Several unconventional myosins, including myosin I, V, and VI, have been implicated in membrane traffic in the actin cortex (Eichler et al., 2006; Buss and Kendrick-Jones, 2008; Kim and Flavell, 2008), which is generally found to be polarized, with the actin filament barbed ends being attached to the plasma membrane. Their role in these processes has been inferred from loss-of-function experiments in vivo (Buss et al., 2001; Varadi et al., 2005; Krendel et al., 2007). Myosin VI has been implicated in moving vesicles from the cell surface to the interior of cells, through the cortical actin meshwork (Buss et al., 2001), which is in keeping with its ability to move toward the pointed ends of actin filaments (Wells et al., 1999).

Cortical F-actin is generally organized as a thin (< 0.5 µm) dense network below the plasma membrane (Mooseker, 1985). There are many examples of using an exposed cellular cytoskeleton for in vitro experiments (Sawada and Sheetz, 2002; Sivaramakrishnan et al., 2008a; Brawley and Rock, 2009). The choice of cell type that is used to obtain a native actin cytoskeleton is critical, depending on the type of information on myosin function one is interested in obtaining. Microdissection of Nitella axillaris to expose the polarized actin cytoskeleton was used to study movement of 1-µm-sized microspheres coated with muscle myosin II (Sheetz and Spudich, 1983). The uniform polarity of the long actin bundles in N. axillaris enabled the establishment of the first quantitative in vitro motility assay with muscle myosin speeds comparable to those seen in muscle (Sheetz and Spudich, 1983). Detergent extraction is another way to reveal and maintain an intact cytoskeleton (Svitkina et al., 1995). Brawley and Rock (2009) recently examined the movement of single dimeric forms of myosin V, VI, and X on detergent-extracted Drosophila S2 and mammalian COS-7 and U2OS cells. They found that myosin V selectively moves toward the cell periphery, that myosin VI moves toward the cell center, and that myosin X is recruited to filopodia. These cell types have an actin cortical region that extends in toward the cell interior by <1 µm. The goal of our study was to define quantitatively the behavior, on a cellular F-actin interlaced network, of nanospheres coated with multiple copies of myosin VI in both monomeric and dimeric states, and to compare and contrast them. Thus, we needed a cell type with an extensive lamellipodial actin cortex. We chose the keratocyte lamellipodium, which has been characterized using light (Schaub et al., 2007) and electron microscopy (Svitkina and Borisy, 1998). It spans ∼30 × 10 µm, providing a large surface for observing interaction of unconventional myosins. The actin polarity is thought to be uniform (Small et al., 1995), with the pointed ends of the actin filaments pointing inward toward the nucleus. This widely held hypothesis is supported by the findings of our study.

Myosin VI is necessary for timely transport of endocytic cargo such as transferrin to early endosomes (Hasson, 2003). Because actin filaments in the terminal web are polarized with their pointed ends away from the cell edge, myosin VI–based movement of early endosomes along actin filaments has been inferred to facilitate endocytosis (Mooseker and Tilney, 1975). However, individual myosin VI molecules isolated from cells are monomeric and therefore are nonprocessive (Lister et al., 2004). Indeed, several unconventional myosins, including myosin I, VI, and VII, have been shown to be nonprocessive when isolated from cells, leading to questions about whether or how they can transport vesicles/organelles in vivo (Loubéry and Coudrier, 2008). Myosin VI, although monomeric as a pure protein, binds to endosomes through adaptor proteins such as GIPC and Dab2, which can dimerize on endocytic vesicles (Lou et al., 2001; Xu et al., 1998); this suggests that myosin VI may function as a dimer in vivo. This suggestion is supported by anisotropy Förster resonance energy transfer measurements in vivo that suggest that myosin VI molecules are precisely positioned in close proximity on endosomes (Altman et al., 2007). However, whether dimerization is required for processive movement of cargo is unclear.

Complicating the issue, myosin VI and V spend most of their time bound to F-actin (high duty ratio), whereas myosin I transitions from a low to high duty ratio motor under load (De La Cruz et al., 2001; Moore et al., 2001; Laakso et al., 2008). Multiple myosin heads with high duty ratio could interfere with each other, preventing the movement of cargo bound to multiple motors along the cortical actin mesh.

In this study, we examine the coordinated interaction of multiple copies of myosin VI molecules with the isolated keratocyte actin network. We address an outstanding question in the unconventional myosin field, namely the ability of multiple myosin molecules to interact, regardless of their oligomerization state, in order to bring about the transport of cargo.

Fig. 1 a shows a fluorescence image of the keratocyte actin cytoskeleton after extraction, displaying the characteristic densely packed interlaced actin network (Schaub et al., 2007). High-resolution images obtained using the platinum replica technique (Fig. 1 b) confirm that the dense interlaced actin network after detergent extraction is maintained in the absence of fixatives. Although endocytosis may or may not occur in the keratocyte lamellipodium, this dense actin meshwork is likely to be similar to the smaller actin cortex found in other cells. To model the movement of endocytic vesicles on such a network, we use myosin VI coupled to 200-nm-diameter fluorescent nanospheres.

