Genetic screen in Drosophila muscle identifies autophagy-mediated T-tubule remodeling and a Rab2 role in autophagy

Transverse (T)-tubules make-up a specialized network of tubulated muscle cell membranes involved in excitation-contraction coupling for power of contraction. Little is known about how T-tubules maintain highly organized structures and contacts throughout the contractile system despite the ongoing muscle remodeling that occurs with muscle atrophy, damage and aging. We uncovered an essential role for autophagy in T-tubule remodeling with genetic screens of a developmentally regulated remodeling program in Drosophila abdominal muscles. Here, we show that autophagy is both upregulated with and required for progression through T-tubule disassembly stages. Along with known mediators of autophagosome-lysosome fusion, our screens uncovered an unexpected shared role for Rab2 with a broadly conserved function in autophagic clearance. Rab2 localizes to autophagosomes and binds to HOPS complex members, suggesting a direct role in autophagosome tethering/fusion. Together, the high membrane flux with muscle remodeling permits unprecedented analysis both of T-tubule dynamics and fundamental trafficking mechanisms. DOI: http://dx.doi.org/10.7554/eLife.23367.001


Introduction
Differentiated muscle cells, or myofibers, are highly organized in order to coordinate the roles of specialized subcellular structures involved in contraction. Myofibril bundles of sarcomeres provide the contractile force. The power of contraction, however, requires synchronous sarcomere function under control of the 'excitation-contraction coupling' system that includes two membranous organelles, the sarcoplasmic reticulum (SR) and Transverse (T)-tubules (Al-Qusairi and Laporte, 2011).
The T-tubule membrane network is continuous with the muscle cell plasma membrane, with tubulated membranes that invaginate radially inward in a repeated pattern at each sarcomere. With excitation-contraction coupling, neuromuscular action potentials are transmitted along the muscle T-tubule membrane to the SR junction, or dyad/triad, triggering coordinated SR Ca 2+ release and synchronous sarcomere contractions (Al-Qusairi and Laporte, 2011). Formation of organized T-tubule membranes is thus critical for muscle function (Takeshima et al., 2015). Mechanisms must also remodel the T-tubule membrane network with ongoing myofiber reorganization in response to muscle use, damage, atrophy and aging. However, the extent and mechanisms of T-tubule remodeling remain largely unknown, in part due to challenges with observing T-tubule membrane network dynamics within intact mammalian myofibers.
The T-tubule network includes both transversal and longitudinal tubular membrane elements that form and mature with myofiber differentiation and growth. In mouse skeletal muscle, mostly longitudinal tubular membranes initially present in embryonic muscle are remodeled postnatally with expansion to predominantly transversal tubular elements (Takekura et al., 2001). In contrast, both longitudinal and transversal T-tubule elements are maintained in adult mammalian cardiac muscle (Brette and Orchard, 2003) and in insect muscles (Razzaq et al., 2001). Relatively few molecular factors are known to shape the T-tubule network, and perhaps not surprisingly, all of which so far encode for membrane-associated functions (CAV3, DYSF, BIN1/Amph2, MTM1, DNM2) (Butler et al., 1997;Hnia et al., 2012;Lek et al., 2012;Morlot and Roux, 2013;Tang et al., 1996). Mutations in each also are associated with human myopathy and/or cardiomyopathy with T-tubule disorganization (Bashir et al., 1998;Betz et al., 2001;Bitoun et al., 2005;Laporte et al., 1996;Liu et al., 1998;Minetti et al., 1998;Nicot et al., 2007), pointing to the critical importance of membrane-mediated mechanisms to maintain the T-tubule membrane network.
Drosophila is a powerful system for insights into the functional requirements for T-tubule formation and remodeling. The BIN1 BAR-domain protein has a conserved function involved in membrane tubulation required for T-tubule formation that was first described for the single Drosophila homolog, Amphiphysin (Lee et al., 2002b;Razzaq et al., 2001). The amph null mutant flies lack transversal T-tubule element membranes in myofibers at all developmental stages, corresponding with both larval and adult mobility defects (Razzaq et al., 2001). In contrast, the myotubularin (mtm) fly homolog of mammalian MTM1/MTMR2/MTMR1 subfamily of phosphatidylinositol 3-phosphate phosphatases is required only at later stages in development for T-tubule remodeling. While mtm loss of function has no obvious effects on larval muscle T-tubule organization or function, mtm-depleted post-larval stage muscles lack transversal T-tubule membranes with adult mobility defects in eclosion and flight (Ribeiro et al., 2011). Together, the amph and mtm mutant conditions that both lack transversal T-tubule elements in post-larval stage muscle yet different early development requirements underscores that distinct mechanisms are involved in T-tubule formation (amph-dependent) versus maintenance/remodeling (amph-and mtm-dependent).
In Drosophila, a set of larval body wall muscles that persist as viable pupal abdominal muscles, called dorsal internal oblique muscles (IOMs), are essential for adult eclosion (Kimura and Truman, 1990). During metamorphosis, changes in IOM cell size and myofibril content have been noted (Kuleesha et al., 2014(Kuleesha et al., , 2016. We previously showed that wildtype IOMs undergo dramatic cortical and membrane remodeling with costamere integrin adhesion complex disassembly and reassembly at discrete pupal stages (Ribeiro et al., 2011). In contrast, the mtm-depleted IOMs exhibited persistent disassembly or a block in reassembly of integrin costameres along with the loss of transversal T-tubule membranes at late pupal stages, but without any precocious cell death (Ribeiro et al., 2011). A striking feature in the mtm-depleted IOMs was the accumulation of endosomal-like membranes decorated with integrin and T-tubule markers, Amph and Discs large (Dlg1, a PDZ protein). Altogether, these results suggest that T-tubule membranes may undergo disassembly-reassembly with normal myofiber remodeling, including the delivery of disassembled T-tubule membrane into an endomembrane trafficking pathway. The role for a molecular-cellular program in control of T-tubule remodeling that is at least partially distinct from that involved in initial T-tubule formation raises many questions about possible mechanisms, including the regulation of T-tubule organization and dynamics, the membrane fate(s) and source(s) with disassembly-reassembly, respectively, and the specific membrane trafficking routes and effectors involved. Possible hints may come from studies of other specialized dynamic cell membrane invaginations shown to involve endosomal and Golgi membrane trafficking pathways, such as cellularization of Drosophila syncytial embryos Harris, 2013, 2014;Pelissier et al., 2003) and the tubulated demarcation membrane system in megakaryocyte platelet formation (Eckly et al., 2014).
Membrane trafficking relies on the large family of Rab GTPases, with over sixty Rabs in humans and thirty in flies (Klö pper et al., 2012). The different Rabs are under distinct spatiotemporal regulation for recruitment, activation and functions at specific membrane compartments or domains. Guanine nucleotide exchange factors (GEFs) convert specific inactive GDP-bound Rabs to an active GTPbound form. Active Rab-GTP then recruits a range of specific effector proteins to the membrane that mediate key trafficking functions, including cargo selection, vesicle budding, transport, tethering and fusion. Subsequently, GTPase-activating proteins (GAPs) deactivate Rabs by promoting GTP hydrolysis. Many membrane compartments have been defined by well-established localized functions of specific Rabs, for example: ER (Rab1), Golgi (Rab1, Rab6), secretory vesicles (Rab8), early endosomes (Rab5, Rab21), recycling endosomes (Rab11, Rab35), late endosomes (Rab7, Rab9), lysosomes (Rab7) and others (Jean and Kiger, 2012;Stenmark, 2009). Thus, identifying the specific Rabs required for a cellular process can provide clues to potential underlying membrane trafficking mechanisms involved. However, examples exist of Rabs with multiple known sites of function or yet unknown functions, and conversely, certain cellular processes -like T-tubule remodeling -lack defined roles yet for any Rabs.
Here, we utilized the advantages of Drosophila IOMs to screen for Rab GTPases and related membrane trafficking functions required for T-tubule remodeling in intact muscle. Our results show that the entire contractile and excitation-contraction coupling system, including T-tubules, are disassembled and reassembled in IOMs during Drosophila metamorphosis. We found that autophagy, the membrane trafficking process for degradation of cytoplasmic contents by delivery to lysosomes, is upregulated with IOM remodeling where it plays an indispensable role for progression through T-tubule disassembly to reassembly. Our genetic analysis of IOM remodeling also reveals an unexpected and broad role for Rab2 in autophagy in flies and mammals. From our data, we propose that Rab2 localizes to autophagosomes where it interacts with the HOPS complex, which in turn, mediates tethering and trans-SNARE complex formation with Rab7-marked lysosomes to promote autophagosome-lysosome fusion. Together, these results show that Drosophila IOM remodeling provides an unprecedented in vivo context for discovery and analysis of T-tubule dynamics with relevance to human myopathy, as well as an ideal system due to high membrane flux to study fundamental trafficking pathways.

