Here, we describe, at low resolution (≈4 Å), two structural states previously not seen in intact scallop S1. These results, together with the unusual conformation we have recently reported (6), demonstrate how the intact head of a single myosin isoform undergoes critical conformational changes that correspond to three distinct steps in the contractile cycle. One of the states we describe corresponds to a nucleotide-free structure with the lever arm in the postpower stroke (≈45°) orientation. New crystallization results do not support earlier interpretations (7) that this structure might correspond to a stable ATP state. We find that, for each structural state, the position of the subdomains and the conformation of the joints within the MD (see Fig. 1 and ref. 6) are relatively invariant between the available crystal structures of different myosin isoforms. In contrast, the position of the lever arm varies significantly when structures of different myosin isoforms are compared in both the nucleotide-free and transition states. This finding shows that the position of the lever arm can be modulated by various interactions with little change in the MD conformation. This study supports the conclusion that the distinct structural states observed in an intact, unmodified myosin head depend on the specific nucleotide or nucleotide analog that is bound. Moreover, we have identified a short pliant junction at the beginning of the lever-arm helix that may play a critical role in the mechanical properties of the myosin head.

Figure 1 Schematic drawing of the subdomains within the myosin head in the nucleotide-free scallop S1 structure. The four subdomains of the MD are labeled, as are three structural elements (joints) that articulate movements within the head: switch II (residues (more ...)

Crystallization and Data Collection. S1 from scallop (Argopecten irradians) striated muscle myosin was obtained as described (6). Crystals of nucleotide-free S1 (near-rigor state) grew in a day at 4°C in hanging drops containing equal amounts of reservoir solution [50 mM cacodylate, pH 7.0/50 mM (NH₄)₂SO₄/10 mM MgCl₂/15% (vol/vol) glycerol/≈7% (vol/vol) polyethylene glycol 8000] and protein solution (20 mg/ml). These crystals belong to space group P2₁ (a = 86.5 Å, b = 52.1 Å, c = 164.6 Å, β = 100.3°) with one molecule per asymmetric unit. Crystals of S1 complexed with MgADPVO₄ (transition state) were obtained at 4°C by microseeding sitting drops containing equal amounts of reservoir solution [7% (vol/vol) polyethylene glycol 8000/50 mM Mes, pH 6.0/3 mM MgCl₂] and protein solution (20 mg/ml of S1 and 0.5 mM MgADPVO₄). The crystals belong to space group P2₁ (a = 53.3 Å, b = 243.5 Å, c = 124.6 Å, β = 100.2°) with two molecules per asymmetric unit. Crystals of the third or detached state, as discussed in ref. 6, grew under different conditions, by microseeding sitting drops containing equal amounts of reservoir solution [6% (vol/vol) polyethylene glycol 20,000/100 mM Tris, pH 8.0/12% (vol/vol) glycerol] and protein solution (20 mg/ml S1 complexed with 4 mM MgADP, 4 mM MgADPBeF_x, or 4 mM MgAMPPNP (adenylyl-imidodiphosphate).

By using glycerol as a cryoprotectant, a data set was collected at 100 K (Cornell High Energy Synchrotron Source beamline A1; λ = 0.90 Å) to 4.2-Å resolution for a nucleotide-free crystal and to 3.8-Å resolution for two S1-VO₄ crystals. The data sets were indexed and scaled with denzo and scalepack (9). For the nucleotide-free S1 crystal, scaling of 139,608 observations resulted in a 97.9% complete data set of 10,676 reflections with an R_merge of 6.3%. For the transition-state crystals, scaling of 426,824 observations resulted in a 95.5% complete data set of 29,173 reflections with an R_merge of 9.8%. The averaged I/σ ratio is 27 in the case of the nucleotide-free crystal for reflections between 30 and 4.2 Å and 20.4 in the case of the transition-state crystals for reflections between 40 and 3.8 Å.

