Atg9 is required for intraluminal vesicles in amphisomes and autolysosomes

ABSTRACT Autophagy is an intracellular recycling and degradation process, which is important for energy metabolism, lipid metabolism, physiological stress response and organism development. During Drosophila development, autophagy is up-regulated in fat body and midgut cells, to control metabolic function and to enable tissue remodelling. Atg9 is the only transmembrane protein involved in the core autophagy machinery and is thought to have a role in autophagosome formation. During Drosophila development, Atg9 co-located with Atg8 autophagosomes, Rab11 endosomes and Lamp1 endosomes-lysosomes. RNAi silencing of Atg9 reduced both the number and the size of autophagosomes during development and caused morphological changes to amphisomes/autolysosomes. In control cells there was compartmentalised acidification corresponding to intraluminal Rab11/Lamp-1 vesicles, but in Atg9 depleted cells there were no intraluminal vesicles and the acidification was not compartmentalised. We concluded that Atg9 is required to form intraluminal vesicles and for localised acidification within amphisomes/autolysosomes, and consequently when depleted, reduced the capacity to degrade and remodel gut tissue during development.


INTRODUCTION
Macroautophagy or autophagy is an intracellular process that is highly conserved from yeast through to mammals, and involves the encapsulation of cytoplasmic components into a double membrane structure called an autophagosome (Ohsumi, 2014). Autophagosomes can then undergo a sequential maturation process by interacting with endosomes to form amphisomes and then lysosomes to generate autolysosomes; which eventually results in the degradation of engulfed cytoplasmic material (Gordon and Seglen, 1988;Berg et al., 1998;Lamb et al., 2013a;Shen and Mizushima, 2014). This catabolic process is an important mechanism for the bulk degradation of cytoplasmic constituents, the clearance of protein aggregates, the recycling of aged or defective organelles and for combatting intracellular pathogens (Mizushima et al., 2008). Autophagosomes also have critical functional roles in cellular homeostasis, being specifically involved in energy/nutrient sensing, glucose/glycogen metabolism, and lipid transport, storage and metabolism (Singh et al., 2009;Singh and Cuervo, 2011). Autophagy operates at a basal level in most cell types, but can be specifically induced in response to hormonal and developmental signalling, nutrient restriction, aberrant protein folding, altered homeostasis/physiological stress, and various pathological conditions including neurodegenerative disease, infection, aging and cancer (Kundu and Thompson, 2008;Deretic, 2009;Kroemer et al., 2010;Liu and Ryan, 2012). While the core molecular machinery involved in autophagy has been identified (Tsukada and Ohsumi, 1993;Thumm et al., 1994;Harding et al., 1995), the mechanisms orchestrating this critical intracellular process and the precise role for each component of the molecular machinery are yet to be fully defined.
Over 30 autophagy related genes (Atg) have been discovered, mainly from studies in yeast (Tsukada and Ohsumi, 1993;Nakatogawa et al., 2009). Upon the induction of autophagy, by for example starvation, most of these Atg proteins are localised to a perivacuolar structure called the pre-autophagosome structure (PAS; Suzuki et al., 2001Suzuki et al., , 2007. The PAS is a precursor for autophagosome formation and involves the recruitment of the kinase complexes Atg1/ULK1 and Atg6/Beclin1 to a structure that generates an autophagosome-specific pool of phosphatidylinositiol-3-phosphate (Rubinsztein et al., 2012). This lipid pool drives the nucleation of the phagophore and recruitment of other autophagyrelated proteins to the isolation membrane (Lamb et al., 2013b;Wirth et al., 2013). Membrane expansion and closure of the phagophore involves the Atg5-Atg12 complex, which acts as an E-3 ligase to mediate the lipidation of Atg8/LC3, and the latter remains associated with the outer autophagosome membrane during maturation (Klionsky and Schulman, 2014). The multiplespanning integral membrane protein Atg9 is an early target of the Atg1/ULK1 kinase complex and Atg9 phosphorylation is required for the efficient recruitment of Atg8 and Atg18 to the site of autophagosome formation ; making it an essential component of the autophagic molecular machinery.
