Filopodia are conduits for melanosome transfer to keratinocytes

Melanosomes are organelles unique to melanocytes that function in the synthesis of melanin, a complex pigment involved in photoprotection of the skin through its ability to absorb and scatter light and reduce reactive oxygen species (Marks and Seabra,2001). In order to accomplish this, melanosomes must be transferred to epidermal keratinocytes, where they are found in autophagic vacuoles in a perinuclear or cap-like distribution in vitro and in vivo(Corcuff et al., 2001). Melanosomes are elliptical or spheroidal organelles that contain melanogenic enzymes and co-factors, including the tyrosinase gene family of proteins and have been categorized as stage I-IV on the basis of their electron microscopic appearance and degree of melanization(Kushimoto et al., 2001). Recent studies suggest that premelanosomes (stage I and II) and late stage melanosomes (stage III and IV) represent a distinct lineage of organelles that are separable from conventional endosomes and lysosomes within pigmented cells(Raposo et al., 2001).

It is now known that melanosome trafficking is mediated in part by microtubular motor myosin Va, the product of the dilute locus, which traps melanosomes at the actin-rich periphery of the dendrite, and rab27a, the product of the ashen locus(Mercer et al., 1991;Provance et al., 1996;Wu et al., 1997;Wu et al., 1998;Wei et al., 1997;Wilson et al., 2000;Bahadoran et al., 2001;Wu et al., 2001). Rab27a is involved in the transport of melanosomes through its ability to recruit myosin Va to the tip of the melanocyte dendrite(Hume et al., 2001). These important and relatively recent insights into melanosome trafficking were made possible through the use of mutant mouse strains and time-lapse video microscopy of cultured cells, which allowed direct visualization of melanosome movement and modifiers of actin, microtubules and their motor proteins. In contrast with melanosome trafficking, much less is known about melanosome transfer. A major hurdle that has severely limited progress in understanding the molecular basis of melanosome transfer has been the lack of a model system. The majority of studies of melanosome transfer to keratinocytes have been based on co-cultures of non-human cells observed by electron microscopy. Studies performed utilizing time-lapse video microscopy have been limited by the relatively poor resolution achieved(Mottaz and Zelickson, 1967;Cohen and Szabo, 1968;Wolff, 1973). Other more recent studies have utilized gold particle uptake by keratinocytes, melanin uptake or transfer of cytoplasmic dyes from melanocytes to keratinocytes to measure transfer (Seiberg et al.,2000a; Seiberg et al.,2000b; Sharlow et al.,2000; Minwalla et al.,2001). In toto, these prior studies led to important observations that suggested phagocytosis of melanocyte dendrites by keratinocytes as the major mode of melanosome transfer, although exocytosis of melanosomes into the extracellular space with uptake by keratinocytes and insertion of melanocyte dendrites and transfer of melanosomes to keratinocytes have also been proposed(Yamamoto and Bhawan, 1994). Although the more recent studies using particle uptake provide insight into the role of the keratinocyte in granule uptake, the use of a model system in which melanosome transfer is being studied directly provides an opportunity to examine the potential role of the melanocyte in melanosome transfer.

