Myosin S2 Origins Track Evolution of Strong Binding on Actin by Azimuthal Rolling of Motor Domain

Abstract

Myosin crystal structures have given rise to the swinging lever arm hypothesis, which predicts a large axial tilt of the lever arm domain during the actin-attached working stroke. Previous work imaging the working stroke in actively contracting, fast-frozen Lethocerus muscle confirmed the axial tilt; but strongly bound myosin heads also showed an unexpected azimuthal slew of the lever arm around the thin filament axis, which was not predicted from known crystal structures. We hypothesized that an azimuthal reorientation of the myosin motor domain on actin during the weak-binding to strong-binding transition could explain the lever arm slew provided that myosin’s α-helical coiled-coil subfragment 2 (S2) domain emerged from the thick filament backbone at a particular location. However, previous studies did not adequately resolve the S2 domain. Here we used electron tomography of rigor muscle swollen by low ionic strength to pull S2 clear of the thick filament backbone, thereby revealing the azimuth of its point of origin. The results show that the azimuth of S2 origins of those rigor myosin heads, bound to the actin target zone of actively contracting muscle, originate from a restricted region of the thick filament. This requires an azimuthal reorientation of the motor domain on actin during the weak to strong transition.

Introduction

The swinging lever arm hypothesis describes acto-myosin motility as arising from tilting of the myosin lever arm domain, while the motor domain (MD) is strongly bound to actin with an essentially fixed orientation (1). Various crystal structures of myosin subfragment 1 (S1, the myosin head) in different nucleotide states show a large axial tilt of the lever arm domain and support this hypothesis (2–5). However, the published crystal structures represent states not attached to actin, and thus might more properly describe the repriming of myosin by ATP hydrolysis, and not the actin-attached, force-producing steps (1). Previously, we directly imaged force-producing acto-myosin cross bridges in three dimensions using electron tomography of cryo-trapped contracting insect flight muscle (IFM), three-dimensional class averaging, and quasi-atomic model building (6,7). The results obtained suggest a more complex conformational change during force production in situ than suggested from the crystal structures.

IFM is ideal for studies of cross-bridge structure in situ because of its unique filament arrangement. The IFM filament lattice consists of a hexagonal filament array in which the thin filaments are placed midway between pairs of thick filaments (Fig. 1 A, left). A thin longitudinal section 25–30 nm thick can be cut in which all the myosin heads attaching to the intervening thin filament can be visualized because they must originate from the neighboring thick filament pair. Such a section is called a “myac layer” because of the alternation of myosin and actin filaments. Vertebrate striated muscle, on the other hand, has the thin filaments placed at trigonal positions between three thick filaments (Fig. 1 A, right). A longitudinal section of similar thickness would contain parts of two superimposed thin filaments, and still not include all of the bound cross bridges attaching to each.

Figure 1.

(A) Comparison of the filament arrangement in IFM (left) and vertebrate skeletal muscle (right). The placement of thin filaments between two neighboring thick filaments in the IFM lattice facilitates symmetric attachments, whereas the placement of thin filaments at trigonal positions in the vertebrate produces asymmetric attachments to the nominally twofold symmetric actin filament from three neighboring thick filaments. (B) Longitudinal view of all strong-binding cross bridges from Wu et al. (6) (gray); actin subunits (green and blue); and tropomyosin (yellow). The two crystal structures used as starting models are shown (red, chicken skeletal muscle S1, PDB: 2MYS; and magenta, scallop transition state structure, PDB: 1DFL). This view shows the axial range of the lever arms. (C) Orthogonal view, looking Z-ward down the thin filament axis. The crystal structures have similar azimuthal lever arm orientations that are, roughly perpendicular to the long axis of the MD in this view. Compared to the crystal structures, however, the azimuthal spread of the lever arms in contracting IFM is both large and systematically anticlockwise with respect to the crystal structure lever arms (clockwise with respect to the thick filament axis) when viewed Z-ward. (D) Diagram illustrating how the S2 domain flexibly links S1 to the thick filament backbone and allows the heads to search for nearby actin subunits. LMM is light meromyosin; HMM is heavy meromyosin. Although proteolysis of monomeric myosin typically defines S2 length as ∼50 nm, only ∼11 nm can be exposed via lattice swelling in IFM (10). This length is nevertheless sufficient to allow independent heads originating 14.5 nm apart axially to converge on and bind two adjacent actin subunits within a single target zone (7).