The integral membrane properties of Atg9 and its detection in different membrane compartments, including small vesicles that reside in close proximity to the Golgi, mitochondria and the PAS, have led to the suggestion that Atg9 is involved in membrane delivery to the expanding phagophore (Mari et al., 2010;Lamb et al., 2013b). In yeast, for example, Atg9 can be detected on small 30-60 nm vesicles at the PAS  and may interact with the Atg17 scaffolding complex assembled by Atg1/ ULK1 to facilitate Atg9 vesicle fusion (Sekito et al., 2009;Ragusa et al., 2012). This vesicular fusion may provide a membrane platform on which the isolation membrane can then be formed Shibutani and Yoshimori, 2014). In mammalian cells, Atg9 also localises with Atg1/ULK1 and Atg16L at recycling endosomes (Longatti et al., 2012;Puri et al., 2013) and is trafficked to the recycling endosome from the plasma membrane, possibly through the early endosome (Ravikumar et al., 2010;Longatti et al., 2012;Puri et al., 2013). While Atg9 might have a role in the formation of membrane platforms, it is sequestered to but not integrated into autophagosomes (Orsi et al., 2012), raising concerns over the ideas on membrane recruitment and suggesting that it has an alternate functional role.
Atg9 co-locates with endosomes following the induction of autophagy (Young et al., 2006;Tang et al., 2011), adding confusion about the precise role of Atg9 and leading to speculation about its involvement in autophagosome maturation. Amphisome formation, involving the heterotypic fusion of autophagosomes and endosomes, is known to utilise a range of vesicular machinery including, Rab11 (Savina et al., 2005;Fader et al., 2008), the SNARE protein VAMP3  and the ESCRT proteins Vps 28, Vps25, Vps32 and Deep Orange (Lindmo et al., 2006;Rusten et al., 2007;Vaccari et al., 2009;Spowart and Lum, 2010). However, the exact involvement of this vesicular machinery and the regulatory mechanism is not clear. For example, while Rab11 is a marker for recycling endosomes it is also detected on multivesicular endosomes, and is involved in amphisome formation (Fader et al., 2008;Richards et al., 2011). In addition, Rab11 has a role in Atg9 trafficking from the plasma membrane to autophagic compartments (Ravikumar et al., 2010;Longatti et al., 2012;Puri et al., 2013). This trafficking of Atg9 by Rab11 compartments appears to be vital for autophagosome initiation, but has not been fully investigated in relation to amphisome formation (Ravikumar et al., 2010;Longatti et al., 2012;Puri et al., 2013). Other autophagy initiators also have functions in autophagosome maturation; for example, the Atg12-Atg5 initiation complex ( part of the Atg12-Atg5-Atg16 complex), which is involved in autophagosome maturation via an interaction with the tethering protein TECRP1 during autolysosome formation (Chen et al., 2012). Lysosomal fusion with the maturing autophagosome to form autolysosomes, also involves specific vesicular machinery, including Rab7 and the SNARE protein Syntaxin 17, which binds to SNAP-29 and VAMP7 to form a ternary complex Itakura et al., 2012;Takats et al., 2013). As the only integral membrane protein in the core autophagy machinery, Atg9 may play a role in endosome and lysosome recruitment, acting to facilitate vesicular fusion in manner similar to that proposed for its role in membrane recruitment to the phagophore (Takahashi et al., 2011).
