Mechanisms of dihydropyridine agonists and antagonists in view of cryo-EM structures of calcium and sodium channels

Calcium channels play key roles in the physiology of electrically excitable cells by mediating currents that trigger excitation–transcription coupling, neurotransmitter release, hormone secretion, gene transcription, and other processes (Zamponi et al., 2015). L-type calcium channels (Cav1.1–Cav1.4), which are expressed in cardiac, skeletal, and smooth muscles, contain the pore-forming α₁ subunit and ancillary subunits Cavβ, Cavα2δ, and Cavγ. The pore-forming α₁ subunit folds from a single polypeptide chain of four homologous repeats. Each repeat includes six transmembrane helices and an extracellular membrane-reentering P-loop with selectivity-filter residues between helices P1 and P2. In each repeat, helices S1–S4 form a voltage-sensing domain, while helices S4–S5, S5, S6, and the P-loop contribute a quarter to the pore domain. The extracellular part of the pore domain has similar conformations in the open and closed channels. In contrast, conformations of the cytoplasmic half of the pore domain, in particular, the activation gate region, are significantly different in the open and closed states. The first cryo-EM structures of the rabbit Cav1.1 channel (Wu et al., 2015, 2016) enabled structure-based interpretations of experimental data on physiology, pathophysiology, and pharmacology of calcium channels.

1,4-Dihydropyridines (DHPs), phenylalkylamines, and benzothiazepines are three major classes of L-type calcium channel ligands (Hockerman et al., 1997; Lacinová, 2005; Catterall, 2011; Catterall et al., 2020). DHPs are used to treat hypertension, cardiac ischemia, pain, and tremor (Zamponi et al., 2015; Wang et al., 2017; Parthiban and Makam, 2022; Crossley et al., 2022). Large efforts have been made to explore structure–activity relations of different DHP agonists and antagonists, map their binding site in calcium channels, and understand mechanisms of action. Unlike phenylalkylamines and benzothiazepines, which directly block the ion permeation pathway, DHP agonists and antagonists allosterically decrease or increase calcium currents.

The structure of a DHP molecule can be described as a flattened boat (Fig. 1 A), with the axial aryl ring as the bowsprit, the NH group at the stern, and various substitutions at the port and starboard sides (Goldmann and Stoltefuss, 1991). Structure–function analyses revealed that the portside group is a major determinant of the agonistic or antagonistic activity. Compounds with a short hydrophilic group (like NO₂) act as agonists, whereas compounds with a larger hydrophobic group act as antagonists. The opposite action of enantiomeric DHPs Bay k 8644 (Franckowiak et al., 1985; Schramm et al., 1985) or 202–791 (Kongsamut et al., 1985) exemplify this intriguing phenomenon. Detailed analysis demonstrates that DHP agonists stabilize long openings, whereas DHP antagonists promote inactivated states (Hess et al., 1984; Kokubun and Reuter, 1984).

Numerous mutational studies unambiguously determined the DHP binding site. The DHP-sensing residues were found in helices S5_III, P1_III, S6_III, and S6_IV (Hockerman et al., 1997; Lacinová, 2005). These residues face the fenestration between repeats III and IV that was earlier proposed to form a hydrophobic access pathway to the pore in sodium and calcium channels (Tikhonov and Zhorov, 2005; Tikhonov et al., 2006). The existence of such a pathway in sodium channels has long been predicted by Hille (1977). Atomic-scale structures directly demonstrated that in sodium and calcium channels the fenestrations are much wider than in potassium channels, implying that they could accommodate small molecules such as local anesthetics (Payandeh et al., 2011). In the absence of high-resolution structures of calcium channels with DHPs, numerous experimental data on the action of DHPs (including mutational analyses) were initially rationalized using computational models (Lipkind and Fozzard, 2003; Cosconati et al., 2007; Tikhonov and Zhorov, 2009), which were based on available structures of potassium channels. Despite some differences in the models, they suggested that in the closed channel, the DHP portside has a hydrophobic environment, whereas in the open channel, the portside has a hydrophilic environment.

Crystal structures of engineered prokaryotic homotetrameric calcium-permeating channel CavAb complexed with various ligands revealed a DHP-binding site in the fenestration (Tang et al., 2016). However, the DHP molecules are more distant from the pore lumen as compared with what was suggested in mutational studies of eukaryotic calcium channels. More recently, cryo-EM structures of DHP complexes with the Cav1.1 channel were obtained (Zhao et al., 2019a; Gao and Yan, 2021). In these structures, the binding positions of different DHP antagonists and agonists are very similar (Fig. 1, B–D), indicating the common DHP binding pose. The stern, bowsprit, starboard, and portside approach, respectively, residues in helices P1_III, S6_III, S5_III, and S6_IV, while a portside moiety faces the pore lumen. This binding mode generally agrees with available mutational data (Fig. 1 E), although not all experimentally defined DHP-sensing residues are in close contact with the molecules.

