Computational Methods Elucidate Consequences of Mutations and Post-Translational Modifications on Troponin I Effective Concentration to Troponin C

Abstract

Ca²⁺ binding to cardiac troponin C (cTnC) causes a conformational shift that exposes a hydrophobic patch (cTnC_HP) for binding of the cTnI switch peptide (cTnI_SP), ultimately resulting in contraction of the heart. The inhibitory peptide (cTnI_IP), attached at the N-terminal end of the cTnI_SP, serves as a tether for the cTnI_SP to the rest of the troponin complex. Due to this tethered nature, the cTnI_SP remains within proximity of the hydrophobic patch region, resulting in the cTnC_HP experiencing an elevated “effective concentration” of the cTnI_SP. Mutations to the cTnI_IP region have been hypothesized to cause disease by effecting the ability of the cTnI_SP to ‘find’ the hydrophobic patch, resulting in alterations to the heart’s ability to contract normally. We tested this hypothesis using molecular dynamics (MD) simulations of the troponin complex using a model that contained all three subunits of troponin: C, I, and T (cTnT). We developed methods that allowed us to quantitatively measure the effective concentration of the cTnI_SP from the simulations. A significant reduction in cTnI_SP effective concentration was observed when the cTnI_IP was removed from the system, showcasing the importance of a tethered cTnI_SP. Through accelerated MD methods, we proposed the minimum effective concentration of a tethered cTnI_SP to be approximately 21mM. Modification of the cTnI_IP via PKC mediated phosphorylation of T143 was shown to significantly increase the estimated effective concentration of cTnI_SP, help the cTnI_SP find the cTnC_HP more effectively, and maintain the relative shape of the cTnI_IP when compared to the native model. All of this data indicates that pT143 may be able to help promote binding of cTnI_SP to the cTnC_HP. We then tested six mutations within the cTnI_IP region that are known cTnC Ca²⁺ sensitizing mutations, and have been linked with cardiomyopathy. We did not observe a significant reduction in effective concentration upon introduction of these mutations, however we did observe increased variability in the flexibility and dynamics of the cTnI_IP region when compared to native. Our observations led us to hypothesize that the mechanism by which these cardiomyopathic mutations effect Ca²⁺ sensitivity is by altering the off rate of cTnI_SP from the hydrophobic patch.

Graphical Abstract

Introduction

The cardiac troponin complex (cTn) is responsible for modulating the interaction between myosin and actin necessary for muscle contraction.¹ cTn is made up of three subunits: Ca²⁺ binding subunit (cTnC), inhibitory subunit (cTnI), and tropomyosin-binding structural subunit (cTnT) that anchors the complex to the thin filament. Although there are four EF-hand Ca²⁺ binding sites in cTnC, only Ca²⁺ binding to site II regulates cardiac muscle contraction. When Ca²⁺ binds to this site, it causes a conformational shift within the N-terminal region of cTnC that exposes a hydrophobic patch (cTnC_HP, residues 20, 23, 24, 26, 27, 36, 41, 44, 48, 57, 60, 77, 80, 81). This patch promotes binding of the cTnI switch peptide (cTnI_SP, residues 149–164), removing the C-terminal end of cTnI from its binding site on actin, allowing myosin to interact with actin for muscle contraction.² Differences between the Ca²⁺ bound and unbound structures can be seen in Figure 1. The sequence directly N-terminal of the cTnI_SP is the largely unstructured inhibitory peptide (cTnI_IP, residues 137–148) that essentially acts as a tether for the cTnI_SP to the rest of the cTn complex.

Figure 1. Ribbon representation of cardiac troponin.

A) Ca²⁺ free, cTnI_SP unbound, cardiac troponin (PDB: 6KN7, chains a, b, and c). B) Ca²⁺ bound, cTnI_SP bound, cardiac troponin (PDB: 6KN8, chains a, b, and c). cTnT is shown in purple, cTnC is shown in red, helices that line the cTnC_HP are shown in orange the IT arm of cTnI is shown in cyan, with the cTnI_IP in dark blue and the cTnI_SP in green, Ca²⁺ ions added by Autodock Vina are in yellow.

