The Dark Side of Lysosome-Related Organelles: Specialization of the Endocytic Pathway for Melanosome Biogenesis


  • Sachse M, Urbé S, Oorschot V, Strous GJ, Klumperman J. Bi-layered clathrin coats on endosomal vacuoles are involved in protein sorting towards lysosomes. Mol Biol Cell, submitted.

Corresponding authors: Graça Raposo, and Michael Marks,


Melanosomes are lysosome-related organelles within which melanin pigments are synthesized and stored in melanocytes and retinal pigment epithelial cells. Early ultrastructural studies of pigment cells revealed that melanosomes consist of a complex series of organelles; more recently, these structures have been correlated with cargo constituents. By studying the fate of melanosomal and endosomal cargo in melanocytic cells, the effects of disease-related mutations on melanosomal morphology, and the genes affected by these mutations, we are beginning to gain novel insights into the biogenesis of these complex organelles and their relationship to the endocytic pathway. These insights demonstrate how specialized cells integrate unique and ubiquitous molecular mechanisms in subverting the endosomal system to generate cell-type specific structures and their associated functions. Further dissection of the melanosomal system will likely shed light not only on the biogenesis of lysosome-related organelles but also on general aspects of vesicular transport in the endosomal system.

The organelles of the endocytic pathway serve a housekeeping function in all cells, allowing for the selective uptake of nutrients, breakdown of macromolecules, and control of intracellular signaling pathways. Selected cell types have imposed additional functions upon endosomal organelles, some of which are linked to both secretory and endocytic processes. This specialization of the endocytic pathway is exemplified by the collection of cell-type specific organelles, known as secretory lysosomes (1) or lysosome-related organelles (2). Lysosome-related organelles exhibit diverse morphological features and functions, ranging from the calcium- and nucleotide-laden dense granules of platelets (3) to the onion-like, surfactant-laden lamellar bodies of type 2 lung epithelial cells (4) to the cytolysin-enriched cytotoxic granules of T lymphocytes and natural killer cells (5). These organelles all share a variety of features with lysosomes, such as low intralumenal pH and resident lysosomal hydrolases and membrane proteins. Most also share functions with secretory granules, including sequestration, intracellular storage and regulated release of lumenal contents.

The existence of such structurally and functionally varied organelles in distinct cell types raises several questions about the mechanisms by which they are formed and maintained as distinct entities within their host cells. Are lysosome-like organelles merely modified lysosomes or are they distinct from conventional endocytic organelles? If they are distinct, how are resident proteins sorted among tissue-specific and conventional organelles? Are ubiquitous sorting pathways subverted for use by specialized organelles or do specialized cell types possess novel sorting pathways? The answers to these questions are largely unknown, but will likely provide novel insights into the molecular mechanisms regulating the biogenesis of both tissue-specific and ubiquitous organelles.

One particular lysosome-related organelle, the melanosome, is emerging as an excellent model with which to address these questions (6). Melanosomes are specialized for the biosynthesis and storage of melanins, the major pigments made by mammals, in choroidal melanocytes and retinal pigment epithelial cells of the eye and in epidermal melanocytes. Melanosomes in retinal pigment epithelial cells likely function in the detoxification of phagocytosed photoreceptor outer membranes (7), whereas melanosomes in epidermal melanocytes serve as a secretory organelle, facilitating transfer of melanins from melanocytes to neighboring keratinocytes and generation of the characteristic pigmentation of skin and hair (8). Melanosomes were among the first lysosome-like organelles to be characterized morphologically at the ultrastructural level (9–11), yet their molecular composition and a mechanistic understanding of their origins is only now coming to light. This review highlights recent advances in our understanding of the relationship of melanosomes with the endocytic pathway and discusses mechanistic models for their biogenesis within the context of the greater endosomal network.

Why Do Melanin Biosynthesis and Storage Require a Specialized Organelle?

