Vesicle-associated brain myosin-V can be activated to catalyze actin-based transport

With the exception of actin-based organelle movements in plants, many membrane-trafficking events in higher eukaryotes are known to be driven by microtubule-based molecular motors (reviewed in Hirokawa, 1998). Recent observations of membrane movements on actin filaments in animal cells have prompted a search for membrane-associated myosins in these organisms (Bearer et al., 1993; D’Andrea et al., 1994; Fath et al., 1994; Langford et al., 1994; Mermall et al., 1994; Morris and Hollenbeck, 1995; Simon et al., 1995; Evans and Bridgman, 1995). Myosins are mechano-enzymes that use the energy of ATP to move along actin filaments. The myosin superfamily consists of at least 14 structurally distinct classes of actin-based motor proteins (Mermall et al., 1998). Although each myosin contains a conserved motor domain with actin and ATP binding sites, each class has a unique tail domain that may dictate the cellular localization and function of that class (Mooseker and Cheney, 1995). Considerable evidence has accumulated to support the involvement of several myosin classes, including class I, II, III, V, VI and VII myosins, in membrane trafficking (reviewed in Mermall et al., 1998; Hasson and Mooseker, 1997). However, the most compelling evidence linking unconventional myosins to intracellular transport exists for the class-V myosins.

Myosin-V is expressed in a variety of species (reviewed in Titus, 1997). Class V myosins are two-headed motor proteins that contain a unique globular tail domain. Genetic analyses of myosin-V mutations indicate that some class V myosins have essential functions. For example, yeast MYO2 (a myosin-V gene) mutants are unable to form mature buds, arrest as unusually large cells with an accumulation of small vesicles in the mother cell, and are deficient in vacuole transport (Johnston et al., 1991; Govindan et al., 1995; Hill et al., 1996). Mutations in at least one isoform of mammalian myosin-V also produce defects in vesicle trafficking. Melanocytes from mice lacking this isoform of myosin-V (dilute lethal d^l mice) have a defect in melanosome transport that results in an abnormal accumulation of pigment-containing organelles in the perinuclear region (Provance Jr. et al., 1996; Mercer et al., 1991; Silvers, 1979). Recently, defects in the distribution of endoplasmic reticulum in Purkinje cell spines were observed by electron microscopic analysis of dilute lethal brain tissue (Takagishi et al., 1996). This is particularly interesting because these mice exhibit severe neurological abnormalities that worsen from birth until death in the third postnatal week (Searle, 1952). Recent genetic studies demonstrate that mutations in myosin-Va lead to Griscelli disease, a rare human disorder characterized by pigmentary dilution, and immunodeficiency (Griscelli et al., 1978; Hurvitz et al., 1993; Pastural et al., 1997). These abnormalities in organelle distributions have led to the hypothesis that myosin-V participates in some forms of vesicle trafficking in these cell types.

The localization of myosin-V in melanocytes and neurons has provided evidence that myosin-V is organelle-associated (Espreafico et al., 1992; Evans et al., 1997; Wu et al., 1997; Nascimento et al., 1997). Recently, myosin-V-associated pigment granules were shown to move on actin, suggesting that this motor might drive these movements (Rogers and Gelfand, 1998). However, these studies have not directly demonstrated that myosin-V is involved in organelle trafficking. In the present study we have isolated and characterized an organelle fraction enriched in and tightly associated with brain myosin-V. A majority of myosin-V vesicles in this fraction are pleomorphic vesicles that colocalize with synaptic vesicle markers. Using an adaptation of an in vitro actin filament motility assay (Simon et al., 1995), we provide direct evidence that this vesicle-associated myosin-V can be activated to translocate along actin filaments. Our results suggest that vesicle components may regulate the function of organelle-associated brain myosin-V.

Electrophoresis chemicals, imidazole, ATP (grade II), EDTA, EGTA, dithiothreitol (DTT), aprotinin, benzamidine, bovine brain calmodulin (CaM), rabbit skeletal muscle calpain and trichloroacetic acid (TCA) were purchased from Sigma. Chemiluminescent western blotting kit and Pefabloc were purchased from Boehringer-Mannheim. Triton X-100 and a BCA protein assay kit were purchased from Pierce.

