Coordination of autophagosome–lysosome fusion and transport by a Klp98A–Rab14 complex in Drosophila

Macroautophagy (hereafter ‘autophagy’), is a multistage process capable of mediating the destruction of diverse cellular structures, including organelles, through lysosomal degradation. As a result, autophagy is crucial to the maintenance of homeostasis in living cells, preventing accumulation of harmful materials and allowing generation of vital macromolecules. Autophagy begins by encapsulating bulk cytosol in a double-membraned vesicle known as the autophagosome, which initiates at multiple sites throughout the cell, including endoplasmic reticulum (ER)–mitochondria junctions, the ER–Golgi intermediate compartment (ERGIC) and the plasma membrane (Shibutani and Yoshimori, 2014). Once autophagosomes are fully formed, their subsequent fusion with the lysosome forms the acidified single-membraned autolysosome, in which lysosomal hydrolases promote cargo degradation.

How the trafficking of autophagosomes and lysosomes is regulated as they fuse to form an autolysosome is poorly understood. Analysis of lysosomal movement has revealed a role for microtubule motor proteins in their intracellular trafficking (Rai et al., 2013; Santama et al., 1998). Cytoplasmic acidification has also been shown to influence lysosomal positioning in cultured cells (Heuser, 1989; Korolchuk et al., 2011), and movement of autophagic vesicles is required for their acidification during axonal transport (Maday et al., 2012). The minus-end-directed dynein–dynactin motor complex is required for the centripetal movement of autophagosomes, endosomes and lysosomes (Burkhardt et al., 1997; Johansson et al., 2007; Kimura et al., 2008). In contrast, most kinesins are plus-end-directed motor proteins, which drive their cargo towards the cell periphery, often playing a role in recycling at the plasma membrane (Hoepfner et al., 2005). KIF5b, a kinesin 1 family member, has previously been shown to influence autophagosome positioning in cancer cells (Cardoso et al., 2009), and KIF2, a kinesin 13 family member, has been suggested to modulate lysosome positioning in response to nutrients (Korolchuk et al., 2011). Kinesin-3 family members are unique in harboring a C-terminal pleckstrin homology (PH) or Phox homology (PX) motif, which binds to phosphoinositide lipids (Hoepfner et al., 2005; Miki et al., 2005), essential elements along the autophagic cascade for initiation, membrane elongation and lysosomal reformation (Dall'Armi et al., 2013; Knaevelsrud et al., 2013; Sridhar et al., 2013).

The Rab family of small GTPases are crucial for vesicular trafficking, often compared to ‘zip codes’ capable of identifying vesicle type. As such, their role in autophagy has been investigated, and various Rabs, including Rab5 and Rab7, have been implicated in the regulation of autophagy (Bento et al., 2013; Stenmark, 2009). Movement of vesicles to their appropriate destinations relies on interaction between Rabs and motor proteins (Delevoye et al., 2014; Janssens et al., 2014; Lee et al., 2015).

Here, we identify the Drosophila kinesin 3 family member Klp98A as a new protein involved in autophagy. We demonstrate that Klp98A regulates the positioning of autophagic vesicles and lysosomes, and independently affects autophagosome formation, autophagosome–lysosome fusion and cargo degradation. We show that Klp98A physically interacts with the autophagic-vesicle-associated protein Atg8a and with Rab14, which is also necessary for autophagosome growth, positioning and fusion. We find that Klp98A acts upstream of Rab14 by controlling its intracellular distribution, thus identifying Klp98A as a new Rab14 effector that coordinates the formation and maturation of autophagosomes with their intracellular positioning to promote functional autophagy.

To screen for phosphoinositide-binding proteins with potential roles in autophagy, we expressed a collection of RNA interference (RNAi) transgenes in clones of larval fat body cells under autophagy-inducing starvation conditions. In this screen, we identified Klp98A, the Drosophila ortholog of human KIF16B (Fig. 1A). To study the function of Klp98A in autophagy, we examined the effects of Klp98A loss- and gain-of-function on mCherry-tagged Atg8a protein (mCh–Atg8a), a well-characterized marker of autophagosomes and autolysosomes (Klionsky et al., 2012; Mauvezin et al., 2014). Klp98A depletion in fat body cells led to a reduction in the overall number of vesicles marked by mCh–Atg8a, and to their redistribution towards the perinuclear region of the cell (Fig. 1B; Fig. S1A,B). Conversely, overexpression of Klp98A skewed the distribution of mCh–Atg8a punctae towards the cell periphery, and these vesicles also tended to be reduced in size (Fig. 1C). Thus, Klp98A plays an essential role in the intracellular positioning of autophagic vesicles, as well as in their formation and/or growth.

