Myosin isoforms show unique conformations in the actin-bound state

Abstract

Crystallographic data for several myosin isoforms have provided evidence for at least two conformations in the absence of actin: a prehydrolysis state that is similar to the original nucleotide-free chicken skeletal subfragment-1 (S1) structure, and a transition-state structure that favors hydrolysis. These weak-binding states differ in the extent of closure of the cleft that divides the actin-binding region of the myosin and the position of the light chain binding domain or lever arm that is believed to be associated with force generation. Previously, we provided insights into the interaction of smooth-muscle S1 with actin by computer-based fitting of crystal structures into three-dimensional reconstructions obtained by electron cryomicroscopy. Here, we analyze the conformations of actin-bound chicken skeletal muscle S1. We conclude that both myosin isoforms in the nucleotide-free, actin-bound state can achieve a more tightly closed cleft, a more downward position of the lever arm, and more stable surface loops than those seen in the available crystal structures, indicating the existence of unique actin-bound conformations.

The structure of myosin in the actin-bound state is central to our understanding of contractility. However, high-resolution atomic structures of myosin subfragments have only been obtained for the detached states. The myosin head consists of a globular motor domain that contains the nucleotide-binding pocket, and a light chain binding domain that is believed to act as a lever arm during the contractile cycle. A prominent cleft divides the actin-binding face of the motor domain into an “upper” and a “lower” 50-kDa region. A small domain, the converter, is thought to amplify small conformational changes in the nucleotide-binding pocket into a large movement of the light chain binding domain. The available crystal structures fall into two major conformational classes: in the first, presumed to represent a post power stroke, the cleft that connects the actin-binding surface to the nucleotide-binding site is “open,” and the light chain domain is positioned in a “downward” orientation. In the second class, this cleft is “partially closed,” and the lever arm is oriented in an “upward” position. The latter structure, obtained in the presence of ADP-P_i analogs, is generally thought to represent a transition state for ATP hydrolysis. The existence of these two conformations and their coupling to particular nucleotide states seems to be isoform-independent in the detached states (1–9). A “power stroke,” the force-producing step of the contractile cycle, is believed to occur during isomerization from the weak-binding states to the strong-binding actomyosin states (reviewed in ref. 10). A power stroke of ≈10 nm can be generated by superimposing the myosin heads of the two detached conformational classes and assuming a swing of the lever arm from the upward to the downward position. Electron cryomicroscopy (cryoEM) of actin-bound smooth-muscle S1 provided the first direct demonstration of a lever-arm swing (11).

We have used cryoEM to obtain structural insights into the actomyosin complex. By computer-based fitting of the crystal structure of smooth-muscle myosin subfragment-1 (S1) into 3D reconstructions of decorated actin filaments, it has been possible to derive atomic models of actin-bound smooth-muscle S1 in the presence and absence of MgADP (12). A principal conclusion from this study was that the cleft that splits the 50-kDa region at the actin-binding interface is more closed in rigor than in any of the crystal structures. This closure is achieved by a movement of the upper 50-kDa region upon MgADP release. We have now extended these studies to actin decorated with skeletal muscle S1. By comparing the actin-bound states of a fast striated muscle myosin with those of the slower smooth-muscle myosin, we are able to generalize our conclusions regarding conformational changes in S1 upon binding to actin. A unique, isoform-independent actin-bound conformation of myosin can be characterized by a closed cleft, lever arm in a downward orientation, and stabilized surface loops. Mapping of structural and dynamic changes in the actomyosin complexes suggests a common pattern, in which a similar sequence of conformational changes leads to cleft closure in skeletal muscle myosin as well as in smooth-muscle myosin.

Methods

Protein Preparations.

