Myosin VI undergoes a 180° power stroke implying an uncoupling of the front lever arm

Before discussing the details of myosin VI data, it is important to keep the previous defocused imaging measurements on myosin V in context when thinking about what might be expected for myosin VI. In the case of myosin V, when the orientation of the lever arm was measured via defocused imaging, the β angle of the trailing head was 57° and when the head was leading it was 128°: hence, the change in β was approximately 71° (13). As demonstrated in Fig. 3A, if myosin VI walks in a hand-over-hand method through a rotation of its light chain domain (i.e., the lever arm) similar to myosin V, then one would expect the bifunctional probe to alternate between one β angle for the post-power stroke state (the trailing head) and a second β angle for the pre-power stroke state (the leading head). For example, the rotation shown in Fig. 3A would result in a rotation of the LCD of myosin VI during the power stroke that is near Δβ = 70°. If the lever arm rotates as shown, then one would expect to see a rotation of 90–100° in the φ angle within the lab frame while the θ angle will remain fairly constant. However, no such rotation in β is observed for all three labeled positions on the calmodulin.

While at first glance the data for all three labeled positions might seem contradictory, the explanation that best fits the data are that the light chain of myosin VI undergoes a 180° rotation with each step. The first evidence for a 180° rotation is from the observation that the β angle for all three labeled positions on the calmodulin remains constant throughout the motion of myosin VI. Fig. 3B demonstrates how a 180° rotation will result in no change in the measured value of the β angle for the [83, 90] and [137, 143] labeled calmodulin positions, although the plot is also applicable to the β angle of the [66, 73] position. The dipole of the fluorophore is shown with the orange/red arrow on the left hand side of Fig. 3B in the pre-power stroke state. In this state the red arrow is pointing down. As the lever arm rotates during the power stroke the dipole rotates 180° so that the red arrow is now facing up (right hand side of Fig. 3B). While the red and orange arrows flip position during the power stroke, the emission pattern from a dipole is degenerate when rotated by 180° so that the defocused pattern would remain unchanged. That is exactly what is observed for the [83, 90] and [137, 143] labeled calmodulin positions. Since the dipoles of these two positions are parallel to the plane formed by the light chain domain (i.e., where the calmodulins bind) of myosin VI and the actin filament the defocused pattern remains unchanged as myosin VI walks along actin resulting in the α and β angles remaining unchanged as well.

While the [66, 73] labeled position on the calmodulin is slightly more complicated, the β angle remained unchanged as well during myosin VI motion again supportive of a 180° rotation of the light chain domain. However, the [66, 73] labeled position of calmodulin yields even further evidence supportive of a 180° rotation than the [83, 90] and [137, 143] labeled positions. Fig. 2B showed that while β remains constant, α changed with each step of myosin VI. Fig. 3C demonstrates how this is further evidence of a 180° rotation of the light chain domain. Unlike the other two labeled calmodulin positions, the [66, 73] position resulted in the dipole not being parallel to the plane formed by the light chain domain and the actin filament. This results in the dipole having a different azimuthal angle, α, as the light chain rotates during the power stroke. This is measured in the lab frame through a rotation in φ of 180°. This was what was observed in Fig. 2B.

Finally, why did the α angles for the [83, 90] and [137, 143] labeled positions rarely change as myosin VI walked along actin? Since the dipole of these two positions is in the same plane as the light chain domain and actin filament we would expect the α angle to remain unchanged during the 180° power stroke of the lever arm. However, since myosin VI takes such a broad distribution of step sizes as shown in Fig. S3, then eventually the α angle of the probe should change as the myosin VI binds to different actin monomers with different azimuthal angles. Yet, the data does not support an individual myosin VI binding to actin monomers with significantly different azimuthal angles and hence the motor does not appear to rotate around the actin filament as it walks. Rather it appears that myosin VI walks in a relatively straight line along actin thereby keeping a relatively constant α angle. Remember that the broad distribution of step sizes in Fig. S3 is a result of many myosin VI molecules and not a single molecule. Hence it is plausible that an individual molecule will only access the actin monomers directly in front of the motor and rarely attempt to bind to monomers that would require a significant rotation of the head around the actin. However, we cannot rule out the fact that the experimental setup played a role in our observations of no change in α for the [83, 90] and [137, 143] labeled calmodulin positions. The α angle for both labeled positions is centered near the top of the actin filament. This is most likely a result of the actin being attached to the glass surface and myosin VI only having access to the top half of the filament due to steric constraints from the surface, thereby preventing or at least reducing the likelihood of myosin VI rotating around actin as it walks along the filament.

