Cryo-atomic Force Microscopy of Unphosphorylated and Thiophosphorylated Single Smooth Muscle Myosin Molecules*

The purpose of this study was to determine whether steric blockage of one head by the second head of native two-headed myosin was responsible for the inactivity of nonphosphorylated two-headed myosin compared with the high activity of single-headed myosin, as suggested on the basis of electron microscopy of two-dimensional crystals of heavy meromyosin (Wendt, T., Taylor, D., Messier, T., Trybus, K. M., and Taylor, K. A. (1999) J. Cell Biol. 147, 1385–1390; and Wendt, T., Taylor, D., Trybus, K. M., and Taylor, K. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 4361–4366). Our earlier cryo-atomic force microscopy (cryo-AFM) (Zhang, Y., Shao, Z., Somlyo, A. P., and Somlyo, A. V. (1997) Biophys. J. 72, 1308–1318) indicates that thiophosphorylation of the regulatory light chain increases the separation of the two heads of a single myosin molecule, but the thermodynamic probability of steric hindrance by strong binding between the two heads was not determined. We now report this probability determined by cryo-AFM of single whole myosin molecules shown to have normal low ATPase activity (0.007 s–1). We found that the thermodynamic probability of the relative head positions of nonphosphorylated myosin was approximately equal between separated heads as compared with closely apposed heads (energy difference of 0.24 kT (where k is a Boltzman constant and T is the absolute temperature)), and thiophosphorylation increased the number of molecules having separated heads (energy advantage of –1.2 kT (where k is a Boltzman constant and I is the absolute temperature)). Our results do not support the suggestion that strong binding of one head to the other stabilizes the blocked conformation against thermal fluctuations resulting in steric blockage that can account for the low activity of nonphosphorylated two-headed myosin.

One of the most fundamental, but not yet understood, questions about smooth muscle and nonmuscle myosin II is how phosphorylation of the regulatory light chain (RLC), 1 located C-terminally in the myosin head, activates the N-terminal motor domain at a nearly 15-nm distance; or conversely, how dephosphorylated RLCs maintain myosin in the "off state." Myosin II is composed of two heavy chains each having at its N terminus a catalytic (motor) and a regulatory domain bound with essential light chain (LC 17 ) and RLC. Whereas the hexamer containing unphosphorylated RLCs is inactive and requires RLC phosphorylation for ATPase activity, a single head (S1) (4) or a construct of a single head with the heavy chain tail and two associated light chains is active without RLC phosphorylation (e.g. see Refs. [5][6][7]. The structural basis of the inhibitory mechanism has been difficult to decipher partly because the high resolution crystal structure of the smooth muscle myosin S1 fragment does not include the regulatory light chain (8); the scallop muscle protein structure contains both essential and regulatory light chains, but the protein is regulated by calcium binding rather than by phosphorylation (9,10); and the residue (Ser-13) equivalent to the phosphorylation site (Ser- 19) in smooth myosin RLC is not seen in the structure of striated muscle myosin (11). Therefore, before atomic structures of both phosphorylated and dephosphorylated RLC bound to twoheaded smooth muscle myosin are obtained, information at less than atomic resolution may lead to critical insights about the regulation of smooth muscle myosin or could at least rule out some proposed mechanisms.
The two approaches used recently to obtain structural information about the relationship of the two heads of smooth muscle myosin II, cryo-AFM of single myosin molecules (3) and EM of heavy meromyosin (HMM) (1,2), examined, respectively, the full-length molecules in the nonfolded (6 S) conformation present in smooth muscle in vivo and the "short tail" hexamer in two-dimensional crystals. Cryo-AFM showed that thiophosphorylation of RLC favored greater separation of the two heads, but the change was not dramatic (3). In contrast, analysis of two-dimensional crystals by EM (1,2) showed the actin-binding domain of one of the two heads to be bound to the converter domain of the second "blocked head" in dephosphorylated HMM. This intramolecular steric hindrance of one by the other head was proposed to be responsible for the inhibited state of dephosphorylated two-headed myosin (2).
