|Home | About | Journals | Submit | Contact Us | Français|
Specialized cells exploit adaptor protein complexes for unique post-Golgi sorting events, providing a unique model system to specify adaptor function. Here, we show that AP-3 and AP-1 function independently in sorting of the melanocyte-specific protein tyrosinase from endosomes to the melanosome, a specialized lysosome-related organelle distinguishable from lysosomes. AP-3 and AP-1 localize in melanocytes primarily to clathrin-coated buds on tubular early endosomes near melanosomes. Both adaptors recognize the tyrosinase dileucine-based melanosome sorting signal, and tyrosinase largely colocalizes with each adaptor on endosomes. In AP-3-deficient melanocytes, tyrosinase accumulates inappropriately in vacuolar and multivesicular endosomes. Nevertheless, a substantial fraction still accumulates on melanosomes, concomitant with increased association with endosomal AP-1. Our data indicate that AP-3 and AP-1 function in partially redundant pathways to transfer tyrosinase from distinct endosomal subdomains to melanosomes and that the AP-3 pathway ensures that tyrosinase averts entrapment on internal membranes of forming multivesicular bodies.
Sorting of integral membrane proteins among post-Golgi organelles is facilitated by cytoplasmic coats, including the ubiquitously expressed, heterotetrameric adaptor protein (AP) complexes. By binding cytoplasmic sorting signals on cargo proteins, AP complexes recruit cargo to patches on donor membranes that bud to form vesicles or tubules destined to fuse with target membranes (Bonifacino and Traub, 2003 ; Robinson, 2004 ). Genetic analyses have shown that whereas deficiency of AP-1 or AP-2 results in embryonic lethality, AP-3 deficiency is not lethal and results in severe phenotypes only in specific tissues (reviewed in Boehm and Bonifacino, 2002 ). Thus, the transport pathways in which AP-3 functions are likely redundant in most cell types but necessary for unique events in certain cells. Defining these pathways and those involving the essential APs will be necessary to fully understand endosomal maturation and the formation of tissue-specific organelles.
How and where AP-3 functions in “nonspecialized” cells is controversial. Yeast AP-3 binds to cytoplasmic dileucine-like sorting signals of cargo, such as alkaline phosphatase and Vam3p, and facilitates their biosynthetic traffic to the vacuole in a pathway that bypasses the prevacuolar compartment, analogous to the mammalian multivesicular late endosome/multivesicular body (MVB) (reviewed in Burd et al., 1998 ). Mammalian AP-3 also binds preferentially in vitro to dileucine- and tyrosine-based sorting signals of lysosomal transmembrane proteins (Bonifacino and Traub, 2003 ; Janvier et al., 2003 ), implying a function in lysosomal sorting (Odorizzi et al., 1998 ; Robinson and Bonifacino, 2001 ). Consistent with this function, a subset of lysosomal membrane proteins in AP-3-deficient fibroblasts undergo increased recycling via the cell surface (reviewed in Starcevic et al., 2002 ). Whereas some evidence may support a role for AP-3 in direct transport from the trans-Golgi network (TGN) to lysosomes (Rous et al., 2002 ; Ihrke et al., 2004 ), other data are most consistent with a primary role in sorting from early endosomes. AP-3-containing coats have been localized not only to the TGN (Simpson et al., 1997 ) but also more prominently to endosomal membranes in a number of cell types (Simpson et al., 1996 ; Dell'Angelica et al., 1997 , 1999 ; Simpson et al., 1997 ) and in particular to tubulovesicular domains of early endosomes in HepG2 cells; in these cells, AP-3 is thought to divert membrane proteins such as lysosomal-associated membrane protein (LAMP)-1 away from recycling to the cell surface and toward lysosomes (Peden et al., 2004 ). However, loss of AP-3 function does not affect the gross distribution of LAMP-1 and other cargo to their target late endosome/lysosome destination in HepG2 and fibroblast systems. The dispensability of AP-3 function in these cells for late endosome/lysosome localization has made it difficult to dissect the interrelationships among AP-3-dependent and -independent sorting and the pathways in which they participate.
AP-3-deficient animals are characterized by severe defects in organelle function in specific tissues, including dense granules in platelets, lytic granules in cytotoxic T-cells, a subset of synaptic vesicles in some neurons, and melanosomes of pigment cells (Starcevic et al., 2002 ; Stinchcombe et al., 2004 ; Seong et al., 2005 ). These phenotypes imply that AP-3-dependent sorting is uniquely required for the biogenesis of specialized structures, collectively referred to as lysosome-related organelles (LROs) (Dell'Angelica et al., 2000 ; Blott and Griffiths, 2002 ). LROs thus provide a context in which AP-3 function is best defined. A candidate LRO cargo for AP-3-dependent sorting is tyrosinase, a type I integral membrane protein whose enzymatic activity mediates the limiting step in melanin biosynthesis within melanosomes. A dileucine-based signal within the cytoplasmic domain of tyrosinase binds in vitro to AP-3 (Honing et al., 1998 ) and is required for efficient sorting in nonpigment cells (Blagoveshchenskaya et al., 1999 ; Calvo et al., 1999 ; Simmen et al., 1999 ; Wang et al., 1999 ). Moreover, tyrosinase is mislocalized in melanocytes derived from AP-3-deficient Hermansky-Pudlak Syndrome patients (Huizing et al., 2001 ). Although these data imply a direct role for AP-3 in sorting tyrosinase to melanosomes, several questions remain unanswered. Does AP-3 mediate transport from the TGN or from endosomes? Because AP-3-deficient melanocytes are not totally devoid of mature melanosomes (Nguyen et al., 2002 ), can an alternative AP-dependent pathway direct tyrosinase to the melanosome? A candidate is AP-1, recently implicated in the biogenesis of another LRO, the Weibel-Palade body (Lui-Roberts et al., 2005 ). Finally, given the unique dynamics of the endosomal system in melanocytes (Raposo et al., 2001 ), are AP functions in sorting to melanosomes related to those in nonspecialized cell types? To address these questions, we have quantitatively localized AP-3, AP-1, and tyrosinase within melanocytes and defined how AP-3-deficiency affects tyrosinase transport. The results indicate that tyrosinase traffics through the early endocytic system and that mammalian AP-3 functions on early endosomes to divert tyrosinase away from a pathway leading to multivesicular endosomes and toward one leading to melanosomes. Moreover, they reveal a dynamic interplay between AP-3 and AP-1 in endosomal sorting to provide alternative routes to the melanosome.
