Introduction

Structural characterizations of voltage gated ion channels often lack data describing the cytosolic components. AlphaFold prediction1 of an intact structure are often insufficient for understanding allosteric communication between cytosolic events and structural features (such as voltage sensors and selectivity filters) that govern ion conduction. Our initial modeling efforts focused on a complete description of the NaV1.5 alpha subunit. For this, we combined the latest AlphaFold model with improvements in the Amber forcefields (ff19SB2, lipid213, OPC water model4). These efforts improved predictions of the cytosolic components (phi/psi backbone torsion angles) as well as enhanced classification of intracellular, transmembrane, and extracellular regions5.

NaV1.5 has three predominant channel configurations. At resting membrane potential (− 80 mV), the majority of channels adopt a closed configuration. With an increase in membrane potential (depolarization), NaVs open and allow the passage of an ion through the lipid bilayer. Subsequently (1–5 ms), NaVs become non-conducting through the fast inactivation process. This mechanism involves the cytosolic linker that connects domains III and IV, which is often referred to as the inactivation gate (IGATE)6. Permeation of a Na+ ion through the entirety of the channel requires voltage sensor translocation, leading to conformational change of the activation gate. This activation gate (AGATE) is comprised of four amino acids that are located at the intracellular end of DI-DIVS67.

Wingo et al.8 posited an EF hand-like motif in the cytosolic C-terminal portion of the NaV1.5 channel. Shortly thereafter, an IQ motif (sequence of amino acids ([FILV]Qxxx[RK]Gxxx[RK]xx[FILVWY]) that often interacts with the calcium sensing protein calmodulin in the absence of calcium) was identified downstream from the CTD. It has been posited (based on fluorescence and NMR data) that these components function as a molecular switch that coordinate an interaction between CaM and the CTD9. Later on, an NMR structure of the CTD was determined and four helices were observed to pack into a globular domain (helix I and helix IV in antiparallel orientation)10. Titration of this construct with an IQ motif peptide yielded Chemical Shift Perturbations (CSPs) consistent with a CTD–IQ motif interaction10. Isothermal titration calorimetry data revealed modest affinities in the absence and presence of calcium (apo = 27 µM, Ca2+ = 38 µM)10. Subsequently, crystallographic investigation of an intact CTD–IQ motif construction captured the IQ motif bound to the CTD of a neighboring molecule11. Shortly after, a CryoEM structure revealed the CTD bound to the inactivation gate (IGATE)12. Solution NMR data then validated this interaction for the human NaV1.5 sequence13. Our recent modeling efforts were consistent with a moderate CTD–IGATE affinity that is predominantly generated through electrostatic interactions5.

To evaluate the feasibility of a CTD–IQ motif interaction occurring within the context a full-length NaV1.5 alpha subunit, we performed MD simulations of the individual components and then evaluated them within a full-length channel. Similar to our previous investigation5, we leverage recent advancements in MD simulations (ff19SB, lipid21, and OPC water model). This afforded an opportunity to further explore the energetic landscape of cytosolic components as well as the potential allosteric communication to transmembrane regions (voltage sensors S4 and pore forming S5 and S6).

Results

Release of CTD–IGATE afforded opportunity for additional interactions

Notably, two out of four previous MD simulations captured dissociation of the C-Terminal Domain (CTD) from the Inactivation Gate (IGATE)5. To us, the release of the CTD–IGATE interaction raised a potential for interaction between the CTD and other cytosolic components. Here, we evaluate the potential for the CTD to engage the cytosolic IQ motif as posited by Chagot et al.10.

