Visualization of Melanosome Dynamics within Wild-Type and Dilute Melanocytes Suggests a Paradigm for Myosin V Function In Vivo

Control primary melanocytes were prepared from the skin of C57BL/6J (D/D) pups, while dilute melanocytes were prepared from the skin of C57BL/6J pups that were homozygous for the dilute lethal allele d^l20J, which represents a true null allele at dilute (see Strobel et al., 1990 and Fig. 1 in Wu et al., 1997). These latter pups were obtained by crossing heterozygous parents of the genotype d^l20J/d^vse (short ear [se]¹) (a generous gift of Dr. Neal Copeland and Dr. Nancy Jenkins, NCI, NIH), and were distinguished from their d^l20J/d^vse and d^vse/d^vse littermates based on Western blot analysis, as described previously (Wei et al., 1997; Wu et al., 1997). Only homozygous d^l20J pups are devoid of detectable myosin Va heavy chain protein (Wu et al., 1997). Cultures of primary melanocytes were prepared from the epidermis of individual mice by a modification of the method of Tamura et al. (1987), as described previously (Wei et al., 1997), and were grown on gelatin-coated 24.5-mm round No. 1 1/2 coverslips in Ham's F-10 medium supplemented with 10% horse serum, 2% fetal calf serum, 1% (vol/vol) penn strep, 0.1 μM dbcAMP, and 85 nM tetradecanoyl phorbol acetate (TPA) (No. P8139; Sigma Chemical Co., St. Louis, MO). The nontransformed but immortal mouse melanocyte cell line melan-a (D/D) was a generous gift of Dr. Dorothy Bennett (St. George's Hospital Medical School, London, UK) and was maintained as described previously (Wu et al., 1997).

The dynamics of melanosomes in melanocytes was monitored using a Zeiss Axiovert 135 microscope (Carl Zeiss, Thornwood, NY) equipped with differential interference contrast (DIC) optics, a Plan Apochromat 100× 1.25 NA oil immersion lens, and illumination provided by a halogen lamp through a green interference filter. Bright field or DIC images were recorded using a Newvicon camera (model C2400 Newvicon; Hamamatsu, Hamamatsu City, Japan) and were captured and stored at a rate of one frame (1/30 s) every second using an optical disc recorder (model TQ303IF; Panasonic) controlled through a PC. Background subtraction and contrast enhancement were performed in real time before recording images using an Argus-10 image processor (Hamamatsu).

Cells grown on 24.5-mm round coverslips were introduced into a Dvorak-Stotler perfusion chamber (Nicholson Precision Instruments, Gaithersburg, MD) that was fitted with a custom-made elliptical chamber spacer (∼35 μl internal volume), and with a stainless steel, two-way inlet manifold to allow for changes in the perfusion media. Perfusion medium was identical to the tissue culture medium used for culturing the cells and was delivered to the chamber by gravity feed at a flow rate of 0.2 ml/min, which allowed complete exchange of the chamber solution every ∼10 s. Perfusion medium was warmed to 37°C and equilibrated with humidified 5% CO₂ using specially designed perfusion bottles (gift of Dr. Ken Springs, NHLBI, NIH). Solenoid-actuated pinch valves (Custom Electronics, Gaithersburg, MD) were used to switch between perfusion media of different composition. To maintain the chamber at 37°C, the entire microscope was placed inside a custom made enclosure connected to a forced air heater (Air Therm; World Precision, Inc., Sarasota, FL). The stage temperature was monitored using a temperature probe (model YSI 400; Coparma) placed adjacent to the perfusion chamber. Using this arrangement, the cells could be maintained for many hours in standard culture medium at 37°C. To depolymerize microtubules, cells were perfused at 37°C with medium containing 9.6 μg/ml (16 μM) nocodazole (No. M1404; Sigma Chemical Co.) diluted from a freshly prepared 5 mg/ml stock dissolved in DMSO. To drive depolymerization further, cells that had been perfused with nocodazole at 37°C for at least 10 min were placed on ice for 30 min in the presence of nocodazole, and then perfused at 37°C in the presence of nocodazole.

To prepare plasmid EGFP-MC-ST, a cDNA clone encoding the COOH-terminal ∼75% of mouse myosin Va was obtained from a B16 mouse melanoma cDNA library (gift of Dr. Murray Brilliant, University of Pennsylvania, Philadelphia, PA) and digested with SstI and EcoRI, and the ∼2.2-kb SstI/EcoRI fragment that begins at residue 1258 (numbering according to Mercer et al., 1991) and ends at the first poly A⁺ site was cloned into plasmid EGFP Cl (No. 6094-1; CLONTECH Labs, Palo Alto, CA). The fusion protein generated by this plasmid contains the COOH-terminal 619 amino acids of the predominant spliced isoform found in melanocytes (including both melanocyte-specific exons—exons D [27 residues] and F [25 residues]—identified in Seperack et al., 1995) fused to the COOH terminus of an enhanced version of green fluorescent protein (EGFP; 238 amino acids). To prepare plasmid FLAG-MC-LT, the cDNA clone described above was digested with Eco47III and converted to an EcoRI site by linker addition, and the ∼2.7-kb EcoRI fragment that begins at residue 1091 and ends at the first poly A⁺ site was cloned into plasmid pCMV2-FLAG (IBI Eastman Kodak, Rochester, NY). The fusion protein generated by this plasmid contains the COOH-terminal 786 amino acids of the predominant spliced isoform found in melanocytes fused to the COOH terminus of the FLAG epitope tag (nine amino acids). For transient transfections, melan-a melanocytes were grown to 70% confluence on 12-mm round No. 1 1/2 coverslips (for fixation/immunostaining), 24.5-mm round No. 1 1/2 coverslips (for live cell imaging in the Dvorak-Stotler Chamber), or chamber slides with No. 1 coverslip bottoms (No. P136307; Lab-tek, Naperville, IL) (for short-term viewing of live cells on the confocal microscope) and transfected with 1 μg of CsCl-purified plasmid DNA per 2.5 ml of culture media using FuGENE 6 transfection reagent (No. 1-814-443; Boehringer Mannheim, Indianapolis, IN), as described by the manufacturer. Cells were imaged or fixed/stained at various intervals between 24 and 60 h after transfection.

