Structural bases of inhibitory mechanism of CaV1.2 channel inhibitors

Architecture of the Ca_V1.2 channel complex

To investigate the inhibitory mechanism of Ca_V1.2-targeting drugs, we co-expressed the human Ca_V1.2 with its physiological auxiliary subunits α2δ1 and β2b in HEK293 cells. The genes encoding these three subunits were subcloned into pEG-BacMam vector. We also carried out whole-cell patch clamp using HEK293T cells to characterize the functional properties of these constructs (Supplementary Fig. 1a). The subunits expressed with these constructs form a functional complex, showing an activation curve similar to those from previous studies^31,32, and the half-activation voltage (V_1/2) of the recombinant channel is +1.0 (±0.5) mV. Subsequent purification experiments resulted in a sharp and symmetrical SEC profile, indicating homogeneity of the sample of the channel complex. The SDS-PAGE gel confirmed that the purified sample contained all three subunits (Supplementary Fig. 1b, c). Next, we collected cryo-EM data using Titan2 Krios and carried out single-particle cryo-EM analysis. The three-dimensional cryo-EM maps of Ca_V1.2 in the apo state (denoted as Ca_V1.2^apo), tetrandrine bound state (Ca_V1.2^TET), and benidipine bound state (Ca_V1.2^BEN) were determined at 3.5-Å, 3.4-Å, and 3.3-Å resolutions, respectively. These maps are rich in structural features, including densities for sidechains, N-glycans, and lipid molecules, and allowed us to reliably build atomic models of these Ca_V1.2 complexes (Supplementary Figs. 2–3) (Supplementary Table 1).

The structure of the Ca_V1.2 channel complex is composed of a pore-forming α1 subunit and auxiliary subunits α2δ1 and β2b (Fig. 1a, b). Its α1 subunit consists of four repeat transmembrane domains (D_I−D_IV). Each domain contains six transmembrane helices (S1−S6) and forms the channel in a domain-swapped fashion. The S1−S4 helices constitute the voltage sensing domain (VSD) to detect changes in the cross-membrane electrostatic potential_. Consistent with all known structures of voltage-gated channels, the S4 helix assumes a 3₁₀-helical conformation. The S5−S6 regions of all four domains form the pore domain to facilitate and control the passage of ions (Fig. 1c, d). The selectivity filter (SF) is contributed by the re-entrant loops from each subunit that connect S5 and S6 and include P1 and P2 helices (Fig. 1e). In contrast to the conformational heterogeneity of the α-interaction domain (AID) found in Ca_V1.1 (ref. ³³), AID is well resolved in the current Ca_V1.2 structure, with its N- and C-termini adjacent to the intracellular gate and VSD_II, respectively (Fig. 1a). The selectivity filter contains four acidic residues (E363^DI, E706^DII, E1135^DIII, and E1464^DIV; i.e. the EEEE motif), one from each domain, producing a strongly negatively charged zone to attract cations and determine selectivity for Ca²⁺ ions³³ (Fig. 1e). On the cytoplasmic side, the four S6 helices bundle together and act as a gate to control the access of ions into the pore (Fig. 1f). The pore diameter profile indicates that the current structure represents an inactivated state with a closed intracellular gate (Fig. 1d). Although all S4 helices in the four VSDs are in an ‘up’ conformation, the intracellular gate remains closed, suggesting that the structure of Ca_V1.2 was captured in an inactivated state. In addition, the α2δ1 subunit was clearly resolved and sits on the extracellular side of Ca_V1.2, interacting with the E149, D150, and D151 residues from the D_I repeat. The β2b subunit is composed of a GK domain and an SH3 domain and is associated with the cytosolic AID region of Ca_V1.2 (Fig. 1). A lipid molecule was observed in the D_I−D_II fenestration of Ca_V1.2^apo, consistent with other structure studies on Ca_V1.2 (refs. ^34,35,36).

