Hypertrophic cardiomyopathy mutations Y115H and E497D disrupt the folded-back state of human β-cardiac myosin allosterically

Abstract

At the molecular level, clinical hypercontractility associated with many hypertrophic cardiomyopathy (HCM)-causing mutations in β-cardiac myosin appears to be driven by their disruptive effect on the energy-conserving, folded-back ‘OFF’-state of myosin, which results in increased number of heads free to interact with actin and produce force. While many characterized mutations likely act by directly perturbing intramolecular interfaces stabilizing the OFF-state, others may function allosterically by altering conformational states of the myosin motor. We investigate two such allosteric HCM mutations, Y115H (Transducer) and E497D (Relay helix), which do not directly contact OFF-state interfaces. Biochemical analyses and high-resolution crystallography reveal that both mutations increase active myosin head availability likely by destabilizing the pre-powerstroke conformation required for OFF-state formation. We propose that destabilization of the folded-back state of myosin, either directly or allosterically, represents a common molecular mechanism underlying hypercontractility in HCM across a broader spectrum of pathogenic mutations than previously recognized.

Introduction

Hypertrophic cardiomyopathy (HCM) is a heart muscle disease that affects more than 1 in 500 individuals and is associated with significant morbidity in the form of arrhythmia, heart failure and sudden death^1,2. It is clinically characterized by diastolic abnormalities and hypertrophic remodelling of the left ventricle. HCM is caused overwhelmingly by autosomal dominant mutations in sarcomeric genes, most commonly MYH7 (encoding the heavy chain of β-cardiac myosin, the predominant ventricular motor) and MYBPC3 (encoding cardiac myosin binding protein-C, cMyBPC)^3,4. It is not completely clear how these point mutations in myosin and other proteins lead to hyperdynamic contraction of the heart and increased energy consumption^5,6.

Muscle contraction is fuelled by ATP-dependent cyclic interactions between myosin-based thick filaments and actin-based thin filaments in sarcomeres. Myosin molecules that are not bound to actin filaments exist structurally in at least two distinct states: an ‘ON’ state in which heads are away from the thick filament backbone and available to interact with actin, and a folded-back ‘OFF’-state known as the interacting heads motif (IHM) where the heads are not available for interaction with actin^7,8. These structural states are thought to correlate with two distinct enzymatic states of myosin: a disordered relaxed state (DRX) with heads that hydrolyse ATP at the normal basal ATPase rate (which is ~100-fold slower than the actin-activated ATPase rate), and an energy conserving state called the super-relaxed state (SRX) with heads turning over ATP ~10-fold slower than DRX myosin^9,10. Myosin autoinhibition has been proposed to be an energy conserving mechanism by decreasing ATP consumption when myosin molecules are in their IHM state^11,12,13; therefore, the conversion between heads stabilized in IHM or free to participate in contraction is expected to be a key determinant of cardiac contractility and energy utilization¹³. Not surprisingly then, perturbation of this conversion is linked to cardiovascular diseases^5,14,15. Indeed, many HCM-causing mutations in β-cardiac myosin have been shown experimentally to result in hypercontractility by destabilizing the IHM ‘OFF’-state, which results in more myosin heads participating in systolic force production^16,17.

In the IHM ‘OFF’-state, the two catalytic Subfragment 1 (S1) heads of the myosin dimer interact with each other and are folded back onto their proximal coiled-coil Subfragment 2 (S2) tail, which inhibits both ATP hydrolysis and actin-binding^18,19. Over 700 HCM mutations throughout the sequence of MYH7 have been reported²⁰. Many of these are found on the surface of the interacting domains of the S1 heads, which includes a relatively flat surface termed the mesa²¹, and the proximal S2 tail of myosin⁷. Mapping these mutations on homology model(s) of the ‘OFF’ state of human cardiac myosin had previously suggested that many of them might destabilize the ‘OFF’ state by disrupting potential intramolecular head-head and head-tail ionic interactions sequestering the myosin heads, as well as intermolecular interactions with other sarcomeric proteins like cMyBPC that may further stabilize the ‘OFF’ state in the thick filament^7,22,23,24. Using recombinant human β-cardiac myosin protein and/or intact myofibers, several such myosin mutations have indeed been shown to increase the number of heads available for actin-binding using various biochemical and biophysical assays (reviewed in Spudich et al.¹⁶). This prediction is further supported by the recent high resolution (3.6 Å) cryo-EM structure of the human β-cardiac myosin IHM, which demonstrated that the two heads adopt a pre-powerstroke state (PPS) to form the IHM via interfaces that greatly differ from those previously proposed from homology modelling²⁵.

It is immediately clear from the cardiac IHM structure that many HCM mutations are not directly located at the intramolecular interfaces stabilizing this sequestered state. How do such mutations lead to hypercontractility? While altered myosin chemo-mechanoenzyme activity might be one mechanism, it is unlikely to be the unifying theme across different mutations, as previous studies with recombinant mutant forms of human β-cardiac myosin have shown that HCM mutations do not consistently lead to hypercontractile changes in the fundamental biomechanical properties and kinetics of actively cycling myosin¹⁶. We therefore hypothesized that such HCM mutations can instead affect the stability of the sequestered IHM state allosterically. The myosin motor domain is highly allosteric²⁶ and HCM mutations could affect the allosteric communication networks between different subdomains of the motor that lead to small conformational changes in the S1 heads, which results in weakening of the IHM structure. Predictions from a recent in silico study lend support to this hypothesis²⁴, but there is little biophysical data exploring the effects of such mutations on the motor function of myosin and its ability to form the folded-back state.

In this work, we investigate two previously uncharacterized pathogenic myosin HCM mutations, Y115H and E497D, neither of which lie on the intramolecular interfaces stabilizing the cardiac IHM structure. Tyr-115 lies on the first β-strand of the Transducer, close to the active site, and Glu-497 lies in the Relay helix (Fig. 1). Both Y115H and E497D HCM mutations are known to be clinically pathogenic^27,28. The E497D mutation has been associated with an apical hypertrophy pattern²⁷. Unfortunately, similar descriptions of specific HCM phenotypes, including markers of severity and hypertrophy patterns are unavailable for the Y115H mutation, as is the case with the vast majority of HCM-associated mutations. Here, using human β-cardiac myosin protein, recombinantly expressed and purified, we find that mutations in these critical regions of the motor domain result in significant changes in the fundamental contractility parameters of myosin, although the two mutants are altered in their contractility parameters in drastically different ways. Remarkably, both the Y115H and E497D mutations severely compromise autoinhibition of myosin. Crystal structures of the pre-powerstroke state, the structural state adopted by both heads in the IHM, reveal that minor structural changes are sufficient to destabilize the formation of the IHM. The mutations lead to an increase in the number of myosin heads functionally accessible for actin-binding. These results show that the disruption of the IHM ‘OFF’-state of myosin by HCM mutations is not limited to mutations in residues that are directly involved in stabilizing the IHM interfaces, in keeping with our unifying hypothesis that most myosin HCM mutations decrease the number of heads in the IHM ‘OFF’-state, which results in more heads available for interaction with actin and hypercontractility of the heart¹⁶.

Fig. 1: Location of the residues Y115 and E497 on the three-dimensional structure of β-cardiac myosin.

a Residues Y115 and E497 are indicated on the motor domain; X-ray crystal structure of bovine cardiac myosin S1 fragment in the pre-powerstroke state is used for the illustration (PDB ID: 5N69). Heavy chain (light grey), ELC (dark grey) and key relevant structural elements (central β-sheet: coral, Relay helix: light blue and Converter: green) are highlighted. Left panel on the bottom shows the residue Y115 (red) located in close proximity to Y134 (yellow) in the purine-binding loop and the nucleotide (ADP). Right panel on the bottom shows the interaction between E497 (blue) on the Relay helix and R712 (green) on the Converter. b Residues Y115 and E497 are indicated on the cryo-EM structure of the folded-back state of human β-cardiac myosin (PDB ID: 8ACT). Heavy chain (two shades of grey), ELC (two shades of orange), RLC (two shades of yellow) and proximal S2 tail (two shades of cyan) are indicated. Protein images were generated using UCSF Chimera.

Results

Selection of HCM mutations not located at the intramolecular interfaces stabilizing the cardiac IHM

The motor function of myosin depends on allosteric communication between its nucleotide-hydrolysis site, actin-binding surface, and the lever arm, which is facilitated by several loops and deformable connectors within the motor head²⁶. HCM-causing mutations linked to MYH7 could result in hypercontractility by affecting the fundamental molecular parameters of the motor (such as intrinsic force or velocity of movement along actin) and/or by destabilizing the IHM ‘OFF’-state. For the purpose of this study, we focused on pathogenic HCM mutations that do not appear to affect residues involved in the intramolecular interactions stabilizing the cardiac IHM. After examining the high-resolution cryo-EM structure of the cardiac IHM (PDB 8ACT)²⁵, we chose two such mutations, Y115H and E497D (Fig. 1a, b), both of which are known to be clinically pathogenic^27,28 but remain uncharacterized.

