Human GABARAP can restore autophagosome biogenesis in a C. elegans lgg-1 mutant

Abstract

We recently described in C. elegans embryos, the acquisition of specialized functions for orthologs of yeast Atg8 (e.g., mammalian MAP1LC3/LC3) in allophagy, a selective and developmentally regulated autophagic process. During the formation of double-membrane autophagosomes, the ubiquitin-like Atg8/LC3 proteins are recruited to the membrane through a lipidation process. While at least 6 orthologs and paralogs are present in mammals, C. elegans only possesses 2 orthologs, LGG-1 and LGG-2, corresponding to the GABARAP-GABARAPL2/GATE-16 and the MAP1LC3 families, respectively. During allophagy, LGG-1 acts upstream of LGG-2 and is essential for autophagosome biogenesis, whereas LGG-2 facilitates their maturation. We demonstrated that LGG-2 directly interacts with the HOPS complex subunit VPS-39, and mediates the tethering between autophagosomes and lysosomes, which also requires RAB-7. In the present addendum, we compared the localization of autophagosomes, endosomes, amphisomes, and lysosomes in vps-39, rab-7, and lgg-2 depleted embryos. Our results suggest that lysosomes interact with autophagosomes or endosomes through a similar mechanism. We also performed a functional complementation of an lgg-1 null mutant with human GABARAP, its closer homolog, and showed that it localizes to autophagosomes and can rescue LGG-1 functions in the early embryo.

The double-membrane autophagosome is the key organelle of macroautophagy (thereafter called autophagy). Yeast genetic screens have identified Atg8, a ubiquitin-like protein associated with the membrane of the phagophore and autophagosome, as essential for phagophore initiation and autophagosome biogenesis.¹ Atg8 is highly conserved and presents numerous orthologs and paralogs in mammals, which have been classified as MAP1LC3/LC3 (microtubule-associated protein 1 light chain 3) and GABARAP (GABA[A] receptor-associated protein) subfamilies.² LC3 and GABARAP proteins act in diverse intracellular trafficking and autophagy processes where they have multiple roles from membrane elongation and fusion, to recognition and addressing of cargoes in the autophagosome.²

Atg8 and its mammalian LC3 protein orthologs have the particularity to interact with numerous partners and a LC3-interacting region has now been described in more than 20 proteins.³ Numerous data have already demonstrated that LC3 proteins can interact differentially with adaptive proteins involved in selective autophagy, in particular during mitophagy and aggrephagy. On the other hand, the mechanisms of interactions between lysosomes and autophagosomes have not been extensively studied and almost no data have been reported concerning the roles for LC3 proteins in the degradation of autophagosomes. However, several studies in metazoans have shown that autophagosomal maturation shares some characteristics with endosomal maturation. First, autophagosomes can fuse with multiple endosomal compartments and generate an intermediate vesicle called amphisome.⁴ Second, part of the machinery involved in endosomal maturation appears to be important for autophagosome degradation (e.g., RAB7, ESCRT, FYCO1, among others).

The presence of multiple mammalian Atg8 orthologs and paralogs raises different hypotheses about their functions, from high redundancy to very selective functions, but adds a level of complexity in their individual analysis. In this regard, a proteomic analysis of 6 human Atg8 orthologs and paralogs reveals that two-thirds of their interactome are specific for either the LC3 or the GABARAP subfamilies.⁵ In a recent paper,⁶ we have characterized the C. elegans LGG-1 and LGG-2, which are the unique orthologs for the GABARAP-GABARAPL2 and LC3 subfamilies, respectively. Electron microscopy and immunofluorescence studies revealed that both LGG-1 and LGG-2 localize to autophagosomal membranes and define 3 different populations of autophagosomes during embryonic development.

We have further characterized the respective functions of LGG-1 and LGG-2 in allophagy, a stereotyped and selective autophagic process for degrading paternal mitochondria and organelles in the early embryo.^7,8 During allophagy, autophagosomes positive for both LGG-1 and LGG-2 form a cluster, which was easy to analyze during the whole autophagic process. Genetic analyses combined with time-lapse imaging revealed that LGG-1 and LGG-2 act sequentially during this autophagic flux and that LGG-1 function is essential for the localization of LGG-2 to the autophagosomal membrane. We discovered that LGG-2 is implicated in the acidification and maturation of these autophagosomes by facilitating their interaction with the lysosomal compartment.

A Common Tethering Machinery for Autolysosome and Endolysosome Formation?

