Cryo-EM structure of the folded-back state of human β-cardiac myosin

Summary:

During normal levels of exertion, many cardiac muscle myosin heads are sequestered in an off-state even during systolic contraction to save energy and for precise regulation. They can be converted to an on-state when exertion is increased. Hypercontractility caused by hypertrophic cardiomyopathy (HCM) myosin mutations is often the result of shifting the equilibrium toward more heads in the on-state. The off-state is equated with a folded-back structure known as the interacting head motif (IHM), which is a regulatory feature of all muscle myosins and class-2 non-muscle myosins. We report here the human β-cardiac myosin IHM structure to 3.6 Å resolution. The structure shows that the interfaces are hot spots of HCM mutations and reveals details of the significant interactions. Importantly, the structures of cardiac and smooth muscle myosin IHMs are dramatically different. This challenges the concept that the IHM structure is conserved in all muscle types and opens new perspectives in the understanding of muscle physiology. The cardiac IHM structure has been the missing puzzle piece to fully understand the development of inherited cardiomyopathies. This work will pave the way for the development of new molecules able to stabilize or destabilize the IHM in a personalized medicine approach.

Muscle contraction depends on myosins from class 2. In the sarcomeres, they associate in filaments: the so-called thick filaments that interact with the thin filaments of filamentous actin to produce force during contraction (reviewed by ¹). Muscle contraction is energy consuming and fueled by ATP, and skeletal muscles represent between 30 and 40% of the body mass for an adult human². In order to avoid energy loss between contractions, the myosin heads adopt a super-relaxed state (SRX) characterized by a low ATP turnover^3,4,5, promoting energy saving^3,5. This functional state is present in all muscle fibers including skeletal³; cardiac^5,6 and smooth⁷ muscles.

While the structural basis underlying force production by myosins has been intensively studied (reviewed by ¹), the structural basis of the SRX remains poorly understood. From early studies on smooth muscle myosins (SmMyo2), the SRX has been linked to the asymmetrical Interacting Heads Motif (IHM) where the two heads of a myosin molecule fold back onto their own coiled-coil tail^8,9. The IHM was further recognized from structural studies of human cardiac myo2¹⁰, non-muscle myo2¹¹ and muscles from animals such as sea sponges and jellyfish¹¹, worms, insects, arachnids, mollusks, fish and human¹². From the low resolution examined (>2 nm), all types of muscle myo2 IHM look alike and this motif has been assumed to be conserved¹¹. Three higher resolution cryoEM structures of the smooth muscle myosin (SmMyo2) IHM have been reported at resolutions of 6 Ź³, 4.3 Å ¹⁴ and 3.4 Å ¹⁵. These structures reveal the intra molecular interactions within the dimer (IHM interfaces) and the different hinges of flexibility required for the stabilization of this sequestered state for smooth muscle myosin.

The role of myosin filament-based regulation to control the time course and strength of cardiac contraction has been demonstrated ¹⁶. Hypertrophic cardiomyopathy (HCM)-causing point mutations in human β-cardiac myosin, the motor that drives heart contraction, have been reported to destabilize the SRX and increase the number of heads available to engage force production, leading to hypercontractility (eg.,¹⁷). Many of these HCM mutations are located on a flat surface called the “Mesa” and have been proposed to lead to the destabilization of the IHM interfaces, thus increasing force^18,19,20. Other mutations may affect the stability of diverse conformational states of the motor leading to different ways of modulating the force produced²¹. Interestingly, the recent SmMyo2 IHM structures challenges the Mesa hypothesis¹⁴ in that these structures are significantly different from the human β-cardiac homology models²⁰. Moreover, the activator Omecamtiv mecarbil (OM, in phase 3 clinical trials) and the inhibitor Mavacamten (Mava, approved by the FDA for treatment of HCM) have been reported to destabilize²² and stabilize²³ respectively the sequestered state. The sequestered state of β-cardiac myosin has been described at 28 Å resolution¹⁰, but that is insufficient to describe precisely the IHM interfaces. Together, these results emphasize the urgent need for a high-resolution structure of the cardiac myosin IHM to understand the development of inherited cardiac diseases, and to treat these diseases via the development of new modulators acting on the stability of the sequestered state.

Here, we present the high-resolution structure of the human β-cardiac myosin IHM solved at 3.6 Å resolution, with a highest resolution in the heads of 3.2 Å. Importantly, the comparison with SmMyo2 IHM reveals some major differences in the interfaces and the hinges in the lever arm and the coiled-coil region, challenging the claims that the IHM are conserved amongst the different myosins 2 and that SmMyo2 IHM structure would have been a good model to discuss the cardiomyopathy mutations. Altogether, our results provide an accurate framework for studying the effects of HCM mutations, and for understanding small molecule modulator effects on the stability of the sequestered state. Finally, the structure reveals how the differences between the cardiac and the smooth IHM may be an adaptation to fit different muscle functions and organization.

Structure of the cardiac myosin IHM

We solved the near-atomic resolution IHM structure of human β-cardiac myosin-2 via single-particle cryo-EM using a native uncrosslinked protein preparation with ATP bound (see Methods and Extended data Fig. 1). In SmMyo2 but not Cardiac myo2, a distal coiled-coil region (residues 1410–1625) adds to the stability of the sequestered, auto-inhibited state in addition to the S2 coiled-coil ¹⁰(Extended data Fig. 2). The dimeric cardiac myosin construct we used comprised two heads (human β-cardiac motor domain followed by the lever arm with human β-cardiac essential (ELC) and regulatory (RLC) light chains bound to the target HC sequences IQ1 and IQ2 respectively) and 15 heptads of the proximal S2 coiled-coil. The IHM classes used for the final 3D reconstruction represented only a portion of the picked particles (~200 000 particles over 0.5 million). We thus obtained the high-resolution structure of the sequestered off-state (IHM) of human β-cardiac myosin at a global resolution of 3.6 Å (map 1) (Extended data Table 1, Extended data Fig. 1, Fig. 1a). The IHM motif was immediately recognizable in the cryo-EM map, with the so-called blocked and free heads (BH and FH respectively), the lever arm and the proximal S2 coiled-coil. In map 1, the head regions correspond to the highest resolution and the lever arm/coiled coil to the lowest resolution (Fig. 1a, Extended data Fig. 1, Extended data Movie 1). By performing a focused refinement on the head/head region, we improved the map for the motor domains to a highest resolution of 3.2 Å (map 2, Fig. 1b, Extended data Movie 2), with clear indications for the side chains (Fig. 1c). The nucleotide bound in each head is clearly ADP and P_i, which are trapped after hydrolysis of ATP (Fig. 1c, 1d). Interestingly, despite the high resolution of the map for both motor domains, some side chains are not visualized although they are likely involved in intra-molecular interactions stabilizing auto-inhibition (Fig. 1e). Flexibility exists indeed in the formation of these interactions, as shown also from the movies of the 2D classes (Extended data Fig. 1f). The lability of these interactions is an important feature of the IHM stabilization that is required so that the head can be readily activated when they are needed for contraction. We thus computed realistic interfaces with side chain interactions using molecular dynamics (see Methods). Finally, two essential regions of regulation, the N-term extensions of the ELC and RLC were partially rebuilt in density (Fig. 1f). In contrast to the SmMyo2 IHM structures, the density of the FH is equivalent to that of the BH in our maps, indicating that both heads are equally stabilized in the cardiac IHM (Extended data Fig. 1).

Figure 1 – CryoEM density map of the IHM of β-cardiac myosin.

(a) Complete map of the interacting head motif (IHM) at a resolution of 3.6 Å (map 1). (b) Map masked on the two heads improved the resolution at 3.2 Å (map 2). (c) and (d) map 2 allows to see the ADP, the hydrolyzed phosphate and the magnesium. The side chains of the residues in the active site are well defined in density. (e) Interface between the two heads of the IHM (map 1). (f) Coiled-coils, light-chains and N-terminal extensions are seen without ambiguity in map 1, the N-terminal extension (Nterm) are shown by an arrow.

In the cardiac IHM, both heads have the lever arm primed and their motor domains are highly superimposable (rmsd 0.7 Å) (Extended data Fig. 3a). Their structural state corresponds to a classical pre-powerstroke (PPS) state (rmsd ~0.84 Å with 5N6A,²⁴), defined by the structure of the motor domain from the N-terminus through the Converter ending at the pliant region (Extended data Fig. 3b, 3c). The major and only difference is the larger kink in the pliant region²⁵ of the BH lever arm that alters the ELC/Converter interface (Extended data Fig. 3d, 3e, 3f) and the so-called musical chairs (Extended data Fig. 3d), a network of labile electrostatic interactions involved in the control of the lever arm dynamics at this interface²¹. Our results demonstrate that there is no significant conformational change of the Converter orientation upon formation of the sequestered state from the PPS state. The closed nucleotide pocket and the myosin inner cleft do not change either (Extended data Fig. 3g), as suggested by previous spectroscopy experiments ²⁶. Despite its overall asymmetry, the IHM has two motor domains essentially in the same conformation.