Our results do not support a small rotation of the lever arm of myosin VI as it walks along actin as was seen for myosin V but a large rotation of nearly 180°. While Sun et al. did support our results of a 180° rotation of the lever arm of myosin VI (8), our work differs from Sun et al. in that we never saw differing β angles for different step sizes of myosin VI for all three labeled positions on the calmodulin. It is hard to imagine how a 180° rotation would result in different β angles for the light chain domain of an emitting dipole due to the degeneracy of the dipole emission. However, because Sun et al. collected data on a much faster time frame, approximately 10-fold faster than our work, it cannot be ruled out that they observed the lever arm in some unsettled position while we measured the lever arm in its final resting position.

The properties of myosin VI during processive movement have been difficult to explain, and all likely emanate from the fact that it is a reverse direction myosin that must coordinate, or gate, its heads to allow both processive movement and anchoring (22, 23). Myosin VI was found to have an irregular, but large, step-size (2), and achieved this using an unusual extension of the lever arm (5, 24). However, this lever arm extension of myosin VI is largely alpha-helical (5) and cannot be so compliant as to allow the lead head to release its hydrolysis products, bind ATP and detach from actin while the rear head is attached (i.e., no “gating” of the lead head) in a processive dimer (18). Furthermore, if the lever arm extension was too compliant, we might not see a change in the lever arm angle. The lever arm of the lead head would rapidly approach its rigor or ADP-bound positions since phosphate release, which limits the rate of transition out of the pre-power stroke conformation, occurs at approximately 70 s⁻¹. This rate is much faster than the imaging time of our experiment (18). [Note that based on cryo-EM, the rigor and ADP positions of the lever arm are similar (1).] This view is inconsistent with the data in Fig. 2B, in which the α angle is observed to change with each step, which is consistent with a 180° lever arm swing when the myosin VI lead head becomes a rear head (i.e., during stepping).

A recent paper by Sun et al. (8) proposed that myosin VI walked on actin with large and variable tilting of the LCD. Our data shows no such variability in tilting of the LCD of myosin VI. One possible explanation, as mentioned earlier, of the inconsistency between our results and their results is the approximate 10-fold better time resolution of their technique. However, our traces showed clear stepwise movements confirming that our time resolution was adequate to capture steps and lever arm reorientations at the low ATP concentrations (10 μM) that we used. Furthermore, being able to accurately track myosin VI's movements along both longitudinal and transversal axis of the actin filaments allowed us to perform an extra analysis step to examine the suggested “wiggly” movement of myosin VI. Here we measured the side-to-side distances of the bifunctional rhodamine probes from the actin axis (Fig. S4). Overall, the average distance from the actin axis was approximately 0 nm. However, if the lever arm of myosin VI was going through a wiggly movement as described by Sun et al., the distance of the probe from the actin axis would fluctuate between approximately −5 and +5 nm. Hence, our result provides evidence against the suggested model by Sun et al.

Sun et al. also saw a difference between the β angle for the forward and trailing head of myosin VI. We saw no evidence of a change in β as myosin VI walks along actin for all three of the labeled positions, which, again, is most consistent with a rotation of the lever arm through 180°. While Sun et al. also claim a 180° rotation is most consistent with their data, they further state that the distance between the trailing head and the leading head can affect the orientation of the leading head's lever arm and hence this is why they see two distributions in β. The most likely explanation for the apparent inconsistencies between our work and their work is a difference in data analysis.