Differentiated myofiber remodeling includes regulated T-tubule membrane disassembly and reassembly
To monitor T-tubule remodeling in Drosophila abdominal muscles, we expressed the mCD8:GFP transmembrane fusion protein as a marker of the muscle cell and T-tubule membranes (Peterson and Krasnow, 2015). When observed in live IOM persistent larval muscles through the cuticle at 4 days after puparium formation (4d APF), mCD8:GFP showed a mesh-like pattern ( Figure 1A . The brightness of mCD8:GFP enabled us to monitor membrane dynamics during IOM remodeling in undissected live animals. As previously reported (Kuleesha et al., 2014;Wasser et al., 2007), the IOMs remodeled during metamorphosis with myofiber thinning through 2d APF followed by rethickening from 3-4d APF ( Figure 1E-F, top and middle rows). Along with these cell morphology changes, the well-organized mCD8:GFP-marked membranes detected in third instar larval precursor muscles were disassembled in IOMs by 1-2d APF and then reassembled by 4d APF ( Figure 1F, middle and bottom rows).
To more specifically investigate myofiber remodeling, we monitored Dlg1:GFP (T-tubules), Reticulon:GFP (Sarcoplasmic Reticulum; SR), and GFP:actin (myofibrils) at 24h intervals during metamorphosis ( Figure 2A). Each of these organelles was disassembled by 1-2d APF, and then reassembled by 4d APF in IOMs ( Figure 2A). Furthermore, ultrastructual analysis of myofiber remodeling by transmission electron microscopy (TEM) imaging of IOM transverse sections ( Figure 2B) confirmed both the timing and extent of the disassembly and reassembly stages during metamorphosis ( Figure 2C-G). This reveals that differentiated myofiber structures critical for muscle function, including T-tubule membranes, undergo regulated and stereotypical remodeling in IOMs during metamorphosis.
A unique T-tubule remodeling phenotype with knockdown of a set of genes In order to identify functions involved in T-tubule membrane remodeling, we performed a muscletargeted RNAi screen of candidate membrane trafficking-related genes, including all fly genes Source data 1. Relates to Figure 1D. DOI: 10.7554/eLife.23367.003 Figure 1 continued on next page predicted to encode Rab GTPases, Arf GTPases, sorting nexins, BAR domain proteins, SNARE proteins, and phosphoinositide regulators (Supplementary file 1). Since T-tubule organization is required for muscle function (Al-Qusairi and Laporte, 2011;Razzaq et al., 2001), we first screened for muscle-targeted RNAi effects on fly mobility. Among 300 RNAi lines tested, 77 lines showed a defect in adult eclosion or mobility, 151 lines resulted in normal viability and mobility, and 83 lines were unscored due to pre-adult lethality ( Figure 3A). As a secondary screen, we tested the 77 RNAi lines with eclosion or mobility defects for mCD8:GFP organization by live cell imaging in IOMs at 4d APF. We identified 10 RNAi lines targeting a set of 5 different genes, representing two Rab GTPases and three soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins, with a similar phenotype ( Figure 3A-B and Supplementary file 1). Instead of the organized T-tubule network seen in control IOMs, RNAi of Rab2, Rab7, Stx17, SNAP29 or Vamp7/8 each resulted in an accumulation of mCD8:GFP-positive small vesicles that filled often misshapen or swollen myofibers ( Figure 3B). Unlike in controls, T-tubules (Dlg1) and most myofibrils (F-actin) were absent throughout these RNAi-treated IOMs at 4d APF ( Figure 3C-D).
The 'Rab2 class' of shared phenotypes suggested that these five genes all function in a shared process or pathway in IOMs. Indeed, it has been shown that four of the five genes play known roles together in lysosome fusion: Stx17, SNAP29, and Vamp7/8 form a trans-SNARE complex involved in autophagosome-lysosome fusion (Itakura et al., 2012;Takáts et al., 2013), while Rab7 functions in late endosome-lysosome fusion as well as late steps in autophagy (Gutierrez et al., 2004;Jäger et al., 2004;Hegedűs et al., 2016). It was unexpected, however, to find a shared RNAi phenotype between this set of known functions and Rab2, which had been implicated with functions at the ER and Golgi (Saraste, 2016). The specificity of the Rab2 RNAi phenotype was confirmed by rescue with co-expression of a wildtype Rab2 transgene (Figure 3-figure supplement 1A-B). From this, we speculated that the 'Rab2 class' of shared phenotypes was in each case a result of a block in autophagosome-lysosome fusion. To explore this possibility, we tested a requirement for the homotypic fusion and protein sorting (HOPS) complex that is known to mediate SNARE-dependent autophagosome-lysosome fusion (Jiang et al., 2014;Takáts et al., 2014). As expected, disruption of the HOPS complex with RNAi of subunits, Vps39, Vps18, and Vps11 each phenocopied the specific Rab2 class of IOM defects ( Figure 3E and Figure 3-figure supplement 1C). These results suggest that a defect in autophagosome-lysosome fusion leads to the unique Rab2 class of RNAi phenotypes in IOM remodeling and predicts a novel role for Rab2 in autophagy.

