Structure of mavacamten-free human cardiac thick filaments within the sarcomere by cryoelectron tomography

Significance

Cardiac muscle contains two filament types: thin filaments composed of actin and calcium regulatory proteins and thick filaments composed of myosin to produce force, titin to define the filament length, and myosin-binding protein C to modulate force production. During each heartbeat all cardiomyocytes contract, which requires a complex mechanism to match cardiac output to physiological requirements. Separate mechanisms control the quantity of myosin heads released for thin filament binding and the degree of thin filament activation that permits myosin binding. Using cryo-EM of frozen human cardiac muscle lacking exogenous drugs, we show that the thick filament is structured to provide three levels of myosin activation. This hierarchy of control may explain many properties of length-dependent activation in cardiac muscle.

Abstract

Heart muscle has the unique property that it can never rest; all cardiomyocytes contract with each heartbeat which requires a complex control mechanism to regulate cardiac output to physiological requirements. Changes in calcium concentration regulate the thin filament activation. A separate but linked mechanism regulates the thick filament activation, which frees sufficient myosin heads to bind the thin filament, thereby producing the required force. Thick filaments contain additional nonmyosin proteins, myosin-binding protein C and titin, the latter being the protein that transmits applied tension to the thick filament. How these three proteins interact to control thick filament activation is poorly understood. Here, we show using 3-D image reconstruction of frozen-hydrated human cardiac muscle myofibrils lacking exogenous drugs that the thick filament is structured to provide three levels of myosin activation corresponding to the three crowns of myosin heads in each 429Å repeat. In one crown, the myosin heads are almost completely activated and disordered. In another crown, many myosin heads are inactive, ordered into a structure called the interacting heads motif. At the third crown, the myosin heads are ordered into the interacting heads motif, but the stability of that motif is affected by myosin-binding protein C. We think that this hierarchy of control explains many of the effects of length-dependent activation as well as stretch activation in cardiac muscle control.

The heart, with its continuous rhythm of contraction and relaxation, is powered by a precise interplay between the actin-containing thin filaments and the myosin-containing thick filaments, situated within the cardiac muscle sarcomeres (1) (Fig. 1A). The vast majority of familial cardiomyopathies occur in just three sarcomeric proteins, myosin (2), myosin-binding protein C (MyBP-C) (3, 4), and titin (5, 6), which are constituents of the thick filament. Although the thin filaments have been characterized structurally at high resolution (7–9), the molecular arrangement in the vertebrate thick filament, unlike its invertebrate counterparts (10–13), has been more challenging to solve. Early attempts made in different species such as the zebrafish, mouse, and human (14–17) achieved structures with limited resolutions. Experimental difficulties include the greater intrinsic heterogeneity of vertebrate thick filaments compared to the highly ordered invertebrate ones (10, 18) and other obstacles from sample preparation procedures. The negative stain technique may obscure the structural details beneath the backbone surface (15) and even lead to radial collapse of the constituent molecules (19).

Fig. 1.

Schematic molecular arrangement of sarcomeric proteins and a tomogram from a human cardiac sarcomere. (A) Schematic showing the relative placement of cardiac sarcomeric proteins on the thick filament, together with the thin filament, Z-disc, cMyBP-C, and titin. (B) The major structural components of a myosin II molecule. (C) A slice of the reconstructed tomogram obtained by the combination of cryo-FIB and cryo-ET showing the thick and thin filaments within a sarcomere, as well as Z-disc, A-band, and Bare Zone containing M-line. (Scale bar: 200 nm.) A movie of one tomogram is shown in Movie S1.

Striated muscle thick filaments are mainly composed of myosin II molecules, each of which have two heavy chains, two regulatory light chains (RLC), and two essential light chains (ELC) (20). Each heavy chain consists of an N-terminal “head,” dubbed S1 (Fig. 1B). Myosin head densities protruding from the thick filament backbone in the same axial level are referred to as “crowns.” Each head is followed by a long α-helix which together with the same helix of another myosin forms a nearly 1,600 Å long α-helical, coiled-coil “tail.” The myosin tail length is highly conserved (12) across all species and is just short of 11 × 143 Å in length, i.e., 11 crowns. The initial ~1/3 of the coiled coil is called the S2, and the rest is called LMM (21). The initial segment of the myosin tail closest to the heads is referred to as the proximal S2.

The catalytic “off” state of myosin is the interacting heads motif (IHM) (22). The IHM function is apparent from its structure; the F-actin-binding site of one head, the “blocked” head, interacts with the companion head, the “free” head. When formed on a myosin filament, the blocked head binding onto the proximal S2 region orients the free head actin-binding domain away from the thin filament (22). The IHM has been identified under relaxing conditions in individual myosin II molecules (23–25) and isolated thick filaments from many species including cardiac thick filaments preserved in negative stain (10, 16, 18). Recent studies showed that the IHM exists in high proportion within native mouse cardiac muscle as well as isolated human cardiac thick filaments when treated with mavacamten (26, 27).

Two nonmyosin proteins of vertebrate thick filaments also participate in controlling cardiac muscle contraction. One is myosin-binding protein C (cMyBP-C) that modulates force production (28, 29). The interactions between cMyBP-C’s N-terminal domains and the thin filament have been characterized by cryo-EM (30, 31). In contrast, its interactions with the thick filament, especially with the myosin heads, until recently, were mostly deduced through biochemical approaches (32, 33) which showed that dephosphorylation of cMyBP-C promotes the OFF state of the myosin heads (34, 35) and phosphorylation promotes the ON state (36). The other is titin, which consists mostly of a string of Ig-like and Fn3-like domains. As the largest natural protein known, titin provides sarcomeric viscoelasticity and determines the myosin filament length (5, 37).