It is well established that Cdc42, a member of the Rho family of GTP-binding proteins, is involved in filopodia and microspike formation in many cell types. Filopodia are actin-based structures that arise from neuronal growth cones and function in neuronal pathfinding(Davenport et al., 1993;Rosentreter et al., 1998). The recent demonstration that Cdc42 is associated with coatamer proteins in the Golgi apparatus, that it regulates exit of apical and basolateral proteins from the Golgi network and is involved in exocytosis of secretory granules in mast cells is indicative of the diverse roles that Cdc42 plays in cells(Brown et al., 1998;Hong-Geller and Cerione, 2000;Wu et al., 2000;Müsch et al., 2001). The downstream effectors of Cdc42 fall into six families most of which contain a CRIB-binding domain and include Cdc42-binding kinase, myotonic dystrophy kinase-related Cdc42-binding kinase, mixed lineage kinase, p21-activated kinase (PAK), WASP (Wiscot-Aldrich Syndrome Protein), IQGAP and MSE55/BORG/CEP(Burbelo et al., 1995). At least four closely related isoforms of PAK (PAK1, PAK2, PAK3 and PAK4) exist in mammalian cells (Manser et al.,1994; Manser et al.,1995; Martin et al.,1995; Dan et al.,2001). PAK-family kinases are activated by GTP-Cdc42 or GTP-Rac1 as well as G-protein-coupled receptors and cytokines and phosphotidyl-inositol 3-kinase (PI3-kinase); this leads to a change in conformation of the kinase inducing autophosphorylation on multiple serine and threonine residues and activation (Knaus et al.,1995; Manser et al.,1997; Wang et al.,1999; Chung et al.,2001). Activation of PAK results in effects that mimic Rac1 and Cdc42 and include lamellipodia and filopodia formation, activation of the c-Jun N-terminal kinase MAP kinase cascade and NKκB, alteration in cell motility and inhibition of apoptosis and stimulation of macropinocytosis(Sells 1997; Sells et al.,1999; Frost et al.,1998; Frost et al.,2000; Dharmawardhane,2000). Non-kinases that interact with Cdc42 include the WASP family, which consist of WASp, N-WASP and related Scar proteins isolated in Dictyostelium. WASP, in concert with WIP (WASP-interacting protein)participates with the Arp2/3 complex to induce actin nucleation and filopodia formation (Symons et al.,1996; Miki et al.,1996; Miki et al.,1998; Rohatgi et al.,1999; Banzai et al.,2000; Martinez-Quiles et al.,2001). WASP is only expressed in hematopoietic cells and is mutated in patients with Wiscot-Aldrich syndrome, whereas N-WASP is ubiquitously expressed but is enriched in the brain(Fukuoka et al., 1997). In a cell-free system, addition of active Cdc42 significantly stimulates neuronal-WASP (N-WASP) by exposure of N-WASPs' actin depolymerizing region,creating free barbed ends from which actin polymerization can take place(Suzuki et al., 1998).

In this report we used have high resolution movies made from digital images to directly observe melanosome transfer to keratinocytes in human cells. These movies, along with electron microscopy of cells in vitro and skin in vivo provide evidence that suggests that melanosome delivery to keratinocytes occurs along filopodia. We show that expression of activated Cdc42 in human melanocytes accentuates filopodia formation and melanosome transport and that melanosomes are enriched in PAK1 and N-WASP, Cdc42-effector proteins. In combination with previous data showing SNARE and rab proteins on melanosomes(Scott and Zhao, 2001), these observations suggest a novel model for melanosome transfer to keratinocytes.

Polyclonal antibodies to PAK1 were purchased from Zymed Laboratories (San Francisco, CA); polyclonal antibodies to Cdc42 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); polyclonal antibodies to N-WASP were a generous gift of Dr Rohatgi (Harvard Medical School, Boston, MA) and have been described previously (Rohatgi et al.,1999); polyclonal antibodies to chick brain myosin 5a tail domain(clone 32a) were a generous gift of Richard Cheney (Chapel Hill, NC) and have been described previously (Espreafico et al., 1992); monoclonal antibodies to TRP-1 (mel-5) were purchased from Signet Laboratories (Dedham, MA); monoclonal antibodies to transferrin receptor were purchased from Zymed Laboratories (San Francisco, CA);polyclonal antibodies to tubulin were purchased from Sigma Co (St. Louis, MO);fluorescein isothiocyanate and Texas Red goat anti-rabbit and anti-mouse antibodies were purchased from Molecular Probes (Eugene, OR);horseradish-peroxidase-conjugated goat anti-rabbit and anti-mouse antibodies and normal rabbit serum were purchased from Sigma Co; vitrogen was purchased from Cohesion (Palo Alto, CA). Membrane dyes DiI and DiO and Alexa Fluor 594 phalloidin were purchased from Molecular Probes. Nocodazole and cytochalasin D were purchased from Sigma Co.

Neonatal foreskins were obtained according to the University of Rochester's Research Subject Review Board. Co-cultures of human melanocytes and keratinocytes were initiated from human foreskins as previously described(Scott and Haake, 1991) and maintained in Keratinocyte Growth Media (KGM, Gibco BRL, Gaithersburg, PA). In primary skin cultures this media sustains melanocyte growth through the production of melanocyte growth factors by proliferating keratinocytes(Halaban et al., 1988). For growth of melanocytes, cells were cultured in Melanocyte Growth Media (MGM,Gibco-BRL).

Co-cultures of melanocytes and keratinocytes (approximately 10⁵cells total) were subcultured on vitrogen-coated 25 mm glass coverslips for 1-2 days and placed in a closed heated chamber (Warner Instruments, New Haven,CT) maintained at 37°C. The cells were viewed on a Nikon Eclipse Microscope 800 under differential interference contrast (DIC) optics with a 100× objective. The chamber was perfused with KGM maintained at a constant temperature of 37°C by an in-line heater (Warner Instruments)using gravity flow. The rate of flow was approximately 166 μl/min and imaging lasted 45 minutes. Cell viability was checked following experiments with trypan blue and no cytotoxicity was observed.

