Blog

By Yanli Dong, Yiwei Gao, Shuai Xu, Yuhang Wang, Zhuoya Yu, Yue Li, Bin Li, Tian Yuan, Bei Yang, Xuejun Cai Zhang, Daohua Jiang, Zhuo Huang, and Yan Zhao

Excerpt from the article published in Cell Reports, Volume 37, Issue 5, 2021, 109931, ISSN 2211-1247 https://doi.org/10.1016/j.celrep.2021.109931

Editor’s Highlights

• N-type voltage-gated calcium (Ca_V) channels mediate Ca²⁺ influx at presynaptic terminals in response to action potentials and play vital roles in synaptogenesis, and release of neurotransmitters.
• The Ca_V2.2 is a High-voltage activated (HVA) channel that is exclusively expressed in central and peripheral neurons.
• The structure of the Ca_V2.2 complex contains α1, α2δ1, and β1 subunits.
• The α2δ1 subunit is divided into α2 and δ subunits, which are linked by a disulfide bond and associate with the α subunit by interacting with extracellular loops (ECL).
• The α2δ1 subunits in both Ca_V2.2 and Ca_V1.1 complexes display similar binding geometry.
• The α2δ1 exhibits a little tilt relative to the remaining complex because the ECL2 and ECL3 are directly packed with the α2δ1 subunit.

Authors’ Highlights

• W768 on W-helix is a structural determinant for CSI of Ca_V2.2 channel
• Voltage-sensing domain II was trapped in the resting state by a PIP₂molecule
• Ziconotide is located on the top of the selectivity filter blocking ion influx
• Small molecules, PD173212 and Ca_V2.2 blocker 1, occupy the D_III–D_IVfenestration site

Summary

N-type voltage-gated calcium (Ca_V) channels mediate Ca²⁺ influx at presynaptic terminals in response to action potentials and play vital roles in synaptogenesis, release of neurotransmitters, and nociceptive transmission. Here, we elucidate a cryo-electron microscopy (cryo-EM) structure of the human Ca_V2.2 complex in apo, ziconotide-bound, and two Ca_V2.2-specific pore blockers-bound states. The second voltage-sensing domain (VSD) is captured in a resting-state conformation, trapped by a phosphatidylinositol 4,5-bisphosphate (PIP₂) molecule, which is distinct from the other three VSDs of Ca_V2.2, as well as activated VSDs observed in previous structures of Ca_V channels. This structure reveals the molecular basis for the unique inactivation process of Ca_V2.2 channels, in which the intracellular gate formed by S6 helices is closed and a W-helix from the domain II–III linker stabilizes closed-state inactivation. The structures of this inactivated, drug-bound complex lay a solid foundation for developing new state-dependent blockers for treatment of chronic pain.

Graphical abstract

Introduction

Voltage-gated calcium (Ca_V) channels are essential mediators to convert action potentials (APs) into an influx of Ca²⁺ ions, a crucial secondary messenger to regulate a variety of types of cellular event, such as muscle contraction, secretion of neurotransmitters, cell division, differentiation, and apoptosis (Catterall, 1991; Catterall and Few, 2008; Reuter, 1979; Tsien et al., 1988). Ca_V channels are generally categorized into two groups according to their activation threshold, i.e., high-voltage activated (HVA) and low-voltage activated (LVA) Ca_V channels. Based on their sequence homology, Ca_V channels in mammals contain 10 members, which are further classified into three subfamilies (Ca_V1, Ca_V2, and Ca_V3) that conduct six types of Ca²⁺ current (L-, P/Q-, N-, R-, T-types) (Catterall et al., 2005; Nowycky et al., 1985; Snutch and Reiner, 1992). Ca_V2.2 is a HVA channel that is exclusively expressed in central and peripheral neurons (Olivera et al., 1994). It is predominantly located in the presynaptic terminals and mediates the Ca²⁺ influx that triggers neurotransmitter release at the fast synapses (Westenbroek et al., 1992, 1998). Dysfunctions of Ca_V2.2 channels alter neuronal functions and lead to diseases, such as myoclonus-dystonia-like syndrome (Weiss, 2015). Moreover, the Ca_V2.2 channel plays a critical role in spinal nociceptive signaling, and thus it has become an important drug target for chronic pain treatment (Patel et al., 2018).