CTD–IQ motif interaction in a reduced system

Our initial modeling efforts utilized a CTD (E1784–S1865) construct and an IQ motif peptide (E1901–S1925) to explore complex formation via a naïve docking protocol. This approach explored the entirety of both the CTD and IQ motif surfaces. Plotting the Rosettadock free energies yielded two binding funnels (two energetic minima) (Fig. S1). These minima correspond to two architectures where the IQ motif peptide engaged the same CTD binding surface (helix I and IV) in (i) parallel or (ii) antiparallel orientations (CTD-helix 1 relative to IQ motif helix). Clustering analysis failed to predict a preferential orientation (based on population in pool of ~ 100,000 decoys). A second round of docking that sampled an additional 100,000 decoys provided clarity. For this round of docking each decoy contained a randomly oriented IQ motif in the region of the CTD binding interface. These results supported an antiparallel CTD–IQ motif orientation as the 30 lowest-energy models contained 27 antiparallel (Fig. 1a) and 3 parallel (Fig. 1b) orientations. To interrogate the feasibility of an antiparallel binding orientation in the context of an intact CTD (as found on the NaV channel), we used Chimera Modeller (loop-building routines) to construct the remaining amino acids (in silico). This connected the C-terminus of the CTD construct (helix V) with the N-terminus of the IQ motif (helix VI) (Fig. 1c). Subsequent microsecond MD simulations of this E1784–S1925 construct predicted that this self-complex was stable. Oscillations of 4–5 Å were observed during the last 500 ns of the MD simulation (RMSD calculated with Rosetta/Modeller starting structure). A few short (20–30 ns) excursions up to 7 Å RMSD occurred that quickly returned to 4–5 Å (data not shown).

Fig. 1
figure 1

Molecular dynamic simulations explored the CTD–IQ motif interaction. Predicted models of CTD–IQ motif interaction in (a) antiparallel and (b) parallel binding orientations. (c) MD simulation of a monomeric (CTD linked to IQ) construct predicted feasibility of CTD–IQ motif interaction (antiparallel orientation, defined by Helix I–Helix VI (IQ motif)) across four independent one microsecond timescales.

Full size image

CTD–IQ motif interaction within an intact NaV1.5 alpha subunit

To ascertain the feasibility of the CTD–IQ complex in the context of a full-length ion channel in a lipid bilayer, we incorporated this complex into our model where the CTD had dissociated from the inactivation gate5. We performed four independent one microsecond MD simulations (CTD–IQ motif) as well as an additional four independent one microsecond MD simulations with the CTD unattached to the IGATE or IQ motif (CTD–FREE) (Fig. 2). Consistent with our CTD–IGATE simulations5, backbone torsion angles (phi/psi) of the CTD–FREE and CTD–IQ motif models display values consistent with alpha helical, beta sheet, and random coil secondary structures (Fig. 2a). A representative snapshot of each model (final coordinate file post MD) is shown in Fig. 2b. Visual inspection of the CTD–IQ motif complex provided insight into the intermolecular forces that held this complex together (Fig. 2c–f). Several investigations have explored varying lengths of the IQ motif construct (E1901-S1925 and E1901-D1940)9,10. In our CTD–IQ model, the IQ motif helix broke at G1918 to adopt a coil configuration allowing for bending around helix V. This contrasted with our CTD–IGATE and CTD–FREE models, where the N-terminus of the IQ motif region (1925–1940) retained an elevated alpha helical content (Fig. S2).

Fig. 2
figure 2

NaV1.5 models in a lipid bilayer predicted feasibility of CTD interaction with appropriate phi/psi torsion angles. (a) Ramachandran plots of NaV1.5 models post one microsecond MD simulations. (b) Representative models of each CTD configuration. Enlarged inset of CTD–IQ motif interaction highlighting intermolecular forces at the CTD–IQ motif binding interface. (c) Surface rendering of CTD (helix I–IV) highlighting the hydrophobic pocket that engaged the IQ motif (helix VI). (d) Ribbon diagram of predicted side chain contacts at CTD–IQ motif binding interface. Local side chain contacts between (e) helix I and IQ motif and (f) helix V and IQ motif.

Full size image

Predominantly hydrophobic interactions predicted at CTD–IQ motif binding interface

A hydrophobic interaction between the IQ motif and the CTD is predicted to form where CTD helices I and IV engage the IQ motif (helix VI) in an antiparallel orientation (helix I relative to helix VI) (Fig. 2c–e). At the core of this interaction, F1912 on the IQ motif is sandwiched between F1791 and Y1795 (helix I) (Fig. 2e). The side chain aromatic rings were predicted to adopt a “parallel displaced” conformation14. Y1795 also packed against I1908 from the IQ motif. Additionally, two salt bridges (E1784–R1919 and D1792–R1913) were predicted to contribute to this interaction, yielding a total of five contacts between the IQ motif and helix I (Fig. 2e). Helix V was predicted to provide a more modest contribution to IQ motif binding, with a single hydrophobic contact (I1892–V1907) (Fig. 2f).

