Myosin VI (myoVI) and myosin Va (myoVa) serve roles both as intracellular cargo transporters and tethers/anchors. In both capacities, these motors bind to and processively travel along the actin cytoskeleton, a network of intersecting actin filaments and bundles that present directional challenges to these motors. Are myoVI and myoVa inherently different in their abilities to interact and maneuver through the complexities of the actin cytoskeleton? Thus, we created an in vitro model system of intersecting actin filaments and individual unipolar (fascin-actin) or mixed polarity (α-actinin-actin) bundles. The stepping dynamics of individual Qdot-labeled myoVI and myoVa motors were determined on these actin tracks. Interestingly, myoVI prefers to stay on the actin filament it is traveling on, while myoVa switches filaments with higher probability at an intersection or between filaments in a bundle. The structural basis for this maneuverability difference was assessed by expressing a myoVI chimera in which the single myoVI IQ was replaced with the longer, six IQ myoVa lever. The mutant behaved more like myoVI at actin intersections and on bundles, suggesting that a structural element other than the lever arm dictates myoVI's preference to stay on track, which may be critical to its role as an intracellular anchor.
Myosin VI (myoVI) and myosin Va (myoVa) are double-headed, processive molecular motors that have the capacity to transport intracellular cargo over long distances by stepping in a hand-over-hand fashion along actin filament tracks [1-6]. MyoVI is the only known class of myosin that moves toward the minus end of actin filaments . The actin cytoskeleton is polarized in cells with their plus ends pointing toward the cell periphery, so it is not surprising that myoVI is involved in endocytosis while myoVa is critical for exocytosis [8, 9]. The actin cytoskeleton is a complex network of intersecting filaments, and bundles such as those found in filopodia and stress fibers, that act as an intracellular highway for myoVI and myoVa transport. This actin network presents a physical and directional challenge for efficient intracellular cargo delivery. Do differences in the structures and stepping dynamics of myoVI and myoVa provide an advantage to either motor when maneuvering through the complexities of the actin cytoskeleton?
MyoVI and myoVa (Figure 1A) are each composed of two heavy chains, containing an N-terminal motor domain (i.e. head) followed by a rigid lever arm that amplifies small conformational changes in the motor domain associated with the hydrolysis of ATP. Following the lever is a C-terminal tail with both dimerization and cargo-binding capacity. MyoVI and myoVa differ in all three domains. The myoVI motor domain contains a unique insert (insert 2) which redirects the lever arm swing so that the motor travels in the opposite direction to other myosin motors . The conventional lever arm of myoVI is short and comprised of a calmodulin bound to insert 2 and a single calmodulin-binding IQ motif, but is functionally extended by elements distal to the IQ motif [11, 12]. In contrast, myoVa contains a longer lever arm made up of six IQ motifs and associated calmodulins. For both motors, alternating swings of the individual lever arms result in a hand-over-hand stepping pattern [2, 4-6]. It was surprising that myoVI was able to take >30 nm steps, comparable to that of myoVa, despite its short conventional lever arm [3, 13]. To accomplish this, it has been suggested that some part of the myoVI tail domain effectively serves as part of the lever arm [11, 12], providing the additional length needed for myoVI to take its large but variable steps. The striking difference between their lever arm designs raises the possibility that the two motors explore their three-dimensional space within the cell quite differently [14, 15], which in turn could help these motors cope with the complexities of the actin cytoskeleton in unique ways.
The roles of unconventional myosin motors in cells go beyond single motor vesicular trafficking. For example, myoVa and Vb may serve as dynamic tethers to ensure that vesicle localization and retention occur at the plasma membrane through motor binding to the actin cytoskeleton [16-18]. However, it is unclear if dynamic tethering is a property of a single motor being able to attach to two different actin filaments simultaneously, or an emergent property of an ensemble of myoVa motors each attaching to different actin filaments while bound to the same cargo. In addition to endocytic trafficking, myoVI may serve as an intracellular anchor with the best example being in stereocilia where myoVI is thought to anchor the apical membrane between individual stereocilia to the cuticular plate . This anchoring role was first proposed given that load imposed on a single myoVI molecule when attached to a single actin filament in the optical trap resulted in prolonged attachment times . Therefore, we and others hypothesized that the unusual myoVI lever arm design promotes anchoring as a result of both heads preferably interacting with the same actin filament [21, 22].