Melanin pigments are derivatives of L-DOPA that are immobilized, likely by polymerization onto a heterogeneous protein scaffold (12). They are synthesized within melanosomes in sequential reactions, initiated by the conversion of tyrosine to L-DOPA catalyzed by the integral membrane melanosome constituent, tyrosinase (13). Red, brown, and yellow pigments, or pheomelanins, are synthesized under certain conditions by subsequent reduction reactions requiring cysteine (13); pheomelanogenic melanosomes differ from those making predominantly black pigments, or eumelanins (14, 15), and will not be discussed further. Eumelanins result from sequential oxidation and hydroxylation reactions, mediated in part by tyrosinase and the related enzymes, Tyrp1 and Tyrp2 (also known as TRP1 and TRP2). The end-result of these reactions are reactive indoles. Exposure of cellular components to these oxidative reactants would be deleterious; thus, melanosomes serve to sequester the reactions and their products to render eumelanin synthetic intermediates harmless.

Melanogenesis is dependent on a number of properties normally associated with lysosomes, such as a low pH environment (16, 17). Although the nature of the acid-requiring steps is not fully known, disruption of the proton gradient either by treatment with inhibitors of the vacuolar ATPase, such as bafilomycin A or concanamycin B (GR, unpublished observations), or by mutagenesis of melanosomal membrane transporters, such as P (OCA2p) (18) or Underwhite (OCA4p) (19), results in a disruption of melanosome morphology and hypopigmentation. Proteolytic events, likely mediated by lysosomal hydrolases such as cathepsins D and L, may also contribute to the maturation of certain melanosomal constituents [MSM, unpublished observations and (20)]. These requirements likely explain the link to a specialized environment tied to the endosomal/lysosomal system.

Developmental Progression of Melanosomes

Just as the organelles of the endocytic pathway can be subdivided into compartments with distinct characteristics, eumelanogenic melanosomes progress through defined, sequential morphological steps (21) (Figure 1). Early stages, often referred to as premelanosomes, lack pigment but have other defining characteristics. Stage I premelanosomes are electron-lucent, membrane-delimited structures with variable amounts of internal membranes. They are similar in morphology to early multivesicular endosomes, as discussed in more detail below. The unique features of the melanosomal system become more evident in stage II premelanosomes. These elliptical structures harbor intralumenal fibers that run the length of the organelle in an organized array, producing a striated appearance. It is upon these fibers that melanins are deposited as they are synthesized, resulting in blackened and thickened striations in stage III melanosomes and eventual masking of all intralumenal structure in stage IV. The striations likely function to detoxify eumelanin intermediates, to sequester and concentrate melanins, and in epidermal melanocytes to prevent diffusion of melanin during transfer to keratinocytes.

Figure 1.

Figure 1.

The melanosomal organelles. (Left) An electron micrograph of plastic-embedded MNT-1 melanoma cells, illustrating the four stages of melanosome development. (Center) A schematic diagram of each of the four stages of melanosome development and their morphological features, as described in the text. (Right) A schematic diagram of the organelles of the ‘conventional’ endocytic pathway for comparison with those of the melanosomal system. EE: Early Endosomes; LE: late endosomes; PLys: prelysosomes; Lys: lysosomes.

In order to effect the morphological changes during maturation, melanosomes of different stages are enriched in distinct cargo molecules. Tyrp1, and likely tyrosinase and Tyrp2, are enriched in later stage, pigmented melanosomes (11, 17, 22). Only one cargo protein has been conclusively shown to be preferentially enriched in premelanosomes – the integral membrane protein Pmel17 (also known as gp100, ME20, or the product of the mouse Silver locus) (17,23) – but others are likely to exist. The differential enrichment of proteins implies that distinct sorting processes must govern protein localization to melanosomes and premelanosomes of distinct stages. Indeed, steady-state morphological analyses show that Pmel17 and Tyrp1 accumulate in distinct intermediates prior to their ultimate destination (17), a notion that confirmed earlier biochemical studies (24). This suggests that different routes are used for intracellular transport to premelanosomes and mature melanosomes (Figure 2A). The protein transport routes to mature melanosomes are not yet well understood. However, a relatively clear and perhaps generally applicable picture is emerging for Pmel17 transport to premelanosomes.

Figure 2.

Figure 2.

Model for transport within the melanosomal and endosomal systems of melanocytic cells.(A) A model for sorting of premelanosomal and melanosomal constituent proteins among the indicated organelles. Arrows represent putative pathways for transport of the indicated proteins, as discussed in the text. Each protein represents cargo constituents of distinct compartments: Pmel17, premelanosome; TRP1 and tyrosinase, mature melanosome; LAMP1, resident late endosome/lysosome; and BSA-gold, lysosome-bound endocytic cargo. Modified from (17) (The Journal of Cell Biology, 2001; 152: 809–823) with copyright permission of the Rockefeller University Press. (B) Hypothetical role of effector proteins in melanosome protein transport. The indicated effector proteins may act within any of the indicated transport steps as defined in panel A. Tyrp1 is referred to here as TRP1.