Myosin-V was purified from chick brain as described previously (Nascimento et al., 1996). Rhodamine-actin was purchased from Cytoskeleton, Inc. Two different affinity-purified myosin-V antibodies were used. (1) A polyclonal antibody raised against the C-terminal 109 kDa of chicken myosin-V, which includes 931 amino acids extending from within the coiled-coil domain to the C terminus in the globular tail domain (Espreafico et al., 1992), for western blots and most immunoelectron microscopy. (2) Another polyclonal antibody, raised against the head (motor) domain of chicken myosin-V (amino acids 5-752), was a generous gift of Dr Roy Larson (University of Sao Paulo, Ribeirao Preto, Sao Paulo, Brazil 14049; Espreafico et al., 1992) and was used in motility assays. This antibody, which inhibits the motor activity of myosin-V (Wolenski et al., 1995), exclusively reacts with a region of relatively low homology to other known myosins (amino acids 1-136) (F. Espindola and M. S. Mooseker, unpublished observations). Furthermore, this antibody does not recognize other myosins present in brain, based on immunoblot analysis of brain from myosin-V-null mice (results not shown). Synaptic vesicle antibodies directed against synaptophysin (polyclonal) and SV2 (monoclonal acites fluid) were kindly provided by Dr Pietro DeCamilli (Yale University, New Haven, CT 06511, USA). Gold-conjugated secondary antibodies were purchased from Jackson Immunochemicals, Inc.

Myosin-V-enriched membranes were prepared using the initial fractionation steps of an established protocol for myosin-V purification (Nascimento et al., 1996; Fig. 1A). Chick or rat brains were homogenized at 4°C in 40 mM Hepes, pH 7.8, 10 mM K-EDTA, 5 mM ATP, 2 mM DTT and 2 mM Pefabloc in an Omnimixer blender. Homogenates were centrifuged at 40,000 g (at 4°C) for 40 minutes. The salt concentration of the supernatant (S1) was raised to 0.6 M NaCl, and the solution was incubated on ice for 1 hour. A microsomal pellet was produced by centrifugation of S1 at 40,000 g (at 4°C) for 40 minutes. The pellet (P2) was resuspended in vesicle buffer (20 mM imidazole, pH 7.2, 75 mM KCl, 5 mM MgCl2, 1 mM EGTA, 4 mM ATP, 2 mM DTT and 1 mM Pefabloc). In fractionation studies, the supernatant S2 was further centrifuged at 150,000 g for 1 hour to produce an ultraspeed pellet (P3) and soluble supernatant fraction (S3). Although the salt precipitation step is essential for enrichment of myosin-V in the P2 pellet, it is not required for sedimentation or vesicle association: myosin-V associated vesicles can be sedimented from S1 at high speed (150,000 g for 2 hours) but contamination with small vesicles lacking myosin-V is much greater (results not shown).

The amount of total protein in each fraction was determined from the BCA protein assay (Pierce) and volume measurements. The amount of myosin-V in each fraction was determined by quantitative immunoblotting using purified myosin-V as standards for densitometry. Enrichments were determined by comparing the amount of myosin-V in each subcellular fraction with that of the homogenate.

Formvar-coated, carbon-stabilized grids were incubated in a dilute vesicle solution and then immunostained (Lewis and Bridgman, 1996). Myosin-V tail antibodies were used at 10 µg/ml. Ascites fluid against SV2 was diluted 1:10. Control grids were incubated in 10 µg/ml nonimmune rabbit IgG. After labeling with appropriate 6 or 10 nm gold-conjugated secondary antibodies, grids were stained with 1% uranyl acetate in water, and wicked dry. In some instances grids were fixed in 0.25% glutaraldehyde before immunolabeling; however, no significant differences were observed between fixed and unfixed specimens. Electron micrographs were obtained using a Zeiss 10C electron microscope. Samples were imaged at 80 kV.

