The Central Role of the Tail in Switching Off Myosin II in Cells

Myosin II is a motor protein playing an essential role in cell motility. The molecule can exist as a polymer that pulls on actin to generate motion, or as an inactive monomer with a compact structure, in which its tail is folded and its two heads interact with each other. This conformation functions in cells as an energy-conserving storage and transport molecule. The mechanism of inhibition is not fully understood. We have carried out a 3D reconstruction of the switched-off form revealing for the first time multiple interactions between the tail and the two heads that trap ATP hydrolysis products, block actin binding, obstruct head phosphorylation, and prevent filament formation. Blocking these essential features of myosin function can explain the high degree of inhibition of the folded form of myosin, serving its energy-conserving, storage function in cells. The structure also suggests a mechanism for unfolding when activated by phosphorylation.


Introduction 35
Myosin II is a motor protein that, together with actin filaments, generates mechanical force and 36 motion using the chemical energy of ATP hydrolysis [4]. In muscle, myosin II is responsible for 37 shortening and force production, while in nonmuscle cells it is essential in cell adhesion and 38 division, intracellular transport and cell migration [5]. The myosin II molecule is a hexamer 39 composed of two heavy chains, two essential light chains (ELC) and two regulatory light chains 40 (RLC). The light chains and the N-terminal halves of the heavy chains form two globular heads, 41 containing ATP-and actin-binding sites, while the heavy chain C-terminal halves form an α-helical 42 coiled-coil tail extending from the heads. The tails self-associate to form the backbone of thick 43 filaments, which are the functional form of myosin II in muscle and cell motility. Mutations in the 44 heads and tail impair myosin function and cause muscle disease [6,7]. Myosin II molecules can 45 exist in two different conformations, 6S and 10S, named for their sedimentation coefficients [8].

46
10S molecules have a compact structure [2,[9][10][11][12], in which the tail is folded into three segments of 47 similar length, and the heads are bent back on the tail and interact with each other (Fig. 1). This 48 folded conformation traps the products of ATP hydrolysis, causing strong inhibition of ATPase 49 activity [13], and inhibiting assembly into thick filaments. Phosphorylation of the RLC in 10S 50 molecules favors unfolding to the 6S form [10,14,15], in which ATPase activity is switched on [16] 51 and the tail has an extended conformation, favoring filament assembly. 52 The inhibited, 10S molecule plays a fundamental role in nonmuscle cells, where it functions as an 53 energy-conserving, storage form of myosin, and provides assembly units for thick filament 54 formation when and where contractile activity is required [14]. Its compact conformation would 55 facilitate rapid transport to the sites of filament assembly and thus motility. In smooth muscle, 10S 56 myosin may also provide a pool of inactive monomers, which can unfold and augment myosin 57 filament length or number when smooth muscle is activated [17][18][19]. In striated muscles, myosin is 58 almost exclusively filamentous, but there is nevertheless a small concentration of 10S molecules, 59 Free head shows heavy chain (blue), ELC (magenta) and RLC (red). The tail is folded into three segments: segment 1 (= subfragment 2), originating from the heads and ending at hinge 1; segment 2, starting at hinge 1 and ending at hinge 2; and segment 3, starting at hinge 2 and ending at the tip of the tail. whose compact form has been proposed to facilitate transport to sites of filament assembly during 60 muscle development and turnover of myosin filaments [20][21][22]. 61 A unique structural feature of 10S myosin is an asymmetric interaction between its two heads (the 62 interacting-heads motif, or IHM), which inhibits their activity [1]. Head-head interaction was first 63 observed in two-dimensional (2D) crystals of smooth muscle heavy meromyosin in the switched-off 64 state (RLCs dephosphorylated) [1], and later confirmed by electron microscopy (EM) and 2D 65 classification of single myosin II molecules [9]. Inhibition was suggested to occur by different 66 mechanisms in the two heads. In one head ("blocked"), actin-binding is hindered by proximity of 67 the other head ("free") to its actin-binding site. In contrast, ATPase activity of the free head is 68 inhibited through stabilization of its converter domain, which interacts with the blocked head [1]. 69 Three-dimensional (3D) reconstruction of thick filaments subsequently showed that 6S myosin, 70 when polymerized into thick filaments, has a similar head-head interaction to that in folded, 10S 71 molecules [23]. The IHM in thick filaments may underlie the super-relaxed state of muscle [24], in 72 which myosin activity is highly inhibited, thus serving as an important energy-conserving 73 mechanism for striated muscle. 74 The IHM structure has been compared in a variety of evolutionarily diverse species. EM and 2D 75 image classification of negatively stained molecules shows that the compact, folded form of 10S 76 myosin has been conserved since the origin of animals: head-head interactions in the IHM appear 77 similar in all species studied [2,25,26]. Similarly, the IHM first observed in thick filaments in cryo-78 EM studies of tarantula muscle [23] has been confirmed in thick filaments from a variety of 79 vertebrate and invertebrate species [27][28][29][30]. The high level of conservation of the IHM in isolated 80 molecules and in thick filaments over hundreds of millions of years implies that it is a fundamental 81 structure, critical to the function of muscle and nonmuscle cells [2,25,26]. 82 While the inhibitory interactions between the heads in the IHM are similar in thick filaments and 83 10S molecules, there are likely to be critical additional interactions with the folded tail that account 84 for the order of magnitude greater inhibition of 10S molecules compared with filaments [21,31].