Premelanosomes Derive from Endosomes

Pmel17 is a type I transmembrane glycoprotein that localizes to the ‘matrix’ of stage II premelanosomes, where it aligns with the intralumenal striations (17, 23, 24). Thus, by following its fate in melanocytic cells, it is possible to trace the development of premelanosomes.

Premelanosomes are commonly quoted as emerging from the smooth endoplasmic reticulum, based primarily on the conclusion from an early ultrastructural study that predated our understanding of the relationship of the secretory and endocytic pathways and the nature of the trans-Golgi network (TGN) (25). The failure to observe Golgi-processed forms of Pmel17 in the mouse (24) and the apparent cosegregation of immature Pmel17 isoforms and other ER proteins with premelanosomes by subcellular fractionation (20,24) has been taken as support for this view. However, Golgi- and post-Golgi processed forms of Pmel17 are clearly detectable in human melanocytic cells (26). Moreover, mature Pmel17 isoforms accumulate in endosomal structures prior to their appearance in stage II premelanosomes (17, 26). Within these endosomes, Pmel17 is present on both the limiting membrane and intralumenal membranes, suggesting that this is the site in which Pmel17 accesses the lumen of intracellular compartments for eventual incorporation into the striations (Figure 3). Thus, endosomal organelles, and not the smooth endoplasmic reticulum, are the direct precursors to premelanosomes.

Figure 3.

Figure 3.

Localization of distinct cargo to endosomes (stage I premelanosomes), stage II premelanosomes and stage III/IV melanosomes by immunoelectron microscopy. Shown are electron micrographs of ultrathin cryosections of the pigmented human melanoma cell line, MNT-1, immunogold labeled for Tyrp1 (A), Pmel17 (B) or Pmel17 and EEA1 (C). Tyrp1 and Pmel17 are detected with protein A coupled to 10 nm gold particles, and EEA1 with protein A coupled to 15 nm gold particles. (A) Note the endosomal network, consisting of electron-lucent compartments surrounded by small tubulovesicular structures as well as multivesicular bodies (stars). Labeling for Tyrp1 is absent from these structures, but is found on the limiting membrane of the electron dense mature melanosomes (Stage IV). (B) Labeling for Pmel17 is observed on the limiting membrane of endosomes as well as on the internal membrane vesicles. CE, coated endosome. Note the electron-dense coat with inward curvature (arrows). Dense labeling for Pmel17 is also observed in stage II premelanosomes (II). (C) A coated endosome (CE) labeled for Pmel 17 and EEA1. EEA1 labeling is restricted to the noncoated area of the endosome. Coated area is indicated by arrows. Bars: 100 nm

What type of endosomal structures harbor premelanosomal proteins? As in other cell types, the early endocytic network of melanocytic cells consists of a network of vacuolar and tubulated structures that are enriched in ‘housekeeping’ proteins such as EEA1. At least two distinct early endosomal domains can be distinguished: tubulovesicular structures containing the bulk of EEA1 and accessed by transferrin and endocytic tracers after 5–10 min, and globular structures, accessed by endocytic tracers only after 15–20 min, with emanating tubules that seem to mediate transferrin recycling. In melanocytes, Pmel17 (Figure 3) and a melanoma-associated protein of unknown function, MART-1 (A. De Mazières and J. Slot, personal communication), are particularly enriched at steady-state in the latter compartments, and at least Pmel17 can enter them after internalization from the plasma membrane with kinetics consistent with receptor-mediated endocytosis (17). Inhibition of endosomal V-ATPases with bafilomycin or concanamycin B inhibits the processing of Pmel17 (26) and induces its accumulation in endosomes (GR, unpublished results). Thus, organelles of the early endocytic pathway are intermediates in the trafficking of premelanosome constituents, and the globular domains are likely equivalent to stage I premelanosomes (Figure 2A).