The structure of Ca_V1.2^apo was compared with that of Ca_V1.1 (PDB ID: 5GJW)³³, giving rise to an RMSD of 1.35 Å for 1772 C_α-pairs. The overall structure, including VSD, SF, and extracellular loops (ECLs), is fairly superimposable (Supplementary Fig. 4a). In both structures, in the absence of applied membrane potential, all VSDs show an “up” conformation of the S4 helices and are thus in the “activated” state. However, some structural differences are observed between the structures of Ca_V1.2^apo and Ca_V1.1. For example, two π-bulges occur on the S6^DI and S6^DIII helices of Ca_V1.2^apo, whereas both helices assume a canonical α-helix conformation in Ca_V1.1. The sidechains of residues on the intracellular segment, starting from F394^S6I and F1175^S6III, undergo an approximate 90° rotation (Supplementary Fig. 4b). The residues forming the central cavity and the intracellular gate differs from that in Ca_V1.1. In specific, the side chain of F394 and F1175 face the central cavity, causing a shrinkage of the inner radius of the pore domain (Supplementary Fig. 4c, d). In Ca_V1.1, the intracellular gate is constituted by V329, L333, F656, A660, F1060, V1064 F1376, and I1380. With the local conformational shifts on the S6 helices in Ca_V1.2, the intracellular gate is formed by L401, S405, L749, V753, V1182, I1186, F1519, and I1523 (Supplementary Fig. 4e). A previous study on TRPV6 channel suggested that the transition from an α-helix to a π-helix may prompt channel opening³⁷. However, despite such transition being observed in Ca_V1.2, its intracellular gate remains closed (Supplementary Fig. 4f).

Pathogenic mutations of Ca_V1.2

Dysfunction of Ca_V1.2 usually causes heart and vessel diseases and neurological disorders^18,19,20. At least 86 mutations of the Ca_V1.2 α1 subunit have been identified in various cases of human diseases, including Timothy syndrome (TS), long QT syndrome 8 (LQTS), short QT syndrome (SQTS), Brugada syndrome (BrS), autism, and schizophrenia, among others (Supplementary data). Our current Ca_V1.2 structure provides a template to map these disease-related mutations into the 3D structure. Among these mutants, 39 sites are distributed on the four VSDs, pore domain, and C-terminal domain (CTD) (Fig. 2). The VSD_II and pore domain harbor most of the pathogenic mutations. The remaining 47 mutation sites are mainly distributed on the N-terminus, cytosolic long linkers between different domains (mainly D_II−D_III), and the C-terminal tail, and they were not determined due to conformational heterogeneity.

a–f Disease related mutations are located on D_I (b), D_II (c), D_III (d), D_IV (e), and CTD (f). D_I, D_II, D_III, D_IV, and CTD of α1 subunit are colored in deep green, light green, deep blue, mauve, and orange, respectively. Subunits α2, δ1, and β2b are colored in turquoise, pink, and magenta, respectively. Mutation sites are marked with spheres and colored in blue (gain of function), yellow (loss of function), and red (other), respectively.

Three gain-of-function (GOF) mutations have been identified and verified by electrophysiological experiments. G402S and G406R were found in Timothy syndrome cases, and they inhibit voltage-dependent inactivation of Ca_V1.2 (refs. ^18,38). Since all G406 substitutions to other amino acid residues, such as alanine, serine, and valine, impair the inactivation process, the function of G406 is proposed as irreplaceable³⁸. In our Ca_V1.2 structure, both G402 and G406 are located on the S6^DI helix and close to the intracellular gate (Supplementary Fig. 5b). We speculate that the absence of a sidechain in glycine at these two positions is essential for the conformational transition from the open state to the inactivated state, and replacement by a sidechain-containing residue at either of these two positions probably hinders the transition.

In a previous study, L762F associated with LQTS was found to result in a GOF defect with slower inactivation³⁹. In the current Ca_V1.2 structure, L762 on the S6^DII helix is located around the cytosolic membrane surface. Its sidechain points toward the AID motif and the S4-S5^DII linker helix (Supplementary Fig. 5c). We speculate that the mutation L762F results in steric hindrance with AID or S4-S5^DII helices, leading to a conformational change in these helices and affecting channel inactivation kinetics.

