The Vici Syndrome Protein EPG5 Is a Rab7 Effector That Determines the Fusion Speciﬁcity of Autophagosomes with Late Endosomes/Lysosomes

Article

The Vici Syndrome Protein EPG5 Is a Rab7 Effector that Determines the Fusion Speciﬁcity of Autophagosomes with Late Endosomes/Lysosomes

Graphical Abstract Authors Zheng Wang, Guangyan Miao, Xue Xue, ..., Du Feng, Junjie Hu, Hong Zhang

Correspondence [email protected]

In Brief Wang et al. demonstrate that the Vici syndrome protein EPG5 acts as a tethering factor that determines the fusion speciﬁcity of autophagosomes with late endosomes/lysosomes. EPG5 stabilizes and facilitates assembly of the trans-SNARE complex that mediates autophagosome maturation.

Highlights d EPG5 is a Rab7 effector that mediates fusion of autophagosomes with late endosomes d EPG5 facilitates the assembly of trans-SNARE complexes for autophagosome fusion d EPG5 is recruited to late endosomes/lysosomes by binding to Rab7 and VAMP7/8 d EPG5 interacts with LC3/LGG-1 and assembled autophagosomal SNARE Qabc complexes

Wang et al., 2016, Molecular Cell 63, 781–795 September 1, 2016 ª 2016 Elsevier Inc. http://dx.doi.org/10.1016/j.molcel.2016.08.021 Molecular Cell Article

The Vici Syndrome Protein EPG5 Is a Rab7 Effector that Determines the Fusion Speciﬁcity of Autophagosomes with Late Endosomes/Lysosomes

Zheng Wang,1,6 Guangyan Miao,1,2,6 Xue Xue,1,6 Xiangyang Guo,3 Chongzhen Yuan,1 Zhaoyu Wang,1 Gangming Zhang,1 Yingyu Chen,2 Du Feng,4 Junjie Hu,1 and Hong Zhang1,5,* 1National Laboratory of Biomacromolecules, CAS Center for Excellence in Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, PRC 2Department of Immunology, Peking University School of Basic Medical Science, Beijing 100191, PRC 3College of Life Sciences, Nankai University, Tianjin 300071, PRC 4Institute of Neurology, Afﬁliated Hospital of Guangdong Medical College, Zhanjiang 524001, PRC 5College of Life Sciences, University of Chinese Academy of Sciences, Beijing 100049, PRC 6Co-ﬁrst author *Correspondence: [email protected] http://dx.doi.org/10.1016/j.molcel.2016.08.021

SUMMARY organelles (Stenmark, 2009; Mizuno-Yamasaki et al., 2012). For example, Rab5 presents on early endosomes, while Rab7 lo- Mutations in the human autophagy gene EPG5 cause cates on late endosomes. Tethering factors, which are generally the multisystem disorder Vici syndrome. Here we recruited to cognate membranes by Rab GTPases, function in demonstrated that EPG5 is a Rab7 effector that de- the initial attachment of transport vesicles to their target mem- termines the fusion specificity of autophagosomes branes for fusion (Bro¨ cker et al., 2010; Yu and Hughson, with late endosomes/lysosomes. EPG5 is recruited 2010). The tethering factor p115 is essential for fusion of COPII to late endosomes/lysosomes by direct interaction vesicles with the Golgi apparatus, while the multisubunit HOPS tethering complex is required for several fusion events in late with Rab7 and the late endosomal/lysosomal endosomes and the vacuole (Bro¨ cker et al., 2010; Yu and Hugh- R-SNARE VAMP7/8. EPG5 also binds to LC3/LGG-1 son, 2010). The membrane fusion process is driven by trans-as- C. elegans (mammalian and Atg8 homolog, respec- sembly of cognate sets of membrane-anchored SNARE pro- tively) and to assembled STX17-SNAP29 Qabc teins on the vesicle and target membranes. The core SNARE SNARE complexes on autophagosomes. EPG5 complex, consisting of a bundle of four a helices, is formed by stabilizes and facilitates the assembly of STX17- the 60-residue SNARE motif in each SNARE protein (Jahn SNAP29-VAMP7/8 trans-SNARE complexes, and and Scheller, 2006; Su¨ dhof and Rothman, 2009). The assembly promotes STX17-SNAP29-VAMP7-mediated fusion of cognate SNARE complexes under physiological conditions is of reconstituted proteoliposomes. Loss of EPG5 ac- facilitated by tethering factors (Cai et al., 2007; Yu and Hugh- tivity causes abnormal fusion of autophagosomes son, 2010). with various endocytic vesicles, in part due to Autophagy is an evolutionarily conserved lysosome-mediated degradation process (Nakatogawa et al., 2009; Feng elevated assembly of STX17-SNAP25-VAMP8 com- SNAP25 et al., 2014). It involves the formation of a crescent-shaped plexes. knockdown partially suppresses phagophore (also known as the isolation membrane, IM), which EPG5 the autophagy defect caused by depletion. expands and closes to form the double-membrane auto- Our study reveals that EPG5 is a Rab7 effector phagosome. The autophagosome is then delivered to the involved in autophagosome maturation, providing lysosomes/vacuole for degradation of the sequestrated mate- insight into the molecular mechanism underlying rials. Genetic screens in yeast and worm have identified a Vici syndrome. group of autophagy proteins that act at distinct steps of autophagosome formation (Feng et al., 2014; Lu et al., 2013). The Atg1/Atg13 Ser/Thr kinase complex and the Vps34/Atg14 INTRODUCTION PI(3)P kinase complex are involved in the induction and initia- tion of the IM (Nakatogawa et al., 2009; Feng et al., 2014). Membrane-enclosed organelles are intricately connected via The ubiquitin-like protein Atg8, which is conjugated to phos- vesicle-mediated membrane trafficking (Bonifacino and Glick, phatidylethanolamine (PE), associates with outer and inner 2004). The specificity and efficiency of membrane trafficking de- membranes of autophagic structures (Nakatogawa et al., pends on the concerted action of Rab GTPases, tethers, and 2009; Feng et al., 2014). The membrane-associated Atg8-PE the SNARE complex (Cai et al., 2007; Wickner and Schekman, interacts with a variety of factors at multiple steps in the auto- 2008). Different Rab GTPases associate with distinct membrane phagy pathway (Birgisdottir et al., 2013; McEwan et al., 2015).

Molecular Cell 63, 781–795, September 1, 2016 ª 2016 Elsevier Inc. 781 Figure 1. Loss of Function of epg-5/hEPG5 Causes Coalescence of Autophagosomes with Endocytic Vesicles (A–D) DIC images of the intestine of WT L4 larvae (A), epg-5(tm3425) L2 larvae (B), epg-5(tm3425) L4 larvae (C), and atg-2 mutants (D). epg-5 mutants have enlarged vesicles (arrows), which are visible in L1 larvae and are most abundant in L4 larvae. Small granules (arrowheads) in WT intestine are lysosome-related organelles (LROs) or lipid droplets (LDs). epg-5 mutant larvae still contain LROs and LDs, which are visible in other focal planes. (E–L) Enlarged vesicles (arrows) in epg-5 mutants are labeled by GFP::LGG-1 (E and F), 23 FYVE::GFP (G and H), RFP::RME-1 (I and J), and LAAT-1::GFP (K and L). epg-5 mutants carrying an RFP::RME-1 transgene have fewer enlarged vesicles (I and J). (E)–(L) are confocal images. (M–P) Localization of GFP::LGG-1 and Cherry::RAB-7 (M and N) and RFP::RME-1 (O and P) in WT and epg-5 mutant intestine. WT intestine contains a few small GFP::LGG-1 puncta that are distinct from Cherry::RAB-7 (M) or RFP::RME-1 (O) puncta. In epg-5 mutant intestine, GFP::LGG-1 puncta accumulate and partially co-localize with enlarged Cherry::RAB-7 (N) or RFP::RME-1(P) punctate structures. The focal planes in (N) and (P) are different from those in (C), (E), (G), and (K). Vacuoles are not evenly distributed in epg-5 mutant intestine. GFP::LGG-1 puncta are evident on focal planes with fewer vacuoles. (Q and R) In control cells, GFP-Rab5 puncta are distinct from LC3 puncta, detected by anti-LC3 (Q). In hEPG5 KD cells, enlarged GFP-Rab5 punctate structures partially co-localize or closely associate with LC3 puncta (R). (S and T) In control cells, EHD1-GFP labels tubular structures and also forms a few puncta (S). In hEPG5 KD cells, EHD1-GFP forms many more punctate structures that partially co-localize with LC3 puncta (T). (M)–(T) are confocal ﬂuorescent images. (U and V) Efﬁciency of hEPG5 siRNA KD. hEPG5 protein (U) and mRNA (V) levels are dramatically reduced in hEPG5 shRNA cells. The mRNA level, analyzed by real-time PCR, was normalized to that in control cells, which was set to 100. Data are shown as mean ± SEM (n = 3 independent experiments). *p < 0.05.