Drosophila provides an ideal model system to investigate the role of Atg9 in autophagy; as in the fly, autophagy is induced in response to physiological stresses, such as nutrient restriction (Mulakkal et al., 2014), and Atg9 RNAi silencing can reduce this autophagic response (Pircs et al., 2012;Low et al., 2013;Nagy et al., 2013Nagy et al., , 2014. Autophagy is also up-regulated during Drosophila metamorphosis from larvae to adult-hood (Butterworth et al., 1988;Rusten et al., 2004;Lindmo et al., 2006;Denton et al., 2009Denton et al., , 2013 and autophagosomes increase in abundance in the fat body tissue as the larvae approach puparation (Rusten et al., 2004;Lindmo et al., 2006); enabling the investigation of autophagy under natural conditions without an exogenous stimulus. Here we have used the large size of Drosophila fat body cells and organelles, and the capacity for genetic manipulation in the fly, to further investigate the role of Atg9 in autophagy. In this model we observed intraluminal vesicles in Atg8-GFP amphisomes/ autolysosomes, which co-located with the endosome marker Rab11 and lysosome marker Lamp1. Upon Atg9 depletion these intraluminal vesicles were no longer detected, suggesting that Atg9 has a specific role in intraluminal vesicle formation in autophagic compartments.

RESULTS
Atg9 depletion reduced the number and size of autophagosomes at a time point in Drosophila development when autophagy is normally up-regulated Drosophila Atg9 has previously been investigated in the autophagic response to starvation and hypoxia (Pircs et al., 2012;Low et al., 2013;Tang et al., 2013), but its involvement in developmental autophagy has yet to be defined. Here we investigated Atg9 in relation to either Atg8 (another autophagy marker), Rab11 (an endosomal marker) or Lamp1 (an endosomal-lysosomal marker), in fat body tissue at puparium formation (0 h PF), when autophagy is known to be up-regulated (Rusten et al., 2004;Lindmo et al., 2006). There was an increased amount of the Atg9 protein, detected by western blotting, in wild-type fat body tissue at 0 h PF when compared to −4 h PF (supplementary material Fig. S1A). At 0 h PF, Atg9 co-located with Atg8a-GFP in fat body tissue, but not all Atg8a-GFP compartments were positive for Atg9 ( Fig. 1A-A II ). At this time point Atg9 was also detected in association with large Rab11-GFP compartments that in most cases contained intraluminal Rab11-GFP positive vesicles ( Fig. 1B-B II ). Small Rab11 positive vesicles were also observed in close proximity to larger Rab11-GPF compartments and some of these compartments contained Atg9 ( Fig. 1B-B II ). Atg9 was detected in association with Lamp1-GFP compartments that contained intraluminal Lamp1-GFP positive vesicles ( Fig. 1C-C II ). Atg9 was mainly detected as discrete punctate staining when associated with Atg8, Rab11 and Lamp1 compartments (Fig. 1).

Atg9 is required for Rab11 intraluminal vesicles and compartmentalised acidification in amphisomes during developmental autophagy
In control Drosophila fat body cells at 0 h PF, Atg8a-mCherry colocated to 75±8% of large Rab11-GFP compartments (>5 μm 2 ; Fig  Vps20 by RNAi silencing also resulted in a loss of Rab11-GFP intraluminal vesicles, but the LysoTracker ® was not sequested into compartments and had a cytoplasmic distribution (Fig. 3I-I II ). As LysoTracker ® is a acidotrophic dye this may suggest that the endosomal compartments are not appropriately acidified to cause accumalation of the dye in late endosomes or lysosomes.