To determine which steps of autophagy are regulated by Klp98A, we examined the intracellular localization of three vesicle populations in cells co-expressing mCh–Atg8a and the late endosome and lysosome marker LAMP1–GFP (autophagosomes are only positive for mCh-Atg8a; late endosomes and lysosomes are only positive for LAMP1–GFP; and autolysosomes are positive for both LAMP1–GFP and mCh–Atg8a). Each of these vesicle populations was markedly redistributed towards the cell center in response to Klp98A depletion, and towards the cell periphery in response to Klp98A overexpression (Fig. 2A; Fig. S1C,D). To ask whether these effects on vesicle intracellular localization are specific to Klp98A, we depleted kinesin heavy chain (Khc), a ubiquitously expressed motor. Loss of Khc led to a similar perinuclear localization of mCh–Atg8a vesicles, consistent with previous studies showing that kinesins play an important role in autophagic vesicle transport (Cardoso et al., 2009; Korolchuk et al., 2011) (Fig. 2A; Fig. S1E). Overexpressed Khc–GFP colocalized with a subpopulation of mCh–Atg8a vesicles near the cell periphery, but did not appreciably affect their overall distribution (Fig. S1F). We quantified these effects by measuring the relative distance of each vesicle population from the nucleus, taking into account the average cell size in each independent image. In both Klp98A- and Khc-depleted cells, autophagosomes and autolysosomes were significantly shifted towards the cell center compared to control (Fig. 2B; Fig. S1B). In addition, depletion of Klp98A, but not of Khc, similarly affected the positioning of late endosomes and lysosomes (Fig. 2B).

The overall direction of vesicle transport is determined by the orientation of microtubules and their associated motor proteins. As polarization of the microtubule network in fat body cells is not well defined, we examined the localization of Khc-nod–LacZ, a minus-end microtubule reporter (Bitan et al., 2010). This marker specifically decorated the central perinuclear area, indicating that microtubules are arranged into a well-organized network in these cells, with the minus ends proximal and positive ends distal to the nucleus, similar to in fibroblasts and other mesoderm-derived cells in mammals (Fig. 2C). Accordingly, the distribution of autophagic vesicles shifted towards the cell periphery upon depletion of Dhc64C, the heavy chain subunit of the minus-end-directed cytoplasmic dynein motor complex (Fig. S1G). Our results thus suggest that autophagic vesicles move along the microtubule network through both minus-end- and plus-end-directed motors.

As Klp98A and Khc had similar effects on autophagic vesicle positioning, we performed reciprocal rescue experiments to determine whether these proteins perform shared or unique functions. Expression of Klp98A or Khc fully restored normal localization of mCh–Atg8a vesicles in Khc- or Klp98A-depleted cells, respectively, suggesting that these motor proteins share a common function in promoting vesicle positioning (Fig. 2D–G). Taken together, these results indicate that Klp98A regulates intracellular transport of autophagic vesicles, a role that is shared with other kinesin family members.

Previous studies have shown that microtubule- and dynein-dependent transport of autophagosomes and lysosomes affects not only their intracellular location but also their fusion and activity (Batlevi et al., 2010; Jahreiss et al., 2008; Köchl et al., 2006; Maday et al., 2012), although similar requirements for kinesin-related proteins have not yet been reported. We therefore characterized the functionality of the redistributed autophagic vesicles resulting from Klp98A depletion. To monitor autophagosome–lysosome fusion, we analyzed the colocalization and quantified the Pearson's correlation coefficient (Rr) between mCh–Atg8a and LAMP1–GFP. In control cells, the majority of LAMP1–GFP- and mCh–Atg8-positive vesicles colocalized, with a correlation coefficient of 0.64, consistent with normal autophagosome–lysosome fusion (Fig. 3A,B). Depletion of Klp98A caused a substantial decrease in colocalization (Rr=0.37), similar to the effect of depletion of Syntaxin17, a recently described Q-SNARE protein specifically required for autophagosome–lysosome fusion (Itakura et al., 2012; Takats et al., 2013) (Fig. 3A,B). Similar results were observed with a second dsRNA targeting a distinct region of the Klp98A transcript (Fig. S2A). In contrast, colocalization of LAMP1–GFP and mCh–Atg8a was normal in Khc-depleted cells (Fig. 3A,B), indicating that the autophagosome transport defect due to kinesin motor protein depletion does not necessarily impact on vesicle fusion. The significant differences between the Khc- and Klp98A-knockdown phenotypes suggest that the role of Klp98A in autophagosome–lysosome fusion is at least partially independent of its effects on vesicle localization.

Fusion of autophagosomes with lysosomes is required for the acidification of autophagic vesicles and the subsequent degradation of their cargo. To ask whether the fusion defects in Klp98A-depleted cells interfered with these processes, we monitored vesicle pH using a tandem-tagged GFP–mCherry–Atg8a construct (Kimura et al., 2007; Nezis et al., 2010) and quantified the green:red fluorescence intensity ratio (FI_G/R) in these cells. Both the GFP and mCherry fluorophores of this reporter are visible in the neutral lumen of mature autophagosomes, whereas the GFP signal is quenched in acidic autolysosomes upon autophagosome–lysosome fusion. In starved control cells, the majority of Atg8a-positive vesicles were labeled strongly with mCherry but not GFP, with a FI_G/R of 0.13 (Fig. 3C,D). As a positive control, depletion of the lysosomal proton pump V-ATPase subunit V1H led to strong retention of the GFP signal (FI_G/R=0.95; Fig. 3D; Fig. S2B). Klp98A depletion also resulted in a significant increase of the green:red ratio, consistent with a moderate defect in vesicle acidity (FI_G/R=0.36; Fig. 3C,D and Fig. S2B). Taken together, these results indicate that Klp98A is required for normal fusion between autophagosomes and late endosomes or lysosomes.