85
However, it is not known what these interactions might be, as no 3D reconstruction of the entire 86 molecule has been reported. The interactions between the myosin heads in 10S myosin were well 87 defined in studies of 2D crystals [32], but the tail was poorly visualized and the 3D reconstruction 88 provided no insights into its potential role in regulation. The tail was clearly delineated in 2D class 89 averages of isolated 10S molecules, but its course in three dimensions was not studied [9]. Here 90 we present a 3D reconstruction of 10S myosin molecules using single particle analysis. were chosen for image processing. The tail in these molecules was folded into 3 segments, and 104 wrapped around the heads, as previously described [9]. The heads showed different apparent 105 shapes and sizes, corresponding to different orientations of the molecules on the EM grid (Figs. 2  106 B-E, S1). Compared to the relatively rigid arrangement of the interacting heads, the tail was quite 107 flexible where it extended away from the heads. In this region, segments 1, 2 and 3 ( Fig. 1) run 108 closely together, and this combined 3-segment rod bends within a range of ~ 60º from its point of 109 emergence from the heads, as previously described [9]. Typical 2D class averages of the 110 interacting heads and the proximal portion of the 3-segment tail are shown in Figs. 2 B-E and S1. 111 The asymmetric architecture of the blocked and free heads was consistent with the results of 112 previous 2D classification analysis [2,9,25,35]. Mirrored class averages, where molecules lie on 113 the grid facing down or up, were observed as previously reported [35] (Fig. S1). The class 114 averages showed that the myosin molecules typically lie parallel to the grid surface; the different 115 orientations about their long axis enabled us to analyze the 3D structure of the molecules using 116 single particle analysis. 117

3D reconstruction 118
An initial 3D structure of the folded molecule was computed using the random conical tilt method 119 [36], in which images are recorded at 0º and 50º tilts. This model was then used for projection 120 matching to determine the final single particle reconstruction (EMDB accession code EMD-20084). 121 Computing the initial structure ab initio excludes potential model bias in the final reconstruction. 122 The latter had a moderate resolution (25 Å; Fig. S2) due significantly to the flexibility of the 123 molecule [9]. 124 The head region of the reconstruction has the appearance of a flat disk, with an irregular polygonal 125 "front" view (as viewed in Fig. 3A, D; Movie 1), enclosing a central hole, and a narrow side view 126 (Fig. 3C, F, Movie 1), similar to the motifs seen in the 2D crystals of HMM and myosin [1,32] and 127 in thick filaments [23]. The atomic model of the interacting-heads structure computed from cryo-128 imaged smooth muscle HMM 2D crystals (PDB-1I84) [1] docked well into the reconstruction using 129 rigid-body fitting (Fig. 4, Movie 2), confirming this similarity. This was clear in both front-and side-130 views (Fig. 4D, F). The fit implies excellent preservation of the IHM in negatively stained single 131 molecules (cf. [33]), without significant flattening, and enables us to interpret head structure and 132 interactions at higher resolution (<7 Å) than the reconstruction itself [37]. 133 The reconstruction also reveals, for the first time, the course of the folded tail as it wraps around 134 the heads -a feature that was mostly missing from the only previous reconstruction of 10S myosin 135 [32]. Only a small portion of the folded tail (~150 Å) extending as three merged segments above 136 the heads was included in the reconstruction, due to flexibility of the distal region. However, the 137 essential part of the tail that interacts with the heads was included. Three segments of the tail 138 could be identified in this region (Fig. 3). Strikingly, all three appear to bind to the heads: segment 139 2 along the left side of the blocked head (Figs. 1, 3A, D), and segments 1 and 3 on the back side 140 of both heads (Fig. 3B, E, Movie 1). Segment 1 (subfragment 2 of the tail) would be expected to 141 start at the merge point of the lever arms of the blocked and free heads (blue ribbon, Fig. S3).