We compared the high resolution cardiac IHM structure with the three previous homology models (PDB code 5TBY²⁷; MS03²⁸; MA1²¹) generated using distinct approaches. Since MS03 and 5TBY share similar features, we used only MS03 in the illustration (Extended data Fig. 4a). As expected, these models could only approximate the overall position and interface of the heads. The exact positioning of the structural elements engaged in this interface were difficult to predict in the absence of high-resolution structural information (Extended data Fig. 4b). The models could not list correctly the residues on the BH head that could be close enough to interact with the S2 coiled-coil (Extended data Fig. 4c). The major differences between the models and the structure reside in the lever arm, which gives rise to the overall asymmetry in the structure (Extended data Fig. 4d, 4e). Indeed, higher resolution was absolutely required to describe correctly the hinges of flexibility.

The MA1 model correctly positioned the BH and FH Converters by keeping their position as found in the single head PPS state (5N6A²⁴) ((Extended data Fig. 4f), but the lever arms were modelled as straight, which is true only for the FH (Extended data Fig. 4e). The other models positioned the Converter in a more primed position for both heads by proposing that the Relay orientation would change to further prime the Converter by 26° (Extended data Fig. 4f). In fact, in MS03, large kinks are proposed for the pliant region of both heads (Extended data Fig. 4d, 4e). The high resolution IHM structure now reveals that the Converter position is not modified compared to what is found in an active head but the priming of the lever arm occurs only in the BH head and comes from the modification in the pliant region (Extended data Fig. 4c). The MS03 model had detected the requirement of kinks in the heads and thus had proposed that the heads would adopt a so-called “pre-prestroke” conformation, with a Converter more primed than the conventional prestroke²⁰. This ‘pre-prestroke’ concept implied that slower release of products could result from these conformational changes. The high-resolution IHM structure now reveals that the slow product release (SRX) is not a result of a change in the motor conformation. In the IHM structure, both heads are in the same conventional prestroke conformation and it is the interactions between them that slows their activity.

The interactions stabilizing the IHM

The structure reveals five main stabilizing regions within the asymmetrical IHM: the head-head interface, the coiled-coil/BH, the RLC/RLC and the BH and FH ELC/RLC interfaces (Fig. 2, Extended data Table 2). Extended data Fig. 5 defines the structural elements involved in these interfaces. Both the head-head and the coiled-coil/BH interfaces involve part of a large and flat surface of myosin heads called the Mesa (Fig. 2a; Extended data Fig. 6), which was originally defined as a large and relatively flat surface containing highly conserved residues among β-cardiac myosins across species, which are hot-spots for cardiomyopathy mutations, and which was predicted to be a docking site for other protein elements that would put the myosin head in an OFF state¹⁸. More specifically, the mesa was proposed to be essential in the interaction with S2, and possibly MyBPC or other components of the thick filament²⁰. Our high-resolution human β-cardiac myosin IHM structure now provides a precise definition of the Mesa residues of the BH head that interact with the S2 coiled-coil, and indicates which Mesa residues of the FH head are involved in a distinct stabilization surface. Mutation of some Mesa residues, such as R453, can thus affect two stabilization surfaces if both the BH and FH heads possess the mutation.

Figure 2 – The IHM is stabilized by multiple interfaces.

(a) Footprint of the interfaces stabilizing the sequestered state. The subdomains involved in the interaction are colored and the contours of interface surfaces indicate which region of the IHM is involved in recognizing this surface : L50 (wheat); U50 (marine blue); Relay (yellow); Converter (green); coiled-coil (black); HCM-loop (dark red); Loop-4 (cyan). The large flat surface called “Mesa” is delimited by purple dotted lines. (b) Cartoon representation of the head-head interface. Key-residues involved in the interaction are represented as stick. The two views on the left and on the right represent the interaction involving the HCM-loop and the interactions mediated by the primary head-head interaction site (PHHIS, in red dotted lines) and Loop-4 respectively. The key-interactions mediated by the HCM-loop through R403 are contoured in a black circle. The connectors and elements involved in the interface are shown and named in Extended data Fig. 5. (c) Coiled-coil/BH interface. (d) ELC/RLC interface on the BH, the differences with the FH are illustrated in Extended data Fig. 3. (e) RLC/RLC interface involving the N-terminal extension of the RLC from the FH (_FH-RLCNterm ext).

The head-head interface is formed by close interactions between the Mesa of the FH head with two BH actin-binding elements: loop-4 and HCM-loop as well as residues on the surface of the U50 subdomain previously designated as the primary head-head interaction site (PHHIS) (Fig. 2a, Extended data Fig. 6, Extended data Table 2). The structural elements of the Mesa in the FH head are the Relay helix (a connector essential for allosteric communication between the active site and the lever arm)¹, the β-bulge and HO-linker, two connectors of the Transducer^1,29 previously reported to influence myosin transient kinetics and force production (Extended data Table 2). Interestingly, we find that the HD-linker, an element of the Transducer adjacent to the HO-linker, is also part of the Mesa and interacts with the BH PHHIS (Fig. 2b). In addition, the Converter and the ELC loop 3 also interact with the BH PHHIS. This interface includes in particular the specific _ConverterTop-loop, whose flexibility and conformation greatly depends on the sequence of the Converter and its interactions with the ELC²¹ (Fig. 2b). Three hallmarks of the head-head interface can be reported (Extended data Table 2): (i) a high number of electrostatic interactions; (ii) the central hydrophobic lock-and-key interaction involving the FH _Top-loopI736; (iii) the importance of the BH _HCM-LoopR403 that is involved in electrostatic interactions and a cation-π stacking with _HO-linkerY455, a residue of the FH Mesa (Fig. 2b). The BH loop-4 establishes few interactions with the FH _RelayK503 (Fig. 2b). The coiled-coil (cc)/BH head interface involves the HO-linker and the _L50HW helix, comprising some residues of the Mesa of the BH, which both interact with electrostatic surfaces of the S2 coiled-coil while some hydrophobic residues consolidate the interactions (Fig. 2c).

The hinges and interfaces within the lever arm participate in the stabilization as well as the asymmetric orientation of the two heads of the IHM. The role of flexible hinges in the lever arm for the formation of the IHM had been anticipated³⁰ but they had never been described precisely. Here the model built in clear electron density demonstrates that asymmetry in the various hinges within the lever arm are positioned (i) at the pliant region, located at the end of the Converter (residues 778–782) where the main helix of the lever arm is kinked by ~55° in BH but remains straight in FH (Extended data Fig. 3a); (ii) at the ELC/RLC interface since differences between the two heads in the ELC loop1 conformation and in the nearby HC kink near R808 promote different electrostatic and apolar interactions at the ELC/RLC interface (Fig. 2d, Extended data Fig. 7a, 7b); (iii) within the RLC lobes in which we find differences between the two heads in the positioning of the HC helices in particular for the bulky residues W816 near the inter-lobe RLC linker as well as W829 and Y823 within the N-terminal lobe of the RLC, (iv) at the RLC/RLC interface that involves direct asymmetric apolar interactions between helices 1 and 2 of the BH RLC and the N-term extension (M20, F21) and loop2 of the FH RLC (Fig. 2e); (v) at the RLC/cc interface where HC residues F834-E846 adopt drastically different positions (Extended data Fig. 7c). While all these _BHHC residues mediate interactions within the _BHRLC, only the first part of the _FHHC residues interact within the _FHRLC via different contacts than those formed in the BH head. After Pro838, _FHHC residues form stabilization contacts at the surface of the _BHRLC, in particular via L839 and R845 (Extended data Fig. 7c).

Interestingly, all these interfaces in cardiac IHM are quite modest in surface area and involve several electrostatic interactions, as shown by the electrostatic profile of the footprint (Extended data Fig. 8). Their formation at different places of the IHM maintains the two heads together, yet their relative position is distant which allow variability and leads to lower resolution for the lever arm region compared to the motor domains of the IHM. In addition, most of the contacts made are dynamic in nature making it difficult to visualize side-chains at the interface. As previously described for the Converter/ELC interface²¹, these electrostatic interactions are labile and can arrange differently. The structure demonstrates the generality of the concept of “musical chairs”²¹ as it is found at different interfaces and hinges of flexibility of the cardiac IHM. Finally, the interfaces formed in the IHM include key-elements for actin interaction and mechano-chemical transduction (Transducer and Converter). While interactions with S2 and constraints between the lever arms efficiently result in positioning the heads away from the actin filament, blocking the dynamics of regions involved in force production also contribute to the efficient shutdown of myosin activity.

Cardiac and Smooth IHM differ

Surprisingly, at high resolution, the β-cardiac IHM and the SmMyo2 IHM strongly diverge in structure. This was totally unexpected since from the low and medium resolution structures, the IHM was thought to be conserved in all class-2 myosins (reviewed by³¹).

