Disruption of the MT Alters Processivity But Not Step Size Under ∼Zero Load.
We used total internal reflection fluorescence microscopy (TIRF) (
26) and gold nanoparticle tracking (GNT) to observe dimeric swivel constructs stepping processively along actin (
27). Single-molecule TIRF observations demonstrate that all three dimeric swivel constructs (GGSGGSGGSGG inserted at residues Leu
914, Gln
931, and Arg
941) are processive but with run lengths that decrease as the swivel position is moved closer to the PT (
Fig. 1B and
Table 1). Higher temporal and spatial resolutions than those provided by TIRF microscopy are necessary to elucidate the stepping mechanism for both M6dimer and the swivel constructs. We therefore used GNT, which enables nanometer-resolution tracking at ∼1,000 times higher frame rates than single fluorophore imaging (
28). We used evanescence darkfield microscopy (
29) to image gold particles functionalized with single myosin VI molecules as they moved along surface-immobilized actin (
Fig. 2A and
Fig. S2). Total internal reflection (TIR) illumination markedly decreased background from out-of-focus particles (
30), allowing us to work at higher particle concentrations than in previous experiments (
27).
Step size and dwell time distributions for M6dimer are in good agreement with previously reported data, demonstrating that the attachment of the gold particle does not perturb stepping (
Figs. S3 and
S4) (
31–
37). All three swivel constructs take forward steps of between 30 and 34 nm in the GNT assay, nearly indistinguishable in size from those of M6dimer (
Fig. 2B and
Fig. S3 and Table 1). This observation is surprising given that the inclusion of the swivel disrupts the lever arm. However, our data are similar to observations made with a myosin V construct containing flexible breaks in its lever arm. This myosin V construct takes processive steps similar in size to those produced by control constructs (
38). It is possible that structural constraints imposed by actin favor ∼30 nm steps for both myosin V and VI even in the presence of flexible lever arms.
Individual steps for both M6dimer and the swivel constructs are rapid (
Fig. 2C). In the absence of an observable diffusive substep, we infer that the timescale for rebinding of the front head is fast (> 200 s
-1), consistent with rebinding rates measured for myosin V, a processive plus-end directed motor (
27). These free head rebinding rates are fast relative to the rate of the weak-to-strong binding transition in the new lead head (reported as ∼40 s
-1) (
25), which should help the swivel constructs to maintain a high duty ratio and hence processive stepping.
Dimeric Myosin VI Exhibits Multiple Kinetic Steps at Rate-Limiting ATP Concentrations.
The hypothesized roles of myosin VI in cargo transport and anchoring would likely be enhanced by mechanisms that prevent premature detachment from actin. In myosin V, intramolecular strain slows ADP release from the front head, thus inhibiting ATP-mediated detachment from actin (
40,
41). We sought to determine whether a similar mechanism, termed ADP gating, is also part of the myosin VI stepping mechanism.
If intramolecular tension sensing significantly increases the affinity of the lead head for ADP, the observed dwell time distribution will follow a sequential exponential model in which ADP must leave the rear head before ATP can bind (
Fig. 4A). In contrast, if ADP gating is absent (unaltered lead head ADP binding and release kinetics), ADP would leave the front head at ∼5 s
-1 (
2,
21,
35). In the latter case, if the stepping rate is sufficiently slow (in the presence of < 250 μM ATP) ADP is likely to leave the front head before the next step occurs, resulting in a new trailing head that is devoid of ADP (
Fig. 4B). In this second scenario the single rate-limiting step would be ATP binding, resulting in a single-exponential dwell time distribution.
The distribution of waiting times between steps for M6dimer in either the presence or absence of applied load and ∼100 μM ATP is well-fit by a model containing two sequential processes (
Fig. 4 and
Tables S1 and
S2). The slow and fast rates are consistent with previous measurements of ATP binding and ADP release at this ATP concentration (
21). Because the stepping rates under these conditions are considerably slower than the rate of ADP release, we infer that ADP remains bound to the front head prior to the step.