Cardiomyopathies are a collection of diseases that can thicken, stiffen or thin out the heart muscle, which can lead to heart failure or sudden cardiac death.³ Hypertrophic cardiomyopathy (HCM) is characterized by increased thickness of the interventricular septum and decreased left ventricular chamber volume, leading to impaired diastolic function of the heart.⁴ HCM affects 1 out of 500 people with more than 70% of cases being familial. Restrictive cardiomyopathy (RCM) is a rarer disease state, characterized by impaired ventricular filling and a decreased diastolic volume. The prognosis for RCM is worse than HCM given that of individuals identified with restrictive cardiomyopathy, 10% die by age 19 and 50% die before age 35.⁵ HCM and RCM mutations have been identified within all subunits of cTn and have been experimentally shown to cause a shift in the Ca²⁺ sensitivity to site II of cTnC.⁵

How exactly cardiomyopathic mutations outside of the Ca²⁺ binding site of cTnC can cause a shift in Ca²⁺ sensitivity to this site has been the focus of intense research.^6–18 A previous study by Siddiqui et al.¹⁹ proposed that the availability of the cTnI_SP to the cTnC_HP may be one explanation to the variability in Ca²⁺ sensitivity. It was suggested that due to the tethered nature of the cTnI_SP via the cTnI_IP, the cTnC_HP experiences an elevated effective concentration of cTnI_SP. During traditional molecular dynamics (MD) simulations of the cTn complex, a study by Dvornikov et al. observed a significant reduction in contacts between cTnI 145 and cTnC E56, E59, and E63 when the R145W RCM or T143E phosphomimic mutations were present.¹⁵ Given that cTnC has been shown to have a higher sensitivity for Ca²⁺ when cTnI_IP is tethered to the complex, in addition to the observed reduction of contacts between cTnI_IP and cTnC observed in MD simulations, we hypothesized that alterations to the cTnI_IP region may change the effective concentration, and therefore, the Ca²⁺ sensitivity.

The focus of this study was to quantify how the cTnI_IP works to help the cTnI_SP ‘find’ the cTnC_HP and potentially elucidate a mechanism. A recent structure proposed by cryo-EM studies from Yamada et al.²⁰ was selected as the starting model for MD simulations as it contained cTn within its natural environment of the thin filament. A cartoon representation of this model, created using Illustrate,²¹ can be seen in Figure S1 in Supplementary Material. We investigated six different mutations within the cTnI_IP region that have been associated with cardiomyopathy: L141Q, L144P, L144Q, R145G, R145Q, and R145W. The L144Q and R145W mutations have been associated with restricted cardiomyopathy (RCM), while the other four have been associated with hypertrophic cardiomyopathy (HCM).⁵ All mutations have been experimentally shown to increase Ca²⁺ sensitivity,^{22, 23} with the RCM mutations causing a more drastic increase in Ca²⁺ sensitivity than HCM.²⁴ In addition to these cardiomyopathic mutations, we also performed simulations with a phosphorylated cTnI pT143. The introduction of this phosphothreonine had conflicting reported evidence on Ca²⁺ sensitivity effects, but has been shown to disrupt cooperativity of activation and decrease relaxation rate.²⁵

We geometrically investigated cTnI_SP effective concentration by developing a computational methodology to explicitly measure cTnI_SP effective concentration from molecular simulations. This method can effectively estimate the volume sampled by atoms within the cTnI_SP throughout the course of all our simulations. This allowed us to test how alterations to the cTnI_IP region affected the effective concentration. Our results showed virtually no change in effective concentration for any of the cardiomyopathic mutations, but we did observe a significant increase in the effective concentration for the phosphothreonine models.

Methods

Model Preparation

The starting model representing cTn was extracted from PDB entry 6KN7, a cryo-EM structure, representing a Ca²⁺-free, cTnI_IP unbound, and cTnI_SP tethered conformation of the complex. Residues extracted from PDB 6KN7 included: cTnT 199–272 (chain a), cTnI 41–166 (chain b), and cTnC 2–161 (chain c). To generate a Ca²⁺-bound model, Ca²⁺ ions were added to site II and the two sites in the C-terminal region of cTnC using AutoDock Vina.²⁶ The same process was followed to create a Ca²⁺-bound, cTnI_SP bound, and cTnI_SP tethered model using the PDB entry 6KN8. The addition of Ca²⁺ ions via AutoDock Vina to these structures was a necessary step as neither 6KN7 nor 6KN8 contained the positions of Ca²⁺ ions. Standard protonation states for residues contained within the Ca²⁺ binding domains of cTnC were used as it has been shown that these residues are highly resistant to alterations in their protonation state.²⁷ A Ca²⁺-bound model with an unbound and untethered cTnI_SP was created by deleting the cTnI_IP residues 137–148 from the Ca²⁺ bound, cTnI_SP unbound, and cTnI_SP tethered starting model. Cardiomyopathic mutation models were generated using the Ca²⁺ bound, cTnI_SP unbound and cTnI_SP tethered starting model using the Mutagenesis Wizard function available in PyMOL.²⁸ A phosphorylated version of each cTnI_SP tethered model was then created using the same Ca²⁺ bound, cTnI_SP unbound and cTnI_SP tethered starting model using the PyTMs²⁹ plug-in available in PyMOL.