The first major difference concerns the orientation of the heads within the motif: when the cardiac and the SmMyo2 IHM are aligned on the BH, the cardiac FH is rotated ~20° anticlockwise compared to the smooth FH (Fig. 3a). At the head/head interface, the conformations of the FH β-bulge and HO-linker differ and the interaction they mediate are distinct among these two IHMs (Fig. 3b, Extended data Table 2). In addition, the _FHTop-loop adopts different conformations in Smooth and Cardiac FH heads and distinct interactions are formed by the _FHConverter with the similar _BHPHHIS region of the U50 subdomain in the two IHMs (Fig. 3b). In fact, the sequence alignment of elements involved in the interfaces shows major differences among Myo2s (Fig. 3c), which are linked to changes in structure. For example, the Top-loop of SmMyo2 is one residue shorter and comprises (i) a charge reversal at the position E732 (replaced by a lysine in SmMyo2); (ii) replacement of the key I736 by a methionine (Fig. 3c). Such differences prevent conservation of interactions. Major sequence differences in regions involved in the head/head interfaces are also found in the HO-linker and the β-bulge. Interestingly, from sequence comparison, we can anticipate that the IHM of skeletal muscle Myo2 (SkMyo2) is close to that of cardiac myosin while that of non-muscle myosin 2a (NM2a) would be similar to the SmMyo2 IHM.

Figure 3 – The sequestered state of β-cardiac myosin and smooth muscle myosin (SmMyo2) differ greatly.

(a) Relative orientation of the BH and FH heads in Smooth (Sm, PDB code 7MF3, ¹⁵) and Cardiac (Card) IHMs. SmMyo2 and cardiac myosins are aligned on the BH (green), the FH of cardiac myosin is rotated 22° anticlockwise compared to the FH of SmMyo2. (b) Comparison of the head-head interfaces in SmMyo2 and cardiac myosin, two views are represented. On the left, aligned on the FH motor domain, on the right, aligned on the BH motor domain. The cardiac IHM is represented in transparency. Note the difference in the HCM-loop on the left, and on the _FHRelay (yellow) and _FHConverter positions (Blue) on the right. (c) Sequence alignment of the different connectors and elements involved in the interactions showing the divergence: Card_Hs (Homo sapiens β-cardiac myosin); Sm_Gg (Gallus gallus smooth muscle myosin); Sk_Hs (Homo sapiens fast skeletal muscle myosin); NM2a_Hs (Homo sapiens nonmuscle myosin 2a). Regions of interactions are in colored box: conserved in red, divergent in green. (d) Comparison of the position of the coiled-coils on the surface of the heads between cardiac myosin and SmMyo2. (e) Comparison of the RLC-RLC interface, the models are aligned on the IQ2 of the FH. A sequence alignment of the N-term extension of the RLCs is shown with conserved regions between smooth muscle RLC (SmRLC) and cardiac RLC (CarRLC) boxed in red. The phosphorylation is represented in red with “P”. (f) Comparison of the angles at the hinges of flexibility: pliant region, aligned on the Converter (on the left), FH and BH lever arm, aligned on IQ1 (on the right).

The second major difference is the positions of the S2 coiled-coil (S2). In the SmMyo2 IHM structure, S2 is located close to the FH head and the IHM is strengthened by interaction with two additional coiled-coil segments (Seg2 and Seg3). No such segments stabilize β-cardiac myosin, rendering the IHM motif less stable when isolated from filaments^13,14,15. Surprisingly, the cardiac S2 interacts extensively with the Mesa of the BH, in drastic contrast to the position S2 adopts in the SmMyo2 IHM, which only allows for a few interactions with the _FHLoop2. In fact, it is Seg3 that interacts with the BH Mesa in SmMyo2 IHM (Fig. 3d).

More importantly, the detailed comparison of the RLC/RLC interface reveals the structural differences underlying the distinct modes of regulation of the cardiac and SmMyo2 IHM formation. Both _CarRLC and _SmRLC are phosphorylatable in a charged and disordered N-terminal extension (N-term extension) which is not conserved in sequence (Fig. 3e), although from the globular N-terminal lobe of the RLC, the position of the two phosphorylatable serines S19 in _SmRLC and S15 in _CarRLC is conserved (Fig. 3e). In SmMyo2, RLC phosphorylation acts as an on/off switch to disrupt the SRX and activate the muscle³². This was explained in the Smooth IHM structures^13,15where the _BHRLC N-term extension interacts with seg3 and the _FHRLC N-term extension is directly part of the RLC/RLC interface. S19 Phosphorylation thus disrupts these IHM interfaces, switching on the motor (Fig. 3e). In cardiac myosin, RLC phosphorylation is not strictly necessary to activate cardiac muscle but modulates its activity³³). The cardiac IHM structure explains this more moderate effect on IHM interfaces, beyond the fact that seg3 is not part of stabilizing the cardiac IHM. The RLC/RLC interface is drastically different in the cardiac IHM: the _BHRLC extension is not part of it and the _FHRLC N-term extension is only found at the periphery of this interface (Fig. 3e). No density allows precise positioning of the _FHRLC S15 as it is not a major part of the RLC/RLC interactions. Thus, phosphorylation would only modulate the stability of the IHM in cardiac. This major structural difference in how the smooth and cardiac IHM form underlies the distinct regulation in striated and smooth muscles.

Finally, the differences in the hinges of flexibility of cardiac and SmMyo2 myosins are required to allow the distinct but precise asymmetric orientation of the heads in the two IHMs. The flexibility and the kink at the pliant region are much larger in the BH head compared to the FH, which adopts a relatively straight lever arm. However, the BH pliant region is 10° more kinked in SmMyo2 compared to cardiac (Fig. 3f). The kinks of the HC at the IQ1/IQ2 region are different in the two IHMs. There is an angular difference of 5° between smooth and cardiac FH heads and 30° between the BH heads respectively (Fig. 3f). The ELC/RLC interactions are thus different both for the BH and FH heads. This validates that these regions are specific ankles of flexibility required for the formation of the IHM (Fig. 3f).

Consequences on muscle physiology

In the thick filament, the IHM motifs are organized in crowns with different symmetries depending on the muscle and organism considered¹². This different organization implies distinct inter-crown interactions that could not be resolved precisely with low-resolution models of the relaxed thick filaments^34,31 The 28 Å EM structure of the human cardiac filament revealed repeats of three crowns, crown 2 being less ordered¹⁰. Regarding the tilt or orientation of each crown, crown 1 is rotated 19° in a clockwise direction as viewed from outside the filament compared to crowns 3 and 2, which are similar¹⁰ ( Fig. 4a). This allows a strong interface to be formed between crowns 1 and 3, but not between crowns 3 and 2. The distance between the crown 2 and crown 1 heads is too large for interactions to be formed, thus resulting in a perturbation of the helical symmetry^35,36,37.

Figure 4 – Organization of the thick filament.

(a) Fit of the IHM motifs in crowns 1 and 3 using the 28 Å resolution map of the relaxed human cardiac thick filament (EMD-2240¹⁵). Crown 2 is less stabilized and less defined in density. A tilt of 19° is observed for crown 1 compared to crowns 2 and 3. (b) Zoom in the interface between crown 1 and 3 that was previously thought to involve mainly the RLC of crown 3. Our model indicates that the ELC N-lobe and the N-terminal extensions (Nterm extensions) of both the ELC and the RLC of crown 3 could interact with the L50 HTH and loop 3 of crown 1. The 49 aminoacid Nterm extension of the ELC can reach the activation loop (A-loop) and the helix-turn-helix (HTH). (c) Analysis of the sequence conservation of the regions involved in the inter-crown interface. (d) Schematic overview of the differences stabilizing the sequestered states in striated and smooth muscle. In cardiac muscle (upper panel), the inter-crown interface stabilizes the IHM of crowns 1 and 3, while crown 2 is less stabilized and less defined in the density because (i) it is not oriented to establish this interface and (ii) the crown 2/crown 3 inter-crown distance is longer compared to crown 3/crown 1. A fit of the IHM cardiac model in the purely helical relaxed cardiac skeletal muscle³⁸. The helical symmetry induces different inter-crown interfaces compared to cardiac relaxed filament.

