|Home | About | Journals | Submit | Contact Us | Français|
The biosynthetic sorting of acid hydrolases to lysosomes relies on transmembrane, mannose 6-phosphate receptors (MPRs) that cycle between the TGN and endosomes. Herein we report that maintenance of this cycling requires the function of the mammalian Golgi-associated retrograde protein (GARP) complex. Depletion of any of the three GARP subunits, Vps52, Vps53, or Vps54, by RNAi impairs sorting of the precursor of the acid hydrolase, cathepsin D, to lysosomes and leads to its secretion into the culture medium. As a consequence, lysosomes become swollen, likely due to a buildup of undegraded materials. Missorting of cathepsin D in GARP-depleted cells results from accumulation of recycling MPRs in a population of light, small vesicles downstream of endosomes. These vesicles might correspond to intermediates in retrograde transport from endosomes to the TGN. Depletion of GARP subunits also blocks the retrograde transport of the TGN protein, TGN46, and the B subunit of Shiga toxin. These observations indicate that the mammalian GARP complex plays a general role in the delivery of retrograde cargo into the TGN. We also report that a Vps54 mutant protein in the Wobbler mouse strain is active in retrograde transport, thus explaining the viability of these mutant mice.
Lysosomal acid hydrolases are a structurally diverse group of enzymes that are efficiently targeted to mammalian lysosomes by virtue of a shared posttranslational modification: the acquisition of mannose 6-phosphate residues on their N-linked carbohydrate chains (Kornfeld and Mellman, 1989 ). This modification is catalyzed by the sequential action of two enzymes, a UDP-N-acetylglucosamine:lysosomal enzyme N-acetylglucosamine-1-phosphotransferase localized to cis-Golgi cisternae (Lazzarino and Gabel, 1988 ) and an N-acetylglucosamine-1-phosphodiester-N-acetylglucosaminidase localized to the trans-Golgi network (TGN; Rohrer and Kornfeld, 2001 ). At the TGN, the mannose 6-phosphate-modified hydrolases bind to two transmembrane mannose 6-phosphate receptors (MPR), a cation-dependent MPR (CD-MPR) and a cation-independent MPR (CI-MPR), leading to the concentration of the hydrolases within clathrin-coated areas of the TGN (Ghosh et al., 2003 ). Transport carriers then form that deliver the hydrolase-MPR complexes into endosomes. Exposure to the acid pH of the endosomal lumen causes the release of the hydrolases from the MPRs, after which the hydrolases continue on to lysosomes while the receptors are retrieved to the TGN to be reutilized in further rounds of hydrolase sorting (Ghosh et al., 2003 ).
It is now well established that sorting of the hydrolase-MPR complexes at the TGN involves recognition of specific signals in the cytosolic tails of the receptors by the clathrin-associated, GGA proteins and AP-1 complex (Ghosh et al., 1998 ; Bonifacino, 2004 ; Ghosh and Kornfeld, 2004 ). The subsequent retrieval of unoccupied MPRs from endosomes to the TGN, on the other hand, depends on other components of the protein trafficking machinery. Among these are Rab9 and TIP47, which retrieve MPRs from late endosomes (Diaz and Pfeffer, 1998 ; Carroll et al., 2001 ), and epsinR (Saint-Pol et al., 2004 ) and the retromer complex (Arighi et al., 2004 ; Seaman, 2004 ; Carlton et al., 2005b ), which do so from early endosomes or early-late endosomal intermediates. In the case of retromer, retrograde transport involves passage through tubules that emanate from the vacuolar part of endosomes (Arighi et al., 2004 ; Carlton et al., 2004 , 2005a ,b ).
Despite the identification of several key components of the lysosomal transport machinery in mammalian cells, there is reason to think that many more components remain to be identified. Indeed, more than 60 different vacuolar protein-sorting (VPS) gene products have been shown to participate in the sorting of acid hydrolases to the vacuole of the yeast, Saccharomyces cerevisiae (Bowers and Stevens, 2005 ). The phylogenetic conservation of the core trafficking machinery and the diversification of lysosome function in mammalian cells predict that an even larger number of proteins must be involved in sorting acid hydrolases to lysosomes.
To identify novel or uncharacterized proteins that are involved in acid hydrolase sorting in human cells, we performed an RNA interference (RNAi) screen for the requirement of 39 candidate proteins in this pathway. These candidates were selected based on their homology to yeast VPS gene products or their involvement in other endocytic or lysosomal targeting events. This screen revealed a key role for the human homolog of the yeast Golgi-associated retrograde protein (GARP; Conibear and Stevens, 2000 ) or Vps fifty-three (VFT; Siniossoglou and Pelham, 2001 ) complex in the sorting of the acid hydrolase, cathepsin D (CatD), to lysosomes. RNAi-mediated depletion of GARP subunits caused secretion of unprocessed CatD into the culture medium due to impaired recycling of MPRs from endosomes to the TGN. We also found that GARP depletion blocked endosome-to-TGN transport of the TGN protein, TGN46, and the B-subunit of Shiga toxin (STxB). In the absence of GARP, all of these proteins accumulated in a population of small vesicles that likely correspond to endosome-to-TGN retrograde transport intermediates. Finally, we observed that a missense mutation in the GARP-Vps54 subunit found in the Wobbler motor neuron disease mouse strain does not preclude its function in retrograde transport of the CI-MPR. This explains why, unlike mice with ablation of the Vps54 gene, Wobbler mice are viable. These findings thus identify the GARP complex as a novel component of the molecular machinery involved in retrograde transport of various cargo proteins in human cells, probably by enabling the fusion of endosome-derived transport carriers with the TGN.