To assess whether the divergent lever arm designs of myoVa and myoVI lead to different actin interactions when presented with multiple actin filaments, we defined the stepping dynamics of single molecules of expressed, Qdot-labeled myoVI and myoVa motors (Figure 1A) when challenged by in vitro models of the actin cytoskeleton. These in vitro models included intersecting actin filaments (Figure 1C), and actin filament bundles (Figure 1D) either of unipolar actin filament polarity (fascin-actin) or of mixed polarity (α-actinin-actin). A myoVI mutant in which the short single IQ motif was replaced with the longer six IQ myoVa motif was also expressed (Figure 1A). The stepping dynamics of this chimeric motor on various actin tracks helped define the extent to which the lever arm affects the ability of the motors to maneuver along complex actin tracks. We show that myoVa explores the cytoskeletal landscape more freely, and thus switches between filament tracks more often than myoVI, which prefers to maintain its direction of travel along the actin track on which it is stepping. This difference between the two motors is not simply because of the elements comprising their lever arms.
Results and Discussion
Single motor maneuverability at actin–actin intersections
The actin cytoskeleton is a dense meshwork of intersecting actin filaments, which is most apparent in lamellopdia . Here, we determined how myoVI might deliver cargo along such a complex actin network using an in vitro model system of actin filament intersections created by first adhering Alexa660-labeled actin filaments to a glass surface followed by TRITC-labeled actin filaments as described previously . Knowing the actin filament color and the sequence in which the filaments were introduced into the experimental flowcell, we could determine which filament was on top (see Figure 1C). Given the ∼7 nm actin filament diameter and the size of the myosin motors and their attached Qdots (∼20 nm), the only way a motor could pass through the intersection is by stepping onto or over the top of the intersecting actin filament.
We previously demonstrated  that when a single myoVa encountered an actin filament intersection, the majority of motors (48%) switched to the intersecting track, going equally left or right, suggesting that the polarity of the intersecting filament was random and that there was no structural bias to the stepping pattern of myoVa. Only 15% of the myoVa motors crossed over the intersection, while the remaining 37% terminated at the intersection . We proposed that the high percentage of track switching, relative to crossing, was because of the diffusive search of the leading head mapping out a target zone (Figure 1D) that encountered more available actin-binding sites on the intersecting filament [14, 15, 24, 25]. In contrast, single myoVI motors (N = 67) were almost twice as likely (26%) as myoVa to crossover at actin intersections at the expense of switching tracks (39%), but had a similar termination probability (35%) as myoVa. Assuming a target zone model, the diffusive search of myoVI may be laterally restricted, which increases the probability of the motor crossing over the intersecting filament (Figure 1D). As previously observed for myoVa, single myoVI motors can turn at intersections having angles as great as 150° (see Figure S1). These data suggest that both motors are flexible enough to allow each of their two heads to bind stereospecifically and simultaneously to different actin filaments at obtuse angles when such events occur. This inherent flexibility must exist in structures beyond their lever arms (i.e. the tail) as X-ray crystal structures do not indicate the presence of a freely rotating joint at the head–lever arm junction .
The inherent bias for myoVI not to switch filaments as often as myoVa was also observed by Brawley and Rock  when tracking individual myoVI and myoVa motors on detergent exposed COS-7 cell actin cytoskeletons, which are predominantly composed of single intersecting actin filaments. To determine if the substantial difference in the lever arm designs between these two motors contributes to this bias, we expressed a chimeric myoVI motor that was developed previously to determine the structural domains involved in myoVI directionality . This mutant contained the myoVI motor domain with its insert 2 and tail, but its 1 IQ lever was replaced with myoVa's sixfold longer lever (Figure 1A). The chimeric myoVI maintained its opposite directionality , and was processive with velocities and run lengths that were the same as wild-type myoVI on single actin filaments (Figure 2; p > 0.05).
The myoVI chimera (N = 40) showed probabilities for switching (42%) and crossing (18%) that were intermediate between either wild-type motor with a 40% termination at the intersection. Thus the long lever arm of myoVa is not sufficient to allow the chimera to mimic the maneuverability of myoVa at intersections.