Atg9 depletion reduced the size of autolysosomes and altered the compartmentalisation of LysoTracker ® and Lamp1 intraluminal vesicles in fat body cells during developmental autophagy In control Drosophila fat body cells at 0 h PF, Atg8a-mCherry colocated with Lamp1-GFP compartments and many of these Lamp1-GFP/Atg8a-mCherry positive compartments contained intraluminal Lamp1-GFP vesicles (Fig. 4A-A II ). Co-location of Atg8a-mCherry with Lamp1-GFP was also detected in Atg9 depleted Drosophila fat body cells, but there was a more uniform distribution of Lamp1-GFP in these compartments ( The Drosophila larval midgut is degraded by autophagy during metamorphosis (Lee et al., 2002;Denton et al., 2009). In Atg9 depleted Drosophila midgut cells at 0 h PF, the Rab11-GFP and LysoTracker ® compartments were small and the morphology was altered when compared to control midgut cells (Fig. 5A-B II ). This difference in compartment morphology was particularly evident for LysoTracker ® , with control midgut cells having larger compartments that contained intraluminal vesicles (Fig. 5A I ,A II ) and Atg9 depleted midgut cells having smaller compartments with little evidence of compartmentalisation (Fig. 5B I ,B II ). TEM analysis of Drosophila midgut cells at 0 h PF supported this altered compartment morphology, with control midgut cells having larger vesicles with multiple intraluminal vesicles (Fig. 5C,D), whereas the compartments in Atg9 depleted midgut cells had either a uniform granular appearance or only limited evidence of intraluminal vesicles (Fig. 5E,F). Consequently, there were significantly more multivesicular structures (P<0.03) identified in control midgut cells (12.2±1.9) than Atg9 RNAi Line2 midgut cells (6.9±1.2). This appeared to affect autophagic degradation as at +4 h PF the control Drosophila pupae had shorter midguts with an average perimeter of 2273±57 µm and gastric caeca that were almost completely degraded (Fig. 5G), whereas the midguts from Atg9 RNAi Line2 pupae were significantly longer than controls (P<0.05) with an average perimeter of 6363±368 µm and gastric caeca that remained intact (Fig. 5H).

DISCUSSION
Autophagy is a cellular degradation and recycling process that is important during energy metabolism, lipid metabolism, physiological stress, tissue remodelling and organism development (Singh and Cuervo, 2011). Changes in autophagy progression have been shown to be important in numerous human pathologies including neurological disorders, cancer and infectious diseases (Choi et al., 2013). Defining the mechanism and regulation of autophagy is essential to fully understand the pathophysiology of these diseases and this may lead to the design of new targeted therapies. Atg9 is the only transmembrane protein known to be involved in autophagy and it is thought to function in early autophagosome formation, however, it localises to multiple cellular locations (Young et al., 2006;Orsi et al., 2012;Puri et al., 2013) and is not thought to be integrated into autophagosome membranes (Orsi et al., 2012). We have demonstrated that Atg9 has a function relating to intraluminal vesicle formation and is required for compartmentalised acidification within amphisomes and autolysosomes.
In yeast, Atg9 is required for autophagy and its depletion prevents the formation of autophagosomes that are induced by rapamycin treatment or through nitrogen/amino acid starvation (Suzuki et al., 2001). In mammals, the depletion of Atg9 also reduces autophagic activity, limiting LC3-I to LC3-II lipidation, reducing the number of early autophagic structures or LC3 puncta formed, and decreasing the turnover of long lived proteins (Yamada et al., 2005;Young et al., 2006;Saitoh et al., 2009;Orsi et al., 2012). Similarly, in Drosophila, the depletion of Atg9 by RNAi silencing reduced the number of autophagic structures detected in fat body cells following starvation of third instar larvae and also reduced protein clearance (Low et al., 2013;Tang et al., 2013;Nagy et al., 2014). The reduction in both the number and the size of autophagosomes during development, which were observed upon RNAi silencing of Atg9 in the fat body, would be compatible with previous reports for roles of Atg9 in membrane recruitment to the phagophore and for membrane expansion (Suzuki et al., 2001;Yamada et al., 2005;Young et al., 2006;Saitoh et al., 2009;Orsi et al., 2012;Lamb et al., 2013b;Low et al., 2013;Tang et al., 2013;Nagy et al., 2014). Although there was reduced formation of autophagosomes in Atg9 RNAi silenced fat body tissues during development, autophagic compartments were still formed allowing us to track autophagosome maturation in Atg9 depleted tissue.