142
However, there was missing density for this initial portion of subfragment 2 in the reconstruction 143 (circle, Figs. 3E, S3), a known issue for this very flexible region of the tail [28,38]; more density 144 , but with space-filling model. The atomic model fits well into the reconstruction in all views, except for the narrow part of the lever arm in the blocked head (red arrow) and part of the ELC in the free head (yellow arrow). Unfilled densities mostly represent portions of the three segments of the tail, for which there is no atomic model. There is also some unfilled density in the blocked head (green arrow) that would be filled by repositioning the flexibly-connected SH3 domain (not done with the rigid-body docking procedure; see Methods). In a previous study, molecular dynamics flexible fitting of cryo-imaged tarantula thick filaments demonstrated this point [3]. Note: (C) and (F) rotated in opposite direction from (C) and (F) in Fig. 3. becomes visible at low contour cutoff (Fig. S4). The first visible part of segment 1 density was 145 assumed to occur where the tail density (cyan) widens, as segment 1 joins segment 3 (above the 146 circle, Fig. 3E). Segment 1 then runs next to segment 3 across the blocked head (in a region now 147 known as the mesa [39]) towards the top, where segments 1 and 3 merge with segment 2 (Fig. 3D, 148 E). Segment 2 (magenta in Fig. 3) branches from the merged segments near the top and wraps 149 around the blocked head. The end of this segment and the starting point of segment 3 form the 150 second hinge point of the tail (Figs. 1, 3D), thought to occur at Glu1535 [9]. This region of density 151 (marked with a star in Fig. 3D) is unique to 10S myosin, and not found in HMM (lacking segments 152 2 and 3) or thick filaments (where the tail is unfolded). Segment 3 (red arrow in Fig. 3E) starts from 153 the hinge point, then runs, merged with segment 1, over the back of the blocked head (Fig. 3D, E) 154 to the top of the IHM where they join segment 2. This arrangement of segments 1 and 3 on the 155 same side of the IHM (Figs. 3, 5) would allow for rapid opening of the molecule upon activation. An 156 alternative arrangement, consistent with 2D projection images (Figs. 2, S1; [9]), but not with the 3D 157 structure, would have segments 1 and 3 on opposite sides, sandwiching the blocked head and 158 restricting opening of the IHM (cf. [9]). 159

Intramolecular interactions within 10S smooth muscle myosin molecules 160
The 3D reconstruction establishes key interactions occurring within the 10S myosin molecules 161 (summarized in Table 1). Three types of interaction were found: head-head, head-tail, and tail-162 163 contact between segments 1 and 3, and segments 1, 2 and 3, respectively. From these initial points, these segments run 172 173 tail. The interaction between the blocked (B) and free (F) heads (BF) has been described before [1, 174 32], and is confirmed here. Eight putative interactions between the heads and the tail (T) have not 175 previously been observed in 3D: these include the free head with tail segments 1 and 3 (TF), and 176 the blocked head with segments 1, 2 and 3 (TB). Interactions between different segments of the 177 tail were also observed (TT). 178 Head-head interactions: As described previously, the overall conformation of the IHM is 179 produced by the blocked and free heads (red and green respectively in Fig. 3) interacting with 180 each other (BF, blue arrow in Fig. 5A). The atomic model from this earlier work (PDB1i84) can be 181 fitted well into the volume (Fig. 4, Movie 2). 182 Tail-head interactions: Three interactions are found between segment 2 of the tail and the 183 blocked head as the tail wraps around its perimeter (Fig. 5A). TB2 is the interaction between tail 184 Figure 5. Interactions between the tail and heads in 10S myosin. (A) Front view shows the interactions between segments 2 (magenta) and 3 (cyan) and the blocked head (TB2, TB3, TB4, TB5), between the two heads (BF), and between segment 2 and the other tail segments (the interaction extends upwards from TT1). TB2 is between the SH3 domain (blue) in the blocked head and segment 2, which lies under it. TB3 is between segment 2 and the converter domain (cyan) of the blocked head. TB4 is between segment 2 and the ELC (orange) of the blocked head. TB5 is between segment 3 and the blocked head RLC. (B) Back view shows the interaction between the free head and the tail (TF1, TF2, TF3), between segments 1 and 3 (unresolved from each other) and the blocked head (TB1), and between segments 1 and 3 (starting at TT2). Interaction TF1 is between the actin-binding loop in the free head and the tail, while TF2 is between the free head and segment 1 or 3. TF3 may represent merging of tail and blocked head densities and not be a real interaction (see text). TB1 is an extended interaction between the blocked head and segment 1 and/or segment 3. (C) Magnified view of TB5. Cys108 is represented by a cyan sphere. (D) The view in (C) is rotated 135º to visualize the geometric relationship between Cys108 and segment 3. segment 2 and the blocked head SH3 domain (blue in Fig. 5A). The blocked head is likely to be 185 stabilized by this interaction. Interaction TB3 was found between segment 2 and the converter 186 domain of the blocked head (cyan in Fig. 5A). The third interaction, TB4, was found between 187 segment 2 and the ELC of the blocked head (orange in Fig. 5A). 188 Four interactions were observed between the tail and the underside of the heads (Figs. 5A, B, 189 Table 1). TB1 is an extensive interaction, where the tail (including segments 1 and 3, which are not 190 resolved from each other) runs over the back-surface of the blocked head motor domain (Fig. 5B).

191
TB5 involves interaction between the RLC of the blocked head and segment 3, close to the second 192 tail hinge point (Fig. 5A). TF1 occurs between the tail and the actin-binding loop of the free head, 193 while TF2 is the interaction between tail segment 1 or 3 and the upper 50K domain of the free 194 head. TF3 is a possible interaction between the tail and the free head RLC, but could alternatively 195 represent low resolution merging of densities without actual contact ( Fig. S4; see Discussion).