MD Simulations and Enhanced Sampling Methods

For the MD simulations, the models were solvated with explicit TIP3W water molecules and NaCl counterions were added to neutralize the entire system and bring it to a final salt concentration of 150mM. This system was restrained, followed by the energy of the water molecules and protein being minimized over two separate subsequent 10,000 step minimizations, with a step size of 2fs. The restraints were then removed in an initial equilibration of 190,000 steps, with a final equilibration of 10,000 steps occurring production run conditions. All preparation steps and MD simulations were conducted using NAMD 2.13³⁰ with the Charmm36 force field.³¹ The MD simulations were conducted using an NPT ensemble at 310K with Langevin temperature and pressure dampening. All bonds with hydrogen were constrained using the ShakeH algorithm which allowed for a 2 femtosecond timestep, with structures being saved every 2 picoseconds. The simulations were carried out on the Owens Cluster of the Ohio Supercomputer Center³² using 28 processors on 1 node, for 50,000,000 steps, resulting in 100 nanosecond simulations for each system. After the simulation was completed, the system was stripped of all water molecules and Na⁺/Cl⁻ ions. Then, each saved structure was aligned to the first frame to create a new DCD file. This resulting DCD file was then used for all analysis methods.

All preparation steps for Accelerated MD³³ (AMD) mimicked those of the traditional MD preparation steps and were performed using NAMD 2.13 with the Charmm36 force field. Additional parameters for running AMD included applying the boost energy to the dihedral potential of the system. The threshold energy (E) and acceleration factor (α) were applied after they were determined to be 5437 kcal/mol and 408.8 kcal/mol, respectively, using data from an equilibrated traditional MD simulation of the cTn complex. The resulting DCD files were stripped of waters and aligned as described above. Brownian Dynamics (BD) simulations were performed using BrownDye,³⁴ with electrostatic grid input being generated using APBS.³⁵ Normally, BrownDye is employed to observe the association between two entities, whether it be enzyme-substrate or protein-peptide.³⁴ BrownDye allows the user to input criteria that would tell the program that a ‘reaction’ has occurred if the substrate/peptide were to occupy a given space. However, we wanted to employ Brownian Dynamics for enhanced sampling of the conformational space available to the switch peptide, and thus we did not want a reaction to occur. Hence, we have chosen a ‘phantom’ atom approach to allow the cTnI_SP to ‘escape’ during every trajectory. This allowed the cTnI_SP to essentially sample as much volume as possible throughout the simulation time. This approach was carried out by creating two ‘phantom’ OD1 atoms within the cTnC_HP and designating that for a reaction to occur, the CA atoms of residues cTnI 149 and 164 must be within 1.0Å of these phantom atoms. Since this was impossible to occur during the simulation, the cTnI_SP continued to sample space for the desired steps in the trajectory unless the cTnI_SP ‘escaped’ the cTn complex. 5000 trajectories were created with 1,000,000 maximum steps per trajectory, and the coordinates of the cTnI_SP were saved every 10 steps.