Recently, two papers reporting similar results have significantly advanced the structural knowledge of vertebrate striated muscle thick filaments. One, like this work, used cryofocused ion beam milling (cryo-FIB) and cryoelectron tomography (cryo-ET) but applied to mouse cardiac muscle (26); the other used isolated thick filaments from human cardiac muscle (27). Both studies used mavacamten to stabilize the myosin heads into the IHM (38), thus removing a major source of heterogeneity. By comparison, our mavacamten-free structure is more disordered. We observed a nonrandom distribution of ordered myosin heads within the 429 Å C-zone repeat, along with variations of the conventional IHM, which we call semi-IHMs. Our results together with the two recent reports are broadly in agreement except for the IHM distribution and structure.

Results

The Human Cardiac Sarcomere under Relaxing Conditions.

About 2 mm³ of cardiac tissue was isolated from the left ventricle of a donor human heart. The vitrified fragments of the tissue were directly cryo-FIB milled by Ga ions (SI Appendix, Fig. S1A). The resulting lamellas after ion polishing resemble a thin line when viewed from the direction of the focused ion beam (SI Appendix, Fig. S1B). Sarcomeres along the myofibrils in a lamella can be observed directly from the scanning electron micrograph with a near-perpendicular view (SI Appendix, Fig. S1C). The length of a sarcomere is approximately 2.0 μm in vitrified areas (SI Appendix, Fig. S1D). In areas presumably ruined by crystalline ice formation, the myofilaments in the sarcomeres appear disturbed (SI Appendix, Fig. S1E). Rich details are observed in one of the reconstructed tomograms selected from vitrified areas that contains approximately half a sarcomere, revealing the A-band, a zone where thick filaments and thin filaments overlap each other, the bare zone where antiparallel myosin tails interact with each other without projecting heads, and the Z-disc, which contains α-actinin cross-links between antiparallel thin filament lattices (Fig. 1C and Movie S1).

C-Zone Structure of the Human Cardiac Thick Filament.

Myosin head densities are revealed in the unmasked map at a resolution about 26 Å with imposed C3 symmetry (Fig. 2A). Their overall positions along the thick filament agree with previous studies (15, 16, 26, 27). The densities divide into three crowns along the thick filament axis which together constitute one axial C-zone repeat measured to be 429.6 Å, a value in good agreement with X-ray fiber diffraction 429 Å (16, 39). The bottom crown has only a single ordered myosin head, a structure we call semi-IHM; the middle crown shows a conventional IHM with two ordered heads; the topmost crown shows no ordered myosin heads. The bottom IHM’s “free” myosin head appears to interact directly with cMyBP-C (see below), so it is named the IHM-C (C for cMyBP-C). The myosin heads immediately above IHM-C unequivocally adopt the complete IHM form representing the super-relaxed state (SRX), so it is called IHM-S (S for SRX). The topmost crown shows almost no myosin head density suggesting they are mostly disordered. Given that the muscle tissue was frozen under relaxing conditions of MgATP with low calcium, these heads apparently adopt a disordered-relaxed (DRX) state (40, 41). We refer to these myosin heads as IHM-D (D for DRX). Compared with the nomenclature adopted by Dutta et al. (27) and Tamborrini et al. (26), IHM-C corresponds to CrH and Crown 1, both of which only adopt the classical IHM conformation; IHM-S agrees well with CrT and Crown 3; IHM-D is similar to the CrD structure but different from the complete IHM form of Crown 2.

Fig. 2.

The C-zone of the cardiac thick filament structure under relaxing conditions is depicted by reconstructed 3D maps and segmentation of the myosin tails. (A) Side view of the repeating unit from the unmasked global average reconstruction (26 Å resolution), with the top of the segment pointing to the M-line. The IHMs are labeled with black circles. (B) The head densities have been masked out and subsequent refinement with the heads removed reveals details of individual myosin II tails (16.5 Å resolution). The circles indicated the same locations as in panel (A). (C) Individual myosin II tails are segmented out after enforcing C3 symmetry and grouped according to the myosin tail layer to which they contribute. The tails of IHM-C, IHM-S, and IHM-D myosin tail layers are colored red, yellow, and blue, respectively. This reconstruction is extended so that it covers the length of a single myosin II tail. (D) Cross-section positions indicated by numbered black lines in extended reconstruction of panel (C) reveal myosin tail layers formed by the three distinct myosin II tails. The connected black dots identify the myosin tail layer assignments of their member myosin tails. The approximate S2 regions separated from the backbone are indicated by black arrows. Note that the thresholds of the tails are adjusted in each cross-section to show approximately the same diameter on the cross-sections. Movie S2 steps through the global average map.

The backbone without head densities possesses C3 symmetry, an observation confirmed during data processing by generating a reconstruction with only C1 symmetry. Details of the myosin tails with a resolution about 16 Å are revealed when the myosin head densities are masked off (Fig. 2B and Movie S2). The filament backbone shows a complex structure consisting of rod-shaped densities, the myosin tails, and strings of larger, ~40 Å diameter domains, which we identify as being titin and cMyBP-C (see below). With the myosin head densities removed, the proximal S2 segment is readily identified where it emerges from under the blocked head (within circles in Fig. 2 A and B). The features are not spaced along a regular helix and differ in both the axial and azimuthal separation. The segment of the proximal S2 which participates in IHM formation is not resolved (42). We generally resolve the myosin tails when they emerge from under the blocked head.