N300D is a loss-of-function (LOF) mutation and has been identified in Brugada syndrome patients⁴⁰. This mutation leads to a decrease in the current density by impairing the expression level of Ca_V1.2 in the membrane⁴⁰. N300 is located in the extracellular loop between S5^DI and S6^DI (Supplementary Fig. 5a) and is exposed to the interface between the α1 and α2δ1 subunits, which is essential for trafficking of the Ca_V1.2 complex^41,42. N300D may thus affect traffic by disturbing the interaction with α2δ1. In addition, E1135K is another LOF mutation identified in patients with idiopathic QT prolongation, bradycardia, and autism spectrum disorder^43,44. This mutation is positioned in the EEEE motif of the selectivity filter in the current Ca_V1.2 structure (Fig. 1e and Supplementary Fig. 5d). Electrophysiological experiments demonstrated that the E1135K mutation makes Ca_V1.2 a nonselective monovalent cation channel, in line with functional roles of the EEEE motif in determining ion selectivity^44,45.

Previous studies have indicated that R511Q, R518C, and R518H are associated with LQTS and cardiac only Timothy syndrome^46,47. In the current Ca_V1.2 structure, R511 and R518 are located at a cytosolic amphiphilic helix N-terminal to S1^DII (denoted as S0^DII). The S0^DII helix is rich in positive charged residues (⁵¹¹RFCRRKCRAAVK⁵²²) that electrostatically interact with the negatively charged C-terminus of AID (Supplementary Fig. 5e). Electrophysiological experiments demonstrated that pathogenic mutations associated with R511 and R518 result in slower inactivation^46,47. We speculate that these mutations interfere with the interaction between S0^DII and AID and thus affect the conformation of AID relative to the pore domain, resulting in reduced inactivation. This speculation is consistent with the notion that AID is important for regulating channel inactivation^48,49.

Recognition of tetrandrine

Tetrandrine is a traditional Chinese clinical agent for autoimmune disorders, cardiovascular diseases, and hypertension. It is extracted from the root of Stephania tetrandra S Moore and inhibits both L-type and T-type Ca_V channels^25,50,51. To explore the molecular basis how tetrandrine blocks the activity of Ca_V1.2, tetrandrine (10 μM) was added during protein expression and purification. We determined the tetrandrine-bound Ca_V1.2 complex (Ca_V1.2^TET) at 3.4-Å resolution (Supplementary Fig. 2). In our cryo-EM map, we found a well-resolved triangle-shaped density in the central cavity of the pore domain, which is well fitted with the tetrandrine molecule and positioned proximal to the selectivity filter (Fig. 3a–c). The two major moieties of the 6,7-dimethoxy-2-methyl-1-benzyl-isoquinoline ring are positioned close to the D_II−D_III fenestration site and D_IV, and they are stabilized by forming hydrophobic or Van Der Waals interactions with surrounding residues from the selectivity filter and S6 helices (Fig. 3d, e). In particular, the residues F1175^S6III, M1178^S6III, A1512^S6IV, I1516^S6IV, and F1519^S6IV play important roles in stabilizing tetrandrine. The 6,7-dimethoxy and 2/2’-N groups from the isoquinoline ring form hydrogen bonds with the N741^S6II and Y1508^S6IV/N1179^S6III residues. In addition, in agreement with observations in drug-bound Ca_V3.3 complex structures⁵², a lipid molecule penetrates through the fenestration site between D_I and D_II and participates in hydrophobic interactions with the tetrandrine molecule (Fig. 3d, e). Compared with other drug binding modes observed in the previously reported Ca_V1.1 structures, we found that the tetrandrine binding mode is distinct from those of nifedipine (a DHP derivative), verapamil (PAA), diltiazem (BTZ), and cinnarizine (CIN) (Supplementary Fig. 6a). In particular, tetrandrine completely squeezes the binding site of BTZ (Supplementary Fig. 6b), consistent with previous experimental results showing that tetrandrine and diltiazem cannot simultaneously bind to L-type Ca_V channels⁵³. To validate the binding site of tetrandrine, we designed two mutants, N741W and N1179W, with the potential to create steric clash between tetrandrine and the binding site. We determined the dose-response curves of tetrandrine for both wild-type and these two mutants. The curves for both mutants exhibited a rightward shift compared to that of wild-type Ca_V1.2, indicating that these mutants reduce the sensitivity of these mutants to tetrandrine. The derived IC₅₀ values for tetrandrine with the N741W and N1179W mutants are 71.3 nM and 41.8 nM, respectively, which are higher than that of wild-type (~14.2 nM) (Fig. 3), further confirming the binding site of tetrandrine.