Regrettably, these atomic-scale structures do not provide an unambiguous rationale for the opposite action of DHP agonists and antagonists. The channel has practically the same conformation in the presence of agonists or antagonists. In the cryo-EM structures, the channel is captured in a supposedly inactivated state (Zhao et al., 2019a). It remains unclear how DHP agonists stabilize the open state, whereas DHP antagonists stabilize the closed/inactivated state. A possible explanation is that these DHP ligands cause only moderate changes in the probabilities of different functional states, while the most preferable state remains the same.

Recent structural studies revealed an interesting peculiarity of S6 segments in sodium, calcium, and TRP channels. In some structures, these segments are not entirely α-helical but contain a π-helical turn at a position where the glycine gating hinge is located in potassium channels (Tikhonov and Zhorov, 2017, 2021, 2022). Since a π-helix has an extra residue per turn as compared with an α-helix, the π-helical turn is seen as a bulge in the helix. Furthermore, the π-bulging causes a drastic (about 100°) reorientation of residues downstream of the helix. Although the exact role of π-helical turns is still unclear, available data suggest that transitions between the α and π forms of S6 may be important for gating (Zubcevic and Lee, 2019). Another interesting observation is that such transitions can be induced by ligand binding in the pore (for review, see Tikhonov and Zhorov, 2021). Importantly, the π-helical bulges appear exactly at the DHP binding region. These observations prompted us to suggest an initial hypothesis that the binding of DHPs can favor or disfavor π-helical elements in calcium channels, which, in turn, modulate the channel gating. Here, we employed molecular modeling of the Cav1.1 channel with DHP agonists and an antagonist with the aim to understand different mechanisms of their action on the channel gating. We used as templates cryo-EM structures of sodium and calcium channels with or without π-helical bulges in segments S6_III and S6_IV and optimized the complexes with Monte Carlo energy minimizations. Based on these calculations, we elaborated a hypothesis that dynamic π-bulging of S6 helices and their return to completely α-helical structure affects the gating in calcium channels due to dramatic reorientation of S6 residues C-terminal to the bulge. Our models suggest that DHP agonists stabilize a π-bulged helix S6_IV, whereas DHP antagonists stabilize entirely α-helical structure of S6_IV.

Our methodology of molecular modeling and ligand docking is described in Garden and Zhorov (2010) and Tikhonov and Zhorov (2017). Briefly, we used the method of Monte Carlo energy minimizations (Li and Scheraga, 1987) in the space of internal (generalized) coordinates and the AMBER force field (Weiner et al., 1984, 1986), as realized in the ZMM program (https://www.zmmsoft.ca). Electrostatic interactions were calculated with the distance- and environment-dependent dielectric function (Garden and Zhorov, 2010). Atomic charges in ligands were calculated by MOPAC (Dewar et al., 1985). No distance cutoff was used to calculate electrostatic interactions involving ionized groups. For other interactions, we used the distance cutoff of 9 Å and a shifting function (Brooks et al., 1985).

The Monte Carlo minimization (MCM) sampling protocol randomized torsion angles in ligands and the channel side chains, as well as generalized coordinates that govern the position, orientation, and conformation of the ligand. However, during energy minimizations, all generalized coordinates (both backbone and side chain torsions of the channel, torsion angles of the ligand, and its bond angles) were treated as flexible. In the MCM trajectory, random displacements of the ligand from the experimental position were limited to 4 Å.

We used “pin” constraints to ensure the similarity of the backbone conformation in the model and the cryo-EM template. A pin is a flat-bottom parabolic energy function that imposes the energy penalty if a C^α atom in the model deviates by more than 1 Å from the template position. For all constraints, the energy penalty was calculated with the force constant of 10 kcal∙mol⁻¹∙Å⁻².

We used the following protocol for ligand docking. We imposed a starting position and orientations of the ligand as seen in respective cryo-EM structure, randomly sampled positions and orientations of the ligand, as well as side-chain conformation of residues within 10 Å from the ligand, and MC-minimized the energy until 1,000 consecutive energy minimizations did not improve the energy. MCMs ensure intensive sampling of the conformational space and provide merely approximate values of the enthalpy contributions to the free energy of ligand–channel interactions. Therefore, only relative values of these energies are meaningful when different ligand binding poses in models with different structural templates are compared.