In Squire’s “curved molecular crystalline layer” model of the thick filament backbone, which we will refer to as “myosin tail layers” (43), adjacent myosin tails within each layer are offset by three crowns, which facilitates tracking a myosin tail for its entire length. The segmented tails of three myosin tail layers fill the inner structure of the extended thick filament backbone reasonably well (Fig. 2C and Movie S3). Each of the three different myosin tail layers corresponds to one of the three IHM types. In cross-section (Fig. 2D), each myosin tail layer always shows at least three densities. Sometimes a fourth is seen separated from the other three, which we identify as part of the proximal S2. This density is not a part of the backbone but instead acts as a tether allowing myosin heads to search for thin filament-binding sites (44). Thus, there are 30 small densities representing myosin tails in every transverse map section.

The arrangement of myosin tails has a direct relationship to the myosin head conformations. The IHM-D tail layer lies solely on the outer surface of the myosin tail annulus, which is the region of the filament backbone containing just myosin tails (Fig. 2D). As such, its myosin tails interact with other myosin tails on only the inner side. The outer side binds cMyBP-C. The myosin tail layer linked to IHM-S is the only layer that reaches the center of the filament and thereby defines the narrow and empty central channel. In this capacity, IHM-S most closely resembles the myosin tail arrangement of invertebrates (10, 11, 13). The myosin layer linked to IHM-C also lies mostly on the surface but is covered in a few locations by myosin tails of IHM-D (Fig. 2D). In another departure from invertebrate striated muscle thick filaments, the proximal S2 of myosins incorporated into IHM-C and IHM-S are closely associated and distinctly separated from the myosin tails incorporated into the filament backbone. The proximal S2 of IHM-D is also separated from the rest of the thick filament backbone (Fig. 2D) but is not associated with tails from the other two myosin tail layers.

The three distinct groups of myosin II tails when segmented and combined with their respective head densities (Fig. 3 A–C) agree with recent studies of vertebrate cardiac muscle (26, 27). The atomic models of TaH, TaT, and TaD (27) also fit well with the overall conformations of individual tails of IHM-C, IHM-S, and IHM-D, respectively (Fig. 3D). Among the three human cardiac myosin tail layers, the one formed by IHM-D has a configuration most similar compared with its invertebrate counterpart, a myosin layer from the flight muscle of Bombus ignitus (13) (Movie S5) though its placement in the filament backbone is quite different.

Fig. 3.

Stacks of segmented myosin molecules showing their overall stacking conformations in the C-zone of the thick filament. (A) Five crowns of heads and tails of IHM-C (red) displayed in parallel with the three myosin molecules constituting one crown. The head conformation used is the semi-IHM form. (B) IHM-S (yellow). (C) IHM-D (blue). The head densities used in panels (A–C) are from the focused refinement (see below). (D) Fitting the cardiac myosin II tails TaH, TaT, and TaD from the human cardiac atomic model (PDB 8G4L) into the three segmented tails of IHM-C, IHM-S, and IHM-D, respectively. The color scheme is consistent with panel (A).

Dynamic Structures of the Interacting Head Motifs.

Focused classification including more subtomograms than used in the global average (Fig. 2 A and B) reveals that the IHM-Cs assume two recognizable conformations. One class of IHM-C adopts a complete IHM conformation with ordered blocked and free heads (Fig. 4A). Another form of IHM-C has an ordered “free” head and a disordered “blocked” head (Fig. 4B). The ratio of the complete and the semi-forms is roughly 3:2 (SI Appendix, Fig S2 A and D). IHM-S only adopts the classical SRX conformation when the myosin heads are ordered (Fig. 4C), constituting about 11% of the subtomograms (SI Appendix, Fig. S2 B and D). However, the double heads axis of IHM-S is tilted compared to the horizontal one in the complete IHM-C form, agreeing with recent observations (26, 27). IHM-D does not form a recognizable head shape in the global average (Fig. 2A). Focused classification revealed that no complete IHM formed at this location. Instead, a semi-IHM occurs approximately 5% of the time with an ordered blocked head and a disordered free head (Fig. 4D and SI Appendix, Fig. S2 C and D), suggesting largely a DRX state at this location. Rigid body docking reveals that the atomic model of human cardiac IHM (PDB 8G4L) fits the IHM-D density after deleting the free head from the model. To better illustrate the locations and conformations of the three IHMs, their densities are segmented from the focused refinement results and placed on the thick filament backbone where the myosin head densities had previously been masked off (Fig. 4E). Their locations and conformations agree well with the density map from isolated thick filaments (27).

Fig. 4.

The complete IHM-C, semi-IHM-C, IHM-S, and IHM-D classes from focused analysis, fitted with the human cardiac IHM atomic model (PDB 8G4L). (A) Complete IHM from the IHM-C location, with the free head directly interacting with cMyBP-C and the blocked heads interacting with the proximal S2. (B) Semi-IHM at the IHM-C location reveals only the free head contacting cMyBP-C. The “blocked” head is presumably mobile and disordered. The blocked head is deleted from the atomic model. (C) IHM-S assumes the classical IHM structure. (D) A semi-IHM is found as a minor class at the IHM-D (blue) location consisting of only an ordered blocked head, stabilized by binding the proximal S2. The IHM-D free head is presumably disordered. The free head is deleted from the atomic model. (E) The densities of IHM-C complete form (red), IHM-S (yellow), and IHM-D (blue) obtained from the focused classification are installed on the global average backbone (transparent with outline) to illustrate their locations compared with the isolated human cardiac thick filament map (EMD-29722, dark gray).