In view of the importance of the mechanism of regulation of smooth muscle and nonmuscle myosin II, we re-examined the relationship of smooth muscle myosin heads by cryo-AFM of whole myosin molecules obtained by various preparatory methods, including the charge of the surface on which myosin is deposited. We further made the (reasonable) assumption that for the steric model to operate in vivo, the blocked conformation * This work was supported by National Institutes of Health Grants EB002017 and HL48807. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence may be addressed:  1 The abbreviations used are: RLC, regulatory light chain; EM, electron microscopy; AFM, atomic force microscopy; HMM, heavy meromy-would have to be sufficiently stable in the absence of activation, and, consequently, the energy cost of the transition from the blocked to the open state should be significantly higher than the thermal energy. Otherwise, if regulation were solely or primarily dependent on the blocked-head model, thermal fluctuation would separate the two heads, leading to nearly continuous activity (12). Our results suggest only a minimal energy difference between the closed and the open forms of dephosphorylated myosin and do not support the mechanism of regulation based on intramolecular occlusion by strong forces binding one head to the other.

EXPERIMENTAL PROCEDURES
Protein Preparation-Turkey gizzard myosin was purified as described previously (13) and kept at 4°C in buffer A (0.3 M KCl, 10 mM KH 2 PO 4 , 1 mM EGTA, 1 mM EDTA, 1 mM dithiothreitol, pH 6.8) for less than 2 weeks or stored in buffer A with 50% glycerol at Ϫ20°C without noticeable differences in both the function and the structure of the protein. Before use, myosin was extensively dialyzed at 4°C against buffer A or low salt buffer (20 mM K 2 PO 4 , 1 mM MgCl 2 , 1 mM EGTA, 120 mM KCl, pH 7.2) to remove glycerol. To eliminate undesirable "unregulated" molecules, the unphosphorylated myosin was incubated with equimolar actin in buffer A containing 1 mM ATP and 4 mM MgCl 2 and subjected to centrifugation at 200,000 ϫ g for 1 h (see "dead heads" removal protocol in Ref. 14). Myosin was thiophosphorylated by a standard procedure (15) as described previously (3); thiophosphorylation efficiency was determined for each preparation by one-dimensional urea-gel electrophoresis according to Ref. 16 and Fig. 1. More than 95% of myosin was thiophosphorylated, whereas only a single band of nonphosphorylated myosin was found in the absence of ATP and ATP␥S.
ATPase Activity Assay-The ATPase activity of the purified myosin (with or without the dead heads removed, as well as nonphosphorylated and thiophosphorylated) was measured under single turnover conditions (myosin Ն ATP) by monitoring the rate of inorganic phosphate release with the aid of linked phosphorolysis reaction (Molecular Probes, EnzChek kit E-6645). Following the procedure described by Webb (17), solutions in two Hamilton syringes containing 50 l of myosin (3-5 mg/ml) in 300 mM KCl and separately 50 l of a stoichiometric amount of ATP together with the linked enzyme system (100 M 2-amino-6-mercapto-7-methylpurine riboside, 10 units/ml purine nucleoside phosphorylase) were injected simultaneously into a 100-l spectrophotometric cuvette, and the time course of absorption at 360 nm was monitored. The ATPase rate was determined from single exponential fitting of the absorption increase at 360 nm over time. The "dead time" of the system was less than 1 s, giving adequate time resolution for the present study. The lower limit of rate measurements determined by the sensitivity to [Pi] changes (1 M) was determined as 0.001 s Ϫ1 . Unphosphorylated smooth muscle myosin hydrolyzed MgATP at a rate of 0.007 Ϯ 0.001 s Ϫ1 (n ϭ 7) and at 0.04 Ϯ 0.02 s Ϫ1 (n ϭ 5) after thiophosphorylation by myosin light chain kinase. The ATPase rate of unphosphorylated myosin without the dead heads was below the detection limit (Ͻ0.001 s Ϫ1 ), in agreement with reports in the literature (6,18,19).
Specimen Preparation for Cryo-AFM-Mica was used as the substrate for preparing all AFM specimens. The freshly cleaved mica surface was negatively charged. Positively charged surfaces were prepared by treating mica with 1 mM spermidine followed by washing with deionized water. The treated mica was blown dry with nitrogen gas leaving sufficient spermidine retained on the surface to efficiently adsorb negatively charged molecules. After 20 l of myosin was deposited on the substrate (1 g/ml in buffer A or low salt buffer), excess protein was removed by washing with the same buffer. To avoid possible structural effects, including separation of the heads because of changes in ionic strength during evaporation of the volatile buffer, the surfaceadsorbed molecules were cross-linked with the incubation buffer containing 0.5% glutaraldehyde for 2-3 min. The sample was then washed again with the incubation buffer before finally being flushed with 250 mM or 125 mM ammonium acetate, pH 7.2 (volatile buffer). The specimens were quickly transferred to the cryo-AFM for equilibration and imaging.