Y115 resides on the first strand of the conserved central seven-stranded β-sheet, lying very close to the ATP-binding pocket of myosin. Crystal structures of the WT β-cardiac myosin motor head (in both the pre-powerstroke state²⁹ and the post-rigor state complexed with MgADP²⁴) show contacts between Y115 and the conserved purine binding loop (residues 126-134, NPXXXXXXY), specifically with residue Y134 which also forms a hydrogen bond with the adenine ring (Fig. 1a). The Y115H HCM mutation might alter the nucleotide binding properties of myosin, but how this would affect IHM stability cannot be predicted from the high-resolution cryo-EM structure of the cardiac IHM²⁵. E497 lies in the Relay helix (Fig. 1a) which transmits rearrangements occurring within the motor head to the Converter, which in turn transmits them to the lever arm that effects the power stroke. E497 forms a highly conserved salt bridge with R712 in the Converter (seen in both WT PPS²⁹ and post-rigor state²⁴ crystal structures), and mutations at both these positions cause HCM in humans. Mechanical uncoupling of the Relay helix-Converter interface via disruption of this interaction in Drosophila indirect flight muscle myosin led to sarcomere assembly defects and impaired myosin function³⁰, and more recently, the R712L HCM mutation was shown to result in a defective working stroke of human β-cardiac myosin³¹. The E497D HCM mutation is linked to a non-obstructive apical hypertrophy morphology, but this mutant variant of human β-cardiac myosin has not been extensively characterized³¹. E497 was proposed to be involved in major IHM interactions based on the analysis of the IHM model (PDB 5TBY), and a mutation at this position was predicted to impair IHM formation²³. However, the homology model was based on a low-resolution model of tarantula skeletal myosin thick filaments and the recent high-resolution structure of human β-cardiac myosin IHM reveals that E497 is in fact not part of the interactions stabilizing this conformation of cardiac myosin²⁵.

Since both Y115H and E497D mutations are expected to significantly affect the enzymatic activity of myosin by perturbing key long-range communication networks within the motor head, we first characterized the effect of these mutations on the fundamental contractility parameters of human β-cardiac myosin.

The HCM mutations Y115H and E497D have varying effects on myosin ATPase activity and actin-sliding velocity

For this set of experiments, we used the recombinant single-headed short Subfragment-1 (S1) construct of human β-cardiac myosin (sS1; amino acids 1-808 of the myosin heavy chain), consisting of the catalytic head domain (also named motor domain, MD) plus a portion of the lever arm that allows for essential light chain (ELC) binding (SDS-PAGE gels in Supplementary Fig. S1). We first examined the effect of these mutations on the actin-activated ATPase activity of myosin (Fig. 2a and Supplementary Table 1). Steady-state actin-activated ATP hydrolysis rates of wild type (WT) and Y115H mutant myosin were comparable (WT, k_cat = 3.1 ± 0.2 s⁻¹; Y115H, k_cat = 3.2 ± 0.1 s⁻¹), but the mutant protein had a higher apparent actin affinity compared to WT protein (mean ± SEM is reported, see Supplementary Table 1 for details). On the other hand, the E497D myosin had a significantly faster ATPase rate (WT, k_cat = 3.1 ± 0.2 s⁻¹; E497D, k_cat = 4.7 ± 0.3 s⁻¹) and the apparent actin affinity remained comparable to WT myosin (Supplementary Table 1).

Fig. 2: HCM mutations Y115H and E497D alter the actin-activated ATPase activity and actin sliding velocity of human β-cardiac myosin sS1.

a Steady-state actin-activated ATPase data for WT, Y115H and E497D sS1 ensembles. A representative dataset is shown from one side-by-side preparation of the three proteins; mean ± SD from three technical replicates at each actin concentration is plotted. The data is fit to the Michaelis–Menten equation (solid line) to determine k_cat and K_app for each myosin and the shaded area indicates the computed 95% confidence intervals. A summary of values and statistical comparisons from ATPase experiments from four independent protein preparations is provided in Supplementary Table 1. b Actin sliding velocities for WT, Y115H and E497D sS1 myosin ensembles at 23 °C. Filtered mean velocity (MVEL₂₀, see methods and Supplementary Fig. S2) measurements from 8 independent experiments across 4 protein preparations are reported, mean ± SEM is plotted; all values of velocities are given in Supplementary Table 1. A two-sample, two-sided unequal variance t-test (Welch’s t-test) was used to compare WT and each mutant (pairwise comparison: WT vs. Y115H, p ≤ 0.0001; WT vs. E497D, p = 0.0078). ** indicates p ≤ 0.01, **** indicates p ≤ 0.0001. Source data are provided as a Source Data file.

We next examined the effect of both mutations on myosin’s ability to move actin filaments in an unloaded motility assay. Mean actin-sliding velocity was significantly reduced by both mutations: ~55% by the Y115H mutation and ~20% by the E497D mutation (Fig. 2b and Supplementary Table 1; Fast Automated Spud Trekker or FAST analysis³² in Supplementary Fig. S2, and additional motility parameters from FAST analysis and discussion in Supplementary Fig. S3 and Supplementary Note 1). Overall, we observed that, as expected based on their locations, both Y115H and E497D mutations significantly altered the ensemble properties of myosin. Under the detachment-limiting conditions typically employed for the motility experiments described here, decreased actin-sliding velocity by the mutations can be explained by a change in either the actin bound time and/or myosin’s stroke size. We next sought to determine the effect of the Y115H and E497D mutations on these parameters using single-molecule optical trapping experiments.

Both mutations affect the load-dependent actin detachment kinetics, and, unexpectedly, the Transducer mutation Y115H reduces myosin’s stroke size

Motor properties of single myosin molecules were determined by harmonic force spectroscopy³³ (HFS) using the three-bead optical trap system (Supplementary Fig. S4). At physiological ATP concentrations, the rate of ADP release limits the detachment of cardiac myosin from actin, and as expected from the force-velocity relationship of cardiac muscle contraction, this rate depends exponentially on applied load. Using HFS, the lifetimes of binding events between a single myosin sS1 molecule and an actin filament were measured under a range of assistive and resistive load forces, applied randomly and automatically over the course of many binding events by harmonic oscillation of the stage upon which myosin is fixed to a bead platform for interaction with the trapped actin filament (Supplementary Fig. S5, a–c). Each binding interaction was characterized by its duration (t_s) and the corresponding mean force (F) (Supplementary Fig. S5d–f). The resulting events were sorted into force intervals (Supplementary Fig. S6), and for each force bin, the detachment rate ${k}_{\det }(F)$ was determined using maximum likelihood estimation, under the assumption of an exponential distribution of bound times (Supplementary Fig. S6a–e). HFS has been successfully used in the past to quantify the effects of several pathogenic mutations as well as small molecule myosin modulators on the load-dependent actin detachment kinetics of human β-cardiac myosin^34,35,36,37.

Experimentally obtained force dependent detachment kinetics of the WT and mutant molecules were fit to the harmonic force-corrected Arrhenius equation (Eq. (1)), as described previously^33,36,38, to obtain detachment rates at zero external force $({k}_{0})$ and force sensitivity ($\delta$) of individual molecules (Supplementary Figs. S6f and S7a–c):

$${k}_{\det }(F,\Delta F)={k}_{0}{I}_{0}\left(\frac{\Delta F\delta }{{k}_{B}T}\right)exp \left(\frac{-F\delta }{{k}_{B}T}\right)$$

(1)

where $\Delta F$ is the amplitude of the sinusoidal force, ${k}_{B}$ is the Boltzmann’s constant, and T is the temperature. A zero-order Bessel function $({I}_{0})$ is used to correct for the sinusoidal nature of the applied force. ${I}_{0}$ is a function of $\Delta F$.

In agreement with the previously published values, WT sS1 myosin had a detachment rate at zero load k₀ = 119 ± 38 s⁻¹ (Fig. 3a) and force sensitivity of the detachment rate $\delta$ = 1.03 ± 0.23 (Fig. 3b). Y115H and E497D mutations oppositely affected the actin-detachment kinetics at zero load (Fig. 3a and Supplementary Table 1), while the force sensitivity was significantly different only for the E497D mutant molecules (Fig. 3b and Supplementary Table 1). Simply put, the Y115H mutation significantly increased the average time myosin molecules were bound to actin under zero load (~16 ms for Y115H myosin vs. ~8 ms for WT myosin), without affecting the load sensitivity of the bound times. On the other hand, E497D myosin molecules bound actin only for ~3 ms on average under zero load and displayed a much shallower/less sensitive dependence on applied load.

Fig. 3: HCM mutations Y115H and E497D alter human β-cardiac myosin sS1 single molecule biomechanics and ensemble properties.