Structure Determination and Refinement. The molecular replacement solution was found with the program AMoRe (10). The search models used were two separate components of the crystallographic model of scallop S1-ADP (PDB-code 1B7T): the lever arm and the MD; in the latter component, 25 residues of the relay and 83 residues of the SH1 helix region and the converter were excluded. Side chains were removed from the models because of the limited resolution of the data. The fiting routine of AMoRe was used for rigid body refinement of the model, which was divided into 12 parts for the nucleotide-free structure and 16 for the transition-state structure. After convergence, the subdomains of the MD adopt positions similar to those observed in other structures (such as chicken skeletal S1 for the nucleotide-free scallop S1 and smooth muscle MDE-AlF₄⁻ for scallop S1-VO₄). The electron density map calculated by using phases from these models revealed clear density that allowed us to position both the converter and the relay and indicated that the SH1 helix was intact in both structures. Model building was performed with the program o (11). Near the nucleotide-binding site, the density was of poor quality in the case of the nucleotide-free structure, and 26 residues were removed, including switch I. In contrast, the density was of good quality for this region in the case of the S1-VO₄ structure, including that of the MgADPVO₄ ligand. The final model for the nucleotide-free structure has an R_factor of 41.5% at 4.2-Å resolution; that of scallop S1-VO₄ has an R_factor of 39.4% at 3.8-Å resolution. (Note that positional refinement cannot be carried out because of the limited resolution of the data.) The quality of the electron density map, which allowed us to visualize the position of the relay, the SH1 helix, and the converter, validates these low-resolution models. Note also that starting from various atomic models, in the detached, transition-state, or near-rigor conformations, the rigid body refinement always converged for new positions of the so-called “N-terminal,” “upper,” and “lower” 50-kDa subdomains of the MD (see Fig. 1) that correspond to a near-rigor conformation in the case of data from the nucleotide-free crystal and a transition-state conformation in the case of S1-VO₄ crystals. Thus, the rms deviation for about 500 Cα positions of equivalent residues of these three subdomains is only 0.97 Å between nucleotide-free chicken and scallop S1 and 1.0 Å between smooth MDE-AlF₄⁻ and scallop S1-VO₄. The rms deviation for these same atomic positions between scallop S1-ADP (used as the first model to solve the structure by molecular replacement) and nucleotide-free S1 is 1.72 Å and is equal to 2.64 Å when scallop S1-VO₄ and scallop S1-ADP (detached state) are compared.

Other Crystallization Experiments. Using each of the three crystallization conditions discussed above (producing the near-rigor, detached, and transition-state structures), we examined a number of other nucleotides or their analogs to determine which crystals they produced (if any). By using the near-rigor conditions, crystals similar to the near-rigor state could be obtained with MgADP. However, S1 complexed with either ATP analogs (MgAMPPNP, MgADPBeF_x) or ADPPi analogs (MgADPVO₄, MgADPAlF₄⁻) did not produce crystals. In contrast, by using the appropriate crystallization conditions, transition-state crystals were obtained for S1 complexed with MgADPAlF₄⁻. However, no crystals grew without nucleotide present or when S1 was complexed with MgADP or ATP analogs. By using detached-state conditions, crystals of S1 complexed with ATP analogs were obtained, but no crystals were obtained with scallop S1 when no nucleotide was bound or when the S1 was complexed with transition-state analogs.

