Drug specificity and affinity are encoded in the probability of cryptic pocket opening in myosin motor domains

Read the full article


The design of compounds that can discriminate between closely related target proteins remains a central challenge in drug discovery. Specific therapeutics targeting the highly conserved myosin motor family are urgently needed as mutations in at least 6 of its members cause numerous diseases. Allosteric modulators, like the myosin-II inhibitor blebbistatin, are a promising means to achieve specificity. However, it remains unclear why blebbistatin inhibits myosin-II motors with different potencies given that it binds at a highly conserved pocket that is always closed in blebbistatin-free experimental structures. We hypothesized that the probability of pocket opening is an important determinant of the potency of compounds like blebbistatin. To test this hypothesis, we used Markov state models (MSMs) built from over 2 milliseconds of aggregate molecular dynamics simulations with explicit solvent. We find that blebbistatin’s binding pocket readily opens in simulations of blebbistatin-sensitive myosin isoforms. Comparing these conformational ensembles reveals that the probability of pocket opening correctly identifies which isoforms are most sensitive to blebbistatin inhibition and that docking against MSMs quantitatively predicts blebbistatin binding affinities (R 2 =0.82). To test our ability to make blind predictions, we predicted blebbistatin’s binding affinity for an isoform (Myh7b) whose blebbistatin sensitivity was unknown. Encouragingly, we find good agreement between the predicted and measured IC50 (0.67 µM vs. 0.36 µM). Therefore, we expect this framework to be useful for the development of novel specific drugs across numerous protein targets.


Drug development requires the discovery of compounds which specifically target one member of a protein family without triggering side effects that arise from interactions with other related proteins. Myosins are a family of motor proteins that are drug targets for heart diseases, cancer, and parasitic infections. Here, we investigate why the compound blebbistatin specifically inhibits some myosins more potently than others, even though its binding site is closed in all known experimental structures. We find that the blebbistatin binding pocket opens in molecular dynamics simulations of certain myosin motors, and that the probability of opening predicts how potently blebbistatin inhibits a particular motor. Our work suggests that differences in cryptic pocket formation can be exploited to develop specific therapeutics.

Article activity feed

  1. eLife assessment

    This important study presents insights into how conformational dynamics differentially influences drug specificity and affinity in myosin isoforms using computational approaches. The evidence supporting the conclusions is convincing, establishing a relationship between inhibition and protein dynamics using state of the art computational techniques followed by experimental validation. The work will be of broad interest to computational biophysicists and medicinal chemists.

  2. Reviewer #1 (Public Review):

    Targeting allosteric sites, including cryptic sites holds great potential for achieving drug design that distinguishes between isologous protein targets. Here, Meller et al seek to reveal the mechanisms by which blebbistatin, a selective allosteric inhibitor of myosins achieves selectivity between proteins with high structural and sequence similarity. Blebbistatin binds in a supposed cryptic pocket, and authors explore the hypothesis that this selectivity is modulated by dynamics of opening in the cryptic pocket. Studies use MD simulations to show that while cryptic pockets do not exist in experimental structures, they appear in simulations. Markov state models (MSM) generated from simulations are used to quantify probability of pocket opening. The same methods are used to show that ADP-bound myosins are more likely to open than ATP-bound state, consistent with higher blebbistatin binding affinity observed in the ADP-bound state. Myosin-II proteins are shown to have higher probability of opening than non-myosin-IIs, along with an observed correlation of probability with IC50. By docking blebbistatin into structures derived from MSMs, authors show a correlation between predicted binding affinity from docking and experimental binding. Binding was correctly predicted for a new isoform in a blind study, further establishing the utility of using conformational ensembles to predict sensitivity of blebbistatin binding to myosins.

  3. Reviewer #2 (Public Review):

    This study utilizes extensive molecular dynamics simulations to probe the binding of a widely-used myosin II inhibitor to several closely-related myosin isoforms. The authors focused on so called 'cryptic' drug binding site, which is not apparent in isolated 'apo' states of the proteins, but are unveiled in simulations. The probability of unveiling these sites was implicated as the factor that distinguished myosins that bind blebbistatin from those that do not. Importantly, they focus on targetting an allosteric site, which can circumvent issues with targeting the binding site of the cognate ligand that can lead to nonselective binding in other targets. These simulations were accompanied by markov state model decompositions of those trajectories to isolate states conducive to drug binding, which were assessed using molecular docking to yield aggregated drug binding free energies. To demonstrate the reliability of their model, they performed a blinded prediction for an as-of-then uncharacterized myosin variant and found strong agreement with experimentally measured affinity (micromolar). Another finding of note includes identifying the ADP/Pi-bound myosin state as the preferred conformation for blebbistatin, which is line with the drug's inhibition of myosin ATPase activity.