Effective Concentration Measurements using VECA

For each of the MD and AMD simulations, 50,000 frames were created over the course of 100 nanoseconds (1 frame every 2ps). For analysis purposes, we extracted every 4^th frame from the trajectories, totaling 12,500 PDB files, or one frame every 8ps. For the BD simulations, every 25^th frame from the trajectories was extracted for analysis. The effective concentration of the cTnI_SP was directly measured for all these simulations using a method we refer to as ‘Volume Estimation using Cube Approximation’, or VECA. It was implemented as an in-house python script where a sphere with a given radius is generated using 1Å × 1Å × 1Å cubes. The center of this sphere is the average position of the CA atom of cTnI 136, the last residue before the cTnI_IP region that serves as an anchor for the tether. For all atoms in the cTnI_SP region (residues 149–164) in every extracted frame of the simulation being measured, the cubes that these atoms occupy are marked within the sphere. To account for the size of the atoms, we marked six additional points within the sphere in the positive and negative x, y, z directions using the following values for covalent radii: hydrogen (0.31Å), carbon (0.71Å), oxygen (0.66Å), nitrogen (0.71Å), sulfur (1.05Å), phosphorous (1.07Å).³⁶ The creation of these additional points helped better account for atom size and sample cubes that may have otherwise been missed with just marking the center position of the atom. After all frames of a simulation had been analyzed, the total number of uniquely sampled cubes was totaled to yield a sampled volume in ų. This volume was converted into liters and subsequently used to determine an effective concentration assuming 1 molecule of cTnI_SP in the volume sampled. A detailed example of this calculation can be seen in the Results section.

Trajectory Analysis of Simulations

For each extracted frame of the MD simulations, the distance between the cTnI_SP and cTnC_HP was measured by determining the average position of cTnI C-alpha atoms in residues 149–164 and calculating the distance of the average position of those cTnI atoms to cTnC residues 20, 23, 24, 26, 27, 36, 41, 44, 48, 57, 60, 77, 80, 81. These cTnC residues were selected to match those in a previous study of the exposure of the hydrophobic patch.³⁷ The radius of gyration (R_g) for the cTnI_IP was calculated for each frame and then graphed as a function of time over the course of the simulation. The formula used to calculate the radius of gyration was: ${\text{R}}_{\mathrm{g}}=\sqrt{\frac{{\sum }_{i=1}^N{m}_i{r}_i^2}{M}}$, where N = total number of atoms of the cTnI_IP; M = total mass of the cTnI_IP; m_i = mass of atom i; and r_i = distance between center of mass of the cTnI_IP and atom i.

Results and Discussion

Geometric Estimation Determined low Micromolar Minimum Effective cTnI_SP Concentration

We first geometrically estimated the maximum volume that a tethered cTnI_SP could theoretically sample. For infinitely efficient sampling and neglecting any steric hinderance by the remainder of the troponin complex, this would set a lower limit to the effective switch peptide concentration. Based on geometric arguments, we assumed that there will be one molecule of cTnI_SP within a volume defined by a sphere with a radius equal to the maximum possible length of the stretch of amino acids involved in the cTnI_IP and cTnI_SP regions. We further assumed that the cTnI_IP region (12 residues, 137–148) can take on a fully extended conformation with a contour length per residue of 3.5Å,³⁸ and that the cTnI_SP region (16 residues, 149–164) will remain in a helical confirmation with a rise per residue of 1.5 Angstroms³⁹, as observed in PDB 6KN7. Under these assumptions the minimum effective concentration can be calculated as follows:

$$\mathit{\text{Max}}\mathit{\text{Radius}}=\left(12\mathit{\text{residues}}\times 3.5\AA \right)+\left(16\mathit{\text{residues}}\times 1.5\AA \right)=66\AA$$

$$\mathit{\text{Max}}\mathit{\text{Volume}}=\frac{4}{3}\pi {\left(66\AA \right)}^3=1,204,260{\AA }^3\times \frac{m^3}{{10}^{30}{\AA }^3}\times \frac{1000L}{1m^3}=1.20\times 10^{-21}L$$

$$\mathit{\text{Moles}}=\frac{1\mathit{\text{cTnIsp}}}{6.02\times {10}^{23}\frac{\mathit{\text{cTnIsp}}}{\mathit{\text{mol}}}}$$

$$\mathit{\text{Min}}.\mathit{\text{Eff}}.\mathit{\text{Concentration}}=\frac{\mathit{\text{Moles}}}{\mathit{\text{Max}}\mathit{\text{Volume}}}=1.38\mathit{\text{mM}}$$

