Structural differences in Apo-, D132E- and Holo-Calmodulin bound to the IQ domain of the cardiac L-type calcium channel

Abstract

Proper regulation of the cardiac L-type calcium channel (Ca_V1.2) involves calcium-dependent inactivation (CDI) of the channel. CDI is mediated by the calcium-sensing protein calmodulin (CaM), which interacts with the IQ domain of Ca_V1.2. CaM mutations have been implicated in dangerous cardiac arrhythmias such as long-QT syndrome (LQTS), but it is not clear how structural changes at the CaM-Ca_V1.2 binding interface can lead to LQTS. In this study, we examine structural differences in holo-, apo- and D132E-CaM bound to the IQ domain of Ca_V1.2 using molecular dynamics (MD) simulations and molecular mechanics Poisson-Boltzmann surface area (MMPBSA) binding free energy calculations. Calcium (Ca²⁺) binding to CaM is shown to stabilise key binding interaction triplets in the C-lobe of holo-CaM-Ca_V1.2. We additionally find that D132E-CaM-Ca_V1.2 is able to bind with higher affinity than holo-CaM-Ca_V1.2 due to subtle structural changes taking place at the CaM-Ca_V1.2 binding interface. The differences in the CaM-Ca_V1.2 binding interface are larger when calcium is not bound, as opposed to the differences induced by the mutation, which suggests that the D132E mutation may lead to LQTS via defective binding of Ca²⁺ to CaM as opposed to structural changes occurring at the CaM-Ca_V1.2 binding interface.

1. Introduction

The opening of the cardiac voltage-gated L-type calcium channel (Ca_V1.2) on the outer cell membrane of a cardiac cell produces an inward calcium current that acts as an initiation signal in cardiac excitation-contraction coupling (ECC). ECC is a calcium-dependent signalling cascade, which is responsible for triggering a heartbeat at the physiological level [1]. Defects in the calcium-dependent regulation of ion channels involved in ECC, such as Ca_V1.2 and the Ryanodine Receptor Type 2 (RyR2), have been implicated in dangerous cardiac arrhythmias such as long-QT syndrome (LQTS) and catecholaminergic polymorphic ventricular tachycardia (CPVT) [2–5]. LQTS and CPVT can lead to serious diseases, including sudden cardiac death.

After the initiation of ECC, Ca_V1.2 must close in the presence of a raised intracellular calcium concentration level to terminate the initial signal. This process is referred to as calcium-dependent inactivation (CDI). In CDI, Ca_V1.2 interacts with the calcium-sensing protein calmodulin (CaM), which facilitates the closure of Ca_V1.2 in a calcium-dependent manner [6–12]. The structure of CaM consists of two globular lobes, an N-terminal lobe (N-lobe) and a C-terminal lobe (C-lobe), that are connected by a thin flexible linker (Figure 1(a,c)) [13]. Each lobe consists of two EF hand motifs that are capable of binding to a calcium ion (Ca²⁺), each for a total of four Ca²⁺ ions bound to calmodulin when it is fully saturated (holo-CaM) (Figure 1(b,d)). When the intracellular Ca²⁺ concentration is low, CaM is believed to pre-associate with the Ca_V1.2 in a low-affinity state with its Ca²⁺ binding sites unoccupied (apo-CaM) [14,15]. As the Ca²⁺ concentration rises, Ca²⁺ binds to the four EF hand sites in CaM. This process increases the binding affinity between CaM and a specific binding region in Ca_V1.2, acting as a trigger for CDI. This specific binding region in Ca_V1.2 is referred to as an IQ domain [13,16–18]. It is located within residues 1665–1685 in Ca_V1.2, and its name comes from the identification of a highly conserved, characteristic amino acid residue sequence: IQXXXRXXXXR, where X is a variable amino acid.

Figure 1.

(Colour online) Structural features of CaM bound to the Ca_V1.2 IQ domain peptide. (A) Top-down surface depiction of holo-CaM (residues 2–149) bound to the Ca_V1.2 IQ domain peptide (1660–1685). Here the N-lobe of CaM (residues 2–64) is coloured yellow, the central linker of CaM (residues 65–92) is coloured red, the C-lobe of CaM (residues 93–149) is coloured orange, and the Ca_V1.2 IQ domain peptide is coloured gray. (B) Top-down view depicted as ribbons. Ca²⁺ ions are coloured green. (C) Side-on surface depiction of holo-CaM bound to the Ca_V1.2 IQ domain peptide. The C-terminal end of the Ca_V1.2 IQ domain peptide is pointed out of the page. (D) Side-on view depicted as ribbons. This image is of the initial structure of holo-CaM-Ca_V1.2 just prior to running an MD simulation (PDB ID: 2BE6) [37]. These images were produced using UCSF Chimera X 1.8 [56].

Biochemical and structural studies both indicate that stable binding of the Ca²⁺-bound C-lobe of holo-CaM to the IQ domain of Ca_V1.2 is a critical component of CDI [11,19–23]. The C-lobe of CaM forms a C-shaped pocket which can grip onto the C-terminal end of the Ca_V1.2 IQ domain peptide, and this binding event facilitates channel closing. As the calcium concentration rises further, it is believed that the Ca²⁺-bound N-lobe of holo-CaM binds to an independent domain referred to as an N-Terminal Spatial Ca²⁺ Transforming Element (NSCaTE) domain, which acts to fully close the Ca_V1.2 channel [24–26].

In terms of the molecular underpinning of the disease phenotypes, LQTS has been primarily associated with defects in the interaction between Ca_V1.2 and CaM [27], while CPVT has been primarily associated with defects in the interaction between RyR2 and CaM [2]. Several mutations within CaM itself have been implicated in either LQTS, CPVT or in both LQTS and CPVT [2,28–31]. At present, 20 out of 22 known LQTS mutations within CaM are at residue locations in the C-lobe of CaM that directly bind to Ca²⁺ ions and 15 out of the 22 mutations that cause LQTS are located within EF Hand IV in the C-lobe, involving residues within the sequence 130–142 in CaM (Table 1).

Table 1.

Key residues in the interaction between CaM and the Ca_V1.2 IQ domain peptide, and CaM LQTS mutation locations.

Category	Key residues
CaV1.2 IQ Domain Peptide	E1660, K1665, Q1673, Y1675, R1677, K1678, R1682, K1683
CaM	E12, E15, Q42, E88, E121, E124, M125, E128, M145, A148
CaM (LQTS)	D94, D96, N98, E105, D130, D132, D134, Q136, N138, E141

Key residues involved in the interaction between CaM (residues 2–149) and the Ca_V1.2 IQ domain peptide (residues 1660–1685) are listed in the table above. CaM residues involved in mutations that lead to LQTS are also listed in the bottom row of the table.

In a prior study [32], we identified residues involved with several key interactions at the binding interface between Ca_V1.2 and holo-CaM using molecular dynamics (MD) simulations (Table 1). These key interactions generally act in residue triplets, where two residues on CaM bind on opposite ends of a single residue in the Ca_V1.2 IQ domain (Figure 2). Triplet interactions make optimising binding highly dependent on the overall structural architecture of the complex, which is regulated by Ca²⁺ binding to CaM. The majority of key interaction residues were found in the C-lobe (residues 93–149), with a few residues identified in the N-lobe (residues 1–64) and the central linker (residues 65–92). None of the key interactions at the binding interface between Ca_V1.2 and CaM that were identified bind directly to Ca²⁺ ions, but the majority of key interactions in the C-lobe were clustered together between residues 120–128 and 144–149 in CaM, and these residues flank the EF hand IV Ca²⁺ binding site within CaM (residues 130–142) on opposite sides.

Figure 2.

(Colour online) Key interaction triplets between CaM and the Ca_V1.2 IQ domain peptide. Key interaction triplets involved in the binding of holo-CaM to the Ca_V1.2 IQ domain peptide are highlighted in the image above. These are: E12-R1677 and E15-R1677 (red), Q42-K1665 and E88-K1665 (orange), E124-R1682 and E128-R1682 (blue), M125-Y1675 (yellow) and M145-Y1675 (purple-yellow), and M145-K1678 and A148-K1678 (purple). (A) Side-on surface depiction of holo-CaM bound to the Ca_V1.2 IQ domain peptide. The C-terminal end of the Ca_V1.2 IQ domain peptide is pointed out of the page. (B) The image in (A) after undergoing a 90° rotation clockwise about the vertical axis. (C) The image in (A) after undergoing a 90° rotation counterclockwise about the vertical axis. (D) The same as image as (A) with the surface removed. (E) The same image as (B) with the surface removed. (F) The same image as (C) with the surface removed. All images are of the initial structure of holo-CaM-Ca_V1.2 just prior to running an MD simulation (PDB ID: 2BE6) [37]. These images were produced using UCSF Chimera X 1.8 [56].

