Unconventional myosin traffic in cells reveals a selective actin cytoskeleton

Myosin V and myosin X both move toward the barbed end of actin filaments (17, 18) while myosin VI moves toward the pointed end (19). Therefore, we may use our ex vivo motility assay to dissect cellular actin architecture, as a motor trajectory reports the location and polarity of the underlying filament without ambiguity. The general features of actin organization are clearest in the Drosophila S2 cells. This macrophage-like cell line spreads on concanavalin-A surfaces to yield thin, circular cells (20). The circular symmetry of this cell type is apparent from the MV and MVI trajectory maps and orientation angles (Fig. 1 B, C, E, and F). Of the motors taking straight processive runs, MV generally moves to the cell periphery while MVI generally moves toward the interior. These directions are consistent with the known orientation of actin-barbed ends toward the plasma membrane (21–23). Although processive MV runs become less likely as the orientation angle decreases (and vice versa for MVI), in each case roughly 10% of the runs travel in the opposite direction with a similar velocity. This surprising motility of MV toward the cell center and MVI toward the periphery suggests that at least some actin filaments are reversed, with their barbed ends toward the cell center. This population of reversed filaments may be as high as 10%, but can be lower if these cells also contain a large population of inert actin filaments that do not support myosin motility, which would effectively dilute the reversed filament population. It is possible that the actual population of reversed filaments is sufficiently small that they have been missed in prior EM and fluorescence studies. We propose that these reversed filaments may be used in a return traffic pathway to move cargo in the opposite direction with the same motor.

Previously, we have shown that myosin X is highly processive on bundled actin filaments (6) like those found in filopodia (24). Because filopodia are bundled by the actin-bundling protein fascin (25, 26), which is encoded by singed in Drosophila (27), we analyzed extracted cells that displayed filopodia (Fig. 2A). Here, we find 3 categories of MX directionality in S2 cells. First, we find MXs that move radially outward along filopodia, with an orientation angle near 180° (n = 46/184, 25%; θ_min = 107°; Fig. 2 C and D). We also find MXs that travel along the cell circumference around the lamella edge, with a corresponding orientation angle near 90° (n = 34/184, 18%; θ_min = 55°, θ_max = 120°; Fig. 2 C and E). Tokuo et al. also observed motility of myosin X clusters along the cell edge, where it is likely involved in reorganizing circumferential fascin-actin bundles into filopodial projections (28, 29). The remainder of MX runs occurs throughout the cell body on actin structures that are difficult to distinguish due to the density of filaments (n = 104/184, 57%; Fig. 2F). This last set of runs does not show a preferred direction. MX motors making runs on filopodia and around the lamella move at an average velocity of 0.58 ± 0.21 μm/s (± SD, n = 26; Fig. S3), with an unusually broad distribution compared to MV and MVI (Table 1). This velocity is 2 times faster than we observed previously in vitro (6), suggesting that filopodial MX events in the ex vivo motility assay are on more organized bundles.

Unlike the trajectories in S2 cells, myosin motility in the fibroblast COS-7 cell line is far more isotropic, with all motors showing runs sampling all orientation angles (Fig. S7). This suggests a lack of the global actin organization apparent in S2 cells, although we cannot rule out more locally organized arrays of actin. However, we do not observe large populations of runs that align with obvious actin bundles or stress fibers. In the osteosarcoma cell line U2OS, we see little motility of any sort (discussed further below). The most active motor, MVI, seems to track along the prominent stress fibers that are found in this cell line (Fig. 3 and Fig. S7).

When presented with a high density of crossed actin filaments, myosin motors can switch actin filament tracks in the middle of a processive run and travel in a new direction. We define a track-switching event as a trajectory with an internal bend of at least 20°, and infer that the motor switched filaments when such a sharp bend is observed (e.g., Fig. 1H). A single actin filament is unlikely to bend over these short lengths. Using this criterion, we generally find that 5–15% of motors switch tracks within a processive run (Table 1). But MV has been suggested to switch actin tracks with nearly a 50% probability at crossed filaments that are in direct contact (30, 31). We propose that cellular actin filaments, although crowded, have sufficient separation to inhibit this track-switching behavior. Curiously, we found that track switching is particularly common for MX motors in COS-7 cells for unknown reasons (Table 1).