Cells transfected with plasmid EGFP-MC-ST were fixed for 30 min at room temperature in freshly prepared 4% paraformaldehyde (No. 00380; Poly Sciences, Warrington, PA) in serum-free culture medium prepared using MEM powder (No. 41600-016, GIBCO BRL; Life Technologies, Gaithersburg, MD), washed briefly in PBS, and mounted on glass slides using Slow Fade (No. S-2828; Molecular Probes, Eugene, OR) and rubber cement. Cells transfected with plasmid FLAG-MC-LT were fixed as above, washed with PBS, blocked for 15 min in three changes of PBS containing 10% (vol/vol) fetal calf serum, incubated for 1 h at room temperature with anti-FLAG monoclonal antibody M5 (No. IB13091; Eastman Kodak) diluted 1:250 in blocking buffer supplemented with 0.2% saponin (No. S790; Sigma Chemical Co.), washed with PBS, incubated for 1 h with Texas red–labeled goat anti–mouse IgG (No. 115-025-146; Jackson ImmunoResearch Labs, Inc., West Grove, PA) diluted 1:100 in saponin-containing blocking buffer, washed with PBS, and mounted using Slow Fade and nail polish. To stain for F-actin and microtubules, cells were fixed in 4% paraformaldehyde dissolved in cytoskeletal stabilization buffer (Conrad et al., 1993) and processed as described above using Oregon green 488-labeled phalloidin (No. 0-7466; Molecular Probes) at a final concentration of 30 nM, a rat monoclonal antibody to tubulin (No. MAS 077P; Accurate Chemical) at a dilution of 1:100, and Texas red–labeled donkey anti–rat IgG (No. 712-075-153; Jackson ImmunoResearch Labs). For conventional light microscopy, samples were viewed with a Zeiss Axioplan microscope using either a 63× Plan-Apochromat Phase 3 objective or a 100× Plan Neofluar objective, and were recorded using Kodak TMAX film (400 ASA). For confocal microscopy, samples were viewed with a Zeiss LSM 410 confocal microscope using either a 63× Plan Neofluar objective for fixed cells or a 63× water-immersion objective for living cells in chamber slides. Images, which are either single, 1-μm-thick optical sections, or superimposed stacked images representing the full thickness of the cell (see figure legends), were transferred to Adobe Photoshop and printed using a Fuji (Tokyo, Japan) digital color printer. The fluorescent image in Fig. 7 H was obtained using an intensified CCD camera (model C2400; Hamamatsu) attached to the microscope for live cell imaging.

Melan-a melanocytes were permeabilized for 5 min at 37°C in PHEM buffer (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 2 mM MgCl₂, pH 6.9) supplemented with 0.02% saponin and 80 nM phalloidin (No. P-3457; Molecular Probes). The cells were then incubated for 30 min at room temperature in PHEM buffer containing 0.5 mg/ml rabbit skeletal muscle subfragment 1 (a generous gift of Dr. James R. Sellers, NHLBI, NIH), rinsed twice (5 min each) in PHEM buffer, and rinsed once (3 min) in 0.1 M sodium phosphate buffer, pH 6.8 (PB) (Lewis and Bridgman, 1992). The cells were then fixed for 30 min at room temperature in PB containing 2% glutaraldehyde, rinsed briefly with PB, and treated as described by McDonald (1984). In brief, cells were postfixed for 15 min on ice in PB containing 0.5% OsO₄, 0.8% K₃Fe (CN)₆, and 1 mM MgCl₂, rinsed three times (5 min each) in distilled water, mordanted for 1 min in PB containing 0.15% tannic acid, rinsed in PB, and stained en bloc for 2 h at room temperature in 2% aqueous uranyl acetate. The cells were then dehydrated in graded alcohols and embedded in Epon. Thin sections were examined using a Phillips 410 electron microscope (Mahwah, NJ). Actin polarity was scored as described in the text. Actin filaments within 2 μm of distal dendrite membrane were used to determine actin orientation at dendritic tips.

Individual video frames acquired at 1-s intervals were transferred from the OMDR to a Macintosh Quadra 950 computer, and melanosome centroid positions in successive frames were hand digitized. Path plots were assembled using Dias software (version 1.2; Solltech, Inc., Iowa City, IA). The speed of each 1 s step was calculated using the direct method as follows:

where i is the i^th frame, Δt is the time interval (1 s), and P is the centroid position. The length of each step was the distance between consecutive centroid positions. The speed and path lengths of microtubule-dependent movements reported in Table I were calculated based on individual path plots representing one number. The maximum speed of each path plot is the average of the fastest step and any additional steps that were at least 70% the speed of the fastest step. The average speed for each path plot is the average of all the steps that were faster than 0.35 μm/s. Path lengths are the sum of all the individual steps in the path. Averages, standard deviations, and significance values (performed using the Student's t test) were calculated using Excel.