We label residues by using a scheme that is universal for P-loop channels (Zhorov and Tikhonov, 2004; Table 1). While the major target for dihydropyridines is Cav1.2 channels, we modeled Cav1.1 channels for which cryo-EM structures in complexes with DHPs are available. The amino acid sequences of the Cav1.1 and Cav1.2 channels are practically identical in those parts of helices P1_III, S5_III, S6_III, and S6_IV that contribute to the DHP binding site (Table 1).

As the first step, we MC minimized available structures of the Cav1.1 channel with agonists and antagonists to obtain energetic characteristics of DHP–channel interactions. DHP-bound structures with the following PDB indexes were used: 6jp8 (S-Bay k 8644), 6jp5 (nifedipine), 7jpk (S-Bay k 8644), 7jpl (S-Bay k 8644), and 7jpw (R-Bay k 8644). The MC-minimized structures slightly deviated from the experimental ones. The RMS deviations for α carbons in S5 and S6 segments varied from 0.6 to 0.7 Å. The largest deviations were obtained for sodium channel templates 5x0m (0.7 and 0.71 Å for S5 and S6 segments, correspondingly) and 7fbs (0.71 and 0.69 Å for S5 and S6 segments, correspondingly). The smallest RMS deviations were found for template 7jpl (0.61 and 0.60 Å for S5 and S6 segments, correspondingly). In all the structures, the DHP position was stable, and the pattern of interactions was conserved although random reorientations and translations up to 4 Å were sampled during the MCM trajectory. This stability of drug binding pose suggests that it corresponds to an energy minimum. The major contributors to the interaction energy were the same in all models: Val^3o10, Thr^3o13, Gln^3o17, Phe^3p44, Ser^3p47, Ile³ⁱ¹⁴, Met³ⁱ¹⁸, Met³ⁱ¹⁹, Phe³ⁱ²², Tyr⁴ⁱ¹¹, and Ile⁴ⁱ¹⁹. Most of these residues surround the DHP molecule providing multiple specific interactions (Fig. 1 C).

Close inspection of the portside group contacts, which are critical for the agonist/antagonist action, showed that the methoxy group of nimodipine and R-Bay k 8644 approached the side chain of isoleucine Ile⁴ⁱ¹⁹ (Fig. 2 A). The attraction between the methoxy group and the side chain was rather weak (−0.2 to −0.5 kcal/mol). The hydrophobic methoxy group fit nicely into the hydrophobic pocket formed by the side chains of Ala⁴ⁱ¹⁵, Phe⁴ⁱ¹⁶, and Ile⁴ⁱ¹⁹. The NO₂ group at the portside of agonist S-Bay k 8644, which is much smaller than the methoxy group, did not fill up this pocket (Fig. 2 B). The void space at the NO₂ group of S-Bay k 8644 would be filled up by a water molecule, which is sterically comparable with the methyl group at the portside of R-Bay k 8644. We placed a water molecule in the NO₂ group of S-Bay k 8644 (Fig. 2 C). However, MCMs demonstrated that the water molecule is unstable in this position. When the water molecule was constrained in the starting position with a possibility to rotate and shift up to 2 Å, it experienced a noticeable repulsion (+1.8 kcal/mol) from the hydrophobic pocket and S-Bay k 8644. When the distance constraint was removed, the water molecule escaped from the hydrophobic pocket (data not shown). Thus, completely α-helical S6_IV attracts antagonist R-Bay k 8644 much stronger than agonist S-Bay k 8644.

3-D alignment of available structures of P-loop channels suggests that calcium and sodium channels have similar folding of the pore domain, which differs from that in potassium, TRP, and glutamate-gated channels (Tikhonov and Zhorov, 2021, 2022). Therefore, crystal and cryo-EM structures of sodium and calcium channels can be used as templates to model channel Cav1.1 in different functional states. The Protein Data Bank contains numerous crystal structures of homotetrameric prokaryotic sodium and calcium channels and cryo-EM structures of pseudo-heteromeric eukaryotic sodium and calcium channels. 3-D alignment of the structures shows that the most striking differences in the region of DHP binding (helices S5_III, P1_III, S6_III, and S6_IV) are due to the presence or absence of π-helical bulges in helices S6_III and S6_IV. Upon π-bulging, the helical residues, which are C-terminal to the bulge region, turn anticlockwise by ∼100° when viewed along the helix from its N-end. Therefore, residues in completely α-helical S6_III and S6_IV that contribute to the DHP binding site would turn away from the binding site in π-bulged helices.