Our recent work identified a direct interaction between LGG-2 and the “clathrin and Vps domain” of VPS-39, a subunit of the HOPS complex. Together with the small GTPase RAB7, the HOPS complex has been well described for its role in the tethering between endosomes and lysosomes, the preliminary step before fusion of membrane by SNARE proteins.⁹ However, HOPS implication in autolysosome formation has not been extensively studied. Our in vivo analysis demonstrated that LGG-2 facilitates the interaction between autophagosomes and lysosomes, and involved RAB-7 and the 2 HOPS specific subunits VPS-39 and VPS-41. We proposed that these proteins participate in the tethering of autophagosomes and lysosomes to allow the formation of autolysosomes. Our results are also supported by 2 recent publications demonstrating in human cells and Drosophila that the HOPS complex is important for autolysosome formation.^10,11 However, it is still unclear whether the lysosome uses a unique tethering machinery for fusing with autophagosomes or endosomes. Noticeably, the presence of several homologs of HOPS units in mammals provides additional possible combinations to form tethering complexes⁹ and the existence of a specific HOPS complex for autolysosomes has been hypothesized.¹² As a first step to address this question, we compared the localization of autophagosomes, endosomes, and amphisomes in wild-type, lgg-2, rab-7, and vps-39 early embryos (Fig. 1).

Figure 1. Localization of endosomes, amphisomes, and lysosomes in lgg-2, rab-7, and vps-39 depleted embryos. (A–D) Single confocal planes of 4-cell stage embryos in control (A), lgg-2(tm5755) (B), rab-7(RNAi) (C) and vps-39(ok2442) (D) showing nuclei (DNA in blue) and endosomes (HGRS-1/VPS-27 in white). Arrowheads indicate the perinuclear enrichment of endosomes in vps-39(ok2442) embryos. (A’–D’) Same confocal planes than (A–D) showing autophagosomes (LGG-1 in red), endosomes (HGRS-1/VPS-27 in green) and amphisomes (in yellow). Insets are 5-fold magnifications of amphisomes indicated by white arrows. Embryos were prepared as described previously⁶ and incubated with the following primary antibodies: rat anti-LGG-1 (gift from Zhang lab, Beijing, China)¹³ at 1:500 and rabbit anti-HGRS-1/VPS-27¹⁴ at 1:250. As secondary conjugated antibodies, Alexa Fluor^® 488 (Molecular Probes, A11034) and Alexa Fluor^® 568 (Molecular Probes, A11077) at 1/500 were used and DNA was labeled using TO-PRO^®3-iodide 10 µM (Molecular Probes, T3605). (E and F) Quantification of endosomes (E), realized on 3 nonconsecutive planes for each embryo, and amphisomes in whole embryo (F) in control, lgg-2(tm5755), rab-7(RNAi) and vps-39(ok2442) embryos from 2- to 4-cell stages (numbers of planes or embryos are indicated above the graphs; Student t test, *P < 0.05 **P < 0.005 ***P < 0.0005). (G–J) Epifluorescence images of late endosomes/lysosomes labeled by LysoTracker Red in control (G), lgg-2(tm5755) (H), rab-7(RNAi) (I) and vps-39(ok2442) (J) 4-cell stage embryos. Pictures shown are maximum projections of 1 µm thickness. Adult animals were incubated 4 h in 150 µL of saline buffer containing 10 µM of LysoTracker Red (Molecular Probes, L7528) and the progeny analyzed. (K) Quantification of the percentage of embryos with perinuclear accumulation of lysosomes in control, lgg-2(tm5755), rab-7(RNAi) and vps-39(ok2442) embryos from 2- to 4-cell stages (numbers of embryos are indicated above the graphs; Chi-square test, *P < 0.05.) Scale bar: 10 µm. Error bars are ± SDM. (L) Model of interaction of autophagosomes, endosomes and lysosomes in the early embryo.

In wild-type embryos, fertilization triggers a rapid formation of autophagosomes that cluster around the male pronucleus^6-8 and progressively disappears between the 2- and the 4-cell stages. The cluster also contains amphisomes that were almost undetectable after the 4-cell stage (Fig. 1A′). At this stage, endosomes were uniformly distributed in the embryos (Fig. 1A). In the lgg-2 mutant, autophagosomes and amphisomes accumulated and persisted in a cluster, indicating that LGG-2 was dispensable to form either autophagosomes or amphisomes, but was involved in their degradation (Fig. 1B′ and F). In these embryos, we observed a slight increase in the number of endosomes, but their localization remained unaffected (Fig. 1B and E). Interestingly, a recent paper identifies the recycling endosomes as a possible source for biogenesis of autophagosomes.¹⁵ The presence of endosomes in the allophagic cluster raises the possibility that they are involved in the formation of the autophagosomes rather than in their degradation.