196
These interactions between the tail and the free head were not reported in 2D analysis of smooth 197 muscle myosin [9,35], probably because the tail in these interactions is superimposed on the 198 heads and cannot be identified in the 2D images. coiled-coil complex [9]. A second extensive interaction appears likely between segments 1 and 3 203 (unresolved from each other), as they pass together over the blocked head (Fig. 5B). This 204 interaction starts at TT2, where the first part of segment 3 (cyan in Fig. 5B), would merge with the 205 initial portion of S2 (not visible in the reconstruction, as discussed earlier). Other parts of the tail 206 (the region of segment 2 peeling away from the triple segment and wrapping around the blocked 207 head, and the start of segment 3 after the second tail hinge) exist as single coiled-coils, consistent 208 with their narrower diameter in the reconstruction. 209

Heterogeneity in the reconstruction 210
The flexibility of the merged tail segments extending beyond the head region is visualized in raw 211 images, and was investigated using 2D classification by Burgess et al. [9], who showed that the tail 212 flexed within a range of ~ 60 o where it left the heads. The heads are also flexible, though more rigid 213 than the tail. 3D classification was performed to explore flexibility and heterogeneity in the 10S 214 molecules. 15833 particles were classified into 6 classes using RELION [40] (Figs. 6, S5). Classes 215 3 and 4, comprising 51% of the particles, exhibited very similar structures (Fig. 6C, D). Both tail 216 and IHM were visualized clearly in 3D reconstructions and could be fitted well with the 3D atomic 217 model of chicken smooth muscle myosin. The particles in these classes were combined and used 218 for 3D refinement to produce the reconstruction described above. The other 4 classes were 219 smaller and exhibited different tail and IHM structures in their reconstructions (Fig. 6A, B, E). 220 The classification results suggest that the tail plays a crucial role in stabilizing the compact 221 structure of 10S myosin. When the tail can be traced explicitly (classes 3, 4 and 6), the IHM shows 222 more consistent structural features than in the other three classes. 30% of myosin molecules 223 (classes 1, 2 and 5) had flexible or disordered tails. Segment 2 density was discontinuous in class 224 1 and almost completely disappeared in classes 2 and 5, and the second hinge point was not 225 visible in the latter two (Fig. 6B, E). Although these classes exhibited a similar basic asymmetric 226 organization of heads, their IHM structure was somewhat varied, and the atomic model of the IHM 227 (PDB 1I84) could not be docked into them very well. This implies that the tail directly influences the 228 stability of the molecules, helping to maintain a compact structure: the weak interaction between 229 the blocked and free heads is strengthened by the interactions of the heads with the tail. Disruption 230 of these interactions results in conformational changes and less compact myosin molecules. Class 231 6 molecules exhibited less detail than classes 3 and 4, and were therefore also excluded from the 232 final reconstruction. 233 234

Discussion 235
While the head interactions that define the IHM in 10S myosin have previously been studied in 236 detail [1,32], the role of the tail in inhibiting function has not been well appreciated. Our 237 reconstruction shows the 3D organization of the tail for the first time, providing crucial insights into 238 the mechanism of inhibition that underlies the storage function of the folded molecule. The 239 densities of the tail running across and wrapping around the two heads can be traced explicitly 240 (Fig. 3D, E, F, Movie 2), in contrast to earlier studies where the tail could not be identified [1] or 241 was not analyzed in 3D [9]. These earlier studies illustrate the difficulties of visualizing the myosin 242 tail by cryo-EM and the ease with which it can be seen by negative staining, our technique of 243 choice (see [33,34] and Methods). While the resolution of the reconstruction is not high (due partly 244  binding, by head-head and head-tail interaction; 3. Activation by phosphorylation, through RLC-tail 252 interaction; and 4. Filament formation, by tail-tail interaction. Inhibition of every aspect of myosin 253 function appears to be nature's failsafe method for conserving energy in the inactive state. 254 The tail plays a central role in switching off 10S myosin activity 255 The reconstruction shows evidence for eight possible interactions between the tail and the heads 256 in 10S myosin ( Table 1). Five of these involve the blocked and three the free head, suggesting that 257 the blocked head is the more stable [9]. All eight interactions may be important in creating a fully 258 inhibited 10S molecule, but two stand out as being the most stable. . This is approximately the distance segment 1 must travel from its origin at the head-269 head junction to its interaction (TB1) with the blocked head mesa (Fig. 5B). This supports the view 270 that the S2-blocked head interaction plays a critical role in inhibition [23]. 271