We placed the high-resolution structure of the cardiac IHM in the thick filament electron density designated for crown 1 and 3¹⁰, which allowed us to describe the precise structural elements involved in the most stable inter-crown interface (Fig. 4a). Interestingly, crown 1 and crown 3 heads can interact via three actin-binding elements of the FH head of crown 1, (i.e. the helix-turn-helix (HTH), activation loop (A-loop) and loop-3), and the N-terminal extensions of the _BHRLC and the _BHELC of crown 3 (Fig. 4b). These extended regions contain several charged residues that can mediate long-range interactions and concur in stabilization of the two crowns in their respective orientation (Fig. 4b). Sequence alignment of elements of the interface confirm a specific conservation between striated muscle myosins (cardiac and skeletal) that is not found in smooth muscle and non-muscle Myo2s (Fig. 4c). The ELC that binds SmMyo2 and NM2a have no N-terminal extension and their RLCs have longer and divergent N-term extensions compared to cardiac RLC (Fig. 4c). As was previously suggested from a low resolution model of the relaxed muscle thick filament¹⁰, our model shows that the RLC is part of the Crown1/Crown3 interface, providing now in addition a potential role for the N-terminal lobe of cardiac ELC for stabilizing these IHMs. The BH RLC S15 phosphorylatable residue would also be found close to the inter-crown interface, suggesting how phosphorylation of the crown 3 BH RLC head might weaken the cardiac IHM stability and modulate the number of active myosin heads. In other muscles with different symmetry for the thick filament (e.g. tarantula skeletal muscle which harbors a perfect helical symmetry³⁸, the inter-crown interface differs and does not involve the RLC (Fig. 4d; Extended data Fig. 9). Thick filament geometry may thus induce different types of inter-crown regulations.

The N-terminal extension of _CarELC is also involved in direct interaction with actin and increases cross-bridge kinetics (reviewed by ³⁹). Sequestering it in the inter-crown interface would be an additional mechanism to block interaction with actin. Interestingly, most of the elements involved in actin binding are part of either the intra-molecular interactions within an IHM or the inter-crown interface. The inter-crown stabilization interactions challenges the concept of a free-head⁸ and the idea that the IHM of cardiac myosin is less stable than that of SmMyo2. In the cardiac thick filament, indeed stabilizing partners such as MyBP-C or Titin¹⁰ could further stabilize the heads and the inter-crown interactions. Thus, as many as half of the cardiac heads may be shut down in the off state even during systolic contraction of the heart^10,6. In contrast, strain upon contraction could disrupt these thick filaments stabilization interactions and would increase the number of heads recruited for optimal heart contraction.

Inherited cardiomyopathies and therapies

One of the major predictions to explain the hyper-contractility of the cardiomyocytes during the early stages of hypertrophic cardiomyopathies (HCM) is the increase in the number of heads available for contraction¹⁸, and specifically due to the destabilization of the IHM by point mutations responsible for the disease^28,21,40,27. Exploiting structural knowledge of the myosin allostery promoting force production and low-resolution models of the IHM, HCM mutations have been grouped in different classes²⁷ to predict how mutations may influence contractility. Consistent with this hypothesis, many HCM mutations studied with purified human β-cardiac myosin constructs have indeed been shown to directly decrease the SRX and diminish the stability of the IHM^41,17,28,42Differences in the number of heads available to interact with actin are thus likely to be a major cause of the disease, although each mutation can also affect differently the intrinsic force produced, the ATPase rate, and/or the velocity of movement of actin over myosin in the in vitro motility assay^{21,20,41,17,28,42}.

The cardiac IHM structure shows that a number of mutations indeed lie close to the IHM stabilizing interfaces (Extended data Fig. 10). With this long awaited structure, the prediction for mutations in critical regions such as the Converter can be clarified (Extended data 1), and a more accurate prediction of the effects of HCM and DCM mutations can be released (Extended data 1). This high-resolution structure now provides a precise distinction between residues directly involved in the intra-molecular IHM interactions (Extended data Table 2, Extended data 1) and those found nearby that affect these interfaces indirectly via their influence on the conformation of critical structural elements required for the stabilization of the IHM. The high-resolution IHM structure provides also details on how mutations in the light chains or at the main interfaces can disrupt the IHM, that could not previously have been interpreted. Previous low-resolution models used for mutation prediction⁴²were all unable to model correctly the hinge of flexibilities and light chains and how mutations could affect the inter-crown interface. An example here is _ELCA57G, a severe infantile mutation in the ELC light chain⁴³ that leads to a decrease of the SRX in mouse cardiac fibers⁴⁴. The IHM structure indicates how the replacement of A57 to a glycine would destabilize helix _ELCα1 and _ELCLoop1, and thus the ELC/RLC interface. In fact, these ELC structural elements are crucial for stabilizing distinctly the hinges of the two lever arms of the IHM. In addition, this _ELCA57G mutation can directly affect the inter-crown interactions (Extended data 1, Fig. 5a). Our analysis also indicates that the inter-crown and the RLC/RLC interfaces will also be weakened by HCM mutations in the RLC N-term extension, such as A13T, F18L and E22K (reviewed by ⁴³) (Extended data 1; Fig. 5a). Mutations in the actin binding elements of the L50 (Loop3, activation-loop (A-loop) or helix-turn-helix (HTH)), initially interpreted as disrupting motor function, are in fact in the inter-crown interface and could thus also dysregulate the number of heads participating in contraction (Fig. 5a). Interestingly, while six HCM mutations (two located in the A-loop, two in loop-3) are predicted to alter the inter-crown interface, this region is enriched with five DCM mutations (Fig. 5a). The consideration of the inter-crown interface leads to the definition of a new class of cardiomyopathy causing mutations, which mainly affect the IHM stability by its stabilization on the filament level via interactions between IHM crowns or with thick filament associated proteins such as MyBP-C and titin. Higher resolution structures of the thick filament will be required to further analyze the effect of such mutations.

Figure 5 – Molecular determinisms underlying hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathies (DCM).

(a) DCM (circled in red) and HCM causing mutations predicted to destabilize the inter-crown interface or the IHM are represented on a cartoon representation of crown 1 and 3. Zoom at the inter-crown. (b) and (c) show a superimposition of the structure of the S1 fragment of cardiac myosin bound to the activator Omecamtiv mecarbil²⁴ (PDB code 5N69, PPS-S1-OM) superimposed on the motor domain of each head FH and BH respectively. PPS-S1-OM is represented as a black ribbon and shows how OM induces an orientation of the lever arm that is not compatible with the requested orientations for the FH or the BH when the IHM is formed.

Small molecules modulating cardiac myosin force production are among the most promising therapies for inherited cardiac diseases. The activator Omecamtiv mecarbil (OM) was reported to destabilize the IHM²². The OM-bound cardiac myosin structure ²⁴ showed that OM binds in a pocket of the cardiac pre-powerstroke state located between the motor domain and the Converter. The closure of this pocket upon OM binding introduces a hinge that shifts the Converter orientation prior to the pliant region. This now makes these two OM-bound heads incompatible with the sequestered state (Fig. 5b, 5c), explaining the reduction in the IHM formation in the presence of this activator. The IHM structure will be essential to identify different classes of inhibitors and whether they are compatible or not with the formation of the IHM depending on which pockets they bind to.

Conclusions

The high-resolution structure of the sequestered state of β-cardiac myosin is paramount for understanding thick filament regulation. The unexpected divergence with the SmMyo2 IHM structure reveals that while myosin regulation is conserved in evolution, distinct sequences have evolved from a common ancestor to fulfill the requirements of distinct fiber types and muscle functions (Fig. 4d). Interestingly, most of the interfaces involve actin-binding elements that are variable and these differences were reported to tune (i) myosin force production and (ii) the orientation on the actin tracks^1,45,46. It is tempting to speculate that actin-binding elements are part of a super-regulation mechanism, also regulating the sequestered state stability. Future high-resolution studies of other IHMs will provide insights to decipher how the divergence of the IHM motif evolved to also fit different types of Ca²⁺ or phosphorylation regulation, together with fulfilling different types of muscle contraction properties.

The IHM structure provides the missing puzzle piece to understand the molecular mechanism of inherited cardiomyopathies. As for β-cardiac myosin, MyBP-C is also a hotspot of HCM mutations⁴⁷. It was reported to increase the SRX through direct interactions with cardiac myosin^48,28. MyBP-C and Titin are positioned to influence the myosin crowns in the relaxed thick filament¹⁰. Future structural studies with these partners will further complete the modelling of how mutations can alter SRX regulation. With a near-atomic model of the IHM, the mode of action of future treatments of cardiomyopathies can also be deciphered more precisely. The availability of the cardiac IHM structure will pave the road towards the rational design of novel therapeutics against distinct families of inherited muscle disease.

Materials and methods

15-hep production and purification.

The human β-cardiac myosin construct 15-hep HMM was produced using a modified AdEasy^™ Vector System (Qbiogene, Inc, Carlsbad, California, USA). The 15-hep cDNA consists of residue 1 to 942 of MYH7, encompassing S1 and the first 15 heptad repeats of the proximal S2 coiled-coil region. This sequence is followed by a GCN4 leucine zipper to ensure dimerization, a flexible GSG linker, and finally a carboxy-terminal PDZ-binding octa-peptide (RGSIDTWV). The adenoviral vectors expressing the heavy chain myosin and the human ventricular essential light chain (ELC) (MYH3) containing a TEV protease cleavable N-terminal FLAG tag (DYKDDDDK) were co-expressed in differentiated mouse myoblast C2C12 cells (purchased from ATCC). C2C12 cells were infected with the adenoviral vectors 48–60 hours after differentiation and harvested 4 days post infection in a lysis buffer with the following composition: 20 mM imidazole at pH 7.5 containing 50 mM NaCl, 20 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 10% sucrose, 1 mM DTT, 3 mM ATP, 1 mM PMSF and Roche protease inhibitors. Harvested cells were immediately flash frozen in liquid nitrogen and stored at −80°C.