Sun et al., because they did not have access to translational data, chose an artificial (wrong) choice. In our case, we consistently chose the dipole angle to be in the upper hemisphere (0 ≤ α ≤ 180°; 0 ≤ β ≤ 180°), which was set by the angles that were measured in the lab frame. On the other hand, Sun et al. chose a hemisphere that lies in between the two hemispheres (i.e., −60° ≤ α ≤ 120°; 0 ≤ β ≤ 180°), which imposes a considerable ambiguity on their analysis. In our experiments, where we used the same CaM labeling as they did (Cys-66–73), the β values did not change and the average value for β was approximately 70 ± 15°. The α values alternated between 33 ± 15° and 152 ± 15°. If we choose the same hemisphere as Sun et al. did, our angular values would alternate between (α = 33°, β = 70°) and (α = −28°, β = 110°), which are very similar to the values reported by Sun et al. (see Fig. S5). The advantage of our imaging technique is that we can correlate translational movement with any change in the LCD's angle. Again, for the time scale of our images, we simply do not see a change in the value of β as myosin VI moves.

Ultimately, the gating of the heads of myosin VI must be different from that of any other processive, plus end-directed, myosins (18). With myosin V, the gating of the heads is largely due to the intramolecular strain, which prevents the lead head from rotating its converter into a position that allows the active site to release MgADP (21). This traps the lever arm of the lead head of myosin V in a position that has a marked angular change as compared with the rear head, which has rotated sufficiently to release all products and can rapidly rebind MgATP (13). For myosin VI, the situation has to be different, since the vector of the intramolecular strain felt by the lead and rear heads is reversed. If the converter subdomain of the lead head of myosin VI was prevented from rotating, then it could not release ADP. However, it has been shown that the lead head of a processive myosin VI dimer releases both phosphate and ADP at essentially the same rate as in the absence of strain (18). Instead of ADP release being blocked, the lead head is stalled on actin until the rear head detaches due to extremely slow binding of ATP (18). This implies that while the converter domain of the lead head can rotate sufficiently to release the hydrolysis products, intramolecular strain greatly slows the lead head from attaining a rigor conformation.

We are left with a paradox. There needs to be a mechanism to allow significant rotation of the converter subdomain of the lead head of myosin VI to allow product release, but it does not involve a highly compliant linkage between the heads (since we observe a 180° angular change). At the same time, our current experiments suggest that the lead head has its lever arm held at an angle similar to that seen in the crystal structure of the pre-power stroke (ADP^.P_i) state, even though it must be in a very different state to explain the product release kinetics.

A plausible explanation of our results is that in the pre-power stroke state in the absence of strain, the lever arm of myosin VI is only weakly associated with the converter. In a processive dimer, the development of intramolecular strain upon attachment of the lead head results in a complete uncoupling of the lever arm from the converter. This result allows the product release steps on actin to occur at a rate similar to the unstrained rate (18). Due to the intramolecular strain and the position of the rear head, the uncoupled lever arm of the lead head is held parallel to the actin filament in the states that allow product release. This is consistent with the structural data we have to date (16, 17). While the lever arm appears tightly coupled to the converter in the rigor-like state of myosin VI (16), the model of the lever arm position in the pre-power stroke state published by Ménétrey et al. (17) had very weak interactions between the converter and first calmodulin of the lever arm. Strain could break such interactions and allow the lever arm of the lead head to come free of the converter (see Fig. 4). This contrasts with the model of Sun et al. (8), which postulated that there is a significant compliance at the base of the lever arm to account for the variable lever arm angles that they observed. Furthermore, they ruled out the type of uncoupling that we are proposing based on the logic that the lever arm would be floppy, but we counter this argument with the notion that the intramolecular strain holds the lever arm parallel to the actin filament.

Furthermore, note that unlike other myosins that have been characterized, the converter of myosin VI is not found in a single conformation (17). There is a major rearrangement of the converter itself between the pre-power stroke and rigor states of the motor. What our new data suggests, consistent with the structures, is that the pre-power stroke conformation of the converter does not maintain a tight coupling of the lever arm, while the rigor conformation does. If one postulates that ATP can only bind rapidly to myosin VI once it achieves its rigor conformation, then this explains the paradox for myosin VI movement. The converter can rotate when the lever arm becomes free, releasing ADP at the unstrained rate, but the converter rotation is prevented from fully attaining its rigor conformation due to the composite stiffness of the linkage between the lever arm and the converter (which may be comprised of all or part of the helix immediately preceding the insert 2-CaM complex) and the unfolded three-helix bundles that constitute the extensions of the myosin VI lever arms (5).