To ask whether the vesicle fusion defect observed in Klp98A-depleted cells ultimately affects the overall rates of autophagic degradation, we examined the levels of GFP-tagged Ref(2)P (a p62 ortholog), a selective autophagy substrate whose levels are highly sensitive to autophagic flux (Chang and Neufeld, 2009; Lindmo et al., 2008; Nezis et al., 2008; Pircs et al., 2012). In control cells, GFP–Ref(2)P protein levels rapidly decreased during starvation-induced autophagy. In contrast, the rate of GFP–Ref(2)P turnover was markedly lower upon Klp98A depletion, consistent with the observed defects in autophagic vesicle fusion and acidification (Fig. 3E). In mammalian cells, amino acid starvation leads to an initial drop in the activity of mechanistic target of rapamycin (mTOR), followed by its reactivation in response to nutrients generated through autophagic degradation (Yu et al., 2011). Interestingly, similar mTOR reactivation occurred in control fat body cells but not in Klp98A-depleted cells, consistent with decreased autophagic flux (Fig. 3F). Depletion of Klp98A also led to a significant reduction in the size of starved cells, consistent with a decrease in mTOR activity (Fig. S2C). Taken together, our results demonstrate that loss of Klp98A in fat body cells induces a perinuclear accumulation of defective autophagic vesicles, leading to a reduction of autophagic flux.

In addition to its effects on the distribution, fusion and function of autophagic vesicles, Klp98A depletion also resulted in a marked reduction in the overall number of these vesicles in starved fat body cells (Fig. 1B). To analyze this phenotype, we quantified the size and number of autophagosomes and autolysosomes in control and Klp98A-depleted cells. Syx17 was included in this analysis to control for the effects of Klp98A depletion on autophagosome–lysosome fusion. In control cells, induction of autophagy by amino acid starvation led to an increase in the numbers of both autophagosomes and autolysosomes per cell, as well as an increase in autolysosome size (Fig. 4A,B,D). As expected, depletion of Syx17 reduced the number and size of autolysosomes, and increased the number and size of autophagosomes, consistent with a block of autophagosome–lysosome fusion and a concomitant accumulation and homotypic fusion of autophagosomes (Fig. 4A,D). Depletion of Klp98A led to a similar reduction in the number and size of autolysosomes, further confirming a requirement for Klp98A in autophagosome–lysosome fusion (Fig. 4B,D). However, in contrast to control and Syx17-depleted cells, the number of autophagosomes was not increased by starvation in Klp98A-depleted cells (Fig. 4A), and Klp98A depletion did not lead to an increase in autophagosome size as observed in cells depleted for Syx17 (Fig. 4C). These effects were specific to Klp98A, as depletion of Khc did not lead to significant changes in the size or number of autophagic vesicles (Fig. S3A,B). Depletion of Klp98A also decreased the number of vesicles marked by the early autophagosome marker Atg5–GFP compared to wild-type and Syx17-depleted cells (Fig. S3C). This reduction in autophagosome number in Klp98A-depleted cells cannot be explained by the defects induced in later stages of autophagy upon loss of Klp98A (e.g. autophagosome–lysosome fusion or degradation or TOR reactivation, Fig. 3), each of which promote accumulation of autophagosomes, as observed in cells depleted for Syx17. Rather, these results suggest that Klp98A has additional functions in earlier stages of autophagosome formation and growth, and this might provide an explanation for previous observations of reduced autophagy in cells treated with microtubule inhibitors (Köchl et al., 2006). Taken together, our results indicate that Klp98A plays a multifaceted role in autophagy, directly or indirectly controlling the formation, fusion and intracellular positioning of autophagic vesicles.

Rab GTPases are well-known regulators of endosomal fusion and trafficking (Jordens et al., 2005; Wandinger-Ness and Zerial, 2014). Among their cellular functions, they regulate the association of specific motor proteins with vesicles trafficking to and from the endocytic pathway. To initiate a mechanistic characterization of Klp98A, we took note of previous observations showing that association of human KIF16B with its cargo can be modulated by direct interaction with the small GTPase Rab14 (Ueno et al., 2011). We confirmed a physical interaction between Rab14 and Klp98A in Drosophila S2 cells by co-immunoprecipitation (Fig. 5A). To examine the specificity of this interaction, we tested whether other Rabs also interact with Klp98A. Both Rab10 and Rab11 failed to readily co-immunoprecipitate with Klp98A whereas Rab3, which forms a complex with the other kinesin-3 family members (KIF1Bβ and KIF1A) in mammalian neurons (Niwa et al., 2008; Okada et al., 1995), did readily co-immunoprecipitate with Klp98A (Fig. S3D–F).