Sequential images were obtained at 8 second intervals using the green filter of a Spot digital camera (Diagnostic Instruments, Sterling Heights,MI). The resulting 8 bit/pixel megapixel (1315X1033) images yielded a resolution of 10 pixels/micron when combined with the 100× microscope objective. A series of operations to reduce noise and artefacts were performed on the images using the Matlab GUI facility (Mathworks, Natick, MA). To further reduce size, the movies were created using QuickTime Pro (Apple Computer Inc, Cupertino, CA).

Melanocytes grown in MGM and keratinocytes grown in KGM were labeled with DiI (0.6 μM) and DiO (0.2 μM) respectively for 10 minutes at 37°C followed by extensive washing. 24 hours later, melanocytes were trypsinized from the dish and added to keratinocytes on vitrogen-coated 100 mm glass coverslips at an approximate ratio of 1:1 in KGM. To stimulate melanosome transfer, co-cultures were irradiated with a single dose of ultraviolet (UV)irradiation using a solar simulator at a dose of 4 J/cm² as previously described (Scott and Zhao, 2001). 24 hours after irradiation, live cells were viewed on a Nikon Eclipse Microscope 800 and images were captured with a Spot digital camera. To arrive at an approximate percentage of cells with membrane fusion, the number of keratinocytes with yellow fluorescence viewed under a filter to detect both DiI and DiO was counted in 10 random fields (100× objective). Experiments were repeated three times. Digital images were postprocessed using Adobe PhotoShop 5.0.

Melanosomes were isolated from human melanocytes essentially as described for isolation of melanosomes from cultured B16F1 cells, with a few modifications (Scott and Zhao, 2001). The postnuclear supernatant was centrifuged for 10 minutes at 10,000 g at 4°C to obtain a large granule and a small granule fraction. The large granule fraction, which is enriched in melanosomes, was then layered onto a sucrose gradient and centrifuged at 85,000 g at 4°C for 1 hour. The melanosome-rich fraction was collected from the 2M layer of the gradient and lysed in buffer (150 mM NaCl, 10 mM Tris-HCL, pH 7.8, 1% Triton-X) plus protease inhibitors (Complete TM Mini, Boehringer Mannheim, GmbH, Germany). Protein samples were quantified using the Bio-Rad Dc protein assay kit(Bio-Rad Laboratories, Hercules, CA) with bovine serum albumin as standard. Equal amounts of protein were electrophoresed on 10% or 15% precast SDS gels(Jule Inc, New Haven, CT) and blotted to nitrocellulose membranes (Bio-Rad Laboratories) using standard procedures. Full range rainbow molecular weight markers were purchased from Amersham Life Sciences (Arlington Heights, Ill). Visualization of the immunoreactive proteins was accomplished using an enhanced chemiluminescence reaction (Amersham Life Sciences). Positive controls for Cdc42 and N-WASP consisted of mouse brain extracts (StresGen Biotechnologies, Victoria, Canada); positive controls for PAK1 consisted of Jurkat cell lysates (Upstate Biotechnology, Lake Placid, NY).

Melanocytes grown in MGM were subcultured onto vitrogen-coated 2-well glass chamber slides (Nalge Nunc International Corp., Naperville, IL). Cell monolayers were fixed in cold methanol/acetone (1/1) followed by permeabilization in 0.5% triton-X-100 in stabilization buffer (PBS, 100 mM MgCl₂, 1 mM CaCl₂) for 15 minutes, and non-specific binding of antibody was blocked by incubation of the slides in 10% normal goat serum. Primary antibodies were applied overnight at 4°C followed by incubation with appropriate Texas-Red- or fluorescein-conjugated secondary antibodies for one hour at room temperature. For double labeling experiments,the second primary antibody was applied for one hour at room temperature followed by the appropriate secondary antibody. DAPI (Vector Laboratories,Burlingame, CA) was used to stain nuclei. To stain actin, cells were fixed in formalin, permeabilized as described above and incubated with Alexa Fluor 595 for 2 hours at room temperature. Images were captured with a Spot digital camera and post-processed using Adobe PhotoShop 5.0.