For purification, frozen cell pellets were thawed at room temperature, and supplemented with 3 mM ATP, 1 mM DTT, 1 mM PMSF, 0.5% Tween-20 and Roche protease inhibitors before lysis with 50 strokes of a dounce homogenizer on ice. The lysate was incubated on ice for 20–30 mins to allow for the assembly of native mouse full length myosin into filaments before clarification; filament formation is promoted by the high MgCl₂ and low salt content of the lysis buffer. The lysate was clarified by spinning at 30,000 RPM in a Ti-60 fixed-angle ultracentrifuge rotor for 30 minutes at 4°C. All subsequent steps were carried out in a cold room set at 4°C. The supernatant was incubated on a nutating rocker with anti-FLAG resin for 1–2 hours, followed by low-speed centrifugation to settle down the protein-bound resin. The lysate was decanted, and protein-bound resin was washed with >10-fold excess of a wash buffer (20 mM imidazole at pH 7.5, 5 mM MgCl2, 150 mM NaCl, 10% sucrose, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 3 mM ATP, 1 mM PMSF and Roche protease inhibitors). Native mouse regulatory light chain (RLC) was then depleted by incubating the resin with RLC depletion buffer (20 mM Tris-HCl at pH 7.5, 200 mM KCl, 5 mM CDTA pH 8.0, 0.5% Triton-X-100 supplemented with 1 mM ATP, 1 mM PMSF, and Roche protease inhibitors) for 75 mins on a nutating rocker. The resin was washed with wash buffer and incubated with human RLC, purified from E. coli as previously described⁴, for ~3 hours. Finally, unbound human RLC was washed away, and the resin was nutated overnight in wash buffer containing TEV protease to cleave the hexameric myosin assembly from the resin. The next morning, the protein was further purified using anion-exchange chromatography. The myosin in solution was separated from the stripped-off resin using a Micro Bio-Spin^™ Chromatography Column (Bio-Rad) and loaded on a 1 mL HiTrap Q HP column (Cytiva) attached to an FPLC. The protein was eluted with a gradient of 0–600 mM NaCl (in a buffer containing 10 mM imidazole at pH 7.5, 4 mM MgCl₂, 10% sucrose, 1 mM DTT and 2 mM ATP) over 20 column volumes. The fractions corresponding to the protein peak were run on a 12.5% SDS-PAGE and pure fractions (devoid of the contaminating full length mouse myosin) were pooled. The protein was concentrated to ~10 mg/ml using Amicon Ultra centrifugal filter units, and flash frozen as 20–40 uL aliquots.

Single ATP turnover experiments from Jim Spudich lab (manuscript in preparation) and others⁷ confirm that 15 heptads of the proximal S2 tail allow myosin to adopt the folded-back SRX state in solution.

Cryo-EM data collection and processing.

The sample, was diluted at 0.3 mg/ml concentration in 10 mM imidazole pH 7.5, 2 mM MgCl₂. 0.5 mM Mg.ATP was added and the mix was incubated 15 minutes at room temperature. 3 μl was applied to glow discharged UltrAuFoil 1.2/1.3 holey gold grid and vitrified in a Mark IV Vitrobot (ThermoFisher). Grids were imaged in ESRF’s CM01 facility using a Titan Krios microscope (ThermoFisher) at 300 keV ⁸ with a K3 direct electron camera at 0.84 Å per pixel. 5154 movies were collected with 40 frames each and a dose of 0.98 e−/Ų per frame. After beam-induced motion correction with Motioncor2⁹ and Contrast transfer function (CTF) estimation with Gctf¹⁰, 4423 micrographs were selected for further particle analysis performed in cryoSPARC¹¹. A total of 493179 particles were selected after automatic picking and the first round of 2D classifications. These particles were used to generate several ab initio reconstructed 3D models that were subsequentially used for a 3D heterogeneous refinement. At the end of this first round of refinement, 213596 particles were unambiguously attributed to the sequestered state. These particles were further processed with 3D refinement, using the cryoSPARC homogeneous refinement algorithm¹¹, giving final global resolutions based on the gold-standard Fourier shell correlation (FSC = 0.143) criterion¹² of 3.6 Å (map 1). To further improve the resolution of the heads region, a homogeneous refinement using a mask covering the heads region alone was carried out. This final refinement produced a 3.2 Å resolution map of this region (map 2). Local resolutions of the density maps were calculated in cryoSPARC¹¹.

Model building and refinement.

The head/Converter/ELC region was rebuilt based on the bovine β-cardiac myosin PPS complexed to Omecamtiv mecarbil² (PDB code 6N69) without solvent or ligand. The coiled-coil S2 region was fitted from the crystal structure¹³ (PDB code 2FXM). The IQ2/RLC region was obtained by homology modeling based on the SmMyo2 IHM¹ (PDB code 7MF3). Each part was fitted in the density map and the model was built and optimized with the module Quick MD Simulator from CHARMM-GUI^14,15. Steps of real space refinement were performed against map 1 with Phenix¹⁶. During the procedure, the head region (residues 1–781) which was at higher resolution was subjected to reciprocal space refinement against map 2 with Refmac¹⁷. An independent run of molecular dynamics was performed in order to improve the statistics at the interfaces and to compute realistic interactions. The system was built with the CHARMM-GUI/Quick MD simulator module^14,15. Only the interfaces (head/head, coiled-coil/head) of the model were relaxed in a box containing explicit water (TIP3P) and salt (150 mM KCl) in the CHARMM36m¹⁸ force field. 10 ns of simulation was performed in GROMACS (version 2018.3)¹⁹. A final run of real space refinement was performed against map 1. Quality indicators were calculated with Phenix, including Molprobity²⁰ and EMRinger²¹.

Interface analysis.

The analysis of the interfaces was performed automatically with PDBsum²². Interactions were manually checked and extended up to 4.5 Å to discuss long range weak interactions. We chose to discuss the realistic interfaces of the IHM because we computed it with molecular dynamics. While some ambiguity exists at times in these interfaces, the electron density is good for the main chain and the large hydrophobic side-chains. As discussed in the text, most of the interfaces of cardiac IHM contain however labile side-chains (musical chairs) that can adopt more than one conformation. Lower EM density for such side chains are then apparent, even in the high-resolution regions such as the head-head regions.

Extended data 1:

With a detailed analysis of the high-resolution structure of the β-cardiac IHM, we can provide here a detailed evaluation of the effects of cardiomyopathy mutations. The precise analysis of these mutations on the interfaces was impossible in the absence of a high-resolution structure for the IHM. The fact that mutations can be located in connectors such as the Transducer or the Converter also complicate their analysis as they have an impact in motor function⁴⁶. With the IHM structure, the impact on the stabilization of the heads can be precisely assessed as we have access to the conformation these flexible elements adopt to form the IHM interfaces. With a detailed analysis of a few HCM mutations and the structural impact for the IHM and for the function of force generation, we can now more clearly distinguish the different classes of mutations that can lead to the cardiomyopathies.

Mutations in the heads

In the last years, several mutations located in the motor domain have been analyzed in the Spudich laboratory and in other groups. Some mutations were directly reported to decrease the SRX state: R249Q; H251N; D382Y; R403Q; V606M; P710R and R719W^{34,4,47,23,48}. The IHM structure now provides detailed structural insights to identify how mutations can lead to the development of the HCM phenotype.

Some of these mutations are located at the interfaces and would thus directly impact the IHM stabilization:

R403Q is a hotspot of mutations^46,49, which would directly affect the head-head interface by destabilizing contacts between the _BHHCM-loop, the _FHHO- and the _FHHD-linkers (see main text and Figure 2b; Extended data Table 2). This explains the significant decrease of the SRX measured with this mutant^23,47. Even if R403 is not directly involved in the actomyosin interface in Rigor⁵⁰, the alteration of the local structure and dynamics of the loop is responsible for the 2-fold reduced binding affinity of the mutant for the actin track⁵¹.

D382Y is not located on the Mesa but on the so-called _BHPHHIS surface of the _BHU50 subdomain. This surface residue is directly involved in polar interactions with the _FHHD-linker (Figure 2b; Extended data Table 2) and we predict that the mutation would have little effect on the structure of the head itself. The fact that the IHM structure predicts that the mutation would destabilize the head-head interface is consistent with data indicating a 46% reduction of the SRX³⁴, with only modest alteration of the mechanochemical properties of the motor carrying the mutations⁵².

R453C (HO-linker) is also a direct part of the interface between the _BHmotor domain and the S2 coiled-coil (Figure 2c; Extended data Fig. 10a). Interestingly, this mutation would greatly affect a second interface of the IHM when present in the FH head: the interactions between the HO-linker of the Mesa of the FH and the _BHHCM-loop (Extended data Table 2). While the effect of the mutation on the SRX has not been measured yet, it decreases the affinity of the S1 fragment for the S2 coiled-coil⁴, consistent with our IHM structure. Since this mutation is located in a linker of the Transducer, it is predicted to alter the structural transitions between the states and thus it is not surprising that it changes motor properties²⁹.