Correspondence Between the Conformation of the Myosin Head and the Nucleotide Content of Its Active Site. We have shown that scallop S1 crystallizes in three different crystal forms depending on the content of the nucleotide-binding pocket. Different ionic and pH conditions (see Methods) are required for the growth of each of these crystal forms, probably reflecting differences in interactions between neighboring molecules. We have solved these stuctures, which correspond to three distinct conformational states of the motor. In this respect, scallop is a privileged system, because it allows one to test which conformation exists with a particular nucleotide bound to S1. The first structure (called near-rigor by us) was that of nucleotide-free scallop S1, determined to 4.2-Å resolution. This structure could also be obtained in the presence of MgADP, as reported for Dictyostelium myosin II (4). The second structure is that of S1 complexed with MgADPVO4 (S1-VO₄), determined to a resolution of 3.8 Å, and corresponds to a transition-state structure, required for hydrolysis of ATP (2). Crystals of this form could be obtained only when scallop S1 was complexed with ADPPi analogs. The third structure, obtained under different conditions (see Methods and ref. 6), is that of S1 complexed with MgADP (S1-ADP), determined to 2.5-Å resolution (6), which we believe corresponds to a new detached ATP state of the actomyosin cycle. Crystals with unit cell parameters similar to this form also grew when S1 was complexed with ATP analogs but not in the absence of nucleotide or when S1 was complexed with ADPPi analogs. Using scallop S1, we have thus shown the link between a particular nucleotide (or its absence) in the active site and the conformational state of an intact unmodified myosin head (i.e., with a complete lever arm) from a single isoform. We must emphasize, however, that these crystallographic results, obtained in the absence of actin, including the case of the near-rigor nucleotide-free S1, do not define the conformation of the strong binding states of S1.

A Simple Description of the MD Arrangement. As proposed in Houdusse et al. (6), the motor of the S1 structure is most simply described as being made up of four major subdomains (see Fig. 1) that are linked by three single-stranded joints. In the transition between states, coordinated changes in the conformation of the joints are associated with rigid body movements of the four subdomains. The region of the motor that shows the greatest positional change in the three conformations is one of these subdomains, which is called the “converter” (12), that immediately adjoins the lever arm. The other subdomains show only small changes in position in different states. The movements of the converter are directly coupled to and amplified by the extended lever arm. One of the joints, the “relay”, has strong connections to the converter and controls its movement. Different orientations of the converter/relay module (produced by conformational changes at both ends of the relay) characterize the different structural states. A 15-residue-long region before the converter constitutes a second joint—the so-called “SH1 helix region”—whose conformation depends on the different orientations of the converter. Nucleotide binding promotes the unwinding of the helix, uncoupling the converter from the rest of the MD. Switch II is a third joint, near the γ-phosphate pocket and positioned at the bottom of the actin-binding cleft. This joint plays a critical role in the communication between the nucleotide-binding site and the actin-binding interface. In the transition between states, the three joints act together to control the overall organization of the myosin head. This arrangement allows direct communication between distant critical regions of the head, such as the nucleotide-binding site, the actin-binding interface, and the lever arm.

The Scallop MD in the Three States. Despite the low resolution of the two new structures of scallop S1 (Fig. 2A), which precludes identification of side-chain interactions, the quality of the electron density allows us to establish the relative positions of the subdomains within the MD as well as the overall conformation of the joints within the motor (Fig. 2 B and C). These features allow us to assign conformational states without ambiguity. The nucleotide-free structure corresponds to a near-rigor conformation previously observed at higher resolution in the methylated nucleotide-free chicken skeletal S1 structure (1) and in the truncated MD of Dictyostelium myosin II complexed with ATP analogs (2, 4). Note, however, that in contrast to the results of Rayment and colleagues (refs. 2 and 4; see also ref. 7), scallop S1 complexed with ATP analogs did not crystallize in near-rigor conditions (see Methods), suggesting that this near-rigor conformation is not compatible with a stable ATP state. The scallop S1-VO₄ structure corresponds to a transition-state conformation previously described at higher resolution in the structures of Dictyostelium myosin II MD complexed with ADPPi analogs (2, 3) and smooth muscle MD and MDE bound to MgADPAlF₄⁻ (5).

Figure 2 (A) Ribbon diagrams of the nucleotide-free scallop S1 structure (Lower) and of scallop S1-VO4 (Upper) oriented such that the lower 50-kDa subdomains of these two structures superimpose. An arrow indicates the approximate direction of the actin filament (more ...)