Intriguingly, helix V is also predicted to position an aromatic side chain (F1879) near the binding interface; however, the closest interacting partner (F1912) is preoccupied with F1791 and Y1795 (Fig. 2f). Additionally, a salt bridge between R1929 and E1876 stabilized the coil region downstream of the IQ motif. In the CTD–FREE simulation, the distances between the side chains of the aforementioned contacts were significantly elevated (Fig. S3). We note that the one microsecond time scale of our simulations do not afford a comprehensive picture of all possible CTD interactions. Further investigation will be required to understand if the IQ motif can spontaneously bind to the CTD or if other factors are required to facilitate this interaction (such as conformational changes by voltage sensors or accessory proteins).

Our modeling efforts are in close agreement with empirical NMR data of reconstituted fragments (CTD (E1773–S1865) and IQ motif peptide (E1901–S1925)). Chemical Shift Perturbations (CSPs) were consistent with our predicted (MD simulation) InterMolecular Forces (IMFs). Our simulations agreed with the observed CSPs for hydrophobic F1791 and Y179510 (predicted as anchoring residues for the CTD–IQ motif interaction). Moreover, negatively charged side chains with notable CSPs (E1784 and D1792) were predicted to have appropriate distances from positively charged side chains (R1919 and R1913), consistent with salt bridge formation between the IQ motif and CTD. L1786, I1797, F1855, and L1862 are also predicted to pack and form hydrophobic pockets; however, direct interaction between these side chains and the IQ motif residues were not observed. This was consistent with CTD–IQ motif binding allosterically influencing the local environment of these hydrophobic side chains.

Position of the CTD is not predicted to allosterically modify lipid or water contacts

To understand if CTD position influenced the location of the transmembrane helices within the lipid bilayer, we quantified hydrophobic (OL and PA) (Fig. 3a), hydrophilic (PC) (Fig. 3b), and water (Fig. 3c) contacts to NaV1.5 alpha carbons for the CTD–IQ motif and CTD–FREE MD simulations (Fig. 3d–f). Similar to our CTD–IGATE simulations5, the core of each transmembrane helix had a high number of consistent contacts with the OL and PA molecules for the duration of each one microsecond simulation. As anticipated, the edges of each transmembrane helix also had occasional contact with the OL and PA molecules. These data were not significantly influenced by the position of the CTD (CTD–IGATE, CTD–IQ, CTD–FREE) (Fig. 3a,d). Analysis of PC molecules relative to NaV1.5 alpha carbons revealed contacts between the edges of the transmembrane helices as well as some of the cytosolic components. Notably, other cytosolic amino acids (not contained within CTD) were able to engage with PC molecules during the CTD–FREE simulation (Fig. 3b, e). This was consistent with the CTD being translocated away from the PC membrane, leaving these molecules more exposed for interaction with other residues located in the cytosol. We did not detect any gross changes in solvation that could be clearly attributed to changes in CTD interaction (Figs. 3c,f and S4).

Fig. 3
figure 3

Lipid and water contacts. Post MD model of NaV1.5 CTD–IQ motif in a lipid bilayer. Lipid headgroups (PC) shown as transparent pink surface. Hydrophobic tails omitted for visual clarity. Contacts mapped by color (a) hydrophobic (< 7 Å from OL or PA), (b) hydrophilic (< 7 Å from PC), (c) water (< 4 Å from water oxygen). NaV1.5 CTD–FREE contacts (d) hydrophobic (< 7 Å from OL or PA), (e) hydrophilic (< 7 Å from PC), (f) water (< 4 Å from water oxygen). All contacts measured from alpha carbon of each amino acid.