Single motor transport on actin filament bundles
Directionality of travel on unipolar bundles and mixed polarity bundles
Actin filament bundles, such as in filopodia and stress fibers, are also critical components of the actin cytoskeleton. Depending on the actin cross-linking protein, bundles can be either unipolar or mixed polarity. Bundles formed in the presence of fascin, such as in filopodia, have their actin filaments aligned with their pointed ends in the same direction and are unipolar . Mixed polarity bundles, such as in stress fibers, are formed by α-actinin cross-linking . We created such bundles in vitro (see Materials and Methods) that were 22 ± 9 µm (n = 52) long for both unipolar and mixed polarity bundles (Figure 1B). To estimate the number of actin filaments contained within these bundles, we determined the ratio of the total fluorescence intensity from a cross-sectional area scan through a bundle relative to a single actin filament within the same visual field (Figure S2). This ratio, 11.8 ± 5.4 (n = 22) for unipolar bundles, and 9.3 ± 3.2 (n = 19) for mixed polarity bundles, suggests that approximately 10 actin filaments make up each bundle.
The capacity for myoVI, myoVa and the myoVI chimera to interact with bundles compared to single actin filaments was characterized by the number of Qdot-labeled motors bound/µm of actin track, and the percentage of moving motors under identical experimental conditions (i.e. 0.25 nm of available motor; Table 1; Figure S3). For all three motors, approximately 0.45 motors were observed per micrometer of single actin filaments, with 88% of these moving processively. In comparison, approximately 2.5-fold more motors of each type were bound on both unipolar and mixed polarity bundles (average = 1.16 motors/µm of bundle; Table 1). This approximate 2.5-fold increase is expected even though each bundle is composed of approximately 10 actin filaments. A significant number of the bundle's filaments are inaccessible by being either within the bundle core or those on the bottom surface that adhere to the coverslip . There is no apparent difference in track selectivity between myoVa and myoVI for single actin filaments versus bundles, as may be the case for myosin X [31, 32]. The average percentage of processive moving motors on unipolar bundles (79%) and on mixed polarity bundles (70%) was less than that on single actin filaments (Table 1), which most likely results from some motors spanning two actin filaments in a conformation that cannot move and which remain bound for the entire 16.7-second observation period.
Table 1. Motor motile parameters on filaments and bundles
Single actin filaments
Unipolar fascin-actin bundles
Mixed polarity α-actinin bundles
Processive motors % of total
Processive motors % of total
Switch direction % of total
Processive motors % of total
Switch direction % of total
Motors bound on actin filaments and bundles determined at a motor loading concentration of 0.25 nm (see Figure S3 for additional loading conditions). Switch direction refers to a motor traveling along a bundle that changes its direction of travel towards the opposite end of the bundle.
91 (total N = 72)
83 (total N = 160)
3 (total N = 160)
74 (total N = 79)
12 (total N = 79)
89 (total N = 62)
80 (total N = 274)
2 (total N = 274)
69 (total N = 153)
21 (total N = 153)
84 (total N = 69)
74 (total N = 112)
2 (total N = 112)
66 (total N = 56)
15 (total N = 56)
At least 97% of the myoVI, myoVa, and the myoVI chimera maintained a constant direction of travel toward one end of a given unipolar bundle (Table 1; Figure 3). In contrast, in mixed polarity α-actinin bundles, 12% of the myoVI, 21% of the myoVa and 15% of the myoVI chimera motors switched direction toward the opposite end within a given run (Table 1; Figure 3). MyoVI has also been reported to switch tracks less frequently than myoVa on mixed polarity stress fibers in detergent-exposed U2OS cell cytoskeletons . The percentage of motors that switch directions on a mixed polarity bundle should depend on the target zone subtended by the diffusional search of the unbound head, and the spatial distribution of actin filament polarities within the bundle. On the basis of probabilities described above for myoVI and myoVa motors at single filament intersections, myoVI should switch between filaments on mixed polarity bundles less often than myoVa. In this regard, the myoVI chimera behaves more like myoVI by maintaining its travel direction on a mixed polarity actin bundle (Table 1).