Light immunofluorescence micrographs of melan-a melanocytes (D/D) stained with fluorescent phalloidin (Fig. 1, C and E) typically reveal sparse staining in the central cytoplasm, intense edge staining along the length of the dendrite, and prominent staining at dendritic tips (see also Lacour et al., 1992; Wu et al., 1997). Confocal microscopy revealed that the edge staining in dendrites is due to a cortical shell of F-actin under the plasma membrane (data not shown). To determine the orientation of actin filaments within this cortical shell, melan-a melanocytes were permeabilized with detergent, labeled with rabbit skeletal muscle myosin subfragment 1, fixed, processed for transmission electron microscopy, and sectioned such that the cell body and as much of the dendritic extension as possible were present in a single section (Fig. 2,A). As in the light micrographs, the electron micrographs show that actin filaments are scarce in the central cytoplasm and positioned in dendrites almost exclusively just under the plasma membrane, forming a cortical shell that surrounds a microtubule-rich, actin-poor central zone. Actin polarity in dendrites was scored in a series of 1-μm-wide cross-sectional bands that spanned the short axis of the dendrite and that ranged from the base of the dendrite to where the dendrite left the plane of the section. A total of 28 such bands from seven dendrites, each from a different melanocyte, and each spanning a distance of between 20–30 μm from the nucleus, were scored. Decorated filaments were grouped into one of three categories, as described in the schematic in Fig. 1 B. Those in which the arrowhead pointed within a 90° arc whose center points directly away from the nucleus were scored as pointed end–out. Those in which the arrowhead pointed within a 90° arc whose center points directly towards the nucleus were scored as barbed end–out. Those that fell into the two remaining 90° arcs were scored as “other.” Three main points summarize what was found. First, the number of barbed end–out filaments per 1-μm cross-sectional band (4 ± 2.3) was not significantly different from the number of pointed end–out filaments (3.3 ± 2.2) (n = 28). Second, in several instances where the number of barbed end–out filaments exceeded pointed end–out filaments by a factor of greater than 3:1, the filaments in the adjacent 1-μm cross-sectional areas always exhibited an inverse ratio (data not shown). Third, of the 367 filaments scored, 112 (31%) were barbed end–out, 92 (25%) were pointed end–out, and 163 (44%) fell into the “other” category. These three results indicate that actin filament orientation in dendrites is essentially random. Using other thin sections that contained a total of 19 melanosome-rich dendritic tips, we also found that the number of barbed end–out filaments per tip (5.3 ± 3.5) was not significantly different from the number of pointed end–out filaments per tip (4.2 ± 2.9) (n = 19). Of 227 filaments scored in tips, 100 (44%) were barbed end–out, 80 (35%) were pointed end–out, and 47 (21%) were “other.” Therefore, as in the dendrite proper, the actin in dendritic tips exhibits a largely random orientation. Together, these results indicate that if myosin Va is responsible for the long-range centrifugal transport of melanosomes, it would be via a mechanism wherein random walks of melanosomes catalyzed by myosin Va on actin filaments of random orientation drive the outward spreading or “facilitated diffusion” of the organelle.

In contrast to the mechanism described above, melanosomes in primary dilute null melanocytes were found to be undergoing rapid, bidirectional translocation between the cell center and the periphery. Fig. 3,A shows a still image of one such melanocyte (d^l20J/d^l20J) possessing a broad, well-spread cell extension projecting to the upper right. As is typical of heavily melanized dilute melanocytes (Wei et al., 1997), the majority of melanosomes are perinuclear, although some are clearly present in the broad flat cell extension. Time lapse video microscopy revealed that many of these latter melanosomes are moving rapidly (>1 μm/s) and bidirectionally between the cell center and the periphery. Two examples of these melanosome movements, which are very obvious when viewed at ∼15–30 times real speed, are shown in successive video stills (1-s time intervals) in Fig. 3, B1–B10 (centrifugal) and C1–C10 (centripetal). D and E show the paths (1-s time intervals) for some of the melanosomes whose centrifugal (D) and centripetal (E) movements within the boxed area in A were obvious to the eye over a 3-min time span (see video segment No. 1).² While not all melanosomes traverse the entire cell extension in one continuous movement, some do (for example, see melanosome numbers 5 and 11 in D and E, respectively). Such long-range persistent movements are even more obvious in the lightly melanized mutant cell shown in Fig. 3,F, where melanosome numbers 2–4 (G, closed circles) move from near the nucleus to near the margin of the cell without significant pause, while melanosome numbers 7 and 13 (G, open circles) move from near the cell margin to near the nucleus with only a few short pauses. In typical fashion, essentially all of the melanosomes in the periphery of this cell are undergoing rapid bidirectional movement (or are temporarily pausing; see video segment No. 2), and individual melanosomes can be seen to move by this fast component in rapid succession from the center to the periphery and back again. One of these melanosomes is shown in successive video stills in Fig. 3, I1–I11, while H shows the paths for four different melanosomes undergoing such bidirectional movements. Fig. 3, J–L, shows that this bidirectional melanosome movement is also evident within longer cell extensions that are more typical of melanocyte dendrites in vivo (see video segment No. 3). Finally, Fig. 3, M and N, shows very striking bidirectional melanosome movements within a large dendritic extension emanating from a mutant cell (see video segment No. 4). These movements are notable in that they persist for long distances (melanosome 1, 48 μm; melanosome 6, 24 μm).

From analysis of these and other dilute melanocytes, we obtained the following information regarding this bidirectional component of melanosome motility. First, individual melanosomes move on average 13.3 ± 6.9 μm (n = 34) and 11.9 ± 8.5 μm (n = 34) between stops for centrifugal and centripetal movements,³ respectively (Table I). Second, stops may be short pauses on the order of seconds or longer pauses lasting for more than a minute, and pauses may or may not be associated with reversals in direction. Third, melanosomes that reverse direction sometimes appear to follow the same track in reverse (for example, see the path in the upper left corner of the Fig. 3,H). Fourth, while long (i.e., 20–30 μm) movements can occur without significant pauses, it is more common to see such movements interrupted by several pauses lasting on the order of a few seconds. Fifth, the average and maximum speeds of melanosomes undergoing this movement are 1.04 ± 0.29 and 1.67 ± 0.64 μm/s (n = 34), respectively, for centrifugal movements, and 1.13 ± 0.35 and 1.86 ± 0.67 μm/s (n = 34), respectively, for centripetal movements (Table I). These values for centrifugal and centripetal movements are not significantly different. Finally, while we did not attempt to quantitate the numbers of centrifugal and centripetal movements per unit time, their frequencies appear to be essentially the same. Indeed, they must be, given the phenotype of dilute null melanocytes, where net accumulation of melanosomes in the cell's periphery does not occur. Therefore, while this component of melanosome motility provides an efficient means to move melanosomes to the periphery, it does not by itself generate the net peripheral accumulation of melanosomes characteristic of wild-type melanocytes.