Therefore, in the limit of infinitely efficient sampling and neglecting steric hinderance by cTn, a tethered cTnI_SP would be able to sample all available space in a sphere with a 66Å radius, leading to an effective concentration of 1.38mM. Our estimated minimum effective concentration for a cTnI_SP is comparable to the 3mM estimated effective concentration proposed by Siddiqui et al.¹⁹ A representation of the maximum volume available to the tethered cTnI_SP, in the context of the cTn complex is shown in Figure 2. This value is an estimation of the total amount of space available to the cTnI_SP and does not account for the volume occupied by the remaining cTn residues in that sphere or for unrealistic backbone torsional angles within the cTnI_IP residues that could potentially prevent the cTnI_SP from sampling certain parts of space centered around anchor residue cTnI 136. When creating a sphere of 66Å radius using the VECA method, the representative sphere was comprised of 1,204,466 empty cubes, a 206ų (0.02%) difference from the theoretical volume calculated above. Likewise, the volume of the cTn complex without the cTnI_IP or cTnI_SP was determined to be 16,862ų. By removing the unsampleable space occupied by the cTn complex, the maximum volume available to the cTnI_SP was then 1,187,604ų, thus leading to a minimum effective concentration of 1.40mM.

Figure 2. Representation of VECA Method.

A) ‘Side’ view (same angle as Figure 1) and B) ‘Top-down’ view of the cTn complex within a sphere of 66Å created by the VECA method. The black dot in the middle of the sphere is the position of the CA atom of TnI 136.

Brownian Dynamics Simulation Verified VECA Method and Determined Minimum Effective Concentration of cTnI_SP

We performed a BD simulation using a Ca²⁺ bound, cTnI_SP unbound and untethered model to test our VECA method for accuracy and to determine a minimum effective concentration of an untethered cTnI_SP. If enough trajectories are simulated, theoretically the cTnI_SP should be able to sample every cube within the 66Å radius sphere that does not contain the cTn complex. The results from the BD simulation determined that the cTnI_SP sampled a volume of 1,187,604ų within the sphere, which is 100% of the sampleable space available to the cTnI_SP, resulting in an estimated effective concentration of 1.40mM. This value would therefore represent the lowest effective concentration for the cTnI_SP given the volume of the cTn complex is unsampleable and restricting the cTnI_SP within the 66Å radius. This evidence shows that our VECA method is efficient at representing the 66Å sphere and it can accurately measure sampled volumes during a simulation. If the radius restriction is removed from the BD volume estimation, the cTnI_SP is essentially unrestrained within the simulation. Troponin molecules are spaced out approximately 386.4Å from each other along the thin filaments.²⁰ When the VECA calculation was performed in a relatively large cube with 300Å sides to represent this physiological spacing (with a maximum volume of 27,000,000ų) we observed a total volume of 12,017,637ų sampled during the BD simulations. This correlated to a 10-fold increase in the volume sampled compared to when the evaluation was limited to a sphere of radius 66Å. This allowed us to estimate the minimal effective concentration of the untethered cTnI_SP to be 0.14mM. Hence, we estimate that the cTnI_IP tether increases the cTnI_SP effective concentration by at least an order of magnitude.

MD Simulations Showed Significant Reduction of Effective Concentration When cTnI_IP Tether is Removed

After using BD and geometric considerations to establish that the VECA method could accurately estimate the cTnI_SP effective concentration, we measured the effective concentration of tethered and untethered cTnI_SP models (both Ca²⁺ bound) during 100 nanosecond MD simulations. Using the restriction of a 66Å radius from TnI 136, the average effective concentration for the tethered models was determined to be 29.7mM ± 5.5mM, whereas the effective concentration for the untethered models was determined to be 13.4mM ± 8.0mM. Standard deviations were calculated based on 25 trials for each system. This data corresponds to volumes sampled by the cTnI_SP of 55,930ų and 123,964ų, respectively. It is physically impossible for the tethered cTnI_SP simulation to have the cTnI_SP extend beyond the 66Å radius, however, it did occur within the simulations with an untethered cTnI_SP. When the 66Å restriction was removed from the VECA method, we determined the untethered cTnI_SP to experience an effective concentration of 9.8mM ± 9.0mM, which correlated to an average sampled volume of 169,503ų. The large difference between observed tethered and untethered effective concentrations within only 100ns, a short time compared to that of contraction, quantifies the importance of the cTnI_IP tether for the function of the cTn complex. Figure 3 illustrates the difference between the space sampled by each of these models within 100 ns. These figures were generated using the most representative trials, which exhibited the closest measured effective concentration to the model average. Figure 3 shows that the untethered switch peptide was free to sample space all around the cTn complex, whereas the tethered switch peptide stayed within the general vicinity of the N-terminal region of cTnC.

Figure 3. Representations of the space sampled using the VECA method.

A) Space sampled by a tethered switch peptide with a volume of 55,146ų. B) Space sampled by an untethered switch peptide with an estimated volume of 141,528ų. All figures were made using the most representative trial for each model, determined by having the closest measured effective concentration to the model’s average.