Actin was prepared from chicken pectoralis acetone powder (13) and stored at 4°C as F-actin (10–15 mg/ml) in 5 mM KCl/5 mM imidazole, pH 7.5/2 mM MgCl₂/3 mM NaN₃. It was generally used within 2–3 weeks of preparation. F-actin was diluted to 0.025–0.03 mg/ml with 20 mM NaCl/5 mM NaP_i, pH 7.0/1 mM MgCl₂/0.1 mM EGTA/2 mM NaN₃ just before application to the grids. Skeletal myosin S1 was prepared by chymotryptic digestion of chicken pectoralis myosin, as described (14). It was used at a concentration of 0.5 mg/ml in 10 mM NaCl/10 mM imidazole, pH 7.0/1 mM MgCl₂/1 mM DTT. In the presence of 0.5 mM MgADP in the same buffer, a higher protein concentration of 2 mg/ml was used to compensate for the reduced binding affinity.

The protein preparation, microscopy, image analysis, and docking of the smooth-muscle myosin reconstructions used for comparison here are described (12).

Microscopy.

To prepare frozen-hydrated specimens for cryoEM, F-actin was applied to holey carbon grids, followed by incubation with myosin S1 in the corresponding buffer. The grids were blotted and plunged into liquid-nitrogen-cooled liquid ethane (15). Low-dose images were recorded with a Philips CM12 electron microscope (Philips Electronic Instruments, Mahwah, NJ) and a Gatan cryoholder (Gatan, Pleasanton, CA) at a nominal magnification of 60,000 (at 120 keV) and ≈1.5 μm defocus (electron dose ≈10e⁻/Ų).

Image Analysis.

Generation of 3D maps in real space, difference maps, and flexibility mapping followed the procedure described (12). Quality indicators are listed in Table 1. In short, all maps were reconstructed in real space. Twenty-three layer lines trimmed to a resolution of 21 Å were used in all of the reconstructions. Because this is within the first node of the contrast transfer function, no phase correction was necessary. The real space averaging provides a single averaged value and its associated variance for each voxel in the 3D map. The fact that we have access to the variances allows us to check the features in the averaged maps for statistical significance. The significance of the features in the 3D maps was assigned at a confidence level of 99.5% by a Student's t test procedure (16).

Table 1.

Quality indicators for skeletal actoS1 reconstructions

	Actin	ActoS1, rigor	ActoS1, ADP
No. of data sets	2	9	2
Minimal no. of subunits in data set	3249	3701	5690
Subunits per turn	2.1606 ± 0.0013	2.1601 ± 0.0021	2.1599 ± 0.0020
Average filament length, in subunits	162	168	167
Average of phase residual, °*	17	23	26
Average of up/down residuals, °†	25	44	40
Resolution cutoff, Ň	21	21	21

^*The mean phase residuals for this study are at the lower end of the range appearing in the literature (20–45°). The lower the phase residual, the higher the signal-to-noise ratio. 

^† The up-down difference in residual is a measure for the polarity of the filaments and should depend more on the structure than the signal-to-noise ratio. 

^‡ The layer line data sets were trimmed to a resolution of 21 Å. Because this value is within the first zero of the contrast transfer function, no phase correction was applied. 

Difference maps were calculated by subtracting the MgADP map from the rigor map. Before subtracting, the maps were aligned to each other by using a new hybrid real-space reciprocal-space alignment procedure that includes common features of the two maps in the alignment (17). Again, we tested the significance of each difference between corresponding pairs of real-space voxels by using a classical t test at a confidence level of 99.5%. No scaling of size was applied during the alignment or subtraction procedures as the microscopy, scanning, and image processing for all data sets was done under the exact same conditions. The pixel size was calibrated by using the known height of actin layer lines. Twenty individual difference maps were calculated by combining independent data sets (2 ADP and 10 rigor), thus allowing extensive crossvalidation of the difference-map features. Densities corresponding to a single myosin molecule were isolated by using an automatic segmentation approach (18).

The absolute value of individual differences (AVID) was used to map intrinsic structural flexibility of individual actomyosin units within the filaments (19). This procedure was applied to each individual data set from rigor, MgADP, and undecorated-actin. The procedure identifies regions within one 3D reconstruction that adopt multiple conformations or are partially occupied. In conjunction with difference mapping, this independent source of information can pinpoint changes in stability of domains within the complex. Two AVID maps each were calculated for undecorated actin and the ADP state, and ten maps were calculated for rigor. This procedure allowed crossvalidation of the map features.