(legend continued on next page)

782 Molecular Cell 63, 781–795, September 1, 2016 In higher eukaryotes, nascent autophagosomes fuse with endo- RESULTS cytic vesicles to form amphisomes that further fuse with lysosomes to produce degradative autolysosomes, a process In epg-5 Mutant Worms, Autophagosomes Fuse known as autophagosome maturation (Lamb et al., 2013). Abnormally with Endocytic Vesicles Fusion of autophagosomes with late endosomes/lysosomes In contrast to wild-type (WT) and animals with mutations in is mediated by a SNARE complex consisting of Syntaxin 17 genes acting during autophagosome formation, epg-5 mutants (Qa SNARE) on autophagosomes, SNAP29 (Qbc SNARE), accumulated numerous enlarged vesicular structures in the and the endosomal/lysosomal R-SNARE VAMP7/8 (Itakura intestine, as judged by differential interference contrast (DIC) et al., 2012). This fusion process is regulated by several factors, microscopy (Figures 1A–1D, S1A, and S1B, available online). including the late endosomal/lysosomal-localized protein The number of enlarged vesicles increased as development PLEKHM1 (which directly interacts with LC3 [mammalian proceeded and peaked at the L4 larval stage (Figure S1A). Atg8 homolog], HOPS, and Rab7) and TECPR1 (which inter- The enlarged vesicles in epg-5 mutants were labeled by acts with PtdIns(3)P and the Atg12-Atg5 conjugate) (McEwan markers for autophagosomes (GFP::LGG-1), early endosomes et al., 2015; Chen et al., 2012). The HOPS complex (which is (GFP::RAB-5 and 2xFYVE::GFP), late endosomes (GFP::RAB- recruited to autophagosomes by Syntaxin 17) and the autopha- 7), and recycling endosomes (RFP::RME-1, an EH-domain-con- gosomal protein ATG14 (which interacts with STX17 and the taining ATPase associated with recycling endosomes) (Figures STX17-SNAP29 Qabc complex) also regulate autophagosome 1E–1J, S1C, and S1D). The lysosomal membrane protein maturation (Jiang et al., 2014; Diao et al., 2015). The role of LAAT-1::GFP labeled the vesicles (Figures 1K and 1L), while these factors in autophagosome maturation under physiolog- NUC-1::Cherry (lysosomal-localized DNaseII homolog) was de- ical conditions, however, has not been demonstrated. Auto- tected inside the vesicles (Figure S1E). The enlarged vesicles, phagy is intimately integrated with intracellular trafficking however, could not be stained by lysotracker (Figure S1F), (Lamb et al., 2013). Impairment of early or late endosome func- suggesting that they are not acidic. RME-1-labeled recycling tion results in accumulation of non-degradative autolysosomes endosomes showed both tubular and spherical structures in (Lamb et al., 2013; Eskelinen, 2005). The mechanisms that WT intestinal cells, while epg-5 mutants contained fewer tubular coordinate endocytic and autophagic processes are largely and more spherical RME-1-labeled structures (Figures S1G unknown. and S1H). In WT intestine, RFP::RAB-5-, Cherry::RAB-7-, and The metazoan-specific autophagy gene EPG5 was identified RFP::RME-1-labeled structures were separable from a few small by genetic screens in C. elegans. Loss of epg-5 function in GFP::LGG-1 puncta (Figures 1M, 1O, 1W, and S1I). As well as la- worms causes defective autophagic degradation of a variety of beling enlarged vesicular structures, several endosome markers protein aggregates during development (Tian et al., 2010). Unlike formed enlarged punctate structures in epg-5 mutant intestine. other genetically identified ATG and EPG genes, which act dur- GFP::LGG-1 puncta dramatically accumulated and partially co- ing autophagosome formation, EPG5 deficiency in C. elegans, localized with the enlarged punctate structures labeled by re- mice, and human blocks the maturation of autophagosomes porters for RAB-5, RAB-7, RME-1, and NUC-1 in epg-5 mutants into degradative autolysosomes (Tian et al., 2010; Zhao et al., (Figures 1N, 1P, 1W, and S1J–S1L) (we define two puncta as co- 2013a). Loss of Epg5 activity also impairs endosomal trafficking localized if their fluorescent signals overlap by >70%, and as (Zhao et al., 2013a). Epg5 knockdown (KD) slows endocytic associated if their fluorescent signals are in close proximity but degradation and delays endocytic recycling (Zhao et al., with a limited overlap). BFP::LGG-1 puncta were labeled by 2013a). Recent studies revealed that human EPG5 mutations both RFP::RME-1 and LAAT-1::GFP (Figures S1M and S1N). cause the multisystem disorder Vici syndrome (Cullup et al., These results indicate that autophagosomes and various endo- 2013). The molecular role of EPG5 in autophagy and endocytic somes coalesce in epg-5 mutant intestine, and EPG-5 may trafficking remains unknown. Here we demonstrated that EPG5 determine the fusion specificity of maturing autophagosomes. directly interacts with Rab7 and VAMP7/8 on late endosomes/ly- TECPR1 and PLEKHM1 participate in autophagosome matu- sosomes, and also binds to LC3/LGG-1 (C. elegans Atg8 homo- ration in mammalian cells (Chen et al., 2012; McEwan et al., log) and the assembled STX17-SNAP29 Qabc complex on auto- 2015). We examined whether their C. elegans homologs are phagosomes. EPG5 regulates the stability and the assembly of involved in autophagy. Putative null mutants of t01h3.2 and STX17-SNAP29-VAMP8 trans-SNARE complexes. Loss of func- y51h1a.2, encoding TECPR1 and PLEKHM1, respectively, tion of EPG5 causes non-specific fusion of autophagosomes were created using CRISPR/Cas9. In t01h3.2(bp1286) and with other endocytic vesicles, resulting in formation of non- y51h1a.2(bp1287) mutants, there was no accumulation of degradative enlarged vesicles. Our results demonstrate that SQST-1 aggregates or LGG-1 puncta in embryos or larvae (Fig- EPG5 promotes specific fusion of autophagosomes with late en- ures S2A–S2H; data not shown). Accumulation of SQST-1::GFP dosomes/lysosomes. in epg-5 mutants was not further exacerbated by loss of function

(W) Quantification of co-localization of indicated markers with LGG-1 puncta in WT and epg-5 mutant intestine. (The number of tiny GFP::LGG-1 puncta varies among animals. Puncta R1 mm in diameter were examined in five different animals.) (X) Quantification of co-localization of indicated markers with LC3 puncta in control and hEPG5 KD cells. Data are shown as mean ± SEM in (W) and (X). *p < 0.05, **p < 0.01, ***p < 0.001. Data were compared with two-tailed unpaired Student’s t tests. Scale bars, 10, 5 (inserts in M–P), and 2 mm (inserts in Q–T). See also Figures S1 and S2.

Molecular Cell 63, 781–795, September 1, 2016 783 of t01h3.2 and y51h1a.2 (Figures S2I–S2L). These results indi- from endosome/lysosome fusion (Figures 2G, 2H, and S3I– cate that homologs of TECPR1 and PLEKHM1 are not essential S3N). Together, these results show that hEPG5 is mainly local- for autophagy in C. elegans. ized on late endosomes/lysosomes.

Loss of Function of hEPG5 Causes Abnormal Fusion of EPG5 Directly Binds to Rab7 Autophagosomes with Endocytic Vesicles in Mammalian The subcellular localization of EPG5 prompted us to investigate Cells whether EPG5 interacts with Rab7 on late endosomes/lyso- We next investigated whether hEPG5 is also involved in control- somes. Recombinant full-length (FL) worm EPG-5 bound ling autophagosome fusion speciﬁcity in HeLa cells (epg-5 and strongly to RAB-7 in an in vitro pull-down assay, while neither EPG-5 refer to the C. elegans gene and protein, while hEPG5 RAB-5 nor RAB-11, which localize on early and sorting endo- and hEPG5 indicate the human gene and protein. EPG5 refers somes, respectively, showed strong binding to EPG-5 (Fig- to both C. elegans and human proteins.). In control cells, the ure 3A). EPG-5 bound to the constitutively active RAB-7(Q68L) few LC3 puncta were largely distinct from early endosomes mutant (GTP locked form, corresponding to Q67L in human labeled by Rab5 and EEA1 under replenished conditions (Figures Rab7), but not the dominant inhibitory RAB-7(T23N) mutant 1Q, 1X, and S1P). In hEPG5 KD cells, LC3 puncta were enlarged (GDP locked form, corresponding to T22N in human Rab7) and more abundant even under growth conditions (Figures 1R, (Figure 3B). An EPG-5 fragment containing amino acids 1–760, 1U, 1V, and S1Q). Rab5/EEA1-labeled vesicles were also EPG-5(1–760), but not EPG-5(691–1,599), bound to RAB-7 in a enlarged and partially co-localized with LC3 puncta in hEPG5 pull-down assay (Figures S4A and S4B). The minimal RAB-7- KD cells (Figures 1R, 1X, and S1Q). EEA1 puncta were separable binding region was delineated to EPG-5(500–530) (Figure S4C). from LAMP1-labeled late endosomes/lysosomes in control cells, Flag::EPG-5 was speciﬁcally co-precipitated by GFP::RAB-7 in while a few enlarged EEA1 puncta were LAMP1 positive in hEPG5 a GFP-TRAP assay (a co-immunoprecipitation [co-IP] assay, in