Segment 3-blocked head RLC interaction:
We suggest that interaction TB5 (Fig. 5A), 272 between segment 3 and the blocked-head RLC, is the strongest (and therefore the key) interaction 273 that pins the tail to the heads, creating the folded, inhibited structure of 10S myosin, and helping to 274 hold segments 2 and 3 in position where they interact with key sites on the blocked and free 275 heads. The existence of this interaction is supported by photo cross-linking, showing that Cys108, 276 in the C-terminal half of the RLC, is cross-linked to the tail between Leu1554 and Glu1583, which 277 lie in segment 3 [35]. The tail must therefore fold back on to the light chain domain. The calculated 278 distance from Glu1535 to Lys1568 (the middle residue in the cross-linking region) is ~ 48 Å, 279 assuming 1.485 Å rise per residue along a coiled-coil. In the reconstruction, the distance measured 280 from Glu1535 (hinge 2 [9], star in Fig. 3D) to Cys108 (cyan in Fig. 5C, D; Movie 3) in the blocked 281 head RLC is similar (~ 55 Å). In the crosslinking studies, only one of the two RLCs is crosslinked to 282 the tail [35,43]. This is readily explained by the reconstruction, which shows that only blocked 283 head Cys108 is close enough to the tail for crosslinking (Movie 3). 284 The crucial importance of TB5 to the folded structure is illustrated by previous biochemical and EM 285 observations of 10S myosin.
(1) When observed by rotary shadowing or negative staining, 286 molecules that are not crosslinked by glutaraldehyde show a loosely folded structure, in which the 287 only (and therefore strongest) intramolecular contact occurs between the start of segment 3 and 288 the neck region (containing the RLC) of one head [10,35]: the other interactions observed in our 289 work are presumably weaker, and, in the absence of the glutaraldehyde treatment that we used, 290 are disrupted by binding of the molecule to the negatively charged mica surface used for 291 shadowing [26].
(2) In Dictyostelium and many insect flight muscle myosin II molecules, segment 3 292 does not bind to the blocked head; in these molecules, the tail and free head are quite flexible (Fig.  293 S6A, C).

296
They thus lack segment 3 entirely and the last 100-150 Å of segment 2. Neither of these myosins 297 folds [26], further supporting the idea that segment 3 is essential for forming the 10S structure. (4) 298 Biochemical observations also demonstrate the importance of the RLC for folding. to the heavy chain, and the N-terminal for folding. TB5 in our reconstruction shows segment 3 303 positioned over the C-terminal lobe and immediately next to the N-terminal lobe of the blocked 304 head RLC, providing a structural explanation for these findings (Fig. 7E, F). The first twenty-four 305 residues of the RLC are absent from the N-terminus in PDB 1i84, probably due to mobility. This 306 region (dubbed the phosphorylation domain (PD) [46]) stabilizes tail folding in smooth muscle 307 myosin [47]. We have estimated the location of these residues by superimposing the RLC of 308 scallop (PDB 3JTD, which includes eleven additional N-terminal residues) on to the smooth muscle 309 structure (Fig. S7). Residues 1-11 in scallop (14-24 in smooth muscle) of the blocked head RLC 310 appear to contact segment 3, consistent with this stabilizing interaction. Based on sequence 311 analysis and modeling, it has been suggested that the N-terminal 24 amino acids, including a 312 specific group of positively charged residues, may lie over a cluster of negatively charged residues 313 in segment 3, trapping this region of the tail on top of the RLC C-lobe and accounting for the 314 strength of this interaction [35]. An alternative proposal is that these N-terminal RLC residues 315 interact with helix A of the ELC in the switched off molecule [48]. 316

Other interactions:
While the blocked head RLC interacts with segment 3, a putative 317 interaction (TF3) with the free head RLC is less certain (Fig. 5C, D). The density in this region is 318 very weak and fitting of PDB 1i84 suggests that there is no RLC density to fill the volume (Fig. S4); 319 the relatively uniform cylindrical tail also lacks such density (Figs. 4, 5, S3, S4). This apparent 320 interaction could therefore represent low resolution merging of the tail and head densities without 321 actual contact. TF3 is close to the point where the first portion of segment 1, missing from the 322 reconstruction, as discussed earlier (Figs. 3E, S4), would meet segment 3. This may account for 323 the sudden widening of the tail at this location (Fig. 5B), leading to possible fusion with the RLC 324 density. 325 In addition to interacting with the heads, the folded tail appears to interact with itself, where 326 segments 1 and 3 run next to each other over the blocked head (interaction starting at TT2; Fig.  327 5B), and at the top of the reconstruction, where all three segments form a compact, rod-shaped 328 structure, extending upwards from TT1 (Figs. 3, 5A). These interactions appear to play an 329 essential role in the function of 10S myosin. First, folding sequesters the tail, inhibiting 330 polymerization with other tails to form filaments, the active form of myosin II in cells. Second, 331 extending up from the TT1 contact site, the complex of the three tail segments is quite flexible and 332 can be significantly curved [9]. However, the interacting-heads structure remains relatively rigid.

