptor with similar domain architecture to SQSTM1, that mediates pexophagy (selective degradation of peroxisomes through autophagy), contains within its sequence a PB1 domain for heterodimerization, LIR motifs, a coiled-coil domain for homo-oligomerization and the phospholipid-binding, amphipathic α-helical J domain with membrane binding capacity [38,39]. Similarly, an LIR motif in the FYCO1 protein is not sufficient for FYCO1 targeting to the autophagosome, and dimerization of the phosphatidylinositol-3-phosphate-binding FYVE domain via the coiled-coil region in FYCO1 is important for its association with the membrane [40,41]. RavZ, a Legionella anti-autophagy effector, is another great example of the cooperative action of multiple domains, because it requires both LIR motifs and a phosphatidylinositol-3-phosphate-binding membrane-association domain to target to autophagosomes [42,43]. Therefore, Stolz et al. [27] and Lee and colleagues [26] improved the avidity of fluorescence-based LC3/GABARAP-specific sensors to the LC3/GABARAP protein by using multiple LIR motifs, altering charge distribution or by introducing the oligomerization (PB1) domain of SQSTM1, as well as by adding a short hydrophobic domain (HyD), respectively.

Taking the concepts applied in each of the sensor types into consideration, it is impossible not to wonder if using a flexible HyD at the N terminus of the LIR motif (or any other membrane-specific domain), to enhance phagophore or autophagosomal membrane targeting, might be a better strategy than using a multi-interacting domain such as PB1 for sensor construction, because the PB1 domain is present in at least 13 human proteins functioning as interaction modules in various pathways through PB1-mediated hetero-dimerization or homo-oligomerization [44]. Because peptides engineered by Stolz et al., recognize almost each individual mAtg8 isoform in contrast to those proposed by Lee et al. which can only distinguish diverse LC3 proteins from GABARAPs (Figure 1) [26,27], it will be interesting in the future to create a second generation of sensors with narrowed specificity to each individual mAtg8 isoform and test if the HyD domain, in combination with the peptides engineered by Stolz et al. [27], is sufficient for sensor recruitment to mAtg8-positive structures.

The enormous advantage of new sensors lies in the fact that before the creation of AIM/LIR-based fluorescent sensors, monitoring individual mAtg8s was strongly limited to the expression of the respective isoform as a fusion with a fluorophore. Such a strategy ignored the endogenous population of mAtg8 isoforms and raised several concerns. For example, it was previously reported that overexpression of LC3 may cause abnormal neurite branching [26] or affect its proper subcellular localization causing its aggregation or association with ubiquitin-positive aggregates in autophagy-deficient cells [45]. The usage of specific AIM/LIR-based fluorescent sensors for individual mAtg8s opens up the exciting opportunity to finally monitor, in a time-dependent manner, populations of endogenous mAtg8s in vivo and study their distinct function in cells without causing cellular abnormalities. Lee et al. [26] have already proven that their new sensor does not affect the morphology of cultured cortical neurons.

However, it is also necessary to point out some limitations of these sensors. First of all, these sensors were developed specifically for mammalian LC3 and GABARAP subfamilies. Therefore, it is unlikely that they will be suitable for other Atg8-family proteins (from other species) and a new set of sensors is most likely required for plant and yeast Atg8-family proteins. Second, with these sensors it seems difficult to distinguish populations of LC3A from LC3B as well as GABARAP from GABARAPL1 and GABARAPL2. It is plausible that these proteins have weak binding specificity. For instance, GABARAP has many diverse interactors with no single dominating LIR motif. Although screening a phage-displayed peptide library against GABARAP revealed some preferences at certain positions within the LIR motif sequence, such as tryptophan (W) at sequence position W₀ corresponding to the position of W/F/Y in the xLIR motif, and aliphatic amino acids at positions X₁ and L₃ (corresponding to the position of I/L/V in the xLIR motif), an aromatic residue at position X₂ and a proline at position X₄ or X₅ [46], such primacy may be not enough to distinguish GABARAP AIM/LIR-binding preferences from those of GABARAPL1 and GABARAPL2 [47]. Furthermore, both types of sensors do not seem to influence autophagy and autophagic flux, which means that endogenous receptors would outcompete sensors if they are not highly expressed, but it also means that these sensors might not be able to reveal events where only low amounts of mAtg8s are involved, or where tight binding to a receptor/effector is required. Therefore, in the future, it might be important not only to further optimize these sensors but also to develop ones that are not AIM/LIR-dependent and/or could be used in all species. Moreover, the human LC3 family has, besides LC3A, LC3B and LC3C, a fourth member LC3B2, as well as 2 variants of the LC3A protein. [4] Therefore it could be interesting to test sensors' specificity toward these proteins, too.

Developing other types of sensors or further studying and optimizing AIM/LIR-based sensors is of importance for several reasons. It was noted that AIM/LIR motifs can be recognized by proteins other than Atg8/LC3/GABARAP, such as the association of Atg5 with Atg19 that occurs in an AIM-dependent manner [48]. It is true that some AIM/LIR motifs are exquisitely selective, while others are rather promiscuous. The LIR motif present in CALCOCO2/NDP52 is highly selective toward the LC3C protein [49], whereas the LIR motif from SQSTM1 can bind all LC3/GABARAPs with the same efficiency [10]. Yet, such specificity seems not to correlate to their predisposition towards binding other proteins because both CALCOCO2 and SQSTM1 efficiently co-precipitate with ATG5 [48]. In addition, Habisov et al. [50] reported recently that UBA5 by its short linear motif at the C terminus, termed LIR/UFIM, can bind, besides GABARAPL2, UFM1—a distant ubiquitin-like protein that is involved in the ufmylation process. These examples prove that AIM/LIR motifs are not limited to interact strictly and only with Atg8/LC3/GABARAP proteins. Last, but not least, it is important to highlight that post-translational modifications on Atg8/LC3/GABARAP protein surfaces may highly affect recruitment of AIM/LIR-based sensors. For instance, 2 of the 3 recently mapped phosphorylation sites in the LC3 proteins (Thr6 and Ser12 recognized by PRKC and PRKA, respectively) lie directly in, or in the vicinity of the LIR-binding site [51,52]. Although the mechanism of autophagy inhibition through phosphorylation of LC3 proteins is still unclear, the possibility that phosphorylation of these sites interferes with the interaction of LC3 proteins with LIR-containing proteins is highly attractive and therefore may have serious consequences in the recruitment of AIM/LIR-dependent sensors.

In conclusion, the newly developed sensors reported here are very promising and valuable new tools that will help our understanding of the autophagy process and in the elucidation of the functions of individual mAtg8 members. Nevertheless, these advances should be viewed as part of an ongoing progression in the necessity of new tools required to overcome limitations that still persist in the development of sensors that recognize Atg8-family proteins.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Additional information

Funding

This work was supported by the National Institutes of Health [grant number 5RO1 DK41737].