With 1 notable exception, we find run lengths comparable to the in vitro values for all motors across all cell types that exhibited significant runs (i.e., excluding the rare MV and MX runs on U2OS, see below) (Table 1, Fig. S4). Thus, the cellular actin architecture neither enhances nor inhibits processivity. Since our runs are generally straight, with few track-switching events, it seems likely that runs begin and end on the same filament. Thus, the motors must walk along filaments that are longer, on average, than the observed run length, or about 1–1.5 μm on S2 and COS-7 cells. Although a parallel bundle of short actin filaments could serve as a potential substitute, these did not enhance the processivity of myosin V in vitro (6). Surprisingly, we found an unusually long run length of 0.81 ± 0.07 μm (± SE, n = 196; Fig. S4) for MVI in S2 cells. This run length is 3 times longer than observed previously in vitro (15), suggesting that the architecture of actin within these preserved cells enhances processive MVI runs. One possibility is that an unknown actin-ABP complex promotes myosin VI motility, much like fascin does for myosin X (6). Since MVI seems to run along stress fibers in U2OS cells, stress fibers may present a favored architecture for MVI. Curiously, MVI run lengths drop significantly from S2 to COS-7 cells, while the run lengths for the other 2 motors are nearly the same in these 2 cell lines (Table 1 and Fig. S4). This implies that in the COS-7 cells, myosin VI moves along a distinct population of actin. These actin filaments are either shorter or have roadblocks that disrupt processive motility.

We also found ample evidence for regioselective motility, where motor activity is modulated from region to region across a single S2 or U2OS cell. As discussed above, in S2 cells with prominent filopodia and circumferential fascin-actin bundles, we find nearly half of the MX events in these narrow zones (Fig. 3). Surprisingly, we find that myosin VI is excluded from the same regions. In the S2 cell shown in Fig. 1E, only 1 filopodium out of 11 supports MVI motility (n = 3 motor events) even though numerous motors land on the other filopodia. We believe this filopodium has a unique feature that allows MVI motility, since statistically there is only a 1:121 probability that all 3 events would occur on 1 filopodium (i.e., the probability that 2 subsequent events occurred on the same filopodium as the first event). One possibility is that there are general features of filopodial actin that exclude MVI, but these features have been lost from the single filopodium. Alternatively, the other 10 filopodia may have remnants of plasma membrane that block all motors, but we note that we still observe MVI landing events on these filopodia, and that myosin X walks into filopodia and is able to escape at the tips under the same conditions. Even though MVI is apparently excluded from filopodial fascin-actin bundles, it does seem to walk along α-actinin bundles found in stress fibers (Fig. 3).

Even more surprisingly, we find that motor velocities can vary across the surface of a single cell. For MV in S2 cells shown in Fig. 1B, the motors move at a velocity of 0.36 ± 0.08 μm/s (± SD, n = 138), while the subset of motors in the center of the cell move slower (0.17 ± 0.09 μm/s, ± SD, n = 25) (Fig. 1D and Fig. S4). The S2 cells are thicker in the center, but a cosine error from oblique actin filaments cannot account for a 2-fold difference in velocity, since we do not observe the motor moving out of focus. Previously published in vitro TIRF velocities for MV agree with the faster motors at the periphery (14, 32). Thus, it seems that some feature of the central actin is either inhibiting the rate-limiting ADP release step (33, 34), or introducing a new rate-limiting process without significantly affecting the MV run length. Alternatively, the arrangement of actin filaments in the central region may force MV to take a shorter step, with unaltered kinetics.

The COS-7 cells provide a counterexample of cells that lack apparent regioselectivity, as they display uniform motility of all 3 motors across the entire cell surface (Fig. 3). Any regioselective motility that does exist in a COS-7 cell must be confined to small areas that are effectively randomized over the whole cell and not immediately apparent in our assay.

We also found evidence that each cell line differentially regulates global myosin motor traffic, distinguishing between motor classes. We find that both S2 and COS-7 cells support active MV motility, reflected in the fraction of motors that move once they have landed on the cellular actin. We find a lower likelihood that MVI moves in these cells, and MX is the least likely to move (although the difference is insignificant between MVI and MX in COS-7 cells). Although all of these motors likely play multifaceted roles, the general trend is that vesicle traffic (MV) (35) dominates over organelle anchoring and endocytosis (MVI) (36), followed by integrin or filopodial traffic (MX) (37). U2OS cells show a different pattern, with moderate levels of MVI motility, and a statistically significant reduction of MV and MX motility.