GARP subunit cDNAs were amplified by PCR from human or mouse fetal brain cDNA libraries or by RT-PCR from HeLa cell total RNA. The resulting cDNAs were subsequently cloned in-frame with the V5 and hexahistidine tags at their C-termini in TOPO pEF6-V5-His (Invitrogen, Carlsbad, CA). Primer sequences are shown in the Supplementary Methods section. The mouse Vps54 cDNA was further subcloned into the XhoI-BamHI sites of pEGFP-N1 (Clontech Laboratories, Mountain View, CA). To obtain a small interfering RNA (siRNA)-resistant human Vps52-V5, five of the nucleotides targeted by the siRNA (232-GTAGATCTCGt cacTAtT-250) were mutated (232-GTAGATCTCCGcgcaTAcT-250) by site-directed mutagenesis using the QuikChange kit (Stratagene, La Jolla, CA). Mouse Vps54-V5 was also mutated at codons for D145E and I151T to match them with the human epitope selected for the generation of rabbit polyclonal antibodies. To create an allele similar to that of the Wobbler mouse, we introduced the L967Q mutation in the mouse Vps54-V5 and Vps54-GFP plasmids.
RNAi was performed using siRNAs from Dharmacon (Lafayette, CO). Initial screens were performed with siGENOME SMART pools or ON-TARGET plus SMART pools. Subsequently, the four different duplexes of each SMART pool for GARP subunits were tested and oligonucleotides Vps52.1 (GUAGAUCUCCGUCACUAUUUU, D-011806-01), Vps53.4 (GGAUGUAAGUCUGAUUGAAUU, J-017048-08), Vps54.3 (UCACGAUGUUUGCAGUUAAUU, J-021174-07, targeting only human Vps54), and Vps54.4 (CCAGAUCUCUCUUACGUUCAUU, J-021174-08, targeting both human and mouse Vps54) were selected and used in knockdown (KD) experiments. Transfection of oligonucleotides (typically 40 nM) was done using Oligofectamine (Invitrogen) according to the manufacturer's protocol.
Human HeLa epithelial or H4 neuroglioma cells (American Type Culture Collection, Manassas, VA) were cultured on 24- or 6-well plates at 37°C in DME/high-glucose medium (Invitrogen) supplemented with 10% (vol/vol) FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin. When cells reached 80% confluency, they were transfected with 0.8–3.2 μg plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. For stable expression of Vps54-GFP in H4 cells, cells were transfected and selected in medium containing 0.5 mg/ml G418 (Geneticin, Invitrogen). Positive clones were identified by expression of Vps54-GFP by fluorescence microscopy and expanded. For immunoprecipitation experiments, cells were lysed in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 5 mM EDTA, 1% (vol/vol) Triton X-100, and a protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN), on ice for 30 min, and microcentrifuged. Lysates were further cleared with 30 μl protein A-Sepharose beads (Amersham Biosciences), before adding specific antibodies (2 μl sera) bound to protein A-Sepharose beads and rocking at 4°C for 2 h. Immunoprecipitated material was washed three times in PBS and eluted from the beads by heating at 90°C for 3 min in Laemmli sample buffer. Samples were subsequently analyzed by SDS-PAGE and immunoblotting.
Polyclonal antibodies were raised by immunization of rabbits with peptides corresponding to amino acids 79-96 of human Vps52, 61-78 of human Vps53, and 138-155 of human Vps54 (Quality Controlled Biochemicals, Hopkinton, MA). Polyclonal antibodies to human Vps53 were purified by affinity chromatography on immobilized peptide. The three antibodies were able to immunoprecipitate the corresponding proteins. Only antibodies to Vps52 and Vps53 worked by immunoblotting, and none detected the endogenous proteins by immunofluorescence microscopy. In addition, the following antibodies were used for immunofluorescence and/or immunoblotting: mouse monoclonal antibodies to the V5 epitope (Invitrogen); p230 (golgin-245), Vti1a, GS28, BiP, and actin (BD Biosciences); CI-MPR (clone 2G11; AbCam, Cambridge, MA); TfR (clone H68.4; Invitrogen), and CD63 (clone H5C6; Developmental Studies Hybridoma Bank, Iowa City, IO). Rabbit polyclonal antibodies to human SNX2 (Haft et al., 2000 ), p230 (Yoshino et al., 2005 ), and CI-MPR (Kametaka et al., 2005 ) have been described previously. Other polyclonal antibodies used were sheep antibody to human TGN46 (Serotec, Raleigh, NC); rabbit antibody to giantin (Covance Research Products, Denver, PA), GFP (Invitrogen), and CatD (Calbiochem, San Diego, CA); Alexa-488– or -594–, or -647–conjugated donkey anti-mouse IgG, Alexa-488– or -594–, or -647–conjugated donkey anti-rabbit IgG, and Alexa-488–, -594–, or -647–conjugated donkey anti-sheep IgG (Molecular Probes, Eugene, OR); horseradish peroxidase–conjugated mouse anti-goat IgG (Pierce, Rockford, IL); and horseradish peroxidase–conjugated donkey anti-mouse and donkey anti-rabbit IgG (Amersham Biosciences).