To quantify the distribution of immunogold label, gold particles were counted in random-field micrographs (for each experiment n=20), taking note of whether each particle was substrate- or organelle-associated and whether it appeared as a single particle or a cluster. Gold particles were considered organelle-associated if within 20 nm of a negatively stained vesicle profile, and were otherwise considered substrate-associated. The grid substrate is defined as the colloidon/carbon film support. A substrate cluster is one or more gold particles that appear within 20 nm of each other on the substrate. This allows comparison of the average number of gold particles per organelle with the average number of gold particles in a similarly sized region on the substrate.

Brains were homogenized and fractionated as reported above. Enzyme assays were performed on the homogenate, P1, P2 and S2 fractions using common protocols. Enzymes assayed were: ouabain-sensitive Na+/K+-ATPase activity (plasma membrane; Blanco et al., 1995), glucose-6-phosphatase (endoplasmic reticulum; Aronson and Touster, 1974), α-mannosidase II activity (Golgi; Storrie and Madden, 1990) and succinate dehydrogenase (mitochondria, assayed as succinate INT-reductase activity; Pennington, 1961).

Samples of myosin-V-associated vesicles (100 µg total protein) were incubated under various conditions known to solubilize membrane proteins (see Fig. 4 legend) for 10 minutes, and then sedimented by ultracentrifugation (150,000 g). The amounts of myosin-V in pellet and supernatant fractions were assayed by SDS-PAGE and western blotting.

P2 vesicles were washed twice in vesicle buffer without EGTA (20 mM imidazole, 75 mM KCl, 2.5 mM MgCl₂, 4 mM ATP, 2 mM DTT and 1 mM Pefablock), and were resuspended in this buffer to a final concentration of approximately 0.5-1 mg/ml total vesicle protein. CaCl₂ was added to give a final concentration of 2 mM Ca²⁺, and calpain was added at a 1:10 ratio of calpain to myosin-V (estimated from quantitative immunoblots). This mixture was incubated at 4°C overnight, then for 4 hours at room temperature, and then for 10 minutes at 37°C. Before inhibiting the reaction with 2 mM EGTA, 50 µg/ml calmodulin and an additional 2 mM ATP were added. Excess

ATP was added to decrease the likelihood of the myosin-V motor domain binding to vesicle-associated actin, and excess calmodulin was added in an attempt to saturate light chains in the myosin-V regulatory domain, as this region of the molecule has potential for nonspecific binding when calmodulin is depleted. After digestion, vesicles were ultracentrifuged (100,000 g) to separate soluble myosin-V fragments from vesicle bound fragments. These rigorous digestion conditions were required to achieve complete digestion of myosin-V.

Nitrocellulose-coated flow cells holding 30-50 µl were prepared using standard methods (see Sellers et al., 1993). Vesicle-coated substrates were used as in Simon et al. (1995). Here, ice-cold 10% Triton X-100 was added to a suspension of microsomes (approx. 2 mg/ml vesicle protein) in vesicle buffer, to give a final concentration of 0.1% Triton X-100. This solution was incubated on ice for 2 minutes, and then diluted in 10 volumes of vesicle buffer and centrifuged at 40,000 g for 30 minutes at 4°C. The pellet was washed twice and resuspended in approx. 2 volumes of vesicle buffer for storage as a concentrated vesicle solution. Vesicles were stored on ice and used within 3 days. A sample of the vesicle stock was diluted 5- to 10-fold, and perfused into a motility chamber until a thin, uniform, particulate surface was observed by differential-interference-contrast (DIC) microscopy. Motility buffer (vesicle buffer containing 4 mM ATP, 0.6% methyl cellulose, 100 µg/ml calmodulin, 2.5 mg/ml glucose, 0.1 mg/ml glucose oxidase, 0.02 mg/ml catalase, 10 nM rhodamine-actin, and 40 µg/ml nonimmune rabbit IgG) was prepared just before use. An Olympus IMT-2 microscope (Olympus Optical Co., Tokyo, Japan) equipped with a 100×1.4 NA objective was used to view rhodamine-actin on the coverslip surface. A random field was chosen, and 2×2 binned images were captured every 3 seconds using a cooled, charge-coupled-device (CCD) camera with a 512×512 back-illuminated, thinned chip (Photometrics CH250). The camera recordings were activated and controlled by a network (macro) created in ISee Unix-based software (Inovision Corporation, Durham, NC). Each filament in a recording was tracked manually by zooming the playback image 4×, stopping the recording at each frame, and using the cursor to determine the coordinates of a filament centroid in each frame. Filament centroids were calculated by averaging the x-y coordinates of the pixels forming the filament edges. The centroid coordinates were adjusted for curved filaments to be the estimated midpoint between filament ends. The Adobe Photoshop (Adobe Systems, Inc.) layers tool was used to superimpose recording frames and create summary figures of filament centroid movements during a recording.