Chemicals and tissue culture reagents were from Sigma-Aldrich (St. Louis, MO) unless otherwise noted. Monoclonal antibodies (mAbs), their targets, and their sources were as follows: TA99 (Mel-5) to Tyrp1 (American Type Culture Collection, Manassas, VA); 100/3 to γ-adaptin and 100/2 to α-adaptin (Sigma-Aldrich), monoclonal antibody to δ-adaptin (Peden et al., 2004 ), anti-clathrin heavy chain (BD Transduction Laboratories, San Diego, CA); and H68.4 to transferrin (Tf) receptor (Zymed Laboratories, South San Francisco, CA). Polyclonal rabbit antisera to δ-μ3-, and β3A-adaptins (Simpson et al., 1997 ); σ3-adaptin (Simpson et al., 1996 ; Dell'Angelica et al., 1997 , 1999 ; Simpson et al., 1997 ); Pmel17 (αPEP13h; Berson et al., 2001 ); γ-adaptin (Seaman et al., 1996 ); STO 125 (gift from S. Tooze, Cancer Research UK, London, United Kingdom) (Dittie et al., 1996 ); Hrs (Sachse et al., 2002 ) (gift from S. Urbé, University of Liverpool, Liverpool, United Kingdom); and clathrin light chain (gift from F. Brodsky, Department of Biopharmaceutical Sciences, University of California, San Francisco, San Francisco, CA) (Acton et al., 1993 ) have been described previously. Sheep anti-human TGN46 was from Serotec (Raleigh, NC). Rabbit anti-mouse immunoglobulin (Ig)G was from Dakopatt (Glostrup, Denmark); Alexa488- and Alexa594-conjugated secondary antibodies, Tf-fluorescein isothiocyanate (FITC), and anti-FITC antibodies were from Invitrogen (Carlsbad, CA); Tf-horseradish peroxidase (HRP) was from Pierce Chemical (Rockford, IL); HRP-conjugated anti-mouse and anti-rabbit IgG were from GE Healthcare (Little Chalfont, Buckinghamshire, United Kingdom); and FITC- and Texas Red-conjugated isotype-specific secondary antibodies were from Southern Biotechnology (Birmingham, AL). Protein A conjugated to 10- or 15-nm gold particles was from Cell Microscopy Center (AZU, Utrecht, The Netherlands).
MNT-1 human melanoma (Raposo et al., 2001 ) and mouse melan-a cells (Bennett et al., 1987 ) were maintained as described previously. MNT-1 cells (5 × 106 per transfection) were transfected by electroporation using the Equibio Optimix kit (Ozyme, St. Quentin Yvelines, France) and a Bio-Rad electroporation unit (Bio-Rad, Hercules, CA) according to the manufacturers' instructions with a single pulse (140 V, 3.25 μF and infinite resistance). Transfected cells were reseeded on coverslips and cultured for 24 or 48 h. Melan-pe cells were obtained by immortalization of primary mouse melanocytes from B6Pin.C3-Ap3b1pe/Ap3b1pe (pearl) mice (The Jackson Laboratory, Bar Harbor, ME) according to published protocols (Sviderskaya et al., 1997 ) (available at http://www.sghms.ac.uk/depts/anatomy/pages/protocols/msmelpri.html).
HRP-tyrosinase chimerae HRP-tyr and HRP-tyrAA (Blagoveshchenskaya et al., 1999 ) and GAL4AD fusions to γ1, αC, δ, and ε-adaptins in pGADT7 (BD Biosciences Clontech, Palo Alto, CA) (Bonifacino and Traub, 2003 ; Janvier et al., 2003 ) have been described previously. GAL4BD fusion constructs were made by PCR amplification from plasmids encoding mouse tyrosinase (31 C-terminal residues), Tyrp1 (35 C-terminal residues), and Tyrp2 (29 C-terminal residues) and in-frame cloning into the EcoRI-PstI (tyrosinase, Tyrp1) or XmaI-PstI sites (Tyrp2) of pBridge (BD Biosciences Clontech). Mutants of GAL4BD-Tyrosinase were obtained by PCR-directed mutagenesis. The σ AP subunits were amplified by PCR and subcloned into the Not I-BglII sites of pBridge-Tyrosinase, pBridge-Typr1 and pBridge-Tyrp2. GST-Tyrosinase (wild-type; WT) was obtained by in frame cloning of sequence encoding the 31 C-terminal residues of mouse tyrosinase into the EcoRI site of pGEX-5X-1 (GE Healthcare), and the mutants were generated by PCR-directed mutagenesis. Production and purification of GST fusion proteins, GST pull-downs, and yeast three-hybrid assays were performed as described previously (Bonifacino and Traub, 2003 ; Janvier et al., 2003 ).
To prepare GST fusion proteins, a fragment encoding lumenal residues 19-476 of mouse tyrosinase was generated by PCR using pTyrs-J template (gift from H. Yamamoto, Tohoku University, Sendai, Japan) (Yamamoto et al., 1987 ), and subcloned into BamHI/EcoRI-cut pGEX-3X (GE Healthcare) to generate pGST-LD. Fusion protein was expressed in XL-1 Blue cells by induction for 5-7 h at 37°C with 0.1 mM isopropyl β-d-thiogalactoside, and purified from inclusion bodies as described previously (Page and Robinson, 1995 ). Purified protein (0.5 mg/injection) in phosphate-buffered saline (PBS)/0.2% SDS was injected into rabbits at 2- to 6-wk intervals with Complete Freund's Adjuvant by Murex Biotech (Dartford, Kent, United Kingdom), and serum was collected 10 d after the last injection. Affinity-purified antibodies were prepared using purified GST-LD coupled to beads as described previously (Page and Robinson, 1995 ) and adsorbed against GST bound to Sepharose beads before use.