Specialization of the Endosomal System for Premelanosome Biogenesis

How might endosomes contribute to premelanosome biogenesis? One possibility, based on the analogous formation of secretory granules at the TGN (27), is that these organelles facilitate the segregation and concentration of molecules important for premelanosome biogenesis and function, while discarding proteins destined for other sites such as the cell surface, lysosomes or later melanosomal stages. The globular domains associated with early endosomes possess two unique and mechanistically linked features that are likely involved in such sorting and segregation processes: a dense coat complex and invaginating intralumenal vesicles.

Flat clathrin-containing coats are observed at the ultrastructural level on at least one face of the globular region of early endosomes in many cell types (28,29) but such coated structures are particularly abundant and developed in melanocytic cells (17). In ultrathin cryosections, the coat consists of a ‘fuzzy’ region, in which clathrin is a constituent, and an electron-dense layer of unknown composition closely apposed to the endosomal membrane. These coats are distinguished from clathrin-coated buds observed on tubular early and recycling endosomes (30,31) by being planar and thicker. Moreover, unlike the strongly curved clathrin lattices involved in outward budding, the planar coats on globular endosomes appear flat and often confer an inward curvature to the endosomal membrane. Furthermore, budding from the globular endosomal domains seems to take place only from uncoated regions (17), suggesting that the coat functions to maintain molecules within the endosome rather than incorporate them into outwardly budding vesicles.

How might these coats function? The answer may lie in facilitating protein sorting to intralumenal membranes. The limiting membrane of the globular domains appear to invaginate to form intralumenal vesicles of multivesicular bodies [MVBs (17); and M. Sachse, J. Klumperman, personal communication1]. In all cells at steady state, the protein and lipid content of intralumenal membranes are distinct from those of the limiting membrane, and include membrane proteins that are destined for degradation in lysosomes (32,33). In melanocytic cells, the intralumenal vesicles of the coated endosomal domains are enriched in Pmel17 (17) and MART-1 (A. de Mazières and J. Slot, personal communication). Pmel17 also localizes predominantly to intralumenal vesicles of late endosomal MVBs upon ectopic expression in HeLa cells, and it is within these MVBs that premelanosome-like striations form (26). MVB formation at the coated globular endosomal compartment thus appears to be a key step in premelanosome development, perhaps serving to enrich Pmel17 and other premelanosome cargo within the lumen of precursor organelles (see below).

The coats of the globular endosomal domains appear to be functionally linked to MVB formation. In a Chinese hamster cell line, internalized, ligand-activated growth hormone receptors are retained in the planar coat before sequestration in internal membrane vesicles during MVB formation (M. Sachse, J. Klumperman, personal communication1). An analogous process in melanocytes might permit the concentration of Pmel17 and other proteins for subsequent incorporation into intralumenal vesicles of premelanosome precursors. Consistent with this, Pmel17 in the limiting membrane of melanocyte endosomes are mainly observed in coated areas (17). According to this model, molecules not destined for premelanosomes would fail to be retained within the coated region and would be discarded by budding of vesicles or tubules destined for other sites.

The planar coats may also function in regulating fusion events at the limiting membrane of the globular endosomal domains. Syntaxin 7, a t-SNARE that is up-regulated during melanogenesis in differentiating B16 melanoma cells (34), is enriched within clathrin-containing lattices on endosomes (28). As suggested for the coat-like ‘filamentous plaques’ between late endosomes and lysosomes of Hep2 cells (35), regulated release of the coat may unmask syntaxin 7 or other membrane components to permit fusion of the maturing endosome/premelanosome with late endosomes or lysosomes. Thus, the planar coats may function to retain proteins for inwardly budding vesicles of MVBs, for fusion events at the limiting membrane, or both.

The Molecular Machinery at the Coated Endosomal Domain

The molecular machinery involved in protein sorting for premelanosome formation is uncharacterized. However, predictions can be made from related systems. By analogy to the known interaction between cargo proteins and COPI, COPII, or adaptins, cytosol-exposed sorting signals in Pmel17 and other premelanosome cargo may promote association with coat components. The cytoplasmic domain of Pmel17 contains a canonical di-leucine-based sorting motif, and a similar signal in the related QNR-71 melanosomal, protein of quail facilitates sorting and recruitment of the heterotetrameric adaptor protein AP-3 (36). However, neither AP-3 nor the related AP-1 or AP-2 have yet been detected within the planar coats [(17,30) and M. Sachse, J. Klumperman, personal communication1], suggesting that other components may recognize these proteins at the coated endosome.