Cryo-AFM Imaging-All cryo-AFM images were collected at ϳ80 K with instrumentation described elsewhere (20). The scanner was calibrated with a standard grid, and the accuracy was verified by imaging various samples with known dimensions (21,22). Full frame images were collected at a rate of 1 scan line/s. Distortion was minimal because of the high thermal stability of the system (23). No image processing was applied.

RESULTS
We first examined the distribution of different conformations of nonphosphorylated myosin on freshly cleaved mica. A typical image is shown in Fig. 2A. The relative arrangement of the two heads may be roughly divided into three different categories (see Fig. 2B): well separated by 180°(␣); still resolvable, but at a smaller angle (␤); no longer distinguishable even in stereo views and, therefore, may be considered closely associated (␥). Note that in well resolved molecules both the regulatory and catalytic domains of each of the heads can be resolved (e.g. see molecules in Fig. 2B). We also note that, because of the random FIG. 1. Coomassie-stained urea gel (lanes 1, 2, and 3) and corresponding Western blots (lanes 4 and 5) using an anti LC 20 antibody. Lane 1, purified calmodulin; lane 2, unphosphorylated smooth muscle myosin (with dead heads removed); lane 3, thiophosphorylated smooth muscle myosin. Thiophosphorylated LC 20 migrates faster than unphosphorylated LC 20 in urea gels. The lower band is calmodulin used for the thiophosphorylation reaction. orientation of the molecules on the substrate, even the count of the ␥ conformation may yield an overestimate of closed states if, for example, two separated heads were viewed as one on top of the other. Counting the number of nonphosphorylated myosin molecules that were clearly resolved by cryo-AFM (n ϭ 294), we found 25% in the ␣ conformation, 20% in the ␤ conformation and 55% in the ␥ conformation. In other words, the ratio between closely associated heads and open heads is almost 1:1.
However, as noted above, the two-dimensional crystals used for the cryo-EM study were formed on a positively charged lipid monolayer (2), whereas the results of Fig. 2 show molecules deposited on a negatively charged surface. Even though at such high ionic strength mica is only slightly negatively charged, we were concerned that the negative surface could still affect the relative arrangement of myosin heads during adsorption. Therefore, to more closely reconstitute the conditions used for the cryo-EM study, we also used a substrate with a positive surface charge. The adsorption efficiency did not change significantly when the surface charge was reversed, consistent with cryo-EM results. Moreover, the distribution of myosin in the three recognizable conformations remained essentially unchanged. A typical cryo-AFM image of nonphosphorylated smooth muscle myosin molecules on the positively charged substrate is shown in Fig. 3A. Simple counting (n ϭ 349) indicates that in this case, 25% of myosin is in the ␣ conformation, 19% in the ␤ conformation, and 56% in the closed ␥ conformation (see Table I).
To determine whether phosphorylation affects the conformation of myosin heads, we examined samples containing more than 95% thiophosphorylated RLC. The distribution of molecular conformations of thiophosphorylated myosin on the positively charged surface was noticeably shifted between the three conformations, as made more obvious by color-coding the molecules in Fig. 3 Table I, thiophosphorylation reduced the fully closed ␥ conformation by about 50% (to 23%).