Mammalian Rab14 interacts with the C-terminal region of KIF16B, which contains the PX domain (Ueno et al., 2011). To determine whether the PX domain of Klp98A mediates its interaction with Rab14, we asked whether deletion of this domain could disrupt the interaction between Rab14 and Klp98A. Surprisingly, Klp98A^ΔPX could co-immunoprecipitate Rab14 as readily as Klp98A (Fig. 5A). Taken together, these results suggest that the region immediately upstream of the PX domain might be important for the KL98A–Rab14 interaction.

Consistent with the described role of Rab14 in the endocytic pathway (Junutula et al., 2004), we observed clear colocalization of Rab14 with the late endosomal markers LAMP1 and Rab7 under both fed and starved conditions (Fig. 5B; Fig. S4A). Notably, Rab14 also colocalized with a subset of Atg8a-positive vesicles during starvation-induced autophagy (Fig. 5C). Mutation of Drosophila Rab14 has recently been reported to cause perinuclear accumulation of GFP–LAMP1 punctae (Garg and Wu, 2014), similar to the effects of Klp98A depletion shown here. We confirmed this effect on the late endosome and lysosome population in Rab14-depleted cells, as well as a modest but significant shift of autophagosomes towards the cell periphery (Fig. 5D).

To ask whether these effects of Rab14 depletion on vesicle localization were accompanied by defects in fusion, we examined the colocalization of mCh–Atg8a and LAMP1–GFP. Vesicles labeled with these markers showed almost no overlap in Rab14-depleted cells, with a correlation coefficient similar to the effects of Klp98A depletion (Figs 3B, 5E,F). We further examined autophagosome–lysosome fusion in Rab14-depleted cells using the acidic compartment dye LysoTracker Red, which in fat body cells specifically labels autophagy-induced acidic vesicles (Mauvezin et al., 2014; Scott et al., 2004). Consistent with a defect in autolysosome formation, strong punctate LysoTracker Red labeling was observed in starved control cells but not in Rab14-depleted cells (Fig. S4B). In addition, protein levels of Ref(2)P were increased in Rab14-null cells (Fig. S4C), indicating a defect in autophagic degradation and a reduction of autophagic flux that mimic the effects of Klp98A depletion. Finally, similar to what was observed in cells lacking Klp98A, the size of autophagic vesicles was significantly reduced upon depletion of Rab14, whereas Rab14 overexpression led to a modest increase in vesicle size (Fig. 5G,H). Taken together, Rab14 and Klp98A have strikingly similar effects on the growth, positioning and function of autophagic and endocytic vesicles.

The shared phenotypes and physical association of Klp98A and Rab14 led us to examine the intracellular localization of these proteins. Rab14–GFP displayed extensive colocalization with UAS-driven HA-tagged Klp98A, as well as with a Myc-tagged Klp98A protein driven by the ubiquitin promoter that did not affect the distribution of autophagic vesicles (Fig. 6A; Fig. S4D, white arrowheads). Rab14 and Klp98A proteins also colocalized together with a subpopulation of Atg8a-positive vesicles (Fig. 6A; Fig. S4D, yellow arrowheads). In addition, a population of smaller vesicles that were only positive for Rab14 could often be observed juxtaposed to Atg8a- and Klp98A-marked vesicles (Fig. 6A). Given the partial colocalization of Rab14 and Klp98A with Atg8a, and the effects of Rab14 and Klp98A depletion on the distribution of Atg8a-marked vesicles, we examined whether Klp98A and Rab14 can directly interact with Atg8a. Using a GST pulldown assay, we found that GST–Atg8a readily pulled down Flag–Klp98A, whereas no interaction between GST–Atg8a and Rab14–GFP was detected (Fig. 6B). Deletion of the PX domain of Klp98A did not disrupt its interaction with GST–Atg8a (Fig. 6B), suggesting that association of Klp98A with phosphoinositide-positive membranes is not required for this interaction.

Taken together, these results are consistent with a model in which Rab14 acts to recruit Klp98A-associated autophagosomes to the endo-lysosomal compartment; alternatively, Rab14 might function as a cargo whose mobility is dependent upon the motor activity of Klp98A. To distinguish between these possibilities, we examined the intracellular localization of each protein upon depletion of the other. Whereas loss of Rab14 had no effect on the overall distribution of Klp98A, depletion of Klp98A led to a perinuclear accumulation of Rab14-positive vesicles, and Klp98A overexpression resulted in a redistribution of Rab14–GFP to the cell periphery (Fig. 6C,D; Fig. S4E). Taken together, these results support the idea that Rab14 is a Klp98A cargo and acts downstream of Klp98A.

The intracellular distribution and transport of autophagosomes remains poorly understood. How motor proteins bind and regulate autophagosomal function while achieving route specificity during their transport in cells is uncharacterized. The crucial role of molecular motors in regulating cellular pathways like autophagy is evident in disease states linked to their dysfunction, including a wide variety of kidney, neurodegenerative and developmental diseases (Crimella et al., 2012; Hirokawa et al., 2010; Mandelkow and Mandelkow, 2002; Ravikumar et al., 2005). In this study, we begin to uncover the mechanisms driving autophagosome transport in vivo, using the powerful genetic techniques available in Drosophila. We identified Klp98A as a key protein in the regulation of autophagosome formation, positioning and maturation.