AMD Simulations Further Quantitatively Demonstrated Importance of Tethered cTnI_SP

To explore the role of limited sampling in our estimations of effective cTnI_SP concentration from MD simulations and to probe the minimum effective concentration of a tethered cTnI_SP, we performed Accelerated Molecular Dynamics (AMD) on the Ca²⁺ bound, cTnI_SP unbound, tethered model of cTn. These simulations were run in triplicate for 100ns. Using this method, we calculated the effective cTnI_SP concentration to be 21.4mM +/− 1.9mM, with an average volume of 77,622ų. Figure 4 shows volume sampled averaged across the 3 trials over the course of the simulation. The cumulative volume sampled seemed to be plateauing by the end of the 100ns simulations, indicating that 21.4mM may be a near-convergent estimation of the actual effective concentration for a tethered switch peptide. This quantitatively demonstrated that the existence of the tether precludes sampling of large parts of the volume that would be accessible if the cTnI_SP was not being tethered to the cTn complex via the cTnI_IP.

Figure 4.

Volume sampled by AMD trajectories averaged across the 3 trials over the course of the simulation. Standard deviation is shown as the shaded region around the average.

cTnI_IP Cardiomyopathic Mutations Showed No Significant Change to Estimated Effective Concentration

With the importance of a tether being established, the effects of altering the natural state of this tether were tested. First, the effective concentration of a tethered cTnI_SP during 100ns MD simulations was measured for six cardiomyopathic mutations in the cTnI_IP region: R141Q, L144P, L144Q, R145G, R145Q, and R145W. The calculated effective concentrations are listed in Table 1, with the standard deviations calculated based on 3 trials for each model. We did not observe any significant difference in effective concentration for the cardiomyopathic mutations. To analyze differences in cTnI_SP-cTnC interactions, we calculated the distance between the cTnI_SP and cTnC_HP, as described in the methods section, as a function of time over the simulation (Figure 5). Both panels of Figure 5 contain data for the trials that simulated a model of Ca²⁺ bound, cTnI_SP tethered, and cTnI_SP bound cTn complex made from PDB 6KN8. These simulations were performed to establish a baseline minimum distance of these two regions for a cTnI_SP that was bound to the cTnC_HP. This helped put into perspective the relative distances calculated for the cTnI_SP unbound models. We observed that although there was no significant difference in the amount of volume sampled over time, three HCM mutations (R141Q, L144P, and R145G) preferentially sampled space distant from the N-terminal region of cTnC compared to the native 6KN7 model. The other three mutations (R145Q, L144Q, and R145W) did not impact the distance of the cTnI_SP to the N-terminal region of cTnC compared to the native.

Table 1.

Effective Concentration for native and mutated models

cTn Model	Effective Concentration (mM)
Native	29.7 ± 5.5
R141Q	26.0 ± 5.1
L144P	29.7 ± 4.4
L144Q	28.3 ± 4.3
R145G	22.8 ± 5.3
R145Q	30.0 ± 5.6
R145W	30.2 ± 4.3

Figure 5.

Distance between the cTnI_SP and cTnC_HP over the course of the traditional MD simulations of Ca²⁺ bound, cTnI_SP tethered, cTnI_SP unbound. A) Data for four HCM mutations and B) data for two RCM mutations. Both graphs contain the 6KN7 and 6KN8 native data.

Interhelical Angle Analysis Revealed No Significant Opening Events During MD Simulations

The interhelical angle between helices A and B has been shown to be a good indicator of when the hydrophobic patch is exposed or ‘open’.⁴⁰ Using a cutoff of 105 degrees, whereas above this cutoff would indicate a closed conformation and below would indicate an open conformation, during the Ca²⁺ bound, cTnI_SP unbound, cTnI_SP tethered MD simulations, we were unable to see any significant opening events. Given the relatively short timescale of these simulations compared to the estimated timescale of hydrophobic patch opening,^41–44 this was not unexpected, but it did inhibit our ability to see a complete shift from an unbound cTnI_SP, closed patch model to a cTnI_SP bound, open patch model. The apparent asymptote in the cTnI_SP-cTnC_HP distance graphs (Figure 5A) around 14Å arose from the inability of the cTnI_SP in the unbound models to bind to the cTnC_HP since it never opened throughout the MD simulations. Without this opening event, 14Å was the closest possible distance of the cTnI_SP to the cTnC_HP. A model of the cTnI_SP bound to a closed cTnC_HP can be seen in Figure S2 of Supplementary Information. During the time in MD simulations where the distance is shown to be in this asymptotic range (for R145Q), the cTnI_SP was essentially just interacting with the N-terminal region of cTnC, potentially primed to bind to the cTnC_HP if it were to open.