Crystal structures of all CavAb channels have completely α-helical segments S6 (pattern αααα, where Greek characters in positions 1, 2, 3, and 4 indicate, respectively, α-helical conformations of subunits I, II, III, and IV). In contrast, cryo-EM structures of the NavPaS channel show π-bulged S6 helices in all four repeats (pattern ππππ). The majority of eukaryotic Cav and Nav channels, including DHP-bound structures, have completely α-helical segments, S6_II and S6_Iv, but contain π-bulges in segments S6_I and S6_III (pattern παπα). In these structures, DHPs bind between the π-bulged helix S6_III and the completely α-helical segment S6_IV. Importantly, there are cryo-EM structures with various patterns of π-bulges in the four repeats. To explore DHP binding models in different homology models, we selected as templates cryo-EM structures of two calcium channels and four sodium channels. The calcium channel templates are Cav1.1 with verapamil (PDB ID: 6jpa, pattern πααα) and Cav3.1 with ligand Z944 (PDB ID: 6kzp, pattern ππαα). The sodium channel templates are ligand-free channel NavPaS (PDB ID: 5x0m, pattern ππππ), Nav1.7 with a proteinaceous toxin and saxitoxin (PDB ID: 7w9p, pattern παππ), Nav1.7 with the pore blocker XEN907 (PDB ID: 7xm9, pattern παππ), and channel rNav1.5 with putatively open pore domain (PDB ID: 7fbs, pattern παπα). Fig. 3 shows that in these templates the orientations of residues in helix P1_III and N-terminal parts of helices S6_III and S6_IV are well conserved, whereas the orientation of residues in the C-halves of helices S6 are drastically different.

We generated homology models of the Cav1.1 channel by using these templates. Antagonist R-Bay k 8644 and agonist S-Bay k 8644 (with a water molecule attached to the NO₂ group) were placed in these structures. The initial position of the DHPs was imposed as in the cryo-EM structures of the Cav1.1 channel with S-Bay k 8644 (PDB ID: 7jpl) or R-Bay k (PDB ID: 7jpw). The complexes were MC-minimized. The total DHP–channel interaction energies and contributions of individual residues to the total energies were compared.

Superimposition of R-Bay k 8644 and S-Bay k 8644 in the MC-minimized models are shown in Fig. 4, A and B. In all the models, DHP positions are rather close to those seen in experimental structures. This fact implies that the DHP agonists and antagonists have a common binding site and binding mode. In helix S5_III, the strongest interactions are those between the DHP bowsprit and Val^3o10 as well as between the DHP starboard and Thr^3o13 whose hydroxyl group can donate an H-bond to either NO₂ or COOM group of the bound DHP molecule. In helix P1_III, the strongest attraction involved Phe^3p44. The above interactions did not depend significantly on the channel model or the bound DHP enantiomer. In helix S6_III, ligand interaction with Ile³ⁱ¹⁴ was conserved but downstream from this position, the side chain orientations and thus interactions with DHPs strongly depended on the template employed. In all the models, the ligands interacted with Met³ⁱ¹⁸ regardless of the side chain orientation of this residue, which depended on the presence or absence of π-bulge in helix S6_III.

In models with π-bulged S6_III, the side chain of Met³ⁱ¹⁸ approached the DHP bowsprit aromatic moiety from the fenestration, while in models with entirely α-helical S6_III, it extended toward the pore lumen but still interacted with the ligand bowsprit. In contrast, Met³ⁱ¹⁹ interacted with DHPs only in models with π-bulged helix S6_III (Fig. 4 C), where Met³ⁱ¹⁹ occupied the same space as Met³ⁱ¹⁸ in the models with entirely α-helical S6_III. Significant interactions between the DHP bowsprit group and Ile³ⁱ²¹ were found only in models with entirely α-helical S6_III. Only in models with π-bulged helix S6_III, Phe³ⁱ²² at the cytoplasmic side of the DHP binding site interacted with the bowsprit aromatic ring of the ligand in an edge-to-face manner. In the helix S6_IV, Tyr⁴ⁱ¹¹, Met⁴ⁱ¹², Ala⁴ⁱ¹⁵, and Phe⁴ⁱ¹⁶ are located upstream of the π-bulging region and therefore these residues interacted with DHPs in all the models. Ala⁴ⁱ¹⁵ and Phe⁴ⁱ¹⁶ also repelled the water molecule at the NO₂ group of S-Bay k 8644. Ile⁴ⁱ¹⁹ weakly interacted with DHPs, but it significantly repelled the water molecule at S-Bay k 8644 in the models with entirely α-helical S6_IV.