The various ordered IHM structures identified by focused classification do not encompass the total number of myosin heads present in the tomograms but is representative of the conformational heterogeneity of particles included in reconstructions. The missing wedge from the limited tilt angle range results in unequal representation of otherwise identical structures. Head structures that did not form well-defined class averages may exist as disordered heads on the IHM-D model or ordered heads on the IHM-S and IHM-C models. Thus, the relative proportions of different head structures found during the focused classification are only approximate (SI Appendix, Fig. S2 A–D).

In Situ Structure of cMyBP-C.

The three C-terminal domains of cMyBP-C, C8-C10, which are known to interact with the thick filament backbone (45), are visible in the globally averaged map (Fig. 5A). Density for C7 is also visible, but the rest of the domains are disordered. cMyBP-C domains C9 and C10 appear attached to the backbone surface, with C8 contacting the IHM-C free head and also the myosin tails. The conformations of C8–C10 are similar to both of the previous reports (26, 27), but differences are observed in the positions of domains before C8. The N-terminal domains have been reported to interact with both the thin filament and the myosin heads (30, 46, 47).

Fig. 5.

Native cMyBP-C domains binding myosin II molecules. (A) The cMyBP-C domains labeled with white arrows are segmented from the unmasked global reconstruction and displayed in their original position on a backbone reconstruction with the IHM densities masked off (gray). (B) Domains C10 to C5’s atomic structures are fitted into the density map in panel (A) and labeled with purple arrows. The C10 to C7 domains from the map are labeled with black arrows. The model deviates from the density map from its C7. (C) Fitting the free myosin head atomic structure into the semi-IHM-C map to show its interaction with domain C8. The cMyBP-C model labeled with purple arrows is fitted to the cMyBP-C densities from (A) (not shown here). The locations of C9 and C8 of the map are labeled with black arrows. (D) Fitting atomic model of TaD into IHM-D’s tail reveals that C10 contacts the Skip 2 (Glu1385) region. All atomic models used above are from PDB 8G4L.

We compared the in situ conformation of cMyBP-C of this work with its counterpart from the isolated cardiac thick filament (27) by fitting the atomic models (PDB 8G4L) into the density map (Fig. 5B). The agreement for domains C10 and C9 implies their role in anchoring the cMyBP-C onto the thick filament backbone. The position of domain C8 also matches, which indicates the binding to IHM-C’s “free” head is sufficient to fix the position of C8 (Fig. 5C). Differences among the reconstructions begin at C7 where the atomic model for C5–C7 falls onto the filament backbone (27). Our density map at C7 points away from the thick filament, similar to mouse cardiac muscle (26). If the situation in isolated thick filaments also occurs in an intact sarcomere, cMyBP-C would use the flexible linker between C8 and C7 to search for its thin filament-binding site. Possible interactions are observed between C10 and the Skip 2 region from the myosin tail of IHM-D (Fig. 5D), which might explain the fixed position of cMyBP-C’s C terminus in the C-zone.

In Situ Structure of Titin in the C-Zone.

The C-zone titin domains were easily identified by their ~40 Å modulation (5). A pair of eleven-domain titin strings are placed roughly parallel to each other along the axis of the thick filament, assuming a similar overall conformation, within one repeat length of 429 Å (Fig. 6A). Three pairs of these titin strings are extended through the entire C-zone.

Fig. 6.

Native titin domains entangled with myosin II molecules. (A) A pair of eleven-domain titin-C (magenta) and titin-M (green) within one C-zone repeat. The domains labeled Ig1 are the first Ig-like domains of the super-repeats of the two respective titin strings, close to cMyBP-C’s C10. (B) Segmented titin strings installed on the segmented pseudo-thick filament, revealing intimate interactions between titin and myosin II tails. Only one set each of IHM-C, IHM-S, and IHM-D are included, colored red, yellow, and blue, respectively. Titin-M domains (green) are running under the S2 regions of the tails of IHM-C (red) and IHM-S (yellow). cMyBP-C domains are colored light gray. (C) After rotating azimuthally 150° clockwise looking toward the Z-disc direction, titin-C domains (magenta) are revealed running under the S2 region of the IHM-D tail (blue). (D) Titin-C (magenta) and cMyBP-C (white) are separated by IHM-D’s tail (blue). IHM-C (red) is installed to facilitate the illustration of cMyBP-C domains. (E) Atomic models of the two titin strings (PDB 8G4L) are fitted in the density map. The starting Ig-like domains are indicated by arrows in the same location as in panel (A).

Titin is believed to bind myosin and function as a molecular ruler (37). Consistent with this idea, the titin domains visualized here appear to be entangled with myosin II tails and act as constitutive components forming the layers of the thick filament backbone. The titin string close to cMyBP-C we refer to as titin-C, corresponding to the terms titin-β (26) and TB (27), respectively. The other string we dub titin-M as it is more deeply integrated into myosin tail layers, which is also referred to as titin-α (26) and TA (27). The titin domains of human cardiac thick filaments appear to insert between myosin tails of the backbone and the proximal S2 (Movies S3 and S4). Titin-M appears to run beneath the proximal S2s of IHM-C and IHM-S, whereas titin-C is running beneath the proximal S2 of IHM-D (Fig. 6 B and C). By doing so, they appear to lengthen the amount of the myosin tail that could function as a tether for the myosin heads (44).

A long-unsettled question is whether cMyBP-C predominantly binds LMM or titin domains, with evidence supporting the latter showing that areas where titin domains are knocked out also lack the binding of cMyBP-C (48, 49). Our map as well as the previous reconstructions (26, 27) reveal no direct contact between cMyBP-C and the nearest titin domains. Any interaction between the two is relayed through the S2 of IHM-D (Fig. 6D). The lack of cMyBP-C in titin deleted areas is most probably caused by an altered thick filament structure, as titin strings themselves are structurally indispensable for an intact backbone (50).