KD cells (Figures S1O, S1R, and S1S). The enlarged LC3 and which anti-GFP VHH is coupled to agarose beads for immuno- Rab5 puncta, however, were not stained by lysotracker, indi- precipitation of GFP-fusion proteins) using extracts of worms ex- cating that they are not acidic (Figures S1T and S1U). In control pressing Flag::EPG-5 and GFP::RAB-7 (Figure S4D). cells, EHD1-labeled recycling endosomes formed tubular struc- Endogenous hEPG5 was precipitated by endogenous Rab7 in tures and a few punctate structures that were separable from a co-IP assay (Figure 3C). GFP-hEPG5 also immunoprecipitated LC3 puncta (Figures 1S and 1X). In hEPG5 KD cells, tubular endogenous Rab7 in GFP-TRAP assays (Figure 3D). Endo- EHD1-labeled recycling endosomes were reduced, and there genous hEPG5 was precipitated by WT Rab7 and GFP- was dramatic accumulation of spherical structures that co-local- Rab7(Q67L), but not by GFP-Rab7(T22N) (Figure 3E). Levels of ized with LC3 puncta (Figures 1T and 1X). These results are Rab7(Q67L) precipitated by hEPG5 were much higher than consistent with the idea that epg-5/hEPG5 is required for specific Rab7(T22N) (Figure S4E). hEPG5(427–1094) bound to Rab7 in fusion of autophagosomes with endocytic vesicles. in vitro pull-down assays (Figure 3F). hEPG5(546–599), covering the minimal RAB-7-binding region in worm EPG-5, bound hEPG5 Localizes on Late Endosomes/Lysosomes strongly to Rab7 (Figures 3G, S4F, and S4G). Rab7 was not In addition to having a diffuse cytoplasmic localization, GFP- precipitated by hEPG5 with a deletion of amino acids 546–575, hEPG5 formed distinct puncta, mostly in the perinuclear region hEPG5(del 546–575), in a co-IP assay (Figure 3D). (Figure 2A). Staining with endogenous proteins or co-expression with reporters for endocytic vesicle markers revealed that GFP- Localization of hEPG5 to Late Endosomes/Lysosomes hEPG5 co-localized primarily with Rab7- and Rab9-labeled late Depends on Rab7 endosomes (Figures 2A, 2I, S3A, and S3B). Expression of GFP- We next investigated whether hEPG5 is recruited to late endo- hEPG5 did not change the distribution, morphology, and size of somes/lysosomes by Rab7. Co-transfection of Rab7(Q67L) re- late endosomes/lysosomes (Figures 2A and S3A). Some small sulted in formation of enlarged perinuclear GFP-hEPG5 vesicles GFP-hEPG5 puncta, located mainly in the peripheral region, that co-localized with Rab7 (Figure 3H). In cells co-expressing were negative for RFP-Rab7 (Figures 2A and S3A). Structured Rab7(T22N), or treated with Rab7 small interfering RNA (siRNA), illumination microscopy (SIM) revealed that hEPG5 overlapped the diffuse GFP-hEPG5 signal was stronger (Figures 3I, S4H, and with Rab7 on vesicular membranes (Figure 2B). GFP-hEPG5 S4K). A few small cytoplasmic GFP-hEPG5 puncta co-localized also co-localized with LAMP1-labeled late endosomes/lyso- with LAMP1-labeled vesicles (Figures S4J and S4K). The pattern somes (Figures 2C, 2I, S3C, and S4J). Most GFP-hEPG5 puncta of WT and mutant Rab7 was unchanged by co-expression of co-localized with lysotracker-stained acidic vesicles (Figures 2D, GFP-hEPG5 (Figures 3H, 3I, S3A, and S4I). 2I, and S3D). Some hEPG5+Rab7+ vesicles were lysotracker To examine whether hEPG5 is dynamically associated with negative (Figures 2E and 2J). GFP-hEPG5 puncta were largely late endosomes/lysosomes, we performed fluorescence recov- distinct from Rab5/EEA1/2xFYVE-labeled early endosomes ery after photobleaching (FRAP) experiments in HeLa cells and EHD1/TfnR-labeled recycling endosomes (Figures 2F, 2I, expressing GFP-hEPG5. The fluorescence signal from GFP- and S3E–S3H). hEPG5-labeled vesicles recovered 5 s after photobleaching Immunogold electron microscopy analysis of cells transfected in cells co-expressing Rab7 or Rab7(Q67L) (Figures S4L and with GFP-hEPG5 revealed that gold particles recognizing GFP- S4M), but reappeared much more slowly in cells co-expressing hEPG5 were localized on the membranes of organelles that Rab7(T22N) (Figure S4N), suggesting that hEPG5 is dynamically resembled late endosomes and hybrid organelles resulting recruited to late endosomes/lysosomes by Rab7.

784 Molecular Cell 63, 781–795, September 1, 2016 Figure 2. EPG5 Localizes on Late Endosomes/Lysosomes (A) GFP-hEPG5 puncta co-localize extensively with endogenous Rab7 in HeLa cells. A few GFP-hEPG5+Rab7À puncta are present in the peripheral region. (B and C) 3D-SIM images showing that GFP-hEPG5 co-localizes with RFP-Rab7 (B) and LAMP1-mCherry (C) on vesicles (arrows). Images are maximum intensity projections of z stacks for Rab7 (z = 48) and LAMP1 (z = 36). (D) GFP-hEPG5 puncta largely co-localize with lysosomes stained by lysotracker. (E) Localization of GFP-hEPG5, RFP-Rab7, and lysotracker. Some hEPG5+Rab7+ puncta are not stained by lysotracker (arrow). (F) GFP-hEPG5 puncta are distinct from EEA1-labeled early endosomes, detected by anti-EEA1. (G and H) Electron micrographs showing that immunogold particles recognizing hEPG5 are localized on the membrane of late endosomes, but not of mito- chondria (black arrow) and nuclei (white arrows). (H) shows enlargements of the boxed areas in (G). Ultrathin sections of HeLa cells transfected with GFP-hEPG5 were immunolabeled with an anti-GFP antibody followed by gold-conjugated secondary antibody (12 nm). (I) Quantification of co-localization of vesicles labeled by the indicated marker with hEPG5 puncta. (J) Percentage of hEPG5+Rab7+ puncta that are also stained by lysotracker. Each red dot represents data from a single cell. Data are shown as mean ± SEM in (I) and (J). Scale bars, 10 mm (A–F), 2 mm (inserts in A–F), 500 nm (G), 200 nm (H), and 50 nm (inserts in H). See also Figure S3. hEPG5 Localizes with Amphisomes/Autolysosomes mCherry-hEPG5 puncta (Figures 4F, 4G, and S5I). Co-local- upon Autophagy Induction ization of GFP-RFP-LC3 with hEPG5 was also determined. GFP We next examined the localization of hEPG5 under autophagy in- fluorescence is quenched in acidic compartments, so yellow duction conditions. hEPG5 was distinct from IMs labeled by vesicles (GFP+RFP+) represent IMs and immature autophago- WIPI2, ATG16L-GFP, and DFCP1-GFP (Figures 4A, S5A–S5C, somes, while red ones (GFPÀRFP+) are acidified mature autopha- and S5U). GFP-hEPG5 puncta partially co-localized with LC3 af- gosomes (likely amphisomes) as well as autolysosomes. Both ter starvation (Figures 4B, 4E, and S5D–S5G). SIM revealed that yellow and red LC3 puncta co-localized with hEPG5 after LC3 signals were present inside, present on the surface of, or starvation (Figures 4H and S5J), indicating that acidification closely associated with GFP-hEPG5-positive vesicles (Figure 4C). is not required for co-localization of autophagosomes with The stage of the autophagic structures that co-localized with hEPG5 vesicles. Co-localization of LC3 with hEPG5 puncta may hEPG5 was further determined. STX17 is reported to be present result from translocation of hEPG5 to autophagosomes or fusion on closed autophagosomes (Itakura et al., 2012). GFP-STX17+ of autophagosomes with hEPG5-labeled late endosomes/lyso- LC3+ puncta co-localized with mCherry-hEPG5 structures, while somes. Translocation of hEPG5 to autophagosomes would GFP-STX17ÀLC3+ puncta, which largely represent early autopha- result in accumulation of autophagosomes labeled by LC3 gic structures at early starvation stages, did not co-localize with and GFP-hEPG5, but not by LAMP1. Most LC3+hEPG5+ puncta