The fact that the component of melanosome motility described above runs parallel to the long axis of cell extensions, exhibits linear trajectories often exceeding 15 μm in length without pausing, and possesses speeds characteristic of microtubule-dependent motor proteins (Hirokawa, 1998), together with the fact that in melanocytes stained with an α tubulin antibody, microtubules were found to run the length of dendritic extensions (Fig. 1,B), stopping just shy of the dendritic tip (Fig. 1, F and G), all suggest that this component of melanosome motility is microtubule dependent. To verify this, melanosome dynamics within dilute melanocytes were recorded before and after the addition of nocodazole. Fig. 4 B shows the paths for a portion of the fast centrifugal (left) and centripetal (right) melanosome movements that occurred over a 2-min time span within the dendritic extension of the dilute melanocyte in A just before the addition of nocodazole. C shows this same cell 5 min after the addition of 16 μM nocodazole to the media (37°C), while D shows the paths for every melanosome in this dendrite over the ensuing 1-min time interval (see video sequences Nos. 5A and 5B). Clearly, most if not all fast melanosome movements have disappeared (the slow displacements of a few melanosomes are due to local contractions of the cytoplasm). Occasional fast movements that were sometimes seen in dendrites of other similarly treated melanocytes completely disappeared when nocodazole treatment was supplemented with a 30-min incubation at 4°C, and fast bidirectional melanosome movements resumed when cells were returned to 37°C in the absence of nocodazole (see below). These observations, together with immunofluorescence staining of melanocytes incubated at 37°C in 16 μM nocodazole for 5 min, which indicated that this treatment largely eliminated microtubules from the periphery of melanocytes (data not shown), support the conclusion that the fast bidirectional component of melanosome movement is microtubule dependent. In addition, the nearly complete abrogation of this movement within cell extensions produced by nocodazole addition alone (i.e., without a 4°C incubation) suggests that microtubules within dendritic extensions are highly dynamic.

Time lapse video microscopy of untreated and nocodazole-treated dilute melanocytes revealed that essentially all of the melanosomes in the central cytoplasm of these cells are also moving rapidly on microtubules. Fig. 4,E shows a still image of the same lightly melanized dilute melanocyte described in Fig. 3, F–I, while Fig. 4,F shows the paths for a portion of the fast melanosome movements seen in the central cytoplasm of this cell (boxed area in E) over a 2-min time span before the addition of 16 μM nocodazole. These fast movements exhibit nearly the same speed as fast movements in the periphery (1.05 ± 0.44 μm/s, n = 15 from this cell), but differ in that their average excursions are much shorter (3.1 ± 1.3 μm, n = 15 from this cell). G shows this same cell after incubation at 37°C for 10 min, at 4°C for 30 min, and at 37°C for 10 min, all in the presence of 16 μM nocodazole, while H shows the paths for all melanosomes in the boxed area over the ensuing 1 min. Clearly, most if not all of the short, fast movements have disappeared (see video segments Nos. 6A and 6B). We also found that these central, short-range movements reappeared first upon nocodazole washout, with peripheral movements following some time later once cell extensions have reformed. This latter point is demonstrated in I and J, which show this same cell 30 min after nocodazole wash out (see video segment No. 7). We conclude, therefore, that the majority of melanosomes in dilute melanocytes are undergoing microtubule-dependent translocation, with most moving on the large number of short microtubules (see Fig. 1) that terminate in the central cytoplasm, while the rest ride on the smaller number of long microtubules that extend to the periphery of the cell.

Fig. 5,A shows a primary wild-type melanocyte, which exhibits the typical bias in peripheral distribution of melanosomes characteristic of myosin Va⁺ melanocytes, while B shows a portion of the fast centrifugal and centripetal melanosome movements that occurred within this cell over a 2-min time span. As in many wild-type melanocytes, these fast movements, which in well-spread cells like this one are most evident in the region between the nucleus and the melanosome-rich periphery, are considerably less obvious than in dilute melanocytes (see video sequence No. 8). This appears to be due in part to the fact that the peripheral cytoplasm of wild-type melanocytes, unlike dilute melanocytes, contains significant numbers of melanosomes that are not undergoing this fast movement and that block the view of those that are. Striking bidirectional melanosome movements in wild-type melanocytes are not uncommon, however, especially within dendritic extensions. D and G show the paths for a portion of the fast centrifugal (closed circles) and centripetal (open circles) melanosome movements that occurred over a 3-min time span within the boxed area of the dendritic extensions in C and F, respectively. In addition to the melanosomes that are undergoing these fast movements (which occur mostly in the central, microtubule-rich zone of the dendrite), wild-type cells possess large numbers of melanosomes that are at the tip (C) and along the edge (F) of their dendrites and that are undergoing slow, local motions (see video sequences Nos. 9 and 10 and Figs. 9 and 10). Importantly, while microtubule-dependent melanosome movements in these actin-rich zones are rare, melanosomes can be removed from the tip (E) and lateral edge (H) of the dendrite by the fast component, and new melanosomes can be delivered to the tip (E) and lateral edge (H) by the fast component. The extent to which the balance of these dynamics can change the degree of peripheral accumulation of melanosomes was evident in several instances where a surfeit of centrifugal or centripetal movements led to relatively rapid (on the order of minutes) increases or decreases, respectively, in the degree of peripherally accumulated melanosomes. For example, just before the capture of images used for the data in F–H, a surfeit of centrifugal melanosome movements spanning a period of ∼10 min lead to an approximate doubling of the number of melanosomes at the tip (I and J).