Total DHP–channel interaction energies varied between −6.4 and −12.0 kcal/mol. Channel models based on the templates with entirely α-helical S6_III and S6_IV (6jpa and 6kzp) are characterized by rather weak interactions with both S-Bay k 8644 and R-Bay k 8644 (−6.4 to −9.7 kcal/mol) that approached helix S6_III with the bowsprit. For R-Bay k 8644, the binding mode with the strongest interaction energy (−11.4 and −12.0 kcal/mol) was found in models with π-helical S6_III and entirely α-helical S6_IV (templates 7lpw and 7fbs), while models with π-bulged S6_III and S6_IV (templates 7w9p, 7xm9, and 5x0m) had intermediate values of ligand–channel interaction energy (−9.7 to −10.4 kcal/mol). In contrast, agonist S-Bay k 8644 had stronger interaction energies in models where both S6_III and S6_IV were π-bulged (−10.2 to −11.8 kcal/mol). Models with π-bulged S6_III and entirely α-helical S6_IV are characterized by intermediate ligand–channel energies (−9.7 to −10.4 kcal/mol). This difference is due to the reorientation of the residues that surround the water molecule at the NO₂ group of S-Bay k 8644.

The energy difference of the water molecules in various models was big. In models with entirely α-helical S6_IV, the repulsive energy was between +1.6 and +1.8 kcal/mol due to unfavorable desolvation in the hydrophobic pocket. In contrast, in models with π-bulged S6_IV, the repulsion energy of the water molecule was much weaker (+0.3 kcal/mol). Isoleucine Ile⁴ⁱ¹⁹ in these models turned away from the DHP ligand and its position was replaced by Asn⁴ⁱ²⁰ whose polar side chain created a favorable environment for the water molecule (Fig. 4 D). An interesting difference was found regarding the interaction energy of the water molecule with agonist S-Bay k 8644. It was repulsive in the models with entirely α-helical S6_IV, but attractive in the models π-bulged S6_IV. In the former models, the water molecule avoided an unfavorable hydrophobic environment and closely approached the NO₂ group, but repelled from it. In the latter models, the water molecule was at the optimal distance from the NO₂ group and the water molecule energy was negative (attractive). When the water molecule was removed, agonist S-Bay k 8644 had very similar interaction energies in models with π-bulged S6_IV and α-helical S6_IV. Thus, the models with entirely α-helical S6_III are unfavorable for both DHP agonist and antagonist. The models with entirely α-helical S6_IV are favorable for the DHP antagonist while models with π-bulged helix S6_IV favor the DHP agonist with a water molecule in the NO₂ group.

It also should be noted that the DHP binding energy was rather similar in the models based on either Cav or Nav templates. We also did not notice significant energy differences between the models based on putatively open Nav structure (PDB ID: 7fbs) and other structures where the channels are in presumably inactivated states. The energy of antagonist R-Bay k 8644 in the model, which is based on the rNav1.5 template with the open activation gate (PDB ID: 7fbs), is −12.1 kcal/mol. In contrast, in the model based on the NavPaS template (PDB ID: 5x0m) with the activation gate closed at the level of residues in positions i57, the interaction energy is weaker (−9.7 kcal/mol). The binding poses of agonist S-Bay k 8644 in these models had very similar interaction energies (approximately −10.2 kcal/mol). Thus, unlike π-bulging of helices S6, the aperture of the activation gate did not directly affect the binding of DHPs.

In the models with π-bulged S6_IV, it is Asn⁴ⁱ²⁰ rather than Ile⁴ⁱ¹⁹ that contributed to the DHP binding site. The polar side chain of the asparagine created a favorable environment for a water molecule at the NO₂ group of the DHP agonist. Assuming the limitations of static structures, we modeled α to π conformational rearrangement in helix S6_IV with a constraint-driven approach. We used the cryo-EM structure of S-Bay k 8644 bound channels Cav1.1 (PDB ID: 7jpl) as the starting model. A π-bulge in helix S6_IV was imposed by H-bonding distance constraints between backbone carbonyls in positions 4i12–4i14 on the one hand and amide groups in positions 4i17–4i19 on the other hand. After the constraint-driven MCM trajectory converged, we removed the pin constraints from residues downstream from position 4i10 but preserved α-helical H-bonds involving these residues. MCM of the structure resulted in an interesting model (Fig. 5 A). The side chain NH₂ group of Asn⁴ⁱ²⁰ donated an H-bond to the backbone carbonyl of Phe⁴ⁱ¹⁶, which is a “bachelor” (lacks an H-bond donor) in the π-bulge. The same NH₂ group also donated an H-bond to a water molecule at the NO₂ group of S-Bay k 8644. The water molecule donated an H-bond to the NO₂ group of S-Bay k 8644 and another H-bond to the bachelor carbonyl of Ala⁴ⁱ¹⁵. With this H-bonding pattern, the interaction energy of the water molecule was favorable (−1.3 kcal/mol), while the binding energy of S-Bay k 8644 was −11.4 kcal/mol. Thus, modeling of the conformational rearrangement suggests that models with π-bulged S6_III and α-helical S6_IV are more preferable for the binding of DHP antagonists, whereas models with π-bulged S6_III and S6_IV are more preferable for the binding of DHP agonist with a water molecule at the NO₂ group. The latter models have several H-bonds involving the DHP agonist, a water molecule, and S6_IV residues, while Asn⁴ⁱ²⁰ has a key role in the agonist–channel interactions.