Cells were fixed in PBS containing 2% formaldehyde for 15-30 min at room temperature (RT) and then labeled with primary and fluorescent secondary antibodies as described previously (Berson et al., 2001 ). Cells were analyzed on a DM IRBE microscope (Leica Microsystems, Wetzlar, Germany) equipped with an Orca digital camera (Hamamatsu, Bridgewater, NJ), and images were captured and manipulated using Improvision OpenLab software (Lexington, MA) with the volume deconvolution package. Immunoprecipitation from lysates of melan-a cells and immunoblotting were essentially as described previously (Berson et al., 2001 ; Peden et al., 2002 ).
Ultracryomicrotomy, Immunogold Labeling, and Analysis. Cells were fixed with 2% (wt/vol) paraformaldehyde (PFA) or with a mixture of 2% (wt/vol) PFA and 0.2% (wt/vol) glutaraldehyde in 0.1 M phosphate buffer (PB), pH 7.4. Cells were processed for ultracryomicrotomy as described previously (Raposo et al., 1997 ). Ultrathin sections were prepared with an Ultracut fetal calf serum ultracryomicrotome (Leica, Vienna, Austria), and single or double immunogold labeled with antibodies and protein A coupled to 10- or 15-nm gold (PAG10 and PAG15). Sections were observed and photographed under a Philips CM120 electron microscope (FEI, Eindoven, The Netherlands). Digital acquisitions were made with a numeric camera Keen View (Soft Imaging System, Munster, Germany).
Internalization of Tf-FITC. Cells were washed with serum-free medium and starved for 45 min before incubation with Tf-FITC (60 μg/ml) for 45 min at 37°C. After washing with ice-cold medium, cells were fixed and processed for ultracryomicrotomy.
Quantification of Immunogold Labeling. The relative distribution of AP-3 and AP-1 in melan-a cells was evaluated by analyzing directly under the electron microscope randomly selected cell profiles from two distinct grids. In total, 450 and 448 gold particles were counted for AP-3 and AP-1, respectively, and assigned to the compartment over which they were located. The definition of the distinct compartments was based on their morphology and their previous characterization by immunogold labeling with different organelle markers (EEA-1 and Hrs for early endosomes; TGN38 and TGN46 for the TGN; and LAMP-1 for late endosomes/lysosomes) and internalization of endocytic tracers (bovine serum albumin-gold and Tf-FITC). Tubulovesicular membranes that were located at the trans-side of the Golgi were considered as TGN. Vacuolar endosomes were defined as electron-lucent vacuoles with no internal membranes or few internal vesicles. MVBs were compartments de-limited by a membrane with numerous internal vesicles. Electron-dense compartments with no or few internal membranes were classified as lysosomes. Clearly visible tubulovesicular membrane profiles connected to or in close association with endosomal vacuoles, melanosomes, MVBs, and lysosomes were also defined. Melanosome stages were defined by morphology (Raposo et al., 2001 ). Quantification of tyrosinase distribution in melan-a and melan-pe cells was performed on single immunogold-labeled cryosections by counting the number of gold particles in each of the defined compartments (see above) taken randomly in 50 cell profiles. In total, 1033 and 1100 gold particles were counted in melan-a and melan-pe cells, respectively. Results are presented as a percentage of the total number of gold particles in each compartment and represent a mean and SD of two independent experiments. The labeling densities of tyrosinase was estimated in a series of double-labeling experiments to detect tyrosinase and AP-3 or AP-1 in melan-a and melan-pe cells. The monoclonal anti-AP-3 and anti-AP-1 were used in the first step. AP-3- and AP-1-positive membranes were selected under the electron microscope, and 70 images of AP/tyrosinase positive areas were taken with a charge-coupled device camera. Using Adobe Photoshop, images at an end magnification of 80,000× were overlaid with a 5-mm-spaced line lattice. The number of gold particles present on AP-3-positive/AP-3-negative and AP-1-positive/AP-3-negative regions of the compartments was counted. The number of intersections of the distinct membrane domains with the 5-mm line lattice overlay was also counted to give a measure of membrane length (Griffiths, 1993 ; Peden et al., 2004 ). The number of gold particles per intersection reflected the labeling density for tyrosinase in the different areas. The percentage of AP-3- or AP-1-positive vesicles and buds labeled for tyrosinase in melan-a and melan-pe cells was determined by counting directly on the electron microscope the total number of AP-3 or AP-1-coated vesicles of cells double immunogold labeled for tyrosinase and AP-3 or AP-1.
Whole-Mount Electron Microscopy (EM). Immunogold labeling of AP-3, AP-1, Clathrin, Hrs, Pmel17 (using αPep13h antibody to the cytoplasmic domain), and Tf receptor on whole-mounted cells was performed as described previously (Stoorvogel et al., 1996 ) on cells grown for 2 d on Formvar-coated gold grids. In brief, after washing with minimal essential medium, 20 mM HEPES, pH 7.4, cells were starved for 45 min in the same medium at 37°C and then incubated for 45 min with 50 μg/ml HRP-Tf. Cells were cooled to 4°C and incubated in 3,3′-diaminobenzidine tetrahydrochloride (DAB)-H2O2-containing buffer. Soluble cytosolic proteins were removed by permeabilization with saponin, and cells were then fixed with 3% PFA in 0.1 M phosphate buffer, pH 7.4, for 2 h. Immunogold labeling was performed at RT in the presence of saponin and using cold fish gelatin as blocking buffer. After labeling, cells were fixed with 2% glutaraldehyde, rinsed in water, dehydrated in increasing concentrations of ethanol, and dried in a critical point dry apparatus (BAL-TEC, Balzers, Lichtenstein) before EM analysis.
