Drugs exhibit diverse binding modes and access routes in the Nav1.5 cardiac sodium channel pore

Abstract

Small molecule inhibitors of the sodium channel are common pharmacological agents used to treat a variety of cardiac and nervous system pathologies. They act on the channel via binding within the pore to directly block the sodium conduction pathway and/or modulate the channel to favor a non-conductive state. Despite their abundant clinical use, we lack specific knowledge of their protein-drug interactions and the subtle variations between different compound structures. This study investigates the binding and accessibility of nine different compounds in the pore cavity of the Nav1.5 sodium channel using enhanced sampling simulations. We find that most compounds share a common location of pore binding-near the mouth of the DII-III fenestration-associated with the high number of aromatic residues in this region. In contrast, some other compounds prefer binding within the lateral fenestrations where they compete with lipids, rather than binding in the central cavity. Overall, our simulation results suggest that the drug binding within the pore is highly promiscuous, with most drugs having multiple low-affinity binding sites. Access to the pore interior via two out of four of the hydrophobic fenestrations is favorable for the majority of compounds. Our results indicate that the polyspecific and diffuse binding of inhibitors in the pore contributes to the varied nature of their inhibitory effects and can be exploited for future drug discovery and optimization.

Figure 1.

Structure of sodium channel pore module and pore-blocker sampling method. (A) Topology of sodium channel transmembrane helices (S1–S6) and the re-entrant selectivity filter (SF) loop arranged in four pseudo-tetrameric domains (DI–DIV); conserved S6 residues that form the pore binding site are labeled—rat Na_v1.5 numbering. (B) Components of the simulations featuring a drug molecule (blue stick representation) inside the closed sodium channel pore (shown as a cut-through side view with two of the domains visible), embedded in POPC bilayer (sphere representation) and solvated in NaCl ions; dotted blue line indicates the region sampled by the drug in enhanced simulations and red cross represents the drug’s center of mass (COM) and red dotted lines define the central pore axis and the drug COM relative to the central axis. (C) The pore module is formed by the S5 and S6 helices and the selectivity filter (SF), shown in new cartoon representation; positions of the conserved drug binding residues shown in stick representation. (D) Top-down view of a cut-through representation of the pore module showing the regions of the central cavity and four orthogonal fenestrations between the four domains that were sampled by drugs; red dashed lines depict the angle used to define the polar coordinates of the drug COM (red cross).

Figure 2.

Reproducing the cryo-EM (PDB ID 6uz0 ) pore-bound position of flecainide in the rat Na _v 1.5. (A and B) Free energy surfaces (FES, top-down and transverse views) generated from metadynamics sampling of the pore using both neutral flecainide (A) and protonated flecainide (B); FES contours are shown every 4 kJ/mol and lighter colors indicate higher energy or less favorable regions of the pore; center of mass (COM) positions of the representative pose from each of the top five clusters shown in white markers, as well as the COM position of the cryo-EM flecainide pose (cross). (C) Representative snapshot of the top cluster for neutral and protonated flecainide compared to the cryo-EM structure (showing top-down and transverse views of the pore). (D) RMSD of the top neutral and protonated clusters to the cryo-EM pose; the number of data points in each cluster shown above each box/violin. (E) Residue interactions between the neutral (blue) and protonated (red) flecainide identified within the top five clusters. (F) Representative poses of the top five clusters for neutral (left) and protonated (right) flecainide; dotted square highlights the positioning of flecainide’s trifluoromethylated arm in the DII–III fenestration. (G) The pore with resolved flecainide from the cryo-EM structure 6uz0, overlayed with the density around the drug. (H) Close-up of the flecainide cryo-EM density colored by the OccuPy scale value, where values closer to 1 represent better relative local resolution in the cryo-EM map.

Figure S1.

Mobility of drugs in equilibrium and metadynamics simulations . (A) Average RMSD of drug for the 20 replicates over 100 ns of unbiased equilibrium simulations compared to 20 walkers over 100 ns of metadynamics simulations – aligned to the Na_v pore domain backbone (residues 236–285, 313–417, 824–945, 1321–1504, and 1641–1780). (B) Average RMSD of drug within each replicates/walkers of unbiased equilibrium simulations compared with metadynamics simulations – fit to the drug. (C) Percentage of frames out of the metadynamics trajectories in each of the top five clusters.

Figure S2.

All free energy surfaces (top-down view) for nine neutral pore-binding inhibitors, in blue, produced from metadynamics simulations with positions of top five clusters, plus an additional four versions of protonated versions of four of these, in red (flecainide, mexiletine, bupivacaine, and etidocaine). Red circles on the chemical structures highlight the positions of the protonations and green circles highlight the positions of chiral centers. Darkest color represents 0 kJ/mol or the most favorable drug binding sites; contours are plotted every 4 kJ/mol, where lighter colors indicate higher energy or less favorable regions of the pore.

Figure S3.

All free energy surfaces (transverse views along the plane of the bilayer) for nine neutral pore-binding inhibitors, in blue, produced from metadynamics simulations with positions of top five clusters, plus an additional four versions of protonated versions of four of these, in red (flecainide, mexiletine, bupivacaine, and etidocaine). Darkest color represents 0 kJ/mol or the most favorable drug binding sites; contours are plotted every 4 kJ/mol, where lighter colors indicate higher energy or less favorable regions of the pore.