Molecular Cell 63, 781–795, September 1, 2016 785 Figure 3. EPG5 Binds to Rab7, and Localization of EPG5 on Late Endosomes/Lysosomes Depends on Rab7 (A) Worm FL EPG-5 is pulled down by GST-RAB-7, weakly by GST-RAB-5, but not by GST-RAB-11, in an in vitro pull-down assay. (B) Constitutively active RAB-7(Q68L), but not the dominant-negative RAB-7(T23N), pulls down EPG-5 in a GST pull-down assay. (C) Endogenous hEPG5 is precipitated by endogenous Rab7 in a co-IP assay. Extracts were treated with Rab7 antibody and the precipitants were immunoblotted with anti-hEPG5. (D) hEPG5 with a deletion of aa 546–575 fails to precipitate endogenous Rab7 in GFP-TRAP assays. (E) In GFP-TRAP assays, endogenous hEPG5 is precipitated by GFP-tagged WT Rab7 and Rab7(Q67L), but not by Rab7(T22N). Extracts of cells expressing WT and mutant GFP-Rab7 were precipitated by GFP-TRAP beads and immunoblotted with anti-hEPG5. Approximately 4% of extracts used for GFP-TRAP assays are shown as the input. (F and G) Human hEPG5(427–1094) (F) and hEPG5(546–599) (G) bind to Rab7 in GST pull-down assays. (H) Co-transfection of Rab7(Q67L) causes formation of enlarged, perinuclear hEPG5 puncta. (I) In cells co-transfected with RFP-Rab7(T22N), GFP-hEPG5 is more diffuse and forms fewer, less intense puncta. Scale bars, 10 mm. See also Figure S4. were positive for LAMP1 after starvation (Figures 4I and 4J). Co- teracted strongly with LGG-1 in a pull-down assay (Figure 4O). localization of GFP-hEPG5 with LC3 puncta was decreased after EPG-5(500–530), which corresponds to hEPG5(546–575), also RNAi KD of STX17 or SNAP29 (Figures 4D, 4E, S5H, S5S, and bound strongly to LGG-1, while EPG-5(del 500–530) did not S5T). These results suggested that hEPG5 is not recruited to au- bind to LGG-1 (Figures 4P and 5C). tophagosomes before fusion with late endosomes/lysosomes. The LIR Motifs in EPG-5/hEPG5 Mediate LGG-1/LC3 EPG5 Binds to LGG-1/LC3 Binding

The above results suggest that EPG5 participates in autophago- LC3 interacts with proteins via the sequence [W/F/Y]0x+1x+2[I/L/ some maturation. We investigated whether hEPG5 directly V]+3, known as the LC3 interacting motif (LIR) (Birgisdottir et al., interacts with LC3. Co-IP assays revealed that Myc-hEPG5 spe- 2013). hEPG5(546–575) contains two putative LIRs, 550WTLV ciﬁcally precipitated endogenous LC3, and endogenous LC3 and 567WILL (Figure 5A). In in vitro pull-down and co-IP assays, precipitated endogenous hEPG5 (Figures 4K and 4L). In in vitro LC3 binding to hEPG5(427–1094) and FL hEPG5 was reduced by pull-down assays, hEPG5(427–1,094) bound to LC3, and the the W550A or W567A mutation, and abolished by the W550 interacting region was further delineated to hEPG5(546–575) W567 double mutation (Figures 5B and S5X). EPG-5(500–530) (Figures 4M and S5V). hEPG5(del 546–575) did not immunopre- contains two typical LIR motifs, 507WEIL and 523FVTI (Figures cipitate LC3 in co-IP assays (Figure 4N). Flag::EPG-5 was specif- 5A, S5Y, and S5Z). EPG-5(W507A) and EPG-5(F523A) impaired, ically immunoprecipitated by GFP::LGG-1 from worm extracts while EPG-5(W507A F523A) dramatically reduced, LGG-1 bind- in GFP-TRAP assays (Figure S5W). Recombinant FL EPG-5 in- ing (Figure 5C).

786 Molecular Cell 63, 781–795, September 1, 2016 Figure 4. Localization of EPG5 on Amphisomes/Autolysosomes (A) One hour after starvation, GFP-hEPG5 puncta are distinct from early autophagic structures, detected by anti-WIPI2. (B) Localization of GFP-hEPG5 and LC3 puncta 4 hr after starvation. (C) 3D-SIM images of GFP-hEPG5 and LC3 are maximum-intensity projections of z stacks (z = 40). LC3 signal is located inside and on the rim of GFP-hEPG5- positive vesicles (arrows). (D) STX17 KD disrupts the co-localization of GFP-hEPG5 with LC3 puncta 4 hr after starvation. (E) Percentage of hEPG5 puncta positive for LC3 at different times after starvation in control, STX17 KD, and SNAP29 KD cells. Data are shown as mean ± SEM. *p < 0.05. (F and G) Closed autophagosomes (GFP-STX17+LC3+) co-localize with mCherry-hEPG5 (arrow), while unclosed autophagosomes (GFP-STX17ÀLC3+) rarely co- localize with mCherry-hEPG5 (F). Quantification of STX17+LC3+ or STX17ÀLC3+ puncta that are also positive for hEPG5 2 hr after starvation is shown in (G). Each red dot represents data from one cell. The few STX17ÀLC3+hEPG5+ puncta may be autolysosomes. Data are shown as mean ± SEM. (H) hEPG5 co-localizes with both acidified (GFPÀRFP+, amphisomes and autolysosomes) and neutral (GFP+RFP+, IMs and immature autophagosomes) LC3 puncta (arrow). Cells transfected with Myc-hEPG5 and GFP-mRFP-LC3 were serum starved for 1 hr before analysis. (I and J) GFP-hEPG5+LC3+ puncta are also labeled by LAMP1-mCherry (arrow) 2 hr after starvation (I). Quantification of the percentage of hEPG5+LC3+ puncta positive for LAMP1 is shown in (J). Each dot represents data from one cell. Data are shown as mean ± SEM. (K) In a co-IP assay, Myc-hEPG5 precipitates endogenous LC3 (predominantly the lipidated form, LC3-II). Cells were starved for 2 hr before analysis. (L) Endogenous LC3 precipitates endogenous hEPG5 in a co-IP assay. Cells were starved for 2 hr. (M) hEPG5(546–575) binds strongly to LC3 in a GST pull-down assay. (N) Mutant hEPG5 with a deletion of aa 546–575 does not precipitate LC3 in a co-IP assay. (O and P) Worm FL EPG-5 (O) and EPG-5(500–530) (P) are strongly pulled down by GST-LGG-1 in in vitro pull-down assays. Scale bars, 10 and 2 mm (inserts). See also Figure S5.

Molecular Cell 63, 781–795, September 1, 2016 787 Figure 5. The LIR Motifs in EPG-5/hEPG5 Mediate LGG-1/LC3 Binding (A) Sequence alignment of the LC3 binding region in EPG5. The aromatic residues in the two LIR motifs are indicated by an asterisk. Conserved hydrophilic residues are in gray, conserved hydrophobic residues are in cyan, and conserved negatively charged residues are in pink. (B) Binding of hEPG5(427–1094) to LC3 is reduced by the W550A or W567A mutation, and abolished by the W550A W567A double mutation. (C) In a GST pull-down assay, EPG-5(del 500–530) and EPG-5(W507A F523A) are not pulled down by GST-LGG-1. EPG-5(W507A) and EPG-5(F523A) show reduced binding afﬁnity to LGG-1. (D) EPG-5(del 500–530), EPG-5(W507A), and EPG-5(W507A F523A) are not pulled down by RAB-7. EPG-5(F523A) shows reduced interaction with RAB-7. (E–H) Localization of GFP-hEPG5 mutants in HeLa cells co-transfected with RFP-Rab7(Q67L). GFP-hEPG5(del 546–575) (E), GFP-hEPG5(W550A W567A) (F), and GFP-hEPG5(W567A) (H) are mainly diffuse in the cytoplasm, while GFP-hEPG5(W550A) (G) forms some small puncta. (I–L) LGG-1 puncta, detected by anti-LGG-1, are absent in WT embryos at the 4-fold stage (I and J), but dramatically accumulate in epg-5 mutant embryos at the same stage (K and L). (M–T) Accumulation of LGG-1 puncta in epg-5 mutant embryos is rescued by a transgene expressing WT EPG-5 (N), but not mutant EPG-5(del 500–530) (P) or EPG-5(W507A) (R). The EPG-5(F523A) transgene partially rescues the defect in epg-5 mutants (T). (I), (K), (M), (O), (Q), and (S) show DAPI images of the embryos in (J), (L), (N), (P), (R), and (T), respectively. Scale bars, 10 and 2 mm (inserts). The size of C. elegans embryos remains constant, so the scale bar is only shown once.