Autophagosomes interact with endosomes and lysosomes in a maturation process that sequentially generates amphisomes and then autolysosomes. This autophagosome maturation process is essential for the delivery of endolytic and exolytic acid hydrolases, as well as the generation of the acidic conditions, which enable the degradation of the luminal content of autophagosomes (Gordon and Seglen, 1988;Dunn, 1990;Fengsrud et al., 1995). In mammalian cells, Atg9 has been observed to be trafficked via the endosomal network and under normal nutrient conditions is often associated with recycling endosomes (Longatti et al., 2012;Orsi et al., 2012;Puri et al., 2013). The induction of autophagy by starvation decreases the association of Atg9 with recycling endosomes (Longatti et al., 2012;Orsi et al., 2012), but increases Atg9 localisation with late endosomes (Young et al., 2006;Tang et al., 2011). We also observed Atg9 localised to compartments that were positive for the endosome and lysosome markers Rab11 and Lamp1 in Drosophila fat body, at a developmental time point when autophagy is induced by hormone signalling. These observations have led to the hypothesis that Atg9 might have a role in autophagosome maturation (Takahashi et al., 2011;Reggiori and Tooze, 2012). We observed, however, Atg8 autophagosome acidification and co-location of Rab7, Rab11 and Lamp1 with these compartments following Atg9 depletion, indicating that amphisome and autolysosome formation was not abrogated. This implied that the recruitment of endosomes and lysosomes, and the vesicular fusion machinery that mediates the fusion of these degradative organelles to form respectively amphisomes and autolysosomes, was not impaired by Atg9 depletion. Despite the capacity to form amphisomes and autolysosomes, there appeared to be a disruption to developmental autophagy and the reduced turnover of midgut tissue, suggesting that the degradative properties of these compartments might still be impaired.
Impaired autophagic degradation can result from abrogated fusion with endosome and lysosome compartments, causing either the accumulation of autophagosomes, (Eskelinen, 2006;Lee et al., 2007; or the enlargement of autolysosomes (Settembre et al., 2008a,b). Atg9 depletion resulted in smaller amphisomes/autolysosomes and no apparent accumulation of autophagosomes, further indicating that amphisomes and autolysosomes are forming in these tissues. In Atg9 depleted tissues there were, however, significant morphological changes to amphisomes-autolysosomes, with the loss of intraluminal vesicles, which were detectable by Rab7, Rab11, Lamp1 or LysoTracker ® and a loss of intraluminal vesicles was also noted by TEM of Drosophila midgut tissue. In late endosomes, intraluminal vesicles are formed to allow the internalisation of ubiquitinated proteins, membrane rafts, other specific lipid domains and membrane receptors/cargo (Babst, 2011). Amphisomes result from the fusion of autophagosomes with these multivesicular endosomes (Tamai et al., 2007;Fader et al., 2008) and Rab11 multivesicular structures increase in size and co-localise with autophagosomal markers following the induction of autophagy (Fader et al., 2008). Atg9 localises with a number of endosomal markers, which are associated with late endosomes and multivesicular endosomes, including Rab11 and Rab7 (Denzer et al., 2000;Satoh et al., 2005;Savina et al., 2005;Fader et al., 2008;Vanlandingham and Ceresa, 2009); and clusters of Atg9 positive tubules/vesicles have been shown to emanate from these multivesicular endosomes (Orsi et al., 2012). This could suggest that Atg9 functions in intraluminal vesicle formation in endosomes, prior to amphisome formation (Fig. 6A). The ECSRT machinery that is required for intraluminal vesicle formation in endosome maturation is also required for autophagosome maturation; as RNAi depletion or mutations in the ECSRT machinery can cause the accumulation of autophagosomes (Filimonenko et al., 2007;Lee et al., 2007;Rusten et al., 2007;Manil-Segalen et al., 2012). Here, RNAi depletion of the ESCRT-III subunit Vps20, prevented the generation of acidified endosomes and Rab11 amphisomes and autolysosomes, instead generating Rab11 compartments similar to class E-compartments (Doyotte et al., 2005). Paradoxically, while Atg9 RNAi affected intraluminal vesicle formation it did not impair the formation of amphisomes and autolysosomes, suggesting that Atg9 does not affect endosome maturation to the same extent as the depletion of the ESCRT machinery.