Immunofluorescence microscopy was performed as described previously (Mardones et al., 2007 ). Fluorescently labeled cells were examined using an inverted confocal laser scanning microscope (model LSM 510; Carl Zeiss MicroImaging, Thornwood, NY) equipped with Ar, HeNe, and Kr lasers and a 63× 1.4 NA objective. Alexa-488, -594, and -643 fluorescence was visualized using excitation filters at 488, 543, and 633 nm and emission filters at 505–530, 560–605, and 605 nm, respectively. Where indicated, an epifluorescence Zeiss microscope (Carl Zeiss MicroImaging,) equipped with a PlanApo 63× 1.4 NA oil immersion objective and a charge-coupled device (CCD) AxioCam MRn camera (Carl Zeiss MicroImaging) was also used.
Antibody uptake assays were carried out by incubation for 1 h at 37°C of HeLa cells grown on coverslips in the presence of 10 μg/ml mAb to the luminal domain of the CI-MPR diluted in DMEM, 1% BSA, and 25 mM HEPES, pH 7.4. The cells were washed in PBS, chased in complete medium for 1 h, washed again in PBS, and fixed in −20°C methanol. Cy3-STxB (a kind gift from L. Johannes, Curie Institute, Paris, France) was added to cells grown on coverslips at a dilution of 0.5 μg/ml in DMEM, 1% BSA, and 25 mM HEPES, pH 7.4, for 15 min at 37°C. Cells were washed in PBS and chased for 1 h at 37°C in complete medium before fixation in 3.7% paraformaldehyde.
SDS-PAGE and electroblotting onto nitrocellulose membranes were performed using the NuPAGE Bis-Tris Gel system (Invitrogen), according to the manufacturer's instructions. Incubations with primary and secondary antibodies, enzymatic detection, and quantification were performed as described (Mardones et al., 2007 ).
Metabolic labeling of cells was carried out as described (Mardones et al., 2007 ). Briefly, cells grown on six-well plates were pulse-labeled for 2 h at 20°C using 0.1 mCi/ml [35S]methionine-cysteine (Express Protein Label; Perkin Elmer-Cetus, Boston, MA) and chased for 1–20 h at 37°C in regular medium supplemented with 0.06 mg/ml methionine and 0.1 mg/ml cysteine. Chase medium was saved for further use, and cells were rinsed twice in PBS and subjected to lysis in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 2 mM EDTA, 1% (vol/vol) Triton X-100, and a complete protease inhibitor cocktail (Roche Applied Science). Both cell extracts and media were immunoprecipitated and analyzed by SDS-PAGE and fluorography. Quantification was performed on a Typhoon 9200 PhosphorImager (Amersham Biosciences) using ImageQuant analysis software.
Subcellular fractionation on glycerol gradients was performed as described (Zolov and Lupashin, 2005 ) with slight modifications. Briefly, siRNA-treated HeLa cells from one 6-cm plate were collected in PBS-0.5 mM EDTA, pelleted by centrifugation for 5 min at 500 × g, washed once in PBS and once in STE buffer (250 mM sucrose, 20 mM Tris-HCl, pH 7.4, 1 mM EDTA, with protease inhibitors), homogenized by 20 passages through a 25-gauge needle in 0.5-ml buffer STE without sucrose, and then centrifuged at 1000 × g for 2 min to obtain a postnuclear supernatant (PNS). PNS (0.6 ml) was layered on top of a 2%-stepwise, 10–30% (wt/vol) glycerol gradient (11 ml in 20 mM Tris-HCl, pH 7.4, and 1 mM EDTA on a 0.4 ml 80% sucrose cushion) and centrifuged at 280,000 × g for 60 min in a SW40 Ti rotor (Beckman Coulter, Fullerton, CA). Fractions (0.9 ml) were collected from the top. All steps were performed at 4°C. Aliquots of 0.3 ml from each fraction were precipitated with trichloroacetic acid, resuspended in Laemmli sample buffer, and analyzed by SDS-PAGE and immunoblotting.
To identify novel components of the molecular machinery involved in the sorting of acid hydrolase precursors to lysosomes, we conducted a small-scale RNAi screen in HeLa cells using siRNAs directed to 39 candidate proteins (listed in legend to Figure 1). Among the candidate proteins were the human orthologues of several yeast VPS gene products that had not been well characterized in mammals. Cells treated with the different siRNAs were analyzed by immunofluorescence microscopy for the distribution of the acid hydrolase, CatD. In control cells, CatD exhibited its characteristic localization to lysosomes that were scattered throughout the cytoplasm (Figure 1A) and contained the tetraspanin, CD63 (Figure 1B). Although treatment with several siRNAs altered this distribution (data not shown), by far the biggest change was seen for cells treated with siRNAs targeting the human ortholog of yeast Vps52 (Figure 1, D–F). These cells exhibited greatly reduced staining for CatD (Figure 1D) as well as enlargement and clustering of lysosomes stained for CD63 (Figure 1E). The residual CatD staining in Vps52-depleted cells did not colocalize with CD63 but appeared restricted to the Golgi area (Figure 1F).