, B versus A. As summarized in
To exclude the possibility that strong ionic shielding under high salt conditions (buffer A) could favor conformations otherwise disfavored at lower ionic strength, we further compared the distribution of nonphosphorylated and thiophosphorylated myosin in a buffer closer to physiological ionic strength and also similar to that used for the cryo-EM studies (20 mM K 2 PO 4 , pH 7.2, 1 mM EGTA, 1 mM MgCl 2 , 120 mM KCl, 1 mM ATP) (2). These results were essentially identical to those obtained under high salt conditions ( Table I). As shown in Fig. 4, the head conformation of myosin molecules is essentially the same as observed at higher ionic strength (both nonphosphorylated and thiophosphorylated) except that the bends of the tail (characteristic of low salt conditions) are clearly seen. When observed in this buffer even in the absence of ATP, the structure of myosin molecules as well as the distribution of the relative head-to-head positions were not noticeably different (data not shown). However, most of the molecules were not in the completely folded 10 S conformation in this low, but close to physiological ionic strength buffer containing ATP, unlike those found in previous studies where myosin was imaged in 50 mM NaCl (24,25). To further examine the possible sources of this discrepancy, we cross-linked myosin molecules in 60 mM KCl buffer containing 1 mM MgATP, similar to the ionic strength of the buffer used in a previous electron microscopic study (24). Only at this very low (unphysiological) ionic strength did we find a significant number of myosin molecules in the 10 S conformation (Fig. 5, 10 S gallery). Because the partition of molecules into different energy states is determined by the Boltzmann distribution, the statistics shown in Table I can be used to estimate the energy difference between the two states: n 1 /n 2 ϭ exp(Ϫ␦⑀/kT). We found that ␦⑀ is only 0.24 kT for dephosphorylated myosin, and when thiophosphorylated, the energy difference favoring the open state is significantly greater: ␦⑀ ϭ Ϫ1.21 kT.  Fig. 2. In this case, the relative partitioning of the myosin heads into the split and closed conformation remains unchanged compared with negatively charged mica surface, indicating that the charge on the substrate does not have a major influence on the head arrangement. B, thiophosphorylated myosin shifted the distribution into the ␣ and ␤ configurations (see Table I). To show clearly the shift in distribution, the ␣ and ␤ forms are green and the ␥ is red. Image size ϭ 1 m.  Nonphosphorylated  300  25  19  56  349  120  23  25  52  291  Thiophosphorylated 300  64  13  23  336  120  58  18 24 285

DISCUSSION
The questions we attempt to address in this study are relatively simple: 1) do the two heads of inactive myosin containing dephosphorylated RLCs have a high probability of being closely apposed to each other, and 2) does phosphorylation of RLC alter the relationship between the two heads independently of the charge of the surface on which the molecules are deposited? The answers directly address the question whether phosphorylation regulates smooth muscle and nonmuscle myosin II by modifying steric hindrance between the two heads. A high probability of heads that are closely apposed would imply the existence of sufficiently strong binding between the two heads to inhibit catalytic activity (i.e. regulation), whereas a low probability of finding such closely apposed heads would indicate a weak interaction that is easily perturbed by thermal fluctuation.
We found a nearly equal probability (ϳ52%) of closely apposed compared with partially or fully open conformations of the two heads in a population sample of nonphosphorylated myosin molecules. Regardless of the preparatory method, the surface charge of the substrate, or the ionic strength, cryo-AFM of whole nonphosphorylated myosin molecules failed to show a predominant distribution of closely apposed heads. Based on the Boltzmann distribution, this suggests that the energy cost of separating the two heads is small, only about one-fourth of the thermal energy. Our results, obtained by direct imaging of single molecules, do not support the suggestion, based on cryo-EM image processing and modeling of two-dimensional crystals of HMM crystallized at low temperatures, that strong head-to-head interactions are responsible for the inhibited state of myosin in which RLC is not phosphorylated (2). The high proportion (ϳ60%) of thiophosphorylated myosin heads in the completely open state (present study) confirms our original observation (3) about the effect of thiophosphorylation.
The implication that regulation of two-headed myosin is not because of strong head-to-head interactions in the unphosphorylated state is consistent with several relevant studies of myosin in solution. For example, chimeric myosins composed of the motor domain of skeletal muscle and regulatory domains and light chains of smooth muscle are regulated by RLC phosphorylation, arguing against the possibility that special properties of the smooth muscle heavy chain favor intramolecular binding between the two motor domains and are required for regulation (26). Furthermore, the myosin ATPase activity of HMM that contains only one motor domain is regulated by phosphorylation (Ref. 27, but see also Ref. 6). The two heads of dephosphorylated myosin have been shown to be photo-crosslinked through the RLCs (within 8.9 Å of each other) (28), and thiophosphorylation abolishes this cross-linking (29) consistent with the greater separation of the two thiophosphorylated heads in solution (present study and Ref. 3). In the present experiments, because myosin in buffer was deposited on the mica surface prior to cross-linking, only the conformation with closely apposed heads at the time of deposition would be crosslinked independent of the protein substrate interaction. A cryo-AFM image of myosin molecules represents a "snapshot" of all conformations in solution. Indeed, ϳ50% of the unphosphorylated heads were separated, indicating that the two states of different head-to-head interactions are equally present in solution. In other words, myosin molecules fluctuate between the "open" and "closed" states, because of their relatively small energy difference. Following completion of this manuscript, images of two-dimensional crystals of smooth muscle myosin interpreted as being consistent with 10 S conformation were published (30). The relatively high ionic strength of the crystallization buffer used in the study (20 mM phosphate, 90 -120 mM NaCl) that would have further increased during crystallization is difficult to reconcile with conditions producing 10 S myosin in solution (24).