The intraluminal vesicles observed in amphisomes and autolysosomes, were positive for Lamp1 and Atg8, proteins thought to associate respectively with the membranes of lysosomes and autophagosomes; in addition to Rab11 which may also be present on multivesicular endosomes. Though Lamp1 protein has been detected on both endosomes and lysosomes it is enriched in lysosomal compartments and consequently preferentially associated with autolysosomes rather than amphisomes (Schroder et al., 2010). In Drosophila wing discs, Lamp1 has also been detected on the intraluminal vesicles in multivesicular compartments  and in larval fat body Lamp1 intraluminal vesicles have been observed within Lamp1/Atg8 positive compartments following starvation (Kim et al., 2012). These intraluminal vesicles observed in amphisomes-autolysosomes could be formed by the fusion of autophagosomes with multivesicular endosomes (Bright et al., 2005;Yu et al., 2010;Chen and Yu, 2013), but from our data we could not exclude direct formation of intraluminal vesicles by amphisomes or autolysosomes (Fig. 6). Intraluminal vesicle formation in the autophagosome may facilitate cargo sorting, in a manner similar to late endosomes. This cargo sorting may be necessary for autophagosomes, as not all proteins targeted by autophagy are destined for lysosomal degradation, and autophagy can also act as a trafficking pathway for proteins such as, interleukins and lipoproteins (Pan et al., 2008;Deretic et al., 2012;Ouimet, 2013). Intraluminal vesicles may also be important in establishing local pH gradients, as variations in intraluminal acidity were detected by LysoTracker ® following Atg9 depletion.
It is speculated that the localised acidification observed in amphisomes-autolysosomes may also be used to facilitate the degradation of specific cargo incorporated into autophagosomes. The hydrolytic enzymes delivered by endosomes and lysosomes to autophagosomes require an acidic environment to function effectively (Mindell, 2012) and increased lysosome pH is known to decrease protein degradation in autolysosomes (Mousavi et al., 2001;Zhang et al., 2013;Hosogi et al., 2014). The low pH environment of endosomes, lysosomes and autolysosomes is controlled by vesicular H + ATPase complexes, which act as a proton pump to facilitate acidification (Forgac, 2007). We observed compartmentalisation of LysoTracker ® in control, but not Atg9 depleted cells which may explain the reduced degradative capacity and inability to degrade midgut tissue when Atg9 was depleted in this tissue. Inability to distribute the H + ATPase complex to the lumen of degradative vesicles might be central to this impaired degradative capacity (Shen and Mizushima, 2014) and be required to generate a localised low pH environment.
Atg9 appears to have an important functional role in intraluminal vesicle formation and it is clear that this compartmentalisation is required to develop an appropriately localised acidic environment within degradative compartments. While Atg9 has a role in intraluminal vesicle formation in endosomes, it remains to be established whether these intraluminal vesicles can be formed directly in amphisomes and autolysosomes (Fig. 6). Previous studies have implied that Atg9 has a role in autophagosome maturation, but from this study it was evident that this role might not involve vesicular fusion with endosomes and lysosomes as previously suggested. Rather, Atg9 may function in the formation of intraluminal vesicles that are required for correct amphisome and autolysosome function. Atg9 could facilitate the recruitment of the molecular machinery that is required for intraluminal vesicular formation either at the late endosome to form multivesicular endosomes or at the amphisome-autolysosome following endosome fusion. During autophagy initiation, Atg9 has been shown to be involved in the recruitment of a range of molecular machinery including tethering proteins, Rab trafficking proteins and autophagy proteins Wang et al., 2013;Papinski and Kraft, 2014;, and it may play a similar role at the late endosome or amphisome-autolysosome. This effectively adds to the potential role of Atg9 in intracellular trafficking (Saitoh et al., 2009;Popovic and Dikic, 2014), the induction of autophagy and PAS coordination (Orsi et al., 2012), as well as the purported roles in vesicular interaction and vesicle fusion (Webber and Tooze, 2010). It remains to be established how Atg9 facilitates intraluminal vesicle formation and compartmentalised acidification.