To determine whether the altered CatD staining in Vps52-depleted cells was due to biosynthetic missorting, we performed pulse-chase analysis of cells that were metabolically labeled with [35S]methionine-cysteine for 2 h at 20°C (to arrest transport at the TGN) and then incubated in complete medium for different periods at 37°C (to allow release of CatD from the TGN). CatD species were isolated by immunoprecipitation from both cell extracts and culture media and resolved by SDS-PAGE and fluorography (Figure 1G). We observed that, in control cells, the ~50-kDa CatD precursor (p) was cleaved to a ~47-kDa intermediate (i) and, subsequently, to a ~31-kDa mature (m) form over a period of 3–5 h (Figure 1G). This proteolytic processing of the precursor reflects the transport of the protein to lysosomes. A fraction of the precursor was secreted intact into the medium (Figure 1G), as is known to occur in transformed cells such as HeLa. Interestingly, in Vps52-siRNA–treated cells, appearance of the intermediate form was delayed and the mature form was undetectable even after 5 h of chase (Figure 1G). In addition, secretion of intact precursor increased about fivefold (Figure 1G). From these experiments we concluded that depletion of Vps52 caused severe missorting of the CatD precursor.
Vps52 was first identified in yeast as a component of the peripheral membrane protein complex known as GARP (Conibear and Stevens, 2000 ) or VFT (Siniossoglou and Pelham, 2001 ) complex. In addition to Vps52, the yeast complex comprises two other subunits named Vps53 and Vps54, and a more loosely associated component named Vps51 (Conibear and Stevens, 2000 ; Siniossoglou and Pelham, 2002 ; Conibear et al., 2003 ; Reggiori et al., 2003 ). A similar complex has recently been described in humans (Liewen et al., 2005 ), with the notable difference that no Vps51 homolog has yet been identified in humans or other mammals. To build a set of reagents that would allow us to investigate the involvement of other subunits of the human GARP complex in CatD sorting, we made cDNA constructs encoding human Vps52, Vps53, and Vps54 tagged at their C-termini with the V5 epitope and anti-peptide antibodies to each subunit. Notably, the human Vps53 cDNA that we cloned by RT-PCR (sequence deposited in GenBank under accession number EU021218) encoded a protein of 832 amino acids, which differed from the previously reported form of 670 amino acids (Liewen et al., 2005 ). Cotransfection of different combinations of constructs encoding V5-tagged Vps52, Vps53, and Vps54, followed by immunoprecipitation with anti-peptide antibodies to each subunit and immunoblotting with antibody to V5, demonstrated that the three subunits indeed assembled into a complex (Figure 2A and Supplemental Figure 1). The recovery of the coimmunoprecipitated proteins varied depending on the antibody used, with the primary target protein always being recovered in larger amounts (Figure 2A and Supplemental Figure 1). The least efficient coimmunoprecipitation was obtained with the antibody to Vps53, perhaps indicating poor incorporation or partial hindrance of the epitope in the complex. These results were specific, as we did not observe immunoprecipitation of V5-tagged GARP subunits using preimmune serum (Supplemental Figure 1).
V5-tagged Vps53 migrated as a ~100-kDa species in these analyses (Figure 2A), which is consistent with the predicted 832 amino acids encoded by the transcript identified in our studies and with the size of the endogenous protein identified by specific antibody to Vps53 (Figure 2B). These experiments also showed that C-terminal V5-tagging did not interfere with the incorporation of any of the subunits into the complex and that the anti-peptide antibodies specifically immunoprecipitated their intended targets (Figure 2A).
The assays and reagents described above allowed us to assess the role of the other subunits of the GARP complex in CatD sorting. We initially tested four different siRNA duplexes directed to each of the Vps52, Vps53, and Vps54 subunits for their ability to deplete their target proteins and their potential toxicity. Eventually, we selected one siRNA per subunit that caused efficient depletion (Figure 2, B and C) with no obvious toxic side effects (data not shown). Immunoblotting with the anti-peptide antibodies showed that Vps52 RNAi caused virtually complete elimination of this protein and a partial decrease of Vps53 levels (Figure 2B). In addition, we observed that Vps53 RNAi completely eliminated not only this protein but also Vps52 (Figure 2B). These experiments thus demonstrated that the fates of Vps52 and Vps53 are tightly linked such that depletion of one destabilizes the other. Surprisingly, Vps54 RNAi did not alter the levels of Vps52 and Vps53 (Figure 2B). The ability of this treatment to deplete endogenous Vps54 could not be directly determined because the Vps54-specific anti-peptide antibody did not work for immunoblotting (data now shown), even though it worked for immunoprecipitation (Figure 2A). However, we were able to show that Vps54 RNAi completely prevented expression of transfected V5-tagged Vps54 (Figure 2C), indicating that the siRNA was indeed effective. Therefore, it appears that depletion of Vps54 does not destabilize Vps52 and Vps53 (Figure 2B). Finally, we examined by immunoblot analysis the effect of depleting Vps52, Vps53, or Vps54 on CatD processing (Figure 2D). We found that depletion of any of the three subunits greatly decreased the levels of mature CatD within cells (with <10% mature CatD remaining in the cell lysates), and increased 30–40-fold the amount of precursor CatD secreted into the medium (Figure 2D). From these experiments we concluded that the whole GARP complex is required for CatD sorting to lysosomes.