Fitting Procedure and Discrepancy Mapping.

Atomic models of individual components of the skeletal actomyosin complex were fitted to the 3D reconstructions by using the program suite COAN, as described (12, 20). The components used for the final modeling were the myosin motor domain, the essential light chain (ELC), and a model for filamentous actin (21). Nine independent skeletal rigor maps and two ADP maps were used for crossvalidation for the docking as well as for the discrepancy mapping. In addition, nine different structures of the motor domain in the prehydrolysis open cleft state (see Table 2) and nine in the transition partially closed cleft state were docked independently into each of the maps, once with the converter present (complete MD) and once without (truncated MD). Thus, for example, the docking results for the motor domain into the rigor map of skeletal actomyosin are based on 324 independent docking calculations.

Table 2.

Docking procedure, real space correlation, %

Structure	Transition state*	Prehydrolysis state†	Final models
Truncated MD, rigor‡	88.03  ± 0.80	88.03  ± 0.80	89.05
Truncated MD, ADP‡	89.44  ± 0.57	88.87  ± 0.80	89.21
Complete MD, rigor§	88.00  ± 0.88	91.24  ± 0.43	91.55
Complete MD, ADP§	88.67  ± 0.62	91.63  ± 0.46	91.79
Complete S1, rigor¶	86.27  ± 1.41	91.73  ± 3.79	95.33
Complete S1, ADP¶	86.99  ± 0.41	92.84  ± 4.30	95.52
ELC, rigor‖		54.75  ± 0.97	61.12
ELC, ADP‖		50.40  ± 0.67	54.50

^*Five myosin II structures in the crystallographic transition state (PDB codes 1BR1, 1BR2, 1BR4, 1MND, and 1VOM) and four myosin I structures in the transition state (four monomers from PDB entry 1LKX) were independently fitted into each individual map. The results were used to calculate the standard deviations of the correlation values. 

^† Nine myosin structures in the crystallographic prehydrolysis state (PDB codes 1MMA, 1MMN, 1MMD, 1MNE, 2MYS, 1KK7, 1LVK, and 1G8X) were independently fitted into each individual map. The results were used to calculate the standard deviations. 

^‡ Truncated motor MD denotes myosin residues up to residue 710 (in chicken skeletal notation). For both ADP and rigor, the truncated MD of the transition state and the prehydrolysis state fit equally well. 

^§ Complete MD denotes myosin residues up to residue 781 (in chicken skeletal notation). The fit of the prehydrolysis conformations is significantly better than those of the transition state conformations in both rigor and ADP. This indicates that the converter is closer to the prehydrolysis conformation in both actin-bound states. 

^¶ Complete S1 denotes myosin residues up to residue 803 (in chicken skeletal notation) and the ELC. Three prehydrolysis conformation structures were used (2MYS, 1KK7, and 1DFK). In addition, we used the two smooth muscle models for the actin-bound states in the prehydrolysis category. The large standard deviation is due to the wide variety of lever arm positions within this group. Two transition-state structures (1BR1 and 1BR4) were used. 

^‖ ELC denotes the ELC and myosin residues 782–803 (in chicken skeletal notation). Six ELC domains (from 1BR1, 1BR4, 2MYS, 1KK7, 1WDC, and 1SCM) were used independently for fitting and crossvalidation to verify the lack of differences in position between skeletal ADP and rigor states. No distinction between prehydrolysis and transition states was made in this case. The correlation values are relatively low because the complete discrepancy maps were used for docking, and there are several peaks (see Fig. 3) that are not accounted for by the ELC. 

Special care was taken to crossvalidate the docking of the ELC domain to verify the lack of differences in the position of the domain in the skeletal rigor and ADP states. We used 6 different ELC structures (see Table 2) to dock into 11 (2 ADP, 9 rigor) discrepancy maps calculated from the original reconstruction by subtracting the contributions of the docked skeletal motor domains. The original skeletal structure (2) was used to build the final model. There was no experimental evidence for a movement of the converter domain from that of the skeletal S1 crystal structure. Fitting accuracy was estimated from solution sets as described (12), using a confidence level of 99.99%. Density maps from the final fitted models were calculated, filtered to 21 Å, and scaled to the experimental reconstructions. These maps were then subtracted from the experimental maps to generate discrepancy maps. The significance of changes in position of the structural components was calculated by using Student's t test between the respective solution sets with P = 0.005.

