Different effects of Atg2 and Atg18 mutations on Atg8a and Atg9 trafficking during starvation in Drosophila

Highlights • Atg9 and Atg18 are required for autophagy upstream of Atg8a, unlike Atg2.• Atg9 accumulates on Ref(2)P aggregates in Atg8a, Atg7 and Atg2 mutants.• Ultrastructurally, Atg9 vesicles cluster around Ref(2)P aggregates in stalled PAS.• Atg9 does not accumulate on Ref(2)P upon loss of Atg18 or Vps34, while FIP200 does.• Atg18 simultaneously interacts with both Atg9 and Ref(2)P.


Introduction
Autophagy is a major catabolic pathway capable of degrading all kinds of intracellular material including proteins, lipids, polysaccharides, and nucleic acids. During its main pathway, phagophores capture cytosol and organelles to form autophagosomes, followed by the fusion of these double-membrane vesicles with lysosomes [1]. Autophagy was initially considered to be a non-specific, bulk degradation system, in contrast with the ubiquitin-proteasome pathway, in which individual polyubiquitinated proteins are recognized, unfolded and degraded in the inner proteolytic chamber of proteasomes. More recently, multiple studies showed that ubiquitination also signals for selective autophagic degradation, and characterization of ubiquitin-specific autophagy receptors revealed the molecular mechanism involved [2,3]. Proteins such as p62 contain distinct domains mediating multimerization, ubiquitin binding and interaction with Atg8 family proteins. Atg8 is a ubiquitin-like protein bound to phagophores and autophagosomes through a lipid anchor [1,4,5]. Ubiquitinated proteins are captured into aggregates by binding to p62 multimers, and the interaction of p62 with Atg8-positive phagophores is considered to be responsible for their elimination by autophagy [2,3,[6][7][8]. In contrast with this simple model, p62 was found to colocalize with proteins involved in the initiation of phagophores independent of the presence of mammalian Atg8 homologs such as LC3 [9]. These results suggest that additional factors also contribute to the recognition of p62 aggregates by phagophores.
Atg9 is the only transmembrane protein of core autophagy factors, and it likely supplies initial vesicles for phagophore nucleation from multiple membrane sources including endosomes, plasma membrane and Golgi [10][11][12]. Atg9 is considered to be an upstream factor in the hierarchy of autophagy-related (Atg) proteins in yeast, but the molecular determinants of Atg9 recruitment to the phagophore assembly site (PAS) are incompletely characterized [13]. A recent study shows that mammalian Atg9 is recruited to damaged mitochondria independent of lipidated Atg8 homologs and the upstream kinase Atg1/ULK1 during selective autophagic degradation of mitochondria [14]. In yeast, the Atg2-Atg18 protein complex is thought to act in parallel to the Atg8 system, and regulates Atg9 recycling from PAS [13,15]. Diverse Atg18-like proteins are found in eukaryotes. The four mammalian homologs fall into two groups based on bioinformatic analysis: WIPI1/2 and WIPI3/4 [16,17]. Of these, both WIPI2 and WIPI4 were suggested to promote autophagosome formation based on siRNA experiments in cultured cells [16,18]. Yeast Atg18 and its paralog Atg21 belong to the WIPI1/2 group [16,17]. Atg21 only functions in an autophagy-related biosynthetic pathway called cytoplasm to vacuole targeting (cvt) which delivers a subset of hydrolases to the vacuole under growth conditions, whereas Atg18 is required for both autophagy and cvt in yeast [19]. Ygr223c, the third Atg18 family protein in yeast, belongs to the WIPI3/4 group and regulates micronucleophagy, a selective autophagy pathway for degradation of nuclear components [16,20].
Clear orthologs of most Atg proteins are found in the popular metazoan model Drosophila, and the p62 ortholog refractory to Sigma P (Ref(2)P) also promotes ubiquitinated protein aggregation [21,22]. Single genes code for Atg9 and Atg2, and the WIPI1/2 homolog CG7986 is referred to as Atg18, as it is required for autophagy in Drosophila [23][24][25][26]. The roles of another WIPI1/2 family Drosophila protein, CG8678, and of the WIPI3/4-like CG11975 are unknown [16,17].
Here we show that loss of Atg2 or Atg18 have different consequences on Atg8a puncta formation and Atg9 recruitment to Ref(2)P aggregates during starvation in Drosophila, raising the possibility that these proteins act differently during autophagosome formation in Drosophila.