Note our data and model argue against the model of myosin VI recently proposed by Spink et al. (25). In this model, the CaM-containing lever arm of myosin VI is extended by long single alpha helices and dimerization between the myosin VI monomers occurs only in the region of their cargo binding domains (approximately the position of the GCN4 leucine zipper in our HMM construct). This would predict that there be a significant change in beta angle between the lead and rear heads, as shown in the proposed structure in their paper. Our data are consistent with a region of dimerization between two myosin VI monomers that immediately follows the lever arm extension, as was recently shown (5). This is indicated in the diagram of Fig. 4 and further discussed in the SI Text.

In summary, the orientation data for myosin VI coupled with the position data demonstrates that there is a 180° rotation during the power stroke, and not variable angles of rotation. Furthermore orientation/position data suggest that for the two heads to coordinate their processive movement that the lever arm of the lead head must be uncoupled from the converter until the rear head detaches.

As previously described (27), a “zippered” dimeric myosin VI construct was created by truncation of porcine myosin VI at Arg-994, which was followed by a leucine zipper (GCN4) to ensure dimerization, and FLAG tag for purification. This construct was used to create a recombinant baculovirus, which allowed expression in SF9 cells (27).

The bifunctional rhodamine used in the DOPI measurements was purchased from Invitrogen (part # B10621). The labeling or any other treatment of calmodulin was always on ice at approximately 4°C. The dye reacted with the calmodulin for 7 h on ice and then the mixture was dialyzed in 1 L buffer (See SI Text for buffer and further description) twice to remove any free dye. Samples were then flash frozen and stored at −20°C. To ensure that the dye properly attached via both linkers to a single calmodulin the mass spectrum of the labeled calmodulin was obtained. If the dye attached properly then the two iodines would be absent. If it did not label properly then one of the linker-iodines would be replaced by an OH- group.

The labeled calmodulin was exchanged onto the IQ domain of the myosin VI HMM construct in 20 mM HEPES, pH 7.6, 25 mM KCl, 2 mM MgCl₂, 5 mM DTT, 1 mM EGTA, and 50 μM ATP. The concentration of myosin VI HMM in the buffer was 2 μM and the concentration of labeled calmodulin was 8 μM (a ratio of four calmodulins per myosin VI HMM). To remove the native unlabeled calmodulin, 1.2 mM CaCl₂ was added to the solution containing myosin VI and labeled calmodulin. The excess of Ca²⁺ caused the unlabeled calmodulin to be removed and a labeled calmodulin would be favored in binding to the open spot. After 10 min, 5 mM EGTA was added to chelate the free Ca²⁺ and stop the reaction. The sample was then run through a Sephadex G-75 spin column to remove unbound calmodulin. See SI Text for discussion on a second protocol for light-chain exchange.

The actin was attached to the surface via biotin-streptavidin linker system. Actin was polymerized with a mixture of monomers with and without biotin such that approximately every tenth monomer in the filament was labeled with biotin. To attach the actin to the surface of a lab built flow cell, 1 mg/mL BSA-labeled biotin was flowed through the chamber followed by rinsing and 0.5 mg/mL streptavidin followed by another rinse. After BSA-streptavidin was attached, approximately 0.2 μM of polymerized actin was added. After rinsing the flow cell yet again, myosin VI was added to the flow cell and briefly imaged to ensure that it bound to the actin. For motility, the following buffer was added to the flow cell, 20 mM HEPES, pH 7.6, 25 mM KCl, 1 mM MgCl₂, 1 mM EGTA, 10 mM DTT, 10 μM ATP, and 0.05 mg/mL calmodulin, and our oxygen-scavenging system (11). It was found through trial and error that 10 μM allowed for individual steps to be resolved.

A 532-nm diode pumped ND:YAG laser from CrystalLaser was used to excite the bifunctional rhodamine. Imaging was with an inverted Olympus microscope with a 100× NA 1.45 objective. Imaging was with a 512 × 512 pixels, 16 × 16 μM pixel size CCD camera from Andor. The dual view was purchased from Optical Insights.