Rab2 or Rab7 knockdown blocks autophagosome-lysosome fusion with autophagy-dependent myofiber remodeling
To address the underlying role for the 'Rab2 class' of genes in T-tubule remodeling, we characterized autophagy in IOMs with Rab2, Rab7 or Stx17 knockdown at 4d APF. The mCherry:GFP:Atg8a autophagic flux reporter indicates both dually-labeled Atg8-marked autophagosomes and, due to the greater resistance of mCherry than GFP to lysosomal proteases, the successful delivery of autophagosomes to the lysosome by mCherry-labeled autolysosomes (Kimura et al., 2007). In control IOMs, there were dual mCherry-GFP-positive autophagosomes, as well as just mCherry-positive degradative autolysosomes ( Figure 4A). In contrast, the Rab2, Rab7 and Stx17 RNAi IOMs contained a striking increase in dual mCherry-GFP-positive puncta ( Figure 4A-B), showing that autophagosome clearance was severely blocked either due to accumulation of autophagosomes or nondegradative autolysosomes. The Stx17 SNARE localizes to the outer membrane of fully formed autophagosomes then detaches upon lysosomal degradation of the autophagosomal inner membrane (Itakura et al., 2012;Tsuboyama et al., 2016). In the RNAi conditions, Stx17 localized to the vesicle membranes ( Figure 4C) also marked with Atg8 ( Figure 4D; 0.53 Pearson correlation), indicating their identity as primarily mature autophagosomes. Confirming these results, TEM myofiber transverse sections ( Figure 2B) revealed Rab2, Rab7 or Stx17 RNAi-depleted IOMs similarly and uniformly filled with thousands of accumulated autophagosomes carrying nondegraded cytoplasm and organelles   Figure 4E-F). Thus, similar to previous reports of Rab7, Stx17, SNAP29 and Vamp7/8 functions in other Drosophila tissues (Hegedűs et al., 2016;Takáts et al., 2013), with myofiber remodeling, Rab2 also is required for autophagosome-lysosome fusion.
We further validated a requirement for autophagy in T-tubule remodeling by disrupting early steps in autophagy with muscle-targeted Atg1, Atg3 or Atg18 RNAi. In each case, RNAi resulted in disorganized and fragmented T-tubules and sparse myofibrils in IOMs at 4d APF ( Figure 5A-B). Furthermore, TEM analysis of IOMs with Atg1 or Atg18 RNAi showed a lack of T-tubules, disorganized myofibrils and a striking accumulation of mitochondria that filled the cells ( Figure 5C). The accumulation of fluorescently-marked mitochondria ( Figure 5D) confirmed the TEM observations of free mitochondria within the cytoplasm with Atg1 RNAi ( Figure 5C) or inside mature but blocked autophagosomes with Rab2 RNAi ( Figure 4E), suggesting that mitochondria are a major cargo of autophagy during IOM remodeling. As expected for a block in autophagy induction, and unlike the Rab2 RNAi, autophagosomes did not occur or accumulate in IOMs with Atg1 and Atg18 RNAi.

Autophagy is simultaneously upregulated with a requirement for proper T-tubule membrane disassembly
The autophagy requirement for T-tubule remodeling raised the questions whether levels of autophagy change and if required at distinct stages over IOM remodeling. In wildtype animals, we used imaging to monitor the amount of autophagosomes (GFP:Atg8) and endolysosomes/autolysosomes (GFP:Lamp1, GFP:Rab7) in live IOMs over 24h intervals of metamorphosis ( Figure 6A). The number of Atg8 puncta indicative of autophagosomes dramatically increased by 1d APF and then sharply decreased by 3d APF ( Figure 6A, top row and 6B). We confirmed a similar autophagosome distribution detected by immunostaining of endogenous Atg8 in wildtype IOMs at 1d APF ( Figure 6-figure supplement 1A). These results parallel the increased prevalence of uniformly sized mCD8:GFPmarked vesicles ( Figure 1F) and autophagosomes detected by TEM ( Figure 2D-F) over 1-3d APF during wildtype IOM remodeling. Likewise, the number of Lamp1 and Rab7 marked endolysosomes or amphisomes/autolysosomes also increased during pupal stages 1d-3d APF ( Figure 6A). Importantly, the increase in autophagy at 1d APF does not reflect an autophagy role in myofiber cell death, as the IOMs normally survive throughout metamorphosis (Figures 1-2) and assist days later with adult eclosion (Ribeiro et al., 2011).
The autophagy profile 1-3d APF ( Figure 6A) appeared to inversely correlate with the presence of T-tubule membranes ( Figure 2). To better determine the timing and relationship between autophagy and T-tubule disassembly, we performed a more focused timecourse experiment spanning IOM remodeling between late larval to 1d APF stages. We quantified a reproducible onset of autophagy induction between 12h and 18h APF that peaked by 20h APF in IOMs ( Figure 6C, top row and 6D, red). In parallel experiments, we used live imaging of a PI(4,5)P 2 biosensor to detect T-tubules and the plasma membrane. While the T-tubules remained intact from third instar larval through 12h APF stages, the PI(4,5)P 2 -marked T-tubule membrane network became partially fragmented by 18h APF and was completely absent by 20h APF in wildtype IOMs ( Figure 6C, bottom row and 6D, blue). Only transient PI(4,5)P 2 -marked membrane rings that appeared with T-tubule fragmentation could be detected in the central myofiber region up to 1d APF, while PI(4,5)P 2 continued to mark the plasma membrane throughout IOM remodeling. These results show highly reproducible, coincident onsets and peak occurrences of both T-tubule disassembly and autophagy induction between 18-20h APF ( Figure 6C-D), supporting interrelated roles for autophagy and T-tubule remodeling.
Consistent with the timing of autophagy induction in wildtype IOMs ( Figure 6D), a timecourse analysis of T-tubule remodeling in both the Rab2 class (Rab2, Rab7 or Stx17) and Atg1 class (Atg1, Atg18 RNAi) of RNAi phenotypes revealed that defects arose within the first day of IOM remodeling ( Figure 6E and  3IL and 4d APF stages (C and G). At 1d APF, myofibrils were partially lost with mostly disorganized membranes (D). At 2d APF, myofibrils were completely absent with obvious appearance of autophagosomes and electrondense lysosomal compartments (E). At 3d APF, myofibrils were reassembled but not well organized with a lack of obvious T-tubules (F). DOI: 10.7554/eLife.23367.005 Figure 3. A unique T-tubule remodeling phenotype with knockdown of a set of known and unknown gene functions in autophagy. All IOMs imaged at 4d APF. (A) Muscle-targeted RNAi screen of IOM remodeling. In primary screen of 300 selected muscle-targeted RNAi lines (see text), 77 lines exhibited eclosion or adult mobility defects; these lines were used in a secondary screen for mCD8:GFP organization by confocal imaging. Three abnormal phenotype categories were identified for 37 lines. The shared 'small vesicle' phenotype was found for 10 RNAi lines for five genes presented Figure 3 continued on next page observed in the IOM precursor muscles at the third instar larval stage ( Figure 6E, left), in line with the normal larval mobility observed with each muscle-targeted RNAi depletion. However, initial defects in mCD8:GFP organization appeared between 12h and 18h APF ( Figure 6E), at a time when normally both T-tubule disassembly and autophagy induction occur ( Figure 6C-D). These knockdown IOMs remained persistently blocked at an 18-20h stage in remodeling throughout metamorphosis.
In control IOMs at the time of T-tubule disassembly at 18-20h APF ( Figure 6C-6D), mCD8:GFPmarked membranes reorganized into dispersed bright patches and some vesicles ( Figure 6E, top row and Figure 6F). In the case of the Rab2 class of RNAi phenotypes, T-tubule disassembly at 18-24h APF occurred in conjunction with an abnormal accumulation of uniformly-sized mCD8:GFP vesicles ( Figure 6E, middle row; Figure 6F; Figure 6-figure supplement 2B, top rows), many of which were identified as Atg8-marked autophagosomes ( Figure 6G). In contrast, the Atg1 class of RNAi phenotypes showed normal timing in initial T-tubule membrane disassembly at 18-20h APF, but with abnormally large clusters of bright mCD8:GFP-positive membrane patches ( Figure 6F). By 1d APF, the bright membrane clusters increased in size to reveal stacked membrane whirls that persisted to 4d APF ( Figure 6E, bottom row; Figure 6-figure supplement 2B, bottom row; Figure 5A). Together with the lack of T-tubules and autophagosomes in IOMs with autophagy disruption at 4d APF ( Figure 5), these results indicate that autophagy plays a key role downstream of initiation and membrane scission in T-tubule disassembly for subsequent mobilization of disassembled T-tubule-derived membranes.
While the specific phenotypes were distinct with knockdown of genes required either for autophagy induction/autophagosome biogenesis (Atg1, Atg3 and Atg18) or autophagy clearance (Rab2, Rab7 and Stx17), all autophagy functions were consistently required by 18-20h APF when both autophagy induction and T-tubule disassembly normally occurs. Thus, in IOMs, changes in autophagy levels are concurrently regulated and required for myofiber T-tubule membrane remodeling, and specifically, the Atg1 pathway mediates autophagy-dependent progression of T-tubule membrane disassembly.