Conventional Electron Microscopy and DAB Cytochemistry. MNT-1 cells transfected with HRP-tyr and HRP-tyrAA for 24 or 48 h were fixed with 2% PFA/0.5% glutaraldehyde in 0.2 M PB, pH 7.4, for 90 min. After several washes with 50 mM Tris-HCl, pH 7.6, cells were treated for 20 min with 0.03% DAB in the presence of 1 μl/ml H2O2 (30 vol). Cells were fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer for 90 min, postfixed with 2% OsO4, dehydrated in ethanol, and embedded in Epon. Ultrathin sections were counterstained with uranyl acetate before observation.
DOPA Histochemistry. DOPA histochemistry was carried out as described previously (Boissy et al., 1998 ). The cells were then washed and incubated in 1% osmium tetroxide with 1.5% potassium ferrocyanide in 0.2 M sodium cacodylate buffer for 1 h at RT and then washed before dehydration and embedding for sectioning (Reynolds, 1963 ).
To localize AP-3 and AP-1 in melanocytes, we used immunoelectron microscopy (IEM) with immunogold labeling on ultrathin cryosections and “whole-mounted” cells. We first assessed AP-3 localization in ultrathin cryosections of MNT-1 melanoma cells, a faithful model system for melanosome biogenesis (Berson et al., 2001 ; Kushimoto et al., 2001 ; Raposo et al., 2001 ; Berson et al., 2003 ) and in mouse melan-a immortalized melanocytes. In both MNT-1 and melan-a cells, the bulk of labeling for AP-3 was associated with tubulovesicular clusters of coated vesicles or buds (Supplemental Figure S1, A and B) similar to those described in nonpigmented cells (Peden et al., 2004 ). Most of these clusters were in proximity to vacuolar endosomes and were also often found closely apposed to stage III and IV (i.e., more mature) melanosomes (Figures (Figures1A1A and S1, A and B). To evaluate the connection of the AP-3-positive tubulovesicular clusters to endosomes and melanosomes, we used whole-mount immunogold labeling (Stoorvogel et al., 1996 ). MNT-1 cells were allowed to internalize Tf-HRP conjugate for 1 h to fill early sorting and recycling endosomes, and then they were treated with DAB and hydrogen peroxide, permeabilized, and immunogold labeled. Cross-linked early endosomes thus look like electron-dense vacuolar and tubular domains with numerous buds. The endosomal tubules and vacuoles label with antibodies to the cytoplasmic domain of Pmel17 (Figures (Figures1C1C and S1C), an integral membrane protein that passes through early endosomes en route to stage II premelanosomes (Berson et al., 2001 ; Raposo et al., 2001 ). AP-3 is highly concentrated in many of the buds associated with the endosomal tubules and vacuoles, although the buds themselves lack labeling for Pmel17 (arrows in Figures Figures1C1C and S1C). Moreover, in cryosections of MNT-1 cells that had internalized Tf-FITC for 1 h before fixation, buds labeled for AP-3 but not for FITC emerge from endosomal tubules labeled for FITC (Figure 1D). Whole-mount analyses corroborate the proximity to melanosomes of AP-3-coated buds continuous with Hrs-positive endosomal vacuoles (Figure 1B) and also reveal the presence of clathrin on the AP-3-containing buds (Figure 1E; also see Supplemental Figure S1B). Quantitative analysis revealed that 41% of AP-3 vesicles and buds also contain clathrin in agreement with findings in HepG2 cells (Peden et al., 2004 ). The labeling of the associated tubules for Pmel17 and Tf-FITC confirms their endocytic nature, whereas the lack of labeling for Pmel17 and Tf-FITC within the buds indicates that they fail to enrich early endocytic cargo. We conclude that in melanocytes, AP-3 is primarily localized to clathrin-coated buds on tubulovesicular elements adjacent to endosomal vacuoles and melanosomes.
Coated buds devoid of AP-3 labeling are often observed near AP-3-labeled buds on the same tubulovesicular clusters (Figure 1A). Because AP-3 and AP-1 are both present on buds on tubular endosomes in other cell types (Peden et al., 2004 ), we compared their relative distributions in MNT-1 cells. Both AP-3- and AP-1-coated buds are observed on endosomes, as shown by immunogold labeling on budding profiles emanating from DAB-containing tubules by whole-mount EM (Figure 2, A and B) and by double labeling for AP-1 and FITC in cryosections of cells that had previously internalized Tf-FITC (Figure 2, E and F). Like AP-3-coated buds, AP-1-coated buds are often located in the vicinity of Hrs-positive vacuolar endosomes (Figure 2C) and melanosomes (Figure 2, C and D). Although buds and vesicles containing both adaptors are occasionally observed (Figure 2B, inset; ~2% of total labeling for AP-3 and AP-1), labeling for AP-1 and AP-3 was predominantly exclusive, suggesting that the two AP complexes are largely present on distinct endosomal buds (Figure 2, A, B, and D). Quantitative evaluation of the relative distribution of AP-3 and AP-1 in melan-a cells (Table 1) revealed the highest level of labeling for AP-3 in tubulovesicular elements close to endosomal vacuoles (41.5%) and melanosomes (25.8%). In melanocytic cells the highest labeling for AP-1 is also found close to endosomes (39%). Although more enriched than AP-3 in the TGN area and cytoplasmic vesicles, AP-1 is also significantly enriched in tubulovesicular elements close to melanosomes (19%). IFM analysis of MNT-1 cells corroborates with these findings (Supplemental Figure S2). Together, these data indicate that AP-1 and AP-3 most prominently populate distinct buds on endosomal tubules near vacuolar endosomes and melanosomes.