The mean values ± SEM of a actin-detachment rates at zero load (k₀), b force sensitivity (δ), and c stroke-size of single molecules of WT (n = 14 molecules from 2 independent protein preparations), Y115H (n = 12 molecules from 2 independent protein preparations) and E497D (n = 14 molecules from 2 independent protein preparations) β-cardiac myosin sS1 are plotted. A two-sample, two-sided unequal variance t-test (Welch’s t test) was used to compare WT and each mutant. n.s. indicates p > 0.05, *** indicates p ≤ 0.001, **** indicates p ≤ 0.0001. All numbers including exact p-values from multiple pairwise comparisons shown here are given in Supplementary Table 1. HCM mutations Y115H and E497D alter the duty-ratio, average force production, and average power production of β-cardiac myosin sS1 in opposite ways. The d duty ratios (Eq. (2)), e average force (Eq. (3)), and f average power outputs (Eq. (4)) are calculated for the resistive force-range for WT, Y115H, and E497D constructs from the measured values of detachment rate k₀, actin-activated ATPase rate k_cat, force-dependent detachment rate k_det(F), and stroke size (d). Curves for WT are shown in black, Y115H in red, and E497D in blue. Source data are provided as a Source Data file.

Further analysis of the same HFS data was done to obtain stroke sizes for each interacting myosin molecule³⁵ (for details see Methods). Remarkably, the active site adjacent mutation Y115H dramatically decreased myosin’s stroke size by ~50%, while the stroke sizes of WT and E497D myosin molecules were comparable (Fig. 3c and Supplementary Table 1). Under the experimental conditions employed here^39,40, ensemble velocity can be approximated as the stroke size divided by the time myosin is strongly bound to actin (V = d/t_s), and since t_s is inversely proportional to the detachment rate, V = d*k_det. We note that the changes in actin-detachment kinetics and stroke- sizes observed in single molecule experiments account for the change observed in ensemble motility measurements for Y115H myosin, but not for E497D myosin, which is further discussed in Supplementary Note 1.

Single molecule and ensemble measurements do not suggest a clear mechanism of hypercontractility by the Y115H and E497D mutations

The force produced by an ensemble of motor proteins depends on the fraction of its ATPase cycle each motor spends in the force-producing state. The measure of this “duty ratio r(F)” as a function of load force F, can be obtained as follows:

$$r(F)=\frac{{k}_{{attach}}}{{k}_{{attach}}+{k}_{\det }(F)},{\rm{where}}\; {\rm{the}}\; {\rm{attachment}}\; {\rm{rate}},{k}_{{attach}}=\frac{1}{\left(\frac{1}{{k}_{{cat}}}-\frac{1}{{k}_{0}}\right)}$$

(2)

Here, ${k}_{{cat}}$ is the actin-activated ATPase rate measured by the steady-state assay. Myosin molecules that spend less time during each catalytic cycle in the actin-bound, force producing state have lower duty ratios and faster actin detachment rates. The force-dependent duty-ratios for WT, Y115H, and E497D are plotted in Fig. 3d. The duty ratios for Y115H remained higher than WT over the whole force range due to lower ${k}_{\det }(F)$ values for Y115H. The duty ratios for E497D were lower than that of WT; however, near zero external force, the duty ratio for E497D remains comparable to that of WT. This difference in the force dependence of duty ratios between WT and E497D is a result of the lower force-sensitivity of the E497D mutant.

Using duty ratio, the average force produced by a single molecule is calculated as:

$${F}_{{av}}\left(F\right)=F\,r(F)$$

(3)

Here, $F$ is the load force. The plot for the average force against different external forces follows the same trend as that of the duty-ratios (Fig. 3e). Finally, using the ${F}_{{av}}\left(F\right)$, the average power produced by a single molecule, ${P}_{{av}}\left(F\right),$ can be obtained from Eq. (4).

$${P}_{{av}}(F)={k}_{\det }\left(F\right)\times d\times {F}_{{av}}(F)$$

(4)

The average power produced by the myosin molecules against different load-forces is plotted in Fig. 3f. Overall, Y115H and E497D showed different but opposite trends compared to WT in the average power plot, with the significantly higher k_det of E497D leading to greater power despite its relatively decreased F_av, while the lower k_det and smaller stroke-size of Y115H account for its lower average power.

Taken together, single molecule and ensemble experiments show that several changes occur in the biomechanical properties of the myosin motor domain in response to both the Y115H and E497D mutations, but these changes do not clearly explain hypercontractility of the myosins carrying these mutations. So, we next assessed the effects of these mutations on autoinhibition of myosin. For these experiments, we used purified two-headed HMM constructs containing a fragment of the proximal S2 tail that is long enough to stabilize the folded back conformation of myosin in vitro^25,41.

Both Y115H and E497D mutations destabilize SRX myosin and increase the rate of ATP turnover by DRX myosin

Next, we measured the basal rate of ATP hydrolysis by myosin in the absence of actin using the single mant-ATP turnover (STO) assay¹². The kinetics of mant-ADP release reveal the proportion of myosin heads in the ‘open’ DRX state (~0.03 s⁻¹) and the ‘closed’ SRX state (~0.003 s⁻¹). Several HCM mutations have been shown to reduce the SRX fraction in STO assays¹⁶, which is often assumed to be a proxy for the IHM ‘OFF’ state, and we applied this technique to Y115H and E497D mutant myosins. For these measurements, we used short-tailed 8-hep and long-tailed 15-hep myosin constructs, containing the first 8 and 15 heptads respectively of the proximal S2 tail (see Methods). Previous studies with smooth muscle myosin⁴² and human β-cardiac myosin^25,41 have demonstrated that the first 15 heptads of the rod domain are enough to allow the formation of the completely inactive folded-back state of myosin in vitro. The previously uncharacterized 8-hep myosin construct was used as the uninhibited fully active control as its shorter S2 fragment should not support the formation of the folded-back state. Overall, 8-hep and 15-hep constructs are similar in principle to the previously characterized 2-hep and 25-hep constructs, respectively, of myosin⁴³.

We first examined the turnover of mant-ATP by WT 8 and 15-hep proteins at varying salt (potassium acetate, KOAc) concentrations (Supplementary Fig. S8 and Supplementary Table 2). The fluorescence decays for both proteins were bi-exponential (see Supplementary Fig. S9 for residuals from single- and double-exponential fits) and the fitted fast (0.01–0.03 s⁻¹) and slow (0.002–0.004 s⁻¹) rates were similar to the previously measured DRX and SRX rates with WT 2 and 25-hep proteins⁴³. The SRX:DRX ratio of the short-tailed WT 8-hep myosin was uniformly low at all salt concentrations tested, indicating its inability to form the folded-back state, while for the long-tailed WT 15-hep myosin the SRX fraction decreased with increasing salt concentration (Supplementary Fig. S8). These results are consistent with the SRX biochemical state correlating with a structural state with inhibited activity that is stabilized by intramolecular ionic interactions. Thus, the IHM state is the likely structural basis for the SRX state measured here, which is supported by our cryo-EM studies showing that WT 15-hep protein (without glutaraldehyde cross-linking or stabilization by small molecules) forms an IHM structure²⁵, as well as recent FRET experiments where the IHM state was directly quantified for the WT 8-hep and 15-hep proteins⁴⁴.

Unlike previously studied mutations¹⁶, both Y115H and E497D 15-hep proteins displayed significantly faster kinetics of mant-ADP release (Fig. 4a–c; see Supplementary Fig. S9 for residuals from single- and double-exponential fits), which were captured using an automated reagent injector coupled to a plate-based fluorimeter with ~2 s mixing dead-time (see Methods). At low salt concentrations (5 mM KOAc), where the IHM state is most stabilized (Supplementary Fig. S8a), we found that both Y115H and E497D mutations dramatically reduced the percentage of molecules in the SRX state (Fig. 4d and Supplementary Table 2) and presumably in the IHM state. ~55% of WT 15-hep myosin were in the SRX state, compared to <10% of molecules for both Y115H and E497D 15-hep myosin. The percentage of molecules in the SRX state was not significantly different between the WT, Y115H and E497D 8-hep proteins (Fig. 4d), which are unable to form the IHM state. Both Y115H and E497D mutations significantly increased the DRX fast rate compared to WT protein (Fig. 4e), in both 8-hep and 15-hep backgrounds (see Supplementary Table 2 for rates). The SRX rate was not affected by these mutations (Fig. 4e).

Fig. 4: HCM mutations Y115H and E497D reduce the super relaxed state fraction of human β-cardiac myosin HMM.

Representative traces showing the kinetics of mant-ATP turnover by 8 and 15-hep constructs of a WT, b Y115H, and c E497D myosin. The inset in (a) shows a schematic of two-headed short (8-hep) and long (15-hep) tailed HMM protein constructs. The data is fit to a double exponential function (solid line) and fitted fluorescence values at t = 0 and t = ∞ were used to normalize the individual curves. d, e Quantification of the single turnover data showing (d) relative amplitude of the slow rate constants (% SRX) for the 8-hep and 15-hep constructs, and the (e) fast (DRX) and slow (SRX) rate constants for 15-hep constructs (see Supplementary Table 2 for 8-hep rate constants) for all three proteins. Data is combined from 5 independent fluorescence transients from 3 biological protein preparations, mean values ± SEM are plotted. A two-sample, two-sided unequal variance t test (Welch’s t test) was used to compare WT and each mutant. n.s. indicates p > 0.05; * indicates p ≤ 0.05; ** indicates p ≤ 0.01, *** indicates p ≤ 0.001, **** indicates p ≤ 0.0001. All numbers including exact p-values from multiple pairwise comparisons shown here are given in Supplementary Table 2. Source data are provided as a Source Data file.