Analysis of Distance Between cTnI 145 and cTnC E56/E59/E63

To discern differences in interactions between the cTnI_SP and the N-terminal domain of cTnC, we evaluated the distance between residue cTnI 145 and cTnC residues E56, E59, and E63. A previous study observed a decrease in interaction between these residue pairs when comparing native and R145W models during molecular dynamics simulations.¹⁵ The results can be seen in Table 2. The only model to experience a notable increase in residue-residue distance across all three pairings was the R145G mutation. We observed slight decreases in residue-residue distances across all three pairings for the L144P mutation. However, in both cases the ranges of the standard deviations overlapped with that of the native data, preventing us from concluding that these differences were significant enough to investigate further. This data, combined with our effective concentration evaluation, would suggest the mechanism by which the cTnI_IP aids the cTnI_SP in finding the hydrophobic patch involves more than one residue within the cTnI_IP region, and that a singular mutation may not be enough to significantly disrupt this process. The only mutation that caused a notable impact, as measured by these two metrics, was R145G.

Table 2.

Average distance between cTnC 56/59/63 and cTnI 145. All averages and standard deviations are calculated over all trials performed for each system.

cTnC Residue	Metric	Model
		Native	R141Q	L144P	L144Q	R145G	R145Q	R145W
E56	Avg. Distance (Å)	15.7 ± 3.2	15.2 ± 4.5	12.7 ± 4.0	17.2 ± 1.6	21.1 ± 4.9	16.7 ± 1.3	17.7 ± 5.1
E59	Avg. Distance (Å)	19.4 ± 3.7	18.4 ± 4.5	16.8 ± 4.7	21.4 ± 1.6	23.6 ± 5.6	21.3 ± 1.4	21.6 ± 4.8
E63	Avg. Distance (Å)	24.5 ± 3.3	24.1 ± 4.5	22.3 ± 4.5	26.5 ± 1.3	27.0 ± 6.8	25.5 ± 1.5	25.5 ± 3.3

Radius of Gyration Calculation Showed Increased Variability in cTnI_IP in Most Cardiomyopathic Models

We further analyzed the dynamics of the cTnI_IP region by measuring its radius of gyration (R_g) as a function of time throughout the cTnI_SP unbound, cTnI_SP tethered MD simulations (Figure S2). Given that there was no significant change in the estimated effective concentration, we did not expect to observe a higher R_g for the mutated models. However, we did observe an increased variability in R_g in the RCM mutations, indicating that the mutated cTnI_IP region was less stable than the native sequence. A similar observation was made for the HCM mutations, except for the R145Q mutant, which exhibited a native-like trend. The R145G mutation experienced the largest variability in radius of gyration, as well as the largest observed R_g values for any of the models. This was not surprising as this mutation was shown to have sampled the most space throughout the simulations (Table 1). Interestingly, the 6KN8 model that contained a cTnI_SP bound to an open cTnC_HP exhibited the largest R_g values for any of the models. Inspection of the cTnI_IP conformation in Figure 1B reveals that the entire cTnI_IP region is extended with no apparent secondary structure, whereas in the unbound conformation (6KN7, Figure 1A), residues 137–143 have helical character, leading to a decrease in R_g as compared to 6KN8. Elucidation of the timescale of the elongation of the cTnI_IP could be an endeavor of a future project, as it was not observed in any of our MD simulations.

cTnI Phosphothreonine 143 Stabilized the cTnI_IP Region Causing an Increased Effective Concentration