Starting from the Z-disc end, both titin strings possess an eleven-domain super-repeat of the pattern Ig, Fn3, Fn3, Ig, Fn3, Fn3, Fn3, Ig, Fn3, Fn3, and Fn3, starting from their N terminus (5). Based on a recent work of titin structures (27), the first Ig-like domain is identified as the one closest to cMyBP-C’s C10 domain but separated by an IHM-D tail (Fig. 6A). Atomic models of the two titin strings (PDB 8G4L) from isolated human cardiac thick filament (27) fit well into the titin densities on the backbone map (Fig. 6E). Surprisingly, although incorporated into the backbone in different locations, both titin strings assume very similar overall conformations.

Discussion

Comparison with Other Results.

A pair of cardiac thick filament structures similar to ours, but at much higher resolution, have recently been reported (26, 27). Those reports and the present report are quite similar in their identification of the free head binding to cMyBP-C and the position of titin domains relative to the myosin tails. They differ primarily in the ordering of the myosin heads, which was affected due to the addition of the drug mavacamten, a known promoter of IHM formation in cardiac muscle (38, 51). The result obtained with isolated human cardiac thick filaments also found an interaction of the cMyBP-C N terminus with the IHM (27), which supports an equilibrium affected by phosphorylation between the myosin heads, the cMyBP-C N terminus, and the thin filament (34, 52). The isolated human thick filaments (27) also showed comparatively more disorder in the IHM-D crown (CrD) but not the near-total disorder found here. The CrD IHM showed strong density for the blocked head and more disordered density for the free head, similar to here, which when ordered at all showed only the blocked head. The most significant difference is the nonuniform distribution of ordered myosin heads and by implication disordered or active myosin heads in mavacamten-free muscle. Although most of the myosin heads are disordered in the present work, the unusual nature and distribution of the ordered heads suggest unconventional interpretations for cardiac muscle response to external demands.

The Myosin Tail Arrangement.

Two models for the myosin tail arrangement in striated muscle thick filaments have received the most attention, the “curved molecular crystalline layer” (43) and the subfilament (53). In both models, adjacent myosin tails are offset by three crowns and overlap for at least six crowns. With the 3-crown axial offset, the myosin tails within any myosin layer have very complementary shapes (10). Previously, the actual arrangement of myosin tails into layers rather than subfilaments has been demonstrated clearly for only asynchronous insect flight muscle (10, 11, 13). In the truly helical thick filaments, there is only a single myosin layer structure with each myosin tail placed into an identical environment.

In the vertebrate striated muscle thick filament, the 3-crown offset is preserved in each myosin tail layer, but because the repeating distance is also three crowns (429 Å) and not one crown, each myosin tail layer is not required to have the same structure. Indeed, the three myosin-tail layers are as different from one another as are their associated head arrangements. Each of the myosin tail layers of cardiac muscle occupies a unique position in the thick filament. The myosin tails of IHM-S, like CrT (27) and Crown 3 (26), are the only ones positioned to reach the center of the filament. The tails of IHM-D, like those of CrD (27) and those of Crown 2 (26) lie only on the outer surface and partly overly the tails of IHM-C (CrH and Crown 1) which are also partially exposed on the surface.

Our reconstruction lacks the resolution to visualize the myosin tail α-helices but can resolve one tail from another (SI Appendix, Fig. S3 and Movie S6) facilitating comparison between the vertebrate myosin tail arrangement and that from insect asynchronous flight muscle. The myosin tails of the bumble bee (13) form a quasi-planar structure that when viewed from the edge on appears quite thin at 12 Å resolution. None of its counterparts in human cardiac muscle have that characteristic because the myosin tails appear to zig-zag rather than line up laterally as in the flight muscle arrangement (Movie S5). The IHM-D myosin tail arrangement appears most similar to the myosin layers of asynchronous flight muscle but is not an exact copy (10, 11, 13). The myosin tail layers of IHM-C form a zig-zag arrangement (in cross-section) in some regions whose separation appears strongly influenced through interaction with titin.

We generally consider the proximal S2 to be the short crown repeat which in Lethocerus is ~110 Å in length (10, 54), a length that incorporates the part of the S2 that interacts with the blocked head in the classical IHM (10). The other ten crown repeats are constrained to a fixed length of 145 Å (143 Å in relaxed vertebrates) by the periodic structure of the backbone. Titin’s interactions with the segments immediately following the proximal S2, or possibly with the tail segments immediately under this region of the myosin tail, could be responsible for defining how much of S2 interacts with the myosin tails under it and thus its effective length.

The myosin S2 domain is thought to function as a tether about which myosin heads can search for actin subunits to bind (44). Its length has been historically defined by a trypsin cleavage site exposed in myosin molecules solubilized in high salt (55). The length of S2 that functions as a tether has been defined most clearly in insect asynchronous flight muscle (10, 54). The short length of the proximal S2 in asynchronous flight muscle restricts the ability of myosin heads to search for actin subunits due to the bending stiffness of myosin’s coiled coil (56). Here, in a vertebrate striated muscle, a greater length of myosin tail proximal to the heads is distinctly detached from the thick filament backbone apparently under the influence of titin (Fig. 6 B–D). Thus, vertebrate myosin heads may move with a distinctly longer tether than their asynchronous flight muscle counterparts.