Velocity and run length of single motors on actin bundles
The velocity and run length at 2 mm ATP was characterized for myoVI, myoVa and the myoVI chimera on both unipolar and mixed polarity actin bundles (Figure 2). All three motor types exhibited significantly (p < 0.05) slower velocities on unipolar and mixed polarity bundles compared to velocities on single actin filaments, except for myoVI on unipolar bundles (Figure 2A). Previous studies of myoVa on unipolar bundles also reported slower velocities [31, 32]. With regard to run lengths, myoVa had 1.8-fold longer (p < 0.05) run lengths on both unipolar and mixed polarity bundles compared to single actin filaments, which was not the case for myoVI and its chimera (Figure 2B). The lack of any effect on run length for myoVI suggests that this motor is unaffected by the presence of neighboring actin filaments in the bundle, and prefers to track along a single actin filament. The myoVI chimera behaved similarly to myoVI, and thus the presence of the myoVa lever did not confer any advantage to the chimera in terms of enhancing its run length as observed for myoVa (Figure 2B).
For myoVa, the increased run lengths suggest that the presence of parallel actin filaments within a bundle offers the unbound head of myoVa a greater number of available binding sites, thus decreasing the probability of run termination (Figure 1D). This presumes that the myoVa motor switches tracks on the bundle, which would also account for the slower velocities because a fraction of forward motion is given up to lateral movement during track switching (see below).
Single motor stepping behavior on bundles
To better understand the molecular basis for the difference in processive properties between myoVI and myoVa on actin bundles, we characterized the stepping dynamics with high temporal (17 milliseconds) and spatial resolution (6 nm). To enhance our spatial resolution, Qdot-labeled motor displacements were observed at low ATP concentration (≤50 µm). As a control, the stepping dynamics of all three motor types under similar experimental conditions were characterized on single actin filaments.
Although the individual actin filaments within a bundle were not visually resolvable, the x,y stepping trajectories mapped out the location of parallel tracks and provided evidence that the motors were capable of switching tracks (Figure 3). On the basis of these clearly defined stepping trajectories we characterized the following stepping parameters: (i) track-switching probability per step with a switching event defined by a motor traveling down a different filament; (ii) angular displacement between steps within the visual plane (see Figure 4; Materials and Methods); (iii) step-size distributions and lifetimes; (iv) a newly observed phenomenon, head tapping.
MyoVI switched tracks at each step only 13% of the time on unipolar bundles (Figure 3, Table 2), whereas myoVa switched with twice the probability (Table 2). Although the myoVI chimera has a longer lever arm, it exhibited nearly the same track-switching probability (16%) as myoVI on unipolar bundles (Table 2). Although the leading head of the myoVI chimera might explore its landscape as well as myoVa (see head tapping data below), this sampling of neighboring actin filaments did not result in a higher probability of track switching. Interestingly, track-switching probabilities for all three motors match the probability of changing their direction of travel on mixed polarity bundles as might be expected (Table 1).
Table 2. Motor stepping, switching and tapping
Single actin filaments
Unipolar fascin-actin bundles
Step lifetime (ms)
Step lifetime (ms)
Tapping events/total runs
Average taps per tapping event
Tap lifetime (ms)
Tap step size (nm)
Step lifetimes determined as described in Materials and Methods at 50 µm ATP. Switch filaments refers to a motor switching to a different actin filament within the bundle. Average taps per tapping event were calculated by dividing the total number of taps observed for all tapping events by the number of events, where a tapping event consisted of many individual taps. The average lifetime of individual taps was measured (see Materials and Methods).
1282 ± 139 (N = 89)
1428 ± 184 (N = 103)
1020 ± 80 (N = 68)
1265 ± 120 (N = 120)
117 ± 5 (N = 97)
−30 ± 18, 32 ± 15
1036 ± 138 (N = 79)
1136 ± 33 (N = 100)
136 ± 18 (N = 96)
−34 ± 14, 33 ± 19
Given that motors must explore their track landscape in order to switch filaments, we defined the angular displacement of the Qdot-labeled head position along the actin bundle between consecutive steps (Figure 4A). As a control, these angles were also determined for all three motors on single actin filaments [1 ± 11o (SD), N = 69], giving a measurement resolution of 11° (Figure 4B). MyoVI's angular displacement distribution regardless of bundle type was narrower (average SD = 26°, N = 156) than either myoVa (average SD = 44°, N = 180) or the myoVI chimera (average SD = 34o, N = 110; (Figure 4B, C). The distributions of angular displacements are consistent with the track-switching probabilities for the three motors. Thus, it appears that myoVa has the greatest flexibility to allow both of its heads to be simultaneously attached to different filaments. Such flexibility is described by the diffusive search of the unbound head [14, 15, 25, 33] (Figure 1D), which appears to be connected to a universal joint, presumably at the junction of the lever arm and the coiled-coil tail.