Melanocyte-keratinocyte co-cultures were grown on vitrogen-coated glass chamber slides, as described above, in KGM. Cells were fixed for 30 minutes in 2.5% glutaraldehyde in Sorensen's phosphate buffer pH 7.4 and were post-fixed in 1.0% osmium tetroxide in Sorensen's phosphate buffer. Cell membranes were enhanced by incubation of cells for 45 minutes in 0.5% uranyl acetate diluted in 25% ethanol. After dehydrating in a graded series of ethanol, the cells were infiltrated for 30 minutes with a 1:1 solution of 100% ethanol and 100%Spurr epoxy resin and were then infiltrated overnight in 100% Spurr epoxy resin. The next day the slides were inverted onto Spurr epoxy resin filled BEEM capsules and allowed to polymerize. Capsules were trimmed and thin-sectioned with 2.0% uranyl acetate and Reynolds lead citrate and viewed with a Hitachi 7100 electron microscope. To assess purity of melanosome fractions, melanosomes were fixed in glutaraldehyde and post-fixed in osmium tetroxide as described above. Melanosomes were captured in 4% warm agarose and were embedded in Spurr epoxy resin and thin sectioned as described above except that thin sections were not stained with uranyl acetate or lead citrate. Electron microscopy of human skin was accomplished by fixation of a 2 mm punch biopsy of skin from a male subject in glutaraldehyde in Sorensen's buffer overnight. The tissue was processed identically to the cells in culture except that cell membranes were not enhanced by uranyl acetate.

Recombinant adenovirus capable of expressing constitutively active Cdc42(Cdc42^V12) and green fluorescence protein (GFP) in the AdEasy vector (Quantum Biotechnologies, Montreal, Canada), and empty vector expressing GFP alone, were a kind gift of Dr Bambera (Colorado State University, CO) and have been described previously(Brown et al., 2000). Infection efficiency was monitored by viewing the cells in an inverted phase microscope(Nikon Diaphot) equipped with a filter to detect GFP. To assess the effect of Cdc42^V12 on melanosome transfer to keratinocytes, pure populations of melanocytes (10⁵ cells) grown in MGM were infected with either Adeasy Cdc42^V12 or empty vector with a multiplicity of infection(MOI) of 30. 18 hours later keratinocytes were added (10⁵) and the co-culture was allowed to grow in KGM for at least 5 days prior to imaging.

Cell lysates were incubated with GST-PAK-PBD fusion protein according to the manufacturer's instructions (Cytoskeleton Inc., Boulder, CO), and GTP-bound Cdc42 was captured by incubation of the lysate with glutathione beads (BD pharMingen, San Diego, CA). Positive controls consisted of lysates pre-loaded with GTPγS (200 μM). The beads and proteins bound to the fusion protein were washed in lysis buffer, eluted in Laemmli sample buffer,resolved on 15% gels and blotted with antibodies against Cdc42.

To better define the mechanism of melanosome transfer, we utilized DIC optics and time-lapse digital microscopy to directly visualize melanosome movement in human melanocyte-keratinocyte co-cultures and to better define the mechanism of melanosome transfer. The most striking feature observed from time-lapse digital microscopy was the presence of long (up to 16 microns)dynamically active filopodia arising from melanocyte dendrite tips and the melanocyte cell body, many of which contained melanosomes that were easily visualized under DIC optics. Shown in Fig. 1a are images of two melanocyte dendrite tips in which numerous filopodia were observed. Movies of these co-cultures viewed at high magnification allowed one to see the rapid motion of the filopodia. When melanocytes were cultured in the absence of keratinocytes, filopodia moved in a random fashion and made little contact with neighboring cells, although some contact with other melanocytes was seen. Filopodia attached to keratinocyte membranes in many instances and remained attached for the duration of image acquisition (approximately 45 minutes; Fig. 1b,c). Melanosomes were observed to move towards the keratinocyte membrane along filopodia, although in the majority of cases melanosomes remained in filopodia and were not transferred to keratinocytes. Movement of melanosomes along filopodia from one melanocyte to another melanocyte was also observed (not shown). Real time observations confirmed that the structures were filopodia and not retraction fibers, although some retraction fibers were also present. Melanosomes in retraction fibers moved bidirectionally, which is consistent with the presence of microtubules in these structures. To further verify that the structures observed were filopodia, co-cultures were imaged in the presence of nocodazole (10 μg/ml), which would not be expected to alter filopodia movement. Preliminary experiments of nocodazole-treated melanocytes stained with anti-tubulin antibodies established that this dose results in total dissolution of microtubules within 5 minutes; re-establishment of microtubules after washout took up to 30 minutes (data not shown).Fig. 1d shows images of a melanocyte dendrite from a co-culture after 5 minutes of treatment with nocodazole. There was rapid re-distribution of melanosomes to the cell body,with retention of some melanosomes in a cap-like distribution at the actin-rich dendrite tip. These observations are similar to those reported by Wu et al. (Wu et al., 1998) in murine melanocytes treated with nocodazole and are consistent with the role of microtubules in melanosome trafficking to the dendrite tip. The rapid movement of the filopodia was not affected by nocodazole treatment. We attempted to assess the effect of cytochalasin D on filopodia; however these experiments were uninformative because even very low doses of cytochalasin D resulted in rapid collapse of the melanocyte actin network, with retraction of dendrites,as determined both morphologically and by staining of the cells for actin(data not shown).