Aside from identifying the residues involved in key binding interactions between CaM and Ca_V1.2, at present it is not clear how structural changes in CaM due to mutations can lead to the LQTS diseased state. Several recent structural studies were carried out on mutant forms of CaM bound to the RyR2 [33] and Ca_V1.2 [34,35]. Surprisingly, several disease-causing mutations had structural changes that were very subtle in comparison to the wild-type structure. Recently, Gupta et al. characterised the LQTS-associated, pathogenic D132E mutation within CaM [36]. The D132E mutation (along with E141G) differs from most arrythmia-associated CaM mutations, as it brings about an increase in binding affinity between the Ca²⁺ bound mutant form of CaM and the IQ domain in Ca_V1.2. The authors used circular dichroism (CD) and nuclear magnetic resonance spectroscopy (NMR) to compare the structural differences between wild type holo-CaM and D132E-CaM bound to Ca_V1.2, and the authors found that the structural changes between the two were very subtle, with an overall percent similarity of 85.4% between the two structures [36]. The high similarity between the wild type and mutant structures makes it difficult to identify the structural differences that are involved in the disease-causing mechanism for the D132E mutant form of CaM bound to the IQ domain of Ca_V1.2.

In this study, we examine structural differences in holo-, apo- and D132E-CaM bound to the IQ domain of Ca_V1.2 using molecular dynamics (MD) simulations and molecular mechanics Poisson-Boltzmann surface area (MMPBSA) binding free energy calculations. We first look at the effect of removing the Ca²⁺ ions from the structure of holo-CaM to give apo-CaM bound to the Ca_V1.2 IQ domain. Our simulations show that removing Ca²⁺ from CaM leads to a looser binding of apo-CaM to the Ca_V1.2 IQ domain in comparison to the holo-form as indicated by higher RMSF values, a higher surface area, decreased occupancy, decreased binding affinity for key binding interactions, and a lower binding affinity for the overall complex. Our simulations show that D132E-CaM binds to the Ca_V1.2 IQ domain with comparable RMSF values, surface area, occupancy, key binding interactions and a higher affinity for the overall complex compared to the holo-form. This study shows that MD simulations can identify subtle structural and energetic changes that occur when D132E-CaM binds to the Ca_V1.2 IQ domain. Slightly stronger non-polar binding interactions and an increase in the binding affinity of the E12-R1677 and E15-R1677 triplet may help explain the larger binding affinity observed for the D132E-CaM-Ca_V1.2 complex.

2. Methods

2.1. Preparation of CaM-Ca_V1.2 model systems for MD simulations

All of the CaM-Ca_V1.2 model systems used in this study are based on the crystal structure of holo-CaM bound to the Ca_V1.2 peptide provided by Van Petegem et al. (PDB ID: 2BE6) [37]. The crystal structure file was initially edited to contain only CaM, the four bound Ca²⁺ ions and residues 1660–1685 from the IQ domain in the Ca_V1.2 peptide, which has the following amino acid sequence: EVTVGKFYATFLIQEYFRKFKKRKEQ. CaM residues were numbered 2–149 using the HGVS nomenclature [2]. This model system is denoted as holo-CaM-Ca_V1.2. To produce apo-CaM-Ca_V1.2, we removed the four bound calcium ions from the file. The software program Modeller (version 10.4) [38,39] was used to introduce the D132E mutation into the holo-CaM-Ca_V1.2 system to produce the mutant D132E-CaM-Ca_V1.2 system. The nomenclature used for all three model systems is summarised in Table 2.

Table 2.

Model system nomenclature.

Model system designation	Description
holo-CaM-CaV1.2	CaV1.2 IQ domain peptide bound to wild-type CaM with all four EF hands bound to Ca2+ ions in CaM.
apo-CaM-CaV1.2	CaV1.2 IQ domain peptide bound to wild-type CaM with all Ca2+ ions removed from all four EF hands in CaM.
D132E-CaM-CaV1.2	CaV1.2 IQ domain peptide bound to D132E CaM with all four EF hands bound to Ca2+ ions in CaM.

The nomenclature for the three model systems investigated in this study. All models are initially based on the same crystal structure of holo-CaM bound to the Ca_V1.2 IQ domain peptide (PDB ID: 2BE6) [37]. In each system, the IQ domain peptide was truncated down to a 26 amino acid sequence: EVTVGKFYATFLIQEYFRKFKKRKEQ. The Ca²⁺ ions were removed for the apo-CaM-Ca_V1.2 system, while Modeller was used to introduce the D132E mutation into holo-CaM for the D132E-CaM-Ca_V1.2 system.

2.2. MD simulation protocol

Our MD simulations were carried out using the Amber 22 software suite [40,41]. Each of the model systems described above was placed into an isotropic OPC water box with an initial size of 89.947 × 89.947 × 89.947 ų. 21,945 water molecules were initially added to the box using tleap. 3D-RISM was used to determine the optimal number of solvent ions to neutralise our system [42]. To obtain a salt concentration of 150 mM NaCl, 123 water molecules were removed at random and replaced with 67 Na⁺ ions and 56 Cl⁻ ions for the holo-CaM-Ca_V1.2 system. For all other systems, if the net charge of the system differed from the holo-CaM-Ca_V1.2 system, the number of chlorine ions was adjusted to achieve charge neutrality. To parameterise the ions, we used the Li Merz 12-6-4 ion parameters and the C4 correction term was applied using parmed [43,44].

In our prior studies of CaM bound to the RyR2 peptide [45] and CaM bound to the Ca_V1.2 IQ domain peptide [32], we found that using long, 1–2 μs MD trajectories and multiple replicas was necessary to account for the large amount of heterogeneity from replica to replica and to obtain reliable averages that agreed with experimental results [46]. In the present study, we used hydrogen mass repartitioning (HMR) to increase the time step in our MD simulations from 0.002 to 0.004 ps, which allowed for a total simulation run time of approximately 2 μs per replica [47]. To further improve upon our prior studies, we increased the number of replicas in this present study to 10 replicas for each system. In all of our MD simulations, the parameter and topology files were generated using standard residue types in tleap and the Amber 19 force field (ff19SB) was used to run the simulations.

In each MD simulation, a minimisation was carried out using 1,000 steps of steepest descent followed by 1,000 steps of conjugate gradient. The nonbonded cutoff was set to 10.0 Å. The residues in CaM and the Ca_V1.2 IQ domain peptide were restrained to their initial positions using a harmonic potential with a force constant of 2.0 kcal/mol and the SHAKE algorithm was employed. The Ca²⁺ ions were left unrestrained to correct for any suboptimal positioning of the Ca²⁺ ions in the original crystal structure file. The system was heated to 303K in the NVT ensemble using a Langevin thermostat for 120 ps. This was followed by density equilibration in the NPT ensemble for 120 ps using the Monte Carlo barostat with isotropic pressure scaling and a pressure relaxation time of 1.0 ps. After density equilibration, the restraints to CaM and the Ca_V1.2 IQ domain peptide were removed, the pressure relaxation time was increased to 2.0 ps and all other simulation settings were kept the same. The simulation was then extended to approximately 2 μs of total run time for each replica system. Further details on our model selection and method validation of this MD protocol are provided in our previous publication [32]. We carried out this entire protocol for simulation temperatures of 303K, 325K and 350K. RMSD plots for each system are provided in Figures S1–S9 in the Supporting Information.

2.3. RMSF, surface area and percent occupancy calculations

After completing our MD simulations, we used 200 frames that were sampled at 0.5ns intervals over the last 100ns of a 2 μs MD trajectory to calculate RMSF values, the surface area and the percent occupancies of key interactions for each replica system using cpptraj [48]. The results of all 10 replica systems were averaged together to produce a single average result for each quantity. The surface area calculation was carried out using the surf command in cpptraj. This calculation utilises the linear combination of pairwise overlaps (LCPO) method developed by Weiser et al. [49] to calculate the overall accessible surface area (ASA) of all solute atoms in a CaM-Ca_V1.2 complex. For the percent occupancy calculation, we used a distance-based cutoff of 3.0Å, which was chosen to be close to the mean cutoff of hydrogen bond donor–acceptor distances in common protein secondary structures [50]. In our prior study, we also defined high percent occupancy calculations as those interactions that fell within this cutoff distance more than 20% of the time [32]. More information on the development and justification of using high percent occupancy calculations for finding key interactions in systems where CaM binds to a target peptide can be found in our previous studies [32,45].