The HCM mutations Y115H and E497D increase the number of myosin heads available for actin binding

While SRX measurements have been used as a reasonable proxy for the IHM ‘OFF’-state, they are indirect and even single-headed S1 preparations show a fraction of SRX heads in a STO experiment⁴³, showing that SRX heads are not always correlated with the IHM ‘OFF’-state heads^34,45,46,47. We have shown in the past that actin-activated ATPase rate (k_cat) measurements using the Long-tail/Short-tail ATPase Ratio assay (LSAR) is the most direct biochemical method for measuring an increase in the number of heads available for interaction with actin¹⁶. To confirm that the Y115H and E497D mutations increase the number of active heads in a two-headed myosin population capable of forming the IHM ‘OFF’-state, we used the LSAR assay.

Similar to previous observations in LSAR assays with the WT 25-hep (long-tailed) and 2-hep (short-tailed) myosin constructs⁴³, we measured a significant ~40% drop in the k_cat of WT 15-hep (long-tailed) myosin compared to WT 8-hep (short-tailed) myosin (WT 8-hep, k_cat = 4.3 ± 0.4 s⁻¹; WT 15-hep, k_cat = 2.7 ± 0.2 s⁻¹; p-value = 0.02; Fig. 5a and Supplementary Table 2). These results confirm that a large fraction of the WT 15-hep myosin population is in the folded-back state (measured average LSAR = 0.63, Supplementary Table 2), which decreases the number of myosin heads that bind to actin and quickly release ATP hydrolysis products. Consistent with the trends observed with sS1 proteins (Fig. 2a and Supplementary Table 1), ATPase assays with 8-hep myosin showed WT and Y115H mutant myosins as having similar activity while the E497D mutant myosin showed a ~50% increase in k_cat (Supplementary Table 2). Remarkably, both the Y115H and E497D mutations reduced the ~40% drop in ATPase seen for the WT 15-hep construct compared to the 8-hep construct to <10% (Fig. 5b, c). There was no significant difference between the mutant 8-hep and mutant 15-hep myosin ATPase rates (see Supplementary Table 2 for details) and average LSAR value for Y115H and E497D was 0.9 (Supplementary Table 2). LSAR values were used to estimate the additional number of myosin heads made accessible for interaction with actin (additional N_a) as follows (described in a recent review¹⁶): percent additional N_a increase = (x − 0.63)/0.37, where x = the measured LSAR for a particular HCM mutant myosin and 0.63 is the average LSAR value for WT myosin. This metric shows a ~75% increase in the number of myosin heads released by both the Y115H and E497D mutations. An increased availability of myosin heads for actin binding was also evident from the ATPase activity trends for the long-tailed 15-hep constructs. The overall k_cat for both Y115H and E497D 15-hep proteins was significantly elevated (1.5-fold and 2-fold higher, respectively) compared to that of WT 15-hep myosin (Fig. 5d). Overall, these results suggest that both the Y115H and E497D mutations significantly impair myosin’s ability to form the IHM ‘OFF’-state, leading to increased availability of myosin heads for actin binding and force production.

Fig. 5: HCM mutations Y115H and E497D increase the number of active human β-cardiac myosin heads as measured by the actin-activated ATPase LSAR assay.

Actin-activated ATPase curves for 8-hep and 15-hep constructs of a WT, b Y115H, and c E497D myosin. For each mutant, a set of WT 8-hep, mutant 8-hep and mutant 15-hep proteins was grown and processed for purification in parallel to reliably capture the differences in k_cat due to the mutations. Raw data is combined from 12 experimental replicates from 4 independent protein preparations for WT and 9 experimental replicates from 3 independent protein preparations for mutants. Each data point represents the average across all replicates (mean ± SD is plotted). The data are fit to the Michaelis–Menten equation (solid line) to determine k_cat and K_m for each myosin (shaded areas indicate the 95% CI of the fit). ATPase curves in each panel are normalized to the respective 8-hep k_cat (see Supplementary Table 2 for raw values and statistics). Downward arrows in a-c indicate the average % drop in k_cat of 15-hep constructs relative to 8-hep controls. All numbers including exact p-values from multiple pairwise comparisons shown here are given in Supplementary Table 2. d Actin-activated ATPase curves with the Michaelis–Menten fits for the 15-hep constructs of WT (k_cat = 2.9 ± 0.2 s⁻¹), Y115H (k_cat = 4.4 ± 0.1 s⁻¹) and E497D (k_cat = 5.8 ± 0.2 s⁻¹) proteins from one tandem protein preparation; each data point denotes mean ± SD from three independent measurements at each actin concentration. Upward arrows indicate % increase in k_cat of mutant 15-hep constructs relative to WT 15-hep protein. Source data are provided as a Source Data file.

We note that the basal ATPase rates for the Y115H and E497D proteins (both 8-hep and 15-hep constructs) were elevated compared to WT protein (Supplementary Fig. S10), correlating with the increased DRX rates for the mutants (Fig. 4e). For all 6 proteins (WT, Y115H and E497D 8-hep and 15-hep constructs), the measured basal rates correlated well with the amplitude weighted sum of the DRX and SRX rate constants for each protein (calculated basal rates, Supplementary Fig. S10).

Crystal structures of the Y115H and the E497D mutants reveal changes to local interactions in the pre-powerstroke state

We next gained insights into how the mutations impair local interactions and the dynamics within the motor domain. We crystallized both mutant motor domains of human β-cardiac myosin in the pre-powerstroke state using the sS1 constructs. The structures solved are described at 2.48 Å (MD-Y115H-PPS) and 2.60 Å (MD-WT-PPS and MD-E497D-PPS) resolution respectively (Fig. 6; Supplementary Table 3). Similar B-factor profiles were obtained for both mutants compared to WT sS1 (Supplementary Fig. S11) and clear electron density was observed for the mutated residues (Supplementary Fig. S12 and Supplementary Fig. S13). The mutations did not introduce important rearrangements in the pre-powerstroke structures (RMSD 0.246 for Y115H compared to 691 residues of WT protein and RMSD 0.225 for E497D compared to 691 residues of WT protein). The single amino-acid substitutions mainly lead to local changes that result in increasing the dynamics by loss of stabilizing interactions.

Fig. 6: HCM mutations Y115H and E497D alter local interactions in the human β-cardiac PPS structure.

a Localization of residues Y115 and E497 in the motor domain (MD). Regions involved in Y115-Y134-nucleotide interactions are the N-term (grey); Transducer (dark green) and ADP.P_i (orange). Regions involved in E497-R712 interactions are the Relay helix (yellow) and Converter (lime). b, c Electronic density comparison for the Y115H mutation: b Y115 in the X-ray crystal structure of the human cardiac myosin WT MD (PDB 9I8P) and c H115 in the X-ray crystal structure of the human cardiac myosin H115 mutant MD (PDB 9HTF). d, e Electronic density comparison for the E497D mutation: d E497 in the X-ray crystal structure of the human cardiac myosin WT MD (PDB 9I8P) and e D497 in the X-ray crystal structure of the human cardiac myosin D497 mutant MD (PDB 9HTG). f Interactions between Y115/H115, Y134 and the nucleotide (blue lines, H-bonds; grey lines, other distances). No changes in conformation are observed for the ADP.P_i nor for Y134. The Y115-Y134 H-bond is not maintained in the Y115H mutant as the shorter imidazole side chain cannot reach the Y134 hydroxyl group, in contrast to Y115’s larger phenol side chain. g Interactions between E497/D497 and R712 (blue lines, H-bonds; orange lines, salt bridge; grey lines, other distances). E497-R712 forms a strong ionic bond in the WT structure. While the salt bridge between D497 and R712 is maintained in E497D through a change of conformation of R712, it is subsequently weakened as only one of D497’s oxygens is in position to form the ionic bond. The D497-R712 interaction is mediated through Y501 that forms H-bonds with both D497 and R712. The Y501 side chain changes its orientation to accommodate this bonding.

The substitution of Y115 by a histidine leads to shortening of the side chain and introduction of a positive charge closer to a hydrophobic environment made of atoms of residues I114 and H153. In the structural state revealed by the crystal structure, there is no significant effect of the Y115H substitution on the main chain of the overall protein structure, including the Relay helix, the Converter or the position of the nucleotide. Small adjustments of Thr124 and the following residues of the purine binding loop (N126 to Y134) indicate, however, that the purine binding elements are sensitive to this mutation. Indeed, Y115 is at the centre of a network of interactions that stabilize the nucleotide and also involves Y134 and T124. In high resolution structures of the myosin pre-powerstroke state, this network is highly conserved and involves two water molecules²⁹ (Fig. 7a and Supplementary Fig. S14a). The Y115H substitution disrupts this network which thus increases possible dynamics of Y134 and the nucleotide (Fig. 7b and Supplementary Fig. S14b). In short, the Y115H substitution critically affects the active site interactions that stabilize the nucleotide. This can explain the higher basal ATPase rate measured for the Y115H mutant myosin.