Melanosome transfer to keratinocytes in culture is an uncommon event;however Fig. 2a shows a sequence of images in which melanosome transfer to keratinocytes has occurred. A melanocyte and keratinocyte viewed at 100× magnification is shown;next to it are enlarged views of sequential images of an area of melanocyte-keratinocyte contact (images 2-5). A filopodia arising from the lateral aspect of the tip of the dendrite overlays or is attached to the keratinocyte membrane. Sequential images demonstrate melanosomes moving upward towards the keratinocyte in single file over the course of 80 seconds. Melanosomes were also frequently observed in filopodia that arose from the body of the dendrite, even in the absence of contact with a keratinocyte(Fig. 2a; image 6). In addition to filopodia, which were easily recognizable owing to their dynamic motion and thin diameter, we also observed shorter thicker projections arising predominantly from the sides of melanocyte dendrites, which contacted keratinocytes (Fig. 2b). Melanosomes were transported along these structures singly or in pairs towards the keratinocyte membrane.

Because the optical properties of the filopodia and the keratinocyte membrane are similar, we were unable to definitively determine whether membrane fusion occurred using this technique. In an initial attempt to address this question we utilized two lipophilic fluorescent membrane dyes to separately label melanocyte and keratinocytes, followed by co-culture of the cells after a single dose of irradiation to stimulate melanosome transfer. Lipophilic dyes have been commonly used to assess cell fusion in other cell types (Sowers, 1985;Spotl et al., 1995) and show little leakage of one dye to another. DiI absorbs maximally at 546 nm and has a maximum emission at 563 nm. DiO, a closely related compound, absorbs maximally at 489 nm and its peak emission is at 499 nm(Sims et al., 1974;Montecucco et al., 1979; Honig and Hune, 1986). Our preliminary experiments showed that these lipophilic dyes are rapidly incorporated into melanocyte and keratinocyte cell membranes where they persist for weeks in culture and show little if any cytotoxicity. 24 hours after irradiation of co-cultures that had been separately labeled with DiI and DiO, approximately 10% of keratinocytes exhibited yellow fluorescence when viewed with a filter to detect both dyes(Fig. 2c). Yellow fluorescence was observed in an intracellular vesicular pattern, which resembled endosomes,as well as in larger deposits in the Golgi area. The presence of yellow fluorescence within endosome-like structures is consistent with membrane fusion between melanocytes and keratinocytes, with subsequent recycling of the fused membranes into recycling endosomes and transport to the Golgi apparatus. We cannot exclude, however, the possibility that keratinocyte-phagocytosis of melanocyte dendrites resulted in the presence of yellow fluorescence in keratinocytes. In sham-irradiated cells evidence of membrane fusion was observed in approximately 1% of keratinocytes, indicating membrane fusion even in unstimulated cells.

Electron microscopy performed on human melanocyte-keratinocyte co-cultures demonstrated thin projections consistent with filopodia arising from the tips and sides of melanocytes (Fig. 3a). In many cases filopodia were cut in cross section; a longitudinal section of a filopodia is shown inFig. 3b. Melanosomes aligned themselves near the base of the filopodia and osmophilic membrane-bound bodies consistent with melanosomes were observed within the filopodia(Fig. 3c,d). In some cases we observed direct connection between melanocytes and keratinocytes in the form of thin projections that spanned a small space between the two cells(Fig. 3e,f). Melanosomes appeared to be passing between the melanocyte and the keratinocyte along these projections (Fig. 3f). In human skin in vivo we detected structures consistent with filopodia arising from the sides and tips of dendritis which contained melanosomes within their lumina(Fig. 3g-j).