Results
p62 aggregates are selectively captured into autophagosomes under both basal or starvation conditions in mammalian cells, as more than 90% of LC3-positive autophagosomes are also positive for p62 [6]. In line with that, most Atg8a-positive autophagosomes colocalized with Ref(2)P in fat bodies of well-fed, starved or wandering Drosophila larvae, respectively (Figs. 1A and S1). As expected, the number of Atg8a dots increased in response to autophagy induction (Fig. 1B). Strikingly, larger Ref(2)P aggregates seen under low autophagy level in fed cells disappeared during starvation-induced or developmental autophagy (Fig. 1C). Thus, Ref(2)P aggregates are selectively eliminated by autophagy in Drosophila.
Atg2 and Atg18 act in parallel to Atg8 in yeast, whereas Atg2 appears to function downstream of Atg8 family proteins in worms and mammals, respectively [18,30]. We also found that punctate Atg8a structures form in fat bodies of starved Atg2 mutants previously shown to be defective in autophagic degradation ( Fig. 2A, B and E) [23,26]. In contrast, Atg8a dots were rarely detected in starved Atg18 mutants, and they were restored by expression of mCherry-Atg18 (Fig. 2C-E). Similarly, RNAi depletion of Atg2 failed to block the formation of Atg8a-positive structures, while Atg18 knockdown inhibited punctate Atg8a in GFP-marked RNAi cells (Fig. 2F-H). Immunoprecipitation experiments suggested that Atg2 may interact with Drosophila Atg18, and more efficiently with its paralog CG8678 (Fig. S2).
We have recently shown that Drosophila Atg9 is required for basal, Myc-induced or proteasome inactivation-induced autophagy, respectively [24][25][26]. Atg9 is also necessary for the starvation response, as expression of either one of three independent RNAi lines in GFP-marked cell clones prevented punctate LysoTracker staining, a commonly used marker of autolysosomes in the fat body ( Fig. S3A-C and J). Moreover, formation of mCherry-Atg8a positive autophagosomes and autolysosomes were also blocked in knockdown cells, and the selective cargo Ref(2)P accumulated upon Atg9 RNAi during starvation ( Fig. S3D-I, K and L).
We raised polyclonal antibodies against Drosophila Atg9, which specifically recognized Atg9 dots and clusters in starved fat body cells (Fig. S3M). Phagophores assemble at aggregates of the selective cargo pro-aminopeptidase I in yeast, and p62-positive aggregates may play similar roles in metazoans [3]. In wild type cells, the overlap of Atg9 with Ref(2)P was low ( Fig. 3A and G), in line with the transient localization of Atg9 to forming phagophores described in mammalian cells [10,31]. Loss of downstream Atg genes results in the accumulation of upstream factors in stalled PAS, both in yeast and mammals [13,32]. We therefore carried out similar analyses for Atg9 and Ref(2)P. Nearly all Ref(2)P aggregates colocalized with Atg9 in null mutants of Atg7 encoding the E1-like enzyme required for Atg8a lipidation and membrane association, and in Atg8a nulls (Fig. 3B, C and G).
Aggregates of Ref(2)P also overlapped with Atg9 in Atg2 mutant fat cells in starved fly larvae ( Fig. 3D and G), consistent with the role of Atg2 downstream of (or parallel to) Atg8a. Strikingly, colocalization of these two proteins was low in Atg18 mutants ( Fig. 3E and G). Atg18 is a phosphatidylinositol 3-phosphate (PI3P) effector [33], which prompted us to test whether Vps34, the lipid kinase producing this phospholipid, is involved in Atg9 recruitment to Ref(2)P. Colocalization of Atg9 with Ref(2)P was again low in fat bodies expressing dominant-negative Vps34 (Fig. 3F and G). These data suggested that either Atg9 or Ref(2)P is not recruited to the PAS in the absence of Atg18 or Vps34 function. To distinguish between these possibilities, we looked at the localization of FIP200/ Atg17. FIP200 is a subunit of the Atg1 kinase complex, and it accumulates on Ref(2)P aggregates in fat body cells of starved Atg7 null mutant larvae [34]. FIP200 was also enriched on Ref(2)P aggregates in cells lacking Atg2, Atg18 or Vps34 function (Fig. 3H-J), suggesting that Atg9 recruitment to PAS may specifically be affected by the absence of Atg18 or Vps34. Moreover, overlapping mCherry-Atg18 and Ref(2)P structures were readily observed in Atg2 and Atg8a RNAi cells (Figs. 3K and S3N).