The close apposition of the bulk of AP-3- and AP-1-labeled endosomes to melanosomes suggests that these complexes may sort melanosomal cargo such as tyrosinase and Tyrp1, as suggested previously (Huizing et al., 2001 ; Raposo et al., 2001 ). We therefore tested whether AP-1 and AP-3 directly interact with the tyrosinase and Tyrp1 sorting signals. Post-Golgi sorting of tyrosinase is mediated by a dileucine signal (EERQPLL in mice, EEKQPLL in humans) (Blagoveshchenskaya et al., 1999 ; Calvo et al., 1999 ; Simmen et al., 1999 ) (Figure 3A), which conforms to the D/ExxxLL/I motif recognized by both AP-1 and AP-3 (Bonifacino and Traub, 2003 ), and Tyrp1 sorting depends on a related sequence (EANQPLL) (Vijayasaradhi et al., 1995 ). Similar signals bind in a yeast three-hybrid assay to the γ1-σ1A and the δ-σ3A or δ-σ3B hemicomplexes of AP-1 and AP-3, respectively (Janvier et al., 2003 ). Using this assay, the tyrosinase cytoplasmic domain interacts with hemicomplexes from AP-1 and AP-3 but not with those of AP-2 and AP-4 (Figure 3B). By contrast, the Tyrp1 cytoplasmic domain binds only weakly to the AP-1 hemicomplex and not at all to AP-3, and the Tyrp2 cytoplasmic domain, which contains only a tyrosine-based motif predicted to bind to AP μ subunits (Figure 3A), did not bind to any hemicomplex (Figure 3B). These data confirm that the tyrosinase cytoplasmic domain binds AP-3 and demonstrate a previously uncharacterized interaction with AP-1.
The interaction of tyrosinase with AP-3 is known to require the conserved, active dileucine-based targeting signal (Honing et al., 1998 ) (Figure 3A). Similarly, the interaction with AP-1 requires the same signal, because mutagenesis of the two conserved leucine residues (L517 and L518) abrogated interactions with both AP-1 and AP-3 hemicomplexes in the three-hybrid assay (Figure 3C). In contrast, mutation of two nonconserved leucine residues (L527 and L528), which are not required for sorting (Simmen et al., 1999 ), had no effect on the interactions (Figure 3C). GST pull-down assays confirmed the interactions of the tyrosinase cytoplasmic domain with the native AP-1 and AP-3 complexes and the dependence of these interactions on leucine residues 517 and 518 but not those at 527 and 528 (Figure 3D). These data confirm that similar sequences are required to direct both sorting of tyrosinase and binding not only to AP-3 but also to AP-1.
Tyrosinase enzyme activity in melanocytes is detected primarily in mature melanosomes (see below; Seiji et al., 1963 ; Yamamoto and Takeuchi, 1981 ). If localization to melanosomes requires binding to AP-1 and AP-3 in endosomes, then a mutant form lacking the dileucine-based sorting signal should accumulate within the donor endosomal compartments. To test this prediction, we expressed in MNT-1 cells a fusion protein containing HRP in the lumenal domain and the tyrosinase cytoplasmic domain without (HRP-tyr) or with (HRP-tyrAA) a dialanine substitution in the dileucine motif (Blagoveshchenskaya et al., 1999 ). Although not quantitative, the very sensitive peroxidase cytochemical reaction reveals even very low levels of HRP trafficking through an organelle, thus enabling evaluation of the total distribution of the HRP-tyrAA chimera. EM analysis of DAB-treated MNT-1 cells expressing HRP-tyr revealed accumulation of electron-dense DAB reaction product primarily in endosomal tubules and melanosomes, consistent with tyrosinase immunolocalization in mouse melan-a cells (our unpublished data; see below). By contrast, cells expressing HRP-tyrAA accumulate fusion protein primarily in tubular and vacuolar endosomes containing associated buds (Figure 4A) with very little or no accumulation in either the Golgi/TGN area (Figure 4A) or in melanosomes (Figure 4C). HRP-tyrAA also accumulates at the plasma membrane (Figure 4B), as expected for a protein undergoing rapid endocytic recycling. DAB electron-dense deposit was also sometimes observed as clusters in the lumen of compartments that are likely to correspond to MVBs, perhaps reflecting a failure of the mutant protein to be sorted away from a lysosomal destination (Figure 4C). Thus, the distribution of a mutant lacking AP-1 and AP-3 binding sites is consistent with a role for either or both APs in facilitating transport of tyrosinase from endosomes toward melanosomes
To determine whether tyrosinase normally transits through the endocytic pathway, we analyzed quantitatively the steady-state distribution of tyrosinase within melanocytes. Whereas previous analyses have relied on histochemical detection of tyrosinase activity, quantitative analyses require nonenzymatic analyses such as immunogold labeling. Using bacterially expressed GST fusion proteins containing the lumenal domain of mouse tyrosinase, we prepared rabbit polyclonal antisera suitable for IEM analyses of tyrosinase. By immunoblotting, these antisera recognize a specific band of ~70 kDa in whole-cell lysates of melan-a mouse melanocytes but not human melanocytic cells (Supplemental Figure S3A). We therefore used these antisera to analyze tyrosinase localization in mouse melanocytes. Immunofluorescence microscopy and deconvolution analyses of primary mouse melanocytes or melan-a cells with this antibody reveal a vesicular pattern that only partially overlaps with that of other markers of immature (stage II; Pmel17) and mature (stage III and IV; Tyrp1) melanosomes (Supplemental Figure S3B). This partial overlap is consistent with observations using antisera to the tyrosinase cytoplasmic domain (our unpublished data) and may reflect both a steady-state localization to nonmelanosomal compartments (such as endosomes) in addition to melanosomes, and technical limitations of immunofluorescence microscopy for detecting tyrosinase residing in mature melanosomes.
To quantitatively evaluate tyrosinase localization, the antibodies were used for immunogold labeling of ultrathin cryosections of melan-a cells. As shown in Figure 5A and Table 1, tyrosinase is detected primarily in stage III and IV melanosomes, consistent with histochemical analyses for tyrosinase activity using DOPA as a substrate (Yamamoto and Takeuchi, 1981 ; Zhao et al., 1994 ; Boissy et al., 1998 ; also see Supplemental Figure S4D). Stage II premelanosomes with characteristic internal striations but no pigment are not labeled. Tyrosinase is present predominantly at the limiting membrane of stage III and IV melanosomes, although significant labeling is also often detected on membranes in the lumen of less pigmented stage III melanosomes (Figure 5A and Supplemental Figure S4A). Besides melanosomes, some labeling for tyrosinase, like that for Tyrp1 (Vijayasaradhi et al., 1995 ; Raposo et al., 2001 ) and consistent with previous histochemical analyses (reviewed in Jimbow et al., 2000 ), is detected in the Golgi and in vesicles in the TGN area (Figure 5B). This labeling is seen primarily on noncoated membranes that are distinct from those labeled for the TGN resident protein TGN38 (our unpublished data). Interestingly, more substantial tyrosinase labeling is apparent in endosomal vacuoles and particularly within tubular extensions close to the vacuoles (Figure 5C and Table 2) and in tubules and vesicles closely apposed to melanosomes (Figure 5D). MVBs are not substantially labeled (Supplemental Figure S4C). Quantitative evaluation of the immunogold labeling in the different compartments (Table 2) emphasizes the prevalence of tyrosinase accumulation at steady state within endosomal vacuoles and closely apposed tubules.