Because of the well known role of Cdc42 in mediating filopodia formation,we next examined the effect of expression of constitutively active Cdc42 on melanocyte morphology, filopodia formation and melanosome transfer to keratinocytes. To investigate whether the virus-expressed Cdc42^V12would exhibit the expected properties of a constitutively active mutant construct in melanocytes, in vitro binding assays of lysates of cells infected with Adeasy virus expressing Cdc42^V12 and empty vector-infected cells with the PBD were performed (Fig. 4a). The PBD is conserved in several effector proteins of Cdc42 and Rac1 and mediates their interactions in a GTP-dependent manner(Sander et al., 1998). Five days after infection of melanocytes (1×10⁶ cells) with 30 MOI of Cdc42^V12 AdEasy vector, levels of PBD-bound Cdc42 were increased in Cdc42^V12 expressing cells compared with empty vector-infected cells. These results indicate that Cdc42^V12 expressed from adenovirus is functionally active in melanocytes. The amount of Cdc42 activation is likely to be underestimated in this experiment because infection efficiency was only 60% and therefore Cdc42 from non-infected cells diluted the amount of activated Cdc42. Because the GFP and Cdc42^V12 cDNAs are driven from separate CMV promoters, we confirmed that GFP-expressing cells also overexpressed Cdc42 by staining infected cells with antibodies to Cdc42. Virtually all cells that expressed GFP also overexpressed Cdc42 (not shown).

Melanocytes were infected with 30 MOI of Cdc42^V12 Adeasy virus or 30 MOI of Adeasy virus (empty vector), and 1 day later keratinocytes were added at a ratio of 1:1. The morphological features of melanocytes 5 days after infection with Cdc42^V12 Adeasy vector, viewed under immunofluorescence microscopy, are shown inFig. 4b (images 1-3). Melanocytes exhibited multiple arborizing dendrites and some cells displayed a growth-cone like morphology. Cells infected with empty vector maintained a bipolar morphology typical of melanocytes grown in the absence of phorbol esters (image 4). To determine whether expression of Cdc42^V12resulted in increased numbers of melanosome-containing filopodia, infected cells were stained with antibodies against TRP-1(Fig. 4c; images 1 and 2). Filopodia were clearly visible in infected cells owing to the presence of numerous melanosomes within them. Vector-expressing cells stained for TRP-1 showed some melanosomes in filopodia but they were less numerous than in Cdc42^V12-expressing cells. (Fig. 4c; images 3 and 4). Time-lapse digital microscopy of these cultures viewed under DIC optics confirmed that Cdc42^V12-infected cells exhibited multiple long filopodia arising from dendrite tips(Fig. 4d, images 1-3) and in many cases filopodia contained melanosomes. Cdc42^V12 expressing melanocytes showed more extensive contacts with keratinocytes through filopodia. As expected, empty vector expressing melanocytes viewed under DIC optics showed filopodia arising from the tips of dendrites, but the number of filopodia was dramatically fewer than in Cdc42^V12-expressing cells(images 4 and 5).

We next examined the expression and localization of Cdc42 and PAK1 in human melanocytes by immunofluorescence microscopy(Fig. 5). We were unable to perform immunofluorescence staining with the antibodies to N-WASP available to us. Cdc42 was present in a vesicular pattern with prominent localization to the melanocyte cell membrane, as well as in the peri-nuclear area(Fig. 5a; image 1). The localization of Cdc42 to the peri-nuclear area (presumed to be the Golgi apparatus) is consistent with previous reports showing that Cdc42 localizes to the Golgi (Erickson et al., 1996). PAK1 was heavily concentrated in the peri-nuclear area with some vesicular staining in the dendrites(Fig. 5a; image 2). Staining of cells with normal rabbit serum instead of a primary antibody failed to show any labeling (not shown). To determine if Cdc42 or PAK1 colocalized with melanosomes, double labeling with antibodies to the melanosome-specific protein mel-5 (TRP-1) and either Cdc42 or PAK1 was performed(Fig. 5b). PAK1 and TRP-1 colocalized in the perinuclear region as well as focally in the melanocyte dendrites (images 1-3). Because melanosomes are heavily concentrated in the peri-nuclear area, it is difficult to determine whether this staining pattern represents true colocalization or an artefact of overlay of melanosomes and other PAK1-expressing structures in this area. Higher power images of melanocyte dendrites from double-labeled cells (lower panel,Fig. 5b) shows clear colocalization of PAK1 and TRP-1 within the dendrites; however colocalization was not 100%. Cdc42 did not colocalize with melanosomes but did colocalize with the transferrin receptor, indicating a component of Cdc42 in recycling endosomes (not shown). Myosin Va in melanocytes colocalizes with melanosomes,the endoplasmic reticulum, Golgi apparatus and mitochondria(Nascimento et al., 1997;Tabb et al., 1998), and myosin Va has been shown to play an important role in filopodia extension in neuronal cells, and in murine melanocytes myosin Va has been identified at the tips of filopodia (Wang et al., 1996;Tsakraklides et al., 1999). Myosin Va was also present in a punctate or dot-like pattern at the tips of filopodia and along the length of the filopodia(Fig. 5c; image 1 and 2).