Nanospheres coated with a single myosin VI artificial dimer land on the keratocyte network, move a short distance (<1 µm), then detach and diffuse away (Fig. 1 c). Myosin VI artificial dimers were created by the insertion of a GCN4 leucine zipper coiled-coil motif in its tail domain (De La Cruz et al., 2001). Insertion of GCN4 ensures dimerization, as has been observed for artificial dimers in single molecule experiments (Rock et al., 2001; Okten et al., 2004). Fig. 1 d shows trajectories of these nanospheres, starting with their landing on (top) and ending with their detachment from (bottom) the keratocyte actin network. The short run lengths (<1 µm) are consistent with previous measurements (Sweeney et al., 2007; Brawley and Rock, 2009). Myosin VI heads release from the actin filament when bound to ATP. Occasionally, when both heads bind an ATP, the dimer releases from the actin and the nanosphere diffuses into solution, terminating the run.

When multiple myosin VI dimers are coupled together through a nanosphere, the nanosphere will release from the actin network only when all the heads are bound to ATP. The frequency of this depends on the rate at which both heads of a dimer are simultaneously bound to ATP (<1 s⁻¹) relative to the rate at which the myosin transitions back to a state when it strongly binds to F-actin (∼40 s⁻¹; Sweeney et al., 2007). Thus, coupling multiple myosin VI motors is likely to keep a myosin VI–coated nanosphere in contact with the actin network longer before it diffuses away. However, the presence of multiple dimers may also interfere with movement, a result of one dimer trying to step while others are bound with both heads to F-actin. Interference between dimers could prevent the nanosphere from moving along the network.

Therefore, we studied the movement of myosin VI–coated nanospheres with multiple dimers to resolve whether they can coordinate movement along the actin network. We found that nanospheres coated with ∼10 artificial dimers, such that ∼2 dimers are able to interact with the keratocyte actin mesh (Fig. S1), move across the entire keratocyte lamellipodium (Fig. 1, e and f). They move in an almost linear pattern, starting from the cell edge and proceeding toward the cell center (Fig. 1 g). All movement, without exception, is directed away from the cell edge toward the cell center (Fig. 1 g and Video 1, which is a sample video of Fig. 1 f).

Platinum replica electron micrographs of the keratocyte lamellipodium acquired after movement of the nanospheres with ∼10 dimers do not indicate reorganization of or damage to the actin cytoskeleton (Fig. 2 a). This suggests that the movement is brought about by the cooperative interaction between multiple dimers and the actin network and not by reorganization or reorientation of the lamellipodium, as could be possible with multiple myosin heads. This finding was confirmed by cross-linking the actin cytoskeleton using a fixative (2% formaldehyde for 10 min at 22°C) before its interaction with the dimer-coated nanospheres (Fig. 2, b and c). Note that on occasion, a nanosphere coated with myosin VI lands on a retraction fiber at the rear end of the cell (green trace at the bottom of Fig. 2 b) and moves toward the nucleus.

To understand how multiple motors coordinate to bring about linear directed motion of nanospheres, we used our detailed knowledge of myosin VI artificial dimer function, obtained from single molecule and kinetic studies, to simulate their movement on the keratocyte actin mesh. To capture the interlaced nature of the keratocyte lamellipodium, we digitized platinum replica electron micrographs of extracted keratocytes to obtain the actin tracks encountered by the myosin motors (Fig. 3 a). We created a simplified model for movement of nanospheres brought about by two myosin VI dimers (Fig. 3 b) using the following steps: (1) each myosin VI dimer steps stochastically at a rate limited by the release of ADP from its rear head (5 s⁻¹; De La Cruz et al., 2001); (2) each myosin VI dimer takes 36 ± 4-nm steps (Rock et al., 2001); (3) the next binding site of the lead head of the stepping dimer is constrained such that the distance between the points where myosin VI attaches to the nanospheres (yellow circles in Fig. 3 b) is conserved; (4) the nanosphere constrains myosin movement such that more than one step by a dimer, before the movement of its counterpart, is not productive; and (5) the rear head of a dimer always releases before the lead head.

Our simplified simulation shows that coordinated movement of two myosin dimers on the keratocyte mesh is indeed possible and results in linear directed motion toward the cell center (Fig. 3 c). This model also reproduces the quality of the tracks seen by myosin VI dimer-coated nanospheres seen in Fig. 1 (e–g), and the end–end speed of the nanospheres (ratio of distance between the ends of the trajectory to the time for covering that distance; Fig. 3 d). The end–end speed includes the linearity of the movement path, kinetics of the individual myosins, their step sizes, and their coordination, therefore suggesting that our simplified model captures the essential features of the myosin VI–nanosphere–network interaction.