From analysis of these and other wild-type melanocytes (e.g., melan-a cells, which are essentially indistinguishable from primary wild-type melanocytes), we obtained the following information regarding this fast component in the context of wild type cells. First, as in dilute melanocytes, it is dependent on the presence of intact microtubules (see below). Second, the average and maximum speeds of melanosomes undergoing these movements in wild-type cells are 0.68 ± 0.17 and 0.95 ± 0.41 μm/s (n = 38), respectively, for centrifugal movements, and 0.76 ± 0.19 and 1.03 ± 0.33 μm/s (n = 34), respectively, for centripetal movements (Table I). While these values for centrifugal and centripetal movements are not significantly different from each other, they are all significantly slower than the values for dilute melanocytes (P < 0.001). Third, individual melanosomes move on average 4.6 ± 2.5 μm (n = 38) and 5.4 ± 2.7 μm (n = 34) between stops for centrifugal and centripetal movements, respectively (Table I). While these values are not significantly different from each other, they are both significantly shorter than the values obtained for dilute melanocytes (P < 0.001). These differences may be due to a myosin Va–dependent damping of both the rate and persistence of microtubule-dependent melanosome movements in vivo (see Discussion).

The fact that the fast, bidirectional, microtubule-dependent component of melanosome transport identified above can move melanosomes to the periphery efficiently but cannot concentrate them there, together with previous studies showing extensive colocalization of myosin Va and melanosomes in the actin-rich periphery, suggested a mechanism in which a myosin Va–dependent interaction of melanosomes with F-actin in peripheral regions of the cell prevents these organelles from returning to the cell center via the centripetal aspect of this fast component, thereby causing their peripheral accumulation. To test this “capture” model, we expressed the myosin Va tail domain within wild-type melanocytes. If, as conjectured (Mooseker and Cheney, 1996), this portion of the myosin Va molecule contains the site that mediates myosin Va/ melanosome interaction, then it should compete with endogenous myosin Va for binding to the melanosome. This competition should, in turn, serve to uncouple the organelle from the peripheral actin cytoskeleton. Once uncoupled, the melanosome should redistribute to the cell center, the site of greatest microtubule density (see Fig. 1,B), by the action of bidirectional, microtubule-dependent traffic. Melan-a melanocytes (D/D) were transiently transfected with plasmid EGFP-MC-ST (see Materials and Methods), which drives the expression of a fusion protein containing EGFP fused to the COOH-terminal 619 amino acids of the melanocyte-specific isoform of myosin Va. This portion of the myosin Va heavy chain contains the last ∼40% of the central stalk domain (including both melanocyte-specific exons [exons D and F in Seperack et al., 1995]), and all of the COOH-terminal 410-residue globular tail domain. Fig. 6, A and C, shows the appearance of several such transfected cells, each surrounded by a number of untransfected cells. The transfected cells, with their striking perinuclear distribution of melanosomes, are very reminiscent of dilute melanocytes, and contrast sharply with the surrounding untransfected cells in which melanosomes are concentrated in the periphery. Similar results were obtained with plasmid FLAG-MC-LT, which drives the expression of a fusion protein containing the nine-residue FLAG epitope tag fused to the COOH-terminal 786 amino acids of the melanocyte-specific isoform of myosin Va (Fig. 6 E). Conversely, cells transfected with the GFP or FLAG vectors alone did not demonstrate this dilute-like phenotype (data not shown). In agreement with the mechanism proposed above, GFP and FLAG fluorescence in cells exhibiting the dilute-like phenotype was largely coincident with the mass of aggregated melanosomes in the cell center. This coincidence was seen not only in fixed cells (B and F), but in living cells expressing the GFP-tagged myosin Va tail chimera as well (D), indicating that the colocalization is not an artifact of fixation.

Evidence that tail expression actually generates the dilute-like phenotype, and further insight into the mechanism by which it does so, came from analysis of GFP fluorescence in transfected cells that do not show the dominant-negative phenotype. Fig. 7 shows one of these cells, which are common between ∼24 and 36 h of transfection but rare after ∼50 h of transfection, and which are characterized by having peripherally distributed melanosomes (Fig. 7,A) and GFP fluorescence that largely colocalizes with these melanosomes (B, arrows). Given that this pattern is precisely what one would predict must precede the generation of the dilute-like phenotype if the mechanism is as described above, we sought to demonstrate by time lapse microscopy that transfected cells like the one in Fig. 7 do indeed turn into cells exhibiting the dilute-like phenotype. Fig. 8,A shows a transfected melan-a cell that was identified by eye as having peripheral, melanosome-associated fluorescence, while B–F show this cell at five successive 30-min intervals. Over this 150-min time span, melanosomes can be seen to undergo a dramatic redistribution, beginning predominantly at the margin of the cell (Fig. 8,A), and ending predominantly in the cell center (F) (see video sequence No. 11). The distribution of melanosomes and GFP fluorescence in this living cell ∼5 min after the time point corresponding to F are shown in G and H, respectively. Once again, GFP fluorescence is largely coincident with the mass of melanosomes in the cell center. These results, which were replicated in each of three similar transfected cells (all of which took ∼120 min to generate this dominant-negative phenotype), lend strong support to the idea that tail expression initiates the generation of the dilute-like phenotype by displacing endogenous myosin Va from melanosomes in the periphery. The microtubule-based melanosome traffic that drives the redistribution of these melanosomes, which are now uncoupled from the peripheral actin cytoskeleton, is evident in high-magnification time lapse images of transfected cells (data not shown), where melanosomes are seen to undergo rapid, bidirectional movements that gradually result in their accumulation in central regions of the cell, where the density of microtubules is greatest (Fig. 1).