The Drosophila larval fat body is comprised of a monolayer of polyploid cells with little overt polarity. However, the relatively large size of these cells likely dictates the need for an organized cytoskeleton to support intracellular transport. Indeed, we show that the microtubule network is polarized in fat body cells, with their minus ends localized near the nuclear periphery. Within this context, our results demonstrate that the intracellular positioning of autophagic vesicles reflects a balance between inwards movement towards the nucleus driven by the cytoplasmic dynein complex and outwards movement driven by plus-end-directed kinesins. These findings are consistent with previous studies in mammalian cells showing that lysosomal positioning is normally biased towards the central microtubule-organizing center (MTOC), and that this bias is increased upon inactivation of specific kinesin motor proteins and decreased in response to kinesin overexpression (Korolchuk et al., 2011; Matsushita et al., 2004; Nakata and Hirokawa, 1995; Santama et al., 1998). Autophagosomes have also been proposed to move towards the MTOC to meet with the lysosomal compartment following their formation throughout the cell periphery (Jahreiss et al., 2008). We found that Klp98A and Khc depletion resulted in similar mislocalization of autophagic vesicles, implying that Klp98A affects vesicle localization by means of its kinesin motor domain. These individual phenotypes of Khc and Klp98A indicate that these motors are not redundant, yet they appear to share a common function, being able to substitute for each other upon overexpression. The total number of available motors is likely important for this function, and compensatory mechanisms might lead to a decrease of cargo specificity to avoid intracellular stress, as has been reported for specialized transport of mitochondria (Tanaka et al., 1998). Interestingly, Klp98A but not Khc also affected the localization of non-autophagic endosomal and lysosomal vesicles, consistent with the previously reported effects of the Klp98A human ortholog KIF16B on endosomal localization (Hoepfner et al., 2005). Thus, both shared and unique functions of kinesin motor proteins contribute to the orchestration of autophagic vesicle trafficking.

Klp98A and Khc also differ in their effects on the fusion of autophagosomes with lysosomes. These vesicles failed to fuse in Klp98A-depleted cells despite being clustered closely together around the nuclear periphery, whereas Khc depletion had a similar effect on the distribution of autophagic vesicles without disrupting their fusion. These results suggest that the role of Klp98A in fusion extends beyond promoting vesicle proximity. This might involve the transport and localization of components of the fusion machinery such as Syntaxin17 or VAMP7, or a more direct effect on membrane curvature or vesicle tubulation as recently demonstrated for the Klp98A human ortholog KIF16B (Skjeldal et al., 2012). The ability of Klp98A to associate with both Atg8a and Rab14 also points to a potential non-transport function of Klp98A in promoting fusion, by mediating the interaction of autophagic and endocytic vesicles. Consistent with this, Klp98A was required for normal recruitment or localization of Rab14, and depletion of Rab14 mimicked the effects of Klp98A on autophagosome localization and fusion. In mammals, KIF16B interacts with Rab14 only in its GTP-bound state, indicating that this interaction is tightly regulated (Ueno et al., 2011).

The presence of PX or PH lipid interaction domains is a unique feature of most kinesin-3 family members, and how these domains affect kinesin regulation or function is unclear. Our finding that Klp98A interacts with Rab14 independently of its PX domain is consistent with previous findings showing that the kinesin-3 proteins KIF1A and KIF1Bβ associate with Rab3 through a stalk region distinct from their PH domain, despite the essential role of this domain for vesicle transport (Klopfenstein and Vale, 2004; Niwa et al., 2008). The effects of Klp98A are also reminiscent of another PX domain containing protein, FYCO1, which promotes the peripheral movement of autophagosomes and endosomes through association with the Atg8 orthologs of the LC3 family (also known as MAP1LC3), the late endosomal protein Rab7, and Kinesin-1 (Bitan et al., 2010). Interestingly, FYCO1 has also recently been shown to promote fusion of endosomes with the plasma membrane (Raiborg et al., 2015). In addition, the PX protein Snx18, as well as other phosphoinositide-binding proteins, are required for autophagy formation and growth, similar to the early function of Klp98A (Knaevelsrud et al., 2013). Thus, the multifaceted roles of Klp98A in autophagy likely reflect its combination of kinesin motor, phosphoinositide-binding, and Atg8- and Rab14-interaction domains. Our finding that these domains can function independently of one another suggests that Klp98A might act as a hub for multiple regulatory signals, and potentially as a therapeutic target.