Molecular mechanisms of the Ca_V channels have been studied extensively for decades, including recent structural studies of L-type Ca_V1.1 isolated from rabbit skeletal muscle (Gao and Yan, 2021; Wu et al., 2015, 2016; Zhao et al., 2019a) and human T-type Ca_V3.1 (Zhao et al., 2019b). These studies revealed structural features of the Ca_V channels in their inactivated state with all the voltage-sensing domains (VSDs; VSD_I–VSD_IV) adopting an activated “up” conformation. The structures of the Ca_V1.1 complex also elucidate the assembly of Ca_V channel with auxiliary β and α2δ subunits. A range of ligands was determined in complexes with Ca_V1.1 or Ca_V3.1, and these structural studies provide molecular bases for ligand recognition and facilitate further rational drug development targeting Ca_V channels. Despite these advances in structural studies on the Ca_V channels, more structural insight is desirable to fully understand the molecular mechanism of the Ca_V2.2 channel from the structural untouched Ca_V2 subfamily, because it shares only low sequence identity with the available structures from the other two subfamilies, and some fundamental mechanistic questions remain to be addressed at the structural level. For example, although the available structures of Ca_V channels have activated VSDs (Wu et al., 2015, 2016; Zhao et al., 2019a, 2019b), structures of these VSDs in their resting state are required to fully understand the gating mechanism. Moreover, Ca_V2.2 inactivates more rapidly than the closely related Ca_V2.1 isoform, exhibits a distinct closed-state inactivation (CSI) mechanism during repolarization (Jones et al., 1999; Patil et al., 1998; Thaler et al., 2004), and is modulated by phosphatidylinositol 4,5-bisphosphate (PIP₂) molecules (Hille et al., 2015; Keum et al., 2014; Rodríguez-Menchaca et al., 2012b; Suh et al., 2012; Vivas et al., 2013; Wu et al., 2002). Further structural analysis is required to understand the underlying molecular mechanisms for these unique functional properties of Ca_V2.2, which are essential to its key physiological roles in neurotransmitter release and regulation of synaptic transmission.

Here, we purified the recombinant human Ca_V2.2 (also called the α1 subunit) in complex with the β1 and α2δ1 subunits and determined their complex structure using the single-particle cryo-electron microscopy (cryo-EM) method. This structure reveals the asymmetric activation of the four VSDs with the VSD_II in its resting state, evidently stabilized by binding of a PIP₂ lipid molecule. In addition, we identify a unique α helix that locks the intracellular gate of Ca_V2.2 in closed conformation, contributing to the CSI of Ca_V2.2. Specific binding of a state-dependent channel blocker in the pore provides a template for future development of next-generation, nonaddictive analgesic drugs. These results provide insights into the activation and inactivation mechanisms of human Ca_V channels and point the way to future drug discovery. During the preparation of this manuscript, the structure of human Ca_V2.2 complex was published (Gao et al., 2021). Despite high similarity between the Ca_V2.2 complex structures, we provide more mechanistic interpretation about CSI of the Ca_V2.2 validated by electrophysiological experiment and also elucidate a binding site for small-molecule inhibitors targeting Ca_V2.2.

Results and Discussion

Structure determination and architecture of the Ca_V2.2 complex

To investigate the architecture of the N-type Ca_V2.2 complex, we co-expressed human N-terminal truncated Ca_V2.2, full-length wild-type (WT) α2δ1, and full-length WT β1 in HEK293 cells. To monitor the expression of the complex during purification, we fused Ca_V2.2 with a C-terminal GFP-Twinstrep tag. In order to enhance the expression level, we truncated the intrinsically disordered N-terminal region (residues 1–64), and we denote this construct as Ca_V2.2^EM. We carried out whole-cell patch-clamp experiments in HEK293T cells to characterize the channel properties of both human WT full-length Ca_V2.2 and Ca_V2.2^EM constructs in the presence of auxiliary β1 and α2δ1 subunits. The Ca_V2.2^EM complex shows indistinguishable gating properties compared with the full-length Ca_V2.2, in terms of voltage-dependent activation and steady-state inactivation (Figures 1A, 1B, and S1A–S1C). Thus, the Ca_V2.2^EM construct was subjected to further structural and functional studies. The purified Ca_V2.2^EM-α2δ1-β1 complex (Ca_V2.2 complex) displaying monodisperse peaks on both size-exclusion chromatography (SEC) and SDS-PAGE further confirmed that all three subunits were present in the purified complex sample (Figures S1D and S1E).

To elucidate its architecture, we carried out cryo-EM study of the Ca_V2.2 complex and obtained the final reconstruction at 2.8-Å resolution for the Ca_V2.2 complex (Figures 1C and S1F–S1J; Table S1). Cryo-EM map of the Ca_V2.2 is rich in structural features, including densities for side chains, N-glycans, disulfide bonds, and associated lipid molecules, which enabled us to unambiguously build de novoatomic models of the Ca_V2.2 complex (Figures 1D and S1J). Within its overall size of approximately 118 Å × 113 Å × 169 Å, the structure of the Ca_V2.2 complex contains α1, α2δ1, and β1 subunits and closely resembles the classic shape of Ca_V1.1.