MyoVI's lower track-switching probability and tighter angular distribution compared to myoVa suggest that myoVI more faithfully tracks along an actin filament within a bundle. Therefore, its step-size distribution on a bundle should be similar to that on single actin filaments, which was the case (Figure 5C). In fact, the step-size distribution on unipolar bundles [57 ± 27 (SD) nm, N = 111] was indistinguishable (p > 0.05) from that on mixed polarity bundles [60 ± 25 (SD) nm, N = 64] both of which were no different than on single actin filaments [63 ± 21(SD) nm, N = 155]. In contrast, the myoVa step size was significantly reduced (p < 0.05) and more broadly distributed on both unipolar bundles [63± 26 (SD) nm, N = 127]  and mixed polarity bundles [64 ± 24(SD) nm, N = 81] compared to single actin filaments [72 ± 10 (SD) nm, N = 142] (Figure 5F). With myoVa motors switching tracks with a >20% probability on both unipolar and mixed polarity bundles (Table 2), a significant number of steps were short steps (45 ± 24 nm, N = 32) to a neighboring filament, explaining the reduced step size and its broad distribution on actin bundles (Figure 5F). The myoVI chimera generated steps on single actin filaments [70 ± 28(SD) nm, N = 67] that were similar to myoVa in magnitude (p > 0.05), but more broadly distributed (Figure 5I) as previously reported . As for myoVI, the myoVI chimera step size and distribution was unchanged (p > 0.05) while stepping along both unipolar [65 ± 24 (SD) nm, N = 99] and mixed polarity bundles [64 ± 32 (SD) nm, N = 63] (Figure 5I). This suggests that the myoVI chimera would prefer to track along on a filament and that having a longer lever arm is not enough to alter the stepping dynamics of the myoVI motor domain.
Finally, on both unipolar and mixed polarity bundles, myoVa took occasional backsteps (∼3.5%), which was rarely the case for myoVa on single actin filaments (Figure 5F), suggesting that straddling between tracks could place an internal resistive load on the motor which prompted a backstep in the laser trap assay [34-36]. In contrast, the back-step frequencies of myoVI (∼5%; (Figure 5C) and myoVI chimera (10%; Figure 5I) were independent of the actin filament type, adding further support that these motors prefer to stay on track.
Surprisingly, ‘tapping’ events, where a motor head alternated many times between fixed positions on different two actin filaments, were observed for all three motors on unipolar bundles (see Figures 3A, C, 5E, H). Head or ‘foot stomping’, as coined by Ando and coworkers, was recently reported using high speed atomic force microscopy , although stomping in their study occurred between adjacent actin monomers on the same actin filament, i.e. a stomp of approximately 5 nm. For myoVa, tapping events were observed frequently (i.e. 13/14 runs) and for each event the myoVa head alternated approximately 8 times between fixed positions on two parallel actin filaments that were approximately 31 nm apart (Table 2, Figure 3C). Given the 12 nm center-to-center distance between actin filaments within the bundle , the 31 nm tapping step size suggests that the actin-binding sites are located on different actin filaments that are either on both sides of the filament to which the strongly bound head is attached or between an actin-binding site on the filament the motor is traveling on and another at least two actin filaments laterally (Figure 3C). As only one head is Qdot-labeled (i.e. the one tapping), this head must be weakly bound while the unlabeled head is strongly bound to actin and most likely in the nucleotide-free state given the low ATP conditions of these experiments. The single head label also prevents us from determining whether the tapping head is the leading or trailing head. Furthermore, the tapping event frequency is probably an underestimate, as the unlabeled head should be capable of tapping but is silent in our assay. Of the 13 myoVa tapping events, 9 resulted in the motor switching tracks. The lifetime of each tap was an order of magnitude shorter (Table 2; Figure 3C) than the lifetime of a forward step under these low ATP conditions, suggesting that the tapping events are not ATP-dependent. These tapping events were also observed in myoVI, but much less frequently (4/13 runs; Table 2). Although the myoVI chimera tapped as frequently (13/14 runs) as myoVa (Table 2), only 4 tapping events resulted in track switching with the remaining 9 events ending with the motor continuing along on the same track. Thus, the myoVI chimera can explore the track landscape as well as myoVa, contributing to its broad angular displacement distribution (Figure 4), but tapping does not enhance the myoVI chimera's ability to switch tracks.