EPG-5(500–530) binds to LGG-1 as well as RAB-7. EPG-5(del motifs is also involved in LC3/LGG-1 binding (Wu et al., 2015). 500–530) did not bind to RAB-7 (Figure 5D). Binding of RAB-7 L510A mutation also impaired interactions of EPG-5 with LGG-1 to EPG-5(F523A) was reduced, while binding to EPG-5(W507A) and RAB-7 (Figure S5A1), suggesting that the EPG-5 LIR motif and EPG-5(W507A F523A) was even more severely affected (Fig- is involved in RAB-7 binding. The two corresponding aromatic ure 5D). The hydrophobic residue (I/L/V) at the ‘‘+3’’ position in LIR amino acids in hEPG5(546–575) are also important for mediating

788 Molecular Cell 63, 781–795, September 1, 2016 the hEPG5/Rab7 interaction. In co-IP assays, binding to Rab7 between endogenous hEPG5 and STX5-GFP or STX10-GFP was almost abolished by hEPG5(W550A W567A) and was was detected in co-IP assays (Figures S6L and S6M). hEPG5(del impaired by hEPG5(W550A) and hEPG5(W567A) (Figure S5X). 546–575) and hEPG5(W550A W567A) still immunoprecipitated Consistent with the critical role of amino acids 546–575 and endogenous STX17, SNAP29, and VAMP8 (Figures S6J and the two aromatic residues W550 and W567 in hEPG5 binding S6N). In vitro pull-down assays showed that hEPG5(427–1094) to Rab7, GFP-hEPG5(del 546–575), GFP-hEPG5(W567A), and bound very weakly to STX17, SNAP29, and VAMP7/8 (Fig- GFP-hEPG5(W550A W567A) were diffusely localized in the cyto- ure S6O). However, hEPG5(427–1094) bound strongly to pre- plasm, while GFP-hEPG5(W550A) was largely diffuse apart from a assembled STX17-SNAP29 and STX17-SNAP29-VAMP8 com- few small puncta (Figures 5E–5H). plexes (Figures 6I and 6J). hEPG5(427–1094) also bound to We next performed rescue experiments to examine whether pre-assembled SNARE domains of STX17-SNAP29-VAMP7/8 binding of EPG-5 to LGG-1/RAB-7 is critical for EPG-5 function. (Figures S6P and S6Q). Together, these results suggest that LGG-1 puncta are present in early C. elegans embryos, but are EPG-5/hEPG5 binds to assembled Qabc and QabcR SNARE largely absent in late embryos from the comma stage onward complexes that are required for fusion of autophagosomes (Figures 5I and 5J) (Tian et al., 2010). LGG-1 puncta accumulate with late endosomes/lysosomes. and persist in late epg-5 embryos (Figures 5K and 5L) (Tian et al., VAMP8 and VAMP7 localize on late endosomes/lysosomes 2010). The autophagy substrate SQST-1::GFP also greatly accu- (Figures 6K and S6R). GFP-VAMP7 also formed some puncta mulates in epg-5 mutants (Figures S5K–S5N) (Tian et al., 2010). at the cell periphery (Figure S6R). hEPG5 puncta co-localized Accumulation of LGG-1 puncta and SQST-1 aggregates in epg-5 extensively with VAMP8 and VAMP7 (Figures 6K and S6R). mutant embryos was rescued by a transgene expressing WT Mutant VAMP7/8 lacking the membrane-anchoring motif EPG-5 (Figures 5M, 5N, and S5O), but not EPG-5(del 500–530) localized mainly to the peripheral cytoplasm and the plasma or EPG-5(W507A) (Figures 5O–5R, S5P, and S5Q). EPG- membrane (Figures 6L and S6S). In cells expressing mutant 5(F523A) retained partial rescuing activity (Figures 5S, 5T, and VAMP7/8, the diffuse cytoplasmic GFP-hEPG5 signal was S5R). W507 and F523 in EPG-5 are essential for binding to stronger, and the dispersed GFP-hEPG5 puncta were smaller RAB-7 and LGG-1. These results suggest that binding to LGG- and fewer (Figures 6L and S6S), suggesting that the interaction 1 and/or RAB-7 is essential for the role of epg-5 in autophagy. with VAMP7/8 contributes to the retention of hEPG5 on late endosomes/lysosomes. EPG5 Binds to STX17-SNAP29 Qabc and Also to the STX17-SNAP29-VAMP8 QabcR SNARE Complex hEPG5 Promotes the Assembly of FL worm SNAP-29, STX-17, and VAMP-7, or the corresponding STX17-SNAP29-VAMP8 Complexes SNARE domains, assembled into 1:1:1 complexes in an in vitro Levels of endogenous STX17, SNAP29, and VAMP7/8 were co-purification experiment (Figures S6A–S6C). Flag::SNAP-29 much lower in hEPG5 KD cells than controls (Figure 6M), indi- was co-immunoprecipitated by GFP::STX-17 and GFP::VAMP-7 cating that hEPG5 stabilizes STX17, SNAP29, and VAMP7/8. from worm extracts in co-IP assays (Figures S6D and S6E). The formation of STX17-SNAP29-VAMP8 SNARE complexes Flag::EPG-5 was specifically co-precipitated by GFP::STX-17 was also examined in hEPG5 KD cells. Levels of endogenous and GFP::VAMP-7 in GFP-TRAP assays of extracts from animals SNAP29 co-immunoprecipitated by STX17 did not change in expressing tagged EPG-5 and SNARE proteins (Figures S6F and hEPG5 KD cells (levels of precipitated SNAP29 were normalized S6G). We further investigated whether EPG-5 interacts with indi- to the corresponding STX17 in control and hEPG5 KD cells, vidual Syntaxins and/or with pre-assembled SNARE complexes respectively) (Figure 6N). However, much less endogenous in in vitro pull-down assays. FL recombinant EPG-5 bound very VAMP8 and VAMP7 was precipitated by endogenous STX17 in weakly to STX-17, SNAP-29, or VAMP-7 (Figure 6A). However, hEPG5 KD cells (Figures 6O and S6T). Thus, hEPG5 promotes EPG-5 bound strongly with STX-17-SNAP-29 Qabc complexes and/or stabilizes the STX17-SNAP29-VAMP8 trans-SNARE and also with pre-assembled STX-17-SNAP-29-VAMP-7 QabcR complex. complexes (Figures 6B and 6C). Pre-assembled SNARE do- We next determined whether hEPG5 regulates the assembly of mains of STX-17-SNAP-29 and STX-17-SNAP-29-VAMP-7 SNARE complexes in vitro. STX17, SNAP29, and VAMP8 were also bound strongly to FL EPG-5 and EPG-5(375–696) (Figures incubated with or without hEPG5(427–1094), and the assem- 6D, 6E, and S6H). EPG-5(del 500–530) and EPG-5(W507A bled complexes were monitored by SDS-PAGE. The assembled W523A), which impaired RAB-7 and LGG-1 binding, still bound SNARE complex is resistant to SDS but sensitive to heat treat- strongly to STX-17-SNAP-29-VAMP-7 complexes (Figure S6I). ment. Levels of assembled SNARE complexes were increased Endogenous STX17 was co-immunoprecipitated by endoge- about 4-fold in the presence of hEPG5 (Figures 6P and 6Q), indi- nous hEPG5 using starved HeLa cell extracts (Figure 6F). GFP- cating that hEPG5 facilitates trans-SNARE assembly. hEPG5 also immunoprecipitated endogenous STX17, SNAP29, and VAMP8 in GFP-TRAP assays (Figures 6G and S6J). The EPG-5 Promotes Tethering and Lipid Mixing late endosomal/lysosomal R-SNARE VAMP7, which interacts of Proteoliposomes Mediated by with STX17, is also involved in formation and maturation of STX-17-SNAP-29-VAMP-7 autophagosomes (Itakura et al., 2012; Lamb et al., 2013). To test whether EPG-5 directly facilitates SNARE-mediated teth- hEPG5 immunoprecipitated endogenous VAMP7 (Figure 6G). ering and fusion, we performed classic in vitro lipid mixing Endogenous hEPG5 was also co-immunoprecipitated by GFP- assays using purified components. STX-17 and SNAP-29 were VAMP7 and GFP-VAMP8 (Figures 6H and S6K). No interaction reconstituted into vesicles containing the fluorophore NBD and

Molecular Cell 63, 781–795, September 1, 2016 789 Figure 6. EPG5 Directly Binds to Qabc and QabcR SNARE Complexes and Promotes SNARE Assembly (A) Worm EPG-5 weakly binds to STX-17, SNAP-29, and VAMP-7 in in vitro GST pull-down assays. (B and C) EPG-5 binds strongly to assembled STX-17-SNAP-29 Qabc SNARE complexes (B) and assembled STX-17-SNAP-29-VAMP-7 QabcR SNARE complexes (C). (D) EPG-5 binds strongly to assembled STX-17-SNAP-29 Qabc and STX-17-SNAP-29-VAMP-7 QabcR SNARE motifs. (E) EPG-5(375–696) binds strongly to assembled SNARE motifs of the STX-17-SNAP-29-VAMP-7 complex. (legend continued on next page)