Molecular Graphics.

Figures were created with CONSCRIPT (22), MOLSCRIPT (23), and rendered with RASTER3D (24).

Results

Atomic Models for Skeletal actoS1.

Three-dimensional maps were generated from actin filaments decorated with chicken skeletal muscle S1 (this proteolytic subfragment contains the myosin heavy chain complexed with the ELC) in the presence and absence of MgADP (Fig. 1 a and b). The density corresponding to S1 was obtained by subtracting the 3D reconstruction of F-actin from the 3D reconstruction of actin decorated with S1 in real space (25). A difference map between the skeletal ADP and rigor reconstructions did not show any significant differences at a confidence level of 99.5%. The crystal structure of nucleotide-free skeletal S1 (2) was fitted into the density of the actin-bound S1 in the ADP and rigor states by the quantitative fitting procedures described (20). The ELC region of the crystal structure did not fit well into the cryoEM structure and had to be fitted independently as a rigid body. The analysis of the fitting statistics (see Table 3) indicates an accuracy of ≈2.8 Å for the positioning of the motor domain, which is 7.5-fold better than the resolution of the 3D maps (21 Å).

Figure 1.

(a and b) Three-dimensional reconstructions of actin filaments decorated with skeletal S1. The final skeletal S1 models and the corresponding 3D reconstructions of actomyosin in the presence of MgADP (a) and in the absence of nucleotide (b) are shown. The contour levels were chosen to enclose only significant density at a 99.5% confidence level. Two neighboring molecules are shown: motor domain in green and ELC in red. All presentations are with the sarcomeric M line at the top of the figure. Note that there is no detectable rearrangement. (c and d) Comparison of ELC positions of actin-bound smooth and skeletal S1. Molecular surface representations of the fitted models, calculated at 15-Å resolution, are shown. The view in d is rotated by 90° in respect to the view in c. There is a pronounced difference in the position of the smooth-muscle ELC in the presence of ADP (magenta) and the absence of nucleotide (green). There is no such difference for the skeletal ELC (light blue). Only the motor domain of skeletal myosin is shown (light gray). The smooth-muscle motor domain rotates by ≈9° around the axis shown in red (solid gray lines in d). There is no significant rearrangement of the skeletal motor domain that is positioned about halfway between the two smooth-muscle actin-bound structures (dashed gray line in d).

Table 3.

Fitting accuracy of atomic models, Å

Model	Actin	Motor domain	ELC
Rigor	1.8	2.8	5.3
ADP	1.8	2.9	5.2

The accuracy was estimated from the rms deviation of the solution sets. These values were crossvalidated by using the independent data sets. 

The resulting atomic models for skeletal S1 in the MgADP state and in rigor are very similar, with neither the converter region nor the ELC showing a significant change in position upon nucleotide release (Fig. 1 c and d). The docking of the skeletal head on actin lies about halfway between the docking for smooth S1 in the two states. The lack of changes is clearly different from the large movement of the light chain domain and rotation of the motor domain at the actin interface, observed for smooth-muscle S1 upon MgADP release (12).

Actin Binding Changes the Conformation of the Myosin Motor Domain.

The differences between the final atomic models derived from the fitting procedures and the 3D reconstructions can be analyzed by subtracting the density calculated for the atomic models from the density of the 3D reconstructions. The resulting discrepancy maps show regions in the models which do not correspond well to the observed cryoEM structures. Discrepancies can be caused by differences in stability or conformational changes that exist between the crystal structure and the actoS1 complex. The surface loop 1 (at the 25/50-kDa junction near the nucleotide-binding pocket) and loop 2 (at the 20/50-kDa junction and the actin interface) are flexible in the detached state, as these regions are only partially resolved in the crystal structure. Upon binding to actin, the loops presumably become stabilized, which results in additional density in the actin-bound S1 structures (Fig. 2 a–d). The remarkable finding is that differences in density occur, for the most part, in the same regions of the motor domain for both smooth and skeletal muscle S1, particularly in the rigor states. This finding provides strong support for the validity of the computer-based modular-fitting approach for structural studies (20).