Western blots were performed for analysis of expression of Cdc42, PAK1 and N-WASP in melanosome fractions (Fig. 6). The purity of the melanosome isolate was assessed by electron microscopy, which demonstrated a relatively homogeneous population of stage III and stage IV melanosomes with few if any contaminating elements such as mitochondria (Fig. 6a). Transferrin receptor expression, used as a marker for recycling endosomes, was not detected in melanosome-enriched fractions, indicating low amounts of contaminating recycling endosomes in the preparation(Fig. 6b). PAK1 was heavily enriched in melanosome extracts; a single strong immunoreactive band was detected (Fig. 6c; 20μg/lane). Western blotting for N-WASP showed an immunoreactive band at the expected molecular weight for N-WASP in melanosome extracts(Fig. 6c; 70 μg/lane). Cdc42 was not detected in melanosome-enriched fractions even when large amounts of protein (up to 80 μg) were loaded onto the gel (not shown).

Through the use of time-lapse digital imaging of melanocyte-keratinocyte co-cultures, we identified filopodia as a conduit for melanosome transfer to keratinocytes. Our ability to create greatly enlarged movies of areas of interest from digital images acquired with a 100× objective allowed us to directly visualize melanosomes, which have an approximate diameter of 0.5-1.0 μm, in real time. These movies, in combination with DIC optics,also allowed us to detect filopodia through their rapid movement and enhanced appearance under phase contrast. Even though time-lapse video microscopy has been used to examine melanosome transfer in the past, we believe that the relatively low resolution of these movies, compared with the high resolution of digital images, prevented detection of filopodia and the role they play in melanosome transfer. Electron microscopy demonstrated melanosomes within structures consistent with filopodia, confirming observations made from time lapse imaging. In human skin in vivo we detected structures with morphologic features consistent with filopodia, many of which contained melanosomes within them. Expression of constitutively active Cdc42 protein in melanocytes resulted in a marked increase in melanosome-containing filopodia and in filopodia attachment to keratinocytes. We also demonstrate the presence of N-WASP and PAK1, Cdc42 effector proteins, on enriched melanosome fractions. Myosin Va, which is an actin-based motor protein for melanosome transport and for filopodia extension, was detected at the tips of melanocyte filopodia.

The well known role of Cdc42 in filopodia formation, as well as data showing increased filopodia formation in melanocytes expressing a constiutively active Cdc42 mutant, are consistent with a role for Cdc42 in mediating filopodia formation in human melanocytes. Although at this point we can only speculate on the motors involved in moving melanosomes along the length of the filopodia, the presence of myosin Va at the tips of filopodia and the fact that filopodia contain actin but not microtubules makes myosin Va a strong candidate. The rapid movement and attachment of melanocyte filopodia to keratinocyte membranes is highly analogous to growth cone filopodia, which contact nearby axons with subsequent synapse formation and synaptic vesicle transmission. In neuronal cells the primary role of growth cone filopodia is to sample the immediate environment and translate environmental cues to the growth cone, which in turn affects growth cone behavior(Rosentreter et al., 1998;Harris, 1999). Recent reports show that neuronal filopodia respond to exogenous growth factors such as fibroblast growth factor by doubling their length(Szebenyi et al., 2001). In a similar manner, melanocytes respond to keratinocyte-derived growth factors through dendrite extension and possibly melanosome transfer(Imokawa et al., 1995;Hara et al., 1995). It is likely that melanocyte filopodia respond to a gradient of keratinocyte-derived growth factors that direct filopodia growth and attachment and subsequent transfer of melanosomes, although further experiments are needed to establish this. Sabry et al. (Sabry et al.,1991) showed that microtubules invade neuronal filopodia during growth cone steering events subsequent to filopodia attachment to guidepost cells. Therefore, another potential function of melanocyte filopodia may be to serve as a vanguard for dendrite extension. We also frequently observed blunt,short projections arising from the shafts of melanocyte dendrites, which also functioned as conduits for melanosome transfer to keratinocytes. These projections may be analogous to dendritic spines, which, in neuronal cells,are actincontaining, short bulbous projections that arise from the sides of neuronal dendrites (Harris,1999; Kaech et al.,2001).