The helical symmetry of the invertebrate thick filament means all myosin molecules are in identical environments, except at the bare zone and the tapered ends. Here, in the vertebrate thick filament, we see close association between the proximal S2 domains of IHM-C and IHM-S which causes the myosin heads to break from the supposed helical track with a smaller azimuthal displacement. These two levels form ordered head structures more frequently than IHM-D, suggesting they may behave in concert rather than independently.

The Myosin Head Arrangement.

Ordered myosin head arrangements in relaxed striated muscle thick filaments come in two varieties. One, found so far only in the flight muscle of the large water bug Lethocerus indicus, has the IHM plane oriented perpendicular to the filament axis with the free head attached tangentially to the filament backbone, and the blocked head appearing to “pin” the free head in place (10). In this arrangement, the blocked head of Lethocerus, like the blocked head of the IHM-C semi-form, is the more mobile (57). The best-studied form is the helically ordered tarantula filament (18) in which all the myosin heads form the classical IHM formation with the blocked head overlying the proximal S2 and the free head actin-binding domain facing but not contacting the thick filament backbone. When a semi-IHM forms in tarantula thick filaments, the free heads are preferentially disordered and therefore the most active (58, 59).

Here, two different forms of semi-IHM are observed. A semi-IHM arrangement like that of tarantula is found infrequently only in the IHM-D position; the most frequent IHM-D head arrangement has both heads disordered. Myosin heads in the IHM-D position of isolated cardiac thick filaments (CrD), though ordered, were less ordered than the completely ordered levels corresponding to IHM-S (CrT) and IHM-C (CrH) reported here and had more density in the blocked-head position than in the free head position (27). In the present work, the classical IHM arrangement with both heads ordered is found with high probability in the IHM-S position. The head arrangement in the IHM-C position is predominately classical with both heads ordered. An apparently unique semi-IHM-C conformation has the blocked head disordered and the free head ordered through an interaction with the cMyBP-C domain C8, thus flipping the dynamics of the myosin heads.

RLC phosphorylation disorders myosin heads in isolated thick filaments (60, 61) which might explain why the IHM-D heads are preferentially disordered. In this study, we did not track the overall level of phosphorylation. Conceivably, phosphorylation could be responsible for the high level of disordered heads everywhere. Previous studies have shown that lack of RLC phosphorylation results in a lower ejection fraction (60) and that RLC phosphorylation increases contractility indicating increased myosin head disorder (62). Phosphorylation measurements would need to discriminate among the different myosin head levels, which leaves open a mechanism whereby IHM-D heads might be preferentially phosphorylated over the other two IHMs. Alternatively, their unusual arrangement of myosin tails, which lay entirely on the surface of the filament backbone, or their interaction with titin at the proximal S2 may inhibit IHM formation.

cMyBP-C.

Interaction between cMyBP-C and myosin heads has high medical significance in hypertrophic cardiomyopathy (63, 64). In our reconstruction as well as the preceding reports (26, 27), interaction between cMyBP-C and the free head of IHM-C (Crown 1, CrH) is confined largely to cMyBP-C domain C8 (Fig. 5E). Multiple cMyBP-C domains before C8 are visible extending into space in the mouse cardiac tomograms (26). That so much of cMyBP-C is visible in the intact sarcomeres suggests that its domains before C7 are stabilized in the sarcomere by the N-terminal domains binding to the thin filament (47, 65, 66). Absent the actin filament lattice, alternative interactions are observed as cMyBP-C domains closer to the N terminus nestle into spaces on the thick filament surface (27). One of these, between C5 and the CrH (IHM-C) free head RLC, could further stabilize the free head. The visibility of additional myosin-cMyBP-C interactions in isolated thick filaments (27) and their absence in related in situ structures (26) suggest they occur only when binding to the thin filament is inhibited.

The M-domain positioned between C1 and C2 has three frequently phosphorylated sites (67). Previous experiments showed that cMyBP-C could reduce the relative movement between cardiac thick and thin filaments (32), with the dephosphorylated form being a more potent inhibitor (32, 36). Complete ablation of cMyBP-C weakens the super-relaxed state (68). The contact between C8 and the free head of IHM-C is strong enough to reverse the relative stability of heads in the IHM such that the typically less dynamic blocked head becomes the more dynamic. The absence of cMyBP-C would remove this effect. Absent actin binding, cMyBP-C is likely a stronger stabilizer of the IHM due to the additional contacts observed in the isolated thick filament (27). Because the three crowns in the 429 Å repeat are not equivalent, cMyBP-C is mostly affecting IHM-C.

Titin.

The titin arrangement seems to have a significant effect on the myosin arrangement. The myosin head arrangement in the presence of titin is not strictly helical. Titin seems to run approximately parallel to the filament axis while bending back and forth so that the titin domains do not follow a straight path. The departures from the helical symmetry of the myosin heads facilitate this titin arrangement. The arrangement is distinctly different from previous 3-D images of negatively stained vertebrate thick filaments (15, 16), in which the titin domains are approximately linear within the C-zone. The two factors, titin’s approximately linear track and the myosin nonhelical arrangement, need not be independent of one another. Titin interacts with nearly all parts of the myosin tail, in some cases almost to the head-tail junction (27). The departures from helical symmetry on the thick filament surface may be driven by titin’s need to follow an axial instead of the helical path along the thick filament.

Force Regulation of Cardiac Muscles.