Rab2 is required for autophagic flux in other tissues and in mammalian cells
To generalize our findings to other cell types, we tested effects of Rab2 RNAi in third instar larval fat body, a tissue with a well-established starvation-induced autophagy response. Consistent with our data in IOMs, Rab2 RNAi increased the amount of both Atg8-marked autophagosomes and uncleared p62 autophagic cargo (  T-tubule (Dlg1, green) and myofibril (F-actin, pink) organization in IOMs from control and Rab2, Rab7, Stx17, SNAP29, or Vamp7/8 RNAi conditions. (E) RNAi of HOPS components, Vps39, Vps18, or Vps11, exhibited shared phenotypes of (top) many mCD8:GFP-marked small vesicles and (bottom) lack of T-tubules (Dlg1, green) and myofibrils (F-actin, pink). See  Mammals each have two Rab2 isoforms, Rab2A and 2B (Aizawa and Fukuda, 2015). To examine whether Rab2 has an evolutionarily conserved role in autophagy, we deleted Rab2A and Rab2B using the CRISPR/Cas9 system in mouse embryonic fibroblasts (MEFs). We generated both single knockouts (Rab2A_KO, Rab2B_KO) and a double knockout cell line (Rab2A/B_DKO; Figure 7A). In the 'fed' culture condition, the number of LC3 puncta indicative of autophagosomes significantly increased in Rab2A/B_DKO versus in control cells ( Figure  Thus, overall lysosomal maturation appears unaffected by Rab2A/B_DKO despite a block in autophagic clearance. Collectively, a conserved Rab2 function is indispensable for autophagic flux, which for Rab2A/B is required after autophagosome delivery in MEFs.

Rab2 localizes to mature autophagosomes
It is reported that Drosophila Rab2 has an affinity with the HOPS complex in both Drosophila and mammals (Gillingham et al., 2014;Kajiho et al., 2016). We confirmed the interaction using Rab2A/ B co-immunoprecipitation assays. As previously reported, Rab7 but not Rab5 had an affinity with Vps39 and Vps41, two selective Rab-binding components of the HOPS complex ( Figure 8A). Similar to Rab7, both Rab2A and Rab2B were also co-immunoprecipitated with Vps39 and Vps41 ( Figure 8A). The HOPS complex tethers membrane compartments to promote fusion between endosomes and lysosomes (Balderhaar and Ungermann, 2013) or between autophagosomes and lysosomes (Jiang et al., 2014;Takáts et al., 2014), and Rab7 has been localized to late endosomes/ lysosomes in both fusion events (Hyttinen et al., 2013). The following source data is available for figure 4: Source data 1. Relates Figure 4B. DOI: 10.7554/eLife.23367.010 Source data 2. Relates to both Figure 4F and Figure 5C. DOI: 10.7554/eLife.23367.011 . Immuno-EM also identified colocalization of endogenous LC3 and GFP:Rab2A on the same membranes of autophagic structures with the typical appearance of autophagosomes containing undigested contents (Figure 8-figure supplement 1B). It is established that LC3 localizes to mature autophagosomes as well as the isolation membrane, an elongating pre-autophagosome structure (Kabeya et al., 2000;Mizushima et al., 2001) marked by Atg16L1 (Mizushima et al., 2003). To assess to what extent Rab2A/B localizes to the isolation membranes and/or autophagosomes, we examined colocalization between Atg16L1, LC3 and GFP:Rab2A/B. Rab2A/B colocalized well with LC3 but not noticeably with Atg16L1 ( Figure  To characterize the dynamics of Rab2A localization in autophagy, we performed live cell timelapse imaging. In starved MEFs, recruitment of CFP:LC3 preceded YFP:Rab2 by several minutes (Figure 9-figure supplement 1A), consistent with Rab2A localization to autophagosomes but not isolation membranes ( Figure 8G-H). In turn, recruitment of YFP:Rab2A to autophagosomes preceded CFP:Stx17 by several minutes in most cases ( Figure 9A), suggesting that Rab2A localization is independent of Stx17. Consistent with the timing of Rab2A recruitment, Drosophila Rab2 localization at autophagosomes was unaffected by Rab7, Stx17 or Vps39 RNAi in IOMs ( Figure 9B-C). Since Rab2 has affinity with the HOPS complex, we tested whether Rab2 RNAi affects the recruitment of HOPS to autophagosomes. As previously reported (Takáts et al., 2014), Dor, a fly homolog of the Vps18 HOPS complex component, colocalized with GFP:Atg8 in control IOMs ( Figure 9D). In contrast, this colocalization was diminished by Rab2 RNAi (Figure 9E), indicating that Rab2 is important for recruitment of the HOPS complex to autophagosomes. Collectively, Rab2 is an autophagosomeassociated Rab GTPase that promotes autophagosome-lysosome fusion through interactions with the HOPS complex ( Figure 9F), and T-tubule remodeling requires a high volume of membrane flux through the autophagy pathway. RNAi or Atg1 RNAi (as in Figure 1F, bottom). Both Rab2 and Atg1 RNAi show normal T-tubules in 3IL muscle and initial defects in membrane organization by 1d APF that persist as a block in remodeling through 4d APF. (F) mCD8:GFP membrane in central regions of IOMs at 20h APF upon T-tubule disassembly in control, Rab2 RNAi or Atg1 RNAi, with mCD8:GFP membrane reorganization into bright patches (arrows), big membrane patches/stacks (red arrows) or membrane vesicles (arrowheads). (G) GFP:Atg8 at 1d APF in control and Rab2 RNAi IOMs. See Figure 6-figure supplements 1 and 2 for related data. DOI: 10.7554/eLife.23367.013 The following source data and figure supplements are available for figure 6: Source data 1. Relates to Figure 6B. DOI: 10.7554/eLife.23367.014 Source data 2. Relates to Figure 6D. DOI: 10.7554/eLife.23367.015 Source data 3. Relates to Figure 6D.