Full size image

CTD interactions predicted to influence dynamics of specific cytosolic amino acid side chains

To gain insight into the influence of CTD position on amino acid side chain dynamics, we calculated RMSF values across the duration of each respective simulation (frames for CTD–IGATE simulations containing dissociation events were removed from data) (Figs. 4 and S5). Differences in average RMSF values (Runs 1 through 4) were mapped onto post-MD models (Fig. 4a). Average RMSF values were also plotted by amino acid number (Fig. 4b). As anticipated, amino acids that were located outside of the transmembrane region (cytosolic and extracellular space) displayed an elevated fluctuation compared to those that were in close proximity to the lipids. Lower levels of fluctuation were observed for P2 and P3 loops with slightly elevated values predicted for amino acids contained in P1 and P4. CTD position was predicted to influence amino acid side chain dynamics of the DI–DII and DII–DIII linkers as well as the C-terminus of the channel (1765–2016) (Fig. 4b). Further work will be required to understand if these changes in sidechain dynamics influence accessory protein interactions or post translational modification.

Fig. 4
figure 4

Influence of CTD interaction(s) on NaV1.5 side chain dynamics. (a) Post MD CTD–IQ motif model shown as ribbon diagram colored by differences in RMSF of amino acid side chains. (b) RMSF values of amino acid side chains plotted by residue for NaV1.5 IGATE (blue), CTD–IQ motif (orange), and CTD–FREE (green) models. Cytosolic components shaded in light blue and loop regions shaped in gray. Amino acid numbers provided for each shaded region based on visual inspection of trajectory and post-MD files.

Full size image

CTD–IQ motif interaction allosterically alters the position of DIII S1-S4 without influence on pore forming helices (S6)

To probe the influence of the CTD interactions on the predicted structural features of the voltage sensors, we performed RMSD analysis of the transmembrane helices S1–S4 from each domain (Fig. 5a). Overall, voltage sensors structural features were fairly similar irrespective of CTD interaction. Close inspection revealed that DIII had some sensitivity to CTD position (Fig. 5b). Specifically, the CTD–IQ motif model was predicted to have a slight translocation for each of the four helices relative to the CTD–IGATE and CTD–FREE. Similar to our previous CTD–IGATE model5, DIII S4 contained a significant bend, where the top portion of the helix adopted a horizontal orientation that approached the hydrophobic (OL, PA), hydrophilic (PC) membrane interface (Fig. 5b). This was consistent with the amphipathic arrangement of side chains, where positive amino acids (ARG and LYS) were oriented along the top half of the curved helix and hydrophobic amino acids (LEU and ILE) resided within the interior half of the curved helix. During the time span of our simulation, these DIII voltage sensor structural alterations did not evoke change in the position of the helices that comprise the channel pore (S5 and S6) (Fig. S6). Further work will be required to understand if these predicted changes in voltage sensor architecture influence S5 and S6 within a physiological time scale (milliseconds) in a manner that alters the position of amino acids within the activation (ALIL) and inactivation gates.

Fig. 5
figure 5

Structural analysis of voltage sensors (S1–S4). (a) Post MD models (aligned by transmembrane helices) displaying individual voltage sensors overlayed with each of the CTD models. RMSD analysis (calculated from average ensemble coordinates of amino acids contained within S1–S4) reveals largest changes occur within S1–S4 of DIII of CTD–IQ motif model. (b) Side view of each voltage sensor (S1–S4) aligned by average ensemble coordinates.

Full size image

Position of CTD does not overtly influence structural features of conducting and non-conducting configurations

CryoEM characterization has provided structural descriptions of the NaV1.5 channel pore in conducting7 and non-conducting15 configurations. Alignment of the CTD–IQ motif and CTD–FREE models to non-conducting (PDB ID: 6UZ3) and conducting (PDB ID: 7FBS) structures revealed that all three NaV1.5 models had similar structural resemblance to the non-conducting (Fig. 6a) and conducting (Fig. 6b) CryoEM structures (irrespective of CTD interaction). As anticipated, both the CTD–IGATE and CTD–IQ motif models had a closer resemblance to the empirical structures. The CTD–FREE configuration had slightly higher RMSD values (less resemblance to empirical data). This was likely due to empirical data necessitating a defined structure, and in the absence of a binding partner, CTD could translocate in several directions, resulting in loss of CryoEM signal16.

Fig. 6
figure 6

Resemblance of cryo-EM structure suggests a non-conducting configuration (IGATE red, CTD blue, IQ motif green). Comparison of NaV1.5 CTD–IGATE, CTD–IQ motif, and CTD–FREE models with CryoEM (a) non-conducting structure (brown, PDBID: 6UZ3) and (b) conducting structure (tan, PDBID: 7FBS).