Fig. 7: Changes in the water-mediated networks in Y115H and E497D mutants.

a The crystal structure of WT bovine cardiac myosin bound to Mavacamten at 1.8 Å resolution* (PDB 8QYR) shows the network that stabilizes water molecules near the nucleotide purine group in the active site. Water mediated networks that include residues T124, A182 and N187 stabilise the nucleotide purine group position. b Environment of the myosin active site in the Y115H crystal structure solved at 2.5 Å resolution. The tyrosine to histidine substitution disrupts the water mediated stabilization of the purine group by loss of interactions between H115 and Y134 as well as between H115 and N187. Reorientation of the T124 and N187 side chains occur and no water-based network involving these residues occur to participate in the stabilization of the nucleotide purine group. c Interactions between E497 and R712 in WT human β-cardiac myosin structure solved at 2.6 Å resolution. A water mediated network allows for further interactions between the Relay and the Converter subdomains. d The E497D crystal structure solved at 2.6 Å resolution indicates that weaker interactions occur between D497 and R712. Reorientation of R712 NE disrupts the interactions found in WT with residues of the Converter: bonding with the carbonyl of F709 is maintained by NE reorientation whereas the water mediated binding with K762 and E500 is lost. In Supplementary Fig. S14, a more detailed version of this figure including the distance annotations is provided.

The substitution of E497 by an aspartate shortens this negatively charged side chain by a carbon-carbon bond. It is important to note that, in the pre-powerstroke state, this residue of the Relay helix plays a critical role by forming an internal salt bridge with R712 of the Converter, which helps stabilize the primed lever arm. The pre-powerstroke structure of E497D was essential to show that the Relay helix or Converter positions can adopt the conformation that is also seen in the WT structure despite changes in this salt bridge stabilization. The side chain of R712 reorients slightly in order to form a hydrogen bond with D497, and in so-doing it also interacts with the hydroxyl of Y501, which also belongs to the Relay helix. This reorientation subsequently may also weaken intra-molecular interactions within the Converter and with the Relay as the reoriented R712 is no longer involved in water mediated interactions with K762 and E500 (Fig. 7c, d and Supplementary Fig. S14c, d). Thus, the contacts between the Relay helix and the Converter differ without affecting the position of the priming of the Lever arm. However, these differences lead to reducing the stability of the pre-powerstroke state as reflected by the elevated basal ATPase rate measured for this mutant.

Indeed, the pre-powerstroke state stability is linked to the trapping of the hydrolysis products. The elevated basal ATPase activity and a 3–5-fold increase in phosphate (Pi) release rates, measured from single-turnover mant-ATP experiments in the absence of actin for both mutants (Fig. 4e), suggest that the lifetime of the PPS, or the M*.ADP.Pi state (M* denotes distinct fluorescence properties of myosin in the PPS), is lower in case of the mutants since they do not hold on the products of hydrolysis as well as the WT protein. The PPS stability however depends both on formation and decomposition rates, and a decrease in folded-back heads may also arise from compromised formation of the PPS state (M*.ATP and/or M*.ADP.Pi). The formation of the PPS state was measured by following tryptophan fluorescence⁴⁸ using Trp-Mant FRET upon mant-ATP binding to sS1 myosin. These transients were bi-exponential with no difference in the fluorescence intensities or kinetics between WT and mutants (Supplementary Fig. S15), indicating that the formation of PPS itself is likely not affected by Y115H and E497D mutations. Thus, both Y115H and E497D destabilize the pre-powerstroke state. Since the high-resolution cryo EM structure of the human β-cardiac IHM showed that the two heads need to adopt a pre-powerstroke state to form the auto-inhibited interactions, this destabilization of the pre-powerstroke state leads to the reduced ability of these mutants to form the IHM as reflected by the SRX and LSAR measurements (Figs. 4d and 5b, c).

Discussion

Experimental support for the central role played by myosin autoinhibition in the regulation of cardiac contractility, and the association of its dysregulation with cardiomyopathies has grown tremendously in the past decade^14,17. Many HCM-causing mutations appear to affect the interfaces stabilizing the cardiac IHM structure^6,7. Aligned with these observations, an increase in the number of functional myosin heads available for interaction with actin using the LSAR assay and/or a decrease in the SRX fraction has been observed for >15 of these when tested in vitro¹⁶. Not surprisingly, these correlations have led to more detailed analyses of HCM variant locations using homology modelled structures of the cardiac IHM^22,23,24, which have predicted that the pathophysiology of ~30–50% of the known pathogenic HCM mutations may be linked to their direct effects on the stability of the autoinhibited state of myosin. Here we show that Y115H and E497D mutations, which are far from the interfaces that stabilize the IHM structure, greatly reduce the fraction of myosin heads in the SRX state and increase the k_cat of long-tailed myosin constructs. These are not isolated findings—G256E⁴⁹ and P710R³⁵ HCM mutations also appear to interfere with myosin autoinhibition without directly altering the IHM interfaces. Our recent FRET results directly show that the P710R mutation indeed destabilizes the IHM structure⁴⁴. Altogether, these observations indicate that as predicted previously²⁴, HCM mutations not directly located at the interfaces of the IHM can alter the stability of the IHM allosterically, and the unifying hypothesis that increases in contractility caused by myosin HCM mutations generally result from increases in the levels of the ON-state of cardiac myosin has gained strong support¹⁶. Importantly, these results highlight that it is not trivial to predict the effects of a pathogenic mutation on myosin’s enzymatic activity or conformational dynamics. Similarly, it is not possible to predict how the mutation affects the power that can be generated by the myosin. Purified mutant forms of bonafide human β-cardiac myosin must be analysed by multiple assays, including the LSAR assay to directly ascertain the levels of the IHM OFF-state¹⁶.

Quantitative in vitro assays with purified proteins must be integrated with high-resolution structural data to gain the desired mechanistic insights into the nature of the changes induced by HCM mutations. Here we show that the single consistent hypercontractile effect of the Y115H and E497D mutations is an increased availability of active myosin heads for interaction with actin (Figs. 4 and 5), with both mutations resulting in an estimated 75% increase in the number of myosin heads released from the sequestered state. Strikingly, these mutations appear to impair myosin’s ability to adopt the folded-back OFF state through allosteric mechanisms, likely by disrupting the thermodynamic stability of the pre-powerstroke conformation. This is supported by elevated phosphate (Pi) release rates from single-turnover measurements, indicating that the PPS state—with ATP hydrolysis products bound—is destabilized. Although an impairment in the formation of the PPS state itself (with ATP bound) could also hinder IHM formation, our data suggest this is not the case for these mutants. Nonetheless, such a mechanism remains a plausible way in which other pathogenic mutations might disrupt the folded-back OFF state. Remarkably, despite a large effect in reducing myosin autoinhibition, no significant structural change is found in the pre-powerstroke structure of either mutant compared to the structure required to form the IHM. Rather, the local changes in the interactions near the mutations likely explain the difference in the stability of the PPS, and thus the stability of the IHM. This suggests that HCM can result from an alteration in the dynamics and stability of the pre-powerstroke state, rather than a change in its structure, leading to a reduced number of heads capable of forming the IHM. This demonstrates the subtlety of the regulation of cardiac muscle myosin and the difficulty in predicting the effect of cardiomyopathy-causing myosin mutations just from their location in the structure.

A pathogenic mutation can therefore destabilize the IHM either directly by altering a residue at the interface, indirectly by being close enough to induce changes in the dynamics or flexibility of the loops and connectors and other structural elements of the motor that are engaged at these interfaces, or by affecting the stability of the PPS by altering long-range interactions that maintain the lever arm in the primed position²⁴. A previous analysis of 178 pathogenic HCM mutations using the MA1 IHM model predicted ~65% of them to disrupt the IHM state, and roughly half of them were proposed to do so by destabilizing the PPS conformation²⁴. The G256E and P710R mutations have also been proposed to destabilize the PPS conformation²⁴, which is likely the reason for the decrease in autoinhibition measured experimentally for both^35,49. In fact, an elevated DRX rate has also been measured for the P710R (modest elevation)³⁵ and P710H (~3-fold, unpublished results) mutations. Unlike several previously studied HCM mutations¹⁷, Y115H and E497D mutations significantly affected motor function as well, indicating unpredictable and interesting effects of the mutations on the actin-bound affinity and/or powerstroke. E497D increased k_cat by ~50%, which is one of the largest changes measured for human β-cardiac myosin disease variants and reminiscent of the early-onset HCM mutations D239N (50%) and H251N (24%)⁵⁰. But the combination of the observed changes for both Y115H (unchanged k_cat, reduced k_det and stroke-size) and E497D (elevated k_cat and k_det, unchanged stroke-size) resulted in opposite effects on the calculated average force and average power produced by a single myosin molecule. Overall, while these changes cannot be predicted, they indicate that hypercontractility for both Y115H and E497D is linked to the decrease in autoinhibition of myosin. While the PPS appears unperturbed structurally by these mutations, changes are to be expected in the timing, the dynamics or the structure of the actin-bound states. Future structural studies are necessary to reveal how these mutations might affect the actin binding interface or the structures of the post-powerstroke or rigour states of myosin. Such studies will be instrumental in defining the mechanism by which these single missense mutations can distinctly and allosterically change motor properties. It is essential to also investigate how the average force and power are affected by HCM mutations as well, as it is currently unknown how disease progression depends on these characteristics and how treatment with myosin inhibitors would need to be adjusted depending on the force and power produced by these mutant myosins. Studies of mutant motors are essential to develop predictions and personal medicine.