The α subunit is composed of four transmembrane domains, D_I–D_IV, which form an ion conducting pore in a domain-swapped fashion. Each domain is composed of six helices (S1–S6), of which S1−S4 helices form the VSD. Four VSDs (VSD_I–VSD_IV) encircle the central pore formed by S5 and S6 helices from all four domains. The β1 subunit comprises a Src homology 3 (SH3) domain and a guanylate kinase (GK) domain. The GK domain, through its alpha-interacting domain (AID) helix, interacts with the Ca_V2.2 (i.e., the α subunit). Nevertheless, the GK domain was not well resolved in our cryo-EM map, presumably because of conformational heterogeneity (Figures 1C and 1D). The α2δ1 subunit is divided into α2 and δ subunits, which are linked by a disulfide bond (C404-C1059) and associate with the α subunit by interacting with extracellular loops (ECLs; in particular the loops between S5 and S6 helices) ECL_I, ECL_II, and ECL_III, and the S1−S2 loop from VSD_I. The structure of human α2δ1 in our complex is nearly identical with rabbit α2δ1 resolved in the Ca_V1.1 complex, with root-mean-square deviation (RMSD) of ∼1.1 Å.

Ion conduction pore of the Ca_V2.2 complex

The central pore is made up of helices S5 and S6 from all four domains of the α subunit (Figures 2A and 2B). In each domain, a re-entrant P loop is located between the S5 and S6 helices and contains two short helices, P1 and P2, connected by a short linker. Four P loops form a funnel-like shape and line the outer entry to the pore. Four highly conserved glutamate residues, E³¹⁴, E⁶⁶³, E¹³⁶⁵, and E¹⁶⁵⁵, form a selectivity-filter ring at the bottom of the funnel, which is one of the narrowest segments of the central pore (Figure 2C) (Smart et al., 1996). These four acidic residues, together with surrounding negatively charged residues, create a strong negative electric field strength to attract cations and repel anions. This pore region also functions as a selectivity filter to discriminate Ca²⁺ from other cation ions. We did identify a strong density within the selectivity filter and close to the glutamate cluster (Figures 2D and 2E). Presumably, this piece of density represents a calcium ion, consistent with observations in previous structures of Ca_V channels (Gao and Yan, 2021; Wu et al., 2015, 2016; Zhao et al., 2019a, 2019b). Thus, on the extracellular side, the S5 and S6 helices and the P loops contribute to the formation and stability of the selectivity filter.

On the intracellular side, the pore-lining S6 helices comprise an intracellular gate to control ion permeation (Figure 2F). The Ca_V2.2 complex was determined in the absence of membrane potential and thus is likely to represent a depolarized state. The cytoplasmic ends of the S6 helices have converged to form a hydrophobic seal through a cluster of hydrophobic residues (Figure 2F). In each of the four domains, the S5 helix is positioned proximal to the S6 helix and connects to S4 of VSD through an amphipathic horizontal helix (termed S4–S5 helix, ∼16 residues). This S4–S5 helix interacts with the S6 helix and thus couples S4 movement in response to depolarization with pore opening.

The central pore of the Ca_V2.2 channel is further compared with those from the Ca_V1.1 (PDB: 5GJW) and Ca_V3.1 (PDB: 6KZP) channels. The S5–S6 ECLs exhibit significant conformational difference among these structures (Figures S2A, S2B, S2D, and S2E). The Ca_V2.2 α subunit possesses shorter ECL_I, ECL_II, and ECL_III, but this structural difference does not hamper binding of the α2δ1 subunit. The S1–S2 loops of VSD_II, also involved in interactions with the α2δ1 subunit, are nearly identical in both Ca_V2.2 and Ca_V1.1. Therefore, the α2δ1 subunits in both Ca_V2.2 and Ca_V1.1 complexes display similar binding geometry. In contrast, the ECLs of the Ca_V3.1 channel exhibit remarkable structural differences from the corresponding loops of Ca_V1.1 and Ca_V2.2, thus giving rise to incompatibility of the α2δ1 subunit with Ca_V3.1 (Figures S2B and S2E). Despite these structural differences occurring at ECLs, the transmembrane helices S5 and S6, as well as the selectivity filter of Ca_V2.2, are superimposable with Ca_V1.1 and Ca_V3.1, yielding RMSD of ∼1.8 Å for Ca_V1.1 (for 332 C_α atom pairs) and ∼1.6 Å for Ca_V3.1 (for 300 C_α atom pairs), respectively (Figures S2C and S2F). In particular, the intracellular gate formed by the four-helix bundle of S6 helices is nearly identical in all three structures, assuming a closed form (Figure S2G). The S6_II helix of the Ca_V2.2 extends into the cytoplasm and is longer than that of the Ca_V1.1. Consequently, the C terminus of the S6_II helix forms a close contact with the β1 subunit in the Ca_V2.2 complex, which is not observed in the structure of Ca_V1.1. Furthermore, the cytoplasmic part of the S6 helix bends by 13° compared with that in Ca_V1.1 (Figure S2H), which may affect gate opening and represent a potential regulatory mechanism of the β subunit.