The capture mechanism described above requires a myosin Va–dependent interaction of melanosomes with the actin cytoskeleton. We sought to identify such interactions indirectly by comparing the dynamics of melanosomes in wild-type and dilute melanocytes. To focus on these interactions, the microtubule-dependent component of melanosome transport was eliminated by treating cells with nocodazole at 4°C for 30 min followed by incubation at 37°C in the presence of nocodazole. This treatment was also necessary to get reasonable numbers of melanosomes in dilute melanocytes that were not microtubule associated. To quantify melanosome dynamics, the positions of every melanosome within ∼120 μm² areas were digitized over a period of 100 s (at 1-s intervals). Care was taken to choose cells in which there were no obvious changes in cell shape or position over this 100-s time span, and where melanosomes were well spread (a prerequisite for quantitating individual melanosome movements). Fig. 9 shows video stills of the wild-type (A) and dilute null (C) melanocytes used in this analysis, while B and D show the entire paths for all of the melanosomes present in the boxed areas. The difference between these paths, which is reasonably obvious from visual inspection (also see video sequences Nos. 12A and 12B), was borne out by quantitative analysis. Fig. 10 shows histograms in which the speeds of these melanosomes (41 for wild-type, 52 for dilute) were binned in 0.025-μm/s increments and plotted as a percentage of total steps (2,432 for wild-type, 5,174 for dilute). Three main points can be made from this plot. First, the percentage of steps in which melanosomes did not move (zero bin) is slightly more than twice as big in the dilute sample (75.3%) as in the wild-type sample (33.4%). Second, both wild-type and dilute samples have a peak (peak 1) centered around ∼0.05 μm/s. While we do not know the precise nature of these movements, they clearly are not myosin Va dependent. Third, the wild-type sample, but not the dilute sample, has another peak (peak 2) that falls roughly between the 0.075–0.1-μm/s bin and the 0.175–0.2-μm/s bin. This portion of the wild-type histogram contains 17.4% of the total steps. By contrast, only 0.17% of the total steps for the dilute sample fall in these bins, suggesting that the melanosome movements corresponding to peak 2 are myosin Va dependent. The average speed of these movements is 0.14 μm/s. Analysis of individual path plots from the wild-type sample revealed that 24 of the 41 paths (∼60%) exhibited at least three contiguous 1-s steps with a speed exceeding 0.1 μm/s. Whether these events represent myosin Va–dependent movements of melanosomes on actin filaments or the movements of actin filaments to which melanosomes are tethered by myosin Va is unclear. Nevertheless, because both situations require a myosin Va– dependent interaction of melanosomes with F-actin, these results support the capture mechanism.

We also examined the fate of melanosomes in microtubule-depleted cells over a longer time interval to estimate the ability of these movements to drive longer-range translocations of melanosomes. Fig. 9,E shows a melan-a melanocyte in which melanosomes were concentrated in the peripheral cortex (see arrows), and were undergoing rapid, bidirectional movements in the zone between the cortex and the nucleus before microtubule depletion (data not shown). Fig. 9 F shows this cell 5 min after being returned to 37°C (in the presence of nocodazole) following a 30-min incubation at 4°C in the presence of nocodazole, while G shows this same cell 90 min later. Comparison of F and G shows that the melanosomes concentrated in the peripheral cortex of this cell immediately after microtubule depletion (arrows) did not spread back into the cell, but rather remained highly concentrated in the cortex for at least 90 min. These results indicate that myosin Va– dependent melanosome movements, whatever their exact nature, do not drive obvious spreading of the organelles over 90 min.

Our current model (Fig. 11) for the role that myosin Va plays in the centrifugal transport and peripheral accumulation of melanosomes in mouse melanocytes stems from five observations: (a) the accumulation of melanosomes within dendrites and dendritic tips requires the expression of myosin Va (Koyama and Takeuchi, 1981; Provance et al., 1996; Wei et al., 1997); (b) the organization of F-actin in melanocytes is not optimal for efficient, long-range centrifugal transport of melanosomes by myosin Va (Figs. 1 and 2); (c) melanocytes possess a microtubule-dependent component of melanosome translocation that provides a very efficient means to transport melanosomes to the periphery (Figs. 3–5); (d) the characteristics of this component (e.g., bidirectionality) are such that it cannot (and does not) by itself generate a net peripheral accumulation of melanosomes (Figs. 3–5; and (e) myosin Va and melanosomes colocalize extensively in the actin-rich periphery (Nascimento et al., 1997; Wu et al., 1997). Together, these observations point to a cooperative mechanism in which a myosin Va–dependent interaction of melanosomes with F-actin in peripheral regions of the cell prevents melanosomes delivered there by centrifugal microtubule-dependent transport from being returned to the cell center by centripetal microtubule-dependent transport, thereby causing their distal accumulation. Three additional lines of evidence support this cooperative/capture mechanism. First, by expressing myosin Va tail domain fusion proteins in wild-type melanocytes, we demonstrated this cooperative/capture mechanism operating in reverse, where melanosomes, displaced from actin in the periphery by myosin Va tail domains, redistribute on microtubules to the cell center, creating a dilute-like phenotype (Figs. 6–8). While the bidirectional nature of microtubule-dependent melanosome transport might appear at first to preclude its ability to drive this redistribution, the much higher density of microtubules present in the cell body relative to cell extensions (Fig. 1) serve as a sink, causing the melanosomes to accumulate there over time. Indeed, this would be why melanosomes remain largely perinuclear in dilute melanocytes, despite the fact that most appear to be moving most of the time on microtubules.

Second, by analyzing the characteristics of microtubule-dependent movements in wild-type and dilute melanocytes, we uncovered differences that may further reflect the functioning of the cooperative/capture mechanism (Table I). Specifically, the average and maximum speeds of microtubule-dependent melanosome movements in wild-type melanocytes (both centrifugal and centripetal) are ∼60% of that in dilute melanocytes. In addition, the average path lengths (both centrifugal and centripetal) in wild-type melanocytes are ∼50% of that in dilute melanocytes. We think that these differences are probably due to myosin Va–dependent interactions of the melanosome with F-actin as the organelle is being translocated on microtubules. By acting as a drag, these interactions serve to damp, or in other cases stop, the microtubule-dependent movements in wild-type cells. Moreover, we suggest that transient interactions that slow the movement predominate in the microtubule-rich, actin-poor central cytoplasm and core of the dendrite (Fig. 11), while strong interactions that stop the movement tend to predominate in the actin-rich, microtubule-poor zone under the plasma membrane of dendrites and dendritic tips. Inherent in this mechanism, therefore, is the concept (see Langford, 1995; Kelleher and Titus, 1998) that the coordination of the microtubule- and actin-dependent components is driven to a significant extent simply by differences in the spatial distributions of these polymers.