These data suggested that Ref (2)P aggregates with associated upstream Atg proteins may represent stalled PAS in Atg mutants. The ultrastructure of protein inclusions is well-characterized in Atg mutant neurons of adult flies and mice, respectively [28,35,36]. These aggregates contain ubiquitinated protein, and are often surrounded by small vesicles (Fig. 3L). Double immunolabeling experiments revealed that at least a subset of vesicles associated with Ref(2)P-positive inclusions are positive for endogenous Atg9 in neurons of Atg7 and Atg8a null mutant adult brains ( Fig. 3M and N).
The lack of Atg9 accumulation on Ref(2)P aggregates in Atg18 mutants raised the possibility that Atg18 may facilitate Atg9 recruitment to Ref(2)P. Indeed, HA-Atg18 coprecipitated with FLAG-Ref(2)P in cultured cells (Fig. 4A). HA-Atg9 showed strong binding to FLAG-Atg18, but not to FLAG-Ref(2)P (Fig. 4B), suggesting that the colocalization of Atg9 with Ref(2)P may not map a direct physical interaction. We hypothesized that Atg18 may potentially facilitate Atg9 recruitment to Ref(2)P aggregates by simultaneously binding to both proteins. In line with this model, coexpression of HA-Atg18 resulted in coprecipitation of HA-Atg9 with FLAG-Ref(2)P (Fig. 4B).

Discussion
Previous studies established the importance of the interaction between p62 and Atg8 family proteins (such as LC3) during selective autophagy in mammalian cells [2,3,6]. A recent live imagingbased report shows that these two proteins are recruited together to the PAS, but later than the Atg1/ULK kinase complex, the Vps34 lipid kinase complex and its effectors such as WIPI1 [31]. How p62 and LC3 are recruited to the PAS is still not understood completely. Several interactions have been recently described between Atg proteins belonging to distinct functional groups and complexes, which likely cooperately coordinate the recruitment of downstream factors. For instance, LC3-interacting regions have been characterized in ULK1/Atg1 and its binding partners Atg13 and FIP200, which may facilitate LC3 recruitment to PAS [37].
Atg9 likely supplies lipids in the form of small vesicles to initiate phagophore formation, and most of it is normally recycled from phagophores [10,31]. Atg18 is a phospholipid effector, and its complex with Atg2 regulates Atg9 recycling from phagophores in yeast [15]. Atg18 and Atg2 are considered to function together, genetically in parallel to Atg8 in yeast [13]. A recent paper shows that Atg18 recruitment to PAS requires PI3P and Atg2, and the Atg18-related protein Atg21 also appears to facilitate its localization to pro-aminopeptidase I aggregates in yeast [38]. Loss of Atg2 leads to formation of stalled phagophores in worms and mammalian cells, and it does not block the recruitment of worm Atg18 to protein aggregates [18,30], which are entirely consistent with our findings in Drosophila. Loss of Atg18 suppressed Atg8a puncta formation unlike the lack of Atg2 function, indicating that Atg18 acts upstream of Atg8a in Drosophila. Similar to our findings, RNAi depletion of the Atg18 homolog WIPI2 was found to prevent LC3 puncta formation in cultured mammalian cells, and mutation of Atg18 attenuates efficient Atg8 puncta formation in the 50 cell stage worm embryo [16,18]. Interestingly, Atg18 and Atg21 were also suggested to promote the recruitment of lipidated Atg8 to the PAS in yeast, based on experiments carried out in a multiple knockout strain which contains only a subset of Atg genes [19].
Our data raise the possibility that the selective autophagy cargo Ref(2)P may be used as a potential PAS marker in Atg mutant Drosophila cells, as most Ref (2)P aggregates become positive for upstream Atg proteins. The Atg1 complex subunit FIP200 is enriched on Ref(2)P aggregates in the absence of Atg7, Atg2, Atg18 or Vps34 function. Atg9 also accumulates on Ref(2)P aggregates in Atg7, Atg8a and Atg2 mutants, but not in Atg18 mutants. Atg18 family proteins contain a WD40 domain with 7 beta-propellers. This structure likely enables the binding of Atg18 to multiple targets simultaneously, including the autophagy proteins Atg2 and Atg9 that we also show here, and PI3P on autophagic membranes [15,19,33,38]. In line with the interaction of Atg18 with this phospholipid, accumulation of Atg9 on Ref(2)P aggregates also requires the lipid kinase Vps34. Our data suggest that Atg18 may play an important role in phagophore nucleation, potentially by facilitating the recruitment of Atg9-containing vesicles to Ref(2)P for selective capture and degradation of ubiquitinated protein aggregates through autophagy in Drosophila.