There are striking differences that distinguish the three structural states. In particular, the three states differ greatly in the position of the converter (and thus the light-chain binding domain or lever arm) as well as in the conformation of the actin-binding interface. The converter rotates about 65° between the transition and nucleotide-free states of scallop S1 and leads to a movement of the lever arm between these states that has a very small azimuthal component (see Fig. 2C). In both these structures, the SH1 helix is intact, and the four subdomains are closely linked. By contrast, the detached ATP conformation (seen in scallop S1-ADP) is characterized by subdomains that are loosely coupled, because the SH1 helix is unwound (6). Correspondingly, the position of the converter/relay module is not constrained and various orientations of this module compared with the other subdomains of the MD could lead to a number of possible positions of the lever arm relative to the MD. In the scallop S1-ADP crystal structure, for example, the converter has rotated by about 30° from the position seen in the near-rigor structure, and the lever arm is almost parallel to the actin filament axis (Fig. 3). The changes in switch II—which affect only five residues—are surprisingly small considering that their structural effects are very large. Thus, in both the near-rigor and detached conformations, in which the 50-kDa cleft is open, this joint seems to be very similar. In the transition-state conformation, however, switch II has moved toward the nucleotide-binding site, and thus closes the bottom of the 50-kDa cleft, preventing phosphate release (2). These relatively invariant features of the MD in myosins from different isoforms allow one to classify structures in different conformational states.

Figure 3 Ribbon diagrams of the nucleotide-free scallop S1 structure in near-rigor, transition, and detached states, oriented such that the lower 50-kDa subdomains of these three structures superimpose. An arrow indicates the approximate direction of the actin (more ...)

A Pliant Region at the Junction Between the MD and Lever Arm. In contrast to the strong conservation of the features in the MD, the exact position of the lever arm can vary greatly among structures in the same state but from different isoforms and from different constructs. In the near-rigor conformation, the orientation of the lever arm of scallop and chicken skeletal S1 differ by 40°, taking into account both the azimuthal and axial components (see Fig. 2 B and C). Because of the axial component (about 25°), the scallop lever arm is positioned closer to the actin filament than that of chicken skeletal S1 (see Fig. 2B). The difference in the lever-arm position is even more striking when one compares scallop S1 and chicken smooth MDE in the transition-state conformation. The orientations of their lever arm differ by about 65°, with a major component perpendicular to the actin filament axis (see Fig. 2 B and C). These differences arise because of two structural features: a minor one is the small difference (about 12°) in the orientation of the converter/relay module, but the major one is the presence of a relatively small, unstabilized, and hence pliant region (residues Asp-775 to Lys-780, scallop sequence) in the heavy-chain helix at the junction between the converter and the lever arm in all three isoforms.

In the three conformational states of scallop S1, the heavy-chain helix is straight at this point, and the converter/ELC interactions seem to be maintained. Note that in the three scallop crystal forms, the lever arm does not interact strongly with neighboring molecules. In the head of this myosin isoform, the converter thus moves the lever arm without the pliant region flexing when other forces are not applied to the lever arm. In the other two myosin isoforms, the heavy-chain helix is not straight at the pliant region, but because these myosin heads have been observed in only one conformational state, we do not yet know what bends might occur in the other states. In fact, the large heavy-chain bend at the pliant region observed in the crystal structure of the transition state of chicken smooth MDE (5) does not seem to be present in the lower-resolution electron-microscope images of the postpower stroke states of this myosin bound to actin (16). Note, however, that in all cases, the converter/ELC interactions (which affect the degree of bending at the pliant region) are weak and differ in the states observed in these three myosin isoforms (6). The key point illustrated by these differences is that the position of the lever arm may be modulated by factors other than the precise position of the converter. The design of this region of the myosin head could thus play a key role in the elastic properties of the crossbridge (17).

The Three Conformational States and the Actomyosin Cycle. The three crystal structures described herein for scallop myosin are all observed in the absence of actin, and none can represent a strong actin-binding state of the myosin motor. The transition-state conformation, in which the lever arm is primed, corresponds to the prepower stroke state that binds actin weakly, before phosphate release. And, as described in ref. 6, we believe that the conformation of scallop S1 complexed with MgADP (in which the SH1 helix is melted) corresponds to a weakly bound, long-lived ATP state of myosin, which promotes the dissociation of myosin from actin.