333
The TT1 contact site may function as a lock, preventing conformational changes in the curved rod 334 complex from being transferred to the heads, thus helping to maintain their inhibited state. This is 335 suggested by our observation that in the absence of TT1, as in Dictyostelium (Fig. S6B) and many 336 insect indirect flight (Fig. S6D) muscle myosin molecules [26], the free head frequently detaches 337 from the blocked head. The interaction between segments 1 and 3 starts at TT2. Proximity of TT2 338 to interaction TB5 (Fig. 5) may stabilize the interaction between segment 3 and the RLC in the 339 blocked head. 340

Tail segments 2 and 3 stabilize the IHM in 10S myosin 341
The atomic model of the IHM (PDB 1i84, based on 2D crystals of HMM [1]) fits well into the head 342 densities of the reconstruction, without the need for any substantial change in head structure (Fig.  343 4D-F, Movie 2). This shows that binding of tail segments 2 and 3 (absent from HMM) to the heads 344 in 10S myosin is not required to generate the head-head interactions, and that it does not 345 significantly distort them. However, the frequency of formation of the interacting-heads structure in 346 HMM, containing only the first third of the tail and having only one interaction between the tail and 347 heads (TB1 in Fig. 5B), and the stability of the free head in HMM, are both lower than in full-length 348 (10S) myosin [9]. Thus, these segments stabilize the head-head interactions, the compact 349 structure and the inhibited function of 10S myosin. Several observations support this view: (1) a 350 subset of the reconstructions that lack some of the tail density features correspondingly had a 351 more varied head structure (Fig. 6A, B, E).
(3) Myosin molecules from insect flight muscle frequently behaved similarly 354 (Fig. S6A, B). These observations suggest that the interactions of all three tail segments with the 355 heads are crucial in fully stabilizing the structure of 10S myosin. 356

Intramolecular interactions in thick filaments and single molecules: similarities and 357
differences 358 10S myosin and thick filaments share several interactions. These include BF, TB1, TB2, TB3 and 359 TF1 (Fig. 5, Table 1). Conservation of these interactions between filament and molecule suggests 360 that they play an important role in the inhibited state of myosin. In both filament and molecule, it is 361 thought that the BF interaction, between the two heads, constrains movement of the converter 362 domain of the free head, inhibiting phosphate release, and blocks actin-binding by the blocked 363 head [1,23]. TB1 occurs between segment 1 and/or segment 3 and the blocked head where these 364 two segments run over its motor domain (Fig. 5B). In filaments, where the tail is not folded [23], 365 and in HMM, where segments 2 and 3 are absent [9,35], this blocked head-S2 interaction is also 366 present, supporting the view that TB1 involves S2, and consistent with data showing that 367 subfragment 2 binds to subfragment 1 in solution [49]. The reconstruction suggests that in 10S 368 myosin, segment 3 also binds to the blocked head motor domain. The tail is wider where it runs 369 over the blocked head than at the start of segment 3, where it is a single α-helix (Fig. 5B), 370 suggesting that segment 3 runs next to segment 1 over the blocked head. This double interaction 371 with the blocked head would presumably help to stabilize the IHM of the folded molecule. TB2 (Fig.  372 5A) occurs between segment 2 and the SH3 domain (blue in Fig. 5A) of the blocked head. A 373 related interaction is observed in relaxed thick filaments, between the SH3 domain of the blocked 374 head and the tail from the next pair of heads away from the bare zone [23,50]; however, in this 375 case the interaction involves segment 1, and its polarity is the reverse of that in the 10S molecule.

376
TB3 occurs between segment 2 and the converter domain of the blocked head (Fig. 5A). As with 377 TB2, a similar interaction (but with reverse polarity) occurs between the converter domain and 378 segment 1 of the neighboring tail in thick filaments [23]. The fifth interaction, TF1, occurs between 379 the tail and the free head actin-binding loop in 10S myosin. A similar interaction between S2 and 380 the actin-binding loop has been suggested in tarantula thick filaments [51], suggesting that TF1 381 also involves S2. 382 The other interactions in 10S myosin are unique to the folded structure. These tail-head and tail-tail 383 interactions are presumably important specifically to the switched-off, folded structure, which 384 underlies its storage and transport function. The compact structure would make transport more 385 efficient through the crowded intracellular environment, while inhibition of ATPase and actin 386 binding would minimize energy use and futile binding to actin filaments during transport. 387

10S myosin is switched off by steric blocking of ATPase and actin-binding sites on both 388
heads 389 EM studies of 10S smooth muscle myosin [1,32] suggested that inhibition was achieved by 390 different mechanisms in the two heads: (1) by preventing actin binding in the blocked head; and (2) 391 by inhibiting ATPase activity in the free head [1]. But they provided no information on the possible 392 role of the tail in inhibition. Our results suggest that the tail plays a central role in switching off 393 activity, such that both of these inhibitory mechanisms occur in both heads-but in different ways. 394

Inhibition of actin binding: Part of the actin-binding interface of the blocked head is obstructed 395
by interaction with the free head, as shown previously [1] (interaction BF; red ellipse in Fig. 7C). In 396 the free head, it is clear that actin-binding is also sterically blocked, through interaction of its actin-397 binding loop with the tail (interaction TF1; Fig. 5B; green ellipse in Fig. 7C, D), and by interference 398 of the bundle of three tail segments (projecting up from TT1) with the free head actin-binding 399 interface. Thus, each head would be prevented from binding to actin by distinct steric blocking 400 mechanisms, one involving the tail and one the other head. 401