790 Molecular Cell 63, 781–795, September 1, 2016 rhodamine-labeled phospholipids (donor vesicles), whereas hEPG5 Depletion Increases the Formation of STX17- VAMP-7 was reconstituted with unlabeled lipids (acceptor vesi- SNAP25-VAMP8 SNAREs cles). The concentrations of NBD and rhodamine-conjugated We investigated whether loss of hEPG5 activity affects formation lipids are high enough in the donor vesicles that the fluorescence of other SNARE complexes comprising STX17, SNAP29, or of NBD is quenched by rhodamine. Once fusion with acceptor VAMP8. In control cells, the Qbc SNAP25-GFP is mainly local- vesicles occurs, the labeled lipids become diluted, resulting ized to the plasma membrane and also to a few perinuclear in NBD fluorescence (Figure 6R). Consistent with previous puncta under normal growth conditions (Figure 7A). SNAP25- reports, the presence of SNAREs alone caused little vesicle GFP puncta are distinct from STX17 (Figures 7E and S7A). In fusion (Figure 6S). In the presence of purified GST-EPG-5, but hEPG5 KD cells, SNAP25 formed numerous puncta under both not GST alone, SNARE-mediated fusion was increased 2-fold normal and starved conditions (Figures 7B and S7B). SNAP25- (Figures 6S, S6U, and S6V; data not shown). Fusion was not GFP puncta were co-localized with STX17 and VAMP8 and detected when SNAP-29 was absent (Figure 6S). Thus, EPG-5 also with LC3 puncta in hEPG5 KD cells (Figures 7B–7E and promotes lipid mixing that is mediated by STX-17-SNAP-29- S7B–S7F). SNAP25-GFP co-localized with EHD1-mCherry- VAMP-7. labeled structures in control and hEPG5 KD cells (Figures 7E, To further test whether RAB-7 plays a role in recruiting EPG- S7G, and S7H). STX17 and VAMP8 did not non-selectively co- 5 to membranes, we anchored RAB-7 onto vesicles by adding localize with SNARE proteins. STX17 puncta were separable the transmembrane domain (TMD) of Sec61b to its C terminus, from VAMP2-GFP- and VAMP4-GFP-labeled punctate struc- and monitored membrane association of EPG-5 using a flota- tures, and VAMP8 was separable from GOSR2-GFP puncta in tion assay. Indeed, some of EPG-5 protein floated to the top control and hEPG5 KD cells (Figures S7I–S7N, S7Q, and S7R). of the sucrose-density gradient when membrane-anchored, Since SNAP25 co-localizes with STX17 and VAMP8, we tested GTPgS-soaked RAB-7 was present in the vesicles (Figure S6W). whether they form a complex. In co-IP assays, SNAP25-GFP in- Very little EPG-5 co-migrated with vesicles composed of lipids teracted weakly with endogenous STX17 and VAMP8 in control only (Figure S6W). Reconstituted, but not detergent-solubilized, cell extracts, while in hEPG5 KD cells, much more endogenous RAB-7 was enriched in the top fractions, confirming the suc- STX17 and VAMP8 was precipitated by SNAP25 (Figure 7F). cessful incorporation of TMD-containing RAB-7 into proteolipo- Higher levels of endogenous SNAP25 were also co-immunopre- somes (Figure S6W). Together, these results suggest that cipitated by GFP-STX17 in hEPG5 KD cells (Figure S7S). Thus, RAB-7 recruits EPG-5 to membranes when activated by GTP hEPG5 depletion results in the assembly of STX17-SNAP25- loading. VAMP8 SNARE complexes. Loss of hEPG5 activity does not To test whether RAB-7-recruited EPG-5 can facilitate SNARE non-selectively promote formation of SNARE complexes. Levels assembly and subsequent fusion, we co-reconstituted TMD- of endogenous VAMP8 co-precipitated by STX12-GFP or containing RAB-7 and VAMP-7 into acceptor vesicles, and STX10-GFP remained unchanged in hEPG5 KD cells (Figures tested their fusion with STX-17-SNAP-29-containing donor ves- S7T and S7U). icles (Figure 6R). RAB-7 alone does not promote fusion of We next determined whether the increased formation of SNARE-containing vesicles, but fusion was enhanced when STX17-SNAP25-VAMP8 complexes contributes to the autopha- EPG-5 was also present (Figure 6S). These results support the gic defect in hEPG5 KD cells. SNAP25 KD causes slight changes hypothesis that recruitment of EPG-5 by RAB-7, probably to in levels of LC3 and p62 (Figures 7G–7I). Simultaneous KD of late endosomal/lysosomal membranes, is essential for STX-17- SNAP25 greatly reduced p62 and LC3 levels in hEPG5 KD cells SNAP-29-VAMP-7-mediated membrane fusion. (Figures 7G–7I). The suppression effect caused by SNAP25 KD

(F) Endogenous hEPG5 co-immunoprecipitates endogenous STX17 in a co-IP assay. Cells were starved for 2 hr and the resulting precipitants were immunoblotted with anti-STX17. (G) Endogenous STX17, SNAP29, VAMP8, and VAMP7 are precipitated by GFP-hEPG5 in GFP-TRAP assays. (H) Endogenous hEPG5 is precipitated by GFP-VAMP8 in a GFP-TRAP assay. (I and J) hEPG5(427–1094) binds strongly to assembled STX17-SNAP29 Qabc SNARE complexes (I) and assembled STX17-SNAP29-VAMP8 QabcR SNARE complexes (J). The assembled STX17 and VAMP8 lack the transmembrane region. (K) GFP-hEPG5 co-localizes extensively with Myc-VAMP8 puncta. (L) When co-expressed with the SNARE domain of Myc-VAMP8(1–73), which shows peripheral localization, GFP-hEPG5 is more diffuse and forms smaller, more dispersed puncta. (M) Endogenous STX17, SNAP29, VAMP8, and VAMP7 levels are much lower in hEPG5 KD cells in immunoblotting assays. STX17, SNAP29, VAMP8, and VAMP7 levels were set to 1 by normalizing with the corresponding actin input. (N) In a co-IP assay, similar levels of endogenous SNAP29 are precipitated by endogenous STX17 in hEPG5 KD and control cells. Cells were starved for 2 hr before analysis. Levels of co-immunoprecipitated SNAP29 were normalized to the corresponding STX17 (set to 1 in control cells). (O) Levels of endogenous VAMP8 co-immunoprecipitated by STX17 are much lower in hEPG5 KD cells in a co-IP assay. Levels of co-precipitated VAMP8 were normalized to the corresponding STX17 (set to 1 in control cells). (P and Q) hEPG5(427–1094) facilitates assembly of the STX17-SNAP29-VAMP8 SNARE complex, as detected by SDS-PAGE. Heat treatment destabilizes the complex (P). Quantiﬁcation of relative SNARE complex levels in the presence or absence of hEPG5 is shown in (Q). Data are shown as mean ± SEM. **p < 0.01. (R) Illustration of the liposome tethering and fusion assays. (S) EPG-5 enhances lipid mixing of proteoliposomes reconstituted with autophagic SNAREs. Mixing is further increased by RAB-7. Scale bars, 10 and 2 mm (inserts). See also Figure S6.

Molecular Cell 63, 781–795, September 1, 2016 791 Figure 7. hEPG5 Depletion Increases the Formation of STX17-SNAP25-VAMP8 SNAREs (A and B) In hEPG5 KD cells (B), SNAP25-GFP forms more perinuclear punctate structures than in control cells (A). SNAP25 and STX17 puncta co-localize in hEPG5 KD cells. (C and D) In hEPG5 KD cells (D), SNAP25-GFP puncta are more abundant than in control cells (C) and co-localize with LC3 puncta, detected by anti-LC3. (E) Percentage of vesicles labeled by the indicated marker that are also positive for SNAP25-GFP in control and hEPG5 KD cells. Data are shown as mean ± SEM. **p < 0.01; N.S., no difference. (F) Levels of endogenous STX17 and VAMP8 co-immunoprecipitated by SNAP25-GFP are much higher in hEPG5 KD cells in GFP-TRAP assays. (G) SNAP25 protein levels are greatly reduced by SNAP25 siRNA in control and hEPG5 siRNA-treated cells.