Shifting the focus from mutations to post-translational modifications, we measured the sampled volume, and hence the effective concentration, for the cTnI phosphothreonine 143 (pT143) cTnI_SP tethered, cTnI_SP unbound model. pT143 has conflicting experimental evidence on its mechanism,²⁵ but is generally thought to decouple the activation of cardiac muscle and decrease the relaxation rate. A previous computational study used a phosphomimic mutation, T143E, and noticed decreased contacts between cTnI 145 and cTnC residues 56, 59, and 63, which would suggest a decrease in cooperativity.¹⁵ Using the same analytical techniques as described for the cardiomyopathic mutations, we observed a significant increase in effective concentration, an improved ability to find the cTnC, and similar R_g data when compared to the native simulations. pT143 simulations were determined to have an estimated cTnI_SP effective concentration of 40.8mM ± 4.6mM, where the standard deviation was calculated from 3 trials. This effective concentration correlated to an average sampled volume of 40,714 ų. This was a significant decrease in sampled volume and increase in effective concentration from the 29.7mM average for wildtype cTn. Figure S3 illustrates the data collected for the cTnI_SP distance to the cTnC_HP and R_g calculations. The pT143 modification facilitated the cTnI_SP experience a shorter distance to the cTnC_HP than the native unbound model. Furthermore, the R_g data mostly aligned with that of the native, leading us to conclude that the structure of this region remained largely unperturbed. Figure S5 shows a comparison between the conformations sampled by the native and pT143 simulations. During the pT143 simulation, the cTnI_SP was more often in the position to bind the cTnC_HP and mimicked the 6KN8 Ca²⁺ bound, cTnI_SP bound structure (Figure 1B).

Interestingly we observed zero interactions between cTnI R145 and cTnC E56/E59/E63 throughout any of the 3 trials, an observation that aligned with previous experiments on the phenomenon, albeit with a T143E phosphomimic model rather than an actual phosphothreonine.¹⁵ All this evidence suggested that pT143 does not induce any large structural changes to the cTnI_IP region but does promote binding of the cTnI_sp to the cTnC_HP.

Introducing cTnI Phosphothreonine 143 to cTnI_IP Cardiomyopathic Models Did Not Increase Effective Concentration

To see if phosphorylation had a similar effect on the cardiomyopathic mutation models, we introduced pT143 to each of the mutated models and ran the same analyses previously described. We saw virtually no change in the effective concentration between the mutated phosphorylated and unphosphorylated models except a slight increase in the R141Q model (Figure 6). Previous experimental research has shown that pT143 is vastly reduced or does not occur when the R145W mutation is present,¹⁵ possibly due to the binding motif required for PKC mediated phosphorylation being disrupted by a mutation only two residues downstream of the target residue. This observation, coupled with our data, may suggest that pT143 is not possible for any of the tested cardiomyopathic mutations. Disruption of this phosphorylation process could therefore be a mechanism of disease caused by these mutations.

Figure 6.

Comparison of measured cTnI_SP effective concentrations in the unphosphorylated (blue) and pT143 modified state (purple).

Conclusions

Through our MD simulations, we were able to quantify the effect of a tethered switch peptide in the cTn complex by using the in-house VECA method. As hypothesized, removing the tether vastly decreased the measured effective concentration of cTnI_SP by allowing it to sample space that is not within the proximity of the cTnC_HP. This effect was observed in both classical MD and Brownian Dynamics simulations. Through AMD simulations with a tethered cTnI_SP, we observed a plateauing of sampled space towards the end of the simulations, indicating that 21.4mM may be a reliable estimate of the actual effective concentration of a tethered cTnI_SP. Modification of the cTnI_IP via PKC mediated phosphorylation of T143 was shown to significantly increase the estimated effective concentration of cTnI_SP, help the cTnI_SP find the cTnC_HP more effectively, and maintain the relative shape of the cTnI_IP when compared to the native model. All of this data indicates that pT143 may be able to help promote binding of cTnI_SP to the cTnC_HP.

The mechanism by which cardiomyopathic mutations in the cTnI_IP region cause disease remains unclear. Our studies indicated that they likely do not impact the effective switch peptide concentration meaningfully. R145G was the only studied mutation that caused a noticeable decrease in the measured cTnI_SP effective concentration. R145-E56 contacts were unaffected by R141Q, slightly increased in L144 mutations, and significantly dampened by all R145 mutations. R145-E59 contacts were essentially knocked out by all mutations, however, these contacts were only observed around 5% of the time in native simulations. R145Q, L144Q, and R145W all showed an increased ability to help the cTnI_IP ‘find’ the cTnC_HP. R145Q was also observed to have the most native-like R_g throughout the simulations, whereas the other mutations had increased R_g or varying R_g. Our data does not support the hypothesis that the mechanism of disease caused by these mutations is predominantly based on altering the effective concentration of the cTnI_SP to the cTnC_HP.