Wu et al. (6,7) identified and quantified a variety of cross-bridge structures, and established a criterion to distinguish weakly-bound from strongly-bound cross bridges. To fit the myosin models within the observed electron density, the lever arm position in the crystal structures used as starting models typically has to be modified in three dimensions relative to the starting crystal structure. When all atomic models were transformed to a single actin subunit and viewed longitudinally, the cryo-trapped, strongly-bound cross-bridge structures showed a range of axial tilts encompassing those seen in the myosin crystal structures, further supporting the swinging lever arm hypothesis (Fig. 1 B).

When the crystal structures of rigor and transition state analogs of myosin used in the model building are superimposed on actin in the strong-binding position and viewed down the thin filament axis, they have similar azimuthal lever arm angles (Fig. 1 C, red and magenta), implying that the lever arm swing is almost entirely axial (4). However, the cross-bridge structures from contracting IFM show a large range of azimuthal angles that are notably biased to one side of the starting crystal structures (Fig. 1 C, gray). Thus, within the filament lattice of the sarcomere, the lever arm azimuth appears geometrically constrained in ways not reflected in the crystal structures of isolated myosin fragments or even nucleotide-free acto-S1. Supporting this observation is experimental evidence suggesting that the working stroke of myosin likely involves an azimuthal component (8,9).

The conundrum can be illustrated another way, using the filament arrangement within IFM. A view of all the strong-binding cross bridges found by Wu et al. (6,7) positioned on their bound actin subunits in the global average of all repeats shows that the apparent cross-bridge origins are positioned clockwise, with respect to the interfilament axis when the thick filament center is used as the axis of rotation. The direction of view is Z-ward, which is the frame of reference used throughout this report unless stated otherwise (Fig. 2 A). This is contrary to the prediction when the crystal structures are positioned on the same actin subunits, which positions their origins anticlockwise of the interfilament axis (Fig. 2 B). To explain the biased azimuthal lever arm slew, Wu et al. (6) proposed three mechanisms, each of which have different implications for the nature of the transition from weak to strong binding by myosin. As explained below, distinguishing among them requires tracing the myosin S2 domain (Figs. 1 D and 2, B–G) back to the thick filament to determine where each cross bridge emerged from the thick filament backbone, a position that we consider to be the cross-bridge origin.

Figure 2.

Schematic view down the filament axis for three alternative models proposed in Wu et al. (6) to explain the observed azimuthal lever arm skew. View direction is from M-line toward Z-disk and is preserved in all figures. (A) Data obtained by Wu et al. (6) showing the position of the S1-S2 junction for strong-binding attachments in fast-frozen, active IFM. (B–I) Profiles of strongly bound MDs (red); of weakly bound MDs (magenta). The radially oriented arrows on the thick filaments represent three successive crowns, with the length of the arrow indicating relative distance from the observer. The circumferential arrows indicate that successive crowns follow a right-handed helix; the rotation between crowns is +33.75°. (B) Position of the S1-S2 junction of the strong-binding myosin attachments to target zone actins in IFM based on crystal structures and high-resolution cryo-electron microscopy (30). At the beginning of the weak to strong transition of models 1 (C) and 2 (G), the MD (magenta profile) contacts actin (blue and green profiles) in the weak binding state with the cleft in the actin binding face open (5) and the S1 lever arm (ELC, blue; RLC, cyan) approximately perpendicular to the long axis of the MD. S2 (thick yellow line) connects the myosin S1 head to the thick filament. With respect to the thick filament center as the frame of reference, the S2 origin in model 1 is clockwise of the interfilament axis, while for model 2 the S2 origin is anticlockwise of the interfilament axis. Both models 1 and 2 have a transient, strongly bound intermediate state before executing the powerstroke (D, E, and H) represented by the color change (to red) and closure of the cleft (34), and depicted at 50% transparency to emphasize that this state is not observed here. For model 1, there are two versions of the intermediate state, 1a and 1b (D and E), depending on whether compliance resides in the lever arm (D) or the S2 (E). S2 (thick yellow line) is oriented axially and connects to the clockwise quadrant of the thick filament in model 1 (F). In contrast, the S2 domain is azimuthally angled and connects to the anticlockwise quadrant of the thick filament in model 2 (I). The aim of this work is to distinguish between model 1 (a or b) and model 2. At the end of the powerstroke (F and I), S1 binds actin strongly (i.e., rigor), but also has the lever arm azimuthally skewed more nearly parallel to the long axis of the MD as observed experimentally (6). See text for description of transient intermediate states.