To determine whether the tubulovesicular cohort of tyrosinase in melan-a cells was enriched in AP-3-and AP-1-coated regions, ultrathin cryosections were double immunogold labeled. As shown in Figure 6, AP-3-containing tubules and vesicles near both melanosomes and endosomes are also labeled with anti-tyrosinase antibodies (Figures 6, A and B, B inset, and S4A), whereas the bulk of tyrosinase labeling in the TGN area is on uncoated membranes that lack AP-3 (Figure 6C). To evaluate the concentration of tyrosinase in AP-3-positive tubulovesicular elements, we determined the labeling densities of this cargo protein in the different defined areas (Table 3). The enrichment of tyrosinase in AP-3-positive areas was then obtained by dividing the labeling densities in AP-3-positive by those obtained in AP-3-negative areas. These data show that tyrosinase is concentrated in AP-3-coated membranes essentially of endosomal origin.
Consistent with the biochemical interaction of AP-1 and tyrosinase, IEM analysis revealed that tyrosinase is also present in clathrin-coated buds and vesicles that contain AP-1 (Figure 6D). Many of these 80-nm vesicles are close to endosomal tubules and thus resemble the AP-1-coated structures labeled by Tf-FITC in MNT-1 cells (Figure 6D). As described above, in the TGN area, tyrosinase does not seem to be consistently present in coated vesicles (Supplemental Figure S4B). Estimation of the labeling densities for tyrosinase in AP-1-positive and -negative membranes indicated that despite a higher enrichment in AP-3-positive areas, tyrosinase is also consistently concentrated in AP-1-coated elements (Table 3). Therefore, we conclude that the bulk of tyrosinase associates with both AP-1- and AP-3-coated endosomal domains in melanocytes.
AP-1 deficiency is embryonically lethal, precluding an assessment of tissue-specific AP-1 functions (Zizioli et al., 1999 ). By contrast, AP-3-deficient mice are viable, but they have defective pigment granule function. If both AP-1 and AP-3 participate in tyrosinase sorting from endosomes, then tyrosinase trafficking should be only partially impaired in the absence of AP-3. To test this prediction, we analyzed the subcellular distribution of tyrosinase in AP-3-deficient melanocytic cells. The immortalized melanocyte cell line melan-pe was derived from homozygous pearl (Ap3b1pe/Ap3b1pe) mice, which bear a truncating mutation in the AP-3 β3A subunit (Feng et al., 1999 ). As shown by immunoblotting, melan-pe cells lack functional AP-3 (note the disappearance of the μ3 and δ subunits in anti-δ immunoprecipitates; Figure 7B), but they contain levels of tyrosinase comparable with those of melan-a cells (Figure 7A). Although the gross distribution of tyrosinase (or other melanosomal markers) in melan-pe cells seems similar to that in melan-a cells by immunofluorescence microscopy analyses (Supplemental Figure S3C), IEM analysis of ultrathin cryosections shows an abnormal accumulation of tyrosinase primarily in enlarged vacuolar endosomes (arrows) with abundant emanating tubules as well as in MVBs (stars) (Figure 7, C and D, and Table 2). The vacuolar compartments are labeled for Hrs, defining them as endosomes (Figure 7C and inset). Much of the labeling for tyrosinase on MVBs coincides with intralumenal vesicles (ILVs), contrasting with the nearly exclusive labeling of the limiting membrane of endosomes in melan-a cells (compare Figure 7D with Figure 5). Similar enlarged vacuolar structures (Figure 7E) and nearby tubulovesicular elements (Figure 7F) are labeled by DOPA histochemistry in standard thick sections of primary melanocytes cultured from homozygous pearl mice, and they are absent in WT controls (Supplemental Figure S4D). These data indicate that in the absence of AP-3, tyrosinase accumulates in endosomal structures, including vacuoles, tubules, and ILVs of MVBs.
Despite the significant redistribution of tyrosinase, melan-pe cells are pigmented and harbor melanosomes that are indistinguishable from those of WT cells as observed by DOPA histochemical staining and standard EM analyses (Figures (Figures7E7E and and8A).8A). Moreover, although the distribution of tyrosinase in melan-pe cells largely resembles that of dileucine signal-deficient HRP-tyrAA in MNT-1 cells, it differs in that a reduced but significant fraction of tyrosinase in melan-pe cells remains associated with melanosomes (Table 2), and the cohort of tyrosinase near endosomal vacuoles is enriched in vesicles with clearly visible coats (Figure 7G). This suggests that although AP-3 is responsible for much of tyrosinase transport from tubulovesicular endosomes toward melanosomes, the dileucine signal may additionally drive tyrosinase toward a second AP-3-independent pathway, perhaps mediated by AP-1. Consistent with this prediction, double immunogold labeling of melan-pe cells revealed abundant AP-1-coated vesicles and tubulovesicular elements containing tyrosinase near endosomal vacuoles (Figure 8B). As in melan-a, the AP-1-coated tyrosinase-containing vesicles in melan-pe cells were often observed close to melanosomes (Figure 8C). Quantitation of the labeling reveals that the percentage of AP-1-coated buds and vesicles containing tyrosinase in melan-pe is more than double that in melan-a (49 vs. 22%; Table 4). The labeling density of tyrosinase in AP-1-positive and -negative membranes (see Materials and Methods) was 0.46 and 0.06 gold particles/intersection, respectively, which corresponded to a relative enrichment of 7.66 (compared with 5.28 in melan-a cells; Table 3). These observations suggest that tyrosinase is normally sorted within the melanocyte by both AP-1- and AP-3-dependent mechanisms; that in the absence of AP-3, AP-1 “sees” a greater proportion of tyrosinase that would otherwise be preferentially recruited to nascent AP-3-coated domains; and that AP-1 provides an alternative route to the melanosome independent of that mediated by AP-3.