a Chemical structure of tetrandrine. b The cryo-EM density shown in blue mesh for tetrandrine in sticks. c The overall structure of the Ca_V1.2^TET complex. The domains of Ca_V1.2^TET are colored as D_I in deep green, D_II in light green, D_III in deep blue, and D_IV in mauve. The tetrandrine located in the pore domain is presented as yellow spheres. d, e Detailed binding sites for tetrandrine showing interactions between tetrandrine and Ca_V1.2. The sidechains of key residues are displayed in sticks and the hydrophobic side chains are overlaid with transparent surfaces. Black dashed lines indicate potential hydrogen bonds. A lipid located within the fenestration site is shown in gray. f Dose-response curves of tetrandrine inhibition on Ca_V1.2^WT, Ca_V1.2^N741W, and Ca_V1.2^N1179W. Ca_V1.2^N741W and Ca_V1.2^N1179W shift the IC₅₀ of tetrandrine on Ca_V1.2 from 14.2 nM to 71.3 nM and 41.8 nM, respectively. Data in curves are represented as mean ± SEM (n = 10 biologically independent experiments). Source data are provided as a Source Data file.

Antagonism of benidipine

1,4-Dihydropyridine derivatives (DHPs), such as nifedipine, selectively inhibit certain voltage gated calcium channels and are widely used clinically for the treatment of hypertension⁵⁴. However, DHPs frequently cause side effects such as proteinuria, upon chronic use. The specific inhibition of L-type calcium channels is usually an advantage but is a defect in the kidney. The reason is that L-type channels are distributed in the kidney glomerulus afferent arteriole but are absent in the efferent arteriole. This inconsistent inhibition results in unbalanced local blood pressure, leading to kidney damage with proteinuria^55,56. However, benidipine, another DHP derivative, can additionally inhibit T-type calcium channels, which are more abundant in efferent arterioles than in afferent arterioles, making blood pressure more balanced in different parts of the kidney. Compared with nifedipine and amlodipine, benidipine causes fewer adverse side effects, such as proteinuria and leg edema^55,57,58; therefore, it is widely used in the clinic. Compared with nifedipine, benidipine replaces 8-nitrophenyl with 9-nitrophenyl and 3-carboxylic acid methyl ester with 3-(1-phenylmethyl-3-piperidinyl) ester. By supplementing benidipine when solubilizing the membrane and before preparing the cryo-EM sample, we obtained a 3.3-Å resolution benidipine-bound Ca_V1.2 structure (Ca_V1.2^BEN) (Supplementary Fig. 2).