What might these components be? The coats are insensitive to brefeldin A but sensitive to inhibitors of phosphatidylinositol (PI) 3′-kinases, indicating that coat formation and homeostatic maintenance are independent of ARF proteins but dependent on phosphoinositide metabolism [(36); GR, unpublished results; M. Sachse, J. Klumperman, personal communication1]. Consistent with the latter sensitivity, the FYVE-domain containing protein Hrs is the only protein other than clathrin heavy and light chains known to be enriched in the endosomal coats (29,37). Hrs seems to be recruited to endosomes through PI(3)phosphate (PI3P) binding to its FYVE domain (38), and Hrs recruits clathrin through a distinct domain (29); thus, Hrs may be partially responsible for the dependence of the coat on PI3P levels. Interestingly, PI3P may also be required for MVB formation and endosomal recycling through the recruitment of additional proteins with FYVE and PX domains (32,39). Thus, PI3P-binding proteins, together with rab proteins, rab effectors, and SNARE proteins, may orchestrate the formation of endosomal domains specialized in sorting to internal membrane vesicles.

How might cargo be directed to these domains? Di-leucine-based sorting signals, like that of Pmel17, may be involved. Similar signals on sortilin and the mannose-6-phosphate receptors bind to the VHS domain of the GGA proteins (40–43), and could conceivably bind Hrs and/or related proteins through their VHS domains. Another potential sorting signal could be provided by ubiquitination of cargo. In yeast, monoubiquitination signals cargo recruitment for sequestration on intralumenal vesicles of MVBs (44). A link between ubiquitination and the endosomal planar coats in mammalian cells is suggested by a shift in the distribution of a truncated form of growth hormone receptor toward the uncoated endosomal regions and away from intralumenal vesicles in cells depleted of free ubiquitin by treatment with proteasome inhibitors (M. Sachse, G. J. Strous, J. Klumperman, personal communication1). In yeast, the ubiquitinated cargo is recognized and sorted by a protein complex (ESCRT-1) through one of its components, Vps23p (44). The human homolog of VPS23, the tumor-susceptibility gene Tsg101, is also directly involved in budding processes at the MVB and the plasma membrane (45, 46). Characterizing the interactions of Pmel17 and MART-1 with these complex machineries and unraveling the composition of the coats will provide a clearer view of the sorting steps involved in premelanosome formation.

Premelanosome Morphogenesis Is Driven by a Cargo Protein

The intralumenal striations of stage II premelanosomes are perhaps the most distinguishing characteristic of the melanosomal system. Pmel17 localizes to these striations, and its localization to intralumenal membranes of MVBs appears to precede striation formation. Indeed, ectopic expression at high levels of Pmel17 in nonmelanocytic cells such as HeLa results in the formation of striation-like structures within multivesicular late endosomes; the striations appear to recruit Pmel17 from the intralumenal vesicles (26) (Figure 4). These results suggest that Pmel17 is a major biogenetic component of the striations. Because striation formation in HeLa cells requires very high level expression of Pmel17, it is likely that other proteins, such as MART-1, may facilitate striation formation through direct or indirect interactions with Pmel17.

Figure 4.

Figure 4.

Model of formation of striations from MVBs. At top is a schematic diagram showing the presence of Pmel17 within distinct elements of coated globular endosomes, a potential multivesicular intermediate [as observed easily in transfected HeLa cells; see (26)], and striated stage II premelanosomes. Below is a schematic diagram of the primary structure of Pmel17 and its processing intermediates proposed to be present in each structure.