Cells tightly regulate the assembly and disassembly of actin, resulting in specific changes in morphology and cell motility. Our results here suggest that the details of cellular actin organization also impact the activity of unconventional myosins. Using this information on myosin motor motility, we can now construct a map of the roads or actin structures in different cell types, as shown in Fig. 4. This roadmap reveals both the spatial arrangement of the actin filaments, as well as the individual motility preferences in the 3 cell types. We note that the 2 cell types that support the most myosin traffic, S2 and COS-7 cells, are likely to have the greatest requirements for intracellular trafficking. The process of an S2 cell spreading on a concanavalin-A surface has been described as a frustrated attempt to phagocytose the surface (20, 38). Such a cell would need to actively traffic integrins and other receptors to and from the plasma membrane. Likewise, an actively migrating fibroblast such as a COS-7 cell would likely have active trafficking pathways for cell migration. Both of these cell types have active actin polymerization at the lamellopodium, with retrograde flow rates of approximately 4 μm/min determined by speckle microscopy (39). Ostap has proposed that this pool of new dynamic actin filaments can select for certain myosin classes such as myosin Ib, since inhibitory tropomyosins are not found in regions with rapid filament turnover (40). Our myosins may prefer these regions as well for similar reasons. In contrast, the slowly migrating U2OS cell maintains prominent stress fibers and is well-anchored to the substrate, with correspondingly low actin retrograde flow rates at the leading edge (≈0.25 μm/min) (41). The ability of myosin VI to move along U2OS stress fibers may reflect this motor's role in the load-dependent anchoring (42) of organelles such as the Golgi (36) rather than active trafficking. Since we used the same motor constructs on all 3 cell lines, these motors are most likely regulated at the level of the underlying actin architecture. Our future work will identify the actin roadblocks and detour signs that control the flow of motor traffic, in part through the siRNA-mediated depletion of ABPs that are potential players in this system.

The myosin X HMM forced dimer construct containing GCN4, GFP, and FLAG (6) was used to create recombinant baculovirus in Sf9 insect cells, purified via Flag-affinity chromatography. Myosin V and myosin VI HMM-GCN4-GFP-Flag constructs were likewise expressed and purified in Sf9 cells using baculovirus expression system as previously described (15). Motor stock concentrations were determined from GFP absorbance. Exchange reactions with Cy5-labeled calmodulin were performed by a calcium pulse as previously described with a 3-fold molar excess of labeled calmodulin over myosin (13).

Drosophila S2 cells were plated on concanavalin-A coated coverslips and allowed to adhere for 1 h to overnight at room temperature. COS-7 and U2OS cells were plated onto coverslips and allowed to adhere for 12–48 h at 37 °C. Coverslips were treated with extraction mixture [0.25–1% Triton X-100 (Calbiochem), 4% wt/vol PEG, PEM buffer (100 mM Pipes, pH = 6.9, 1 mM MgCl₂, and 1 mM EGTA), and 1 μM TMR-phalloidin] for 2–5 min. Longer adherence time typically allowed for higher detergent concentration without loss of cell adhesion. Treated coverslips were assembled into a flow cell comprised of 2 microscope slide fragments used as spacers adhered to a microscope slide with double stick tape (chamber volume ≈170 μL). Motility solution containing myosin motors, ATP (2 mM or 1 μM with 3 mM free Mg⁺²), and an oxygen scavenging system (25 μg/mL glucose oxidase, 45 μg/mL catalase, and 1% wt/vol glucose) were washed into the flow cell and sealed with VALAP (equal weight of lanolin, Vaseline, and paraffin). Treated S2, COS-7, and U2OS coverslips were also subjected to hot (95 °C) SDS/PAGE sample buffer (60 mM Tris-HCl, pH = 6.8, 25% (wt/vol) glycerol, 0.01% (wt/vol) bromophenol blue, and 1 mM 2-mercaptoehtanol). Collected samples were heated at 95 °C for 5 min, loaded onto 4–20% SDS/PAGE gradient gels (Pierce), and subjected to SDS/PAGE (Fig. S1).

Myosin motility was imaged using a custom-built objective-type total internal reflection microscope. Images were collected with a 100×, 1.45 NA objective (Olympus and Zeiss) and an EMCCD camera (iXon, Andor Technologies). Illumination intensities were 11.6 W/cm² at 532 nm (TMR) and 15.4 W/cm² at 633 nm (Cy5). Frames were collected at 2 Hz with a pixel size of 60 nm. Actin was imaged first and then excitation was promptly switched to the motor channel to image motility. Overlaying the actin image with the motility movie facilitated the identification of moving motors on actin within cells. We imaged 3–4 cells per coverslip and analyzed 2–20 cells per cell type per motor class. We identified moving spots by eye and then tracked them by eye and verified selective runs with spottracker in ImageJ (43, 44). Data for spots that moved <1.5 sec was discarded from analysis to eliminate misidentified diffusive events. We tested for observer bias by having a third-party tabulate movies of U2OS cells with MX motors as a blind trial. When we compared trajectory maps, 95% of the runs were found in both sets, with few differences in runs (missed events n = 3/64; track-switching differences n = 3/64). Near-TIRF movies were also collected and analyzed as above (7, 8). With NTIRF the incidence angle is not totally internally reflected to enable a larger image depth within the sample (≈500 nm). We estimate that trajectory angular resolution is ± 3°. Faster frame rate movies were also acquired as above in COS-7 cells with MVI. These movies were acquired with a 0.05-s exposure for 200 frames. Fifty (25/movie) random runs were tabulated ranging from 8–85 frames in length. We imaged at 23 °C.