Triton-treated P2 vesicles were diluted, applied to grids exactly as they were for motility assays, and immunolabeled as above, using the function-blocking myosin-V antibody at 15 µg/ml. Each myosin-V immunogold particle/cluster in a random-field micrograph was counted and labeled as substrate- or organelle-associated. Three concentric circles, with radii equivalent to 0.5, 1.0 and 2.0 µm, were drawn around each gold cluster. It was noted when another gold particle fell within a radius of 0.5, 1.0 and 2.0 µm, and whether it was substrate- or organelle-associated. The total number of organelle-associated gold clusters was tallied and the proportion with neighboring substrate and organelle-associated clusters within the above radii was calculated. The same was done for each substrate-associated cluster.

Brain myosin-V can be purified using a protocol developed by Cheney et al. (1993). In this protocol, myosin-V is purified from a high-speed supernatant fraction of whole brain homogenate that contains approximately one half of the total brain myosin-V (see Materials and methods; Fig. 1). To purify myosin-V, this supernatant is incubated in high salt (0.6 M NaCl) for 1 hour, and then centrifuged at high speed to produce a pellet (P2) that is enriched in myosin-V (approx. fivefold enrichment over homogenate levels, see Fig. 1). This myosin-V-enriched pellet is the fraction from which brain myosin-V is routinely purified.

Myosin-V becomes greatly enriched upon treatment of this P2 pellet with detergent (Cheney et al., 1993). Because the pellet loses its milky appearance upon detergent treatment, we suspected that it contained membranes. Thus, we chose to investigate this fraction further as a possible source for myosin-V-associated organelles. Electron microscopy demonstrated that pellet P2 consists of small membrane vesicles, ranging in size from 30 to 250 nm in diameter (Fig. 2). SDS-PAGE and immunoblotting demonstrate that actin is present in the P2 pellet (Fig. 1C; actin immunoblot not shown). However, very few actin filaments were visible in this preparation by negative-stain electron microscopy, and there was no evidence of vesicle aggregation or crosslinking by acto-myosin complexes.

To determine whether myosin-V was associated with vesicles in this membrane pellet, immunoelectron microscopy for myosin-V was performed and quantified (Fig. 2; Table 1). Myosin-V immunogold appeared to be almost entirely associated with vesicles in this preparation. In fact, 94% of the total gold in these labeled preparations was organelle-associated, although only 30-40% of the total organelles in this preparation were labeled with immunogold. On average, approximately 2-3 gold particles were present per labeled organelle.