In the first model, model 1, the cross bridge originates from the thick filament clockwise of the interfilament axis when viewed from M-line toward Z-disk (Fig. 2 C). With this initial approach, the MD does not have the correct alignment for strong binding; therefore, the transition to strong binding necessarily involves an azimuthal movement of the MD around the thin filament. However, the MD movement on actin can have two distinctly different effects on the lever arm and S2, depending on their relative compliance to azimuthal bending. If the S2 domain is considered stiffer than the lever arm (Model 1a, Fig. 2 D), the azimuthal movement of the MD will flex the lever arm azimuthally producing an intermediate structure with a straighter myosin head, as observed experimentally (Fig. 2 D). The subsequent powerstroke is purely axial so that the lever arm-S2 junction remains clockwise of the interfilament axis (Fig. 2 F).

The second mechanism (Model 1b) differs from the first in reversing the relative compliance of S2 and lever arm. As the MD moves azimuthally on the actin subunit, the S2 is swung azimuthally but the lever arm relationship to the MD remains unchanged in the intermediate structure (Fig. 2 E). The subsequent powerstroke is both axial and azimuthal so that the lever arm ends azimuthally flexed, as observed, and the S2 domain is returned to its straightened position relative to the filament axis (Fig. 2 F). Both versions of model 1 begin and end identically (Fig. 2, C and F), but differ in the intermediate steps (compare Fig. 2, D and E). However, this work does not attempt to distinguish between these two variations of model 1.

In the third model, the cross bridge originates from the thick filament anticlockwise with respect to the interfilament axis (Fig. 2 G). With this approach, the MD is already correctly oriented for strong binding, and the transition to strong binding only involves closure of the cleft in the actin-binding domain (Fig. 2 H). The subsequent powerstroke involves both axial and azimuthal swings of the lever arm, such that the S1 head is straightened and the lever arm ends below the interfilament axis (Fig. 2 I), as observed experimentally (Fig. 2 A) (6). However, S2 ends the sequence angled with respect to the filament axis rather than aligned with it. Distinguishing between this model and the two versions of model 1 requires tracing the path of S2 (Fig. 2, thick yellow line) to determine from which region of the thick filament each cross bridge originates—clockwise (model 2), or anticlockwise (models 1a,b) of the interfilament axis using the thick filament center as the point of reference. While the S1 lever arms were clearly resolved in our previous work, the thick filament point-of-origin of each cross bridge was ambiguous because S2 lay too close to the thick filament backbone to be resolved. Here, we analyze the thick filament origins of individual, strongly bound cross bridges using rigor muscle swollen by low ionic strength to increase the interfilament distance, which has the effect of pulling the S1-S2 junction away from the thick filament backbone and allowing the S2 domain to be visualized in three dimensions by electron tomography (10).

IFM displays a distinctive pattern in rigor, with two types of strong-binding attachments called lead and rear bridges (11,12). Rigor lead bridges attach to the same actin target zone to which strong-binding cross bridges bind in contracting IFM, specifically the two actin subunits on each side of the filament that are exactly midway between successive troponins (6,13,14). Rigor rear bridges attach Z-ward of the target zone, near the troponin complex, but no strong-binding cross bridges are found in this location in contracting IFM. Lead and rear bridges are easily distinguished by their appearance in either transverse or longitudinal sections (11,12). Because no strong-binding cross bridges have been observed in contracting IFM at the rigor rear bridge position, we do not analyze their cross-bridge origins here.

The S2 azimuth is best resolved in transverse sections but these must be perfectly transverse to the filament axis and very thin (12–15 nm) for S2 to be visualized directly, which is technically difficult to achieve. Instead, we collected seven tomograms of relatively thick (∼800-nm) transverse sections and computationally sliced them into a series of ∼11 nm sections that were perfectly oriented, regardless of the actual orientation of the original section.