In addition to models with S-Bay k 8644, we also explored models with DHP agonists H 160/51 (Gjörstrup et al., 1986), S-202–791 (Kavalali and Plummer, 1994), and CGP-28392 (Brown et al., 1984). Agonist S-202–791 has the NO₂ group at the portside and differs from S-Bay k 8644 only in the bowsprit moiety. In agonist CGP-28392, the portside has a polar five-membered ring. The binding mode of CGP-28392 in the model based on the 7jpl structure with imposed π-bulge in S6_IV is very similar to the binding mode of S-Bay k 8644 (Fig. 5 B). Compound H 160/51 lacks the portside substituent, while a hydrogen atom at the C_sp2 carbon of the DHP ring bears a partial positive charge. To obtain the binding model, which is analogous to the binding model of other DHP agonists, we put the second water molecule to fill up the space between the agonist and helix S6_IV. The binding pose of H 160/51 in the MC-minimized model was similar to that of S-Bay k 8644 and CGP-28392 (Fig. 5 C), and the ligand–channel energy was favorable.

The interesting feature of our constraint-driven modeling of α to π transition is that the pore lumen in the π-form resulting structure is narrower than in the initial α-form structure (Fig. 6 A). The transition caused slight but well-detectable movement of the S6_IV C-end toward the pore axis. Similar changes associated with the drug-induced α to π transition are seen in the experimental structures of Nav1.7 channel (Huang et al., 2022; Wu et al., 2023) and Cav1.1 channel (Zhao et al., 2019a). A comparison of the Nav1.7 channel structures in the apo state (7w9k) and in the presence of bupivacaine (8i5b) is shown in Fig. 6 B. Both S6_IV and S6_I demonstrate slight but noticeable shifts toward the pore axis. However, a comparison of the open (6dvz; Singh et al., 2018) and closed (6mho; Zubcevic et al., 2018) structures of TRPV2 (Fig. 6 C) demonstrates the opposite tendency: in the open structure the S6 segments have π-bulges, while in the closed structure they are entirely α-helical.

To understand the nature of this opposite relationship between the presence of π-bulges in S6 helices and the pore dimensions, we considered the S4–S5 helices that form a “cuff” around the S6 segments bundle. Dimensions of the cuff, which is controlled by the voltage-sensing helices S4, limit the S6s divergence. In our model of local α–π transition and in the Nav1.7 structures, where the transition is caused by drug binding, the S4–S5 cuff is the same in the α-form and π-form structures. In contrast, in the open-pore structure of TRPV2, the cuff is noticeably wider than in the closed structure.

α–π Transition causes reorientation of C-part residues, and thus affects contacts of S6_IV with neighboring S6_I and linker helices S4–S5. Such disruption of the intersegment contacts can destabilize the S6 helix bundle. If the cuff remains rigid, it does not allow the S6 to move outward and only some inward movement can be seen. Such tightening of the pore, however, should not be directly associated with transitions between the open and closed states since the latter should include changes in the S4–S5 cuff dimensions. As the channel opens and the cuff becomes wider, the π-bulged S6s can move outward and stabilize the open structure.

Fig. 7 illustrates this hypothesis. As was noted above, the transition involves not only residues around the DHP binding site but all the S6 residues downstream from the bulge. In particular, the movement involves residues that form tight contacts with the neighboring helix S6_I and linker-helix S4–S5_IV in the gate region. Fig. 7 exemplifies this reorientation for residue Ile⁴ⁱ²⁶. In the entirely α-helical conformation of S6_IV, Ile⁴ⁱ²⁶ faces S6_I, but in the model with π-bulged S6_IV, it turns toward the linker helix. Such a drastic change of the contact patterns should affect the stabilities of different conformations of the gate region. For instance, interactions between neighboring S6s would stabilize the helical bundle in the closed state. On the contrary, tight interactions with the linker helix would allow helix S6_IV to move outward when the cuff formed by linker helices S4–S5 is widened in the open state of the pore domain. Of course, when the cuff is in tight conformation, which is controlled by S4 voltage sensors, even π-bulged segment S6_IV has no room to move outward. It has been reported that DHP agonists stabilize long openings, whereas DHP antagonists promote an inactivated state (Hess et al., 1984). Our hypothesis fully agrees with these aspects of DHP action. Thus, we propose that local transitions between α and π forms of S6_IV induced by DHPs affect the orientation of residues at the C-end of S6_IV that affect channel gating. The transitions affect multiple intersegment contacts, but they cannot directly induce the gate opening.