The benidipine molecule penetrates into the D_III−D_IV fenestration and corresponds to a ‘claw’ shape density in the map (Fig. 4a, b). Unlike the pore-blocker tetrandrine, benidipine does not directly block the pore but allosterically inhibits Ca_V1.2. The main moiety of dihydropyridine is biased towards D_III (Fig. 4c). The nitrogen atom on the pyridine ring forms a potential hydrogen bond with S1132^P1 (Fig. 4d). The nitrogen atom on 9-nitrophenyl and the oxygen atom on 5-dicarboxylic acid of the pyridine ring form hydrogen bonds with T1056^S5III and Q1060^S5III, respectively (Fig. 4d). Whereas similar interactions with T935 and Q939 have been identified in the structure of the Ca_V1.1 complex, benidipine also contains a phenylmethyl piperidinyl group, which is positioned in a hydrophobic pocket formed by residues I1046^S5III, I1049^S5III, V1053^S5III, F1181^S6III, M1509^S6IV, and F1513^S6IV, providing additional hydrophobic interactions to stabilize its binding, in line with the inhibitory IC₅₀ of benidipine being approximately ten times smaller than that of nifedipine⁵⁹ (Fig. 4d, e). Previous studies indicated that nifedipine specifically inhibits L-type Ca_V channels, showing no sensitivity to N- and T-type Ca_V channels^60,61,62. In contrast, benidipine, in addition to L-type Ca_V channel, can also inhibit both N-type and T-type Ca_V channels, with IC₅₀ values of approximately 35 μM and 11 μM, respectively^59,63,64. To understand the molecular basis of how benidipine is recognized by Ca_V3.1, we compared the benidipine-bound Ca_V1.2 structure with that of Ca_V3.1, and the result illustrates that most of the residues participating in benidipine binding are also conserved in Ca_V3.1, such as V1053^S5III, F1129^DIIIP1, and F1513^S6IV. Although I1046^S5III, M1178^S6III, F1181^S6III, and M1509^S6IV in Ca_V1.2 are substituted by L1386^S5III, V1505^S6III, M1508^S6III, and L1813^S6IV in Ca_V3.1, respectively, we speculate that these substitutions by similar residues are tolerable and that the relatively conserved binding pocket underlies the basis for how benidipine acts as an inhibitor in multiple voltage gated calcium channels (Fig. 4f). Moreover, two residues, T1056 and Q1060, play pivotal roles in the binding of DHP drugs to L-type Ca_V channels. However, these residues are not conserved in N-type and T-type Ca_V. Residue T1056 of Ca_V1.2 is substituted by Y1289 in Ca_V2.2, and residue Q1060 is replaced by M1293 in Ca_V2.2 and F1400 in Ca_V3.1 (Supplementary Fig. 7a, b, Fig. 4f). Although benidipine does not inhibit mutations T1056Y, Q1060M, and Q1060F as potently as it does in WT Ca_V1.2, it still retains the ability to block the current (Supplementary Fig. 7c). In contrast, previous studies demonstrated that these same substitutions at positions T1056 and Q1060 almost abolished the sensitivity of Ca_V1.2 to the DHP drug R202-791 (ref. ⁶⁵). The varied effects of these mutations on benidipine and R202-791 might arise from the phenylmethyl piperidinyl group in benidipine. This group establishes more interactions with the channel and appears to tolerate substitutions at positions T1056 and Q1060, especially in N- and T-type Ca_V channels.

a Chemical structure of benidipine. b The cryo-EM density shown in blue mesh for benidipine in sticks. c The overall structure of the Ca_V1.2^BEN complex. The domains of Ca_V1.2^BEN are colored as D_I in deep green, D_II in light green, D_III in deep blue, and D_IV in mauve. The tetrandrine is presented as violet spheres. Two cation ions are shown as green spheres. Phospholipid entering through fenestration is shown as gray sticks. d Detailed binding sites for benidipine showing interactions between benidipine and Ca_V1.2. The sidechains of key residues are displayed in sticks and the hydrophobic side chains are overlaid with transparent surfaces. Black dashed lines indicate potential hydrogen bonds. e Comparison of the DHP ligands binding sites of Ca_V1.2 with nifedipine bound Ca_V1.1 structure (PDB ID: 6JP5) (colored in gray), overlaid with the electrostatic surface potential of Ca_V1.2. f Comparison of the benidipine binding sites of Ca_V1.2 with Ca_V3.1 (PDB ID: 6KZO) (colored in orange).

Several dihydropyridine drugs binding L-type calcium channels complex structures have been resolved in previous study^{21,22,66,67,68}, including amlodipine bound Ca_VAb (7KMD), nimodipine bound Ca_VAb (7KMF), nifedipine bound Ca_V1.1 (6JP5), amlodipine bound Ca_V1.1 (7JPX), (S)-Bay K 8644 bound Ca_V1.1 (6JP8) and (R)-Bay K 8644 bound Ca_V1.1 (7JPW). When Ca_V1.2^BEN and these structures are superimposed, a same binding pocket in DIII-DIV fenestration are revealed to accommodate benidipine in Ca_V1.2 and amlodipine/(S)-Bay K 8644/(R)-Bay K 8644 in Ca_V1.1 (Supplementary Fig. 8a–c). However, benidipine binding site do not overlap with binding sites of amlodipine and nimodipine in Ca_VAb (Supplementary Fig. 8d–f), and the residues involved in interaction are not conserved as well.