2.4. MMPBSA binding free energy calculation protocol

MMPBSA calculations were carried out to assess changes in the binding free energy of CaM to the Ca_V1.2 IQ domain peptide in our model systems [51]. For each system, we calculated MMPBSA values using 200 frames that were sampled at 0.5ns intervals over the last 100ns of the 2 μs MD trajectory. The salt concentration was set to 150mM and we did not include the optional entropy calculation due to high uncertainties associated with the normal-mode analysis. This protocol for MMPBSA calculations was established in the method validation in our prior study, as it led to a high correlation (R² = 0.99) with experimental binding affinities and to standard errors for each replica that were lower than 2 kcal/mol [32]. We averaged the final binding free energy results for all 10 replicas together to obtain an overall average binding free energy that we report for the system as a whole.

In addition to reporting the binding free energy for the overall interaction between CaM and the Ca_V1.2 IQ domain peptide, we also calculated the pairwise MMPBSA binding free energies for key interaction residues in CaM-Ca_V1.2 that we identified in our prior study [32]. Due to the high computational cost, we had to sample a frame every 1 ns for this calculation instead of every 0.5 ns, so we used a total of 100 frames for each replica system in the pairwise calculation. The pairwise MMPBSA binding free energy results for each replica were averaged across all 10 replicas to obtain an average binding free energy for each key interaction for each system. All calculations were carried out for all three model systems at simulation temperatures of 303K, 325K and 350K.

3. Results

3.1. MMPBSA results have a high correlation with experimental binding affinities

To validate the computational method used in this study, in Figure 3 we compare our calculated MMPBSA binding free energies with the experimental binding affinity/free energies reported by Gupta et al. [36] of −10.1 kcal/mol (K_D ~ 40 nM) for holo-CaM-Ca_V1.2, −10.5 for D132E-CaM-Ca_V1.2 (K_D ~ 20 nM) and the binding free energy estimate of ~ −6.6 kcal/mol (K_D ~ 14 μM) for apo-CaM-Ca_V1.2 reported by Kameyama et al. [8] Figure 3(a) shows that at any one of the three simulation temperatures of 303K, 325K and 350K, we observe a reduced binding affinity for apo-CaM-Ca_V1.2 than for holo-CaM-Ca_V1.2, in qualitative agreement with experimental binding affinity data [8,52]. We also observe an increased binding affinity for D132E-CaM-Ca_V1.2 at any of the three simulation temperatures compared to holo-CaM-Ca_V1.2, in qualitative agreement with experiment [36]. At 303K (Figure 3(b)), the MMPBSA results show a high correlation with experiment (R² = 0.96). While the results reproduce the experimental binding affinity trends at any temperature, at 303K the MMPBSA results are numerically very similar to each other, with the average MMPBSA value for the three systems differing by ~ 0.5–1.6 kcal/mol from each other, which was less than the standard error range of ~ 2.9–3.3 kcal/mol for these systems. For this reason, we increased the simulation temperature and repeated the analysis. We note that all three systems that we studied were derived from the same initial crystal structure (PDB ID: 2BE6) and as a result the initial atomic coordinates for each MD simulation are essentially the same. A higher system temperature allows systems to escape kinetic traps more reliably and allows for greater conformational flexibility within the 2 μs simulation timeframe. This was done to exaggerate the structural differences between the three systems in our MD simulations, while hopefully not overly disrupting the high correlation with experimental binding affinities obtained at 303K. At a simulation temperature of 325K, the MMPBSA values differed by ~ 2.5–6.1 kcal/mol which was larger than the standard error range of ~ 1.8–3.8 kcal/mol and the correlation with experiment was still high (R² = 0.90). At a simulation temperature of 350K (Figure 3(c)), the MMPBSA values differed by ~ 5.8–14.1 kcal/mol which was outside the standard error range of ~ 2.5–3.9 kcal/mol, and the data maintained a high correlation (R² = 0.90). For this reason, we will be discussing our results using data from all three temperatures in the sections that follow. In particular, the differences in the data for the three systems are usually much easier to see in the 350K data compared to the data obtained at 303K and 325K.

Figure 3.

(Colour online) MMPBSA binding free energy results for CaM bound to the Ca_V1.2 IQ domain peptide. (A) The calculated MMPBSA binding free energies in kcal/mol are given for holo-CaM-Ca_V1.2, apo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K, and 350 K. MMPBSA results for each replica system were calculated using 200 frames sampled every 0.5ns from the last 100ns of a 2 μs MD simulation. The MMPBSA results for 10 replica systems were averaged together to produce the average binding free energy values reported in the figure. The error bars represent +/− the standard error of the mean. (B) Method validation plots comparing the MMPBSA binding free energies calculated at 303 K and (C) 350 K with the experimental binding free energies of −10.1 kcal/mol (K_D ~ 40 nM) for holo-CaM-Ca_V1.2, −10.5 kcal/mol for D132E-CaM-Ca_V1.2 (K_D ~ 20 nM) reported by Gupta et al. [36], and the experimental binding free energy estimate of ~ −6.6 kcal/mol (K_D ~ 14 mM) reported for apo-CaM-Ca_V1.2 by Kameyama et al. [8] The numerical data used to generate this figure are available in the Supporting Information in Table S1.

3.2. Apo-CaM-Ca_V1.2 has higher RMSF values and surface area than holo-CaM-Ca_v1.2

To study the thermal stability of our three model systems at simulation temperatures of 303K, 325K and 350K, we examined the average RMSF values over the last 100ns of the 2 μs MD trajectory. Figure 4 shows the average RMSF plots for residues in CaM (Figure 4(a)) and the Ca_V1.2 IQ domain peptide (Figure 4(b)) for holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2. Figure 4(a) reveals that at 303K and 350K, there is a general increase in the average RMSF values across CaM residues for apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2, with the differences becoming more pronounced at higher simulation temperatures. While the same general increase in RMSF can be seen at 325 K, there are regions within the N-lobe of CaM at 325K that are exceptions to this trend, such as the flexible loop near residue 40 in the N-lobe. In these regions this trend is reversed, and holo-CaM-Ca_V1.2 shows a higher RMSF than apo-CaM-Ca_V1.2.

Figure 4.

(Colour online) RMSF plots of holo- and apo-CaM bound to the Ca_V1.2 IQ domain peptide at different simulation temperatures. (A) RMSF vales for CaM residues in holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K, and 350 K. CaM residues are numbered 2–149 following the HGVS nomenclature. The approximate locations of the calcium binding site residues are indicated along the horizontal axis in green. (B) RMSF values for the Ca_V1.2 IQ domain peptide residues in holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 at temperatures of 303 K, 325 K, and 350 K. RMSF values were calculated for each replica using 200 frames sampled every 0.5 ns over the last 100 ns of a 2 μs MD simulation. The results for individual replicas were averaged over all 10 replicas to produce the RMSF plots shown above. Since we have presented the data in this figure as overlapping trends, the error bars have been omitted from the figure to improve its readability; the full numerical dataset used to generate this figure, including the standard error values, is available in the Supporting Information in Tables S2–S5.

The RMSF values for the Ca_V1.2 IQ domain peptide in Figure 4(b) follow the same general trend that was observed for both apo- and holo-CaM in Figure 4(a). At simulation temperatures of 303K and 350K, we observe a broad increase in RMSF values, particularly at 350K, for the Ca_V1.2 IQ domain peptide in apo-CaM-Ca_V1.2 compared with holo-CaM-Ca_V1.2. At a simulation temperature of 325K, the IQ domain for apo-CaM-Ca_V1.2 has a generally lower RMSF than holo-CaM-Ca_V1.2, but it is still comparable to the result at 303K; the difference is more substantial at 350K.