Full size image

Position of the CTD does not allosterically modify activation gate, inactivation gate, or ion conduction in our predicted models

Given the absence of pore-forming helix translocation (S6), we next queried if CTD position influenced orientation of amino acid side chains that modulate ion conduction. Alignment of the transmembrane helices of our CTD–IQ motif model with the established conducting and non-conducting structures allowed for comparison of the EEDD, DEKA, and ALIL side chain conformations (Fig. S7). Significant deviations from the empirical data were not observed for the EEDD and DEKA side chains. Additionally, the CTD position or interaction did not influence this result (Fig. S8). Similar to our CTD–IGATE model, the CTD–IQ motif and CTD–FREE configurations predicted a rotational shift in S6 of DIV relative to the CryoEM structures. Origins of this change and the consequences to the Activation Gate (AGATE) are discussed below. Further efforts are required to understand the role of the CTD in different channel configurations (consideration of membrane potential).

Notably, the CTD–IQ motif simulation captured a Na+ ion in between the selectivity filter and AGATE. Permeation through the entire channel appeared to be prevented by the activation gate. Figure 7 highlights several potential Na+-selectivity filter (DEKA) interactions (oxygens from D372). When a Na+ ion in the DEKA-AGATE chamber was located > 5 Å from D372, six water molecules were predicted to form a hydration shell around the ion (Fig. 7a). When the Na+ ions were within close proximity of the D372 oxygens (< 5 Å), the number of waters (oxygens < 5 Å) was reduced (Fig. 7b). Interestingly, D372 was also predicted to have the capacity to desolvate two Na+ ions simultaneously (Fig. 7c). Further investigation is required to discern if the permeation of a Na+ ion past the DEKA selectivity filter depends on CTD interaction or timescale of our simulations.

Fig. 7
figure 7

Select frames illustrating potential Na+-selectivity filter interactions. Ribbon diagrams of CTD–IQ motif models highlighting Na+, selectivity filter, and AGATE. Zoomed inset depicting potential Na+-side chain contacts (top). Zoomed insets showing Na+-water contacts (bottom).

Full size image

Discussion

In the absence of membrane potential, empirical channel structures are often described as a non-conducting configuration

Our modeling efforts prompted us to consider factors that dictate open, closed, and inactivated channel configurations. This led us to speculate about the closed/non-conducting configurations that have been observed in CryoEM structures15,17, predicted by AlphaFold1, and our previous modeling efforts5. It is widely accepted that NaV channels adopt a closed configuration at a resting membrane potential (− 80 mV). Our MD simulations require the use of a periodic box (when an atom moves outside of the box, it is immediately replaced by its image on the other side). Because of this, our MD simulations do not contain a Na+ concentration gradient across the lipid bilayer. Without a concentration gradient, the Nernst equation yields a 0 mV membrane potential. At a depolarized potential (0 mV), NaV1.5 conduction is often observed by whole cell patch clamp electrophysiology. Similarly, it is also worth noting that CryoEM structures are often derived from endogenous preparations of NaV channels (fragments of membrane), where the lipid bilayer cannot maintain a continuous barrier. We speculate that this would prevent the retention of a concentration gradient on either side of the lipid fragment when samples are frozen in vitrified ice16. Further work will be required to understand the details surrounding steady state membrane potential calculations and channel configuration using simulations that expand beyond low microsecond timescales.

Amphipathic nature of S4 hints at a curved/horizontal translational mechanism in response to membrane depolarization

It is widely accepted that membrane depolarization causes translocation of segment 4 each transmembrane domain, leading to an open or conducting channel. Several studies suggested that movement of each individual voltage sensor (S4) can impart unique functional changes to ion conduction18,19,20. Specifically, DIIIS4 and DIVS4 modulate the affinity of the IFM particle of the IGATE for its binding site by their coordinated upward and outward translocations21. Within 1–5 ms, NaVs generally cease to conduct through the fast inactivation process22. Consistent with our previous work, DIIIS4 was predicted to curve in a manner that oriented the charged amino acid side chains into the charged PC lipid head group environment with the hydrophobic side chains on the opposite face trending towards the hydrophilic PA and OL lipid tails. In the absence of a membrane potential (unable to establish a concentration gradient due to periodic box), we see evidence of S4 lying horizontally across this hydrophobic-hydrophilic (PC/OL–PA) interface. Lipid modification mechanisms similar to this have been observed for G-Coupled Protein Receptor (GCPR) systems23. We speculate that a similar type of mechanism also applies to NaVs. Further work will be required to validate this posited mechanism in NaV and CaV channels as well as clarify how membrane charge affects voltage sensor localization within a lipid bilayer.