Materials and Methods

Rigor IFM fibers swollen in low ionic strength buffer were prepared for electron microscopy as previously described in Liu et al. (10). Transverse sections ∼80-nm thick were cut on a Leica Ultracut E ultramicrotome (Leica Microsystems, Buffalo Grove, IL) fitted with an ultrasonic oscillating diamond knife (Diatome US, Hatfield, PA) to minimize distortions due to sectioning compression (15,16). The amplitude and frequency of knife oscillations were manually adjusted so that floating sections had the same width and height as the block face, as measured by the reticule in the stereomicroscope attached to the ultramicrotome.

Single axis tilt series in the angle range ±65° were collected at a specimen temperature of −190°C at 120 keV, 0.3 μm defocus on a Titan-Krios electron microscope (FEI, Hillsboro, OR) using a 2° scheme from Saxton et al. (17). Images were recorded at 26,000× magnification on a Tridem Imaging Filter/Ultrascan charge-coupled device camera (Gatan, Pleasanton, CA) operated in zero-loss mode. The pixel size was 14 μm at the camera and 0.54 nm with respect to the original object. Automated data collection utilized the FEI BATCH TOMOGRAPHY software. Although these specimens were embedded in plastic and stained, they were treated as if they were frozen-hydrated and the total electron dose kept <12,000 e⁻/nm². This dose would be predicted to produce <5% shrinkage in the z direction (the section thickness) due to mass loss (18). Tilt series were merged using marker-free alignment (19) and the tomogram computed by weighted back-projection.

The sections were not cut exactly transverse; therefore, the angle of the filaments within the plane of the section was determined and the tomogram reoriented using trilinear interpolation so that the filament axis was parallel to the z axis of the tomogram. The amount of correction was 7–14°. To make the measurement of S2 origins more efficient, we extracted subvolumes consisting of a pair of thick filaments with the intervening thin filament. These were then aligned so that the filament axis was along the z axis and the interfilament axis aligned along the x axis, as illustrated in Fig. 3. Subvolumes were aligned using the thin filament as reference simply to align the center of gravity of the thin filaments along the z axis. We then used multivariate data analysis to cluster the thin filament segments according to the axial position of the lead and rear cross bridges. The lead bridges follow a left-handed helix around the thick filament. Reorienting the tomogram corrects for the slightly oblique angle, but does not remove the necessity of watching the lead bridges changing direction as the viewer steps along the filament. The clustering procedure had the utility of placing the S2 origins of the group members at a similar place in each raw subvolume, eliminating the need to scroll up and down the complete tomogram in order to find the S2 position. Measurements were nevertheless made from the raw group members and not from averages.

Figure 3.

Sample of swollen rigor images of lead bridges with exposed S2. Each image is 85 × 30 nm. Each lead bridge S2 origin is represented by a broken yellow line originating from the center of the thick filament; the break in the line allows an unobstructed view of the S2 density. A second yellow line extends from the center of the thick filament to the center of the thin filament and represents the interfilament axis. (Red arrows) Six of the eight lead-bridge origins falling on or clockwise of the interfilament axis (outliers). (Blue arrows) Lead bridges associated with a rear bridge form a second group.

To improve visibility of S2 without compromising resolution in the x,y plane, we convoluted the group members with a box function of a thickness of 20 voxels (∼11 nm) along the filament axis. This averages (blurs) the axial density but does no averaging in the x,y plane.

S2 azimuthal positions of the lead bridges were defined and measured using the I3DISPLAY visualization program (20) by manually picking in the following order: the center of the thin filament, the center of the thick filament, and the point of origin of the S2 domain on the thick filament surface. Actin filaments are often slightly offset from the line joining adjacent thick filaments. Therefore, the interfilament axis was defined as the line joining the centers of thick and thin filaments; and left- and right-side data were determined independently (Fig. 3) but combined to obtain the angular distribution. By convention, 0° is along the interfilament axis and positive angles are anticlockwise when viewed from M-line toward Z-disk using the center of the thick filament as the reference point, as illustrated in Fig. 2.