In this work, we employed a molecular modeling approach to compare interaction energies of the DHP agonist S-Bay k 8644 and antagonist R-Bay k 8644 with the Cav1.1 channel using as templates cryo-EM structures of eukaryotic sodium and calcium channels with different patterns of π-bulges in the inner helices S6. Our calculations predict that DHPs interact with the π-bulged helix S6_III stronger than with entirely α-helical S6_III. Antagonist R-Bay k 8644 also stabilized α-helical conformation of S6_IV due to preferable interactions of its portside methoxy group with the hydrophobic pocket in this helix. In contrast, a small polar group NO₂ in agonist S-Bay k 8644 did not fill up this pocket, which has a hostile hydrophobic environment for a water molecule. In the model with π-bulged helix S6_IV, the polar side chain of Asn⁴ⁱ²⁰ turned toward the DHPs molecule, creating a hydrophilic environment that is favorable for a water molecule at the agonist NO₂ group, but unfavorable for a hydrophobic group of the DHP antagonist. Thus, our results allow us to propose that DHP antagonists stabilize the structure with π-bulged S6_III and α-helical S6_IV, whereas DHP agonists stabilize structures where both S6_III and S6_IV are π-bulged. In view of electrophysiological data on DHP actions, our models suggest that structures with π-bulged S6_III and α-helical S6_IV would represent the closed/inactivated L-type calcium channel, whereas structures where both S6_III and S6_IV are π-bulged correspond to the open channels.

For our modeling, we used as templates available cryo-EM structures of eukaryotic sodium and calcium channels with various combinations of π-bulged and entirely α-helical segments S6. Dynamic appearance and disappearance of π-bulges in helices S6_III and S6_IV would dramatically reorient residues at the DHP site and in C-terminal parts of the helices where hydrophobic residues provide leaves for the activation gate. Notably, π-bulges are lacking in the inner helices of available structures of potassium channels, as well as homotetrameric prokaryotic sodium and calcium channels. Thus, the results of the present study are based on rather recent cryo-EM strictures of eukaryotic sodium and calcium channels. In the majority of these structures, the four S6 helices are characterized by the παπα pattern. All DHP-bound structures of the Cav1.1 channel also have this pattern.

Recent structures provide numerous examples where the binding of toxins and drugs affect π-bulging in Cav and Nav channels. Verapamil binding to the Cav1.1 channel induces π-bulging of helix S6_III, which is accompanied by tightening of the activation-gate region (Zhao et al., 2019a). In channel Cav3.1, a selective blocker Z944 binds in the central cavity with a part of the ligand extending into the fenestration between repeats II and III, and another, narrower ligand part bound like a plug in the inner pore above the activation gate. Upon binding of Z944, a helical turn in S6_II transits from the α-helical to π-helical conformation (Zhao et al., 2019b). In the Cav1.3 channel, S6 segments undergo structural shifts upon binding of cinnarizine (Yao et al., 2022). To accommodate cinnarizine, a helical turn in S6_III undergoes both π- to α-transition and an outward motion to avoid a steric clash. Interestingly, another helical turn where the gating residue is located undergoes the reverse α- to π-transition, which appears to tighten the activation gate. The binding of proteinaceous toxins HWTX-IV or ProTx-II to VSDs causes π-bulging in helix S6_IV and tightening of the activation gate. This transition also changes the binding site for the fast inactivation motif (Hite and MacKinnon, 2017). Cryo-EM structures of channel hNav1.7 were obtained with inhibitors that bind in the central cavity and block ion permeation but interact with different parts of the pore (Zhang et al., 2022). Blocker XEN907 binds in the pore and partially protrudes into the III–IV fenestration, causing π-bulging of helix S6_IV and tightening of the activation gate in the inactivated channel. Blocker TC-N1752 induces π-bulging of helix S6_II and closes the activation gate. Taken together, these examples clearly demonstrate that ligand binding can affect the pattern of π-bulging in S6 helices.