As the simulation temperature is increased, the expected general trend is for the RMSF values for all residues to increase compared to lower temperature simulations in both holo- and apo-CaM-Ca_V1.2. This is indeed observed for both systems in Figure 4. However, at a simulation temperature of 350 K, the peak RMSF values in Figure 4(a) are noteably larger for apo-CaM-Ca_V1.2 than for holo-CaM-Ca_V1.2 at several specific locations throughout CaM. At 350K, the largest differences in the peak RMSF values between holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 correspond to the EF hand domains in apo-CaM where the Ca²⁺ ions bind in the holo form. These locations are roughly located between residues 20–32 for EF hand 1, residues 57–68 for EF hand II, residues 94–105 for EF hand III, and residues 130–142 for EF hand IV. We note that in the loop regions, containing the peaks around residues 40 and 80, there is significant overlap in average RMSF values between holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2. Structurally, these regions correspond to flexible linker regions that are not directly stabilised by calcium binding or binding to the IQ domain target peptide. Outside of these regions, we observe broad increases in RMSF values across the N- and C-lobes of CaM in apo-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2.

In addition to the RMSF, we also looked at the average surface area for the apo-CaM-Ca_V1.2 complex in comparison to holo-CaM-Ca_V1.2. A larger surface area can be an indication of a more loosely bound complex. In Figure 5, we see a notable increase in the overall surface area in apo-CaM-Ca_V1.2 when compared with the holo-CaM-Ca_V1.2 system at any of the three simulation temperatures, which is consistent with our RMSF results.

Figure 5.

(Colour online) Average surface area of holo-, apo- and D132E-CaM-Ca_V1.2 at three different simulation temperatures. Average surface area values in Ų are provided for holo-CaM-Ca_V1.2, apo-CaM-Ca_V1.2, and D132E-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K, and 350 K. Average surface area values were calculated for each replica using 200 frames sampled every 0.5 ns over the last 100 ns of a 2 μs MD simulation. The results for individual replicas were averaged over all 10 replicas to produce the RMSF plots shown above. The numerical data used to generate this figure is available in the Supporting Information in Table S6.

A similar comparison of the RMSF values for D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 is provided in Figure 6. In Figure 6(a), for CaM at 325K and 350K, there is a broad, but very subtle decrease in RMSF values for D132E-CaM-Ca_V1.2 in comparison to the holo-CaM-Ca_V1.2 system. In Figure 6(b), at 325K and 350K, we see that the RMSF values of the Ca_V1.2 IQ domain peptide follow the same basic trend, with lower RMSF values for D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2. At 303K, the trend is the opposite, with holo-CaM-Ca_V1.2 broadly having lower RMSF values than D132E-CaM-Ca_V1.2 in Figure 6(a,b). Overall, the RMSF values for holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 are much more similar to each other compared with apo-CaM-Ca_V1.2. Consistent with this observation, in Figure 5 we see that the overall surface area in the D132E-CaM-Ca_V1.2 system is also comparable to the holo-CaM-Ca_V1.2 system, suggesting very subtle changes in the overall conformation of D132E-CaM-Ca_V1.2 in comparison to the holo-CaM-Ca_V1.2 system.

Figure 6.

(Colour online) RMSF plots of holo- and D132E-CaM bound to the Ca_V1.2 IQ domain peptide at different simulation temperatures. (A) RMSF vales for CaM residues in holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K and 350 K. CaM residues are numbered 2–149 following the HGVS nomenclature. The approximate locations of the calcium binding site residues are indicated along the horizontal axis in green. (B) RMSF values for the Ca_V1.2 IQ domain peptide residues in holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at temperatures of 303 K, 325 K, and 350 K. RMSF values were calculated for each replica using 200 frames sampled every 0.5 ns over the last 100 ns of a 2 μs MD simulation. The results for individual replicas were averaged over all 10 replicas to produce the RMSF plots shown above. Since we have presented the data in this figure as overlapping trends, the error bars have been omitted from the figure to improve its readability; the full numerical dataset used to generate this figure, including the standard error values, is available in the Supporting Information in Tables S2–S3 and S7–S8.

3.3. M125-Y1675 has a lower percent occupancy in apo-CaM-Ca_V1.2

Next, we looked at how individual key binding interactions are altered in apo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2. Figure 7(a) gives the average percent occupancy for CaM residues in holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 while Figure 7(b) gives the average percent occupancy values for the Ca_V1.2 IQ domain peptide residues in both cases. When comparing the results for apo-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2, the most striking change is for the M125-Y1675 key interaction. In the holo-CaM-Ca_V1.2 system, this key interaction is maintained at a very high percent occupancy with CaM residue 125, having a percent occupancy of 90.3 + 0.8% at 303K and 80.6 + 5.3% at 350K (Figure 7(a)). In apo-CaM-Ca_V1.2, the steady decrease in percent occupancy for this interaction as the simulation temperature increases is readily apparent; for CaM residue 125 the percent occupancy values are 77 + 8.3% at 303K and 43.8 + 12.3% at 350K (Figure 7(a)).

Figure 7.

(Colour online) Average % occupancy at key interaction locations for holo- and apo-CaM bound to the Ca_V1.2 IQ domain peptide at different simulation temperatures. (A) The average % occupancy of key interactions observed for CaM residues in holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K and 350 K. (B) The average % occupancy of key interactions observed for Ca_V1.2 IQ domain residues in holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K and 350 K. CaM residues are numbered 2–149 following the HGVS nomenclature. The % occupancy for a given residue was defined as the percentage of interactions for that residue found within a 3.0 Å cutoff distance of another residue for 200 frames sampled every 0.5ns from the last 100ns of a 2 μs MD simulation. The results for 10 replicas were averaged together to produce (A) and (B). The numerical data used to generate this figure are available in the Supporting Information in Tables S9–S12.

As the simulation temperature increases, additional reductions in precent occupancy for C-lobe CaM residues are observed. These involve the key interaction triplet of E124-R1682 and E128-R1682 and the triplet formed by M145-K1678 and M148-K1678 (Figure 2). In holo-CaM-Ca_V1.2, CaM residue E124 has a percent occupancy of 20.9 + 4.9% at 303K that decreased to 16.2 + 3.4% at 350K, whereas in apo-CaM-Ca_V1.2, CaM residue E124 has a percent occupancy of 9.5 + 2.1% at 303K that decreased slightly to 8.9 + 2.9% at 350K (Figure 7(a)). In holo-CaM-Ca_V1.2, CaM residue E128 has a percent occupancy of 39.5 + 8.0% at 303K that decreased to 27.1 + 4.3% at 350K whereas in apo-CaM-Ca_V1.2, CaM residue E128 has a percent occupancy of 39.6 + 7.7% at 303K that decreased to 20.0 + 5.9% at 350K (Figure 7(a)). A similar pattern is observed for the percent occupancy values for CaM residues M145 and A148 in the M145-K1678 and A148-K1678 triplet (Figure 7(a)). Taken together, these results indicate a greater loss in percent occupancy for key interactions in the C-lobe of apo-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2 as the simulation temperature increased.

In the N-lobe, the trend is the opposite, as two N-lobe key interactions, E15-Q1673 and Q42-K1665, generally maintained a higher percent occupancy for apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 as the simulation temperature increased. In holo-CaM-Ca_V1.2, CaM residue E15 has a percent occupancy of 23.9 + 3.3% at 303K that decreased to 16.2 + 2.8% at 350K whereas in apo-CaM-Ca_V1.2, CaM residue E15 has a percent occupancy of 26.4 + 3.3% at 303K that decreased to 21.6 + 2.6% at 350K (Figure 7(a)). In holo-CaM-Ca_V1.2, CaM residue Q42 has a percent occupancy of 33.8 + 7.0% at 303K that decreased to 13.9 + 2.6% at 350K, whereas in apo-CaM-Ca_V1.2, CaM residue Q42 has a percent occupancy of 33.3 + 7.7% at 303K that decreased to 17.2 + 3.2% at 350K (Figure 7(a)). The lower decrease in N-lobe CaM percent occupancy values helps offset the loss in percent occupancy for the central linker E88-K1665 interaction. In holo-CaM-Ca_V1.2, CaM residue E88 has a percent occupancy of 39.4 + 3.0% at 303K that decreased to 22.7 + 2.8% at 350K, whereas in apo-CaM-Ca_V1.2, CaM residue E88 has a percent occupancy of 38.8 + 2.7% at 303K that decreased to 16.0 + 4.2% at 350K (Figure 7(a)). Overall, the decrease in C-lobe percent occupancy values and the gain in N-lobe percent occupancy values indicates a subtle shift in percent occupancy away from the C-lobe and towards the N-lobe in apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2.