Activated Cdc42 induced a markedly dendritic morphology in human melanocytes. Because activation of Rac1 results in dendrite extension in melanocytes (Scott and Cassidy,1998) and because others have shown that inhibition of RhoA results in dendrite extension in melanocytes(Busca et al., 1998), signaling for melanocyte dendrite extension may be analogous to N1E-115 cells in which activated Rac1 inhibits RhoA with subsequent dendrite extension(Altun-Gultekin and Wagner,1996; Kozma et al.,1997). A hierarchical, unidirectional cascade of activation of Cdc42, Rac and Rho has been described in a variety of cell types (for a review, see Kjoller and Hall,1999). In most cell types, Cdc42 activates Rac1, which leads to inhibition of Rho activity (Sander et al.,1999; Reid et al.,1999). These studies suggest that in melanocytes Cdc42 may be upstream of Rac1 in dendrite formation through activation of Rac1, which in turn inhibits Rho A. Cdc42 is activated by the inflammatory cytokines tumor necrosis factor-α and interleukin-1(Wojciak-Stothard et al.,1998; Puls et al.,1999), both of which are released by keratinocytes following UV irradiation, a potent stimulus for melanosome transfer(Pathak et al., 1978;Sturm, 1998;Kondo, 1999). Therefore the well known effect of UV irradiation on melanocyte dendrite formation and melanosome transfer may be mediated in part by interleukin-1- and tumor-necrosis-α-induced activation of Cdc42.

The role of PAK1 and N-WASP, which were enriched in melanosome fractions,in melanosome movement is unclear and must await further experiments analyzing the effect of mutants of these proteins on melanosome transport and transfer. A strong association between PAK1 activation and lamellipodia formation, loss of stress fibers, disassembly of focal adhesions and increased cell motility has been demonstrated (Frost et al.,1998; Sells et al.,1999). Daniels et al. (Daniels et al., 1998) have shown that nerve-growth-factor-induced neurite outgrowth in PC10 cells is mediated by PAK1. We have shown previously that Rac1 mediates melanocyte dendrite formation in response to growth factors and to UV irradiation(Scott and Cassidy, 1998). It is possible that in melanocytes, which, similar to PC10 cells, are neuronally derived cells, PAK1 is a downstream effector for Rac1-mediated dendrite extension. Although PAK1 is heavily enriched on melanosomes, activated PAK1,as identified by antibodies to phosphorylated PAK1 (a generous gift of Dr Chernoff, Fox Chase Cancer Center, Philadelphia, PA) was not associated with melanosomes but with the small granule fraction of melanocytes as demonstrated by western blotting (G.S., unpublished). Therefore activation of PAK1 is unlikely to occur on the melanosome membrane.

It is likely that melanosome transfer is accomplished through multiple mechanisms, including phagocytosis of dendrite tips and possibly exocytosis of the melanosome into the extracellular space with uptake by keratinocytes(Yamamoto and Bhawan, 1994). Although we did not directly observe phagocytosis of melanocyte dendrites in time-lapse movies, this may have been due to the relative infrequency of melanosome transfer in culture, lack of appropriate stimulus or both. Although the digital movies presented (jcs.biologists.org/supplemental or www.urmc.rochester.edu/derm/scottmovies.html) provide intriguing evidence of a role for filopodia in melanosome transport,we are unable to definitively conclude that melanosome transfer occurred because of the optical properties of the melanocyte and keratinocyte membranes. Our initial attempt to circumvent this problem using membrane dyes is suggestive of keratinocyte-melanocyte membrane fusion but is not conclusive. We believe that definitive evaluation of melanosome transfer will require in vivo labeling of melanosomes with a marker that would allow one to observe movement of melanosomes from the melanocyte to the keratinocyte, in combination with high resolution digital movies. This would allow one to assess transferred melanosomes within keratinocytes both through direct visualization, and through biochemical means, in response to expression of mutants of a variety of candidate proteins, including PAK1, N-WASP and Cdc42. At the present time we are evaluating the ability of a GFP-Pmel17 fusion protein to label human melanocyte melanosomes.

This work was supported by 1R01AR45427(GS). We thank Karen Jensen for her assistance with electron microscopy.