Most of the structural publications on eukaryotic sodium and calcium channels report tightening of the activation gate upon π-bulging of helices S6. This apparently conflicts with our conclusion that the binding of DHP agonists stabilizes the π-bulged conformation of helix S6_IV and thus the open state of the activation gate. However, the relationships between π-bulging of S6 helices and gating are likely complex and the experimental observations are rather ambiguous. For example, the pore opening upon π-bulging was proposed for TRP channels (Zubcevic and Lee, 2019). Tightening or opening of the activation gate upon π-bulging of the inner helices may depend on the driving force. In other words, π-bulging per se does not necessarily result in the activation gate opening or closure. Such local changes in S6s should not be directly interpreted as corresponding to gating rearrangements. The latter process is much more complex; it involves cooperative movements of voltage-sensor and S4–S5 linkers, which form a cuff, controlling the maximal diverging of S6 helices. Without the cuff widening, significant diverging of S6s is impossible. Maybe this is a reason why a local α–π transition in a single S6 does not cause the outward movement of its C-end but often causes tightening of the pore.

It should be mentioned that π-bulging, besides affecting the activation gate dimensions, may control the channel permeability by changing the hydrophilicity of the permeation pathway. Molecular modeling and MD simulations were performed to test possible open-state models of the NavMs prokaryotic channel (Choudhury et al., 2022). The experimental structure with α-helical S6s was consistently dehydrated at the activation gate, implying an inability to conduct ions. An alternative π-bulged model demonstrated the pore hydration and ion permeation consistent with an open channel state. Recent application of this approach to pseudo-tetrameric Nav channels (Choudhury and Delemotte, 2023) allowed us to propose a scheme in which fully open state of Nav requires π-bulging in both S6_IV and S6_III, in agreement with our hypothesis for Cav channels.

Previously, many mutational studies have been undertaken to reveal the DHP binding site and details of DHP-channel interactions (for reviews see Hockerman et al., 1997; Lacinová, 2005). In general, the DHP binding region between S5_III, P_III, S6_III, and S6_IV was determined. However, the exact role of residues in this region is not so clear, and in some cases it cannot be explained by direct interactions with DHPs. Particularly, mutations of I⁴ⁱ¹⁸, I⁴ⁱ¹⁹, and N⁴ⁱ²⁰ were reported to affect the DHP action. Assuming the α-helical structure of S6_IV, these residues would not directly interact with DHPs simultaneously. Our models suggest direct interactions of the DHP agonists with I⁴ⁱ¹⁹ and N⁴ⁱ²⁰ due to reorientations that are coupled with the α–π transition, but do not suggest direct involvement of I⁴ⁱ¹⁸. However, this residue is involved in strong intersegment contacts. In the entirely α-helical S6_IV, the side chain of I⁴ⁱ¹⁸ contacts F¹ⁱ¹⁶ and I^4o14. In the π-bulged S6_IV, the side chain contacts I^4o14 and L^4o13. Ile⁴ⁱ¹⁹ faces the DHP only in the entirely α-helical S6_IV. However, Ile⁴ⁱ¹⁹ is also involved in intersegment interactions. Besides DHPs, it interacts with F³ⁱ²². In the π-bulged S6_III, F³ⁱ²² turns toward S6_I and tightly interacts with F¹ⁱ¹⁶. Tyrosine residues in helices S6_III and S6_IV, which are located in the structurally conserved part of S6s upstream of the α–π transition region, demonstrate weak interactions with DHP molecules. However, these tyrosines are among the main stabilizers of S6 contacts with neighboring segments because their large sidechains participate in numerous interactions in the tightly packed channel region. Thus, the complexity of DHP–channel and intrachannel interactions and their changes associated with conformational rearrangements prevent straightforward interpretations of mutational data. Indeed, it is difficult to separate the direct effects of mutations on ligand–receptor interactions with the influence of the mutations on the DHP binding site conformation.

In conclusion, the possible mechanisms of action of DHP agonists and antagonists hypothesized here are consistent with experimental data we are familiar with. We suggest that the π-bulging of helix S6_IV, which is seen in several cryo-EM structures of sodium channels, is a dynamic process that may be sensitive to the membrane voltage, environment, methods used to prepare samples for cryo-EM studies, temperature, and other factors.

The data are available from the corresponding author upon reasonable request.

Eduardo Ríos served as editor.

Computations were performed using facilities provided by Compute Ontario (https://www.computeontario.ca) and the Digital Research Alliance of Canada (https://www.alliancecan.ca).

This work has no external grant support.

Author contributions: D.B. Tikhonov: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review & editing. B.S. Zhorov: Conceptualization, Investigation, Methodology, Resources, Software, Writing - review & editing.