Flies were raised at 25°C on standard cornmeal, molasses and agar medium. All RNAi strains as well as UAS-Khc-GFP, UAS-dAtg5-GFP, Ubi-p63E-Klp98A-myc, UAS-Khc-nod-LacZ and UAS-YFP-Rab14 flies were obtained from the Bloomington Drosophila Stock Center (Bloomington, IN) or Vienna Drosophila RNAi Center (Vienna, Austria). The following transgenes were used in this study to deplete the gene of interest: Klp98A RNAi: TRiP HMS01957 (RNAi 1), P{GD11724}v40603 (RNAi 2), Rab14 RNAi: TRiP JF03135, Khc RNAi: P{GD12278}v44337, Syx17 RNAi: TRiP JF01937, Dhc64C RNAi: P{GD12258}v28054 and V1H RNAi: P{GD8795}v47471. The Rab14^null mutant and UAS-Rab14RB–mRFP flies were a kind gift of Louisa Wu (University of Maryland, College Park, MD) and have been described previously (Garg and Wu, 2014). UAS-Klp98A-3xHA was obtained from the FlyORF stock center (Bischof et al., 2013), (University of Zurich, Zurich, Switzerland). Additional UAS lines used in this study were UAS-LAMP1-GFP (Pulipparacharuvil et al., 2005), UAS-mChAtg8a (Arsham and Neufeld, 2009) and UAS-GFP-Ref(2)P (Chang and Neufeld, 2009). UAS-driven transgenes were expressed in the fat body using Cg-GAL4 (Hennig et al., 2006) or were used for clonal analysis as described previously (Arsham and Neufeld, 2009).

Larvae (72 h) were moved to fresh food. After 24 h, the larvae were kept in food (fed condition) or starved for 4 h in 20% sucrose (Macron Fine Chemicals, Center Valley, PA) in 1× PBS. The larvae were bisected, inverted and fixed in 3.7% formaldehyde at 4°C overnight with rocking. After three washes of 15 min each in PBS with 0.1% Triton X-100 (PBS-T) and counterstaining with 1 µg/ml DAPI, the fat bodies were dissected out and mounted in Vectashield (Vectorlabs, Burlingame, CA). All larval samples were imaged at room temperature on a confocal microscope (Zeiss LSM710) equipped with a 40× (W) objective lens (APO DIC III NA 1.2). RGB and grayscale images were further processed with ImageJ (NIH, Bethesda, MD) or Photoshop CS3 and assembled into figures using Illustrator CS3 (Adobe, San Jose, CA). We restricted the acquisition to one unique image per fat body tissue on the nucleus (DAPI) focal plane or above the nucleus (distal plane). Laser lines used in this study were 405, 488, 561 and 647 nm. Images were acquired using Zeiss software Zen 2010. Images were collected with the on-board photomultiplier tubes of the 710 system.

Live images of fat body cells stained with LysoTracker Red (Invitrogen, Grand Island, NY) were obtained on a Zeiss Axioscope-2 microscope after incubation for 3 min at room temperature in LysoTracker Red working solution (1:1000) (Juhasz and Neufeld, 2008).

Indirect immunofluorescence for Klp98A–HA of fixed fat body cells [3.7% paraformaldehyde (PFA)] was performed on bisected third-instar larvae. Carcasses were washed for 2 h in PBS, permeabilized for 15 min in PBTX-DOC (PBS with 0.1% Triton X-100 and 0.05% sodium deoxycholate) and blocked for 3 h in 3% goat serum in PBTX-DOC. Incubation with primary antibody rat anti-HA (1:250; catalog number 11867423001, Roche) in 1% goat serum in PBTX-DOC was performed overnight at 4°C with rocking. After three washes of 30 min each in PBTX-DOC, samples were incubated with secondary antibody goat anti-rat-IgG conjugated to Alexa Fluor 647 (1:500; catalog number 467, Cell Signaling Technology) in 1% goat serum in PBTX-DOC for 4 h at room temperature. After three 15-min washes in PBTX-DOC and one 10-min wash in PBS, fat bodies were mounted with Vectashield (Vectorlabs, Burlingame, CA). For the microtubule polarity assay, antibodies were used as follows: rabbit anti-β-Gal (1:500; catalog number 55976, Cappel), mouse anti-α-tubulin DM1A (1:2000; catalog number T6199, Sigma-Aldrich), anti-mouse-IgG A488 (1:500; catalog number 115-545-166, Jackson ImmunoResearch Laboratories), and anti-rabbit-IgG conjugated to A594 (1:500; catalog number 111-585-144, Jackson ImmunoResearch Laboratories).

Nuclear staining was eliminated using Photoshop CS3 to restrain the analysis to only the cytosol. Red and green channels were split in ImageJ and the specific correlation coefficient between cytosolic red and green spots was measured using ICA Plugin (Li et al., 2004; Mauvezin et al., 2010). Pearson's correlation coefficient was quantified from those images (n>15 animals). Student's t-tests were performed between relevant data sets.

Each sample was scanned using the same laser settings at the LSM710 Zeiss microscope. Identification of autophagic structures was performed using a fixed threshold for cytoplasmic red-only spots in ImageJ and a binary image was saved for each sample analyzed. Red and green channels from the original image were then merged with a red-only binary mask. Selection of autophagosomes in the mask image allowed the measurement of the average fluorescence intensity in the original green and red channels separately. The ratio between red and green was calculated from a minimum of 7 animals. Statistical analysis was performed in Excel using Student’s t-test.