In yeast, the small GTPases Ypt6 and Arl1 have been shown to regulate the recruitment of the GARP complex to membranes (Siniossoglou and Pelham, 2001 ; Conibear et al., 2003 ; Panic et al., 2003 ). We observed that single depletion of the human orthologues of these proteins, Rab6A/A′ (the siRNA oligonucleotides targeted both isoforms) and Arl1, slightly decreased processing and increased secretion of precursor CatD (Figure 2E). Combined depletion of Rab6A/A′ and Arl1, however, exacerbated these defects, although in all cases they were less pronounced than those caused by depletion of GARP subunits (Figure 2E). These observations are consistent with the previously demonstrated role for Rab6A/A′ and Arl1 in retrograde transport (Mallard et al., 2002 ; Medigeshi and Schu, 2003 ; Lu et al., 2004 ; Utskarpen et al., 2006 ), which may be exerted through the recruitment of GARP and other protein tethers to membranes (Munro, 2005 ; Short et al., 2005 ).
Previous use of anti-peptide antibodies for immunofluorescence microscopy analyses showed that the endogenous GARP complex was localized to vesicles that were largely scattered throughout the cytoplasm and costained with endosomal markers (Liewen et al., 2005 ). Some accumulation in the perinuclear area and colocalization with Golgi markers, however, was also noted (Liewen et al., 2005 ). We sought to confirm this localization pattern, but unfortunately, our anti-peptide antibodies failed to immunostain endogenous GARP. To overcome this problem, we made a Vps54 construct that was tagged at the C-terminus with GFP, and expressed it by stable transfection into human H4 cells (Figure 3). Triple-labeling, immunofluorescence microscopy showed that this construct localized to a ribbon-like structure that coincided almost perfectly with the endogenous TGN marker, TGN46 (Figure 3). The structure containing Vps54-GFP also aligned with the Golgi cisternae stained for the marker protein, GM130, although merging of the images showed that the two structures were shifted relative to each other (Figure 3). Similar observations were made in HeLa cells expressing Vps54-GFP (see Figure 9, B–D). A fraction of Vps54-GFP was also consistently found on small puncta distributed throughout the cytoplasm (Figure 3). This fraction may represent cytosolic protein and/or association with small vesicles. These observations thus indicated that human GARP is largely associated with the TGN.
We next investigated how depletion of the TGN-associated GARP complex causes CatD missorting. Like other acid hydrolases, CatD is sorted by the transmembrane MPRs, which follow a cycling itinerary between the TGN and endosomes. Immunofluorescence microscopy showed that, in HeLa cells, the CI-MPR localized to a collection of cytoplasmic vesicles that were more concentrated in the juxtanuclear area (Figure 4A). As previously reported, many of these vesicles showed colocalization with SNX2 (data not shown), a component of the retromer complex that is involved in the retrieval of the CI-MPR from endosomes (Carlton et al., 2004 ; Rojas et al., 2007 ), and only a few overlapped with the TGN-localized TGN46 (Figure 4, A–D). This indicated that the steady-state distribution of the CI-MPR is skewed toward endosomes in these cells. Depletion of Vps52 caused a dramatic redistribution of both the CI-MPR and TGN46. In >95% of Vps52 siRNA-treated cells, the majority of the CI-MPR localized to a profusion of small vesicles scattered throughout the cytoplasm (Figure 4, E, I, and L; compare vesicle size in A and E insets), with the rest being found within a tight juxtanuclear structure (Figure 4E) that roughly colocalized with the Golgi cisternal marker, giantin (Figure 4G). To better characterize this juxtanuclear structure, we costained CI-MPR with the TGN marker p230 (Golgin-245) in Vps52-depleted cells (Figure 4, I–K). Merging of both images showed substantial overlap between the two signals (Figure 4K). We concluded from these experiments that this fraction of CI-MPR localizes to the TGN in GARP-depleted cells. A similar redistribution was observed for TGN46, with the notable difference that the total intensity of staining appeared substantially diminished (Figure 4F). This latter phenotype was quite dramatic and facilitated identification of GARP-depleted cells in subsequent experiments (for instance, see Figures 5, ,8F,8F, and and9,9, G and N). Immunoblot analysis showed that, unlike depletion of retromer components (Arighi et al., 2004 ; Carlton et al., 2004 ; Seaman, 2004 ; Rojas et al., 2007 ), depletion of GARP subunits did not decrease the levels of the CI-MPR (Figure 4P). In contrast to the CI-MPR, and in agreement with the immunofluorescence microscopy results (Figure 4F), total TGN46 levels were decreased about fourfold by GARP depletion (Figure 4P).