We then analyzed the role of 2 essential components of the tethering machinery, RAB-7 and VPS-39, in the formation and the degradation of autophagic and endosomal vesicles. In rab-7 embryos, the number of autophagosomes and amphisomes was increased but they were less clustered than in lgg-2 embryos (Fig. 1C′ and F). Endosomes appeared larger, and more numerous but still localized uniformly (Fig. 1C and E). These results confirm the role of RAB-7 in the final degradation of these 3 types of vesicles¹⁶ and show that it is not essential for the formation of autophagosomes and amphisomes. In vps-39 mutants, endosomes, autophagosomes, and amphisomes accumulated close to the nuclei (Fig. 1D and D′). If the number of autophagosomes is increased,⁶ the number of amphisomes and endosomes is not statistically different, however, these latter appear to be enlarged (Fig. 1D–F). This indicates that in absence of VPS-39, autophagosomes, amphisomes, and endosomes are not properly degraded but are able to migrate toward the nucleus. Next, we analyzed the localization of the late endosomal-lysosomal compartment using LysoTracker Red in wild-type, lgg-2, rab-7, and vps-39 depleted embryos. We observed a partial enrichment of the LysoTracker-positive vesicles around the nuclei of wild-type embryos (Fig. 1G and K).⁶ This localization was unaffected in lgg-2 and rab-7 embryos (Fig. 1H, I, and K), but in the vps-39 mutant, there was a strong perinuclear accumulation of acidified vesicles (Fig. 1J and K). Our new results indicate that amphisome formation requires neither LGG-2, RAB-7 nor VPS-39 and that the localization of endosomes, amphisomes, and autophagosomes is similarly affected after the depletion of the lysosomal tethering machinery. Together, our data suggest that, in C. elegans, lysosomes use common tethering factors to interact with endosomes, autophagosomes, and amphisomes.

Our results are summarized in Figure 1L that presents a model for the degradation of autophagosomes and endosomes in the embryo. Our previous epistatic analysis has revealed that RAB-7 and VPS-39 could act together downstream of LGG-2 to facilitate the degradation of the autophagosome cluster.⁶ We report that lysosomes are detected in close proximity to the cluster of autophagosomes concomitantly to its degradation. Moreover, the use of a sensor of the acidification of autophagosomes shows that acidified autophagosomes are exiting the cluster. Finally, our data suggest that the interaction between autophagosomes and lysosomes is correlated with a centripetal migration, which requires RAB-7 function. This latter has been involved in the movement of vesicles toward microtubule minus-ends and microtubules are key players in both autophagosome¹⁷ and endosome dynamics. Thus, we propose that the tethering of autophagosomes and endosomes to lysosomes allows their degradation but also participates in their movement along microtubules. Studies on mammalian cells have shown that the lysosomes change their intracellular positioning in response to nutrient availability and, upon starvation, cluster near the MTOC, facilitating autolysosome formation.¹⁸ In regard of our data, it would be interesting to analyze whether HOPS complex is involved in this process.

Can Human GABARAP Complement its Nematode Ortholog LGG-1?

The different roles for LGG-1 and LGG-2 during autophagosome biogenesis raise the question whether these functions are conserved for their human orthologs. To highlight structure-function specificities between LGG-1 and LGG-2, we generated independent 3D structural models for both proteins (Fig. 2A). The best models for LGG-1 and LGG-2 were obtained with the bovine GABARAPL2/GATE-16 and the human MAP1LC3B/LC3B structures, respectively. Despite the rather low identity (32%) between their primary sequences, LGG-1 and LGG-2 models superimpose very well, both for the 2 N-terminal α-helices and the ubiquitin core (Fig. 2A), except for 2 differences: the N-terminal part of LGG-2 is much longer than LGG-1 and the third β-strand is not present in LGG-1 model (Fig. 2A′). We then asked whether the closer human orthologs of LGG-1 and LGG-2 could be functional when expressed in C. elegans. Because we have shown previously that GFP-LGG-1 is a functional protein,⁶ we generated transgenic animals expressing GFP-GABARAP and GFP-LC3A in the early embryo. In wild-type embryos, GFP-GABARAP formed a cluster of puncta in close proximity to paternal mitochondria (Fig. 2C–E and J) and presented a very similar dynamic compared with GFP-LGG-1 autophagosomes (Fig. 2B). This suggests that GABARAP is efficiently cleaved and conjugated to phagophore membranes by the nematode lipidation machinery. Conversely, the expression of human LC3A under the same conditions did not result in a punctate pattern, suggesting either its inability to be lipidated in C. elegans or a functional difference with GABARAP (Fig. 2K and L).