Third, by comparing the dynamics of melanosomes in microtubule-depleted wild-type and dilute melanocytes, we identified a class of intermittent, ∼0.14-μm/s melanosome movements that appear to be myosin Va dependent (Figs. 9 and 10). At present it is impossible to tell whether these movements represent myosin Va moving melanosomes on actin, or the movement of actin to which melanosomes are tethered by myosin Va. What we can reasonably conclude, however, is that since both situations require myosin Va–dependent interactions of melanosomes with actin, this data supports the cooperative/capture mechanism.

The data in Figs. 9 and 10, together with the fact that a random orientation of actin does not preclude long-range myosin Va–dependent melanosome transport (via facilitated diffusion), raises the issue of the relative significance of myosin Va–dependent melanosome motility versus the cooperative/capture mechanism identified here in driving the long-range centrifugal transport and peripheral accumulation of melanosomes. Our observation that melanosomes accumulated in the periphery of wild-type cells do not exhibit obvious, microtubule-independent spreading over 90 min (Fig. 9) suggests that the net outward displacement of centrally located melanosomes by the action of myosin Va alone would also be minimal over 90 min. By contrast, melanosomes are seen to undergo microtubule-dependent transport from the center to the periphery in as little as ∼20 s. With regard to peripheral accumulation, while a myosin-only mechanism working in the context of randomly oriented actin can facilitate melanosome diffusion, it cannot generate concentration. By contrast, the actomyosin Va–dependent capture of melanosomes delivered to the periphery by microtubule-dependent movement provides for not only efficient, long-range transport, but also a concentrating mechanism whose kinetics should easily exceed the ability of myosin-driven diffusion to dissipate the accumulated organelles. Evidence that mouse melanocytes actually use this mechanism was provided by forcing the cooperative/capture mechanism to run in the absence of capture (Figs. 6–8), where long-range melanosome movements are seen to be microtubule based. Furthermore, the ∼2-h time interval required for this redistribution to occur, where the sink driving redistribution is simply the difference in microtubule density between the periphery and the cell center, probably underestimates the ability of the cooperative/capture mechanism to generate a peripheral accumulation when working in the correct direction, where the sink is the myosin Va–dependent capture of melanosomes. Based on these and other observations (see below), we conclude that the cooperative/capture mechanism is the predominant mechanism responsible for the long-range centrifugal transport and peripheral accumulation of melanosomes in mouse melanocytes. We say predominant because we cannot rule out some contribution to long-range transport by the action of a slow myosin Va–dependent component. Furthermore, such myosin V–only organelle movements may be very important in other cellular contexts. Having said this, we note that the ability of myosin V to drive long-range displacements of organelles in the context of randomly oriented actin is thwarted not only by the inefficiency inherent in a random orientation of tracks, but also by problems involving (a) anchorage of actin filaments (if not anchored, actin may move over the surface of the organelle and not vice versa), (b) cancellation of forces (caused by myosin V molecules on the organelle's surface interacting simultaneously with actin filaments of opposing orientation), and (c) lack of tensile strength (individual, i.e., not bundled, actin filaments have little tensile strength, so may tend to bend when myosin V–dependent forces are applied). In the case of the melanosome, these problems are probably magnified, relative to 150-nm transport vesicles, by its large size and physical rigidity.

The cooperative/capture mechanism identified here not only overturns existing ideas about the mechanism of melanosome motility in mammalian melanocytes (Hearing and King, 1993; Provance et al., 1996; Wei et al., 1997; Wu et al., 1997; Mermall et al., 1998; Rogers and Gelfand, 1998) but provides the foundation for future efforts directed at understanding how the microtubule- and myosin V–dependent components are regulated to maintain an accumulation of melanosomes at dendritic tips in the context of constitutive, and up-regulatable, transfer to keratinocytes. Success in this area could lead to the identification of agents that augment this transfer, which in turn could have a significant impact on the prevention and treatment of melanoma.

In addition to providing strong support for our cooperative/capture mechanism, the results of expressing GFP- and FLAG-tagged myosin Va tail domain fusion proteins in wild-type melanocytes also provide strong confirmation of previous efforts to demonstrate that myosin Va associates with melanosomes (Nascimento et al., 1997; Wu et al., 1997), and indicate that the portion of myosin Va mediating this interaction resides within the COOH-terminal ∼33% of the molecule. Efforts to further define this domain, which may include two melanocyte-specific exons (Seperack et al., 1995), should aid in identifying the melanosomal protein that serves as the myosin V “receptor.”

This paper represents the first clear demonstration that melanosomes in mammalian melanocytes undergo microtubule-dependent translocation. Our initial appreciation of these movements was facilitated by examination of dilute melanocytes, where essentially all melanosomes are seen to be riding on microtubules most of the time, and where melanosomes moving on microtubules in the periphery stand out against the melanosome-free background (Figs. 3 and 4). While these movements are harder to see in wild-type cells, examples of striking, bidirectional, microtubule-dependent melanosome translocations are not uncommon (see video sequence No. 10) and can cause relatively rapid changes in the degree of peripheral accumulation of melanosomes (Fig. 5). The identity of the microtubule motor(s) responsible for these bidirectional movements, and the polarity of microtubules within melanocyte dendrites, are unknown. In neurons, microtubules are uniformly plus end–out in axons, but almost equally split between plus end–out and minus end–out in dendrites (Baas and Yu, 1998). If the former situation is the case for the melanocyte dendrite, then the melanosome must associate with both plus end– and minus end–directed motors. If the microtubules are of mixed polarity, then either type of motor alone could support bidirectional melanosome traffic. Identification of the motor(s) involved should be possible using function-blocking antibodies and/ or the expression of dominant-negative constructs, and may be facilitated by the characterization of melanosome dynamics in dilute suppressor (dsu) melanocytes. Dsu is an extragenic suppressor of dilute that acts in a semidominant, cell autonomous manner to reverse the effect of dilute mutations on coat color (Moore et al., 1988, 1994). Based on the example in yeast, where overexpression of a kinesin-related protein (smy1p) partially suppresses the phenotype of a myosin V mutant (myo2 66) (Lillie and Brown, 1992), it has been speculated that dsu might encode a kinesin family member that is either turned on or overexpressed in dsu melanocytes (Moore et al., 1994). Our results here make this hypothesis all the more attractive.