L-type selective dihydropyridine (DHP) inhibitors and L/T-selective DHP inhibitors are cataloged in Supplementary Table 2. The structural similarities among Class 1, 2, and 3 necessitate collective consideration. Upon scrutinizing their chemical structures and inhibitory selectivity, we identified primary distinctions in functional group modifications, specifically additional alterations on the C3 and C4 benzene rings. Within Class 1 and Class 2, with the exception of three instances (cilnidipine, amlodipine, nimodipine), all DHP derivatives predominantly featured smaller, relatively hydrophilic C3 groups for L-type selective compounds, while L/T-type selective compounds displayed larger and hydrophobic C3 groups. For Class 4 and Class 5 compounds, characterized by a modification into a ring at the C5/C6 position, the sole difference between the L-type selective Class 4-1 and L/T-type selective Class 4-4 compounds lies in the presence of a benzoic acid group on the C4 benzene ring in Class 4-1, while Class 4-4 has a hydroxyl group at the corresponding position. Thus, for compounds with C5-C6 rings, the hydrophobic nature of C3 modification does not exert a significant impact; the pivotal factor influencing differences is the presence of an additional aromatic group modification on the C4 benzene ring.

To further investigate interactions among compounds of varied selectivity on L-type and T-type channels, CaV1.2^BEN served as a docking template. L-type selective compounds from Class 1 were docked onto L-type channel Ca_V1.2, while L/T-type selective compounds from Class 1 were docked onto both L-type channel Ca_V1.2 and T-type channel Ca_V3.1. Irrespective of L-type or L/T-type selectivity, these compounds bound to the D_III-D_IV fenestration in Ca_V1.2 or Ca_V3.1, exhibiting a favorable superposition of the dihydropyridine backbone (Supplementary Fig. 9b, c). Both L-type selective and L/T-type selective DHPs docked to Ca_V1.2 formed hydrogen bonds with T1056, Q1060, and S1132. However, the C3 group of L-type selective DHPs docked to Ca_V1.2 is confined near M1509 within the hydrophobic pocket formed by D_IIIS6 and D_IVS6. In contrast, L/T-type selective DHPs docked to Ca_V3.1, while unable to form hydrogen bonds at the corresponding positions, not only occupy the binding pocket formed by D_IIIS6 and D_IVS6 but also extend to the hydrophobic pocket formed by D_IIIS5 and D_IVS5, akin to the position occupied by L/T-type selective DHPs docked to Ca_V1.2. This suggests that the ability of the C3 group to occupy the binding pocket formed by D_IIIS5 and D_IVS5 may be pivotal in determining the L-type and L/T-type selectivity of DHPs. In the case of cilnidipine, its C3 benzenyl group occupies the C5 position in the Ca_V1.2 binding pocket, possibly due to the planar rigidity of the phenyl group preventing it from fitting into the C3 position. The excessively large C5 group may clash with F1400 in Ca_V3.1, elucidating why cilnidipine, despite its large and hydrophobic C3 group, cannot inhibit T-type channels. In the case of the other exception, amlodipine docked in Ca_V3.1, the C2 amino group extends towards the pore and forms potential interactions with hydrophilic residues K1462 and Q1868 near the pore, compensating for the modest interaction between the C3 group and Ca_V3.1. Given that L/T-type selective DHPs also exhibit different affinities for various T-type channels, such as nimodipine’s preference for inhibiting Ca_V3.2 and Ca_V3.3 over Ca_V3.1 (ref. ⁶⁴), we aligned the Ca_V3.3 structure with the Ca_V3.1 structure. The alignment results revealed that the hydrophobic pockets formed by D_IIIS5 and D_IVS5 in Ca_V3.3 are smaller than those formed in Ca_V3.1, potentially explaining why nimodipine produces distinct inhibitory effects on T-type channels with different α1 subunits.