To determine the nature of the vesicles in this membrane fraction we assayed for enrichments of common organelle marker proteins. Our initial results, obtained from enzyme assays for typical membrane markers, indicated that this fraction contains Golgi, endoplasmic reticulum and some plasma membrane (Table 2). Prekeris et al. (1997) recently reported that myosin-V associates with synaptic vesicles. To determine whether the myosin-V-enriched microsomal fraction that we had isolated contained synaptic vesicle marker proteins, we used quantitative immunoblotting and immunoelectron microscopy. We found that synaptic vesicle markers are enriched in this fraction (Fig. 3A). In fact, the enrichments of myosin-V and synaptic vesicle proteins in P2 are comparable to the enrichments of these proteins in the crude synaptic vesicle preparation used by Prekeris et al. (1997) (not shown). Immunoelectron microscopy demonstrated that about 70% of the vesicles in this preparation labeled with the synaptic vesicle protein, SV2. Double-labeling for myosin-V and SV2 revealed that a majority (about 80%) of myosin-V label was present on vesicles that also labeled with SV2 (Fig. 3B). However, only 40% of the total SV2-positive vesicles also labeled with antibodies against myosin-V. Many small 30-60 nm vesicles that labeled for SV2 were not labeled with myosin-V-immunogold (average size: 60±30 nm). Although some of the myosin-V/SV2 double-labeled vesicles resembled small synaptic vesicles in shape and size, a majority of these double-labeled vesicles were larger, pleomorphic vesicles (30-250 nm, average size 90±45 nm).

To further examine the association of myosin-V with these P2 vesicles, the conditions that solubilize myosin-V from P2 were examined (Fig. 4). Vesicles were incubated under various conditions, and were then sedimented to separate solubilized proteins from the vesicles. Samples were examined for release of myosin-V into the supernatant by immunoblotting (Fig. 4). Mild agents, such as 0.6 M salt and 5 mM MgATP, were unable to solubilize the myosin-V associated with these vesicles. The small percentage (<10%) of myosin-V solubilized in high salt and ATP may result from the release of actin-associated myosin-V that is not associated with vesicles in this preparation. Strong agents, such as thiocyanate and pH 11.4 carbonate buffer, extracted approximately 85% of the myosin-V from the vesicle surface. Vesicles remained intact under these conditions, as was verified by electron microscopy (not shown). Although incubation in 1% Triton X-100 disrupts the membrane associated with these vesicles, this treatment does not solubilize myosin-V, indicating that myosin-V may be part of a Triton-insoluble complex on the vesicle surface. Solubilization of myosin-V in 1% Triton is increased in buffers with high salt or high pH, further indicating that myosin-V may attach to the vesicle surface by binding tightly to a complex of proteins.

To further investigate the association of myosin-V with P2 vesicles, myosin-V-enriched vesicles were digested with the calcium-dependent protease, calpain. Calpain hydrolyzes myosin-V to produce an 80 kDa globular tail domain that does not exhibit actin binding activity, and a 65 kDa head fragment containing actin binding activity (Nascimento et al., 1996). Digestion of P2 vesicles was performed in the presence of excess ATP to decrease the likelihood of the myosin-V motor domain binding to vesicle-associated actin. Complete digestion of myosin-V was achieved, indicating that myosin-V is exposed on the cytosolic face of these vesicles. Upon calpain cleavage of vesicle-associated myosin-V, at least 60% of both the 80 kDa tail fragment and the 65 kDa head domain remain vesicle-bound after calpain cleavage (Fig. 5).

We next wanted to determine if vesicle-associated myosin-V was capable of actin-based motility. To do so, an adaptation of a commonly used motility assay for myosins was used (Simon et al., 1995). When coverslips were coated with myosin-V-enriched brain microsomes, some rhodamine-labeled actin filaments adhered to the coverslip surface, but little or no actin filament movement was observed in the presence of ATP (2-5 mM) and 100 µg/ml calmodulin (not shown). Serendipitously, we found that exposure of vesicles to ice-cold 0.1% Triton X-100 greatly enhanced actin-filament motility. Triton-treated vesicles were prepared by briefly exposing vesicles to ice-cold 0.1% Triton, sedimenting vesicles by centrifugation, and washing the pellet twice in vesicle buffer without Triton. On substrates with a fine, uniform coating of Triton-treated vesicles (Fig. 6A), some motility (10-30% of the total filaments) was observed in almost every random field. Although this percentage of moving filaments is low compared to motility assays on substrates of pure myosin, it is typical of motility assays on vesicle substrates (Simon et al., 1995). In regions where motility was observed, actin filament movements resembled motility of actin on substrates of pure myosin-V (Fig. 6). Filaments moved across the substrate along their long axis at an average rate of 344 nm/second. Filaments were sometimes observed to spin in place, perhaps indicating regions of lower motor density. The rate distribution of filament motility on Triton-treated microsomal substrates is very similar to that of actin filament motility on pure myosin-V substrates, which has an average rate of 325-375 nm/second (Cheney et al., 1993; Wolenski et al., 1995).