    This study is impactful for several reasons. Firstly, the authors provide a molecular basis for the allosteric inhibition of myosin II by blebbistatin, through extensive GROMACS molecular dynamics simulations. They implicate a cryptic binding site that is obscured in apo state structures of the enzyme, for differences in the ligand's affinity measured for several myosin proteins. Their simulations indicate that the binding site spontaneously opens for apo state myosin isoforms that are inhibited by blebbistatin, but remains closed for other myosins. Moreover, they discovered that the drug's apparent affinity is proportional to the probability of forming the open conformation of the cryptic binding site in the apo state structure. This knowledge is important for guiding selective drug development, because there is generally lesser conservation in allosteric binding sites, e.g. off-target binding is less likely, relative to the primary site for endogenous ligands that are shared for homologous proteins. In addition, they used Markov State Models (MSMs) to identify protein conformational states that are conducive to ligand binding, to which they docked blebbistatin using Autodock Vina. The predicted drug affinities for each state in the MSM ensemble were weighted according to the state's probability, which yielded an aggregate estimate of drug affinity that strongly agreed with experimental data. To further establish the approach's validity, the modeler co-authors predicted the affinity for blebbistatin binding to a myosin protein that had not yet been characterized. The predicted affinity was also found to be in very good agreement with the affinity ultimately reported by the experimentalist co-authors. Overall, this is a strong computational approach applied to a drug/target interaction that is invaluable to the research and clinical community. The researchers' claims are well-supported by the provided data.

    A prominent limitation in the study is that the contributions of entropy in their `multi-state' ligand binding model is not apparent - at the very least I would anticipate an entropic contribution from the states identified from the MSM characterization of the apo myosin simulations. Relatedly, the docking scores likely account for changes in ligand entropy upon binding, but it is unlikely that the 3 structures selected from each MSM state would be sufficient to describe the protein disorder within the state. This limitation does not impact the novelty of the study, but is rather an opportunity to discuss extension of the method in future applications. Additionally, by design the myosins used for the study shared 90% or greater sequence identity. On one hand, this is a great set for testing the limits of predicting selectivity. On the other hand, it would be helpful to know how the approach might work for myosins with lower homology but very similar tertiary structures. Would there still be a cryptic site amenable to drugging, and if so, would its open probability necessarily scale with ligand binding affinity? On a related note, would this approach perform best for well-buried ligand binding domains, or could it also be expected to perform well for more surface exposed sites or those with extensive loops?

    It is expected that this work will be impactful to the scientific community on two fronts. The first of which is establishing a molecular mechanism of selective myosin inhibition, which will be invaluable for drug design efforts targeting the myosin II cardiac isoform in particular. The abundance of ATPases and ATP-responsive proteins in cardiac tissues renders difficult the task of designing molecular species that competitively bind to the ATP pocket - targeting an allosteric site with lesser homology across isoforms is a compelling alternative. The use of markov state models with standard docking techniques to improve binding free energy estimates among closely related proteins has the potential to be broadly used by the computer aided drug design community. The potential for widespread adoption is tempered by the authors' use of a specialized resource, folding at home, to achieve millisecond-length simulations. Enhanced sampling techniques, however, may yield similar results with smaller simulation requirements.

  4. Reviewer #3 (Public Review):

    In this work the authors show, using different computational methods (molecular dynamics simulations, Markov state modeling, docking) that the probability of pocket opening in the isoforms of the protein myosin is an important determinant of the potency of the allosteric inhibitor blebbistatin. The data from the work supports the conclusions, and clearly shows that blebbistatin inhibits more potently myosin isoforms with a higher probability of pocket opening. The authors developed a protocol combining the probability of pocket opening from Markov state modeling with docking scores to estimate the IC50 values of blebbistatin for different myosin isoforms, achieving a good correlation between computed and experimental IC50 values (coefficient of determination of 0.82). The authors also tested their computational protocol prospectively, providing an estimate of IC50 for blebbistatin for the myosin isoform Myh7b which was in line with the experimental results. The computational protocol developed by the authors can be very useful for the community, since it can be applied to any protein containing cryptic pockets.

    A major strength of the work is the prospective test of the computational protocol they developed, and the subsequent conclusion that the IC50 estimated by their method, 0.67 µM, was similar to the value obtained in the experiments, 0.36 {plus minus} 0.08 µM.

    A major weakness of the work is the use of docking scores to compute the IC50 of blebbistatin for the different isoforms of myosin. Docking scores are usually empirical and previous works have shown that they are usually poorly correlated with experimental binding affinities.