In addition to the end–end speed, our model accurately captures the frame–frame speed (the ratio of the distance moved between two successive image acquisition frames to the time between frames acquired during imaging on an epifluorescent microscope) distribution observed in the simulation (Fig. 3 e). Occasional release of the lead head before release of the trailing head of a dimer results in nonproductive steps that slow down the nanosphere. However, occasional multiple steps by a single dimer increase the mean speed of the nanosphere. Effects such as these, which are not modeled in our simulation, likely explain the differences in the width of the end–end speed distributions (Fig. 3 d). Despite the simplifying assumptions, our model verifies that the underlying coordinated movement of the two myosin dimer molecules enables the nanospheres to traverse the entire actin network.

A similar simulation was used to predict the movement of a nanosphere with four bound myosin VI monomers across an interlaced actin network (Fig. 4 a). Each monomer head releases stochastically from the actin network at a rate of 5 s⁻¹. After its release, it undergoes ATP hydrolysis and a search to find the next actin-binding site at random, although the structure of the head and the polarity of the actin filaments lead to heads binding forward in the pointed-end direction (Fig. 4 a, middle). Upon binding to the actin filament, it strokes and attempts to move the nanosphere forward (Fig. 4 a, right). We modeled the flexibility in the myosin molecule, between the nanosphere and the actin attachments, as an elastic spring that allows finite linear distension. The flexibility in the myosin molecule refers to the effective spring constant of all compliant elements in it. The compliant elements are modeled as linear Hookean springs (fixed spring constant) for small distensions with a transition to a very high spring constant for deformations beyond this finite linear distension. This model predicts two extremes, with very little distension (∼1 nm) resulting in an unproductive stroke of the myosin and hence no movement of the nanosphere (Fig. 4 b). The second is for very large distensions of the springs (∼50 nm) comparable to the size of the nanosphere. Under these conditions, the elastic energy of each myosin VI stroke is stored in the system of springs. The full stroke from each of the four motors contributes to movement of the nanosphere, though with a cumulative dwell time that is considerably less than four times the dwell time of each motor. The net effect is a very large simulated speed (∼450 nm/s), which exceeds the speed of a single myosin VI monomer (150 nm/s). We estimate a finite flexibility of ∼3 nm for the myosin VI monomer (Sivaramakrishnan et al., 2008b). This estimation results in a speed of ∼50 nm/s for the monomer-coated nanosphere. The simulated trajectories of the monomer-coated nanosphere, similar to the dimer, are linear and directed toward the nucleus (Fig. 4 c).

These model predictions were verified by experiments using the myosin VI monomer (Fig. 4, d and e). The distribution of speeds from the simulation overlaps well with that observed in the experiment (Fig. 4, f and g). Our simulation does not account for changes in kinetics of monomers when restrained by the finite distensibility of bound heads, which likely reduce the nanosphere speed and lead to broadening of the distribution, as observed in the experiment (Fig. 4 f). The end–end speeds of dimer (65 ± 30 nm/s; n = 156)- and monomer (47 ± 10 nm/s; n = 157)-coated nanospheres are similar. Furthermore, the movement trajectories for dimer and monomer are qualitatively similar (Fig. 1, e–g compared with Fig. 3, d and e). Thus, we infer that dimerization, which gates the two heads of an unconventional myosin and renders a single dimer processive in a single-molecule assay, is not essential for processive movement so long as multiple motor heads simultaneously interact with the actin network.

To gain insight into the effect of motor density on movement of vesicular cargo, we examined the interaction of nanospheres with ∼100 dimers or ∼200 monomers on the actin network. Simulation and experimentation both confirm that nanospheres with large numbers of dimers or monomers are not stalled, but instead creep along at speeds of ∼2-3 nm/s. Thus, although there appears to be no limit to motor density for vesicle movement, large numbers of motors do impede efficiency of cargo transport.

In summary, we used myosin VI to directly test the widely proposed hypothesis that the keratocyte lamellipodium is uniformly polarized, with barbed ends pointed toward the leading edge of the cell. Our study demonstrates that the extracted keratocyte lamellipodium is a model polarized interlaced actin network with a large area (∼10 × 30 µm) to study the coupled function of unconventional myosins. We find that multiple myosin VI molecules, either artificial dimers or monomers, can cooperate to move endosome-sized nanospheres over the entire length of this model F-actin mesh. Using simplified simulations, we show that the known stroke size, duty ratio, and kinetics of myosin VI, combined with a digitized model of the keratocyte lamellipodium, is sufficient to explain the observed movement of nanospheres coupled to myosin molecules. In the context of multiple myosin VI motors interacting with multiple actin filaments in vivo, we find that dimerization of the myosin VI is not essential for processive movement of cargo vesicles under the low load conditions of the assay. The effect of external load on the movement of the dimeric myosin VI–bound nanospheres compared with the monomeric myosin VI–bound nanospheres will be an important future study. This model system will enable the characterization of the function of unconventional myosins, both in their purified form, as in this study, as well as isolated from cells, bound to their native cargo.