Figure 2. Human GABARAP complements LGG-1 function in “allophagy.” (A and A’) Modelization of LGG-1 and LGG-2 structures. The 3D structures of LGG-1 and LGG-2 were predicted independently by using the I-TASSER server http://zhanglab.ccmb.med.umich.edu/I-TASSER/. The 7 C-terminal amino acids of the sequence of LGG-1 used to compute the 3D structure, were deleted (LGG-1 II). The best alignment was with the GABARAPL2/GATE-16 protein of Bos Taurus (1EOC in RCSB-PDB database) for LGG-1 and with LC3B protein of Homo sapiens (2ZJD) for LGG-2. The scores for LGG-1 and LGG-2 are 0.983 (TM-score) 0.54 (RMSD: root mean squared deviation), and 0.915 (TM-score) 0.39 (RMSD), respectively. Visualization and superposition of 3D structures were done with UCSF Chimera. (A’) Rotated and magnified view of the dotted-square region. (B–I) Epifluorescence images of GFP-LGG-1 and GFP-GABARAP in control (B–E) and in lgg-1(tm3489) (F–I) embryos. Insets are 2-fold magnification of GFP-GABARAP (green) and the paternal mitochondria labeled by MitoTracker Red (Molecular Probes, M7512). The staining of paternal mitochondria was performed as described previously.⁶ (J) Quantification of autophagosomes colocalizing with mitochondria in control and in lgg-1(tm3489) embryos (n ≥ 5, Student t test, *P < 0.05). (K and L) Epifluorescence images of GFP-LC3A in control embryo of 1-cell and 4-cell stages. To generate the P_pie-1-gfp-gabarap and P_pie-1-gfp-lc3A fusion constructs, the entire coding frame for human gabarap and lc3A were integrated by Gateway cloning in pID3–01B destination vector (gift from G. Seydoux) containing the pie-1 promoter fragment and the gfp coding sequence. Transgenic strains were obtained by microparticle bombardment.⁶ (M and N) Confocal images of GFP-GABARAP (green in M) or GFP-LGG-1 (green in N) and LGG-2 (red) in lgg-1(tm3489) 2-cell stage embryos. Green and red arrows indicate a GABARAP- or LGG-1-only and a LGG-2-only autophagosomes, respectively. Embryos were prepared as described previously⁶ and incubated with the following primary antibodies: mouse monoclonal anti-GFP (Roche, 1814460) at 1:250 and rabbit anti-LGG-2⁷ at 1:250. As secondary conjugated antibodies, Alexa Fluor^® 488 (Molecular Probes, A11029) and Alexa Fluor^® 568 (Molecular Probes, A11036) at 1/500 were used and DNA was labeled using Hoechst (Serva, 15091) at 1/500. (O–R) Epifluorescence images of GFP-LGG-1 and GFP-GABARAP in rab-7(RNAi) (O and Q) or in rab-7(RNAi); lgg-1(tm3489) (P and R) in 4-cell stage embryos. (S) Quantification of autophagosomes was performed in 2- to 4-cell stage embryos (n ≥ 10, Student t test, *P < 0.05, ***P < 0.0005). All images are maximum projections of the whole embryo, except single plane insets and confocal images (7 µm thickness). Scale bar: 10 µm. Error bars are ± SDM.

To test whether GFP-GABARAP could be functional, we then analyzed its localization in lgg-1(tm3489) null mutant embryos (strain RD192), in which cluster formation is drastically impaired,⁶ and compared with GFP-LGG-1 pattern. In the lgg-1 mutant, GFP-GABARAP was still present in puncta which colocalized with paternal mitochondria (Fig. 2F–J). In addition, colocalization analysis with LGG-2 revealed the presence of 3 populations of autophagosomes, GFP-GABARAP only, LGG-2 only and double-positive (Fig. 2M and N). Finally, GFP-GABARAP-positive puncta accumulated in rab-7(RNAi) embryos (Fig. 2O–S) supporting the conclusion that they are autophagosomes. In these various contexts, the pattern of GFP-GABARAP is very similar to GFP-LGG-1. Altogether, our data demonstrate that human GABARAP can complement the function of LGG-1 for autophagosome biogenesis during allophagy. However, in the early lgg-1 embryos, GFP-GABARAP autophagosomes were less clustered than in control embryos (compare Figure 2C–G). One possibility is that the dynamic of formation of autophagosomes by GABARAP could be slower than with endogenous LGG-1. Previous studies from the Elazar group have shown that the GABARAP-GABARAPL2 family is involved in autophagosome maturation and closure whereas the LC3 family is involved in elongation of the phagophore membrane.¹⁹ Our data, using C. elegans embryo as a heterologous system to analyze human orthologs, support the notion that GABARAP has also an intrinsic capacity to promote autophagosome formation.

Manil-Ségalen M, Lefebvre C, Jenzer C, Trichet M, Boulogne C, Satiat-Jeunemaitre B, Legouis R. The C. elegans LC3 acts downstream of GABARAP to degrade autophagosomes by interacting with the HOPS subunit VPS39. Dev Cell. 2014;28:43–55. doi: 10.1016/j.devcel.2013.11.022.