CTD–IQ motif complex with consideration for point mutations

A mutation of an IQ motif residue (R1913H) was reported to increase late sodium current13,24. Analysis of our model revealed that R1913 forms a salt bridge with D1792 from the CTD that stabilizes the CTD–IQ motif complex. It can be inferred that the loss of this salt bridge due to R1913H would weaken the CTD–IQ motif interaction. Further work will be required to understand how the loss of this salt bridge facilitates an enhancement in late sodium current. Another mutation near the amino end of the IQ motif (E1901Q) elicited a similar increase in late sodium current; however, our model did not predict an interaction between this residue and the CTD or other cytosolic parts of the channel. It is worth noting that the core of the hydrophobic CTD–IQ motif interaction involved F1791, Y1795, and F1912. Only one mutation at these residues (Y1795C) has been clinically observed and associated with disease (LQT3)25. Further investigations are exploring how these IMFs influence channel gating.

F1486–M1766 interaction predicted as a stable contact in the absence of ranolazine irrespective of CTD position

Structural characterization in the absence and presence of ranolazine captured differences in amino acid side chains used for interaction with the IGATE latch (F1486)15,17. Specifically, in the absence of ranolazine, N1765 was in close proximity to F1486 (Fig. 8a, red and pink side chains). In the presence of ranolazine, a twisting of DIVS6 was observed that resulted in translocation of M1766 consistent with formation of a hydrophobic contact with F1486 (Fig. 8b, red side chains)17. Intriguingly, our previous CTD–IGATE model predicted the M1766–F1486 interaction in the absence of ranolazine5. Analysis of our CTD–IQ motif and CTD–FREE models yielded the same result, where F1486 was positioned near M1766 in a manner consistent with hydrophobic contact (Fig. 8a,b, blue side chains). To date, our modeling efforts have suggested that the F1486–M1766 interaction can occur irrespective of CTD interaction or the presence of ranolazine. The lack of rotation to DIVS6 also allosterically controls the position of one of the AGATE amino acids. Specifically, the rotated DIVS6 (observed in all three of our NaV1.5 models) places L1772 in the spatial context of the AGATE (Fig. S9). In the open and closed CryoEM structures7,15, this spatial position is occupied by I1771 (Fig. 8c,d) and only in the presence of ranolazine was DIVS6 rotated to move L1772 into the AGATE orientation (Fig. 8e)17. Further work will be required to understand how conformational changes within the channel (different modes of gating) influence or alter F1486 interactions.

Fig. 8
figure 8

MD simulations predicted rotation of DIVS6 in the absence of ranolazine. Alignment of IFM binding pocket between the NaV1.5 CTD–IQ motif model and non-conducting CryoEM in (a) absence and (b) presence of ranolazine (ranolazine not shown in panel b for image clarity). Alignment of DIVS6 NaV1.5 IQ model and (c) non-conducting, (d) conducting, and (e) ranolazine-bound CryoEM structures. CryoEM side chains (red), postMD (blue, pink and green).

Full size image

Limitations and future perspectives

To date, our modeling efforts have predicted three energetic minima for the NaV1.5 CTD (CTD–IGATE, CTD–FREE, CTD–IQ motif). While these MD simulations provide a wealth of information regarding channel structure and insight into dynamics, there are many questions that merit further investigation. We note that our results are heavily reliant on the Amber forcefields, and our timescale (4 × 1 microsecond simulations) is unlikely to sample a complete energetic landscape. Longer time scales than what are currently feasible (millisecond) will aid evaluating larger energetic landscapes as well as interrogate the potential of our model systems trending towards an open configuration (horizontal DIII-S4 at PC/OL–PA interface). Our MD simulations do not consider the effects of membrane potential (movement of voltage sensors and/or allosteric modification of inactivation gate), lipid composition, post translational modification of NaV, or accessory protein interactions. In the absence of a concentration gradient, our simulations do not recapitulate the Na+ ion driving force (typically 145 mM extracellular, 10 mM intracellular) that aids channel permeation. Lastly, it is worth noting that the CTD–IGATE and CTD–FREE models had a greater degree of structural similarity (Fig. 9).