Our data are most easily interpreted by a class of models in which ADP release must occur before ATP can bind at the rear head. Although several models potentially fulfill this requirement (see
SI Text), the model that most completely and simply accounts for our data is one in which intramolecular strain blocks ADP release in the front head. We estimate an upper bound for the front head ADP release rate of 0.4 s
-1 based on Monte Carlo simulations (see
SI Text).
Our data and model are consistent with prior work in which myosin V dwell time data collected in the presence of rate-limiting ATP concentrations were fit to sequential exponential distributions (
21,
42). Although it was not appreciated at the time, previously collected optical trap data for myosin VI provide additional evidence in favor of gated ADP release (see
Table S2). Importantly, our GNT data suggest that ADP gating is an integral part of the myosin VI catalytic cycle even in the absence of applied load. Our model is qualitatively consistent with the modest deceleration of ADP release seen in single-headed myosin VI molecules under plus-end-directed load (
43). However, a larger change in ADP release rate is required to quantitatively explain our experimental results. This discrepancy may stem from a difference in construct lever arm lengths in the two studies. Alternatively, ADP gating may be strongly geometry dependent, such that it occurs optimally in the context of a dimeric molecule.
Our data are apparently at odds with portions of the model proposed by Sweeney et al., wherein ADP release from the front head is not hindered (
25). One possible explanation for this apparent discrepancy is that Sweeney et al. draw their conclusions from bulk data in which the data reflect the initial encounter between myosin VI and actin, as opposed to the subsequent processive steps. An alternate explanation is that Sweeney et al. use a fluorophore-derivatized ADP as a probe of nucleotide binding kinetics. Fluorophore-derivatized ATP is known to exhibit altered kinetics as compared to unmodified ATP in studies performed with myosin V (
44).
Importantly, Sweeney et al. additionally propose that intramolecular strain blocks ATP binding to the lead head (
25). The presence or absence of front-head ATP gating is not readily tested by our measurements. Thus, while blocked ATP binding to the front head (as proposed by Sweeney et al.) is not necessary to explain our data, it is also not inconsistent with our measurements.
Our data are consistent with a recent report in which single-molecule fluorescence studies were interpreted to support a structural model in which the lead head lever arm is held in its prestroke conformation by intramolecular strain (
45). Although Reifenberger et al. propose a kinetic model that lacks ADP gating, the strained lead head conformation that they report provides an appealing mechanism for gated ADP release, analogous to previous myosin V models (
40,
41).
We additionally measured the dwell time distributions for swivel 1, 2, and 3 at 200 μM ATP using GNT (
Fig. S5). Sequential exponential fits to the swivel 2 and 3 dwell time data yield rates consistent with those measured for M6dimer (
Table S1). Interestingly, the fit to the swivel 1 data yields rates of 12(9.9) ± 6 and 1.5(1.4) ± 0.1 s
-1 (fit rates are derived from bootstrap analysis; MLE value is provided in parentheses). The rate of 12 s
-1 is unlikely to result from fast ADP release as compared to M6dimer: smTIRF velocity data show that swivel 1 steps are slower, not faster, than M6dimer at saturating ATP (
Table 1).
Swivel 1 makes shorter processive runs and takes more backsteps as compared to M6dimer (
Table 1). A possible explanation for these observations is that ADP release is only partially blocked in the lead head of swivel 1 due to the disruption of intramolecular tension. In this scenario, premature release of ADP from the front head of swivel 1 would allow its ATP-mediated dissociation from actin, which would in turn increase the likelihood of both backward steps and complete detachment from the filament. Although considerable caution is warranted given the uncertainty in the fit parameters (see
SI Text), the stepping kinetics we observe for swivel 1 are likewise consistent with partially disrupted ADP gating in the absence of load.
Fits to the swivel 1 dwell time distribution observed in the optical trap under 1.25 pN of backward load and 225 μM ATP yield ATP binding and ADP release rates that are similar to those of M6dimer (
Fig. S6 and
Table S2). Swivel 1 stepping kinetics measured at 1.5 mM ATP and 1.2 pN load are likewise consistent with a swivel 1 rear head ADP release rate that is similar to the M6dimer ADP release rate under these conditions (
Fig. S7 and
Table S2). These data are thus consistent with gated front-head ADP release in the presence of applied load, as in M6dimer.