Some of the mutations are not located at the interface but close enough to induce indirect effects and destabilize the IHM:

H251N and R249Q, two mutations of the β-bulge of the Transducer, lead to a 41% and 59% decrease of the SRX as well as a decrease in the affinity of the S1 head for the S2 coiled-coil ^4,34. Given the essentiality of the Transducer in conformational changes during the motor-cycle, the role of these mutations in force generation has been demonstrated for H251N⁵³ and is also likely for R249Q⁵⁴ although this has not yet been studied in human. The R249 and H251 side chains are in the vicinity but are not involved in direct interactions stabilizing the IHM (Extended data Fig. 10). In the FH head, this Transducer region interacts with the BH-HCM loop (near _BH-HCM-loopE409). In the BH head, the Transducer interacts with the S2 coiled-coil. Destabilization of the IHM may result from local changes in the dynamics of main chain or side chain of the β-bulge, which would then impact the contacts made in the IHM and its stability on two interfaces that involve the Transducer: _BH-Tranducer with the S2 coiled-coil and _FH-Transducer with the HCM-loop. Since H251 and R249 are involved in polar interactions between strands of the Transducer, the presence of the Asn or a Gln side chain instead of H251 or R249 may also alter the stability of the pre-powerstroke state which would indirectly affect the formation and stability of the IHM, as previously proposed⁴⁶. Thus, the structure of the IHM provides much more detailed hypotheses of the role of the mutations than previously possible, which can guide further investigations.

R663H, located at the end of the HW helix on the surface of the myosin head, is close to the Mesas of the FH and the BH. Its surface location would predict that a mutation at this position should not affect the mechano-chemical parameters of the motor. Kinetics experiments demonstrate that this mutation has indeed no effect on the fundamental parameters of the myosin motor, while a major reduction in the SRX is found²³. The BH R663 side chain in the IHM is close to the _BHhead/coiled-coil (_S2D896/D900) interface while the FH R663 side chain is close to the _FHhead/loop4 (_BH-loop4E370) interface, but the current cryoEM map density does not support direct interaction. The charge-removal introduced by the mutation would change the global charge of both Mesa and thus likely remotely affect the IHM formation, possibly also by affecting how this residue interacts with other regulatory proteins of the thick filament.

R719W is located in the Converter, in close vicinity of the Top-loop, a key-element in the head/head interface. While previously debated among IHM models, the IHM structure clearly defines how the FH-Converter is positioned to stabilize interactions with the BH head. This mutation is predicted to destabilize the Top-loop and alter its dynamics and conformation, thus altering the BHPHHIS/_FHConverter interface, resulting in a 68% decrease of SRX^34,47. Interestingly, this mutant only has modest effects on kinetics parameters of the S1⁵² while a 20% decrease of intrinsic force was reported⁵⁵. This is consistent with the location of the mutation that would not disrupt the Converter swing but would alter the Converter/ELC interface and thus the flexibility of the lever arm³ (Robert-Paganin et al., 2018).

R723G and G741R are also close to the Top-loop and would have similar outcomes to R719W. This is confirmed by biochemical assays demonstrating that these mutants have only small effect on S1 mechanochemical properties⁵⁵. Thus, the effects of these mutations are likely only on the BHPHHIS/_FHConverter interface of the IHM by altering the conformation of the Top-Loop, as previously proposed via molecular dynamics predictions of the Converter⁴⁶. However, the quantitative effect of such mutations on the SRX are not yet available.

Finally, a group of mutations affecting the SRX are located far from the interfaces but in regions important for the allosteric conformational changes of the motor upon force production:

P710R is located in the core junction of the Converter, far from all interfaces that stabilize the IHM. However, it can greatly alter the conformation of the SH1-helix leading to drastic alterations of the allostery within the motor. In addition, it can also alter the stability or the conformation of the PPS state that both heads must adopt to form a stable IHM. We predict that in this case, the alteration of the PPS state could also lead to a change in the lever arm position that could result in difficulty in forming the IHM conformation. Thus, the mutation leads to both a decrease of the SRX and a dramatic alteration of motor functions (Extended data Fig. 10a)⁴⁸.

V606M is also far from the interface but located in the U50 core close to the strut and the actin binding cleft and also close to Loop2, an element essential in the first step of actin binding (Extended data Fig. 10a) ⁵⁶. The replacement of a Val by the bulkier Met may alter the structural rearrangements during the powerstroke and the actin binding cleft opening and closure. This may induce severe consequences on mechano-chemical properties of the motor such as association to actin and force output. The V606M also decreases the SRX ⁴⁷ and this effect is likely due to both a potential alteration of the PPS state but also on the nearby _BHPHHIS surface.

As proposed previously, the mutations in the heads can have effects on IHM stability and/or on the ability of the motor to produce force⁴⁶. These two parameters should be taken into account when discussing the molecular consequence of inherited cardiomyopathy mutations. Interestingly, some of the mutations analyzed here are located in the Transducer or the Converter, central elements in force production and allosteric communication of the motor. HCM mutations would thus have two ways of destabilizing the IHM: (i) direct effect on the interfaces or on the conformation of elements of the interfaces and (ii) remote effect on the stability of the PPS. Since the formation of the IHM requires the two heads in PPS, case (ii) alters the ability of β-cardiac myosin to form the IHM, confirming now a previous suggestion based on the hypothesis that the heads would remain in the PPS state while forming the IHM⁴⁶.

Mutations in the lever arm helix and the light chains

Studies of the mutations in the light-chains and the lever arm have been performed (reviewed by ^45,43). HCM causing mutations in the light chains are generally considered as rare: 1% in the RLC and <1% in the ELC⁵⁷. Mutations in the heavy chain from the Pliant and IQ regions were also recently studied: D778V, L781P and S782N in the Pliant; A797T in IQ1 and F834L in IQ2⁵⁸ (Extended data Fig. 10b, 10c). Importantly, none of these mutations have effects on light-chain loading to the heavy chain. The three Pliant mutations have no significant effect for the SRX but alter the mechano-chemical parameters of the myosin⁵⁸. In contrast, mutations A797T and F834L have modest or no effect on mechano-chemical parameters but significantly reduce the SRX⁵⁸. The replacement of the small Ala797 by a bulkier Threonine is predicted to add restrains on the orientation of _ELCα2-helix and may influence the orientation dynamics of the entire N-terminal lobe of the ELC, thus weakening the ELC/RLC interface by long-range effects. In IQ2, the replacement of the aromatic Phe834 with the smaller Leu could have effects on the RLC/RLC interface and may alter the local orientation of the helices involved in this interface, thus weakening it. Both of these mutations in the IQs are predicted to have long-range effects as they would impact the formation of the hinges required in the lever arm for the formation of the IHM.

Mutations in the S2 coiled-coil

Finally, the effect of HCM mutations located in the S2 coiled-coil on the SRX has not yet been investigated. However, the effect of two mutations on the affinity of the S1 for the S2 was evaluated: R870H and D906G⁴. While R870H has no effect on the association of the S1 and S2, D906G reduces the affinity⁴. In our IHM structure, the R870H is located far from the interface and would not affect the coiled-coiled helices association (Extended data Fig. 9a). Modeling of the coiled coil indicates that this is not the case for the D906G mutation which could be relatively close to the interface. The addition of a Gly could weaken the helical secondary structure of the S2 coiled-coil and thus weaken the IHM interfaces with long-range repercussions. A recent study⁵⁹ indicates in fact that R870H has an increase in affinity for phosphorylated MyBP-C, which could lead to the HCM disease by preventing the proper role of phosphorylated MyBPC-C as well as destabilizing heads in the IHM when MyBP-C is phosphorylated.

Conclusions

The structure of the IHM at high-resolution solved in this work enables prediction and analysis of the effect of point mutations that cause HCM. It is an essential tool to understand the molecular basis of cardiomyopathy development. As proposed previously, the point mutations could have different effects on motor function and/or stability of the sequestered state ⁴⁶. Interestingly, the mutations altering the stability of the sequestered state can be classified in different categories: (i) those directly altering the interface; (ii) mutations with indirect effect on the interface, destabilizing elements involved in the interfaces; (iii) mutations located in connectors and subdomains essential to stabilize states with lever arm up, with long range effects. The structure will be of high utility to discuss the outcome of HCM and DCM mutations. Structural studies and dynamic simulations of mutants can be performed based on the protocol or structure this study has provided. This will be essential to reveal how mutations affect force production and the stability of the auto-inhibited state. This knowledge will be beneficial to the patient, as predicting the effect mutations have on motor activity or SRX stability will be critically needed as different types of allosteric modulators of cardiac myosin function become available.