Considering the behavior of muscle itself, the blocked head model of Wendt et al. (2) would have to allow one head to bind to actin to be consistent with the ability of nonphosphorylated myosin in the permeabilized muscle strip to develop rigor stiffness (31), enabling dephosphorylated myosin filaments to hold actin in a rigor-like conformation preventing actin filaments from diffusing away into the solution in motility assays (32), and enabling filaments containing both phosphorylated and dephosphorylated heads to move actin at intermediate rates in comparison to fully phosphorylated heads (33). The blocked model requires the actin-binding domain of one head to be strongly bound to the converter domain of the second head to maintain myosin in the off state. It is also less likely to allow HMM to bind actin in the presence of MgATP, and would not allow the cross-bridge to mechanically cycle (34). The model is FIG. 4. Cryo-AFM images of freshly purified dead head removed myosin observed in buffer (20 mM K 2 PO 4 , 1 mM MgCl 2 , 1 mM EGTA, 120 mM KCl, pH 7.2). The myosin samples were prepared by depositing the protein on mica in buffer containing 120 mM KCl and 1 mM MgATP and subsequently were washed and cross-linked in the same buffer: nonphosphorylated myosin (A); thiophosphorylated myosin (B). In low salt buffer, the partition of the relative head-to-head conformation into the split and closed forms remains similar to that in high salt buffer, but the bends in the tail, characteristic of the low salt conditions, are apparent. Image width ϭ 1 m.

FIG. 5.
A gallery of the completely folded 10 S form of myosin observed with cryo-AFM. Myosin samples were deposited on a positively charged mica surface in buffer (10 mM K 2 PO 4 , pH 7.2, 1 mM MgCl 2 , 1 mM EGTA, 60 mM KCl, 1 mM ATP). Myosin was cross-linked for 2 min in the same buffer containing 0.5% glutaraldehyde before being exchanged into a 60-mM ammonium acetate buffer (pH 7.2). Only when myosin was cross-linked in this low ionic strength buffer did we observe a significant number of myosin molecules in the 10 S form. The tail of these myosin molecules is bent at two regions and folds into thirds of the rod. also difficult to reconcile with the cooperative attachment and cycling of dephosphorylated cross-bridges induced by micromolar ATP or CTP in muscles in rigor, which results in force development of up to 40% of maximal active tension without phosphorylation (31); a recent report (35) showing that both heads of tissue-derived smooth muscle HMM bind to actin irrespective of RLC phosphorylation or the presence of ADP seems also incompatible with a model (2) that requires the actin-binding domain of one head to be strongly bound to the converter domain of the second head to maintain myosin in the off state.
Taking the appropriate Popperian view that hypotheses cannot be proven but only tested and destroyed, we conclude that the available data do not support the hypothesis that requires strong intramolecular head-to-head binding forces to maintain the inhibited state of nonphosphorylated myosin II. In contrast, several studies (13, 36 -39) are consistent with an alternative mechanism in which the activating message from the regulatory to the catalytic domain is transferred through the long ␣ helix of the heavy chain and the converter domain as the result of stiffening of the ␣ helix induced by phosphorylation of RLC. It is plausible to suggest, given the structural homology between calmodulin and myosin light chains (10,11), that phosphorylation of the myosin regulatory light chain causes a conformational change similar to that experienced upon calcium binding by calmodulin, causing it to collapse on the regulatory domain of the lever arm, resembling the interaction of calmodulin with myosin light chain kinase (40). What is not explained by such a mechanism is the original critical observation (4) that two-headed myosin II containing nonphosphorylated RLC is inhibited, but a single head is active. If there are no strong binding forces between the two catalytic domains (2), then what mechanism keeps the catalytic domain of one head informed about the proximity of the other? Is it the inhibitory effect of the rod portion of the molecule on the regulatory domain (41,42), or is it the frequency of contacts (between regulatory and/or catalytic domains) driven by thermal fluctuations?