Results and Discussion

In total, 905 measurements of lead-bridge origins were made from two tomograms, on average one for each of 900 thin filaments. At best, we expect no more than four lead-bridge origins per thin filament based on the specimen thickness, but some that occur near the surface of the section may be poorly preserved. In addition, variations in preservation, staining, and missing wedge position can also affect visibility of the lead bridge origin. Finally, in many instances, the S2 segment from the lead bridge was not clearly distinguishable from the S2 of the rear bridge (Fig. 3, and see Movie S1 in the Supporting Material); we did not include any measurements where ambiguity between lead and rear bridge origins might occur. Though possibly coincidental, the number of measurements clear of rear bridge origins matches the shortage of myosin heads (25%) that are available to fill rear bridge actin targets (21).

The S2 azimuthal angles for lead bridges had a roughly Gaussian distribution centered on a mean of −26° ± 9° (Fig. 4). From previous work with actively contracting IFM, we know that myosin heads originating from two adjacent levels that are axially separated by 14.5 nm can converge on and bind to adjacent actin monomers on the same side of the thin filament (the mask-motif-type cross bridges described in Wu et al. (6,7)). Because the origins of axially adjacent myosins are rotationally separated by 33.75° (22), we expected that the angular range observed here would be at least 33.75°; and indeed, for both groups, 82% of the data fell between 0° and −35°. The 18% lying outside the expected range can be accounted for by the fact that this study examines rigor, as opposed to contracting muscle. In rigor, the absence of ATP powerfully favors strong binding to actin, even on unfavorably oriented actin subunits, resulting in distorted cross bridges such as those observed at the rigor rear bridge sites (23). In contrast, the ATP concentration is high in contracting muscle, which favors the detachment of cross bridges, especially those on unfavorably oriented actin subunits. Indeed, the only strong-binding cross bridges in contracting IFM were found at the four ideally oriented target zone actin subunits where lead bridges of rigor bind (6,7).

Figure 4.

Histograms showing the angular distribution of the S2 origins relative to the interfilament axis for rigor lead cross bridges bound to a single thin filament. Only lead bridges whose origins could be clearly distinguished from rear bridges are included.

Is a Gaussian distribution of this width expected? Myosin head origins follow a four-start, right-handed helical path about the surface of the thick filament with all azimuths equally populated when averaged with respect to the six surrounding thin filaments (22), so it might be expected that the distribution would be broader, even potentially flat over a 33.75° azimuthal range. Our avoidance of any potential ambiguity between lead and rear bridge origins may be one contributor. Ambiguity between closely apposed lead and rear bridge origins most likely occurs near the limits of the possible range of lead bridge S2 origins, i.e., near 0° and −34°. Related to this, the measurements are made visually; the distribution reflects the S2 visibility. Mechanistically, the distribution may represent actual compliance in the lever arm. This reflects lower probability for actin attachment near the limits of the range, in which case a distribution such as that observed would be the expectation.

The average azimuth of the cross-bridge origin observed in three dimensions here is similar to that previously found in two-dimensional averaged images of rigor IFM (12). The need for an azimuthal component to the weak-to-strong transition might have been inferred much earlier, had there been the benefit of a high-resolution crystal structure to guide the interpretation of those images. Although a few cross-bridge origins were observed to have positive angles (8 out of 905), most of them had negative angles, meaning they originated clockwise from the interfilament axis. Therefore, this result is most consistent with an azimuthal movement of the MD on actin during the weak to strong transition, such as depicted in models 1a and 1b in Fig. 2, C–F, and is inconsistent with model 2.

Several structural models for the weak to strong transition have been proposed. Early work from our lab (14) suggested that the weak to strong transition involved an axial rotation of the myosin MD on the target zone actin subunit. That suggestion was further developed into the roll-and-lock mechanism, which included both axial and azimuthal changes (24). Our subsequent reconstructions of contracting muscle (6,7) demonstrated no axial rolling of the MD on actin during the weak to strong transition, although the resolution was insufficient to completely exclude a small axial rotation of the MD. In contrast, the azimuthal changes were large. From these results, a purely axial rotation of the MD fails to explain the observed origins of strong-binding actin-myosin attachments in both rigor and active IFM. A significant azimuthal component to the movement is also necessary.