Figure S4.

Representative poses of the top five clusters (top-down view) generated from all frames of metadynamics simulations for the nine neutral pore-binding inhibitors, and additional four versions of protonated versions of four these inhibitors (flecainide, mexiletine, bupivacaine, and etidocaine).

Figure S5.

Representative poses of the top five clusters (transverse views along the plane of the bilayer) generated from all frames of metadynamics simulations for the nine neutral pore-binding inhibitors, and additional four versions of protonated versions of four of these inhibitors (flecainide, mexiletine, bupivacaine, and etidocaine).

Figure 3.

Binding of bupivacaine is localized to the central pore cavity . (A–E) The pore binding mode of bupivacaine in neutral (A and C) versus protonated (B and D) form – showing the free energy surfaces (FES) reproduced from metadynamics simulations (A and B); representative binding pose of each of the top five clusters (C and D); percentage of the top five clusters where interactions between the mexiletine and protein residues were identified (E). FES contours are shown every 4 kJ/mol.

Figure 4.

Mexiletine shows preference for binding in both the pore cavity and fenestrations . (A–E) The mixed binding modes of mexiletine in neutral (A and C) versus protonated (B and D) form—showing the free energy surfaces (FES) reproduced from metadynamics simulations (A and B); representative binding pose of each of the top five clusters (C and D); percentage of the top five clusters where interactions between the mexiletine and protein residues were identified (E). FES contours are shown every 4 kJ/mol.

Figure 5.

Riluzole and lamotrigine bind promiscuously in two or three of the fenestrations . (A–E) Fenestration binding mode of riluzole (A and C) and lamotrigine (B and D)—showing free energy surfaces (FES) reproduced from metadynamics simulations (A and B); representative binding pose of each of the top five clusters compared to the pose captured in new cryo-EM structures, PDB IDs 8thg and 8thh (C and D); percentage of the top five clusters where interactions between the drug and protein residues were identified (E). FES contours are shown every 4 kJ/mol.

Figure S6.

Residue interactions were identified across the top five clusters of each drug. (A) Summary bar chart showing the most common interacting residues identified for all drugs, showing residues where total occupancy >50% for at least one of the drugs. (B) Locations of each of the residues in panel A shown on the S5/S6/P-loop of the four domains. (C) Stacked bar chart of the number of interacting residues categorized into the different interaction types.

Figure S7.

Residue interactions for neutral and protonated drug species across each of the five different clusters . (A–D) Top interactions stacked by cluster for flecainide (A), mexiletine (B), bupivacaine (C), and etidocaine (D)—neutral and protonated. Interactions are not shown if occupancy <50% in each of the five clusters. Darkest color indicates the top cluster.

Figure 6.

Accessibility via the fenestrations to drug compounds and lipid tails. (A) Percentage likelihood for each drug to occupy the pore or the four fenestrations. (B) Representative 2D averaged volumetric density map of lipid tail atoms (from neutral flecainide metadynamics simulations) showing the varied extent of lipids occupying the four fenestrations, where darker regions indicate a greater presence of lipid atoms throughout all simulations. (C) Representative snapshot of POPC lipid tails occupying fenestrations (in non-biased simulations) in the absence of a pore-blocker. (D) Representative snapshot of the absence of POPC tail in the DII–III fenestration when a drug is bound. (E) Average of the lipid tail density values for each of the four regions defined as the fenestrations; n = 9 measured across the nine different drugs metadynamics simulations; error bars indicate standard error of the mean (SEM). (F) Summarizing the spectrum of drug binding preference in the Na_v channel pore. (G) Simplified schematic of the two modes of drug occupation in the fenestrations as opposed to the central pore cavity.

Figure S8.

Potential correlations between drug promiscuity/percentage of pore occupancy and physiochemical/pharmacokinetic properties. (A–F) Values are plotted from Table 2, with trend lines shown for relationships where Pearson correlation coefficient, r < −0.6, r > 0.6 (i.e., the correlation values highlighted in bold in Table 2). Each panel shows the pairwise comparison between (A) molecular weight of the drugs and percentage of pore occupancy/promiscuity of neutral versions only; (B) octanol-water partition coefficient (LogP) and promiscuity; (C) number of rotatable bonds and percentage of pore occupancy; (D) IC50 and promiscuity; (E) promiscuity and association/dissociation kinetics; (F) promiscuity of neutral versions of each drug and affinity to the inactivated state of Na_v.

Figure S9.

Sequence alignment of pore helices, S5, S6 and the selectivity filter/P-loop of rat Na _v 1.5 and all the human Na _v subtypes. Top drug-interacting residues highlighted accordingly.

Figure S10.

Convergence of metadynamics simulations. (A) Comparison of free energy surfaces (FES) produced at different timepoints (1,000, 2,000, 3,000 and 4,000 ns) for neutral flecainide. (B) The number of FES grid points in the pore and fenestration region sampled over the course of metadynamics simulations for all drugs.