Extended Data

Extended data Figure 1 – CryoEM map and validation.

(a) Shows the steps of processing. A representative micrograph on gold grids, 2D classes and the steps of refinement and masking. (b) and (c) show the Fourier Shell Correlation and the overall resolution on Map 1 and 2 respectively (criterion 0.143). (d) and (e) display the local resolution for each map with two orientations. (e) Shows the flexibility within the different classes used to reconstitute the map. (f) plots the different orientations of the particles.

Extended data Figure 2 – Segments 2 (Seg2) and 3 (Seg3) coiled-coils provide additional stability in the Smooth muscle myosin 2 (SmMyo2) IHM structure.

Seg2 interacts with the ELC from the blocked head (BH) (interface 1) and Seg3 with the N-terminal extension (N-term ext) from the RLC from the BH (interface 2). The phosphorylatable serine (S19) in the _BHRLC is part of the interface with the Seg3. IHM model used here: PDB code 7MF3¹.

Extended data Figure 3 – Comparison of the sequestered state with the pre-powerstroke (PPS) state of cardiac myosin.

(a) The blocked head (BH) and the free head (FH) superimpose on the motor domain with a RMSD of 0.7 Å, the only difference being the kink at the Pliant region in the lever arm, resulting in an angle of 53° between the IQ1 helix of the BH and the FH heads. (b) and (c) show the superimposition of the BH and of the FH respectively with the apo cardiac MD structure in the PPS state (black lines, MD-PPS-Apo, PDB code 5N6A,²). Both motor domains superimpose with MD-PPS-Apo with a rmsd of 0.84 Å. The arrows show the ELC/Converter interface which is altered by the kink in the BH. (d) In the FH, the Converter/ELC interface is maintained by “musical chairs”, a set of labile charged residues located on the ELC (deep teal cyan) or in the Converter, more specifically in the Top-loop³ (orange). (e) and (f) compare the motor/ELC interface and how it is completely altered by the kink present in the BH, the musical chairs are represented on both panels. The kink of the BH shifts the position of the helix. In the BH, the interface is altered by the kink, generating a new interface between the motor (green with Top-Loop in light orange) and the ELC (purple). (g) BH, FH and PPS-S1-OM have all a closed active site in the inner cleft region.

Extended data Figure 4 – Comparison of the IHM structure with the previous homology pseudo-atomic models.

(a) Overall representation of the IHM structure of cardiac myosin with two models, MA1³ and MS03⁴. The pliant region of the BH is shown with an arrow. (b) Zoom in the head-head interface region showing that the same elements are involved in the interface, although the interface was not correctly modeled in the details. (c) S2 coiled-coil in the structure and in the models interacts with the BH. (d) and (e) compare the conformation of the lever arm of the BH and the FH respectively in the structure and in the models. The regions where the ELC and the RLC bind are indicated. (f) Comparison of the orientation of the Relay and the Converter in the BH. The structure (left) is superimposed with the two models (middle, right) for comparison. Note the pre-prestroke position of the Relay (brown) and the Converter (brown, red) in the MS03 model. In (a) and (b), the structures are superimposed on the entire model, in (c), the structures are superimposed on both heads (which include both Motor domains and the lever arms including the two LCs), in (d) and (e) the lever arms are superimposed on the Converter (residues 708–777), in (f), the structures are superimposed on the N-term subdomain (residues 3–202).

Extended data Figure 5 – Elements and connectors of β-cardiac myosin involved in the interfaces stabilizing the IHM.

The different elements are represented on the blocked head of cardiac IHM. The nucleotide indicates the position of the active site while elements of the actin interface are represented : HTH: helix-turn-helix, A-loop: activation-loop, HCM-loop, Loop-4 and Loop-3. Also represented are Sw-2 (switch-2), IQ1 helix: HC sequence involved in binding the ELC after the pliant region. The Transducer (blue) corresponds to the central β-sheet as well as structural elements linked to this β-sheet (β-bulge, HO-linker, HD-linker). The PHHIS is part of the U50 subdomain while the HTH and Loop-3 are part of the L50 subdomains.

Extended data Figure 6 – Definition of the myosin “Mesa” and elements involved in the interfaces of the IHM, as defined by⁵.

Surface HCM mutations are shown as balls and colored depending on the subdomain where they are located. The Mesa was defined as a large and flat surface containing conserved residues and HCM mutations (in purple dotted lines). The Primary Head-Head Interaction Site (PHHIS) and the actin-binding surface are shown with red and orange dotted lines respectively.

Extended data Figure 7 – Asymmetry in the light chain binding regions.

(a) Interactions at the ELC/RLC interface in the free head (FH) differ from the ELC/RLC interface in the blocked head (BH) (Fig. 2d). (b) Lever arm of the BH and of the FH aligned on IQ2 reveals a 18° difference in the orientation as well as major differences in the ELC/RLC interface. (c) Asymmetry in the conformation and contacts of the two RLCs in the IHM structure. The RLC/IQ2 of the BH and FH heads see a different environment. This is specifically true for the region 834–848 of the heavy chain (HC), due to the fact that the coiled-coil triggers a shift in position of the HC residues (arrows between the M849 position). The HC residues (834–848) thus adopt drastically different conformations (yellow for BH, dark blue for FH) that are critical to form the RLC/RLC interface. This shift explains why some residues have different environments in the FH and in the BH: two residues characteristic of this phenomenon are contoured in black, R845 and E846. After residue R846, the phase of the coiled-coil is more canonical. Interestingly, the _BHR845 residue is involved in electrostatic interactions with the _FH-RLCE31 residue, while the _FHR845 residue is buried in the _FH-RLCN-lobe core. See also Movie 3.

Extended data Figure 8 – Electrostatic profile of the footprint of IHM interfaces.

The different surfaces involved in the interactions stabilizing the IHM are represented as electrostatic surfaces. The regions involved in the interactions are indicated.

Extended data Figure 9 – Comparison of the inter-crown interfaces between human β-cardiac myosin IHM (a) and tarantula skeletal myosin IHM (b).

In order to make the comparison, two IHM of β-cardiac myosin were fitted in the low-resolution map of tarantula skeletal myosin IHM (EMDB-1950⁶). The difference in the symmetry of cardiac and tarantula skeletal filaments (pseudo-helical and purely helical respectively) induces different inter-crown interfaces. In tarantula skeletal relaxed filament, the ELC from Crown 3 interacts with elements of the actin binding interface from the adjacent Crown 1: Loop-4, HCM-loop and helix-turn-helix (HTH). There is no contribution of the RLC in the inter-crown interface. In β-cardiac myosin, the inter-crown interface involves the both the ELC and the RLC from Crown 3 and HTH and Loop-3 from Crown 1. These differences in the interfaces are likely to be responsible of the distinct regulatory mechanisms of the two types of filaments.

Extended data Figure 10 – Effects of HCM mutations on myosin function and regulation.

A set of HCM mutations with effect on the SRX or stability of the interfaces are shown with side chains as spheres. Mutations with direct destabilization of the interface are colored in purple; mutations with indirect effect on the interface are colored in yellow sand; mutation without effect on the stability of the IHM are not colored (name in black); mutations with long range effects and major effects on myosin structural changes are colored in red. (a) Head/head and coiled-coiled/head interfaces. P710R is located at the base of the Converter. The internal cleft close to the actin-binding interface (cleft) is shown with dashed lines. (b) Mutations in the Pliant region and in the ELC. (c) Mutations at the RLC/RLC interface, the N-terminal extension (Nterm) has been contoured in black.

Extended data Table 1:

EM validation

IHM β-cardiac myosin (EMD-15353)
Data collection and processing
Microscope	Titan KRIOS G3
Detector	K3
Magnification	105k
Total exposure (e-/Å2)	39.2
Exposure time	2.4 s
Electron exposure per frame (e-/Å2)	0.98
Defocus range (μm)	−0.8,−2.2
Pixel size (Å)	0.84
Symmetry imposed	C1
Initial particles images (no.)	493179
Final particle images (no.)	213596
Map resolution (Å)	3.6
FSC threshold	0.143
Map resolution range (overall)	3–9
Refinement
Initial model used (PDB codes)	6N69, 2XFM, homology modeling from 7MF3
Map sharpening B-factor (Å2)
Model composition
Non-H atoms	18839
Protein residues	2322
Ligands	6
RMSD from ideal geometry
Bond length (Å)	0.002
Bond angles (°)	0.601
Validation
Molprobity score	1.87
clashcore	16.05
Ramachandran (%)
Outliers	0
Allowed	2.89
Favored	97.11
Validation
CC (mask)	0.79
CC (volume)	0.78
EMRinger score (overall mode)	0.81
EMRinger score (1–786, heads regions at high resolution)	1.15
IHM β-cardiac myosin head-head masked map (EMD-15354)
Data collection and processing
Microscope	Titan KRIOS G3
Detector	K3
Magnification	105k
Total exposure (e-/Å2)	39.2
Exposure time	2.4 s
Electron exposure per frame (e-/Å2)	0.98
Defocus range (μm)	−0.8,−2.2
Pixel size (Å)	0.84
Symmetry imposed	C1
Initial particles images (no.)	493179
Final particle images (no.)	213596
Map resolution (Å)	3.2
FSC threshold	0.143
Map resolution range (overall)	2.5–6
CC (mask)	0.61
CC (volume)	0.61
EMRinger score (1–786, heads regions at high resolution)	1.78

Extended data Table 2: Interactions stabilizing cardiac IHM:

Interactions that differ drastically between smooth and cardiac are shown with orange background.