To determine if actin filament motility on Triton-treated microsomal substrates was due to myosin-V, motility experiments were performed in the presence of a function-blocking antibody against myosin-V. This antibody, which recognizes the motor domain of myosin-V and is highly specific for brain myosin-V (see Materials and methods), inhibits motility of myosin-V on actin filaments in vitro (Wolenski et al., 1995). In the presence of nonimmune rabbit IgG, actin filament motility on Triton-treated microsomes is uninhibited; however, no motility is observed after incubation in motility buffer containing the function-blocking myosin-V antibody, even when fresh actin is included in the motility buffer (Fig. 7). Upon addition of the myosin-V antibody, motility is significantly decreased almost immediately. After 1 minute, no motility is observed. This effect was quantified in recordings from three separate preparations of myosin-V enriched brain microsomes, and was found to be highly significant (P<0.00005) (Table 3).

Triton-treated microsomes clearly show myosin-V-mediated actin filament motility. In order to understand how this actin-based motility is activated by the 0.1% Triton treatment, it is essential to understand how this treatment affects both vesicle integrity and the protein molecules associated with these vesicles. To do so, 0.1% Triton-treated microsomes were analyzed both biochemically and visually.

The effect of Triton treatment on vesicle proteins was determined using protein assays and SDS-PAGE. This analysis revealed that approximately 80% of the total protein associated with P2 microsomes is solubilized during Triton treatment (Fig. 8A). This effect is in agreement with previous studies of the effects of detergents on biological membranes, which showed that proteins, rather than lipids, are preferentially solubilized from membrane vesicles at low concentrations of Triton X-100 (Helenius and Simons, 1975). While over 80% of the total vesicle protein is lost during Triton treatment, almost all (about 86%) of the myosin-V is retained in the P2 pellet after treatment. Furthermore, this treatment enriches for myosin-V approximately sevenfold over levels in untreated P2 vesicles (Fig. 8B). Studies not shown here also indicate that the extraction properties of myosin-V are identical before and after 0.1% Triton treatment.

As previously observed in motility assays, Triton-treated microsomal substrates appear particulate by DIC microscopy (Fig. 6A). We have also found that myosin-V-enriched microsomes labeled with the lipophilic dye DiOC₆ retain DiO labeling after Triton treatment (not shown). The integrity of Triton-treated microsomes can be more accurately accessed by electron microscopy (Fig. 9). Negative-stain electron microscopy of Triton-treated microsomes demonstrates that membrane-vesicle profiles are still present after Triton treatment.

Because Triton-treated microsomes were used for motility assays, it was important to determine whether myosin-V remains vesicle-associated after dilute Triton treatment.

Immunoelectron microscopy demonstrated that myosin-V is in fact associated with these vesicles (Fig. 9). Quantification of this labeling (Table 4) showed that over 82% of the total myosin-V-immunogold label is present on vesicle surfaces.

Approximately 12% of vesicles label with myosin-V antibodies. Although this data was obtained with the function-blocking myosin-V head antibody, similar results were obtained with the anti-myosin-V tail antibody. Because of concern that the tail antibody displaces myosin-V from detergent-treated vesicles, only data from myosin-V head antibody labeling is presented here. Quantification of vesicle sizes indicates that a subpopulation of these vesicles fuse during Triton treatment and subsequent centrifugation. Small vesicles (<60 nm) constitute approx. 52% of this preparation. The average size and number of these vesicles changes little after Triton treatment. Larger vesicles, which are often myosin-V-labeled, tend to increase in size after treatment. Before treatment, 40% of myosin-V-labeled vesicles are between 60-100 nm. After treatment this number decreases to 18%, while the number of vesicles larger than 100 nm jumps from 28% to 68%. These fusion events may underlie the decrease in the percentage of organelles labeled after Triton treatment.