Discussion
The importance of T-tubule organization for muscle function is well established. However, the dynamics and mechanisms that shape the T-tubule membrane network are largely unexplored, in part due to methodological challenges in many systems. In the present study, we characterized a wildtype myofiber remodeling program by confocal and electron microscopy in intact muscles in vivo. In Drosophila IOMs during metamorphosis, the entire contractile and excitation-contraction coupling system, including T-tubules, are disassembled and then reassembled. This process highlights that myofibers harbor distinct programs for initial T-tubule formation versus regulated T-tubule remodeling. This likely includes additional mechanisms for T-tubule membrane disassembly and renovation, features that reflect those seen with mammalian myofiber atrophy and recovery (Piccirillo et al., 2014). We demonstrate that the Drosophila body wall muscles provide an unprecedented system permitting a combination of powerful visualization and systematic perturbation analysis, including the first genetic screens, of T-tubule dynamics and organization. We uncovered both that autophagy is indispensable for the remodeling, and that Rab2 plays an unexpected role at autophagosomes for autophagic clearance in this process and in broader contexts.
Autophagy is upregulated with the onset of IOM remodeling during metamorphosis ( Figure 6A-D). Further, disruption of autophagy initiation, autophagosome formation or clearance all induced loss of T-tubules with a block in IOM remodeling at/after T-tubule disassembly ( Figure 6E). As far as we know, this is the first report of a non-cell death role of autophagy in Drosophila metamorphosis. The role of autophagy in IOMs that persist and redifferentiate during metamorphosis is clearly different from its roles in pupal midgut and salivary gland cells that undergo autophagic forms of cell death (Berry and Baehrecke, 2007;Chang et al., 2013;Das et al., 2012;Lee et al., 2002a). There are multiple speculative direct or indirect role(s) for autophagy specifically in T-tubule membrane remodeling: (1) a direct role in T-tubule membrane recycling, as a means to deliver disassembled T-tubule membrane via autophagosomes to lysosomes or related organelles for intracellular storage, then later redeployed to contribute to T-tubule reassembly; (2) an indirect role in cell renovation, including T-tubule membrane clearance, to permit cell space for redifferentiation; or (3) an indirect role in cell metabolism, to support cell survival and/or the energy cost of redifferentiation with starvation during metamorphosis. Most likely, autophagy serves some combination of these roles in IOM remodeling.
How could autophagy play a direct role in T-tubule remodeling? It was surprising that mCD8: GFP-positive small vesicles accumulated to a similar degree as autophagosome numbers in IOMs  when autophagosome-lysosome fusion was blocked (Figures 3 and 4). This suggests that mCD8: GFP localizes to autophagosomes during IOM remodeling. It is possible that T-tubule membranes are a source of autophagosomal membrane, at least in part: mCD8:GFP labels the muscle plasma membrane and T-tubules in larval muscle precursor cells of IOMs ( Figure 1F), and T-tubule disassembly coincides with the upregulation in autophagy early in metamorphosis ( Figure 6C-D). Also, disruption of autophagy induction blocked normal progression in disassembly and remodeling of T-tubule-derived mCD8:GFP-marked membranes ( Figure 6E and Figure 6-figure supplement 2B).
In the absence of autophagy initiation, mCD8:GFP-positive stacked membranes were observed ( Figure 6E and Figure 6-figure supplement 2B), likely retained or partially disassembled T-tubules. We propose that T-tubules are remodeled through autophagosomes. It is important to note that T-tubules are not an apparent autophagic cargo, but instead, a possible source of autophagosome membrane. In this scenario, T-tubules are disassembled into autophagosomes and then reassembled from subsequent autolysosome-related structures ( Figure 9F), both of which successively increased in numbers during wildtype IOM remodeling ( Figure 6A).
Alternatively or additionally, other roles for autophagy could indirectly impact T-tubule remodeling. Extensive IOM atrophy with nearly complete disassembly of the contractile and excitation-contraction systems by 1d APF is followed by a rapid re-differentiation within hours after 3.5d APF. Autophagy could be required to simply clear away and degrade the old contraction systems in order to make space to rebuild and realign new systems, as well as permit the normal central repositioning of nuclei away from the myofiber cortex. However, the persistent block in early IOM remodeling with autophagy disruption suggests that the remodeling normally proceeds through a progression of interrelated steps rather than independent programs for disassembly and reassembly. Autophagy also has a well-established role in metabolic homeostasis through the recycling of amino acids and turnover of damaged mitochondria in the lysosome. Our data suggests that mitochondria are a major autophagic cargo with IOM remodeling. In conditions that disrupted autophagy initiation (Atg1, Atg18 RNAi), the cytoplasm was abnormally filled with mitochondria in IOMs at 4d APF ( Figure 5C). Consistent with that, a significant portion of autophagosomes harbored intact mitochondria when autophagosome-lysosome fusion was blocked (Rab2, Rab7 or Stx17 RNAi; Figure 4E). This is different from observations in larval muscle, in which mitochondria were notably absent in autophagosomes that accumulated with a block in autophagy (Zirin et al., 2013). It is possible that mitophagy, a selective form of autophagy for mitochondrial turnover, is upregulated and could play both metabolic and cell renovation roles in IOM remodeling. Interestingly, the autophagy-blocked IOMs remained viable throughout metamorphosis, suggesting that autophagy is not absolutely required for cell survival through the starvation with metamorphosis.
Through a systemic screen of all Drosophila Rab GTPases, we uncovered an unexpected role for Rab2 in autophagy. The striking Rab2 RNAi IOM phenotype was shared with RNAi of other functions known to be specifically required for autophagosome-lysosome fusion. Genetic blockade of autophagosome-lysosome fusion resulted in a dramatic phenotype, with massive accumulations of autophagosomes within IOMs (Figure 4). Previously, autophagosome-lysosome fusion was shown to involve the cooperative functions of Rab7, the HOPS tethering complex, and a trans-SNARE complex between Stx17, SNAP29 and VAMP7/8 (Shen and Mizushima, 2014). Among these tethering and fusion functions, it has been shown that Stx17 (a hairpin SNARE) is recruited to autophagosomal membranes, while Rab7 and VAMP7/8 localize to endolysosomal membranes. Stx17 localizes to autophagosomes as well as to the ER and mitochondria (Itakura et al., 2012), but the HOPS complex directly associates and colocalizes with Stx17 only at autophagosomes (Jiang et al., 2014;Takáts et al., 2014). This suggests that Stx17 is not a sole determinant for HOPS complex recruitment. We propose that Rab2 is required for the autophagosomal recruitment of the HOPS complex. Rab2 specifically localized to completed autophagosomes (Figures 8-9), and Rab2 had an affinity with the HOPS complex ( Figure 8A), as does Stx17 (Gillingham et al., 2014). We envision that upon completion of autophagosome biogenesis/maturation, Rab2 and Stx17 are recruited to the outer autophagosomal membrane. Then, the HOPS complex is subsequently recruited to autophagosomes in a Rab2-depedent manner through coincident interactions with both Stx17 and Rab2 ( Figure 9D-E). At the same time, the HOPS complex binds Rab7 on lysosomes. In turn, the HOPS complex tethers autophagosomes and lysosomes to promote trans-SNARE complex formation between Stx17, SNAP29 and Vamp7/8 and ultimately autophagosome-lysosome fusion ( Figure 9F).