Figure 2.

(a–d) Differences between the final fitted models and the corresponding reconstructions of smooth and skeletal actomyosin complexes (discrepancy maps). Overview of extra density originating from EM reconstructions of ADP (a) or rigor (b) of smooth-muscle actomyosin overlaid with the motor domain of the docked structures. Overview of extra density coming from EM reconstructions of ADP (c) or rigor (d) skeletal muscle actomyosin. Peaks occur close to loop 1 and loop 2 in all of the maps. The loop 2 peak is close to the resolved portion of the actin-binding loop (loop 2). The resolved ends of loop 1 can be seen close to the loop 1 peak. Note that the loop 1 peak is smaller in the ADP maps for both isoforms. A peak in the actin-binding cleft indicates a more closed conformation than that of the crystal structure used for discrepancy mapping. This peak occurs for all reconstructions except for smooth muscle in the presence of ADP. All cleft peaks shown here occur for all crystal structures regardless of their state of cleft closure (either partially closed or open). (e and f) Test of cleft-state detection procedure using calculated density maps. A 10-Å resolution density map representing a modeled skeletal motor domain with a closed actin-binding cleft (light gray transparent) is shown in e with the best fit of the original, open-cleft motor domain structure. There is no unambiguous indication for a closed cleft at this resolution. A discrepancy map (green) of a calculated closed-cleft density map at 25 Å and the corresponding best fit (gray molecular model) shows a clear peak in the cleft region (f), indicating that, at a resolution of 25 Å or better, the discrepancy mapping procedure is capable of detecting the cleft state.

Judging from crossvalidation studies, there seems to be a consistent change in shape and position of the loop 2 peak between ADP and rigor for both isoforms, suggesting a change in the mode of binding upon nucleotide release. The discrepancy peak close to the resolved ends of loop 1 in the crystal structure is noticeably smaller in the ADP map than in the rigor map for both isoforms, suggesting greater flexibility when nucleotide is present at the active site.

The most notable difference between the smooth and skeletal myosin isoforms is in the cleft region of the maps in the ADP state. The absence of a discrepancy peak in smooth S1 ADP (Fig. 2a) implies that it has a partially closed cleft, similar to the crystal structure used in the fitting, which has a transition-state analog (MgADP.AlF₄-) at the active site (6). The appearance of a discrepancy peak in the rigor state (Fig. 2b) implies that the cleft closes even further upon the loss of ADP. In contrast, skeletal muscle S1 has a discrepancy peak in ADP as well as in rigor (Fig. 2 c and d), irrespective of the crystal structure used in the fitting (see Materials and Methods), suggesting that the actin-binding cleft is tightly closed in both actin-bound states of skeletal myosin.

To test the cleft detection procedure we calculated a 10-Å resolution simulated cryoEM map by using a modeled skeletal myosin motor domain with a closed actin-binding cleft. The closure of the cleft was generated by continuing the rotation of the upper 50-kDa region that occurs between open and partially closed myosin structures (5). Docking the original crystal structure (2) with an open cleft into this simulated map gave a perfectly reasonable fit (Fig. 2e), indicating that even at 10-Å resolution, it is difficult to discriminate between an open or closed cleft structure by relying on a visual inspection of the docking solution alone. However, a distinct discrepancy peak appeared in the cleft region upon comparing the difference in density between the map (closed cleft) and the fitted original crystal structure (open cleft), even if the map was calculated at only 25-Å resolution (Fig. 2f). These calculations imply that discrepancy mapping is well suited to detect structural changes at moderate levels of resolutions, whereas visual inspection of the docking solution can be inconclusive, even at considerably higher resolution than that available for this study.