An azimuthal reorientation of the MD during the weak to strong transition is also consistent with x-ray diffraction changes in vertebrate striated muscle (25). After a rapid temperature jump, the intensity of the first actin layer line at ∼36-nm axial spacing increased concomitantly with tension, which was interpreted as an increase in stereospecific myosin head attachments to the actin filament. However, there was no change in the [1,1] equatorial reflection, suggesting that myosin attachment to actin was already established and had not increased during the shift from poorly ordered to stereospecific ordering. A change in the relative amounts of weak- and strong-binding attachments was sufficient to explain the observation. The 36-nm actin layer line comes from density tracks that have a steep pitch along the actin filament, and thus would be more sensitive to mass movement across these tracks (azimuthal) than it would be to mass movement along the tracks (axial). Thus, the observations of Bershitsky et al. (25) support a large azimuthal component to the weak to strong transition.

Although this result suggests an azimuthal reorientation of the MD during the weak to strong transition, it does not eliminate the possibility of an additional azimuthal movement of the lever arm during the subsequent powerstroke. That is, without some means to trap the intermediate state depicted in Fig. 2, D and E, as of this writing we cannot distinguish between models 1a and 1b. As described in the Introduction, depending on the relative azimuthal stiffness of S2 compared to the lever arm, the azimuthal reorientation of the MD will flex either S2 or the lever arm (Fig. 2, D and E), or perhaps both. Because S2 becomes angled during the weak to strong transition, when force is initiated in the direction of the filament axis, a torque would also be applied to the filaments (see, for example, Fig. 10 of Wu et al. (6)). In this case, an azimuthal movement of the lever arm during the powerstroke would contribute to swinging the S2 back to its alignment parallel with the filament axis and reduce or eliminate any torque on the thin filament at the end of the powerstroke. An azimuthal component to the powerstroke has already been implied by the crystal structure of the minus-end directed myosin VI (26,27). Our results here and elsewhere (6,7) suggest that myosin II uses a similar mechanism and give the direction for the azimuthal component: clockwise when looking Z-ward. Gliding filament assays have also observed a torque with the correct hand to achieve this effect (8,9).

The few crystal structures of myosin II S1 containing the lever arm domain with both light chains are from either chicken skeletal muscle or molluscan muscle (3,4,28,29). Only the post-rigor-state structure with the actin binding cleft open is available for chicken skeletal muscle S1, but the structure of several biochemical intermediates have been solved for scallop S1 as well as for squid (29), thereby permitting an illustration of the coupling between axial and azimuthal lever arm angles. When the scallop S1 transition state structure (Protein Data Bank (PDB) PDB: 1DFL) and the squid rigorlike structure (PDB: 3I5G) are aligned to a single IFM actin target zone subunit using an alignment to the lower 50 kDa domain of the chicken skeletal acto-S1 structure (30), they show an axial lever arm movement of ∼55° (Fig. 5A), but more importantly, an azimuthal rotation of 26° with respect to the center of the thick filament (Fig. 5 B), a value close to the average observed from rigor muscle here (−26° ± 9°).

Figure 5.

(A and B) Three superimposed crystals structures of complete S1s bound to actin illustrating the potential for large azimuthal changes in the lever arm position with progression through the powerstroke. Heavy chains of each S1 have been aligned to the lower 50 K domain of the chicken skeletal myosin S1 structure bound to actin (30), which is not shown in the figure. Actin subunits (green and blue), target zone subunits (darker shades), TM (yellow), and S1s (red, scallop ADP-vanadate transition state, PDB: 1DFL; green, squid ADP, PDB: 3I5F; and blue, squid rigorlike, PDB: 3I5G). The lever arm of the scallop postrigor structure (PDB: 1DFK) using this alignment falls azimuthally midway between PDB: 3I5F and PDB: 3I5G, but axially is close to PDB: 3I5G. Choice of the squid S1 structures provides an intermediate stage both axially and azimuthally. When placed in the context of the filament lattice, the azimuthal angle change (26°) between transition and rigorlike states is in the same direction as posited by Wu et al. (6), and almost sufficient to account for their observations. Note that the lever arm of the strongly bound chicken skeletal atomic model has the same azimuth as the scallop transition state. This may indicate a species difference.