3.4. E12-R1677 has a larger percent occupancy in D132E-CaM-Ca_V1.2 at 350K

The average percent occupancy values for key interactions in D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 are given in Figure 8. In comparison to holo-CaM-Ca_V1.2, the percent occupancy for the M125-Y1675 interaction is slightly increased by raising the simulation temperature. In the holo-CaM-Ca_V1.2 system, CaM residue 125 had a percent occupancy of 90.3 + 0.8% at 303K and 80.6 + 5.3% at 350K whereas in D132E-CaM-Ca_V1.2, CaM residue 125 had percent occupancy values of 91.0 + 0.6% at 303K and 84.5 + 1.4% at 350K (Figure 8(a)). Similarly, M145-K1678 maintained higher percent occupancy values in the D132E-CaM-Ca_V1.2 system. In the holo-CaM-Ca_V1.2 system, CaM residue 145 had a percent occupancy of 39.3 + 5.1% at 303K and 34.5 + 3.9% at 350K whereas in D132E-CaM-Ca_V1.2, CaM residue 145 had percent occupancy values of 48.1 + 2.3% at 303K and 37.8 + 3.1% at 350K (Figure 8(a)). Another notable change in D132E-CaM-Ca_V1.2 is an increase in percent occupancy for the E12-R1677 and E15-R1677 triplet between the N-lobe and the C-terminal end of the Ca_V1.2 IQ domain peptide (Figure 2). In holo-CaM-Ca_V1.2, CaM residue E12 had a percent occupancy of 8.0 + 3.5% at 303K that increased to 27.2 + 4.6% at 350K whereas in D132E-CaM-Ca_V1.2, CaM residue E12 has a percent occupancy of 14.2 + 4.8% at 303K that increased to 35.2 + 5.4% at 350K (Figure 8(a)). Overall, when comparing D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2, the percent occupancy values for certain key interactions, such as M125-Y1675 and E12-R1677, are larger in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2.

Figure 8.

(Colour online) Average % occupancy at key interaction locations for holo- and D132E-CaM bound to the Ca_V1.2 IQ domain peptide at different simulation temperatures. (A) The average % occupancy of key interactions observed for CaM residues in holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K, and 350 K. (B) The average % occupancy of key interactions observed for Ca_V1.2 IQ domain residues in holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at simulation temperatures of 303 K, 325 K and 350 K. CaM residues are numbered 2–149 following the HGVS nomenclature. The % occupancy for a given residue was defined as the percentage of interactions for that residue found within a 3.0 Å cutoff distance of another residue for 200 frames sampled every 0.5ns from the last 100ns of a 2 μs MD simulation. The results for 10 replicas were averaged together to produce (A) and (B). The numerical data used to generate this figure are available in the Supporting Information in Tables S9–S10 and S13–S14.

3.5. Shifts in pairwise binding free energies differentiate apo-, D132E- and holo-CaM-Ca_V1.2

Aside from the average percent occupancy calculations for key interaction residue pairs, we also carried out MMPBSA calculations for each pairwise key interaction to examine differences in the average binding free energy of these interactions. At 303K (Figure 9(a)), interaction triplets such as the Q42-K1665 and E88-K1665 triplet located near the N-lobe of CaM and the N-terminal end of the Ca_V1.2 IQ domain peptide, and the E12-R1677 and E15-R1677 triplet between the N-lobe of CaM and the C-terminal end of the Ca_V1.2 IQ domain target peptide appear to be strengthened in apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2. At 303K, the average binding free energy for E88-K1665 is −9.28 + 0.68 kcal/mol for holo-CaM-Ca_V1.2 while in apo-CaM-Ca_V1.2 it is −10.35 + 0.68 kcal/mol, and in holo-CaM-Ca_V1.2 the average binding free energy for E12-R1677 is −6.36 + 0.68 kcal/mol while in apo-CaM-Ca_V1.2 it is −7.65 + 1.03 kcal/mol.

Figure 9.

(Colour online) Pairwise average MMPBSA binding free energies for key interactions in holo-, apo-, and D132E-CaM bound to the Ca_V1.2 IQ domain peptide. The pairwise average MMPBSA binding free energy in kcal/mol is shown for each key interaction identified in holo-CaM-Ca_V1.2, apo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at simulation temperatures of (A) 303 K, (B) 325 K, and (C) 350 K. The MMPBSA results for each replica were calculated using 100 frames sampled every 1ns from the last 100ns of a 2 μs MD simulation. The results for 10 replicas were averaged together to produce the average MMPBSA values over all 10 replicas shown above. Since we have presented the data in this figure as overlapping trends, the error bars have been omitted from the figure to improve its readability; the full numerical dataset used to generate this figure, including the standard error values, is available in the Supporting Information in Tables S15–S17.

At 303K, the E124-R1682 and E128-R1682 triplet at the far end of the C-lobe shows a net loss in interaction energy in apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2. In holo-CaM-Ca_V1.2, the average binding free energy for E124-R1682 is −7.03 + 1.52 kcal/mol while in apo-CaM-Ca_V1.2 it is −4.42 + 0.96 kcal/mol, and the average binding free energy for E128-R1682 is −6.49 + 1.23 kcal/mol for holo-CaM-Ca_V1.2 and −7.56 + 1.42 kcal/mol for apo-CaM-Ca_V1.2. At 303K, the average binding free energies for the D132E-CaM-Ca_V1.2 key interactions are similar to holo-CaM-Ca_V1.2 except possibly for the E124-R1682 and E128-R1682 triplet (Figure 9(a)). In holo-CaM-Ca_V1.2, the average binding free energy for E124-R1682 is −7.03 + 1.52 kcal/mol while in D132E-CaM-Ca_V1.2 it is −5.69 + 1.36 kcal/mol, and the average binding free energy for E128-R1682 is −6.49 + 1.23 kcal/mol for holo-CaM-Ca_V1.2 and −4.91 + 1.09 kcal/mol for D132E-CaM-Ca_V1.2. Overall, at 303K, apo-CaM-CaV1.2 appears to have slightly increased interaction energy at key interaction sites compared to holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2, which appear similar to one another.

At 325 K (Figure 9(b)), there is a more noticeable loss in interaction energy observed in apo-CaM-Ca_V1.2 for the M125-Y1675 and M145-Y1675 triplet and the E12-R1677 and E15-R1677 triplet in comparison to holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2. The average binding free energy for the M125-Y1675 interaction is −4.88 + 0.02 kcal/mol in holo-CaM-Ca_V1.2 compared to −3.40 + 0.63 kcal/mol in apo-CaM-Ca_V1.2, and the E12-R1677 average binding free energy is −9.60 + 1.13 kcal/mol in holo-CaM-Ca_V1.2 and −6.57 + 0.93 kcal/mol in apo-CaM-Ca_V1.2. On the other hand, the E88-K1665 interaction has an average binding free energy of −7.30 + 0.81 kcal/mol in holo-CaM-Ca_V1.2 compared to −8.25 + 1.51 kcal/mol in apo-CaM-Ca_V1.2 and −8.90 + 0.73 kcal/mol in D132E-CaM-Ca_V1.2. Overall, at 325K, apo-CaM-Ca_V1.2 appears to have a decreased interaction energy at M125-Y1675 and E12-R1677 compared to holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2, which are similar to each other (Figure 9(b)).

At 350K (Figure 9(c)), there are several key interactions with additional reductions in interaction energy for apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2. In particular, M125-Y1675 has an average binding free energy of −2.50 + 0.65 kcal/mol for apo-CaM-Ca_V1.2 compared to −4.51 + 0.29 kcal/mol for holo-CaM-Ca_V1.2 and −4.78 + 0.02 for D132E-CaM-Ca_V1.2. We also saw noteworthy changes for the Q42-K1665 and E88-K1665 triplet and for the E124-R1682 and E128-R1682 triplet. The average binding free energy for the E88-K1665 interaction is −5.92 + 0.74 kcal/mol in holo-CaM-Ca_V1.2 compared to −4.64 + 1.15 kcal/mol in apo-CaM-Ca_V1.2 and −4.43 + 0.77 kcal/mol in D132E-CaM-Ca_V1.2, and the E124-R1682 average binding free energy is −6.23 + 1.28 kcal/mol in holo-CaM-Ca_V1.2, −3.12 + 1.20 kcal/mol in apo-CaM-Ca_V1.2 and −5.47 + 1.37 kcal/mol in D132E-CaM-Ca_V1.2. The E12-K1677 interaction in D132E-CaM-Ca_V1.2 has a noteworthy increase in binding free energy with an average binding free energy of −12.24 + 1.65 kcal/mol compared to −9.47 + 1.22 kcal/mol in holo-CaM-Ca_V1.2 and −8.40 + 1.11 kcal/mol in apo-CaM-Ca_V1.2. Overall, at 350K, apo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 appear to have a decreased interaction energy for the Q42-K1665 and E88-K1665 triplet and for the E124-R1682 and E128-R1682 triplet compared to holo-CaM-Ca_V1.2, while the E12-R1677 and E15-R1677 triplet has an increased interaction energy in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 and apo-CaM-Ca_V1.2 (Figure 9(c)).