Drosophila stocks
Drosophila stocks were maintained in standard medium at 25°C, with a 12:12 h light to dark schedule. The yeast GAL4-UAS system was used for targeted gene expression (Brand and Perrimon, 1993). Fat body specific expression of transgenes from the UAS was driven by CG-GAL4 (Asha et al., 2003) and gut specific expression of transgenes from the UAS was driven by NP1-GAL4 (Drosophila genetic resource center, Kyoto). Wildtype and transgenic stock UAS-Atg9 RNAi (stock #34901) were obtained from the Bloomington Drosophila Stock Center (Indiana University, Bloomington, USA). RNAi silencing stocks UAS-Atg9 RNAi (stock #v10045) and UAS-Vps20 RNAi (stock #v26388) were obtained from the Vienna Drosophila RNAi Centre (Vienna, Austria). UAS-Lamp1-GFP was obtained from Helmut Kramer (Center for Basic Neuroscience, Dallas, USA; Pulipparacharuvil et al., 2005). The autophagy markers Atg8a-GFP and Atg8a-Cherry were driven by an endogenous promoter and kindly provided prior to publication by Erik Baehrecke (University of Massachusetts, Medical School, MA, USA). UAS-Rab11-GFP was obtained from Markos González-Gaitán (University of Geneva, Geneva, Switzerland; Entchev et al., 2000;Wucherpfennig et al., 2003) and Donald F. Ready (Purdue University, West Lafayette, USA; Satoh et al., 2005).

Gene expression and protein analysis
For quantitative real-time PCR (qRT-PCR) analysis, RNA was isolated from the fat body tissue of 20 larvae, using an RNAqueous ® kit (Ambion, Auston, USA) according to the manufacturer's protocol. cDNA was synthesized using a High Capacity RNA-to-cDNA kit (Applied Biosystems, Waltham, USA). Quantitative RT-PCR was performed using a 7500 Fast Real-Time PCR System using a Fast SYBR Green Master Mix kit (Applied Biosystems). PCR primers were obtained from GeneWorks (Adelaide, Australia). The following primers were used to assay the in vivo efficiency of UAS-RNAi transgenes: Atg9 (CG 3615) forward, 5′-AGC AGA AGC ACG GAT TCA CA-3′, and reverse, 5′-GCA GTG CAT CAC AAA GGC AA-3′ and rp49 (CG7939, used as an endogenous control) forward, 5′-CGAGT-TGAACTGCCTTCAAGATGACCA-3′, reverse 5′-GCTTGGTGCGCTT-CTTCACGATCT-3′. Three independent biological samples were analysed for each genotype, and the mRNA expression levels were normalized against the endogenous control gene rp49, using the ΔΔC T method.
For western blotting, the fat body tissue was extracted from 10-15 either late 3rd larval instars [−4 h from puparium formation (−4 h PF)] or newly formed white pupae (0 h PF) using a previously described method (Shandala et al., 2011). Larval development was determined according to the method described by Andres and Thummel (1994). For immunoblotting, protein was separated by SDS-PAGE (30 μg protein load) and then transferred to nitrocellulose membranes (Shandala et al., 2011). The membranes were probed with either rabbit anti-Atg9 antibody at 1/200 (Novus Biologicals, Littleton, USA) or for a loading control goat anti-GADH at 1/1000 (Imgenex, Fig. 6. Potential role of Atg9 in autophagosome maturation. Schematic showing the autophagic progression from initiation at the PAS, membrane elongation to form the phagophore, membrane closure to encapsulate a mitochondrion in the autophagosome which then fuses with a multivesicular endosome to form an amphisome and finally fusion with a lysosome to produce an autolysosome. (A) Atg9 potentially facilitates the formation of intraluminal vesicle at the late endosome to form a multivesicular endosome prior to fusion with an autophagosome. (B) Atg9 also potentially facilitates intraluminal vesicle formation in amphisomes and autolysosome. PAS, phagophore assembly site; Atg, autophagy related gene.