We highlight two unexpected findings from these studies. First, based on the position of Y115H, a reduction in stroke-size is at first an unexpected change. Previously, different mechanisms have been proposed to explain stroke-size changes observed for other mutations based on their location: uncoupling of the biochemical and mechanical activities of myosin (P710R³⁵ and R712L³¹ in the Converter), destabilization of the lever arm (L781P³⁴), and altered lever arm priming in the pre-powerstroke state (R671C in the Transducer³⁷). Judging from the location of Y115H in the motor head, it seems unlikely that this mutation affects lever arm coupling or compliance. An interesting hypothesis is that a difference in the coordination of ADP might affect the extent to which the lever arm rotates as ADP is released, and therefore shorten the second part of the stroke and the overall stroke size.

Second, we note that the E497D and R712L HCM mutations, both affecting the Relay helix-Converter E497-R712 ionic interaction, have dramatically different effects on motor function. While one drastic effect of the R712L mutation is its very small stroke size³¹, our structure of the pre-powerstroke state explains why this is not the case for E497D. The shorter side chain of Asp does not induce an abnormal orientation of the lever arm upon formation of the D497-R712 interaction and the lever arm is stabilized by a different set of bonds upon a small reorientation of the R712 side chain (Fig. 7 and Supplementary Fig. S14). The relatively subtle change in velocity (~20% drop) and lack of change in stroke-size for E497D is consistent with the structural data. Importantly, disruption of the E497-R712 interaction by the mutations E496R (in Drosophila indirect flight muscle myosin³⁰) and R712L (in human β-cardiac myosin³¹) would lead to greater disruption of the priming of the Converter and thus lead to drastically different impairment of motor function, as has been previously reported (~70–80% drop in motility velocity³⁰ and 4-fold reduction in myosin’s stroke-size³¹). Our functional assays show that E497D is more prone to detachment compared to WT and it increases k_cat without an effect on stroke size. This suggests that the Strong-ADP state is less stable with this mutation, leading to a faster ‘Strong ADP to Rigor’ transition, which involves the last part of the lever arm swing. It is interesting to note, however, that despite the mechanistic and structural differences noted here, the common and likely most important effect for E497D and R712L (see Supplementary Fig. S16 for R712L single turnover data) mutations is the reduced ability to form the IHM state. Interestingly, a recent study reported no change in the SRX-DRX equilibrium for the R712L mutant⁵¹, however, the reason for this discrepancy in results is unclear. One possibility is that it may be due to differences in the light chains used in each study. While we use human ventricular ELC and RLC, the other study used human myosin HMM with a mixture of mouse skeletal light chains endogenously expressed in the mouse skeletal myoblast C2C12 cells bound.

Finally, a long-recognized component of HCM pathophysiology is increased energy demands by the sarcomere and inefficient energy utilization^52,53. Impaired myocardial energetics (phosphocreatinine/ATP ratio) have been prominently observed in HCM patients⁵⁴, reportedly evident even before the adverse cardiac remodelling becomes apparent^55,56,57. This deficit appears to be a result of energy depletion induced by increased ATP utilization for sarcomere contraction^52,53, altered myocardial substrate utilization, and reduced ATP synthesis due to mitochondrial dysfunction⁵⁴. Inefficient energy utilization for force generation is attributed to diverse changes in sarcomere properties, including increased ATP utilization due to a shift in the DRX-SRX equilibrium^5,52. Myosins in the SRX state consume 5-fold less ATP compared to DRX myosins, and HCM mutations that reduce SRX proportions would chronically increase ATP consumption and increase the metabolic load on the heart. Profound alterations in cellular metabolism have indeed been observed in studies of human cardiomyocytes carrying pathogenic MYH7 variants destabilizing the SRX conformation^58,59. Our results show that in addition to destabilizing the SRX conformation, HCM mutations can alter the rate at which myosins in the DRX state consume ATP. For the Y115H and E497D mutations, myosins in the SRX state consume 12-fold and 23-fold less ATP, respectively, compared to DRX myosins (Supplementary Table 2). These measurements suggest that increased proportions of Y115H and E497D mutant myosins in the DRX state would result in significantly higher ATP consumption. It remains to be seen how this might affect the cellular metabolism of cardiomyocytes expressing these mutant myosins. Although our experiments are focused on the biophysical properties of mutant proteins, these data offer insights into diverse mechanisms by which an HCM mutation can possibly alter myocardial energetics.

In conclusion, the results we present here highlight that HCM mutations away from the IHM interfaces can allosterically disrupt myosin autoinhibition. More importantly, our results highlight that, unlike HCM mutations that directly lie at the IHM interfaces, predicting the indirect or allosteric effect of an HCM mutation on the stability of the IHM is not straightforward. The effects of HCM mutations on myosin’s regulation by autoinhibition are nuanced and may be evident only upon experimental characterization, as for the Y115H mutation (previously predicted to affect motor activity but not IHM stability²⁴). A previous study suggested that IHM structural models be leveraged to improve molecular prediction of clinically actionable human variants because they found that myosin variants with unknown significance (VUS) that altered residues participating in IHM interfaces (IHM+) were associated with higher rates of adverse clinical outcomes compared to variants that did not affect IHM stability (IHM-)⁵⁸. However, our results show that an accurate classification of a variant as IHM- is not entirely possible due to the indirect and remote effects of mutations on IHM stability that are not easily predicted.

Methods

Protein expression and purification

Three different recombinant human β-cardiac myosin constructs were used for WT and mutant proteins in this study: single-headed sS1 (res 1–808), two-headed short-tailed 8-hep HMM (res 1–893) and two-headed long-tailed 15-hep HMM (res 1–942). Myosin heavy chain (MYH7) was co-expressed with the human ventricular essential light chain (MYL3) carrying an N-terminal FLAG tag cleavable by TEV protease using adenoviruses in murine C2C12 myoblasts (ATCC) differentiated into myotubes; adenoviruses were generated in HEK293T cells (ATCC) using the AdEasy Vector System (Qbiogene Inc, Carlsbad, CA, USA) and purified by cesium chloride gradient centrifugation. MYH7 constructs contained a C-terminal GSG-RGSIDTWV affinity tag, and a GCN4 leucine zipper placed immediately before the C-t tag ensured dimerization in the two-headed constructs. C2C12 cells were grown in 15 cm dishes (~10 dishes per construct) and infected with MYH7 and MYL3 adenoviruses 48 h post initiation of differentiation into myotubes. Cells were harvested into harvesting buffer (20 mM Imidazole, pH 7.5, 50 mM NaCl, 20 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 10% sucrose, 3 mM ATP, 1 mM DTT, 1 mM PMSF and Roche Protease Inhibitors), 1 ml per dish, 96 h post infection using a cell scraper, and immediately flash frozen in liquid nitrogen (LN₂).

All steps for protein purification were carried out at 4 °C (on ice or in a cold room). Cell pellets were thawed and supplemented with 3 mM ATP, 1 mM DTT, 1 mM PMSF, Roche Protease Inhibitors (1 tablet per 50 mL) and 0.5% Tween-20. Cells were lysed by 50 strokes of a dounce homogenizer and the lysate was incubated for 20–30 min to allow for the assembly of contaminating endogenous full length (FL) mouse myosin from C2C12 cells into filaments (aided by the low NaCl and high MgCl₂ content of the harvesting buffer), which pelleted down in the subsequent spin at 30,000 rpm for 30 min at 4 °C. The clarified supernatant was incubated with anti-FLAG resin for 1–2 h to capture β-cardiac myosin protein, followed by washing away non-specifically bound proteins by centrifugation using wash buffer (20 mM Imidazole pH 7.5 containing 150 mM NaCl, 5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF, 3 mM ATP, 10% sucrose and Roche protease inhibitors). For sS1 constructs, myosin was then dissociated from the resin by overnight incubation with TEV protease, cleaving the FLAG tag on ELC. For 8 and 15-hep constructs, protein-bound resin was further incubated with regulatory light chain (RLC) depletion buffer (20 mM Tris-HCl, pH 7.5, 200 mM KCl, 5 mM CDTA, pH 8, 0.5% Triton-X-100 and 1 mM ATP) for 75 min to dissociate endogenous mouse RLC. Resin was extensively washed with wash buffer before incubating it with 10-fold molar excess of His-tagged human RLC (purified from E. coli as described previously²²) for 2–3 h. Excess unbound RLC was washed away and the resin was incubated with TEV protease overnight, cleaving the FLAG tag on ELC and His tag on RLC. The next morning, dissociated myosin (sS1 or 8-hep or 15-hep) in the supernatant was collected by spinning down the resin and the protein was further purified (to remove TEV protease and endogenous mouse FL myosin) by anion exchange chromatography using a 1 mL HiTrap Q HP column (Cytiva) attached to an ÄKTA FPLC system. Myosin typically elutes at ~250 mM NaCl; fractions were pooled based on SDS-PAGE analysis and concentrated using Millipore Amicon® ultra centrifugal filters (50 kDa MWCO for sS1, 100 kDa MWCO for 8 and 15 hep) to a concentration of 10–20 µM which was determined using the BIO-RAD Quick Start Bradford assay kit. Freshly prepped protein was directly used for ATPase assays, and aliquots were flash frozen in LN₂ for later use for in vitro motility and optical trapping experiments. 8-hep and 15-hep constructs were designed because 2-hep myosin has an uncharacteristically tight K_app for actin, and 25-hep myosin has increased protein loss during the purification process because of a significant overlap between the ion-exchange elution profiles of 25-hep myosin and endogenous full length mouse myosin from C2C12 cells.