(legend continued on next page)

792 Molecular Cell 63, 781–795, September 1, 2016 was inhibited by Baf.A1 treatment, which inhibits lysosomal In mediating autophagosome maturation, EPG5 exhibits some degradation (Figure 7I). This indicates that SNAP25 KD pro- properties of tethering factors involved in intracellular membrane motes autophagic flux in hEPG5 KD cells. Co-localization of trafficking. Most tethering factors are Rab effectors. Binding to LC3 puncta with EHD1-positive structures in hEPG5 KD cells Rabs and the SNARE complex contributes to their association was also suppressed by simultaneous SNAP25 depletion (Fig- with distinct membranes (Cai et al., 2007; Stenmark, 2009). For ures 7J–7L). EEA1 puncta were still enlarged in the double KD example, the tethering factor p115 is initially recruited to COPII cells (Figures S7O and S7P). Therefore, SNAP25 inactivation vesicles by Rab1, but is retained by interacting with COPII partially suppresses the autophagy defect caused by hEPG5 vesicle-associated SNAREs (Bentley et al., 2006). Tethering fac- depletion. tors also regulate SNARE stability. The Cog6 subunit of the COG complex interacts directly with and stabilizes Stx6 (Laufman DISCUSSION et al., 2011). Thus, EPG5 may act as a tethering factor to promote fusion of autophagosomes with late endosomes/lysosomes. EPG5 Is a Rab7 Effector that Mediates Autophagosome The concerted action of different tethering factors has been Maturation shown to ensure vesicle transport specificity (Bro¨ cker et al., Here we demonstrated that EPG5 determines the fusion speci- 2010; Yu and Hughson, 2010). For example, interaction of ficity of autophagosomes with late endosomes/lysosomes (Fig- p115 on COPII vesicles with the Golgi-resident GM130/ ure 7M). EPG5 is a Rab7 effector; it binds to the GTP-bound GRASP65 tethering complex is required for endoplasmic reticu- form of Rab7, but only weakly to the GDP-bound form. Recruit- lum (ER)-to-Golgi vesicle transport (Shorter et al., 2002). Auto- ment of EPG5 to late endosomes/lysosomes is enhanced by a phagosomal HOPS and ATG14 have been shown to facilitate constitutively active mutant of Rab7, but is inhibited by a domi- fusion of autophagosomes with late endosomes/lysosomes nant-negative mutant. The membrane association of EPG5 is (Jiang et al., 2014; Diao et al., 2015). Distinct from EPG5, also regulated by its interaction with R-SNAREs VAMP7/8. Over- ATG14 binds to STX17 and STX17-SNAP29 Qabc SNARE com- expression of mutant VAMP7/8 with a deletion of the membrane- plexes on autophagosomes, but not to STX17-SNAP29-VAMP8 anchoring motif causes their mislocalization and also disrupts complexes (Diao et al., 2015), indicating that ATG14 acts prior the late endosome/lysosome localization of EPG5. Multivalent to the assembly of fusogenic SNAREs. EPG5 may act coopera- interactions of EPG5 with Rab7 and VAMP7/8 may determine tively with autophagosomal ATG14 and HOPS to tether and dock the spatial and temporal localization of EPG5 on microdomains autophagosomes with late endosomes/lysosomes. Compared of late endosomes/lysosomes. To mediate capturing, docking, to loss of function of genes acting during autophagosome forma- and fusion of autophagosomes, EPG5 recognizes LC3/LGG-1 tion, the autophagy defect caused by EPG5 deficiency is weaker on the outer membrane of autophagosomes. EPG5 stabilizes (Tian et al., 2010; Zhao et al., 2013a), indicating that autophagic and also facilitates the assembly of trans-SNARE complexes flux can progress at low levels in EPG5-depleted cells. ATG14 for autophagosome maturation. In hEPG5 KD cells, levels of and the HOPS complex may mediate fusion of autophagosomes STX17, SNAP29, and VAMP7/8, and the assembled STX17- with late endosomes/lysosomes in EPG5-depleted cells. STX17- SNAP29-VAMP8 SNARE complex, are dramatically reduced. SNAP29-VAMP7/8 complexes may also be able to facilitate the EPG5 appears not to participate in the initial formation of the slow progression of autophagic flux. STX17-SNAP29 Qabc complex, and formation of this complex is unaffected in hEPG5 KD cells. EPG5 also accelerates trans- EPG5 Coordinates Endocytic Trafficking with SNARE assembly. EPG5 binds simultaneously to STX17- Autophagosome Maturation SNAP29 Qabc- and VAMP7/8 R-SNAREs, and may serve as a Autophagosome maturation involves fusion with early and late platform to allow them to interact directly. VAMP7 and VAMP8 endosomes (Lamb et al., 2013). The detailed itinerary for auto- may function redundantly to mediate autophagosome matura- phagosome maturation, however, remains largely unknown. Au- tion. Binding of EPG5 to the STX17-SNAP29 Qabc complex, tophagosomes may intersect with endocytic vesicles at multiple but not individual STX17 and SNAP29, ensures that closed auto- points. Autophagosomes directly fuse with Rab7-positive endo- phagosomes, but not IMs, are tethered and fused with late endosomes/lysosomes, a process promoted by EPG5. Autophago- somes/lysosomes. somes can also fuse with early endosomes to form amphisomes,

(H) Real-time PCR analysis of siRNA KD efﬁciency. mRNA levels were normalized to that in control cells, which was set to 100. Data are shown as mean ± SEM (n = 3 independent experiments). *p < 0.05, ***p < 0.001. (I) Levels of LC3 and p62 in cells stably expressing control and hEPG5 shRNA transfected with control siRNA and SNAP25 siRNA with or without Baf.A1 treatment. p62 and LC3-II levels are normalized to the corresponding actin input (set to 1 in control cells). (J–L) EHD1-GFP puncta are separable from LC3 in SNAP25 KD cells (K). The enlarged EHD1-GFP puncta and co-localization of EHD1-GFP puncta with LC3 in hEPG5 KD cells are suppressed by simultaneous SNAP25 KD (L). (J) shows quantiﬁcation of co-localization of EHD1-GFP puncta with LC3 puncta. Data are shown as mean ± SEM. *p < 0.05. (M) Model showing how EPG5 mediates fusion of autophagosomes with late endosomes/lysosomes. EPG5 is recruited to late endosomes/lysosomes by associating with Rab7 and VAMP7/8. EPG5 tethers late endosomes/lysosomes with autophagosomes by binding to LC3 and STX17/SNAP29 and facilitates the assembly of the trans-SNARE for fusion. Scale bars, 10 and 2 mm (inserts). See also Figure S7.

Molecular Cell 63, 781–795, September 1, 2016 793 which gradually mature into late endosomes and then fuse with incubated with diluted primary antibodies for 4 hr or overnight at 4C, then lysosomes. Whether EPG5 facilitates endosome maturation with 25 mL pre-cleaned agarose beads for 1.5 hr. Precipitated immune com- has yet to be determined. plexes were washed five times with washing buffer and boiled for 10 min in 23 SDS-PAGE loading buffer for immunoblotting. Equal amount of extracts Depletion of tethering factors impairs fusion of transport vesi- were used for treatment with control IgG and anti-Myc (or other indicated cles with their target membranes. Loss of TECPR1, HOPS, and antibodies). ATG14 also causes accumulation of autophagosomes (Chen et al., 2012; Jiang et al., 2014; Diao et al., 2015). However, Assembly of C. elegans SNARE Complexes Epg5 deficiency results in non-specific fusion of autophago- The STX-17/SNAP-29/VAMP-7 complex was formed by mixing Trx-STX- somes with various endocytic vesicles, leading to the formation 17(1–199) (containing the cytosolic region), Trx-SNAP-29, and GST-VAMP- of abnormally enlarged vesicles with mixed identities. These 7, or Trx-STX-17(105–199), Trx-SNAP29(31–245), and GST-VAMP-7(167– enlarged vesicles are not stained by lysotracker and thus are 233) (containing the SNARE domain of each protein) in a 2:2:1 molecular ratio. After incubating at 4C overnight, the complex was conjugated to not degradative. Non-specific fusion of autophagosomes with GST beads and washed extensively. Bound proteins were boiled in 53 EHD1-labeled recycling endosomes in hEPG5 depletion cells SDS sample loading buffer and analyzed by SDS-PAGE and Coomassie appears to be due to elevated formation of STX17-SNAP25- blue staining. The STX-17-SNAP-29 complex was assembled in the same VAMP8 complexes. KD of SNAP25 partially suppresses the coa- way except that GST-STX-17(1–199) and Trx-SNAP-29(31–245) were mixed lescence of autophagosomes with recycling endosomes and the in a 1:2 molecular ratio. autophagy defect in hEPG5-depleted cells. The abnormal assembly of SNARE complexes in hEPG5 KD cells may be caused Data Analysis by mislocalization or impaired stability of SNARE proteins as a All data are expressed as mean ± SEM. Immunoblot band density was quan- tified using ImageJ software (NIH Image). Unpaired t tests were used to result of abnormal fusion of autophagosomes with endocytic analyze between-group differences. At least three independent repeats were vesicles. Increased assembly of STX17-SNAP29-VAMP8 by performed for each experiment. reduced O-GlcNAc modification of SNAP29 also partially rescues the autophagy defect caused by EPG5 depletion (Guo SUPPLEMENTAL INFORMATION et al., 2014). In addition to autophagy, EPG5 is essential for endocytic recycling and endocytic degradation (Zhao et al., Supplemental Information includes Supplemental Experimental Procedures 2013a). Consistent with this, hEPG5 forms small puncta that and seven figures and can be found with this article online at http://dx.doi. are not labeled by Rab7. The involvement of EPG5 in multiple org/10.1016/j.molcel.2016.08.021. intracellular trafficking processes suggests that EPG5 may coordinate autophagic and endocytic trafficking to maintain AUTHOR CONTRIBUTIONS cellular homeostasis. Zheng Wang, G.M., X.X., Y.C., J.H., and H.Z. designed the experiments, inter- preted the data, and prepared the manuscript. Zheng Wang performed in vitro EPG5 and Pathogenesis of Vici Syndrome and ALS pull-down and SNARE assembly experiments. G.M. and X.X. performed tissue Recessive mutations in human EPG5 cause Vici syndrome, a culture experiments and co-IP assays. C.Y., Zhaoyu Wang, and G.Z. performed C. elegans experiments. X.G. and J.H. performed in vitro tethering multisystem disorder with a wide range of clinical manifesta- and fusion analysis. D.F. contributed to the electron micrographs. H.Z. de- tions, including agenesis of the corpus callosum, myopathy, signed and supervised the study. and combined immunodeficiency (Cullup et al., 2013). Epg5- deficient mice share some phenotypes with Vici syndrome pa- ACKNOWLEDGMENTS tients, including defective autophagy, neurodegeneration, and myopathy (Zhao et al., 2013b). Epg5-deficient mice exhibit se- We are grateful to Dr. Isabel Hanson for editing work. We would like to thank lective motor neuronal damage and recapitulate key features Ms. Shuoguo Li from the Center for Biological Imaging (CBI), Institute of of amyotrophic lateral sclerosis (ALS) (Zhao et al., 2013a, Biophysics, Chinese Academy of Science, for her help in taking and analyzing SIM images. H.Z. was supported by grants from the NSFC (31421002, 2013b). Non-degradative autophagic vacuoles accumulate in 31561143001, and 31225018), the National Basic Research Program of China various tissues in Vici syndrome and Epg5 knockout mice, and (2013CB910100), and in part by an International Early Career Scientist grant this may contribute to the pathogenesis associated with EPG5 from the Howard Hughes Medical Institute. depletion. Our study provides insight into the molecular mechanisms underlying Vici syndrome and ALS. Received: June 14, 2016 Revised: July 29, 2016 Accepted: August 12, 2016 EXPERIMENTAL PROCEDURES Published: September 1, 2016