Finally, we attempted to take a coarse-grained view of our results by consolidating our foregoing analysis into the short summary provided in Table 3. The MMPBSA and Surface Area results from Figures 2 and 5 are listed in numerical form as was presented in the two figures. For the RMSF data in Figures 4 and 6, the percent occupancy data for the key interactions in Figures 7 and 8, and the pairwise MMPBSA data for the key interactions in Figure 9, we attempted to average the data from all residues together to produce a single number representing each measured quantity for each system. However, the difference in the numerical values between individual residues for each of these three quantities can in general be much larger than the differences we see between the three systems when compared with each other. This is reflected by the size of the standard error values in Table 3, which ended up being too large to confidently discriminate between the three systems for these latter three quantities.

Table 3.

Summary of average values for holo-, apo- and D132E-CaM bound to the Ca_V1.2 IQ domain peptide at various simulation temperatures.

Model system	MMPBSA (kcal/mol)	Surface area (Å2)	Avg. RMSF (Å)	Avg. % Occ.	Avg. pair. MMPBSA (kcal/mol)
holo (303 K)	−29.6 ± 3.1	9827.1 ± 66.1	0.9 ± 0.3	35.6 ± 7.4	−3.4 ± 0.8
apo (303 K)	−28.5 ± 2.9	10292.8 ± 64.0	1.0 ± 0.2	36.5 ± 6.5	−3.5 ± 0.9
D132E (303 K)	−30.1 ± 3.3	9918.2 ± 62.0	1.0 ± 0.2	34.1 ± 7.6	−3.0 ± 0.8
holo (325 K)	−27.9 ± 3.8	9881.1 ± 51.5	1.1 ± 0.3	30.2 ± 7.0	−3.1 ± 0.8
apo (325 K)	−24.3 ± 1.8	10308.6 ± 88.0	1.2 ± 0.3	28.7 ± 4.6	−3.0 ± 0.7
D132E (325 K)	−30.4 ± 2.9	9758.6 ± 53.7	1.0 ± 0.3	30.8 ± 6.6	−3.1 ± 0.9
holo (350 K)	−31.1 ± 3.3	9841.4 ± 42.8	1.3 ± 0.3	29.3 ± 6.2	−3.2 ± 0.9
apo (350 K)	−22.8 ± 3.9	10529.3 ± 96.8	1.5 ± 0.3	22.1 ± 3.1	−2.5 ± 0.6
D132E (350 K)	−36.9 ± 2.5	9885.4 ± 53.2	1.2 ± 0.3	29.9 ± 6.9	−3.1 ± 0.9

The average MMPBSA binding free energy (kcal/mol), the average surface area (Ų), the average RMSF (Å), the average percent occupancy for key interactions, and the average pairwise MMPBSA values for key interactions (kcal/mol) are shown at simulation temperatures of 303 K, 325 K and 350 K. Each value was averaged over 10 replica systems using 200 frames from each replica sampled at an interval of 0.5ns over the last 100 ns of a 2 μs MD trajectory. The range is given as ± the standard error of the mean.

Nevertheless, when the average values of all five systems are considered side by side as depicted in Table 3, we can point out some qualitative trends that seem to emerge from analysing the data. The most obvious trend is that the apo-CaM-Ca_V1.2 systems tend to have smaller MMPBSA values, larger surface area values, larger RMSF values, and smaller percent occupancy and pairwise MMPBSA values, particularly at the higher temperatures. All of these observations point to a looser binding of the apo-CaM-Ca_V1.2 complex in comparison to the other two systems. In contrast, we see that holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 are much more similar to each other in these same categories than to apo-CaM-Ca_V1.2. Also, we see that in general, a higher percent occupancy tended to correlate with a higher average pairwise MMPBSA result for the key interactions, with the exception of the holo-CaM-Ca_V1.2 system at 350K, which is another indication of the relationship between these two measures. However, conclusions extrapolated from Table 3 should be weighed against the high standard errors that emerge when taking this coarse-grained approach to the analysis.

4. Discussion

4.1. Calcium ions stabilise the key binding interactions in holo-CaM-Ca_V1.2

To properly control the timing of a heartbeat, the Ca_V1.2 channel exhibits CDI that is mediated by CaM in a Ca²⁺ dependent manner. CDI depends critically on the binding interactions between Ca²⁺, CaM, and the Ca_V1.2 IQ domain. In this study, our data broadly support the view that Ca²⁺ ions stabilise the bound conformation of holo-CaM-Ca_V1.2 in comparison with apo-CaM-Ca_V1.2. Compared to holo-CaM-Ca_V1.2, the overall binding free energies were reduced in apo-CaM-Ca_V1.2 (Figure 3), the RMSF values were generally higher (Figure 4), the overall surface area increased (Figure 5), and at elevated simulation temperatures, we saw a general drop in the percent occupancy of key binding interactions (Figure 7) and in the pairwise binding interaction energies for many key binding interactions (Figure 9(b,c)). In particular, at a simulation temperature of 350K, several key interaction triplets in the C-lobe of CaM, such as M125-Y1675 and M145-Y1675, M145-K1678 and A148-K1678, and E124-R1682 and E128-R1682, had lower percent occupancies and suffered a loss in pairwise interaction energies for apo-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2 (Figure 9(c)). As the only difference between the two systems is that apo-CaM-Ca_V1.2 lacked bound Ca²⁺ ions in the EF hand binding sites in CaM, the data suggest that these differences are due to the lack of bound Ca²⁺ in the apo-CaM-Ca_V1.2 system.

Further support of the notion that a lack of bound Ca²⁺ to CaM leads to increased structural instability in apo-CaM-Ca_V1.2 can be seen mostly directly in the RMSF plots in Figure 4. At 350K, the greatest differences in the RMSF values between apo-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 are seen in the EF hand regions of CaM, which are indicated along the horizontal axis in green in Figure 4(a). In the flexible linker regions near residues 40 and 80, both apo- and holo-CaM-Ca_V1.2 exhibit a very similar increase in RMSF values. The difference between these flexible regions and the EF hand regions is that Ca²⁺ does not bind directly to these two flexible linker regions to stabilise their conformation, and so RMSF fluctuations in apo-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 indicate these flexible loops were similarly affected by the temperature increase. In comparison, the EF hand regions that are directly stabilised by Ca²⁺ binding in holo-CaM-Ca_V1.2 have noticeably lower RMSF values compared to apo-CaM-Ca_V1.2 (Figure 4(a)).

The decrease in the overall binding affinity for apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 can be further analysed by examining the electrostatic energy (EEL + EPB) and non-polar energy (VDWAALS + ENPOLAR + EDISPER) terms shown in Table 4. In Figure 9(a), at 303K, the pairwise key interaction free energy for apo-CaM-Ca_V1.2 is comparable to holo-CaM-Ca_V1.2, and in Table 4 we see that the overall binding free energies differ by only ~ 1 kcal/mol for apo-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 at 303K. At 325K and 350K (Figure 9(b,c)), we see that the pairwise interaction energies for several key interactions are reduced in energy in apo-CaM-Ca_V1.2, and likewise the electrostatic energy term is more unfavourable in Table 4, leading to a wider ~ 3–9 kcal/mol gap in the overall binding free energy between apo-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2. These observations reinforce the general observation that Ca²⁺ stabilises the binding of CaM to Ca_V1.2 in holo-CaM-Ca_V1.2, and in particular, that Ca²⁺ binding increases the stability of several key interaction triplets in holo-CaM-Ca_V1.2 compared to apo-CaM-Ca_V1.2.

Table 4.

Electrostatic and non-polar contributions to the MMPBSA binding free energy results.