The general orientation of the lever arm in the nucleotide-free structures of both chicken and scallop S1 that corresponds to the near-rigor conformation resembles that found in three-dimensional reconstructions of S1-actin complexes (13). This conformation cannot, however, correspond to a strong actin-binding state, because it does not account for actin-induced changes in S1. Kinetic studies show that binding of a nucleotide-free head to F actin is a two-step process that includes a temperature-dependent conformational change (14, 15, 21). Note that whatever structural rearrangements occur on the strong binding of actin to S1, they should greatly reduce the affinity of the myosin head for a nucleotide. With our current understanding of the dynamics in the myosin head, we expect that the actin-binding cleft and the nucleotide-binding pocket might differ between the true actin-bound rigor state and the non-actin-bound near-rigor state, but the conformation of the relay, the SH1 helix, the converter, and the lever arm would be similar. Actin is thus likely to induce subdomain movements that have not yet been described in the three known myosin states, such as movements of the upper 50-kDa subdomain relative to the N-terminal subdomain or movements within the upper 50-kDa subdomain. In our view, the near-rigor conformation might correspond to a transient state of the motor that occurs initially on ATP binding to acto-S1. Our studies on scallop myosin suggest, however, that this state is likely to be followed by a longer-lived ATP state in which the SH1 helix is unwound and the ATP is trapped (6).

The near-rigor conformation is the crystallographic structure that most closely resembles a strong actin-binding state. Thus, the position of the ELC in the nucleotide-free scallop structure seems to be similar to that found in electron-microscope images of actin decorated with nucleotide-free smooth myosin (16). In this view, the transition between the prepower stroke and the near-rigor conformations would correspond closely to the power stroke of the actomyosin cycle. In the case of scallop S1, the step size is about 120 Å, with a very small azimuthal component (see Fig. 2). (Note that this estimate is made by using the same isoform for the two states.) The flexible joint at the pliant region weakens the coupling between the torque generated by the MD and the motion of the lever arm.

In summary, our results indicate that the most stable conformations of scallop S1 seem to be the near-rigor state in the absence of nucleotide, and the transition state when ADPPi analogs are bound. In the absence of actin, ADP can induce both the near-rigor and the detached conformations, and when ATP analogs are bound, the detached state is the most favored conformation. Thus, scallop myosin is an exceptional system, allowing us to detect the influence of the content of the nucleotide-binding site on the conformation of the myosin head in the absence of actin. The current study also reveals that conformations probably common to many isoforms may be experimentally accessible (i.e., by crystallography) only in certain isoforms and under certain conditions. But it is also plain that certain isoforms may exist in specialized conformations not observed for other isoforms, because of structural features related to specific functions. A good example here is ADP release in vertebrate smooth muscle S1 (16) and some unconventional myosins (18) that could lead to the production of movement of the lever arm in two distinct steps (19). Similarly, the study of unconventional myosins, such as myosin VI, should reveal whether the conformational changes observed in the MD of myosin II characterize the whole superfamily and how the insertion of ≈50 residues in this motor allow it to move with reverse polarity (20). Thus, a thorough study of a variety of myosin isoforms is a productive approach to visualizing the multiple steps in the actomyosin cycle.

We thank Hugh Huxley for comments on the manuscript; Yale Goldman, Emil Reisler, and Daniel Himmel for discussions; Roberto Dominguez, Niels Volkmann, and the staff at the Cornell High Energy Synchrotron Source for help with data collection; Carol Palmer and Andrew Brilliant for assistance with photography; and Linda Lynch for devotion to the manuscript. This work was supported by grants from the National Institutes of Health and the Muscular Dystrophy Association (to C.C.), a grant from the National Institutes of Health (to A.G.S.-G.), as well as postdoctoral fellowships from the Human Frontier Sciences Program Organization and the Association Française Contre les Myopathies (to A.H.).