Cardiac muscle contracts in two stages. In diastole, the ventricles first enlarge utilizing the blood pressure in the vena cava or pulmonary vein. On a molecular time scale, this process is slow. When the heart is more active, the blood pressure is greater, thus facilitating more rapid filling of greater volume. At the end of diastole, the atria contract providing a sudden boost to the blood volume in the ventricles and a sudden stretch to the ventricular myocytes. Cardiac muscle is comparatively stiff relative to skeletal muscle so that when the ventricle volume is high, the thick filament is put under tension. When the thick filament is put under tension in diastole, it contracts with more force in systole, which is length-dependent activation otherwise known as the Frank–Starling Law of cardiac muscle contraction. An aspect of length-dependent activation is the sudden stretch of the sarcomere caused by the contraction of the atria at the end of diastole, which after a brief interval leads to even greater force production in systole than was produced by the sudden stretch. This is the phenomenon of stretch activation. The two are related but need not be identical.

During these changes governing the force produced by the muscle, increases in calcium concentration, which govern the degree of thin filament activation, are also occurring. Thus, the ability of the thick filament to produce force must be coupled to the activation state of the thin filament. Our reconstruction, produced as it was under relaxing conditions, suggests different levels of myosin head activation during muscle contraction. A myosin crown that is mostly activated, IHM-D, a myosin crown that is largely inactivated, IHM-S, and a myosin crown linked directly to the thin filament, IHM-C, suggest a complex regulatory mechanism to couple the activation state of the two muscle filaments.

A highly developed stretch activation response occurs in insect asynchronous flight muscle (69) but bears some similarity with that of vertebrate striated muscle. Stretch activation in insects is much simpler than in vertebrates in that the contraction frequency, 40 to 200 Hz (70), is mostly species dependent and tuned to the resonant frequency of the thorax. That frequency is much faster than the rate of cardiac muscle contraction in humans (1 to 3 Hz), rabbits (2 to 5 Hz), mice (5 to 14 Hz), and even hummingbirds (8 to 22 Hz) (71–73). Stretch activation in insects occurs at submaximal calcium concentration and requires some link between thick and thin filaments to fully activate contraction (70). Several peptide extensions combined with myosin heads binding to troponin have been suggested to perform this role (70, 74). The stretch in some insects changes the helical pitch and alters the intensity of the 145 Å meridional reflection (74) suggesting changes in the ordering of myosin heads. The stretch activation response of insects is perfectly adapted to a thick filament in which nearly all the myosin heads are positioned to respond in the same way via a helically ordered thick filament structure. The combination of both thick and thin filament contributions may echo properties found in vertebrate striated muscle.

Vertebrate cardiac muscle contractions have different properties. During each heartbeat, all cardiomyocytes contract, so increases in force must come from increases in contraction strength and frequency, which involves a more graduated thin filament response to increases in calcium concentration (75) and a more graduated increase in the relative number of myosin heads available to bind the thin filament and produce force. Unlike skeletal muscle, live cardiac muscle never achieves a “super-relaxed state” (76) though some populations of myosin heads may assume this state.

In skeletal muscle, changes in the thick filament periodicity are coupled to activation of myosin heads even before tension is generated (77). Tension generated within the muscle also leads to disordering of the myosin heads, a phenomenon dubbed tension sensing (78). Thin filament activation of skeletal muscle is also produced by increases in calcium concentration.

Myosin filaments have a unique structure with extensive overlaps of myosin tails within the filament backbone that are uninterrupted along the entire half sarcomere. Nine crowns of the myosin tail length (1,287 Å, almost 900 residues) are tied up in these interactions with systematic offsets of three crowns between adjacent myosins within a myosin tail layer (79). Changes in one tail can be communicated to myosin crowns along the entire filament. For invertebrates, these interactions are identical for every myosin tail except at the tapered ends and in the bare zone. For vertebrates, these interactions are identical only within those myosin tails associated with a particular crown, which could lead to a much more nuanced behavior.

A mechanism of force regulation by the heart is suggested by the distinct conformations of myosin head pairs among the three crowns that constitute the C-zone repeating structure. Through alterations in the calcium concentration, the degree of activation in cardiac muscle varies (80), and a relaxed state occurs in diastole, in which a population of myosin becomes super-relaxed (51, 76). A necessary feature of striated muscle function is a requirement of a structure to communicate the state of the thin filament to the thick filament. That structure is likely to be disordered (activated) myosin heads. The high degree of disorder of both heads of IHM-D suggests that they have a low probability of forming an ordered relaxed IHM conformation absent external influences. Consequently, they likely serve as constant “sentinels” for even the lowest levels of cardiac contractility (81) and as a sensor of the activation state of the thin filament.

The IHM-S heads are predominately ordered. Their myosin tail arrangement is also most like that of invertebrates in that their tails reach the center of the thick filament backbone. They might form a reserve population, which can be activated by RLC phosphorylation or through tension imposed on the thick filament as the ventricles are filled and following the stretch imposed by contraction of the atria. Either way, for the ventricles to respond to sustained increased activity requires a simultaneous increase in activated myosin heads. If high activity is sustained, RLC phosphorylation of IHM-S heads could increase the number of free myosin heads between contractions.