Co-immunoprecipitation Assay REFERENCES HeLa cells stably expressing control short hairpin RNA (shRNA) or hEPG5 shRNA were transfected with control or hEPG5 siRNA for 72 hr. Forty-eight Bentley, M., Liang, Y., Mullen, K., Xu, D., Sztul, E., and Hay, J.C. (2006). SNARE hours after siRNA transfection, cells were transfected with the indicated con- status regulates tether recruitment and function in homotypic COPII vesicle structs for 24 hr. Cells were lysed in lysis buffer (50 mM Tris-HCl [pH 7.5], fusion. J. Biol. Chem. 281, 38825–38833. 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, and protease inhibitor cocktail). ˚ Lysates were centrifuged at 13,400 rpm and the supernatant (input) was Birgisdottir, A.B., Lamark, T., and Johansen, T. (2013). The LIR motif—crucial collected and used for immunoprecipitation. For GFP-tag proteins, lysates for selective autophagy. J. Cell Sci. 126, 3237–3247. were immunoprecipitated by GFP-TRAP agarose beads (ChromoTek, gta- Bonifacino, J.S., and Glick, B.S. (2004). The mechanisms of vesicle budding 20) for 1.5 hr at 4C. For endogenous or other tagged proteins, lysates were and fusion. Cell 116, 153–166.

794 Molecular Cell 63, 781–795, September 1, 2016 Bro¨ cker, C., Engelbrecht-Vandre´ , S., and Ungermann, C. (2010). Multisubunit Lu, Q., Wu, F., and Zhang, H. (2013). Aggrephagy: lessons from C. elegans. tethering complexes and their role in membrane fusion. Curr. Biol. 20, R943– Biochem. J. 452, 381–390. R952. McEwan, D.G., Popovic, D., Gubas, A., Terawaki, S., Suzuki, H., Stadel, D., Cai, H., Reinisch, K., and Ferro-Novick, S. (2007). Coats, tethers, Rabs, and Coxon, F.P., Miranda de Stegmann, D., Bhogaraju, S., Maddi, K., et al. (2015). SNAREs work together to mediate the intracellular destination of a transport PLEKHM1 regulates autophagosome-lysosome fusion through HOPS complex vesicle. Dev. Cell 12, 671–682. and LC3/GABARAP proteins. Mol. Cell 57, 39–54. Chen, D., Fan, W., Lu, Y., Ding, X., Chen, S., and Zhong, Q. (2012). A mamma- Mizuno-Yamasaki, E., Rivera-Molina, F., and Novick, P. (2012). GTPase net- lian autophagosome maturation mechanism mediated by TECPR1 and the works in membrane traffic. Annu. Rev. Biochem. 81, 637–659. Atg12-Atg5 conjugate. Mol. Cell 45, 629–641. Cullup, T., Kho, A.L., Dionisi-Vici, C., Brandmeier, B., Smith, F., Urry, Z., Nakatogawa, H., Suzuki, K., Kamada, Y., and Ohsumi, Y. (2009). Dynamics Simpson, M.A., Yau, S., Bertini, E., McClelland, V., et al. (2013). Recessive mu- and diversity in autophagy mechanisms: lessons from yeast. Nat. Rev. Mol. tations in EPG5 cause Vici syndrome, a multisystem disorder with defective Cell Biol. 10, 458–467. autophagy. Nat. Genet. 45, 83–87. Shorter, J., Beard, M.B., Seemann, J., Dirac-Svejstrup, A.B., and Warren, G. Diao, J., Liu, R., Rong, Y., Zhao, M., Zhang, J., Lai, Y., Zhou, Q., Wilz, L.M., Li, (2002). Sequential tethering of Golgins and catalysis of SNAREpin assembly J., Vivona, S., et al. (2015). ATG14 promotes membrane tethering and fusion of by the vesicle-tethering protein p115. J. Cell Biol. 157, 45–62. autophagosomes to endolysosomes. Nature 520, 563–566. Stenmark, H. (2009). Rab GTPases as coordinators of vesicle traffic. Nat. Rev. Eskelinen, E.L. (2005). Maturation of autophagic vacuoles in mammalian cells. Mol. Cell Biol. 10, 513–525. Autophagy 1, 1–10. Su¨ dhof, T.C., and Rothman, J.E. (2009). Membrane fusion: grappling with Feng, Y., He, D., Yao, Z., and Klionsky, D.J. (2014). The machinery of macro- SNARE and SM proteins. Science 323, 474–477. autophagy. Cell Res. 24, 24–41. Guo, B., Liang, Q., Li, L., Hu, Z., Wu, F., Zhang, P., Ma, Y., Zhao, B., Kova´ cs, Tian, Y., Li, Z., Hu, W., Ren, H., Tian, E., Zhao, Y., Lu, Q., Huang, X., Yang, P., A.L., Zhang, Z., et al. (2014). O-GlcNAc-modification of SNAP-29 regulates au- Li, X., et al. (2010). C. elegans screen identifies autophagy genes specific to tophagosome maturation. Nat. Cell Biol. 16, 1215–1226. multicellular organisms. Cell 141, 1042–1055. Itakura, E., Kishi-Itakura, C., and Mizushima, N. (2012). The hairpin-type Wickner, W., and Schekman, R. (2008). Membrane fusion. Nat. Struct. Mol. tail-anchored SNARE syntaxin 17 targets to autophagosomes for fusion with Biol. 15, 658–664. endosomes/lysosomes. Cell 151, 1256–1269. Wu, F., Watanabe, Y., Guo, X.Y., Qi, X., Wang, P., Zhao, H.Y., Wang, Z., Jahn, R., and Scheller, R.H. (2006). SNAREs—engines for membrane fusion. Fujioka, Y., Zhang, H., Ren, J.Q., et al. (2015). Structural basis of the differential Nat. Rev. Mol. Cell Biol. 7, 631–643. function of the two C. elegans Atg8 homologs, LGG-1 and LGG-2, in auto- Jiang, P., Nishimura, T., Sakamaki, Y., Itakura, E., Hatta, T., Natsume, T., and phagy. Mol. Cell 60, 914–929. Mizushima, N. (2014). The HOPS complex mediates autophagosome-lyso- Yu, I.M., and Hughson, F.M. (2010). Tethering factors as organizers of intracel- some fusion through interaction with syntaxin 17. Mol. Biol. Cell 25, 1327– lular vesicular traffic. Annu. Rev. Cell Dev. Biol. 26, 137–156. 1337. Lamb, C.A., Dooley, H.C., and Tooze, S.A. (2013). Endocytosis and auto- Zhao, H., Zhao, Y.G., Wang, X., Xu, L., Miao, L., Feng, D., Chen, Q., Kova´ cs, phagy: shared machinery for degradation. BioEssays 35, 34–45. A.L., Fan, D., and Zhang, H. (2013a). Mice deficient in Epg5 exhibit selective neuronal vulnerability to degeneration. J. Cell Biol. 200, 731–741. Laufman, O., Hong, W., and Lev, S. (2011). The COG complex interacts directly with Syntaxin 6 and positively regulates endosome-to-TGN retrograde trans- Zhao, Y.G., Zhao, H., Sun, H., and Zhang, H. (2013b). Role of Epg5 in selective port. J. Cell Biol. 194, 459–472. neurodegeneration and Vici syndrome. Autophagy 9, 1258–1262.

Molecular Cell 63, 781–795, September 1, 2016 795