The data presented here demonstrate that as single motors, both myoVI and myoVa are adept at maneuvering through actin–actin intersections as well as traveling along actin filament bundles. Using high spatial resolution imaging, the stepping dynamics of these motors revealed significant differences between the processive behaviors of these motors, most notably that myoVI prefers to stay on course, traveling along a single actin filament even when opportunities are provided by closely positioned neighboring actin filaments within a bundle. The evidence in support of this conclusion are as follows when myoVI is compared to myoVa: (i) myoVI does not switch tracks as often at actin intersections; (ii) myoVI has a lower probability of changing its travel direction on a mixed polarity bundle (Table 1); (iii) myoVI exhibited fewer head tapping events between two parallel filaments within a bundle (Table 2); (iv) myoVI had a lower probability of switching filaments per step within a bundle (Table 2); (v) myoVI has a tighter distribution of angular displacements between steps on either unipolar or mixed polarity bundles (Figure 4). All of these properties suggest that myoVI more faithfully tracks along an actin filament rather than switching filaments either at an actin-actin intersection or within a bundle. Previously, myoVI was described as walking ‘wiggly’ on single actin filaments . This wiggly behavior refers to the motion of the lever arm relative to the longitudinal plane of the actin filament  and would contribute to a broad step-size distribution associated with linear travel and not to the lesser capacity of myoVI to switch tracks on bundles.
What structural features underlie the apparent differences in maneuverability between myoVI and myoVa? One would have expected that the myoVI chimera, which combined the inherently broader step-size distribution of myoVI on single actin filaments with the longer lever arm of myoVa, to have been better equipped to explore its 3D space for possible actin-binding sites. However, the chimera did not exhibit a higher track-switching capacity. The myoVI chimera must have sufficient flexibility for the free head to tap on neighboring filaments within a bundle (Figure 5H), but these tapping events do not lead to track switching as it did for myoVa (Table 2). Therefore, a structural element other than the lever arm must limit the extent of the lateral movement of myoVI. This element appears to exist within the myoVI head domain, given the results of the chimera.
The tapping events described here are distinct from the stepping behavior of a motor in dynamic stall (i.e. back and forth steps) along a single actin filament  or stomping events seen recently in AFM images . These lateral movements between actin filaments in a bundle may be the underlying process by which a motor determines whether it switches tracks or not. With the lifetime of a tap (∼100 milliseconds) equivalent to the step lifetime of a myoVa motor when stepping at 10–15 seconds−1 at 1–2 mm ATP [35, 39], the motor would only have time to tap once on a neighboring filament in order to switch tracks. The fact that it taps at all, when never before seen on single actin filaments at low ATP, suggests that the conformation of the motor when its two heads are attached to different parallel filaments must create an off-axis intramolecular strain that results in detachment of the leading head. Ishiwata et al.  demonstrate that myoVa processivity is sensitive to off-axis resistive loading, which they attribute to a lower unbinding force of the myoVa head with certain angles of off-axis loading.
Is myoVI or myoVa more efficient at delivering cargo along the complex actin cytoskeleton? As a single motor, myoVa appears to have more options for choosing direction when presented with intracellular obstacles, given its higher probability of switching tracks. However, this flexibility may also lead to cross-linking of actin and tethering, which has been suggested to be a function of myoVa in a dense actin network [16, 41]. This feature may not be necessary, or even desirable, for myoVI when it functions in endocytosis or as an anchor, where having both heads on a single actin filament may be beneficial [21, 22, 42, 43].
Our studies only begin to define cargo transport within the complex intracellular milieu, where multiple motors share the responsibility of cargo transport. The 3D dense actin meshwork provides numerous escape routes, but at the same time presents physical challenges that motors must overcome for efficient cargo delivery to its destination.
Materials and Methods
Protein expression and purification
A double-headed myoVI construct (Figure 1A) was engineered from the porcine myoVI cDNA by truncation at residue Arg-994, following which an eGFP cDNA (Clontech), a leucine zipper (GCN4) to ensure dimerization and a FLAG tag at the C terminus to facilitate purification  were added in sequence. This construct was expressed in Sf9 cells, as described previously [3, 44]. The myoVI lever arm contains a single IQ motif, which binds an exchangeable calmodulin (CaM) [45, 46]. Therefore, expressed biotin-tagged CaM (Bio-CaM) was chemically exchanged for the resident CaM for specific labeling with streptavidin-functionalized Qdots as described previously . Proteins were purified as described [44, 47].