Fig. 9
figure 9

Influence of CTD configuration on transmembrane architecture. NaV1.5 transmembrane helices more closely resemble each other when the CTD is free or bound to the IGATE relative to the CTD–IQ motif interaction. RMSDs calculated for alpha carbons from transmembrane helices.

Full size image

Considering endogenous NaV1.5 function, this hinted that the CTD–IGATE interaction first releases and then an additional energetic input (change in membrane potential or accessory protein interaction) establishes the CTD–IQ motif configuration. Based on the predicted intermolecular forces of the CTD–IQ motif interaction, changes in the binding interface polarity (potential phosphorylation or changes in solvation) could also influence complex formation (kinetics and/or affinity). IQ motif-binding accessory proteins, such as calmodulin (CaM) and fibroblast growth factor (FGF), may influence the favorable binding interfaces of the IQ motif and CTD13,24,26,27. Further work will be required to understand the mechanisms or accessory proteins (CaM and FGF) that could yield changes in the CTD–IQ motif affinity and how the three CTD conformations described by our models integrate into the functional cycling that occurs within a cardiomyocyte.

Methods

Modeling of CTD and IQ motif components

To explore all potential interaction modes of CTD and IQ motif peptide, we carried out a naïve, protein-protein docking study. All calculations used the RosettaScripts application in Rosetta version 3.9, guided with the ref2015 scoring function28. To create the starting model complex, we used residues E1784–S1865 (CTD), taken from the representative model of the 2KBI NMR ensemble10, positioned ≥ 15 Å away from residues 1901–1925 of the IQ motif, (chain A of the 4DCK crystal structure)26. This starting model was minimized with the Relax application and then used as the starting point for creating > 100,000 decoy models with protein-protein docking. We employed a standard protocol, which successively calls the coarse dock_low mover, the srsc sidechain mover, the fine dock_high mover, finishing with the minimize interface mover. The CTD and IQ motif molecules were both randomized at the beginning of each of 112,000 docking calculations (evenly sampled a set of all possible binding interfaces for both molecules). Initial coarse Monte Carlo refinement utilized 3 Å translations and 8˚ rotations, followed with fine refinement using 0.1 Å translations and 5˚ rotation. The ex1 and ex2 extra rotamer bins were both employed during sidechain placement and interface minimization (provided high-resolution sampling at the protein-peptide interface). These 112,000 docking calculations successfully produced 104,744 unique decoys, which were each inspected in detail with the interface analyzer mover, (carried out a comprehensive analysis of the protein-protein interface and produced an estimate of the binding energy in Rosetta Energy Units (REU)). The > 100,000 analyzed decoys were ranked in order of this energy (ΔG separated). The backbone RMSD for each decoy was computed against the lowest-energy decoy. We then plotted RMSD versus ΔG separated in REU (Fig. S1). It is well documented from benchmark examples that when Rosetta is able to construct decoys that reproduce a native-like binding interaction, such a plot produces a “funnel” that points towards the origin, with all of the lowest-interface-energy decoys displaying the lowest RMSD to the best-scoring decoy. This indicated that the calculations both adequately sampled available conformational space, and accurately identified the native-like decoys. We observed this funnel-like pattern in Fig. S1. There were two funnels in this result, approximately 10 Å RMSD apart, corresponding to a “parallel” and “anti-parallel” orientation of the IQ peptide, bound to the same interface on the CTD. To clearly distinguish which of these two orientations is preferred, a second “enhanced-sampling” docking run was computed using the same protocol except that only the IQ motif was randomized and then placed near the previously established CTD interface at the beginning of the refinement. Missing residues (1866–1900) between the CTD and IQ motif were built into the low-energy anti-parallel model from this second run, using the Modeller interface in Chimera software. This model was then used as the starting conformation to three explicit-solvent 1 µs MD simulations carried out with the AMBER18 suite. After minimization, the simulations consisted of an equilibration phase of repeated heating to 298k with thermalization (six iterations for a total of ~ 1ns), followed by 1 µs of production MD at 298 K with constant pressure periodic boundary conditions at 1 atm, SHAKE enabled, and the Langevin thermostat set to a collision frequency of 3/ps.