Late endosomes and lysosomes (LAMP1–GFP-positive vesicles only), autophagosomes (mCh–Atg8a-positive vesicles only) and autolysosomes (vesicles labeled with LAMP1–GFP and mCh–Atg8a only) were isolated from original merged RGB images using Photoshop CS3. Masked images were then analyzed in MATLAB (MathWorks, Natick, MA) using a custom-written code (available upon request). This program did the following to each image. First, images in all color channels were band-pass filtered to subtract out inconsistencies in low-frequency background fluorescence, and to average out high-frequency noise by creating Gaussian filters using the built-in MATLAB function fspecial. Images were then thresholded and converted into black-and-white images using the built-in MATLAB functions graythresh and im2bw. Next, all objects were automatically detected using the built-in MATLAB functions edge and then bwlabel. The centroid of each object was recorded using the built-in MATLAB function regionprops. This analysis was performed in all vesicle populations, such that the nucleus centroids were localized in the blue channel, and then centroids from late endosome and lysosome vesicles (vesicles only positive for LAMP1–GFP) were localized in the green channel, and centroids of the autophagosomes (vesicles only positive for mCh–Atg8a) were localized in the red channel. Note that autolysosome (vesicles labeled with LAMP1–GFP and mCh–Atg8a) detection could be assessed with either the red or the green channel code applied to the mask generated previously in Photoshop CS3. Once all of the centroid positions were recorded for each vesicle population, the distance between the nucleus centroid in the blue channel and all of the centroids in each of identified vesicles was calculated, such that a list was generated of all nuclei to each specific vesicles centroid distances. It was then necessary to assign the objects in each channel to the appropriate nucleus in the blue channel. This was accomplished by assigning a maximum radial distance from the nucleus centroid, and then by using the built-in MATLAB function lt to remove distances longer than the assigned maximum cell radius. If a given spot was within the maximum radius distance for more than one nucleus, the shortest distance was found, and this was used to assign each spot to only one nucleus. Thus, automated object detection was used to generate a list of distances from each nucleus to its associated late endosome or lysosome vesicle, autophagosome or autolysosome.

Average distances for each vesicle population were then corrected relative to cell size in each image, for each genotype. To this end, the average cell area and average nuclear area were determined using Photoshop CS3. Average cell and nuclear radius were calculated assuming perfect circularity. The average nuclear radius was subtracted from each vesicle-to-centroid distance, and the resulting number was normalized to the difference between nuclear and cellular radius. For quantification of reciprocal rescue of Khc and Klp98A to the other in GFP-positive clones, the distance from the centroid of the nucleus to the centroid of the vesicle was measured and corrected by the radius of the cell at that specific location. Average distance from nucleus, standard deviation, standard error, and two-sided t-tests assuming unequal variance calculations were performed in Excel.

Number and size of the three different populations of autophagic vesicles were assessed using the Adobe Photoshop CS3 measurement analysis option. Similar analysis was performed for measuring the size of RNAi-expressing clonal cell compared to surrounding control cells.

To prepare protein lysate, fat bodies from five carcasses were dissected in PBS, lysed in SDS buffer, and boiled for 3 min at 95°C. Lysates were separated by PAGE, and transferred to Immobilon-P membranes (Millipore, Billerica, MA). The membrane was incubated with the primary antibody overnight at 4°C, washed with 1× PBS containing 0.1% Tween 20 (Fisher Scientific), incubated with secondary antibody for 2 h at room temperature, and washed with 1× PBS containing 0.1% Tween 20. SuperSignalWest Pico chemiluminescent substrate (Thermo Scientific, Rockford, IL) was used for visualization of signals on the membrane with HyBlot CL autoradiography film (Denville Scientific, Metuchen, NJ). Quantification was performed using Adobe Photoshop CS3. Antibodies used were rabbit anti-phospho-T398 dS6K (1:250; catalog number 9209, Cell Signaling Technology), rabbit anti-GFP (1:30,000; catalog number A-11122, Molecular Probes), mouse anti-β-tubulin (1:1000; catalog number E7, Developmental Studies Hybridoma Bank, Iowa City, IA) and rabbit anti-Ref(2)P (1:15,000; Pircs et al., 2012).

pUASTattB-3× Flag was generated by PCR-amplifying the 3× Flag sequence from pAFW (Drosophila Genome Resource Center, Bloomington, IN) to incorporate 5′ EcoRI and 3′NotI sites, and then subcloning this product into pUASTattB (Bischof et al., 2007) digested with EcoRI and NotI. The resulting clone was verified by sequence analysis.

Klp98A was subcloned from cDNA clone LD29123 (Drosophila Genome Resource Center, Bloomington, IN) with a KpnI/SpeI restriction digest and ligated into the pUASTattB 3× Flag vector digested with KpnI/XbaI. The N-terminal portion of Klp98A was PCR amplified with a 5′ primer incorporating a NotI site and 3′ primer containing an internal BglII site. The pUAST attB 3× Flag plasmid containing Klp98A and the PCR product were digested with NotI and BglII and ligated to make full-length Klp98A in frame with the 3× Flag tag. The resulting clone was verified by sequence analysis.