Immunostaining and microscopy
For immunostaining, Drosophila fat body tissue was mounted on a microscope slide and fixed in 2% (v/v) paraformaldehyde in PBS, for one hour on ice. The slides were then washed in 0.1% (v/v) Tween 20 (Sigma-Aldrich, St Louis, USA) in PBS for 30 min. Non-specific interactions were blocked by incubating the tissues with 5% (v/v) BSA (Sigma-Aldrich) and 0.1% (v/v) Tween 20 in PBS for 30 min, before incubation with rabbit anti-Atg9 antibody (1/50; Novus Biologicals, Littleton, USA) or goat anti-Rab11 antibodies (1/100; obtained from Robert S. Cohen, University of Kansas; (Dollar et al., 2002), diluted in 0.1% (v/v) Tween 20 and 5% (v/v) BSA in PBS for two hours at RT and then overnight at 4°C. The tissues/ slides were washed in 0.1% (v/v) Tween in PBS for 40 min and then incubated with either goat anti-rabbit IgG or donkey anti-goat IgG, secondary Cy5 labelled antibodies (Jackson ImmunoResearch Laboratories, West Grove, USA), for two hours at RT. The slides were washed in 0.1% (v/v) Tween 20 in PBS for 40 min and then mounted in 80% (v/v) glycerol in PBS. For live cell imaging Drosophila fat body and gut tissues were stained with LysoTracker ® Red (1/200; Life Technology) for two minutes, then mounted in carbomer-940 (Snowdrift Farm, Tucson, USA) based optical coupling gel (Rothstein et al., 2006). Imaging was performed using a Zeiss LSM710 NLO confocal microscope equipped with Argon-gas and 543 nm and 633 nm solid-state lasers (Zeiss, Oberkochen, Germany) and a two-photon Mai-Tai ® , tuneable Ti:Sapphire femtosecond pulse laser (Spectra-Physics, St Clara, USA). Images were captured using a 63× oil immersion lens. Each confocal micrograph represented 1.5 μm thin optical sections.
For gut length analysis, the midguts were dissected from +4 h PF larvae into PBS and fixed in 4% (v/v) paraformaldehyde (Denton et al., 2009). Tissues were stained with Hoechst 33258 (Invitrogen, USA) and imaged with an Olympus IX71 epifluorescence microscope (Olympus, Japan). Tissues for transmission electron microscopy (TEM) analysis were prepared as described previously (Kohler et al., 2009) and viewed with an FEI Tecnai G2 Spirit TEM (FEI, Hillsboro, USA).

Image analysis
Analysis of compartment number and size was performed using Volocity ® 3D Imaging Software (PerkinElmer, Waltham, USA). For each genotype and marker combination, micrographs from a minimum of 10 biological replicates were each analysed over an area of 2025 μm 2 . For the analysis of the number and size of Atg8a-GFP compartments an object finder function was used, allowing automated detection based on compartment intensity at a specific threshold. Lamp1-GFP compartments were measured manually as the sub-compartmentalisation did not allow for automated detection. To assess the changes in Lamp1-GFP intraluminal compartments, a region of interest or ROI was drawn around the Lamp1-GFP compartment, and object finder was used to measure the area of the intraluminal compartments filled with LysoTracker ® . An average number of 13 Lamp1-GFP compartments were measured per micrograph (range 8-38). The data was presented as a mean±s.e.m, and the treatment groups compared by ANOVA analysis in GraphPad Prism with Dunnett post-hoc analysis (Prism software, version 6.01, USA).
The length around the perimeter of the midgut (from the foregut/ midgut to midgut/foregut junction) was used to define midgut length and analysed using AnalySIS Software (Olympus, Shinjuku, Japan). At least 12 midguts from each genotype were examined. The data was presented as a mean±s.e.m. and the different genotypes compared by a Student's t-test using GraphPad Prism. The number of multivesicular structures (intracellular compartments that contained smaller intraluminal vesicles) in each midgut was determined manually from a total of 13 TEM images, from three independent biological replicates for each genotype, and involved a capture field view of 8.5 μm×8.5 μm, at 11,500× magnification.