Inhibition of ATPase activity:
Similarly, ATPase activity of 10S myosin appears to be inhibited 402 not only in the free, but also the blocked head, by stabilization of the converter domain of each, 403 preventing phosphate release (cf. [52]). In the free head, the converter is immobilized by binding to 404 the blocked head [1] (interaction BF; Fig. 5A; blue ellipse in Fig. 7A). In the blocked head, the 405 converter may also be immobilized-in this case by binding to segment 2 of the tail as it wraps 406 around the head (interaction TB3, Fig. 5A; red ellipse, Fig. 7A, B). 407 Inhibition of ATPase activity in both heads of 10S myosin would help explain a previously puzzling 408 observation. Release of ATP hydrolysis products is ten times slower in 10S myosin than in HMM 409 (0.0002 S -1 and 0.003 S -1 respectively) [31]. If the free head is inhibited in both (by binding to the 410 blocked head), blocked head product release must be more inhibited in 10S myosin than in HMM.

411
This could be explained by interaction of tail segment 2 with the blocked head converter in 10S 412 myosin but not HMM, where it is missing. As mentioned previously, tail wrapping around the heads 413 may also stabilize the basic head-head interaction of the IHM [1], enhancing the inhibitory effect of 414 the two heads on each other in the 10S molecule. 415 We conclude that phosphate release and actin-binding are both inhibited in both heads, which can 416 explain why 10S myosin is completely switched off [31], making it so well suited to its energy-417 conserving storage function in cells. The folded tail plays a critical role in this inhibition. 418 In thick filaments [23], both actin binding and ATPase are also sterically inhibited in both heads, 419 again in different ways. Blocked-head actin binding is inhibited by interacting with the free head, as 420 with 10S myosin, while the free-head is inhibited by orientation of its actin-binding face towards the 421 thick filament backbone, away from actin. Both converters are also inhibited [23], impeding 422 ATPase activity: in the free head by binding to the blocked head, as in 10S myosin; in the blocked 423 head by interacting with S2 from the neighboring crown, as discussed above [23]. The greater 424 number of interactions found in 10S myosin than in thick filaments suggests that the IHM of 10S 425 myosin is more stable than in thick filaments, consistent with the greater inhibition observed in 426 molecules [21,31]. 427

The central role of the tail in activation of 10S myosin 428
Smooth and nonmuscle 10S myosins are activated in vitro by phosphorylation of their RLCs. 429 Under physiological ionic conditions this causes conversion to the 6S (unfolded) conformation [10, 430 15], which can form filaments, hydrolyze ATP and bind to actin. Our reconstruction provides insight 431 into the mechanisms of these processes. Biophysical data suggest that phosphorylation of 10S 432 myosin is a sequential process in which one RLC is phosphorylated more readily than the other, 433 and it has been proposed that this might result from an asymmetric arrangement of the myosin 434 heads [53, 54] (others have suggested that phosphorylation is random, but the starting myosin in 435 those studies may not have been in the 10S conformation [55,56]). The reconstruction reveals not 436 only asymmetry of the heads (as previously described), but also dramatically different 437 environments of their phosphorylation sites, which would readily account for a sequential 438 phosphorylation process. Ser19 in the free head RLC is completely accessible to MLCK and would 439 be easily phosphorylated. In the blocked head, Ser19 is sterically hindered by segment 3 of the tail 440 (interaction TB5; Figs Importantly, heavy meromyosin, which lacks segments 2 and 3, requires phosphorylation of only 457 one of its two heads for activation [60]: in this case TB5 does not exist, and phosphorylation of a 458 single head is sufficient to break the weaker head-head interaction in HMM. The requirement for 459 phosphorylation of both heads in 10S myosin (one of which is sterically inhibited) for full activation 460 may be another mechanism by which the cell conserves energy in the quiescent state. If any 461 MLCK molecules in the cell are Ca 2+ -insensitive (e.g. through proteolysis or other damage), 462 resulting RLC phosphorylation will occur preferentially on the free head, but this will not activate 463 the molecule. Only when intracellular Ca 2+ levels rise will there be sufficient activated MLCK to 464 phosphorylate the blocked head RLC as well, with concomitant unfolding of the 10S structure. 465 Based on the above observations and previous work, we propose the following mechanism for 466 regulating 10S myosin activity in cells. "breathing" of the intact 10S structure, which transiently exposes blocked head Ser19 (we assume 475 that the interactions we have described are relatively weak and in equilibrium between interacting 476 and non-interacting states). There is also evidence that the RLCs may interact with each other and 477 with the flexible initial portion of subfragment 2, stabilizing the conformation of this region of the 478 IHM [23,28,50]. Phosphorylation of the free head RLC may disrupt these interactions, leading to 479 increased local flexibility and greater accessibility of Ser19 on the blocked head RLC. 4. 480 Phosphorylation of the blocked head phosphorylation domain weakens or abolishes its interaction 481 (TB5) with the negative charge cluster on segment 3 due to decreased charge attraction. 5. With 482 breakage of TB5 (the key interaction required for the folded state), the binding of segments 1 and 2 483 to the heads is destabilized, allowing the tail to become fully extended. 6. Loss of head-tail 484 interactions during tail unfolding may further destabilize previously weakened head-head contacts, 485 promoting separation of the heads. 7. The extended tails now interact to form filaments, actin-486 binding sites on both heads are exposed, and inhibitory interactions of the converter domains are 487 lifted, leading to full actin-activated ATPase activity and filament sliding. 488

