Super-relaxed or super-stressed: Modeling length-dependent activation in cardiac muscle

Main text

The heart needs to generate sufficient force and power output to perfuse the body over a wide range of physiological conditions ranging from sleep to intense exercise. The heart has evolved complex regulatory mechanisms to dynamically adapt to changing conditions over a range of length scales and timescales (e.g., thin and thick filament-based regulation, mechanosensing molecules, neuronal stimulation, and tissue remodeling). Disruption of these regulatory mechanisms can cause life-threatening diseases. For example, mutation of proteins in the sarcomere, the fundamental contractile unit of the heart, can impair cardiac contraction and cause familial cardiomyopathies, which are leading causes of heart failure and sudden death. Current treatments for cardiomyopathies have improved overall patient outcomes, but many patients do not respond to these treatments, and there is an outstanding need to develop new strategies that improve patient survival and quality of life (1).

Given this pressing medical need, there is an intense interest in dissecting the mechanisms of cardiac regulation and translating these mechanistic insights to treating human disease. Mechanistically probing cardiac contraction requires tools drawn from the nexus of biology, chemistry, and physics, and as such, the biophysical community is ideally positioned to make valuable contributions in this exciting area. Decades of biophysical studies, both computational and experimental, have revealed many of these cardiac regulatory mechanisms and enabled the field to start to decipher the connections between genotype and phenotype for patient-specific mutations (2). Moreover, the community has been involved in leveraging their deep knowledge of molecular mechanism to design new therapeutic strategies that modulate muscle contraction, and there are multiple small molecules in development that target the regulation of cardiac contractility for heart failure and cardiomyopathies (3). These exciting developments have been driven by pairing experimental systems with theoretical models to provide new insights into the fundamental mechanisms regulating cardiac contraction.

One key regulatory mechanism in cardiac muscle is the Frank-Starling relationship, where increasing the volume of the blood returned to the heart during diastole increases the stroke volume during systole (Fig. 1 A) (4). In other words, when the heart fills with more blood during diastole, it increases its contraction during the next beat to meet the body’s increased need. Importantly, the Frank-Starling relationship is blunted in patients with heart failure, impairing the heart’s ability to modulate contractility in response to increased physiological demands. At the level of single cardiomyocytes, the Frank-Starling relationship is manifested as length-dependent activation (LDA), where stretching a cardiomyocyte to longer sarcomere lengths will cause the myocyte to generate more force during contraction (4). The total force exerted by a cardiomyocyte is the sum of the active force generated by myosin-driven contraction and the passive force exerted by stretching of elastic elements, notably titin. Two hallmarks of LDA seen at longer myocyte stretches are increased maximal force production generated at high calcium and increased thin filament activation at submaximal calcium levels (seen as a shift toward lower calcium in the force-calcium relationship). Although LDA is well established in physiology, it has been a challenge for the field to dissect its molecular mechanism, and several competing but not mutually exclusive mechanisms have been put forward to explain LDA, including changes in sarcomeric organization, stretching of titin, increased calcium binding to troponin with length, and activation of myosin binding protein C (4).

Figure 1.

Length-dependent activation (LDA). (A) The Frank-Starling relationship for a healthy and failing heart. Bottom cartoon shows the sarcomere getting stretched with increasing end diastolic volume. (B) Cartoon showing force-dependent recruitment of myosin cross-bridges from the inhibited OFF state to the active ON state.

Recently, a new mechanism for LDA was put forward based on dynamic regulation of the number of myosin heads in the thick filament available to generate force (5). Although thin filament-based regulation of cardiac contraction has been appreciated for decades, the significance of thick filament regulation in cardiac contraction has only become apparent more recently (6). Biochemical studies have demonstrated that cardiac myosin can form a super-relaxed state with slow ATPase activity, and that the presence of the myosin motor domains along with part of the tail significantly increases the population of this inhibited state. Moreover, x-ray diffraction studies and recent cryo-electron microscopy structures have demonstrated that cardiac myosin can adopt an interacting heads motif, where the myosin heads bind to each other, inhibiting their ability to interact with the thin filament. The biochemical super-relaxed state appears to be correlated with the structural interacting heads motif; however, these states are not necessarily strictly coupled (6).

Taken together, both biochemical and structural studies suggest that myosin heads in the thick filament can be dynamically recruited to an active ON state or sequestered to an inhibited OFF state to adjust the number of myosin heads available for force generation (Fig. 1 B). Some cardiomyopathy mutations affect the ability of myosin to form the OFF state, leading to altered contractility (7). Importantly, recent x-ray diffraction studies have demonstrated that additional myosin heads can be recruited to the ON state by mechanically stretching the sarcomere (5). This led to the proposal that the thick filament functions as a mechanosensor in muscle, where mechanical loads shift the pool of myosin heads in the thick filament from the OFF state to the ON state (Fig. 1 B), and that this mechanism is important for LDA; however, new modeling approaches are needed to capture this behavior.