To determine the likelihood of an actin filament encountering a substrate-associated myosin-V molecule in motility assays, electron micrographs of immunolabeled Triton-treated vesicles were examined using a nearest neighbor analysis (Table 5). For filament movements that are greater in

length than the filament itself, the action of multiple motors is required, even for a processive motor. Nearest neighbor analysis indicates that substrate (non-vesicle)-associated myosin-V molecules are too sparsely distributed to support such filament movements. Thus, motility on Triton-treated microsomal substrates is most likely due to organelle-associated myosin-V.

To understand how vesicle-associated myosin-V might function in vivo, it is essential to understand the nature of the organelles with which myosin-V associates. We isolated brain microsomes that contain half of the total brain myosin-V, using methodologies previously designed to purify myosin-V from chick brain (see Materials and methods; Cheney et al., 1993). Immunoelectron microscopy of this myosin-V-enriched fraction clearly demonstrated that myosin-V is associated with vesicles. Very few actin filaments were visible in this preparation by electron microscopy, and there was no evidence of vesicle aggregation or crosslinking by acto-myosin complexes. The actin associated with these vesicles may be in the form of short oligomers, such as the actin oligomers associated with the red cell membrane (Bennett and Gilligan, 1993). Our extraction studies demonstrate that myosin-V is tightly associated with vesicles, even in the presence of 5 mM MgATP and calcium chelators, where myosin-V has a low affinity for actin. Therefore, although some of the myosin-V in this preparation may be associated with actin, the majority appears to be present because of its association with vesicles. Because we wanted to investigate the most significant and enriched subcellular fraction associated with myosin-V, we chose an isolation strategy that enriched for myosin-V rather than a specific organelle fraction. Nonetheless, we have found that most (approx. 80%) of the myosin-V in the membrane fraction that we isolated is associated with vesicles that contain the synaptic vesicle protein SV2. Consistent with our results, Prekeris et al. (1997) have recently reported that myosin-V can be immunoprecipitated using antibodies against synaptic vesicle marker proteins from a crude membrane fraction enriched in synaptic vesicles.

Our quantitative immunoelectron microscopy has provided additional information about the colocalization of myosin-V and synaptic vesicle markers. While a majority of myosin-V in this preparation colocalizes with the synaptic vesicle protein SV2, less than half of the SV2-positive vesicles appear to associate with myosin-V. Most small synaptic-like vesicles are not associated with myosin-V. In preparations of crude synaptic vesicles, myosin-V/SV2 immunoelectron microscopy produces similar results (not shown): the morphology of most myosin-V/SV2 double-labeled vesicles is not consistent with that of small synaptic vesicles. Myosin-V/SV2 vesicles may be dense core synaptic vesicles, which range from 80-200 nm in diameter and are larger than the more uniform, small synaptic vesicles (approx. 50 nm) containing the classical neurotransmitters such as acetylcholine (Calakos and Scheller, 1996). Alternatively, these pleomorphic vesicles may be fragments of larger membrane compartments involved in synaptic vesicle maturation or recycling. Since compartments such as recycling endosomes may contain synaptic vesicle proteins such as SV2 (Wittich et al., 1994), this is also a likely possibility. Although the exact nature of this compartment remains unclear, our results, when considered together with those of Prekeris et al. (1997), indicate that the colocalization of myosin-V with synaptic vesicle proteins is significant. Furthermore, these results suggest that myosin-V may be important in some stages of synaptic vesicle processing.