Model system	EEL + EPB	VDWAALS + ENPOLAR + EDISPER	ΔG	St. error
holo (303 K)	38.6	−68.2	−29.6	±3.1
apo (303 K)	40.6	−69.1	−28.5	±2.9
D132E (303 K)	39.3	−69.4	−30.1	±3.3
holo (325 K)	40.7	−68.6	−27.9	±3.8
apo (325 K)	45.6	−70.0	−24.4	±1.8
D132E (325 K)	39.2	−69.5	−30.4	±2.9
holo (350 K)	38.7	−69.8	−31.1	±3.3
apo (350 K)	44.4	−67.2	−22.8	±3.9
D132E (350 K)	33.0	−69.9	−36.9	±2.5

The MMPBSA binding free energies are broken down into electrostatic and non-polar contributions in the table above. The sum of the electrostatic contributions is the electrostatic energy plus the polar solvation energy (EEL + EPB). The sum of the non-polar contributions is the van der Waals energy plus the non-polar solvation energy plus the dispersion solvation free energy (VDWAALS + ENPOLAR + EDISPER). The sum of the electrostatic and non-polar contributions is the overall MMPBSA (ΔG) free energy value given in the far-right column and in Figure 3. MMPBSA values were calculated for each replica using 200 frames sampled every 0.5 ns over the last 100 ns of a 2 μs MD simulation. The results were then averaged over all 10 replicas to produce the energy values given in the table above. All energies in the table are reported in units of kcal/mol. A more detailed breakdown of the MMPBSA binding free energies by replica and interaction type, including the standard error values, is available in the supporting information in Tables S18–S20.

4.2. Subtle structural changes increase the binding affinity in D132E-CaM-Ca_V1.2

We also looked at the effect of the D132E mutation on the binding interaction between CaM and Ca_V1.2 in terms of the overall binding free energy (Figure 3), RMSF plots (Figure 6), surface area (Figure 5), percent occupancy calculations (Figure 8), and pairwise MMPBSA binding free energy calculations for key interface binding interactions (Figure 9). At all temperatures, we were able to confirm a gain in the overall binding free energy in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 (Figure 3). This indicates that the D132E mutation can stabilise the binding interaction for D132E-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2, which is in agreement with experimental observations [36]. Table 4 reveals that subtle shifts in both the electrostatic energy and non-polar energy terms give rise to this difference in binding free energy. It can be seen that D132E-CaM-Ca_V1.2 is stabilised by ~ −1 to −2 kcal/mol in the nonpolar energy term in comparison to the holo-CaM-Ca_V1.2 system at 303K and 325K. A much larger difference in the overall binding free energy between holo-CaM-Ca_V1.2 (−30.1 + 3.3 kcal/mol) and D132E-CaM-Ca_V1.2 (−36.9 + 2.5 kcal/mol) is observed at a simulation temperature of 350K. Table 4 shows that this gain in interaction energy is due to a noticeable shift in the electrostatic energy term, as the nonpolar energy terms for D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 are comparable at 350K. In Figure 9(c), we see that D132E-CaM-Ca_V1.2 is potentially stabilised due to a gain in interaction energy for the E12-R1677 interaction compared to holo-CaM-Ca_V1.2. This stronger E12-R1677 interaction is likely a factor in the decrease in energy of the electrostatic energy term at 350K for D132E-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2.

We note that the differences in RMSF values between D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 in Figure 6 at any temperature are less dramatic than the differences observed between apo-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 in Figure 4. The RMSF results in Figure 6 do show that at higher temperatures of 325K and 350K, D132E-CaM-Ca_V1.2 has slightly lower thermal fluctuations than the holo-CaM-Ca_V1.2 system. At 303K however, the holo-CaM-Ca_V1.2 system has a lower RMSF than D132E-CaM-Ca_V1.2. The difference at 303K could be because our models are all based on the holo-CaM-Ca_V1.2 system initially. The mutation would initially disrupt the conformation of the system, raising the energy temporarily until a new minimum could be found in the free energy landscape. Raising the simulation temperature would then potentially provide the D132E-CaM-Ca_V1.2 with more conformational freedom to find this new minimum within the short time scale of an MD simulation. Indeed, we observed that the RMSF decreases slightly for D132E-CaM-Ca_V1.2 in comparison to holo-CaM-Ca_V1.2 at 325K and 350K (Figure 6), showing a slight gain in thermal stability for the D132E-CaM-Ca_V1.2 system in comparison to the holo-CaM-Ca_V1.2 system.

The percent occupancies of the key binding interactions depicted in Figure 8 are also comparable between holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2 at any temperature. The biggest difference in percent occupancy for D132E-CaM-Ca_V1.2 is seen at 350K, with the E12-R1677 interaction having a noticeably higher precent occupancy in the D132E-CaM-Ca_V1.2 system compared to holo-CaM-Ca_V1.2. Similar results are generally observed in Figure 9 for the pairwise key interaction energy values for holo-CaM-Ca_V1.2 and D132E-CaM-Ca_V1.2. A notable exception appears at 350K (Figure 9(c)), where we observe a reduction in interaction energy in the Q42-K1665 and E88-K1665 triplet and in the E124-R1682 and E128-R1682 triplets in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 at 303K and 350K. At these two locations, D132E-CaM-Ca_V1.2 more closely resembles apo-CaM-Ca_V1.2 than holo-CaM-Ca_V1.2. The increase in stability for the E12-R1677 interaction in D132E-CaM-Ca_V1.2 presumably helps compensate for these losses. Overall, our data indicate that the D132E mutation confers additional stability to the binding of D132E-CaM-Ca_V1.2, and that subtle, but detectable, structural changes are induced by the mutation at the D132E-CaM-Ca_V1.2 binding interface.

4.3. Structural changes in D132E-CaM-Ca_V1.2 that could lead to LQTS

One of the mysteries surrounding the D132E-CaM mutation is how this mutation can cause a severe loss of CDI, which leads to LQTS. This is a mystery because there is a large degree of structural similarity between D132E-CaM-Ca_V1.2 and holo-CaM-Ca_V1.2 observed in the NMR spectra, and there is an increase, as opposed to a decrease, in binding affinity in the D132E-CaM-Ca_V1.2 complex [36]. The LQTS condition may be brought about by an altered conformation of CaM, which may lead to altered binding interactions at the CaM-Ca_V1.2 binding interface, altered binding of Ca²⁺ ions to the EF hands of CaM, altered binding interactions between CaM and other proteins involved in ECC, or some combination of all of the above factors.

One possibility is that LQTS could be caused by altered binding interactions at the D132E-CaM-Ca_V1.2 binding interface. In a prior study, we showed that key binding interactions between CaM and Ca_V1.2 occur in residue triplets where two CaM residues bind to one Ca_V1.2 residue at a specific location in the Ca_V1.2 IQ domain [32] (Figure 2). These triplets provide geometric constraints to the binding interaction between CaM and Ca_V1.2, and are likely key factors in maintaining the binding specificity, cooperativity, and affinity for CaM and the IQ domain sequence. In this study, we show that subtle changes take place, in terms of both the percent occupancy and the pairwise interaction energies, in these triplets when comparing D132E-CaM-Ca_V1.2 to holo-CaM-Ca_V1.2. At 303K, the biggest difference between D132E-CaM-Ca_V1.2 is observed for the E124-R1682 and E128-R1682 triplet. At 303K, there are lower percent occupancies for these residues in Figure 8, and a loss in binding interaction energy is observed in Figure 9(a). At 350K, in Figure 8 and Figure 9(c), we additionally observe a gain in percent occupancy and binding interaction energy in the residue triplet E12-R1677 and E15-R1677 for D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2. It is noteworthy to mention that these two triplets involve the two highly conserved arginine residues in the IQ domain consensus sequence IQXXXRXXXXR. That these are highly conserved residues suggests that these two arginine residues have a role to play in the CaM-Ca_V1.2 binding interaction. Indeed, one possible explanation for an LQTS disease mechanism might involve a strengthening of the E12 and E15 N-lobe residues to the IQ domain. Both E12 and E15 were identified by Liu and Vogel as possible key residues involved in the binding of the N-lobe to residue R60 in the NSCaTE domain [26], and, therefore, these residues in the N-lobe would need to detach from the IQ domain to bind to the NSCaTE domain at high concentrations of calcium in the proposed mechanism to fully close the channel.

Figure 10 depicts the spatial relationship of residues in the aforementioned residue triplets as well as including residue Q1673, from the IQ residue pair in the consensus sequence, the four Ca²⁺ ions, residues involved in LQTS mutations in EF hand IV, and the M125-Y1675 hydrophobic anchor interaction that was most affected in the apo-CaM-Ca_V1.2 system. As can be seen in Figure 10, these structural elements are carefully positioned to orient and anchor CaM to the C-terminal end of the Ca_V1.2 IQ domain peptide. In particular, we note that E15-Q1673, E12-R1677, and E15-R1677 connect the N-lobe of CaM to the C-terminal end of the Ca_V1.2 IQ domain while M125-Y1675, E124-R1682, and E128-R1682 connect the C-lobe of CaM to the C-terminal end of the Ca_V1.2 IQ domain. EF hand IV binds calcium at the hinge in the C-lobe hydrophobic binding pocket, which supports the M125-Y1675 anchor interaction. All of these interaction sites are in close proximity to each other, and this helps explain how defects in the binding geometry in EF Hand IV, potentially leading to an absence of bound Ca²⁺, can affect their occupancy and binding interaction energies.