The distribution of other TGN markers such as golgin-97 (data not shown) and Golgi cisternae markers such as giantin (Figure 4G) and GM130 and galactosyl transferase (data not shown) was unaffected by GARP depletion, indicating that the altered distribution of the CI-MPR and TGN46 was not due to global disruption of the Golgi complex. In addition, the transmembrane proteins, Lamp-1 (Figure 4M) and CD63 (Figure 1E), still localized to lysosomes in GARP-depleted cells although, as previously mentioned, the lysosomes appeared swollen and clustered in the juxtanuclear area (Figures 1E and and4M)4M) after prolonged depletion of GARP. Therefore, the absence of GARP causes specific defects in the distribution of transmembrane proteins such as the CI-MPR and TGN46 that cycle between the TGN and endosomes, as well as luminal cargo proteins such as CatD, which relies on the CI-MPR for trafficking.
To ascertain the specificity of the effects observed upon depletion of GARP subunits, we performed rescue experiments in which the RNAi-treated cells were transfected with siRNA-resistant cDNAs. After two rounds of treatment with siRNAs, cells were transfected with the corresponding cDNAs and analyzed 24 h later. Use of this protocol for Vps52 resulted in almost complete (>90% of transfected cells) rescue of the steady-state localization of CatD, CI-MPR, and TGN46 over a wide range of expression levels of the transfected construct (Figure 5). Similar results were obtained for depletion and rescue of Vps53 and Vps54 (data not shown). These results demonstrated that the protein localization defects observed in GARP-depleted cells were truly due to the absence of functional GARP and not to off-target effects.
To further characterize the compartment where the CI-MPR accumulates in Vps52-depleted cells, we performed subcellular fractionation of disrupted HeLa cells by sedimentation on glycerol gradients (Figure 6). In control cells, a population of CI-MPR was found at the bottom of the gradient (fractions 11 and 12), which contained a mixture of Golgi (marked by GS28), endosomes (transferrin receptor), lysosomes (Lamp-2), and ER (BiP; Figure 6). Another CI-MPR population was found in lighter fractions that may correspond to transport intermediates. Interestingly, depletion of Vps52 caused a shift of a substantial fraction of the CI-MPR from heavy to lighter fractions that did not contain Golgi or endosomal markers but contained increased levels of the Vti1a, a t-SNARE that also cycles between the TGN and endosomes, and a fraction of the remaining TGN46 (Figure 6). These observations indicated that, in the absence of Vps52, the CI-MPR accumulates in a distinct, light membrane fraction that likely corresponds to the small vesicles visualized by immunofluorescence microscopy (Figure 4, E, I, and L).
The yeast GARP complex has been proposed to function as a tethering complex that enables docking and fusion of endosome-derived, retrograde vesicles with the TGN (Conibear and Stevens, 2000 ; Siniossoglou and Pelham, 2001 ; Whyte and Munro, 2002 ; Conibear et al., 2003 ). To determine whether the altered steady-state distribution of the CI-MPR in Vps52-depleted human cells was due to a defect in retrograde transport, we examined the fate of antibody to the luminal domain of the CI-MPR that was internalized from the cell surface. This analysis was made possible by the presence of ~10% of the total pool of CI-MPR at the cell surface at steady state (Waguri et al., 2003 ). Live control or Vps52-depleted HeLa cells were incubated for 1 h in the continuous presence of antibody to the CI-MPR, chased for 1 h in complete medium, fixed, permeabilized, and processed for immunofluorescence microscopy. In control cells, the internalized CI-MPR antibody exhibited a staining pattern (Figure 7, A and G) that was similar to that of the steady-state CI-MPR (Figures 4A and and7H),7H), namely, juxtanuclear foci and vesicles that exhibited significant colocalization with SNX2 (Figure 7B). In Vps52-depleted cells, on the other hand, the internalized CI-MPR antibody accumulated in myriad small vesicles (Figure 7, D and J; 87% of Vps-52 depleted cells showed this phenotype) similar to those containing the CI-MPR at steady state (Figures 4, E, I, and L, and and7K)7K) and largely devoid of SNX2 (Figure 7, E and F). The similar distributions of internalized and total CI-MPR in both control cells (Figure 7, G–I) and Vps52-depleted cells (Figure 7, J–L), confirm that the mislocalization of CI-MPR in Vps52-depleted cells reflects changes in recycling and not biosynthetic sorting. A shorter pulse without a chase showed significant localization of internalized CI-MPR with SNX2 even in Vps52-depleted cells (Figure 7, M–O), indicating that the small vesicles represent a downstream compartment that is accessible by retrograde transport from the plasma membrane and endosomes. We interpret that accumulation of CI-MPR in this compartment (Figure 7, D and J) prevents it from assuming its normal steady-state distribution to endosomes (Figure 7, A and G). On the basis of the localization of human GARP to the TGN and the function of yeast GARP in mediating docking and fusion with the TGN (Conibear and Stevens, 2000 ; Siniossoglou and Pelham, 2001 ; Conibear et al., 2003 ), we think that the small vesicles correspond to endosome-to-TGN intermediates that are prevented from delivering retrograde cargo to the TGN by the absence of GARP.