Although myosin-V clearly enriches with synaptic vesicle proteins, it is important to note that it may not be exclusively associated with these vesicles. The myosin-V-enriched vesicle fraction used in these studies represents half of the total brain myosin-V. In experiments not presented here we find that the remaining half of the total brain myosin-V may be present on other membrane compartments. This possibility is supported by localization studies which indicate that myosin-V may be on more than one membrane compartment in both melanosomes and neurons (Espreafico et al., 1992; Evans et al., 1997; Nascimento et al., 1997). This is also consistent with recent electron microscopic studies of dilute lethal (myosin-V-null) murine cerebellum, in which defects in the distribution of endoplasmic reticulum in Purkinje cell spines were observed (Takagishi et al., 1996). Given the copurification of myosin-V with synaptic proteins, it is unclear why electron microscopic analysis of myosin-V-null (dilute-lethal) brains has yet to reveal any abnormalities in synaptic vesicle numbers or distribution. If myosin-V associates with only a subpopulation of synaptic vesicles, then defects may be more difficult to detect in thin section electron micrographs of whole brain. More subtle defects might exist if myosin-V associates with recycling endosomes. Since more than one pathway for synaptic vesicle recycling may exist (De Camilli and Takei, 1996; Koenig and Ikeda, 1996), defects in one pathway for synaptic vesicle recycling may be somewhat compensated for by other recycling pathways. Clearly, additional localization, physiological and functional studies must be conducted to address these questions.

One of the most important uses for the myosin-V-enriched membrane preparation described here will be to determine the molecular bases for its association with membranes. Detergent does not solubilize vesicle-associated myosin-V, and only strong chaotropic agents and high pH extract myosin-V from these vesicles. Thus, myosin-V may be tightly associated with a protein or protein complex on the vesicle surface. It is unclear which domain of myosin-V confers binding. Our calpain cleavage results suggest that both the head and tail domains of myosin-V can bind vesicles, even in the presence of excess ATP and calmodulin. We obtained identical results upon calpain digestion of a crude synaptic vesicle fraction. Prekeris et al. (1997) recently reported that the tail of myosin-V binds crude synaptic vesicles; however, they were unable to assess myosin-V head-domain binding, since a probe for this region of the molecule was unavailable to them. Therefore, although it is likely possibility that the tail domain of the myosin-V confers vesicle binding, this cannot be concluded with certainty because the myosin-V head domain also binds to vesicles.

Given the results from our motility experiments, it is interesting that the motor domain of myosin-V associates with vesicles. Our results show that vesicle-associated myosin-V can catalyze actin-based movement. However, we find it must be activated to do so. There are several possible reasons why dilute detergent treatment might activate motility. It is possible that Triton directly activates myosin-V. Alternatively, ATP-insensitive actin-binding proteins (including inactive myosin-V) might produce a load on actin-filaments, preventing their movement by myosin-V. This is unlikely, since motility is not induced by addition of excess unlabeled actin to saturate actin binding proteins. The most interesting possibility, of course, is that this detergent treatment, which extracts over 80% of the total vesicle proteins, activates myosin-V motility by releasing vesicle proteins and/or phospholipids that interact with myosin-V to regulate its function. Such putative regulatory factors may modulate myosin-V-catalyzed vesicle traffic, or may inhibit myosin-V activity until delivery of this myosin to its final target site.

The idea that vesicle factors may regulate the activity of myosin-V fits nicely with emerging models for organelle trafficking. Recent evidence suggests that individual organelles might be capable of movement along both microtubules and actin filaments (Bearer et al., 1993; D’Andrea et al., 1994; Fath et al., 1994; Langford et al., 1994; Morris and Hollenbeck, 1995; Evans and Bridgman, 1995; reviewed in Baker and Titus, 1998). This would require that these organelles posses both actin- and microtubule-based motors, and that these motors be regulated to allow for efficient transport. The vesicle preparation used in these studies will not only provide a means for assessing the colocalization of myosin-V- and microtubule-based motors, but should also provide an excellent preparation with which to investigate regulators of organelle-associated myosin-V.

We thank Dr Foued Espindola for help in purifying myosin-V, Dr Roy Larson for his generous gift of antibody, and Grady Phillips for his technical assistance. We also thank Drs John Cooper, Maurine Linder, Pietro DeCamilli, and Tama Hasson for their helpful comments on this manuscript. This work was supported by grants from the NIH to P. C. B. (NS35162) and M. S. M. (DK25387). L. L. E. is supported by NIH training grant T32DK07017.