The intralumenal striations of stage II premelanosomes are thought to be proteinaceous in nature, and lack the morphological hallmarks of a lipid bilayer. How then is the integral membrane protein Pmel17 recruited to them? Post-Golgi proteolytic events likely play an important role (Figure 4). Pmel17 is cleaved within the lumenal domain, most likely in the TGN and/or endosomes, into two associated fragments (26). The N-terminal fragment consists of 75% of the lumenal domain, and the remaining residues, including the transmembrane and cytoplasmic domains, are present within the C-terminal fragment (26). The stability of these fragments and their accumulation in HeLa cells supports the hypothesis that they represent the cohort of Pmel17 on multivesicular structures (26). Interestingly, Pmel17 within striations is nonreactive with antibodies to the cytoplasmic domain (17, 20), suggesting that either the epitope becomes masked or the C-terminal fragment dissociates prior to striation formation. Our recent results support the latter explanation and suggest that cleavage is required for striation formation to occur (J. F. Berson, D.C. Harper, D. Tenza, G. Raposo and M. S. Marks, in preparation; see Figure 4). The timing of the cleavage event, coupled with microenvironment changes within the newly forming premelanosome, likely regulate conformational changes within Pmel17 that permit ordered oligomerization leading to fibril formation, similar to the formation of fibers by ordered aggregation of prion proteins (47). By analogy, regulated proteolytic cleavage of pro-von Willebrand factor is thought to regulate formation of the morphologically related Weibel–Palade bodies (48), suggesting that proteolytic cleavage of cell type specific proteins may be a common means of regulating biogenesis of lysosome-related organelles. Other proteolytic events within the forming premelanosome and mature melanosome may further shape morphology and functionality (20). Disruption of any of the proteolytic events may contribute to biogenetic defects in diseases of lysosome-related organelles.

Protein Sorting to the Mature Melanosome

Once the striations are formed, melanosome development proceeds with melanin deposits accumulating over the striations in stage III and IV melanosomes. How is melanogenesis timed such that it begins only after complete formation of the fibrous striations? At least one mechanism is the delivery of melanogenic enzymes to preformed premelanosomes. This scenario, supported by their virtual exclusion from stage II premelanosomes (17), would necessitate a delivery pathway for melanogenic enzymes distinct from that used by premelanosome-resident proteins (Figure 2A). Consistent with this, melanogenic enzymes like Tyrp1 are not found to an appreciable extent within endosomal organelles at steady state, and thus may not require transport through the globular coated endosomal domains for trafficking to the melanosome (17). Nonetheless, since both tyrosinase and Tyrp1 can be found at low levels at the plasma membrane and can be internalized (49, 50), a small proportion of melanogenic enzymes likely traffic through the endosomal system. These proteins would not be expected to accumulate in coated regions, but rather may be retrieved from the endosomal system by a budding process.

Rather than obligate trafficking through endosomal intermediates, melanogenic enzymes are likely delivered to preformed premelanosomes directly from the TGN (Figure 2A), as suggested by the accumulation of Tyrp1 within tubulovesicular structures near the TGN (17). Similar Tyrp1-containing, tubulovesicular structures are often coated with clathrin and/or with AP-1, consistent with the copurification of Tyrp1 and tyrosinase with clathrin-coated vesicles by subcellular fractionation (24) and with the cytochemical detection of tyrosinase activity within clathrin-coated vesicles (11, 51). A direct TGN-to-melanosome transport route is supported by the transport defects in mouse melanoma cells deficient in glycosphingolipid biosynthesis (52); in such cells, tyrosinase appears to accumulate in the TGN and Tyrp1 aberrantly cycles through the plasma membrane due to defects in TGN export and sorting.

What molecular components regulate delivery of melanogenic enzymes to melanosomes? The heterotetrameric adaptor proteins, AP-1 and AP-3, play critical roles in this process. Genetic deficiencies of AP-3 subunits underlie the pigmentation and bleeding defects observed in patients with Hermansky–Pudlak Syndrome (HPS) type 2 patients and pearl and mocha mice, as well as the eye pigment defects in several Drosophila mutants (53). The cytoplasmic domain of tyrosinase binds to AP-3 in vitro (54), and tyrosinase is mislocalized in melanocytes from HPS2 patients (55), supporting the notion that AP-3 plays a role in transport of melanosomal components. However, despite the current dogma suggesting that AP-3 mediates cargo sorting and vesicle budding at the TGN, the vectoriality of transport is far from being clear. As indicated in Figure 2(B), AP-3 could function at one of three steps: (i) transport of tyrosinase from the TGN to maturing melanosome; (ii) retrieval of tyrosinase from regions of endosomal membranes devoid of the planar coat into vesicles destined for mature melanosomes; or (iii) removal of tyrosinase from the melanosomal membrane in order to maintain homeostatic levels of enzyme. Interestingly, in AP-3-deficient melanocytes from HPS2 patients, tyrosinase accumulates in multivesicular structures similar in appearance to the coated endosomal domains (55); this could be due to either missorting at the TGN as in model 1 or a failure to retrieve tyrosinase from endosomes as in model 2. Kinetic analyses of tyrosinase localization in normal and AP-3-deficient cells and detailed analyses of AP-3 localization in melanocytes should help to distinguish these models.