Cells harboring LROs use ubiquitously expressed machinery to effect cell type-specific sorting events. We show here that AP-1 and AP-3 both associate on endosomes with tyrosinase and mediate its delivery to melanosomes in distinct but partially redundant pathways. These pathways provide for transport out of early endosomes before or concomitant with endosomal maturation and the generation of MVBs and early stage melanosomes from them. In melanocytes, this mechanism of endosomal export facilitates delivery of tyrosinase at the proper stage during melanosome maturation and “rescues” tyrosinase from being incorporated into maturing MVBs carrying lysosomal cargo. Our data shed light on the trafficking pathways followed by tyrosinase, a key enzyme in melanogenesis, toward melanosomes. Moreover, they have important implications for the general function of both AP-3 and AP-1 in sorting to lysosomes and LROs.
We show here using quantitative, high-resolution EM techniques that the main site of accumulation of AP-3 in melanocytic cells is on clathrin-coated buds of early endosomes through which Tf receptors and Pmel17 cycle. Furthermore, loss of AP-3 association, either by mutation of the AP-3-binding signal in HRP-tyrAA or by AP-3 deficiency, correlates with increased accumulation of the AP-3 cargo protein tyrosinase in endosomes. A smaller but significant pool of AP-3 was observed on uncoated membranes near the TGN in melanocytes (our unpublished data), but it was not possible to determine whether these membranes were derived from the TGN or from nearby endocytic tubules. Our results support a quantitative study of AP-3 localization in HepG2 cells (Peden et al., 2004 ) and suggest that 1) the major site of AP-3-dependent protein sorting is at early endosomes, and 2) AP-3 functions to rescue proteins from a default route toward late endosomal MVBs and to sort them toward lysosomes or LROs.
Tyrosinase activity has long been documented in clathrin-coated structures (Orlow, 1998 ). We show here using quantitative immunogold labeling of mouse tyrosinase that although a fraction can be observed in tubulovesicular structures associated with the Golgi, a much larger fraction is present in clathrin-enriched regions of tubulovesicular endosomes. Many of these structures are near mature melanosomes and are coated with AP-1 or AP-3. These endosomal domains likely represent intermediates in the AP-3- and AP-1-dependent trafficking of tyrosinase toward melanosomes. This is supported for AP-3 by the accumulation of tyrosinase in endocytic compartments in AP-3-deficient melanocytes from congenic pearl mice, shown here, and from human Hermansky Pudlak Syndrome patients (observed by cytochemical EM; Huizing et al., 2001 ), concomitant with a depletion from melanosomes. A direct AP-3-dependent endosome-to-melanosome pathway would explain the selective defect in melanosome biogenesis observed in AP-3-deficient animals and would predict that AP-3 may sort cargo directly from endosomes to LROs in other tissues affected by AP-3 deficiency. Why AP-3 cargo is sorted to melanosomes and not to lysosomes in melanocytes is not yet understood, but it may depend upon cell type-specific distribution of target endosomal SNARE proteins or their regulators, many of which are up-regulated in melanocytes (Wade et al., 2001 ; our unpublished data). Our results highlight the importance of the endocytic pathway in the biogenesis of melanosomes.
Interestingly, by contrast to the enrichment of lysosomal cargo in AP-3-coated buds of HepG2 cells (Peden et al., 2004 ), we were unable to detect an enrichment of the late endosome/lysosome resident LAMP-1 in AP-3-coated buds of melanocytic cells (our unpublished data). This supports potential cell type specificity in cargo trafficking patterns and the incomplete dependence of LAMP-1 sorting on AP-3 (reflected by the steady-state localization of LAMP-1 to lysosomes in AP-3-deficient cells; Starcevic et al., 2002 ) and by the steady-state redistribution of LAMP-1 upon depletion of AP-2, but not AP-3 or AP-1, in HeLa cells; Janvier and Bonifacino, 2005 ). Indeed, the trafficking pathways followed by LAMP-1 in melanocytes have yet to be addressed and are beyond the scope of the current study.
Tyrosinase accumulated abnormally in AP-3-deficient cells on vacuolar endosomes and ILVs of MVBs. This further suggests that AP-3 functions to divert cargo from entering into the “classical” Hrs/ESCRT-dependent MVB pathway in mammalian cells (Babst, 2005 ), a function that would be analogous to that of yeast AP-3. Melanocytes might profit from such a mechanism by 1) ensuring that tyrosinase is removed from the endosomal precursors of stage II melanosomes (Berson et al., 2001 ; Raposo et al., 2001 ) and delivered to stage II melanosomes only after they are fully formed, and 2) preventing tyrosinase partitioning to ILVs within these endosomal precursors and its subsequent default delivery to late endosomes. The missorting of tyrosinase to endosomal ILVs in AP-3-deficient melanocytes may reflect its tendency to associate with invaginating membranes in later compartments, supported by the presence of tyrosinase on internal membranes in stage III melanosomes. A similar AP-3-dependent mechanism in cell types that lack LROs may, by allowing cargo proteins to bypass ESCRT-dependent ILV association in earlier endocytic compartments, provide a means to either avoid ILVs altogether or follow an alternative route to ILVs during later endosome maturation. Thus, the melanosome system may reveal a general property of AP-3-dependent sorting events.