Actin-activated ATPase

Steady-state actin-activated ATP turnover rates were measured using a plate based NADH-coupled assay as previously described^34,48. G-actin was prepared⁶⁰ from acetone powder from bovine cardiac LV tissue and polymerized to F-actin by dialyzing 5x into assay buffer containing 10 mM imidazole, pH 7.5, 5 mM KCl, 4 mM MgCl₂ and 1 mM DTT. F-actin aliquots (130–160 µM) were flash frozen in LN₂ for later use. F-actin, thawed on ice, was thoroughly mixed with 1:100 molar ratio of gelsolin (purified as described previously from E. coli⁶¹) and incubated on ice for ~30 mins. In a 96-well clear plate and 100 µL reaction volume, the actin-gelsolin mixture was diluted in assay buffer to achieve 8–10 dilutions in the 0–80 µM concentration range, and mixed with freshly prepped myosin to achieve 25 nM final concentration (250 nM in the zero actin well to measure basal ATPase rate). Actin and myosin were thoroughly mixed by constant shaking for 10 min in a microplate reader at 23 °C, following which the ATPase reaction was initiated by carefully pipetting in 20 μL of a 5X coupling solution containing lactate dehydrogenase (100 U/mL), pyruvate kinase (500 U/mL), phospho(enol) pyruvate (2.5 mM), ATP (10 mM) and NADH (2 mM). Any air bubbles were quickly eliminated by tapping with a needle and the plate was shaken for additional 5 min before reading absorbance at 340 nm every 15–30 s for 25 min at 23 °C. Data was collected using the SkanItTM software. The concentration of ADP produced was calculated from the absorbance values using an ADP standard curve (0–300 µM) set up on the same plate. ADP produced over time in each reaction well was plotted and the slope of the plot (rate) was divided by myosin concentration in that well and plotted against actin concentration (after subtracting the basal rate from zero actin well from the rate at each actin concentration). ATP turnover rates (k_cat) and the apparent affinity for actin (K_app) were determined by fitting the data to the Michaelis–Menten model of enzyme kinetics using GraphPad Prism versions 9 and 10. For sS1 constructs, 3 technical replicates of each actin concentration were set up on a single plate for WT and two mutants. For HMM constructs, WT 8-hep control was always assayed together with mutant 8 and 15-hep proteins in 3 technical replicates on a single plate. Data from 3 or more independent protein preparations with 3 technical replicates in each assay were averaged (mean ± SEM are reported here). Each biological replicate involved comparison of WT and mutant proteins that were purified in tandem from the same batch of C2C12 cells.

In vitro motility

Motility assay was carried out as described previously³². 5–7 flow channels were created on a glass slide using double-sided tape and #1.5 coverslips coated with 0.1% nitrocellulose/0.1% colloidin in amyl acetate. Myosin molecules were tethered via their C-terminal affinity peptide tag onto nitrocellulose coated coverslips before flowing in fluorescently labelled F-actin filaments. Briefly, 5–6 µL of the following solutions were sequentially flown into the channels and incubated as noted: SNAP-PDZ for 2 min (3 µM in assay buffer containing 25 mM Imidazole pH 7.5, 25 mM KCl, 4 mM MgCl₂, 1 mM EGTA and 10 mM DTT; SNAP-PDZ was purified from E. coli as described previously³²), blocking buffer for 2 min (assay buffer with 1 mg/ml BSA, ABBSA), myosin twice for 2 min (~200 nM in ABBSA), blocking buffer for 1 min and final GO buffer (2.5 nM TMR-phalloidin labelled F-actin, 2 mM ATP and an oxygen-scavenging system containing 0.4% glucose, 0.11 mg/ml glucose oxidase and 0.018 mg/ml catalase in ABBSA). Slides were typically imaged within 20 min of flowing in the GO buffer on a Nikon Ti-E inverted microscope with a 100x TIRF objective on an Andor iXon + EMCCD camera. Three movies were acquired per channel using the Micro-Manager software; images were taken at 2 Hz for 30 s with 300 ms exposure using the 561 nm laser channel. Velocity measurements are extremely temperature sensitive – to minimize the error in measurements arising from temperature fluctuations, objective temperature was maintained at 23 ± 0.3 °C as closely as possible. Filament tracking and velocity analysis was performed by analysing the movies using FAST (Fast Automated Spud Trekker³²) using the following parameters: window size n = 5, path length p = 10, percent tolerance pt = 20, and minimum velocity for stuck classification minv = 80 nm/s. Filtered mean velocity (MVEL₂₀), which is more appropriate for analysing unloaded motility data, is reported here: it is the mean of all filaments moving with a non-zero velocity after filtering out filaments with non-smooth motion (standard deviation in velocity is 20% or more than its mean velocity). When needed, myosin was deadheaded as follows: myosin was mixed with 10–20 fold molar excess of F-actin on ice for ~5 min, followed by addition of 2 mM ATP and further 3 min incubation on ice before spinning the mixture at 95,000 rpm (Ti-100 ultra centrifugal rotor) for 20 min at 4 °C. Unliganded myosin in the supernatant was used for motility experiments (in some cases subjected to another 1–2 rounds of deadheading if stuck percentages remained high, >10%).

Single nucleotide turnover assay

DRX and SRX rates and proportions for WT and mutant proteins were determined from single mant-ATP turnover assays (STO), largely as described previously⁴³. The SRX trends were extracted from experiments with the 15-hep constructs and the 8-hep constructs served as controls. Myosin was buffer exchanged using 100 kDa MWCO 0.5 mL Amicon filters to achieve >10⁵ dilution in STO 100 assay buffer (100 mM KOAc in 10 mM Tris-HCl pH 7.5, 4 mM MgCl₂, 1 mM EDTA and 1 mM DTT), followed by a spin at 40,000 rpm (Ti-100 ultra centrifugal rotor) for 15 min at 4 °C to remove protein aggregates, if any. The protein was diluted to achieve a final concentration of 500 nM myosin heads in 80 µL STO assay buffer containing 6.25 mM KOAc (by mixing appropriate volumes of 0 and 100 mM KOAc containing STO assay buffers) and transferred to a well of a 96-well black plate placed in the Infinite® M Nano+ microplate fluorescence reader attached to the Te-Inject™ injector module. Software controlled injector syringes were pre-loaded with 4 mM mant-ATP ((2’-(or-3’)-O-(N-Methylanthraniloyl) Adenosine 5’-Triphosphate, Thermo-Fisher Scientific) and 40 mM ATP. Single nucleotide turnover reaction was initiated by dispensing 10 µL mant-ATP into the protein containing well from syringe A and the mixture was aged for 15 s before syringe B dispensed 10 µL ATP, which was immediately followed by recording fluorescence emission at 470 nm (excitation: 405 nm) every second for 16 min, using the i-controlTM software. The final concentrations of different components in the reaction mixture were: 400 nM myosin heads, 400 nM mant-ATP, 4 mM ATP and 5 mM KOAc (or 5, 25, 100 mM in salt-dependent measurements). The injector module reduced the dead-time between adding unlabelled ATP and recording the first fluorescence reading to ~2 s, which was manually added to the kinetic data before fitting it to an unconstrained five parameter two phase decay using GraphPad Prism versions 9 and 10. STO assays were performed with either freshly prepped protein (within 24–48 h of protein purification) or flash frozen proteins; the data from the two was indistinguishable. 4–10 technical replicates were performed for each protein construct for each independent protein preparation (2 or more) and ambiguous fits occasionally resulting from mixing artefacts or a bubble in the optical path (<5%) were discarded.

Optical trapping

A dual-beam optical trap was used for harmonic force spectroscopy (HFS) measurements to obtain load dependent detachment kinetics of single myosin heads interacting with a fully stretched actin dumbbell.