We found that the Rab2 role in autophagy discovered in fly muscle relates to a broader autophagy requirement in other cell types and across species. The localization of Rab2 on autophagosomes in Drosophila IOMs was conserved for both Rab2A and Rab2B in mouse embryonic fibroblasts (MEFs). As in flies, the Rab2A/2B double knockout led to a delay or block in autophagy clearance as indicated by accumulation of LC3/Atg8. However, the specific Rab2 loss-of-function phenotypes were not identical. While Rab2 was required for autophagosome-lysosome fusion in fly IOMs, the Rab2A/2B double knockouts in MEFs indicated a requirement at a later step in autophagic clearance. Interestingly, this disparity in autophagy phenotypes across species is also seen with Rab7. In flies and yeast, Rab7/Ypt7 is essential for autophagosome-lyososome/vacuole fusion (Kirisako et al., 1999;Hegedűs et al., 2016), while mammalian Rab7 knockdowns more clearly indicate a required role in autolysosome maturation (Gutierrez et al., 2004;Jäger et al., 2004). Other examples indicate that the autophagosome-lysosome fusion machinery is not highly evolutionarily conserved. The Stx17-SNAP29-VAMP7/8 trans-SNARE complex is conserved in Drosophila and mammals, but not in yeast (Shen and Mizushima, 2014), where no autophagosomal SNARE has been reported so far. Moreover, budding yeast do not encode for Rab2 (Klö pper et al., 2012).
Altogether, it is plausible that Rab2 is required for autophagosome-lysosome fusion efficiency, and Rab2-dependency is variable across different tissues or species. Two possible models could explain the different Rab2 autophagy requirements in flies and mouse cells. First, it is suggested that autophagosomes sequentially fuse with endosomes then lysosomes to become amphisomes and autolysosomes, respectively. If either of the steps requires Rab2A/2B, then intermediates with partially degraded contents could accumulate in double knockout MEFs. Alternatively, an autophagosome may normally fuse with multiple lysosomes to ensure full degradation of its contents (Tsuboyama et al., 2016). In the absence of Rab2A/2B in MEFs, autophagosomes could still fuse but not with a sufficient number of lysosomes, resulting in an accumulation of partially digested autolysosomes.
Rab2 has been previously associated with transport events at the Golgi apparatus, ER-to-Golgi traffic and secretory granule formation (Ailion et al., 2014;Edwards et al., 2009;Saraste, 2016), as well as in a C. elegans endocytic/phagocytic pathway (Chun et al., 2008;Mangahas et al., 2008). Gillingham et al. systematically explored Rab effectors in Drosophila cultured cells, and found that Rab2 interacts with the HOPS complex besides known Golgi-resident effectors (Gillingham et al., 2014). The interaction between Rab2 and HOPS complex is also conserved in mammals (Kajiho et al., 2016) (Figure 8A), and we found the unexpected Rab2 localization to autophagosomes ( Figure 8B-H and Figure 8-figure supplement 1). Thus, it is likely that Rab2 exerts multiple functions through interaction with different effectors at different places. We do not exclude a possible Rab2 function in the endosome-lysosome system that affects autophagic flux, although we did not detect clear lysosomal defects in Rab2A/B knockout MEFs ( Figure 7H-I and Figure 7figure supplement 2F-I). Several other factors that localize to autophagosomes or late endosomeslysosomes, including Atg14, PLEKHM1 and EPG5 (Tian et al., 2010;Tabata et al., 2010;Carlsson and Simonsen, 2015), have been shown to control autophagosome maturation. It is plausible that Rab2 contributes to autophagosome maturation through both a direct role in the fusion mechanism and an indirect role in endo-lysosome maturation, the same as Rab7 and the HOPS complex.
How Rab2 localizes to autophagosomes remains unclear. Localization of Rab2 on autophagosomes in IOMs did not depend on HOPS complex subunits, Vps39 and Vps41, or on Stx17 ( Figure 9B-C). Further studies will be needed to determine the identities of the Rab2 guanine nucleotide exchange factor (GEF) and GTPase-activating protein (GAP) that regulate Rab2 GTPase activity in autophagosome-lysosome fusion. A conserved TBC domain protein, OATL1/TBC1D25, is a strong candidate for a Rab2 GAP, given OATL1 localization to autophagosomes and involvement in autophagosome-lysosome fusion (Itoh et al., 2011). Further, it was reported that OATL1 directly bound to and showed GAP activity for Rab2A (Itoh et al., 2006).
Autophagy is critical for the maintenance of myofiber homeostasis in mammalian skeletal muscle. It is known that several myopathies are associated with excess accumulation of autophagic structures in muscle (Malicdan et al., 2008). Further, loss of autophagy in mouse skeletal muscle shows anomalies, including abnormal mitochondria, disassembled sarcomeres and disorganized triads (Masiero et al., 2009), as also seen in aged muscle (Demontis et al., 2013;Weisleder et al., 2006). It is established that autophagy is down-regulated during the course of aging (Rubinsztein et al., 2011). This evidence points to a possible significance of autophagy in myofiber remodeling and in T-tubule maintenance. Jumpy/MTMR14 PI3-phosphatase and Dynamin-2 (DNM2) GTPase, two causative genes of human centronuclear myopathy, are required for not only T-tubule maintenance but also proper progression of autophagy (Dowling et al., 2010;Durieux et al., 2012). Based on these reports and our findings, we speculate that their roles in T-tubule maintenance are mediated, at least in part, through autophagy.
Signaling pathways that regulate atrophy and hypertrophy in Drosophila have been identified (Piccirillo et al., 2014), however, the mechanisms and direct mediators of muscle remodeling remain largely unknown. IOM remodeling is a good model to study the mechanisms of muscle remodeling, given that the signaling pathways that control muscle remodeling are conserved between Drosophila and mammals (Piccirillo et al., 2014). Advantages of the IOM system are not only its genetic tractability, but also its reproducibility and structure. As a relatively giant single cell along the body wall, IOMs enable tracking of a single cell and its subcellular organization during metamorphosis. Our results show that studies in IOMs can provide new insights into the mechanisms of muscle remodeling as well as regulation of fundamental membrane trafficking pathways, such as autophagy and endocytosis.