The “free” head of IHM-C binds cMyBP-C. Our structure and other work (26, 46) suggest that cMyBP-C simultaneously binds to the thin filament, thereby implicating cMyBP-C in the stretch activation response. We suggest the sudden stretch from the atria contraction pulls cMyBP-C at the thick filament, thereby disrupting its stabilizing effect on the IHM-C free head. This is a fine-tuning mechanism to adapt the amount of force that the cardiac muscle will generate in each heartbeat, which is dependent on the end-diastolic sarcomere length. However, cMyBP-C’s contribution to stretch activation is complex. Complete ablation of cMyBP-C in skinned cardiac muscle leads to an accelerated stretch activation response (28), greater disorder in the myosin heads, and an increase in basal ATPase (68). It is worth pointing out here that Drosophila flight muscle has a strong stretch activation response despite its myosin heads being completely disordered under relaxing conditions and therefore always activated (11, 56). Thus, IHM-C is modulating the “stretch activation” response; in its absence, tension sensing is still operating and possibly recruiting both IHM-C and IHM-S heads. As the sarcomere is being stretched during diastole, cMyBP-C cross-links must “walk” along the thin filament to maintain contact since the range of sarcomere lengths that occur during contraction exceeds the length of those cMyBP-C domains spanning the thick and thin filaments. That this could happen is consistent with the retarding effect observed when isolated cardiac myosin filaments move across isolated thin filaments (32), which suggest that the detachment/reattachment rate is comparable to but slower than the in vitro sliding velocity. If a faster stretch occurs, IHM-C might become activated if its free head’s stabilization by cMyBP-C is reduced or removed. Stretch activating IHM-C via atria contraction occurs every heartbeat, even at low levels of cardiac activity. The size of the response would depend on the number of connections to the thin filament. Higher levels of activity would occur with higher ventricular filling applying substantial tension to the thick filament backbone, which we propose could activate IHM-S.

The response of IHM-C and IHM-S could be coupled because their S2 domains are closely associated with each other apparently influenced by titin (this work and refs. (26, 27)). Without the modulating effect of cMyBP-C, the tension effects on IHM stability may produce twice as many activated heads as needed and at low levels of activity, produce none. The ordered heads of IHM-C and IHM-S constitute recruitable populations but recruitable by different mechanisms. That ablation of cMyBP-C does not eliminate stretch activation suggests that an additional mechanism is present such as mechanosensing by the thick filament (82), the only other structure likely to respond to an external force (78). The two mechanisms may operate under different physiological loads. Activation by cMyBP-C simply requires a sudden relative displacement of thick and thin filaments possibly with minimal tension applied to the thick filament during diastole and thus might only recruit those heads sequestered in the IHM-C position. Mechanosensing would occur under higher loads and recruit heads from IHM-S even absent RLC phosphorylation. This presumably gives the heart a hierarchical and robust way to control its performance.

Methods in Brief

Nonfailing left ventricular tissue was obtained in collaboration with the Lifeline of Ohio Organ Procurement organization from a 42-y-old male organ donor. Homogenized fragments of cardiac tissue were plunge frozen at normal atmospheric pressure in liquid ethane, skipping conventional protocols of releasing myofibrils as well as high-pressure freezing of intact cells but has potential for forming crystalline ice in the specimen. We identified vitrified areas for cryo-FIB milling in an Aquilos cryo-FIB/SEM dual-beam microscope. Tilt series were collected on a ThermoFisher Titan Krios electron microscope equipped with a BioQuantum/K3 direct electron detector. Tilt series were fiducial-free aligned and tomograms calculated by simultaneous iterative reconstruction technique (SIRT) using IMOD and EMAN2 (83, 84). Damage compensated motion correction used MotionCor2 (85). Contrast Transfer Function correction used Gctf (86). Subsequent image processing steps were carried out using EMAN2 (84).

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Slicing through the tomogram shown in Figure 1c reveals sarcomeric structure through the thickness of the lamella. Slicing direction in the tomogram is from map center to bottom, then moving back to top, and then returning to center.

Movie S2.

The first scene depicts the reconstruction being extended to the full length of a myosin molecule. This is followed by slabbing radially through the extended reconstruction a section at a time. The map densities are depicted in black. The last scene shows the segmented myosin II tails of IHM-C, IHM-S, and IHM-D colored red, yellow, and blue, respectively.

Movie S3.

Myosin II tails segmented and colored to illustrate their layers. Each myosin layer is superimposed on an outline of the complete backbone. Myosin layer formed by the tails of IHM-C, IHM-S, and IHM-D are colored in red, yellow, and blue, respectively.

Movie S4.

Segmented myosin tail layers superimposed on a transparent thick filament map together with titin domains to illustrate their interactions. The coiled-coil tails of IHM-C (red) entangled with titin-C (magenta) and titin-M (green) domains. The myosin tail layer of IHM-S (yellow) is disrupted by IHM-C’s tails (red) and the titin-M (green) domains. The myosin tail layer of IHM-D (blue) is relatively undisturbed compared with the other two. Its interaction with titin-C string (magenta) is mostly near the proximal S2 region.

Movie S5.

The three human cardiac myosin tail layers compared with the myosin tail layer from Bombus ignitus. The cardiac myosin tail layers in solid view are overlaid with the Bombus myosin tail layer (gray mesh). From left to right, IHM-C (red), IHM-S (yellow), IHM-D (blue), and myosin tail layer from Bombus (EMD-28208).

Movie S6.

Slicing through the pseudo thick filament. The left panel shows IHM-C (red), IHM-S (yellow), and IHM-D (blue), which are labeled with red, yellow, and blue, respectively. The starting cross section is the cross section 3 in Figure 2c, moving towards the M-line direction for two repeating units along the thick filament. Note that the thresholds of isolated tails are adjusted for better appreciation of tail trace. The right panel shows the cross-sections of titin-C (magenta), titin-M (green) and cMyBP-C (gray). Transparent pseudo thick filaments are shown with outline in both panels.

Data, Materials, and Software Availability

The EM structures generated through this study have been deposited in the Electron Microscopy Data Bank (https://www.ebi.ac.uk/emdb/) under accession numbers EMD-40468 (87), EMD-40471 (88), EMD-40475 (89), EMD-40476 (90), and EMD-40478 (91). Requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Kenneth A. Taylor (ktaylor@fsu.edu).