There are two interpretations pertinent to these results. First, it demonstrates that the myosin lever arm has sufficient compliance to explain our observations, even though crystal packing forces rather than lattice forces as in muscle are enforcing the lever arm positions. Second, it suggests that axial and azimuthal angle changes could be coupled with the direction of the coupling the same as suggested by the in situ cross bridges. Because the transition state structure has an open actin-binding cleft and a raised lever arm characteristic of a pre-powerstroke configuration and the squid rigorlike structure has a closed actin-binding cleft and a rigorlike lever arm now identified with strong binding (30–32), the two states could logically be considered representative of the powerstroke in muscle, even though not from the same species. The lack of a transition state structure for skeletal myosin limits the impact of this comparison, because the chicken skeletal acto-S1 structure (a rigor structure) has the same azimuthal position of the lever arm as the scallop transition state structure when both are aligned to the lower 50-kDa domain (Fig. 1 C). We offer this comparison not as a proof, but only to suggest that crystallography may support our observations in situ.

We have used manual measurements to obtain a distribution for the origins of rigor myosin heads in the same target zone observed for actively contracting muscle, and given evidence that the weak-to-strong binding transition involves an azimuthal reorientation of the MD, which in turn culminates in the azimuthal slew of the lever arm observed previously (6,7). The actin subunits to which strong-binding bridges attach in contracting IFM is precisely defined and the density maps containing apparent strong-binding myosin heads are well fit to the MD density by atomic models of actin decorated with nucleotide free myosin. Thus, the strong-binding actin-MD interaction must reflect well-conserved features across many muscle types, not just IFM (30,32–35). Given these three observations, an azimuthal movement of the myosin MD across actin is a requirement. This motion appears to involve largely azimuthal rotations but may contain some translational movement across the actin subunit as well. While we cannot provide a detailed model at this time, we note that the movement is across subdomain 1 of actin and toward tropomyosin (TM). Thus, TM placement in the blocked state of relaxed muscle would reduce or prevent a weak MD-actin interaction. Movement of tropomyosin to the closed position would establish a more robust actin-myosin interaction (the weak-binding state) that subsequently progresses toward strong binding by pushing TM further to the open position (1).

Recent high-resolution reconstructions of actin-TM filaments decorated with rigor S1 fragments give a detailed picture of the contact surfaces among actin, myosin, and TM (32), and support our model (6) for an azimuthal reorientation of the MD during the weak to strong transition. We noted earlier that a puzzling consequence of the large azimuthal changes in the lever arm is the predicted imposition of a torque on both the thick and the thin filament. For the models 1a and 1b, model 2 being inconsistent with observation, this torque is applied through the S2 domain imposing a twist on the thick filament. If S2 is stiffer than the lever arm (model 1a), the torque is applied as a reaction to the azimuthal bending of the lever arm during the weak to strong transition. If the lever arm is stiffer than S2 (model 1b), the torque is applied as a reaction to the axial force produced by the MD on the angled S2. No such twist has been observed as a consequence of cycling cross bridges, although a small and presumably passive twisting of the thick filament has been observed in both active and relaxed IFM subject to a stretch (36). That the weak to strong transition of model 1b predicts an applied torque, yet none is observed, could be due to the powerstroke having an azimuthal component that compensates for the torque produced by a purely axial force applied to an angled S2 domain. In other words, the motion needed to negate the implied torque is encoded in the conformational changes within the MD itself, as suggested by the S1 structures shown in Fig. 5 and the actin twirling assays of optical microscopy (8,9). Comparison of the converter domain positions from recent actin-TM-myosin image reconstructions with crystals of myosin with ATP bound suggested an ∼20° azimuthal slew of the lever arm (32), in the direction that the lever arm is slewed in our model. Such a movement within the myosin motor would negate the torque applied through an angled S2 domain when force is produced in the direction of the filament axis.

Author Contributions

K.A.T. designed the research; K.A.T. and C.A. performed data analysis; H.W. contributed analytic tools; A.W. collected tilt series data; and K.A.T., C.A,. R.J.P.-E., and M.K.R. wrote the article.