Structural Dynamics in Actomyosin.

Structural flexibility within individual actomyosin units can be detected in filaments by the AVID method (19). This procedure identifies regions within a 3D reconstruction that adopt multiple conformations and thus can provide information on the dynamics within the structure. An example of such a dynamic change is the relocation of an AVID peak close to the cleft in the smooth-muscle ADP state by a much larger AVID peak in the upper 50-kDa region in the rigor state (Fig. 3 a and b). This change is consistent with stabilization of the cleft upon closure and structural rearrangements in the upper 50-kDa region. It is striking that the same general regions of flexibility (AVID peaks) are seen in the two states of the skeletal myosin isoform as in the smooth myosin in rigor (compare Fig. 3 c and d to b). The common pattern, compatible with a movement of the upper 50-kDa region to close the cleft, strongly suggests that a similar sequence of conformational changes leads to cleft closure in skeletal muscle myosin as in smooth-muscle myosin (12). The variability in these particular elements, as well as in the SH1 helix, the cardiomyopathy loop, the lower 50-kDa region, and the ELC seems to be isoform-independent and may reflect their joint participation in communication pathways between the actin interface, the nucleotide-binding pocket, and the light chain binding domain.

Figure 3.

Depiction of the structural flexibility patterns in the ADP state (a and c) and the rigor state (b and d) of 3D reconstructions of smooth and skeletal actomyosin complexes. One peak (D1) is associated with actin (see Fig. 4) and is only shown for reference. The distance between the D1 peak border and the first residue of S1 is >8 Å. For the myosin region, areas of high structural flexibility (AVID peaks) are overlaid with the final docked models (motor domain in gray, ELC in light blue). A number of areas overlap for both the skeletal states (ADP and rigor) and the smooth-muscle rigor state. These peaks are located in the upper 50-kDa region, near the reactive SH groups (SH), close to the ELC, and close to the cardiomyopathy loop (CL). There are peaks in the vicinity of the N terminus (N) in all states with a marked shift in the smooth-muscle ADP state. For the smooth-muscle states, there is an extra peak close to the nucleotide-binding pocket. Two extra peaks appear in the smooth-muscle ADP state, one in the lower 50-kDa domain and one in the upper 50-kDa domain close to the cleft (marked cleft). In summary, the structural flexibility pattern is very similar for the two skeletal-muscle and the smooth-muscle rigor states. This pattern is markedly different from the pattern in the smooth-muscle ADP state.

Domain movements at the interface are not restricted to myosin, but also occur in the actin filament (Fig. 4). In general, pure actin filaments show higher structural flexibility than decorated actin (data not shown). This result suggests that strong myosin binding stabilizes F-actin, consistent with other biophysical studies (reviewed in ref. 26). However, a region of significant structural flexibility remains close to the N terminus of actin in subdomain 1 (D1) in all of the actomyosin reconstructions. This result is supported by a recent NMR study that also shows flexibility at actin's N terminus in both strongly bound states of skeletal actomyosin (27).

Figure 4.

F-actin view of structural flexibility and discrepancy mapping at the actomyosin interface for smooth (a and b) and skeletal (c and d) strongly bound states. The actin subdomains are labeled in b. The N and C termini are labeled in a. The pointed end of the filament is at the top of the figure (as in Fig. 1). Two actin monomers of the filament are shown (upper actin, light pink; lower actin, light cyan). Peaks that are not in proximity to the interface are omitted for clarity. The labeling and color scheme correspond to Figs. 2 and 3. There are corresponding discrepancy peaks (D2) for all four reconstructions between subdomain 2 of the lower actin and the C terminus of the upper actin. There is a corresponding AVID peak in all four reconstructions (D1). This peak is shown for the lower actin only. For both isoforms, there is a change in size and position of the loop 2 discrepancy peak upon ADP release that suggests a change in the mode of myosin binding in this region. The structural flexibility associated with the cardiomyopathy loop of myosin (CL) may also be associated with loop 2.