Because the CI-MPR retrieved from the cell surface does not simply stay at the TGN but rapidly exits toward endosomes, we sought to examine the effect of GARP depletion on retrograde transport of another cargo that does not follow this cycling pathway. The B-subunit of STxB is often used as a model cargo in studies of endosome-to-TGN transport (Mallard and Johannes, 2003 ). STxB is internalized from the cell surface and then undergoes retrograde transport using clathrin, retromer, and other machinery components common to various recycling cargoes (Tai et al., 2004 , 2005 ; Bujny et al., 2007 ; Popoff et al., 2007 ). Unlike the CI-MPR, however, STxB does not recycle back to endosomes but continues its retrograde path to the endoplasmic reticulum (ER; Mallard and Johannes, 2003 ). We observed that 1 h after internalization of the probe, the majority (~65%) of control cells showed Cy3-conjugated STxB staining in a ribbon-like structure typical of the Golgi complex (Figure 8, A and B). In virtually all Vps52-depleted cells, in contrast, none of the internalized Cy3-conjugated STxB localized to the Golgi ribbon but instead remained in a population of small cytoplasmic vesicles (Figure 8, C and D). Rescue with RNAi-resistant, V5-tagged Vps52 restored STxB transport to the Golgi complex (Figure 8, E and H; ~80% of transfected cells recovered the Golgi pattern for STxB). These observations confirmed that depletion of a GARP subunit blocks retrograde transport to the TGN and results in the accumulation of cargo in a vesicular compartment.
A single point mutation near the C-terminus of Vps54 (L967Q) causes the Wobbler phenotype in the mouse, which is characterized by defects in motor neuron function and male sterility (Schmitt-John et al., 2005 ). Complete ablation of the Vps54 gene, however, is lethal at day 12.5 of embryonic development (Schmitt-John et al., 2005 ). This indicates that the Wobbler mutation is hypomorphic. To investigate the effects of this mutation on the function of the GARP complex, we initially examined the assembly of V5-tagged Vps54(L967Q) in transfected cells. We observed that, although this mutant protein was expressed at levels lower than those of the wild-type protein, it nonetheless assembled with Vps52 and Vps53 (Figure 9A). We took advantage of this assay to determine what part of Vps54 assembled with the other two subunits. To this end, we expressed V5-tagged forms of Vps54(1-515) and Vps54(535-977), comprising the N- and C-terminal portions of Vps54, respectively. We found that it was the N-terminal part that assembled into the GARP complex (Figure 9A), a finding that was consistent with the ability of the C-terminally mutated Wobbler allele to be incorporated into the complex. Moreover, immunofluorescence microscopy analysis of cells transfected with GFP-tagged Vps54(L967Q) Wobbler mutant showed that this protein localized to the TGN (Figure 9, E–G). Finally, transfection of Vps54-depleted cells with RNAi-resistant Vps54(L967Q)-V5 or Vps54(L967Q)-GFP restored the normal distribution of CatD, TGN46, and CI-MPR within cells (Figure 9, H–O). Thus, the Vps54 Wobbler mutant appears to retain activity in trafficking between endosomes and the TGN.
The results presented here indicate that the mammalian GARP complex is required for the trafficking of CatD to lysosomes by enabling the recycling of the CI-MPR from endosomes to the TGN. In the absence of GARP, the CI-MPR accumulates in a population of small vesicles downstream of endosomes. The ensuing depletion of MPRs from the TGN causes newly synthesized CatD to be released into the extracellular medium as an uncleaved precursor instead of being sorted to endosomes and then to lysosomes. Loss of CatD, and likely of other mannose 6-phosphate–modified hydrolases, results in swelling of the lysosomes, a phenotype similar to that of lysosomal storage disorders.
Where does GARP act in this process? A previous study had suggested a mostly endosomal localization for mammalian GARP (Liewen et al., 2005 ). However, our analyses of H4 and HeLa cells expressing Vps54-GFP indicate that GARP is mainly associated with the TGN (Figures 3 and and9,9, B–D). It is then at this location that GARP must participate in CI-MPR trafficking. On the basis of the proposed function of yeast GARP (Conibear and Stevens, 2000 ; Siniossoglou and Pelham, 2001 , 2002 ; Whyte and Munro, 2002 ; Conibear et al., 2003 ; Panic et al., 2003 ), we think that mammalian GARP is involved in the tethering or docking of endosome-derived, retrograde transport carriers to the TGN. This would be followed by SNARE-mediated fusion and delivery of the CI-MPR into the TGN. Indeed, depletion of GARP causes a shift in the steady-state distribution of the CI-MPRs from a set of relatively large endosomal structures to myriad small vesicles scattered throughout the cytoplasm (Figure 4, E, I, and L). A similar accumulation in small vesicles is observed for CI-MPR internalized from the plasma membrane (Figure 7, D and J), indicating that these vesicles lie in the retrograde transport pathway. The small vesicles are lighter than endosomes or the Golgi complex (Figure 6) and lack markers of either of these organelles (Figure 4). Importantly, they do not contain SNX2 (Figure 7E), a component of the retromer complex that initiates the retrieval of the CI-MPR from endosomes by diverting the receptor into recycling tubules. Moreover, internalized CI-MPR passes through SNX2-positive endosomes before accumulating in the small vesicles (Figure 7, M–O). This places the small vesicles past the tubular endosomal network (TEN; Bonifacino and Rojas, 2006 ) through which the CI-MPR transits en route to the TGN (Arighi et al., 2004 ). Thus, the small vesicles are likely intermediates in the transport between the TEN and the TGN.