Drosophila strains
Flies were reared at 25˚C, unless stated. For muscle-targeted gene expression, DMef2-GAL4 driver was used. UAS-LacZ was used as a control in RNAi experiments. Genotypes used in this study include the following: (1)

RNAi screens of IOM remodeling in Drosophila
In primary RNAi screen, RNAi inverted repeat (IR) hairpins (Supplementary file 1) were crossed to w; DMef2-GAL4, UAS-Dcr2 at 25˚C. Lethality was checked after two weeks from crossing. The adult progenies were incubated for 1 week at 25˚C. Around twenty flies in the vials were tapped to the bottom, and the speed at which they climbed of the vial was checked. To identify factors required for IOM remodeling, w; UAS-mCD8:GFP; DMef2-GAL4, UAS-Dcr2 flies were crossed with IR lines that induced a defect in eclosion or mobility in the primary screen. Pharate adults were removed from the pupal case and mounted without dissection. IOMs were observed from dorsal side through the cuticle for z-series imaging by confocal microscopy. At least ten IOMs in three animals were checked for each genotype.

Muscle preparations and immunofluorescence in Drosophila
Muscle preparations in pupal abdomens were performed as previously described (Ribeiro et al., 2011) with several modifications. Staged pupae were removed from the pupal case and pinned on a sylgard-covered petri dish in dissection buffer (5 mM HEPES, 128 mM NaCl, 2 mM KCl, 4 mM MgCl 2 , 36 mM sucrose, pH 7.2). Abdomens were opened with scissors, pinned flat, and fixed for 20 min. (4% formaldehyde, 50 mM EGTA, PBS). Then, the samples were unpinned and blocked for 30 min (0.3% bovine serum albuminum (BSA), 2% goat serum, 0.6% Triton, PBS), incubated with primary antibody overnight at 4˚C, washed (0.1% Triton PBS), then incubated for 2h at room temp with Alexa Fluor-conjugated secondary antibodies (Molecular Probes/Thermo Fisher Scientific, Waltham, MA) and counterstained with phalloidin for F-actin as needed. The stained samples were washed and mounted in SlowFade reagent (Life Technologies/Thermo Fisher Scientific, Waltham, MA).

Starvation and immunofluorescence in MEFs
For starvation, MEFs were quickly rinsed with PBS and incubated in EBSS for 1h (immunostaining) or 2h (LC3-II flux assay). For immunofluorescence, cells were cultured on coverslips, fixed with 4% PFA in PBS for 10 min, and permeabilized with 50 mg/ml digitonin in PBS for 5 min. Samples were then blocked with PBS containing 3% BSA for 15 min. Primary antibodies were diluted 1:500 to 1:1000, and Alexa Fluor-conjugated secondary antibodies (Invitrogen/Thermo Fisher Scientific, Waltham, MA) were diluted 1:1000 in PBS containing 3% BSA. Coverslips were incubated with primary antibodies for 1h, washed five times with PBS, and incubated with secondary antibodies for 45 min. Samples were mounted using ProLong Gold Antifade Mountant (Thermo Fisher Scientific, Waltham, MA).

Western blots
Equal amounts of proteins per sample were subjected to SDS-PAGE, and transferred to polyvinylidene difluoride membranes. The membranes were blocked with PBST (PBS and 0.1% Tween 20) containing 1% skim milk and were then incubated overnight at 4˚C with primary antibodies 1000-3000x diluted in the blocking solution. Membranes were washed three times with PBST, incubated for 1h at room temperature with 20,000x dilutions of HRP-conjugated secondary antibodies (GE Healthcare Life Sciences, Pittsbugh, PA) in the blocking solution, and washed five times with PBST. Immunoreactive bands were then detected using ECL plus (GE Healthcare Life Sciences, Pittsburgh, PA) and X-ray films.

Confocal fluorescence microscopy
For imaging of live pupal IOMs, staged pupae were removed from the pupal case, mounted between slide-glass and cover-glass following a protocol (Zitserman and Roegiers, 2011), and imaged from the dorsal side. Live or immunostained IOMs were acquired on a Zeiss LSM 700 microscope with a 10x air/0.45 NA Plan Apochromatic objective or 40x oil/1.3 NA Plan Apochromatic objective. Image acquisition software used was Zen (Carl Zeiss Microscopy, Germany). For imaging in MEFs, fixed and immunostained samples were observed on a FV1000D confocal microscope with a 60x oil/1.4 NA Plan Apochromatic objective. Image acquisition software used was Fluoview (Olympus Life Science, Japan). The images were adjusted using the Photoshop CS4 (Adobe, San Jose, CA) and ImageJ software (https://imagej.nih.gov/ij/).