Interestingly, Tyrp1 localization does not appear to be grossly affected by AP-3 deficiency (55). By contrast, the extensive colocalization of Tyrp1 with AP-1 in melanocytic cells suggests a role for AP-1 in Tyrp1 transport to melanosomes. As for AP-3, AP-1 may participate either in a direct transport pathway from the TGN to the maturing melanosome, retrieval from endosomal membranes before rerouting to later melanosomal stages, or budding processes at the melanosomal membrane. Given the broad distribution of these adaptor complexes in melanocytes and their association with different vesicles and compartments (GR, unpublished observations), it is likely that both AP complexes act at multiple sites. This may complicate the interpretation of the defects observed in cells with mutations in components of adaptors and related complexes.

Segregation of Melanosomal and Lysosomal Proteins

The delivery of melanogenic enzymes and the progressive maturation of the premelanosome into a mature melanosome are likely accompanied to some extent by delivery of lysosomal proteins. For example, Lamp1 and Tyrp1 colocalize in vesicular structures near the TGN of melanocytic cells (17), and AP-3 deficiency affects the trafficking of not only tyrosinase but also Lamp1 and other lysosomal proteins (56–59). This may account for the presence of a cohort of lysosomal proteins within melanosomes, a fact that led to the early notion that melanosomes are merely modified lysosomes [reviewed in (60)]. Nevertheless, whereas melanosomal proteins localize to late endosomes and lysosomes in transfected fibroblasts (50,61–63), the bulk of lysosomal proteins in melanocytic cells accumulate within nonmelanosomal structures with all of the characteristics of bona fide lysosomes (17,64,65). These structures largely exclude melanosomal resident proteins at steady state (17). Thus, melanocytic cells must be uniquely capable of sorting proteins among melanosomal and lysosomal compartments.

How might such sorting be achieved? Several observations suggest that active and direct sorting occurs at least partly between melanosomes and lysosomes. First, the cohort of ‘contaminating’ lysosomal proteins changes throughout melanosome maturation (GR, unpublished observations), consistent with selective delivery of some components and removal of others. Second, whereas tyrosinase in human melanoma cells localizes at steady state predominantly to melanosomes, treatment with lysosomal proteinase inhibitors reveals additional lysosomal localization (MSM, unpublished observations). This suggests that tyrosinase may eventually be degraded in lysosomes, perhaps after transit through melanosomes. Third, lysosomes are often found in close proximity to melanosomes and premelanosomes in melanocytic cells (17). Taken together, these observations support the hypothesis that melanosomes and lysosomes undergo dynamic, regulated interactions, perhaps through ‘kiss-and-run’ fusion (66). These interactions may serve to remove unnecessary material from melanosomes and/or to deliver lysosomal contents to maturing melanosomes. Changes in the degree of this interaction could potentially permit melanosomes and lysosomes to fuse into a single hybrid organelle in some circumstances, such as in cells making pheomelanins.

The coexistence of lysosome and lysosome-like organelle distinguishes the melanosomal system from traditional secretory lysosomes in lymphoid cells, such as cytotoxic T-cell granules (1) and MHC class II antigen processing compartments (67), but is analogous to the situation for other lysosome-related organelles, such as platelet dense granules (3) and Weibel-Palade bodies (68). This subdivision of lysosome-related organelles may underlie the organelle specificity of subtypes of HPS and related disorders (69,70).