Interestingly, the levels of CI-MPR are not changed by depletion of GARP despite its altered distribution (Figure 4P). This is in contrast to the depletion of retromer, which results in lower levels of CI-MPR due to its diversion to lysosomes (Arighi et al., 2004 ). This indicates that the small vesicles where the CI-MPR accumulates in the absence of GARP are past the point where default transport to lysosomes is possible.
The role of GARP in retrograde transport is not limited to CI-MPR trafficking because the recycling TGN protein, TGN46, and the bacterial toxin, STxB, are also prevented from reaching the TGN in GARP-depleted cells. In these cells, both TGN46 (Figure 4F) and internalized STxB (Figure 8, C and D) accumulate in small vesicles similar to those that contain the CI-MPR. However, some differences with the behavior of the CI-MPR are apparent. The total levels of TGN46 are decreased (Figure 4P). In addition, STxB also accumulates in larger structures that colocalize with endosomal markers (Figure 8 and data not shown). This suggests that some cargo proteins back up into endosomal compartments in the absence of GARP. Despite these differences, it is clear that GARP is required for the retrograde transport of different types of protein: recycling transmembrane proteins like the CI-MPR and TGN46, and a glycosphingolipid-binding luminal protein like STxB. GARP thus appears to function as a general mediator of retrograde transport to the TGN.
Previous studies have identified other proteins that function to tether retrograde transport intermediates to the TGN. Among these are the golgins, golgin-97 (Lu et al., 2004 ), golgin-245 (Yoshino et al., 2005 ), GCC88 (Lieu et al., 2007 ), and GCC185 (Reddy et al., 2006 ; Derby et al., 2007 ). It is currently unclear why so many tethering factors would be involved in retrograde transport. One possibility is that they all cooperate to dock the same set of retrograde transport carriers to the TGN. An alternative possibility is that each participates in the docking of a different type of carrier, as defined by its origin or cargo. For example, GCC185 participates, together with Rab9 and TIP47, in retrieval of CI-MPR specifically from late endosomes (Reddy et al., 2006 ). Another variation is exemplified by GCC88, which participates in retrograde transport of CI-MPR and TGN38 (the rat ortholog of human TGN46), but not STxB (Lieu et al., 2007 ). This is consistent with the existence of multiple routes and carriers for retrograde transport. To the extent that we have analyzed it, the role of GARP appears to be general to various cargo proteins.
Although most of the CI-MPR accumulates in small vesicles in GARP-depleted cells, a fraction appears to concentrate in a juxtanuclear structure that colocalizes with Golgi markers (Figure 4, I–K). This may indicate that some CI-MPR molecules are still delivered to the TGN in the absence of GARP, perhaps due to the action of the other tethering factors mentioned above. Alternatively, the juxtanuclear remnant may reflect a certain degree of inhibition of exit from the Golgi complex. Indeed, after prolonged (≥5 d) depletion of GARP, we observed that even the plasma-membrane–targeted VSV-G protein accumulates to some extent in the Golgi complex. This could point to an additional role of GARP in export from the Golgi complex. In this regard, interference with another tethering factor, golgin-97, has been shown to inhibit transport of the adhesion molecule, E-cadherin, from the Golgi complex to the basolateral surface of polarized epithelial cells (Lock et al., 2005 ). However, accumulation in the Golgi complex could also be secondary to impaired retrieval of factors that are required for exit from the TGN, as has been previously proposed for yeast GARP (Conibear and Stevens, 2000 ).
As would be expected for a complex that plays a general role in retrograde transport, GARP is essential for embryonic development and viability in the mouse (Schmitt-John et al., 2005 ). Mouse embryos with homozygous disruption of the Vps54 gene fail to thrive and die at about day 12.5 postcoitum (Schmitt-John et al., 2005 ). However, the mutant Wobbler mouse, which carries the missense mutation L967Q in Vps54 (Schmitt-John et al., 2005 ), is viable though it exhibits motor neuron degeneration similar to that of amyotrophic lateral sclerosis (Boillee et al., 2003 ). The different phenotypes of the Vps54 disruption and Wobbler mutants are likely explained by the ability of the Vps54(L967Q) mutant protein to assemble with the other subunits of GARP and to support sorting of CI-MPR, CatD, and TGN46 (Figure 9). The expression levels of Vps54(L967Q), however, are lower than those of its wild-type counterpart (Figure 9A), perhaps explaining the Wobbler motor neuron defect.
The demonstration of a role of the human GARP complex in retrograde transport further supports the notion that the core machinery for acid hydrolase sorting has been faithfully conserved from yeast to humans, to the point of utilizing similar proteins or complexes at virtually every step of the sorting pathways. This conservation undoubtedly stems from the essential nature of endosomal transport pathways and, in particular, retrograde transport from endosomes to the TGN for the maintenance of cellular homeostasis.
We thank X. Zhu and H.-I. Tsai for expert technical assistance, L. Johannes for Cy3-STxB, M. Marks (University of Pennsylvania, Philadelphia, PA) for antibody to p230, and J. G. Magadán for critical review of the manuscript. This work was funded by the Intramural Program of National Institute of Child Health and Human Development, National Institutes of Health. F.J.P-V. is the recipient of a postdoctoral fellowship from the Ministerio de Educación y Ciencia of Spain.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E07-11-1189) on March 26, 2008.