Recently, Lewalle et al. (8) tested whether it is possible to computationally recapitulate key aspects of LDA by incorporating force-dependent recruitment of myosin heads from the OFF to ON state into a well-stablished model of cardiac contraction developed by Land et al. (9). This model is based on solving a system of time-dependent ordinary differential equations representing transitions between known biochemical states. The original Land et al. model included a nonphysical empirical term that was used to generate LDA. To model LDA based on mechanism, Lewalle et al. removed this term and added an OFF state with load-dependent transitions. They applied this slightly modified model to demonstrate that the introduction of a load-dependent OFF state was sufficient to recapitulate aspects of LDA including increased maximal force production at high calcium levels and a shift in the force-calcium response toward submaximal calcium activation. It is important to note that although the incorporation of load-dependent recruitment of myosin heads is sufficient to produce LDA, it does not exclude potential contributions from other mechanisms (4).

Next, Lewalle et al. further investigated this force dependence. The total force on cardiac muscle is the sum of the active forces from cross-bridge cycling and passive forces due to the viscoelastic nature of muscle, and therefore, Lewalle et al. tested whether LDA in their model depends on the active force, the passive force, sarcomere strain, or the total tension on the muscle. They found that they could best recapitulate LDA by accounting for the total force. In this model, both the active force generated by myosin and the passive force exerted by stretched elastic elements form a feedback loop that regulates contractility, and the authors propose that this feedback loop is an essential feature of LDA.

Finally, Lewalle et al. applied their framework to modeling the effects of the drug, mavacamten, which was developed to treat patients with hypertrophic obstructive cardiomyopathy (3). Many of the mutations that cause hypertrophic cardiomyopathy occur in sarcomeric proteins and cause increased contractility at the molecular scale (2,10). The mutation-induced initial insult of altered contractility at the molecular scale in turn leads to the activation of downstream adaptive and maladaptive pathways that drive the complex and protean disease pathogenesis, including transcriptional and proteomic remodeling at the cellular scale and fibrosis and ventricular hypertrophy at the organ scale (1). Mavacamten was identified through a screen for small molecules that bind to myosin and reduce its ATPase rate (3). This reduction in myosin’s ATPase rate is in part due to drug-induced stabilization of the myosin OFF state, leading to reduced contractility that can help to offset the increased contractility caused by hypertrophic cardiomyopathy mutations. Structural studies have shown that mavacamten binds to a pocket near the converter domain, where it causes several allosteric changes that inhibit phosphate release while stabilizing interactions necessary for the formation of the interacting heads motif. The EXPLORER-HCM trial demonstrated that patients treated with mavacamten showed improvement in symptoms, outflow tract obstruction, and slowed disease progression, and as such, mavacamten became the first FDA-approved medication specifically for hypertrophic obstructive cardiomyopathy (3). Lewalle et al. showed that their model can capture several key features of mavacamten seen experimentally.

A major strength of the approach taken by Lewalle et al. is the fact that features of LDA arise from a relatively simple ordinary differential equation-based approach, which has a lower computational cost compared to other models while still being able to capture key aspects of muscle contraction. Moreover, the authors have made their code publicly available to the broader community to enable others to apply it to their own systems. It should be noted that, as with all models, there are some necessary simplifying assumptions that limit the model’s ability to capture all aspects of muscle contraction, especially with regard to single-molecule behaviors. The authors point out some experimental observations that are not fully captured using their model, and they discuss how inclusion of additional factors might help to recapitulate these behaviors. That being said, it is often easier to test hypotheses and extract insights from computational modeling using the simplest system that can faithfully recapitulate the phenomenon of interest. As such, the model put forward by Lewalle et al. will likely be of use to many in the field.

There are several outstanding questions in the biophysical community that might be addressed by pairing high-quality experimental data with insights gleaned from computational approaches. For example, a frequently cited hypothesis in the contractility field is that hypertrophic cardiomyopathy is caused by molecular hypercontractility, whereas dilated cardiomyopathy is caused by molecular hypocontractility (2). Excellent in vitro biophysical studies have quantified how mutations affect key parameters associated with contractility; however, these studies have shown that a mutation might show gain of function in one assay and loss of function in another. For example, the R403Q mutation in cardiac myosin causes hypertrophic cardiomyopathy; however, it decreases the ATPase and contractile function of individual motors while simultaneously increasing the number of myosin heads in the ON state (7). Computational modeling will be necessary to disentangle these competing effects on contractility, and it is possible that composite endpoints derived from modeling may be useful for refining hypotheses regarding the mechanisms driving different forms of cardiomyopathy. Finally, there are multiple molecular pathways that can lead to cardiomyopathies, and it has been suggested that it might be possible to improve patient outcomes by taking a mechanism-based precision medicine approach (1). In such an approach, subgroups of patients with common molecular dysfunction could be given treatments that optimally reverse the molecular dysfunction driving the disease pathogenesis. Testing this hypothesis will require experimental and computational approaches that enable the community to model the effects of patient-specific mutations and then test what molecules might be optimal to rescue the mutation-induced dysfunction. Taken together, it is an exciting time for both experimentalists and modelers, with opportunities to leverage our biophysical tools to make an impact on human health.

Author contributions

M.J.G. wrote and edited the manuscript.

Editor: Guy Genin.