A double-headed myoVa-heavy meromyosin construct (Figure 1A) was engineered by truncating the full length mouse myoVa heavy chain at residue 1098, followed by YFP and a FLAG epitope tag at the C-terminus to facilitate purification after expression in the Baculovirus/Sf9 cell system, as described previously . The construct also contained a biotin tag either at the N- or C-terminus, which was used for attachment to streptavidin-conjugated Qdots (Invitrogen). The biotin tag consists of 87 amino acids residues from the Escherichia coli biotin carboxyl carrier protein, which is biotinylated at a single lysine during expression in Sf9 cells . Calcium-insensitive CaM (CaMΔall) was co-expressed with myoVa as described previously . Proteins were purified and stored as described .
A mutant myoVI chimera was constructed as described . In brief, the six IQ myoVa lever arm was substituted for the single myoVI IQ in the myoVI construct described above. Therefore, this chimeric motor retained the myoVI head, converter, insert 2 and tail. Bio-CaM was exchanged onto the mutants six IQ lever, presumably at the second IQ , for labeling with streptavidin-Qdots as described above for myoVI.
Buffer A: 25 mm imidazole, pH 7.4, 4 mm MgCl2, 1 mm EGTA, 25 mm KCl, 10 mm DTT; Buffer B: Buffer A plus 1 mg/mL BSA; Buffer C: 25 mm imidazole, pH 7.4, 4 mm MgCl2, 1 mm EGTA, 300 mm KCl, 10 mm DTT.
Actin bundle preparation
Chicken skeletal actin was labeled with TRITC-phalloidin as described previously . Unipolar fascin-actin bundles were formed by adding E. coli-expressed fascin to actin filaments as described in . Briefly, actin filaments and fascin were mixed very gently at a 10:1 actin monomer:fascin ratio in Buffer A and incubated for 4 h to overnight on ice. Mixed polarity, α-actinin-actin bundles were prepared by incubating actin filaments with α-actinin (Cytoskeleton Inc.) at a 5:1 actin monomer:α-actinin ratio in Buffer A for 4 h on ice.
Qdot to motor attachment strategy
For single molecule assays, motors were Qdot-labeled as described previously . In brief, myoVI and the chimera were labeled by initial exchange of a Bio-CaM to the IQ-domain of myoVI and presumably to the second IQ of the myoVa lever of the myoVI chimera , followed by streptavidin functionalized Qdot (655 nm emission; Invitrogen-Molecular Probes) conjugation. The myoVa N-terminus head domain was also labeled using the same streptavidin-functionalized Qdot as described previously [24, 55]. All motors were mixed with Qdots at 1:4 molar ratio and incubated for 15 min on ice in Buffer B. For these conjugation conditions, 91% of the Qdots were attached with a single motor based on binomial statistics .
To prepare microscope coverslips for various processivity assays, an experimental 15 μL flowcell was constructed from glass coverslips and solutions introduced in the following order. First, 20 μL of N-ethyl maleimide (NEM)-modified myosin  at 0.5 mg/mL in Buffer C was introduced for 2 min to create the attachment strategy for actin filaments or actin bundles, followed by a Buffer A wash. Then Buffer B (20 μL) was introduced and incubated for 2 min, followed by a Buffer A wash. Then 50 nm TRITC-phalloidin actin filaments or 200–400 nm TRITC-labeled actin bundles in Buffer A were infused into the flowcell, incubated for 2 min and then washed with Buffer A. The motor:Qdot complex was diluted to a final motor concentration of 0.25 nm in Buffer B (plus ATP and 0.1 mg/mL CaM) and then infused into the flowcell. Movement of Qdot-labeled motors along actin filaments, actin intersections (as prepared previously in  or actin bundles was observed using total internal reflection fluorescence (TIRF) microscopy.