Mosaic analysis in MEFs
To compare parental (wild-type) and Rab2A/B DKO MEFs side by side, parental MEFs and Rab2A/ B_DKO MEFs stably expressing EGFP were mixed and plated on coverslips. The resultant mosaic samples were immunostained for LC3, stained with Lysotracker Red (5000 times diluted) for 5 min or stained with Magic Red (300 times diluted) for 15 min.

Electron microscopy
For TEM of IOMs, staged pupae were removed from pupal cases, pinned on a sylgard-covered petri dish, dissected in fixative (2% paraformaldehyde, 2.5% glutaraldehyde, 150 mM sodium cacodylate, pH 7.4) and fixed for 1.5h at room temp. For TEM of fat body, staged larvae were treated as published (Jean et al., 2015), ripped open in fixative and incubated 25 min at room temp in an eppendorf tube. Fixative was replaced, samples fixed an additional 20 min, and then fat body tissue was dissected from the fixed larval carcass into glass vials. For TEM of MEFs, cells grown in complete medium were fixed in 2% glutaraldehyde (GA) in 0.1 M sodium cacodylate buffer (pH7.4) for overnight at 4˚C, gently scraped in the fixative and then pelleted down by centrifugation. Then, samples were post-fixed in 1% osmium tetroxide in 0.1 M cacodylate buffer and stained in 1% uranyl acetate for 1h. Abdomen fillets were embedded in epoxy resin and 70 nm sections were collected on Formvar and carbon-coated copper grids. Images were acquired on a transmission electron microscope (FEI Tecnai Spirit G2 BioTWIN, Hillsboro, OR) and photographed by a bottom mount Eagle 4 K digital camera (Sony, Japan).

Immunoelectron microscopy
MEFs stably expressing GFP:Rab2A were cultured in EBSS for 90 min and fixed in 4% paraformaldehyde in 0.1 M Sorensen's phosphate buffer pH 7.4 for overnight at 4˚C. Fixed cells were washed with 0.15 M glycine/phosphate buffer, embedded in 10% gelatin/phosphate buffer and infused with 2.3 M sucrose/phosphate buffer overnight at 4˚C. One mm 3 cell bocks were mounted onto specimen holders and snap frozen in liquid nitrogen. Ultracryomicrotomy was carried out at -100˚C on a Leica Ultracut UCT with EM FCS cryoattachment (Leica, Bannockburn, IL) using a Diatome diamond knife (Diatome US, Hatfield, PA). Eighty to ninety nm frozen sections were picked up with a 1:1 mixture of 2.3 M sucrose and 2% methyl cellulose (15cp) as described by Liou et al. (1996), and transferred onto Formvar and carbon-coated copper grids. Immunolabeling was performed by a slight modification of the 'Tokuyasu technique ' Tokuyasu, 1980. Briefly, grids were placed on 2% gelatin at 37˚C for 20 min, rinsed with 0.15 M glycine/PBS and the sections were blocked using 1% cold water fishskin gelatin. Ultrathin cryosections were labeled with mouse anti-GFP (1:25) and rabbit anti-LC3 (1:50) in 1% BSA/PBS for 2h at room temperature, followed by 1h incubation with 12 nm gold conjugated goat anti-mouse IgG and IgM (Jackson ImmunoResearch, West Grove, PA) and 18 nm gold conjugated goat anti-rabbit IgG (both 1:20). Sections were contrasted 10 min in 0.4% uranyl acetate and 1.8% methylcellulose on ice then imaged using a JEOL 1200EX II transmission electron microscope (JEOL, Peabody, MA) and photographed using a Gatan digital camera (Gatan, Pleasanton, CA).

Live imaging
Live-cell fluorescence imaging was performed using DeltaVison Elite (GE Healthcare Life Sciences, Pittsburgh, PA) equipped with a 60x PlanAPO oil-immersion objective lens (Olympus Life Science, Japan, NA 1.42) and a cooled-CCD camera (Photometrics, Tucson, AZ, CoolSNAP HQ2). Cells stably expressing marker proteins fused with GFP variants were placed on a glass bottom dish (Greiner Bio-One, Monroe, NC, 617870) 2 days before imaging. During live-cell imaging, the dish was mounted in a chamber (TOKAI HIT, Japan, INUB-ONI-F2) to maintain the incubation conditions at 37˚C and 5% CO 2 . Images were acquired at intervals of 1 min. Time series of 16-bit images were converted into 8-bit Tiff images using ImageJ software (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997-2015; RRID:SCR_003070) and the LOCI bio-formats ImageJ plugin.

EGFR degradation assay
MEFs were serum-starved for overnight and incubated in serum-free DMEM containing 100 ng/ml EGF and 25 mg/ml cycloheximide for the indicated times (Jean et al., 2015). Cells were lysed and subjected to SDS-PAGE and immunoblotting. Three independent experiments were performed and quantified.

Statistics
Each experiment was performed at least three times as biological and technical replicates (at least three different cohorts of unique flies or cell culture populations were analyzed in repeat procedures performed on at least three different days). One exception was for TEM analyses, which were performed on two parallel replicates with multiple animals each. All replicate experiments were performed in parallel with wildtype controls. RNAi phenotypes for the 'Rab2 class' were fully penetrant. Data shown is representative and inclusive of all results for procedures with properly performing experimental controls. Pearson correlation measurements for co-localization quantification were performed with ImageJ software (https://imagej.nih.gov/ij/) for Figure 1D, Figure 4B and Figure 4D. For each combination, ten 30 Â 30 mm cropped images taken from different animals were used. Quantification of autophagic flux with mCherry:GFP:Atg8a was performed as previously described (Takáts et al., 2013). Quantification of the number of autophagosomes, autolysosomes or mitochondria in TEM images was performed as follows: double-membrane bound structures with undigested cytoplasmic contents (autophagosomes), single membrane bound structures with electron dense cytoplasmic contents (autolysosomes) or mitochondria were quantified manually and normalized to the area of section as shown in Figure 4F, Figure 5C, Figure 7G and Figure 7-figure supplement 1F. For each genotype, ten images taken from multiple animals were used. GFP:Atg8 puncta in IOMs were quantified manually and then normalized to the IOM area for Figure 6B and D. To measure T-tubule area in Figure 6D, we manually set a threshold where the dimmest tubules were visible and plasma membrane and other brighter structures were excluded. More than 10 images were used for the quantification. LC3 puncta quantification in Figure 7C was performed as previously described (Itoh et al., 2011) using ImageJ. Thirty cells were quantified for each condition. Statistical analysis was performed using a two-tailed unpaired t test. Colocalization of puncta was counted manually in Figure 8F and H and Figure 8-figure supplement 1, only taking into account overlapping structures with a similar shape in relevant fluorescent channels. Ten images were quantified for each combination. p-values<0.05 were considered statistically significant.