Klp98A^ΔPX was generated by site-directed mutagenesis of the cDNA clone LD29123 to engineer a stop codon directly before the PX domain. The resulting clone was sequence verified. This clone was then digested with BglII/SpeI and the pUAST 3× Flag KLP98A clone digested with EcoRI/BglII and both fragments were ligated into pUASTattB digested with EcoRI/XbaI to create the pUAST 3× Flag Klp98A^ΔPX construct.

Rab14 was PCR amplified from mRFP-Rab14-RB adult genomic fly DNA (Garg and Wu, 2014) with primers to incorporate a 5′ BamHI site and 3′ XhoI site. eGFP was PCR amplified with primers to incorporate 5′ EcoRI and 3′ BamHI sites. The Rab14 PCR product was digested with BamHI/XhoI and the eGFP PCR product was digested with EcoRI/BamHI and both were subcloned into the pUASTattB vector cut with EcoRI/XhoI. The resulting clone was verified by sequence analysis.

Rab3 was PCR amplified from clone LP05860 (Drosophila Genome Resource Center, Bloomington, IN) with primers to incorporate 5′ KpnI and 3′ XhoI sites. eGFP was PCR amplified to incorporate 5′ EcoRI and 3′ KpnI sites. The Rab3 PCR product was digested with KpnI/XhoI and the eGFP PCR product was digested with EcoRI/KpnI and both were subcloned into pUASTattB digested with EcoRI/XhoI. The resulting clone was sequence verified.

pAGW Rab10 (Actin-promoter, N-terminal GFP-tagged Rab10) and pAGW Rab11 (Actin-promoter, N-terminal GFP-tagged Rab11) were kind gifts from Sally Horne-Badovinac, University of Chicago, IL.

Atg8 was PCR amplified from pUAST-mCh-Atg8 adult fly genomic DNA to incorporate 5′ BamHI and 3′XhoI sites. This PCR product was then subcloned into pGEX-KText vector (Guan and Dixon, 1991) and digested with the same enzymes. The resulting clone was sequence verified. GST and GST–Atg8 (pGEX-KText and pGEX-KText-Atg8) proteins were expressed and purified from BL21 cells in STE buffer containing 0.2 mg/ml lysozyme, CellLytic B (Sigma-Aldrich), and EDTA-free complete protease inhibitor cocktail (Roche), and bound to glutathione–Sepharose-4B beads (Amersham, Pittsburgh, PA).

8.0×10⁶ S2 cells were transfected with Actin-Gal4 and UAS-eGFP-Rab14, UAS-eGFP-Rab3, pAGW-Rab10, or pAGW-Rab11 with or without the UAS-3× Flag-Klp98A or the UAS-3× Flag-Klp98A^ΔPX construct using dimethydioctadecyl-ammonium bromide (DDAB) at 250 µg/ml (Sigma-Aldrich). Flag immunoprecipitations were performed using Flag M2 agarose beads (Sigma-Aldrich) 2 days post transfection. Cells were harvested and lysed in a buffer containing 50 mM HEPES, 150 mM NaCl, 0.5 mM EGTA, 0.9 M glycerol, 0.1% Triton X-100, 0.5 mM DTT and EDTA-free complete protease inhibitor cocktail (Roche). Immunoprecipitation reactions were carried out at 4°C for 2 h. For immunoblotting, 7.5% SDS-PAGE gels were transferred onto nitrocellulose. Antibodies were used at the following concentrations: rabbit anti-GFP (1:2000; catalog number A-11122, Molecular Probes, Grand Island, NY) and mouse anti-Flag M2 (1:20,000; catalog number F3165, Sigma-Aldrich). Fluorescently labeled IRdye800CW (LI-COR Biosciences) and IR680RD (LI-COR Biosciences) secondary antibodies were used at 1:5000.

Similar procedures were carried out for GST pulldown as for immunoprecipitations. S2 cells were transfected with Actin-Gal4 and UAS-eGFP-Rab14, UAS-3× Flag-Klp98A, or UAS-3× Flag-Klp98A^ΔPX constructs. At 2 days post transfection cells were lysed in the buffer used for immunoprecipitations with 1% Triton X-100. Lysates were divided in half and put onto GST- or GST–Atg8-bound beads. Binding reactions were performed at 4°C for 2 h. For immunoblotting, 10% SDS-PAGE gels were transferred onto nitrocellulose. Antibodies were used at the following concentrations: rabbit anti-GFP (1:2000; catalog number A-11122, Molecular Probes), mouse anti-Flag M2 (1:20,000; catalog number F3165, Sigma-Aldrich), and goat anti-GST (1:1000; Pharmacia). The same secondary antibodies were used at the concentration given described in the immunoprecipitation subsection.

Images of the blots from both immunoprecipitation and pulldown assays were obtained using LI-COR Image Studio software, version 3.1 (LI-COR Biosciences, Lincoln, NE).

We would like to thank Dr Michael O'Connor and Aidan Peterson for access and support for confocal microscopy. We are grateful to Sally Horne-Badovinac for providing the Rab10 and Rab11 constructs. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537), Drosophila Genetic Resource Center (Kyoto, Japan), FlyORF (Zurich Switzerland) and Vienna Drosophila Resource Center (VDRC) were used in this study.