All motility experiments were performed at room temperature (23 ± 1°C) using a Nikon Eclipse Ti-U inverted microscope equipped with a PlanApo objective lens (100×, 1.49 n.a.) for through the objective TIRF and epifluorescence (EPI) microscopy. Qdots and TRITC-actin filaments were excited with an argon laser at 532 nm. Actin filaments and bundles were observed in EPI while the Qdots were imaged simultaneously in TIRF and EPI. Typically, 1000 images were captured at 60 frames/second (with 1 pixel = 95 nm) using an intensified CCD camera (Standard Photonics, Turbo-Z, running Piper control v2.3.14 software; Stanford Photonics). Image analysis was performed using Image J 1.41O (National Institutes of Health). The spatial resolution of the imaging system was 6 nm, determined by tracking a stationary Qdot on glass surface over 500 frames as described previously .
Run length and velocity were measured as described previously . In brief, run lengths were measured from the appearance of the Qdot-labeled motor until its detachment from the actin filament using the Image J plugin MtrackJ. However, for runs on mixed polarity bundles where the motor changes direction, the total distance traveled was measured regardless of direction. Run length distributions were fitted to
where λ is the characteristic run length constant, f(x) is the relative frequency of the motor traveling a distance x and A is a constant. Velocity was calculated as the run length/run time with the velocity distribution fitted to a Gaussian: where b is the standard deviation, x0 is the mean velocity and A is a constant. All velocities are reported as the mean ± SD and the characteristic run lengths were reported as mean ± SE. Statistical significance was determined using the Kolmogorov–Smirnov test for run length comparisons  and the Student's t-test for velocity comparisons. For step-size measurement, Image J plugin SpotTracker 2D was used to track the individual Qdot-labeled motor. Then steps were identified in displacement versus time traces by an unbiased, statistically based routine program . The step-size distributions were fitted to a Gaussian as above and reported as mean ± SD.
The step lifetime histograms of myoVI, myoVa and the myoVI chimera were fitted by F(t) = tk2e−kt rather than a single exponential. As one of two heads of the motor is labeled with Qdot, only alternate steps were detected in our experiment. So, the step lifetime was that of the labeled head combined with that of the unobserved step from the unlabeled head, which was assumed to have occurred at the same rate, k [5, 59]. Step lifetimes are calculated as 1/k as the mean ± SE of the parameter estimate of the fit.
To calculate the tapping step sizes and tapping step lifetimes, a fraction of a trace (displacement versus time) was taken where tapping behavior was seen. Then a statistically based step-finding routine  was used to measure the step sizes and step lifetime as described above. The tapping step-sizes distribution was fitted to Double Gaussian:
and lifetime distributions were fitted to
where 1/t is the stepping lifetime, f(x) is the relative frequency of the motor traveling a distance x and A is a constant.
Angular displacement measurements between steps
To determine the turning angle between successive steps within a trajectory, we first identified individual steps as described above. Then the x,y values for every point indentified by the step-finding routine as part of a step were averaged and defined as the position of the motor's head for that given step (red dots; Figure 3). To align the overall trajectory in x,y space to the x-axis, a linear regression was fitted to the step positions. Given the angle, θ, that the regression makes with the x-axis, each of the head's step positions (x,y) were aligned relative to the x-axis using the following equation to generate new position values (x1,y1):
Once aligned, the angle between successive steps (x1,y1 and x2,y2) could be determined using the arctangent function in Excel:
Bundle width estimation
To estimate the average number of actin filaments that form a unipolar fascin-actin or mixed polarity α-actinin-actin bundles in our experimental condition, we selected an area (a rectangular box) on bundle and measured the integrated intensity (Figure S2A) using Image J. We moved this box onto a nearby single actin filament (Figure S2B) and measured the integrated intensity. Integrated intensity was also measured for the background (Figure S2C). Background intensity was subtracted from both the bundle and actin filament intensities, then and the ratio between bundle and actin filament intensity calculated. This ratio was used as an estimate of how many actin filaments exist within a bundle.
We counted the number of Qdot-labeled myoVI, and myoVa that were bound to an actin filament and bundle over 16.7 second (1000 frames). Then the number of Qdot-labeled motors bound per micron was calculated as a measure of actin filament/bundle selectivity (Figure S3).
We thank members of the Warshaw Lab for helpful discussions, Guy Kennedy from the University of Vermont's Instrumentation and Modeling Facility for expert instrumentation and microscopy design and assistance, Elena Krementsova for myoVa expression, Daniel Safer for the myoVI constructs. This work was supported by the American Heart Association (12SDG11930002, M. Y. A.) and National Institutes of Health (GM094229, D. M. W.; DC009100, H. L. S.; GM078097, K. M. T.).