(i) Head-Head
BH HCM-loop / FH HO-linker and β-bulge The BH HCM loop is positioned differently on the FH head in Cardiac and Smooth IHMs. It interacts mostly with the Transducer in Cardiac while in Smooth, interactions reach the HD-linker but no interaction occur with the N-terminus part of the HO-linker
BH	FH	SmHMM	HCM/effect
HCM-loopR403	HO-linkerQ454 (Mesa) (electrostatic) HO-linkerY455 (Mesa) (stacking + electrostatic) B-bulgeA254 (Mesa)	BH-HCMR406, K408 FH-HO-linkerQ456-G457 (electrostatic) FH-HD-linkerR168,E169	R403Q 23 R403L 24 R403W 25 R403G 26
HCM-loopN408	HO-linkerK450 (Mesa) (electrostatic) HO-linkerP452 (Mesa) HO-linkerR453 (Mesa)	No interaction for residues BH-HCMR411 and residues Nter of the FH-HO.linkerT453,R455	N408K 27 R453C/H/S 28,29,30,31
HCM-loopE409	β-bulgeH251 (Mesa) (electrostatic) HO-linkerR453 (Mesa) (electrostatic) HO-linkerQ454 (Mesa)	No interaction BH-HCMD412 FH-β-bulgeN255	H251N 32,33,4,34 R453C/H/S 28,29,30,31
HCM-loopY410	HO-linkerQ454 (Mesa) HO-linkerR453 (Mesa)	Different interaction BH-HCMV413,Q415 FH-HD-linkerE169,R168 FH-Transβ5F460
BH Loop4 / FH Relay Equivalent interaction in Cardiac and Smooth IHMs between Relay and Loop4
Loop4K367 carbonyl	RelayK503 (Mesa) (electrostatic)	BH-Loop4E370, Q375 FH-RelayR507 (electrostatic)
BH PHHIS / FH Converter
Lock-and-key interactions between the FHConverter T-loop and the bhU50 subdomain involve similar residues but in fact differ as the top-loop conformation differ in the two IHMs. The difference in size of the lock and key residue side chain, the top loop size and composition lead to drastically different interactions between the Converter and the BH head.
HM-helixD382 HM-helixK383 HM-helixY386 HJ-helixL302	T-loopI736	FH-T-loopF746 BH-HM-helix Q385 BH-HM-helix K386 BH-HM-helix H389 BH- HJ-helixL306	I736T35 D382Y 34 K383N36 Y386H 37
BH PHHIS / FH Converter Different positioning of the Converter leads to distinct interactions in Cardiac and Smooth IHMs between the U50 subdomain and the Converter. Several polar interactions are found in Cardiac on either side of the I736 lock and key residue while no polar interaction is predicted at this interface in the smooth IHM on either side of the F746 lock and key residue.
HM-helixE379 (electrostatic) HM-helix K383 (electrostatic) HM-helixE382 (electrostatic) HM-helixY386 (electrostatic)	T-loopD737 T-loopI736 carbonyl T-loopS738 T-loopG733 carbonyl	No polar interaction involving the Top loop of Smooth FH head No interaction with G749 No polar interaction	E379K38 ( K383N36 D382Y 34 Y386H 37 G733E 4 Q734E/P39,40
HM-helix T378.E379.D382 HM-helixY386	T-loopD737.S738 apolar interactions T-loopQ734	BH-HM-helix N381.T382.Q385 FH-Top-loopM747mcD748mc
HN-helixN391 (long range) HN-helixA393 No interaction via K397 and converter HM-helixT378	ConvQ720 ConvD717 ConvG716 ConvL714 ConvT761 ConvH760	BH-HN-helix T396 BH- HN-helix R400 (electrostatic)  ONLY IN SMOOTH FH-ConvE729 (electrostatic) FH-ConvV726 FH-ConvQ728 (electrostatic) FH-ConvK773 (electrostatic, long range)	G716A38
BH PHHIS / FH HD-linker and ELC The FH HD-linker interacts with the U50 subdomain but make only non polar interactions in Smooth IHM as it is more involved in interactions with the HCM-loop
HN-helix D394.K397	HD-linkerR169 (Mesa) (electrostatic)	BH-HN-helixD397 (long range) BH- HN-helix R400 FH-HD-linkerR168 Not defined in density but consistent	D394E38 (Kassem et al., 2013) R169G
HU-helix K611	ELc-loop3Res143–145	No interaction	R143G/Q 40
(ii) BH-Coiled-coil
No Conservation between Smooth and Cardiac : Seg3 of Smooth interacts with only a part of the HO-linker of the BH head to mimic some of the interactions that involve the proximal S2 region of Cardiac
BH Mesa	Coiled-coil	SmHMM	HCM
HO-linkerK450 (Mesa) HO-linkerQ451 (Mesa)	S2E874 (electrostatic) S2V878	Interaction with Seg3
HO-linkerP452.R453.Q454 (Mesa)	S2L881-Q882 S2M877 S2V878	Interaction with Seg3	Q882E 44(Mohiddin et al., 2003) R453C/H/S 28,29,30,31
HW-helixK657 (Mesa)	S2L889 (electrostatic, backbone) S2Q892 S2A893 S2D896	No interaction
HW-helixT660 (Mesa) (electrostatic) HW-helixN661 (Mesa) (electrostatic) HW-helixS664 (Mesa) (electrostatic)	S2Q892	No interaction	T660N 27
HW-helixN656 (Mesa) (electrostatic, backbone) HW-helixT660 (Mesa)	S2D896	No interaction	T660N27
No interactions occur in Cardiac IHM between the FH head and the proximal S2 region, whereas a few interactions occur in Smooth IHM
No interaction		FH-HCM loopR411 FH-S2E941.Q945
		FH-Loop2M655.K652.F656 BH-S2E938 (electrostatic) E939.E942
(iii) RLC/RLC interactions (not conserved between Smooth and Cardiac)
BH-RLC/BH	FH-RLC	SmHMM	HCM ?
The interactions between RLC lobes of the BH and FH heads depend on the presence of the N-terminal extension of the FH head in Smooth, but not Cardiac as the RLCs of the two heads interact directly in Cardiac
RLCα2L56, R58 IQ P838 RLCα1I35, Q38 RLCα2D51, A55 IQ L839	Nterm-ext M20-F21 RLC loop2N78 (electrostatic) T80 RLCα4T80-L83 RLCαlK30	FH-Nterm-extRes15–24 Sandwiched between FH-RLCα1- α4 and BH-RLCα1- α2 BH-IQQ852
IQR845 (electrostatic)	RLCα1Q27-E28, E31 (electrostatic)	BH-IQR855 (electrostatic) FH-RLCα1Q27-S28-Q31 (electrostatic)
(iv) ELC/RLC hinge in the BH and FH heads (not conserved between Smooth and Cardiac)
BH-ELC	BH-RLC	SmHMM	HCM ?
ELC-loop1P65-K66	RLCα6T125-Q126	In the BH lever arm, hydrophobic interactions between: ELCRes L17-R20 RLCResT128-M129-R132 IQL819 ELCD19,D23 (electrostatic) RLCR132 ELCR20 (electrostatic) RLCE124- T128 BH-RLCR44/ FH-RLCD45 carbonyl (electrostatic)
ELC-loop1P65 carbonyl (electrostatic)	RLC-linker3R129 (electrostatic)
ELC- α1- D62 carbonyl (electrostatic)	RLCα6Q126 (electrostatic) IQR808 (electrostatic)		ELC A57G, would weaken helix α145,42
FH-ELC	FH-RLC/FH-IQ2	SmHMM	HCM ?
ELC- loop1R63 carbonyl ELC- loop1R63	RLCα6T125 (electrostatic) RLCα6Q126 (electrostatic)	No clear interaction in the FH lever arm.
ELC- loop1P65 carbonyl (electrostatic) ELC- loop1K66	RLC- linker3R129
ELC- α1-D62 carbonyl (electrostatic) ELC- α1-E56 (electrostatic) ELC- α1-L60	RLCα6T125 carbonyl (electrostatic) IQR808 (electrostatic) IQL805 IQL804		ELC A57G, would weaken helix α1 45,42

Data availability

The atomic model is available in the PDB ⁴⁹ under the code 8ACT. The cryo-EM maps Map 1 and Map 2 are available in the EMDB database⁵⁰ under the accession numbers EMD-15353 and EMD-15354 respectively.