Electron microscopy 490
Smooth muscle myosin II from turkey gizzard was purified according to [61] interactively selected from the 100 tilted and untilted micrographs using the program JWEB in 514 SPIDER. 2D classification was performed for the untilted particles. Four classes of particle with 515 similar views in the class averages were chosen for further analysis. For each class, the Euler 516 angle of each particle from the tilt images was calculated based on the tilt angle and the azimuthal 517 angle from 2D classification. Back projection was carried out to compute the 3D reconstruction. 518 Finally, the four class averages were rotated to make the tail parallel to the y-axis. The obtained 519 azimuthal angles were used to modify the Euler angles and to merge particles from the chosen 520 four classes, and the model was built using the merged data set. 521 2D classification and 3D reconstruction. The folded tail of 10S myosin protruding beyond the heads 522 is flexible, and also rotates (within a 60º range) about its junction with the heads [9]. To simplify 523 analysis, only particles with an unbent tail pointing straight up from the head-head junction were 524 chosen for image processing. Particles with dimensions 150 x 150 pixels, including both heads and 525 a small portion of the protruding tail, were manually selected using EMAN [64], and then low-pass 526 filtered to 20 Å. All subsequent image processing was carried out with RELION [40]. Four rounds 527 of 2D classification were performed. Bad or low-resolution particles with smeared densities were 528 removed after each round. The remaining 15833 particles were subjected to 3D classification into 529 6 classes. Classes 1, 2 and 5 showed variable appearances of the tail and heads. Classes 3 and 4 530 had similar 3D reconstructions and were combined for refinement. The resolution of the refined 3D 531 reconstruction was estimated to be 25 Å (Fig. S2). Class 6 molecules showed less detail than 532 classes 3 and 4, and were excluded from the final reconstruction 533 Atomic fitting. The atomic model of the interacting heads of chicken smooth muscle myosin (PDB 534 1I84) [1] was docked into the 3D reconstruction using rigid-body fitting in CHIMERA [65] (Fig. 4). 535 Molecular dynamics flexible fitting was not attempted owing to the low resolution. Densities outside 536 of the heads were attributed to the three segments of the tail. According to 3D classification ( Fig.  537 6), S2 in the 3D reconstruction was quite flexible. We therefore chose not to fit an atomic model of 538 S2 to the reconstruction. 539

Acknowledgments 540
We thank Drs. Edward Korn and Xiong Liu for Dictyostelium myosin and Sanford Bernstein and 541 Floyd Sarsoza for insect flight muscle myosin. We thank the Electron Microscopy Facility and its 542 staff at UMass Medical School for instrumentation and assistance. This work was supported by 543 NIH grants AR072036, AR067279 and HL139883 (RC) and HL075030 and HL111696 (MI). MI is 544 an awardee of University of Texas STARS PLUS Award. The content is solely the responsibility of 545 the authors and does not necessarily represent the official views of the National Institutes of 546 Health. 547

Competing interests 548
The authors declare no competing interests. IHM is clearly demonstrated in the left images, while some averages on the right show the free 769 head separated from the blocked head. 770 Interaction of segments 2 and 3 with the heads in smooth muscle 10S myosin considerably 771 stabilizes the IHM. The above myosins, in which these interactions are weak or absent, serve to 772 illustrate this point. Smooth muscle myosin clearly shows the second hinge point in 2D averages, a 773 characteristic marker of the three-segment folded structure (green arrows, Fig. S1). The absence 774 of this feature in many insect flight myosin molecules suggests that segments 2 and 3 are more 775 weakly bound to the heads. Correspondingly, the free head tends to detach from the blocked head 776 and the blocked head also becomes more mobile, as shown in (C) above. In Dictyostelium, 777 783 Figure S7. The RLC N-terminal interacts with tail segment 3 in the 10S molecule. The N-784 terminal 24 residues of the smooth muscle RLC are absent in crystal structures due to mobility 785 (see text). To estimate their location, we docked the scallop RLC (3JTD; red, which has more 786 residues present) onto the smooth muscle RLC (green), so that the location of the additional 787 residues could be visualized. Residues 1-11 (red spheres, equivalent to residues 14-24 in smooth 788 muscle) suggest that the RLC N-terminal directly interacts with segment 3. A and B are front and 789 back views. 790