Figure 10.

(Colour online) Key interactions between CaM and the C-terminal end of the Ca_V1.2 IQ domain peptide. (A) Key binding interactions are shown between the N-lobe of CaM (residues 2–64, yellow), the central linker of CaM (residues 65–92, red), the C-lobe of CaM (residues 93–149, orange), and the Ca_V1.2 IQ domain peptide (1660–1685, gray). EF hand IV within the C-lobe is coloured cyan, acidic residues are coloured red, basic residues are coloured blue, non-polar residues are coloured green, polar and aromatic residues are coloured purple, and Ca²⁺ ions are spherical and coloured green. This image is of the initial structure of holo-CaM-Ca_V1.2 just prior to running MD simulations (PDB ID: 2BE6). [37] The C-terminal end of the Ca_V1.2 IQ domain peptide is pointed out of the image. (B) Top-down view of the key binding interactions between holo-CaM and the C-terminal end of the Ca_V1.2 IQ domain peptide prior to running an MD simulation. These images were produced using UCSF Chimera X 1.8 [56].

Another possible explanation for D132E causing the LQTS state emerges as a result of the low binding affinity of Ca²⁺ to EF hand IV in the D132E-CaM-Ca_V1.2 mutant. Gupta et al. [36] reported that Ca²⁺ binding to D132E-CaM shows a large decrease in binding affinity (K_D ~ 38.9 microM) compared to the binding of Ca²⁺ to wild type holo-CaM (K_D ~ 2.6 microM). Therefore, we can expect that the Ca²⁺ sensing response of D132E-CaM will be impaired compared to holo-CaM at lower concentrations of Ca²⁺. Our analysis shows that several key interactions at the CaM-Ca_V1.2 binding interface are similarly impaired when Ca²⁺ is not bound in the apo-CaM-Ca_V1.2 system. Thus, a more sluggish response to Ca²⁺ concentration owing to weaker binding of Ca²⁺ to EF hand IV in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2 might account for the development of LQTS in the D132E-CaM-Ca_V1.2 system.

Further support for this suggestion can be obtained by analysing the binding data provided by Gupta et al. [36] for the LQTS associated Q136P-CaM mutant. Gupta et al. [36] reported that the binding affinity of Ca²⁺ saturated Q136P-CaM to the Ca_V1.2 IQ domain (K_D ~ 48 nM) and to the NSCaTE domain (K_D ~ 9.7 microM) are both negatively impacted compared to the binding of wild type holo-CaM to the Ca_V1.2 IQ domain (K_D ~ 40 nM) and to the NSCaTE domain (K_D ~ 2.5 microM). In contrast, the binding affinity of Ca²⁺ saturated D132E-CaM to the Ca_V1.2 IQ domain was measured to be higher (K_D ~ 20 nM) than for the holo-CaM-Ca_V1.2 system (K_D ~ 40 nM), and binding to the NSCaTE domain was found to be slightly impaired compared to the holo-CaM-Ca_V1.2 system (K_D ~ 3.3 microM). If a reduction in binding affinity at the CaM-Ca_V1.2 binding interface was the dominant effect from the mutation leading to thedisease, we would predict that the Q136P mutation (K_D ~ 48 nM) would have a more severe effect on CDI than the D132E mutation (K_D ~ 20 nM). However, the authors found that D132E completely abolishes CDI while Q136P only decreases CDI by about 50% compared to the wild type response [36]. On the other hand, if we compare the binding affinities of Ca²⁺ binding to EF hand IV, D132E-CaM shows a much larger decrease in binding affinity (K_D ~ 38.9 μM) for Ca²⁺ binding than Q136P-CaM (K_D ~ 13.2 μM) compared to holo-CaM (K_D ~ 2.6 μM). If Ca²⁺ binding to EF Hand IV is the dominant factor in reducing CDI, we would expect D132E-CaM to be much more severe at reducing CDI than Q136P, which was observed experimentally. The suggestion that impaired Ca²⁺ binding is the dominant effect in LQTS would also help explain why the CaM mutations leading to LQTS have been to residues coordinating directly to calcium ions as opposed to residues directly involved in binding at the Ca_V1.2-CaM binding interface (Table 1) [2].

Further studies are needed to investigate these possible disease mechanisms further. In particular, an attractive avenue for further study would be to examine the difference in binding of Ca²⁺ to the EF hands of CaM directly while CaM is also bound to Ca_V1.2. Since calcium is a row 4 element in the periodic table with a polarisable valence shell, binding free energy calculations for this interaction would require switching the force field from the standard Amber 19 force field (ff19SB) that we use in this study to a polarisable force field such as the Drude force field (Drude-2019) [53]. Recent studies incorporating polarisability in MMPBSA free energy calculations for calcium binding to calmodulin have been making progress towards this end [54–55]. We should also mention that it is also quite possible that some mutations will lead to multiple mechanistic deficiencies in the Ca_V1.2-CaM interaction that involve changes in both calcium binding to CaM and/or changes at the CaM-Ca_V1.2 binding interface as well. We note that D132E-CaM and E141G-CaM are special cases where the binding affinity for the IQ domain increased rather than decreased, but most other LQTS mutations show a decrease in binding affinity for CaM and the IQ domain [36]. As such, the disease specific details for D132E-CaM might not be generalisable to other LQTS mutants. As an extreme example, Wang et al. [34,35] showed that the D130G-CaM mutation caused a large structural unfolding of EF hand IV leading to both a loss of binding for Ca²⁺ and a loss of binding of the C-lobe to the Ca_V1.2 IQ domain. Both of these effects would need to be taken into account to describe the overall impact on binding to Ca_V1.2 for the D130G-CaM mutation.

A major limitation of this study that can be addressed in the future is the large standard error values seen in many of our calculation results. As we were able to increase the number of replicas from 4 in our previous study to 10 in our present study, we have for the first time included a full statistical analysis of every calculated quantity in our present study. In doing so, we note that the large standard errors prevent us from drawing unambiguous conclusions in several cases, chiefly being the reduction of our RMSF, percent occupancy and pairwise MMPBSA data to a single number, as shown in the summary given in Table 3. The size of the standard error decreases with the inverse square root of the number of replica systems that we use, and so, in principle, increasing the number of replica systems should yield smaller standard errors, improving the analysis. This is an avenue we will explore in future studies.

5. Conclusion

In this study, we examined the structures of apo-, D132E- and holo-CaM bound to the IQ domain of the Ca_V1.2 channel. Our MD simulations and MMPBSA binding free energy calculations reproduced the unusual effect of D132E-CaM having an increased binding affinity for Ca_V1.2 compared to holo-CaM bound to Ca_V1.2. We showed that several key interactions in the C-lobe of apo-CaM-Ca_V1.2 were less thermally stable in comparison to holo-CaM-Ca_V1.2. This helps explain why apo-CaM binds to the Ca_V1.2 IQ domain with a much lower binding affinity than holo-CaM-Ca_V1.2. In contrast, the key interactions were largely maintained or even strengthened slightly in D132E-CaM-Ca_V1.2 compared to holo-CaM-Ca_V1.2. This helps explain why D132E-CaM binds to the Ca_V1.2 IQ domain with slightly higher binding affinity than holo-CaM-Ca_V1.2. To explain the LQTS disease state for the D132E-CaM-Ca_V1.2 system, it seems likely that the decreased affinity of Ca²⁺ in D132E-CaM is responsible for the development of LQTS as opposed to any major alterations taking place in the binding interactions at the D132E-CaM-Ca_V1.2 binding interface. Overall, this study illustrates how computational methods can be used to obtain information on subtle structural changes taking place in the binding of variants of CaM to the same or different cellular target(s). This approach can be particularly useful in situations where the mutant and wild type structures appear to be very similar when examined by other methods, such as NMR. In addition to illuminating subtle structural differences, this approach can potentially be used to help design or test alterations in the binding interactions between CaM and a target protein, which may lead to new treatment strategies in the future.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/08927022.2025.2608026.