MD simulation of intact NaV1.5 alpha subunit

The intact NaV1.5 alpha subunit was constructed using the protein coordinates from our previous CTD–IGATE model, where the CTD had dissociated from the IGATE5 (post MD simulation run 1, coordinates). The C-terminus of the channel (E1784–S1925) was deleted and replaced with coordinates from our linked CTD–IQ motif MD simulation (Fig. 1c). The resulting model was constructed and simulated using the same methodology as our previous CTD–IGATE MD modeling. Specifically, the system was protonated at a pH of 7.4 using H++29, and a lipid bilayer and solvent box were built around the model using CHARMM-GUI Bilayer Builder30. The model was oriented in the lipid bilayer by PPM 2.0, and the dimensions of the bilayer were 22.5 Å thick (Z direction) and 200 Å (total length) in the X and Y directions. The composition of the bilayer was set as heterogeneous with POPC in a ratio of 1:1. The solvent box was set to 200 Å x 200 Å x 180 Å in the X, Y, and Z dimensions, respectively, with a salt concentration of 145 mM of NaCl. The ff19SB forcefield2 was applied to the protein and the Lipid21 forcefield3 was applied to the lipid bilayer with an OPC water model4. The system had 686,972 atoms. The system was minimized, equilibrated, and 100 ns of MD simulation was performed using Amber2331. The resulting model was then sent through H + + again and CHARMM-GUI Bilayer Builder with the same parameters except the box size was reduced. The resulting system had 461,087 atoms and dimensions of 150 Å x 150 Å x 170 Å in the X, Y, and Z dimensions, respectively. This system was again minimized and split into four independent systems. Each simulation was equilibrated and 1000 ns of MD were performed. A step size of 0.004 picoseconds was achieved using SHAKE32 and hydrogen mass repartitioning33. Coordinates were written every 100 picoseconds. The CTD–FREE model was constructed and performed (four replicates of the CTD–IGATE simulation where the CTD had dissociated) using the same parameters as the CTD–IQ motif and CTD–IGATE simulations.

The MD simulations used a constant number of particles with constant pressure of 1 atm and constant temperature of 300 K (NPT ensemble). The Berendsen barostat and Langevin Dynamics thermostat were used to maintain constant pressure and temperature, respectively. The cutoff distance used in calculations was set to 9.0 Å, and the Particle Mesh Ewald algorithm calculated long range electrostatic forces.

The resulting models were analyzed using Amber 23 tools (cpptraj), VMD, gnuplot, and ChimeraX. Figures were created using Chimera, ChimeraX, and Excel. The first 2000 frames (200 ns) of each MD simulation were not used for analysis to minimize potential artifacts that could arise from system equilibration. Analysis of the CTD–IQ motif interaction was performed by visual inspection using ChimeraX and cpptraj distance plots. PC, PA, and OL contacts with amino acids were defined by distances of < 7 Å between the alpha carbon of the amino acid and the lipid molecule. Water contacts with amino acids were defined by distances of < 4 Å between the alpha carbon of the amino acid and the oxygen atom of the water molecule. Contacts with PC, PA, OL, and water were found using Amber23 tools and analyzed using Excel. RMSF values were found using Amber23 tools and analyzed using excel. Comparisons of voltage sensors were made from visual inspection and RMSD calculations using ChimeraX. Similarly, comparison of the MD models to CryoEM structures were made from visual inspection and RMSD calculations using ChimeraX. Ion contacts with the selectivity filter and ion contacts with water were identified using VMD, ChimeraX, and Excel. Run scripts and cptpraj analysis scripts are available from our previous publication (https://zenodo.org/records/13225850). Pre and post-MD pdb files for this current manuscript are available (https://zenodo.org/records/14641335).