Other Components of the Biogenetic Machinery

The biogenesis of a number of lysosome-related organelles, including melanosomes, is disrupted in a group of HPS-related genetic disorders of humans, mice, Drosophila, and other species (6, 70, 71). Genes encoding AP-3 subunits are among at least 16 genes targeted in HPS-related disorders. What roles do other characterized HPS-related gene products play in melanosome biogenesis? The predicted structure of the HPS1/pale ear (72, 73) and HPS3p/cocoa (74, 75) gene products, like that of the Chediak Higashi syndrome CHS1 gene (76), provide no clues to their function, but the phenotypes associated with their defects – enlarged or small melanosomes (72, 74–77) – are suggestive of disregulated organelle fusion or fission. As indicated in Figure 2B, this disregulation could occur at multiple sites, including the melanosome:melanosome, melanosome:lysosome, or coated endosome:late endosome transitions. The pallid gene product, pallidin, which also lacks homology to known proteins, was found in a yeast two-hybrid screen to interact with a t-SNARE, syntaxin-13 (78). Syntaxin-13 is thought to be involved in endosomal fusion (79) and is one of several SNARE proteins that are up-regulated upon differentiation and increased melanization of B16 melanoma cells (34). Thus, pallidin is likely to regulate SNARE-dependent endosomal fusion in some way, perhaps at the level of the coated endosomal domains. The gunmetal gene product is a subunit of the rab geranylgeranyl transferase, and is likely involved in regulating the activity of several rab proteins within the melanosome biogenesis pathway (80). Among these rab proteins are likely to be Rab27a, the product of the gene deficient in ashen mice (81), which together with myosin Va (defective in dilute mice) (82) and melanophilin (defective in leaden mice) (83), regulates melanosome motility and localization to melanocyte dendrites (6). However, gunmetal mice likely have defects in other rab proteins required for melanosome biogenesis, perhaps including Rab7 (84). Elucidating the roles of each of these proteins and the remaining HPS-related gene products will clarify the nature of the transport steps involved in melanosome biogenesis.

Melanosome Biogenesis and Development of Melanoma

Interestingly, cell transformation is associated with changes in melanosome biogenesis and pigment production. During transformation to melanoma, some melanocytes lose their ability to synthesize pigment and most others display disorganized melanosomal structures (13, 85). This effect may in part be due to altered signaling cascades. For example, transfection of the oncogene V-ras into melanocytes abolishes melanosome formation and melanin synthesis (86). Furthermore, melanosomal resident proteins such as MART-1 and Tyrp1 are enriched in ‘exosomes’ derived from melanoma cells, indicating their release into the extracellular environment during fusion of multivesicular compartments with the cell surface (87). This observation suggests that melanosomal proteins may be mistargeted in transformed cells, perhaps due to modifications of the intracellular machinery involved in maintaining the identity of melanosomes within the endocytic pathway. One example of such a modification that may be linked to melanoma development is the cell transformation that results from the loss of expression of Tsg101 (88), a component of the mammalian ESCRT-1 complex (44). While the mechanism by which transformation occurs is unclear, one might imagine, given the role of multivesicular bodies in premelanosome formation, how disregulation of ESCRT-1 function could concomitantly alter melanosome biogenesis. Comparative analyses of normal and transformed cells will likely be fruitful in understanding how changes in intracellular trafficking may be linked to the development of melanoma and other tumors.

Concluding Remarks

The mechanisms of melanosome biogenesis described here build upon our developing notions of the incredible dynamics of the eukaryotic endocytic system and their potential for adaptation. The subversion of the endocytic pathway to form melanosomes in melanocytes parallels similar processes in other cell types possessing specialized organelles and functions, including the modification of the early endosomal system for synaptic vesicle biogenesis in neurons (89) and the development of secretory lysosomes (1). A common theme among lysosome-related organelles, as evidenced for premelanosome formation, may be the exploitation of multivesicular endosomes to concentrate key molecules involved in organelle morphogenesis and function, as has been shown for platelet dense granules (90) and B lymphoid and dendritic cell antigen processing compartments (91). Another common theme may be the use of both conserved and unique mechanisms to generate cell type specific structures and functions. Melanosomes and other complex organellar systems will continue to be valuable tools for better understanding the plasticity and dynamics of the endosomal network and the molecular mechanisms underlying organelle biogenesis.


We are very grateful to J. Klumperman, M. Sachse, G. J. Strous, J. Slot, D. Rimoldi and A. de Mazières for communication of unpublished results, to C. Burd and W. Stoorvogel for critical review of the manuscript and helpful discussions, and to D. Louvard for ongoing support. We especially thank D. Tenza, J. F. Berson, and other members of the Marks and Raposo laboratories for contributions to the figures and models presented here. This work was supported by the C.N.R.S. and the Institut Curie (GR) and by National Institutes of Health grant # R01-EY-12207 from the National Eye Institute (MSM).