There is a significant discrepancy peak between subdomain 2 (D2) of the lower actin and the C terminus of the upper actin in both the skeletal and smooth-muscle isoforms. This discrepancy peak most likely represents a shift in actin density upon myosin binding, consistent with spectroscopic experiments that indicate a movement of subdomain 2 in actin upon strong myosin binding (28, 29). It should be noted that this peak is in the vicinity of the secondary actin-binding loop of the lower 50-kDa region of myosin and may be affected by changes in that region as well.

Discussion

Earlier cryoEM studies on the structure of smooth-muscle myosin S1 were in excellent agreement on a large change in the position of the light-chain domain upon release of MgADP (11, 12). In addition, we concluded that the large cleft that divides the myosin head into an upper and a lower 50-kDa region is more tightly closed than in any of the available crystal structures when myosin is bound to actin in the absence of nucleotide (12). This conclusion was obtained by new methodologies for combining atomic-resolution structures with 3D image reconstructions at lower resolution. Here, we use this approach further to determine the structure of actin-bound skeletal muscle myosin S1.

We find that many features of skeletal S1 are remarkably similar to those of smooth-muscle S1 in the rigor state: skeletal S1 has a closed cleft, the surface loops 1 and 2 which are disordered in the crystal structure become ordered when S1 is bound to actin, and the regions of flexibility in the skeletal structure are sufficiently similar to those in smooth-muscle myosin to suggest that the motor domain (with the exception of the converter) is in a similar conformation for both isoforms. The light-chain/converter position is about half way in between the positions for smooth-muscle myosin in rigor and in ADP. Using the rotation axis previously determined for the smooth-muscle motor domain, the skeletal motor domain also seems to be about half way in between the smooth-muscle myosin positions. This result may indicate that the position of the lever-arm/converter system and the rotation of the motor domain on the actin filament are coupled in the actin-bound states. The coupling between cleft state, nucleotide, and the position of the lever arm, which was observed for the detached myosin structures, is not observed in actin-bound states.

In the presence of ADP, however, there is a distinct difference between smooth and skeletal actomyosin. Whereas the cleft is only partially closed for the smooth-muscle ADP state, the cleft is tightly closed for the skeletal ADP state. The lever arm moves further downward during the transition from ADP to rigor in smooth-muscle myosin (11, 12), but does not produce any significant movement for skeletal muscle myosin (this study and ref. 30). From a kinetic analysis (31) and mechanical measurements of smooth-muscle strips (32), it was argued that ADP release would probably not affect force generation, but a strain-dependent ADP release mechanism would benefit a myosin designed for high forces and slow contractions, as in smooth muscle. Slow, skeletal fibers are also adapted to maintenance of tension with less energy consumption than fast, skeletal fibers. It will be interesting to determine whether there are similar structural changes in actin-bound β-cardiac (slow) myosin upon the release of ADP. If such changes are found, they could well be generally responsible for increasing the dwell times on actin and, thereby, slowing the velocity of shortening.

The first crystal structure published for chicken skeletal S1 (2) had an open cleft structure with a sulfate at the active site. Because the lever arm was in a downward position with an orientation similar to that observed in 3D reconstructions of decorated actin (33), this conformation was considered to represent the strong binding rigor state (34). A subsequent study on Dictyostelium myosin showed that MgATP could bind to myosin with an open cleft, but could not be hydrolyzed (8). This conformation is, therefore, more likely a weak-binding state that occurs just after ATP dissociates the actomyosin complex. An uncoupled conformation has also been demonstrated in the crystal structure of scallop myosin (but in no other isoform), in which the converter/lever arm is uncoupled from the motor domain by the unwinding of the SH1 helix (35). It has been speculated that this conformation follows dissociation but precedes the transition state, in which the lever arm is constrained in an upward position, and the cleft must close partially to allow hydrolysis to occur. After rebinding to actin and releasing products, the true rigor conformation is formed, which, as shown here, seems to be isoform-independent and consists of a myosin with a tightly closed cleft and a lever arm at the extreme end of its swing.

Abbreviations

cryoEM

electron cryomicroscopy

AVID

absolute value of individual differences

ELC

essential light chain