Preparation of the sample chamber: First, ~1.6-μm diameter silica beads (Bangs Laboratories, Fishers, IN, USA) in a solution of 0.1% nitrocellulose and 0.1% collodion containing amyl acetate, were spin coated on top of a No. 1.5 glass coverslip. Then a sample chamber was made by sticking the silica bead coated coverslip on top of a 1-mm thick microscope slide using double-sided tape strips, while keeping the bead coated side facing the microscope slide. The tape strips were applied only on the edges of the microscope slides in order to leave enough space in the middle of the sandwich to form a chamber. Then the inside surface of the coverslip (containing the platform beads) was functionalized by flowing 15 nM SNAP-PDZ solution in AB buffer at pH 7.5 (25 mM KCl, 25 mM imidazole pH 7.5, 4 mM MgCl₂, 1 mM EGTA, and 10 mM DTT) for ~2 min. The low concentration of SNAP-PDZ was used to secure only a sparse presence of the SNAP-PDZ molecules on the surface and the use of the silica beads on the cover-slip platform ensured that some of the SNAP-PDZ molecules remained elevated on top of the beads. We then flowed in 1 mg/ml BSA in AB buffer (ABBSA buffer) solution for ~5 min to block the remaining exposed surfaces. A 50 nM solution of WT or mutant sS1 proteins containing the C-terminal PDZ-binding domain was then flowed into the chamber for ~2 min. During this step, the sS1 molecules get translationally immobilized inside the chamber by binding to the SNAP-PDZ molecules. Then ABBSA buffer was flowed in to wash away any unbound sS1. Finally, a solution of ABBSA buffer containing streptavidin-coated polystyrene beads (1-μm diameter, Bangs Laboratories), filaments of 0.3 nM TMR-phalloidin-labelled biotinylated actin (Cytoskeleton, Denver, CO, USA), 0.018 mg/mL catalase, 0.4% glucose, 0.11 mg/mL glucose oxidase, and 2 mM ATP, was flowed in. The chamber was then sealed with vacuum grease to avoid any evaporation and taken for the HFS experiments. The entire process of chamber preparation and HFS measurements was performed at 23 °C.

HFS measurements: For the HFS measurements, the two optical trap stiffnesses were kept between 0.08 and 0.10 pN/nm. The trap calibration was done by using the power spectrum method, described previously^33,36,38. After the trap calibration was completed, HFS experiments were initiated by trapping a streptavidin-coated bead in each optical trap. A biotin coated actin filament was then snared between the two beads by moving the sample stage in three-dimensions, while probing the location of the filaments by phalloidin fluorescence. Then, by slowly moving the trap beads away from each other (by steering the trap-beams) the actin filament was stretched to form a “dumbbell”. While keeping the stage under 200 Hz oscillation, the actin-dumbbell was then lowered towards a platform bead (containing stationary myosin heads), to find potential myosin-actin interaction events. Upon every successful actin-myosin interaction event, brief but strong displacement would occur in both the trap bead positions simultaneously, due to their direct association with the stage oscillation. When several hundred such interactions were probed from a single myosin molecule, the time-traces of the bead-positions were saved, and the data was considered for analysis. Here we note that, to ensure that we were observing single myosin molecules interacting with any given actin dumbbell, the SNAP-PDZ concentration was kept sufficiently low that, on average, only one out of ~8–12 such trials of fishing for myosin would result in detectable and consistent interactions.

The saved data (time-traces) were processed automatically, as described elsewhere. Each interaction was identified by simultaneous changes in the phase and the amplitude of both the trap beads. During the period of interaction, the oscillation of the stage ensures presence of varying amounts of overall assistive or resistive forces associated with each detachment. By employing maximum likelihood estimation from the event-durations within a user-defined force bin, the rate of detachments in the presence of varied external forces were obtained. The averages of detachment rates obtained against different forces for a single molecule were then fitted with Eq. (1) to obtain ${k}_{0}$ and δ. Further analysis of the same HFS data was done to obtain stroke sizes for each interacting myosin molecule. The average position and the oscillation (both amplitude and phase) of the actin-dumbbell (i.e. the actin filament stretched between two trapped beads) changes upon myosin interaction. Step sizes were determined from HFS data. A sine function was fitted to the trace to model the underlying oscillatory signal. This fitted sine wave was subtracted to isolate the net displacement. The plots of distribution of displacements across all binding events for one representative molecule of each WT, Y115H, and E497D, are given in Supplementary Fig. S7d–f. As shown, the distributions are unimodal and the centres of these distributions correspond to the step sizes, yielding a single step-size value per molecule. Data was analysed using MATLAB software (version R2021b).

ATP binding to myosin sS1

ATP binding to WT, E497D and Y115H myosin sS1 was monitored by observing the increase in the fluorescence of mant-ATP upon binding to the myosin head. Binding experiments were performed in 10 mM Tris-HCl pH 7.5 with 4 mM MgCl₂, 1 mM EDTA, 1 mM DTT, 100 mM KOAc. Experiments were performed on an SFM-3000 stopped-flow module (Biologic, Seyssinet-Pariset, France). Data was collected using the Bio-Kine software. Myosin sS1 (frozen or freshly prepared) was buffer exchanged into experimental buffer using 50 kDa MWCO Amicon filters (Cat # UFC5050BK, Millipore Sigma). Myosin sS1 and mant-ATP were rapidly mixed in 1: 9 or 2: 8 ratio and injected into an observation cell coupled to a MOS-200 detection system. The final concentration of myosin in the reaction mix was 0.1–0.2 µM, and the reaction was carried out at 5 different concentrations of mant-ATP ranging from 0.8 to 5 µM. The mant-ATP bound to myosin head was excited by exciting the tryptophan at 290 nm which excites mant by resonance energy transfer. The fluorescence of mant was collected at 458 nm using a bandpass filter (Cat#FF01 458/64-25, Semrock). The kinetic traces were fit to a double exponential rise equation in GraphPad Prism.

Crystallization of PPS-MD-WT, PPS-MD-Y115H and PPS-MD-E497D

Crystals of PPS-MD-WT (type A) were obtained by the hanging drop vapour diffusion method at 298 K. Hanging drops were made using a 1:1 (v:v) ratio of protein samples (7 mg/ml WT sS1 incubated with Mg.ADP and Vanadate, centrifugated at 21,100 × g for 10 min at 4 °C) against a precipitant solution (25 % PEG 3350 w:v, 0.25 M Lithium sulfate and 0.1 M Tris, pH 8). Optimized crystals were obtained using in situ proteolysis by addition of trypsin with a 1:500 w:w ratio to protein before centrifugation. Optimized crystals were obtained using micro seeding techniques and were cryo-cooled in liquid nitrogen in a cryo-protectant solution (10% Glycerol, 30% PEG 3350 w:v, 0.2 M Lithium sulfate and 0.1 M Tris, pH 8). Crystals of PPS-MD-Y115H (type B) were obtained by the hanging drop vapour diffusion method at 298 K. Hanging drops were made using a 1:1 (v:v) ratio of protein samples (7 mg/ml Y115H sS1 incubated with Mg.ADP and Vanadate, centrifugated at 21,100 × g for 10 min at 4 °C) against a precipitant solution (23 % PEG 3350 w:v, 0.3 M Lithium sulfate and 0.1 M Tris, pH 8). Optimized crystals were obtained using in situ proteolysis by addition of trypsin with a 1:500 w:w ratio to protein before centrifugation. Crystals grew spontaneously ~5 days after and were cryo-cooled in liquid nitrogen in a cryo-protectant solution (27 % Ethylene glycol, 26 % PEG 3350 w:v, 0.2 M Lithium sulfate and 0.1 M Tris, pH 8). Crystals of PPS-MD-E497D (type C) were obtained by the hanging drop vapour diffusion method at 298 K. Hanging drops were made using a 1:1 (v:v) ratio of protein samples (7 mg/ml E497D sS1 incubated with Mg.ADP and Vanadate, centrifugated at 21,100 × g for 10 min at 4 °C) against a precipitant solution (22% PEG 3350 w:v, 0.3 M Lithium sulfate and 0.1 M Tris, pH 8). Optimized crystals were obtained using in situ proteolysis by addition of trypsin with a 1:500 w:w ratio to protein before centrifugation and using micro-seeding. Crystals appeared after 3 days and were cryo-cooled in liquid nitrogen in a cryo-protectant solution (27% Ethylene glycol, 27% PEG 3350 w:v, 0.2 M Lithium sulfate and 0.1 M Tris, pH 8).

X-ray data processing structure determination and refinement

X-ray diffraction datasets were collected at Synchrotron SOLEIL, Gif-Sur-Yvette (Proxima 2 A beamline, λ = 0.95372 Å for type A, λ = 0.98011 Å for type B and λ = 0.98012 Å for type C). Diffraction data were processed using Autoproc⁶². Initial structure factors were obtained by molecular replacement with Phaser⁶³ using a search model generated by Swiss Model⁶⁴ with human β-cardiac myosin sequence and PDB 9F6C⁶⁵ as template. The initial Fo-Fc map showed clear density corresponding to H115 and D497 mutations in type B and type C respectively. The final models were obtained by several cycles of iterative edition with Coot⁶⁶ and model refinement with Buster⁶⁷. The data collection and refinement statistics are presented in Supplementary Table 3. PPS crystals were obtained either using Mg.ADP-BeFx or Mg.ADP-VO4, datasets were chosen based on resolution as the phosphate analogue nature does not influence the PPS conformation. *Figure 7: for the structural analysis of the local interactions changes resulting from missense mutation Y115H, including the water-mediated interactions, we used the high-resolution structure of WT bovine cardiac myosin bound to Mavacamten (PDB 8QYR) determined at 1.8 Å for comparison. This was done since the binding of Mavacamten does not affect the nucleotide binding site²⁹.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.