Fish melanophores generate very rapid changes in surface color through highly coordinated, microtubule-dependent, intracellular dispersion (dark) and aggregation (light) of their pigment granules (Haimo and Thaler, 1994). Recently, Rodoinov et al. (1998) reported that when dispersion was induced in these cells after destruction of the actin cytoskeleton, pigment granules were seen to move all the way to the margins of the cell. This phenotype, which contrasts with the even cytoplasmic distribution of pigment granules normally found in the fully dispersed state, and which is consistent with the unopposed action of a plus end–directed microtubule motor, suggested that an interaction of these granules with the actin cytoskeleton (presumably via a myosin) serves to limit plus end–directed pigment granule traffic by shifting granules from fast, anterograde movement to local movement on short actin filaments of random orientation that were shown to exist between the microtubules. We suggest that this cooperative mechanism, and the cooperative/capture mechanism identified here in mouse melanocytes, are related phenomena tuned to achieve different biological ends. In the case of fish melanophores, the cooperation between plus end– directed granule movements and local, actomyosin-dependent granule movements serves to position granules on actin at points all along the microtubule, thereby generating the even cytoplasmic distribution required for darkening. In mouse melanocytes, the cooperation between fast, bidirectional, microtubule-dependent melanosome transport and actomyosin V–dependent capture of melanosomes at dendritic tips serves to generate the peripheral accumulation of melanosomes required for efficient transfer to keratinocytes. Importantly, the actomyosin-dependent component in both these systems appears to be primarily short-range in nature (Fig. 9; see references in Haimo and Thaler [1994] and Discussion in Rogers and Gelfand [1998]), and, as such, is only apparent when working in cooperation with the long-range, microtubule-dependent component. This may not be the case, however, for frog melanophores, where Rogers and Gelfand (1998) argue that an actomyosin (possibly myosin V)–dependent component, operating in the context of a barbed end–out bias in actin filament organization, functions along with kinesin II (Rogers et al., 1997) to drive the long-range centrifugal transport of pigment granules.

The results obtained in this study add to the growing consensus that microtubule- and actin-based motors cooperate in organelle motility and distribution (Kuznetsov et al., 1992; Lillie and Brown, 1992; Bearer et al., 1993; Fath et al., 1994; Langford, 1995; Morris and Hollenbeck, 1995; Evans et al., 1997; Kelleher and Titus, 1998; Rogers and Gelfand, 1998; Rodionov et al., 1998), agree with the view that microtubules support long-range movements, while actin filaments support short-range movements in the periphery (Bray, 1992), and extend these concepts in several ways. Specifically, the cooperative/capture mechanism identified here generates a polarized (not even) distribution of an organelle and operates in the context of ongoing, bidirectional, microtubule-dependent traffic. This probably represents a common context in which a myosin would seek, in cooperation with microtubules, to determine organelle distribution. This would be the case, for example, in neuronal dendrites, where only a single type of microtubule motor would drive bidirectional transport (Baas and Yu, 1998), or in the case where microtubules are of uniform polarity, and where the organelle possesses both plus and minus end–directed motors, but where retrograde and anterograde movements are ongoing, not reciprocally controlled by hormones as in melanophores. Importantly, because the nature of our study directly implicates myosin Va in the cooperative/capture mechanism, this mechanism may provide insight into the function(s) of myosin V in other cell types. For example, this mechanism may explain the defect in distribution of smooth endoplasmic reticulum (SER) in cerebellar Purkinje cells from dilute rats and mice (Dekker-Ohno et al., 1996; Takagashi et al., 1996), where SER is present in the trunks of the dendrite proper but never penetrates the dendritic spines. By analogy with melanosomes, SER in the dendrite proper may be undergoing bidirectional, microtubule-dependent transport, while capture of SER tubules in the dendritic spines may be actomyosin Va dependent. Myosin Va may also have a general role in influencing the distribution of secretory lysosomes, given that (a) melanosomes are secretory lysosomes (Orlow, 1995); (b) humans with Griscellis' disease, the equivalent of mouse dilute (Pastural et al., 1997), have reported defects in natural killer cell function (Klein et al., 1994); (c) the lytic granules used by natural killer cells to kill target cells are also secretory lysosomes (Griffiths, 1996); and (d) lytic granule secretion is a targeted event involving microtubules (Griffiths, 1995). Finally, our results may provide further insight into the phenotypes of yeast cells lacking either of the two class V yeast myosins (myo2p and myo4p), which are usually classified solely as deficits in the transport of various cargo from mother cell to bud (for reviews see Brown, 1997; Mermall et al., 1998; Titus, 1998). While the steady-state concentration of these myosins at the bud tip (Lillie and Brown, 1994; Govidan et al., 1995) may only represent the pool of myosin V molecules that have completed this transport task, the likelihood that the diffusion of vesicles across the yeast cell is quite rapid (see Bloom and Goldstein, 1998) raises the possibility that the myosin V–dependent capture of vesicles that diffuse to the bud tip represents a significant mechanism driving the accumulation of at least certain cargo there.

We thank Brian Jankowitz for assistance in preparing the manuscript, Dr. Ken Springs and Dr. James Sellers for help with live cell imaging and video microscopy, Dr. Leah Haimo for advice and discussion, Dr. Neal Copeland and Dr. Nancy Jenkins for mouse strains, and Dr. Edward D. Korn for comments and support of this work.

dsu

dilute suppressor

se

short ear

SER

smooth endoplasmic reticulum

EGFP

enhanced green fluorescent protein