Using yeast three-hybrid and GST pull-down assays, we show for the first time that the cytoplasmic dileucine-based sorting signal of tyrosinase binds not only to the δ-σ3 hemicomplexes of AP-3, confirming an AP-3 interaction (Honing et al., 1998 ) but also to the γ1-σ1A hemicomplex of AP-1. Dual binding to AP-3 and AP-1 has also been observed for dileucine-based signals of HIV-1 Nef, LimpII, and CLN3 (Janvier et al., 2003 ; Kyttala et al., 2005 ). The interaction of the tyrosinase and LimpII signals with AP-1 by yeast three-hybrid assays was not observed in surface plasmon resonance analyses (Honing et al., 1998 ), but it is supported by GST pull-down assays (Figure 3B; Janvier et al., 2003 ) and by earlier biochemical data (Kobayashi et al., 1994 ) in which tyrosinase was enriched in clathrin-coated vesicle fractions that are typically devoid of AP-3 (Dell'Angelica et al., 1997 ; Simpson et al., 1997 ). The importance of this interaction was confirmed by the extensive colocalization of tyrosinase on endosomal buds with AP-1 in melan-a cells, by the dramatic increase in the degree of AP-1/tyrosinase colocalization in AP-3-deficient melanocytes, and by the more dramatic effect on steady state distribution of the dileucine mutation (in HRP-tyr) relative to AP-3 depletion. Note that whereas the tyrosinase cytoplasmic domain shows an apparently stronger interaction with the γ1-σ1A hemicomplex than with the δ-σ3 hemicomplexes in the yeast three-hybrid assay, this may not reflect a stronger in vivo interaction because the assay is sensitive to the degree of folding of hemicomplex subunits.
AP-1-containing endosomal buds were largely nonoverlapping with nearby buds containing AP-3, thus defining two separate adaptor-enriched domains on endosomes. The data suggest that AP-1 participates normally in an AP-3-independent physiological sorting pathway for tyrosinase, and that this pathway becomes the “default” in the absence of AP-3. The existence of such a pathway may explain the continued delivery of tyrosinase to melanosomes in AP-3-deficient melanocytes and the partial pigmentation of AP-3-deficient animals (Lane and Deol, 1974 ). Tyrp1, which binds AP-1 but not AP-3 by yeast three-hybrid assay and is enriched in AP-1-coated vesicles (Raposo et al., 2001 ), may be an exclusive cargo of the AP-1-dependent pathway, and other melanosomal cargo, analogous to neuronal ion transporters (Salazar et al., 2004 ), may exclusively use an AP-3-dependent pathway.
How can two adaptors participate in melanosomal delivery of tyrosinase? The increased endosomal accumulation of tyrosinase in AP-3-deficient cells suggests that tyrosinase, like LAMP-1 in HepG2 cells (Peden et al., 2004 ), is sorted by AP-3 at the limiting membrane of early endosomes. The tubules associated with AP-3-enriched buds contain Tf and Pmel17, cargo destined from the vacuolar endosomal domains for both recycling to the plasma membrane and incorporation into ILVs (Raposo et al., 2001 ). Thus, we propose that AP-3 removes endocytic cargo before or concomitant with arrival at the vacuolar domain of early endosomes. Cargo that fails to be sequestered by AP-3 will remain in the vacuolar domain, from which it may subsequently be incorporated in the invaginating membranes. This model parallels that of yeast AP-3 in averting the MVB pathway, is likely to be shared by nonspecialized cells that lack LROs, and would explain the enrichment of tyrosinase on internal membranes of MVBs in AP-3-deficient melanocytes. Cargo that remains on the limiting membrane of early endosomes and that is not immediately retained by AP-3 coats or sorted to ILVs in the vacuolar domains may be “rescued” by recruitment into AP-1-coated domains. The incomplete “rescue” of tyrosinase away from ILVs in melan-pe cells might reflect relative inefficiency of AP-1 recruitment due to a lower affinity for AP-1 than for AP-3 (despite the apparent higher affinity observed in the yeast three-hybrid assay) or saturation of the AP-1-dependent pathway by cargo that would preferentially be sorted by AP-3, or perhaps the possibility that AP-1 acts primarily at a different site, on tubules downstream of the vacuolar domain. The destination for cargo recruited into AP-1-coated vesicles is also not clear. These vesicles may be targeted directly to maturing melanosomes, consistent with models of AP-1 function in lysosomal targeting (Honing et al., 1996 ; Klumperman et al., 1998 ; Reusch et al., 2002 ; Kyttala et al., 2005 ) and the proximity of endosomal AP-1-positive buds and vesicles to melanosomes. Alternatively, AP-1 cargo may be destined to recycle back to either the plasma membrane or to the TGN to undergo a second “attempt” at delivery to melanosomes via AP-3 on endosomes or a third uncharacterized pathway (such as transient fusion with multivesicular late endosomes), consistent with the model of AP-1 as a recycling adaptor (Mallard et al., 1998 ; Meyer et al., 2000 ; Crump et al., 2001 ; Valdivia et al., 2002 ). Despite these unresolved questions, what is now clear is that AP-3 and AP-1 function in partially redundant pathways leading from tubular endosomes ultimately to melanosomes. Further elucidation of these pathways should provide more detailed information about the plasticity of the endocytic pathway of all cells and the adaptation of the endocytic pathway for biogenesis of LROs.
We thank Sylvie Urbé, Sharon Tooze, and Frances Brodsky for generous gifts of antibodies. We are grateful to Terry M. Mayhew (University of Nottingham, Nottingham, United Kingdom) for advice on quantitative analysis on immunogold-labeled sections. This work was supported by grants from the Wellcome Trust (to A.A.P. and M.S.R.) and the Medical Research Council (to D.F.C. and M.S.R.), National Institutes of Health Grants EY015625 and AR048155 (to M.S.M., G. R., and A.C.T.), Institut Curie, Centre National de la Recherche Scientifique, and Vaincre les Maladies Lysosomales (to G. R.). A.C.T. was supported by a Medical Research Council Studentship and T32 CA 09140 National Institutes of Health/National Cancer Institute Training Program.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05-07-0626) on September 14, 2005.
Abbreviations used: AP, adaptor protein; IEM, immunoelectron microscopy, ILV, intralumenal vesicle; LRO, lysosome-related organelle; MVB, multivesicular body; PAG, protein A-gold; PB, phosphate buffer; Tf, transferrin.
The online version of this article contains supplemental material at MBC Online (http://www.molbiolcell.org).