Voltage sensor of the N-type Ca_V channel

A hallmark feature of the Ca_V channel is to open or close the channel in response to changes in membrane potential. Surrounding the central pore, the four VSDs play essential roles in converting the electrostatic signal into a conformational change of the intracellular gate. Each VSD is composed of S1–S4 transmembrane helices. Each S4 helix contains several positively charged residues, either arginine or lysine, spaced at intervals of three. Ca_V2.2 is a depolarization-activated channel. Upon depolarization of the membrane potential, the gating charges on the S4 helix move toward the extracellular cavity of the VSD. The two cavities are separated by a highly conserved hydrophobic constriction site. When the gating charges pass through this hydrophobic seal, existing interactions of the gating charges with hydrophilic and negatively charged residues on surrounding helices are disrupted on one side, and new interactions are formed on the other side. The displacement of the S4 helices across the electrostatic field of the membrane potential would induce lateral movement of the S4–S5 linkers, resulting in disengagement between S5 and S6 helices and thus the channel opening.

In our structure, the S4 helices adopt a 3₁₀-helix conformation, which is consistent with observations from previous structures of voltage-gated ion channel, including Ca_V and Na_V channels, allowing the side chains of gating-charge residues to be aligned on the same side of the helix surface. Using the central pore as a reference, superimposition of the structure of Ca_V2.2 onto that of Ca_V1.1 shows close overlap of VSD_I, VSD_III, and VSD_IV, demonstrating that they are stabilized in the same presumably activated state (Figures S3A–S3E). In contrast, VSD_II of Ca_V2.2 exhibits a strikingly different conformational arrangement compared with that from the Ca_V1.1 structure. Taking a closer look at the four VSDs, we found that four, four, and three gating charges are located at the extracellular aqueous cavities of VSD_I, VSD_III, and VSD_IV, respectively, and interact with polar residues residing on nearby S1–S3 helices, consistent with observations from the available structures of Ca_V channels (Wu et al., 2015, 2016; Zhao et al., 2019a, 2019b) (Figures 3A, 3C, 3D, and S3C–S3E). Strikingly, in VSD_II, only one arginine (R578) on the S4 helix is located on the extracellular side of the hydrophobic seal, but the other four conserved gating charges (R581, R584, K587, and K590) are located in the intracellular cavity, indicating that VSD_II is stabilized in the resting state. Among these basic residues, K590 is located at a position where S4 becomes unwound and is completely exposed to the solvent (Figure 3B). The asymmetric activation of Ca_V2.2 VSDs suggests that each individual VSD may sense membrane potential asynchronously, which is reported to be important for eukaryotic Na_V activation and inactivation (Chanda and Bezanilla, 2002).

To investigate the conformational change of the VSD_II of Ca_V2.2 upon depolarization, we overlaid its structure in the resting state (VSD_II^R) onto that of Ca_V1.1 in the activated state (VSD_II^A) and found that these VSD_II structures are highly superimposable in helices S1–S3 (Figures 3E and 3F). In contrast, the S4 helix slides to the intracellular side in the resting state. Consequently, the extracellular half of the S3 helix bends ∼15° toward S4 in the resting state (Figure 3E). Using the S5 and S6 helices as a superimposition reference, the VSD_II undergoes substantial conformational change relative to the central pore between these two states. In particular, in transition from the activated state to the resting state, the cytoplasmic sides of both S1 and S2 helices in Ca_V2.2 rotate toward the central pore (Figure 3G). Consequently, the intracellular and extracellular termini of the S3 helices are remarkably displaced, moving toward the central pore by 4 and 9 Å, respectively. In addition to sliding inward, the S4 helices also slightly shift toward the pore domain (Figure 3H). Interestingly, even though distinguishable conformational change occurs in VSD_II, the extracellular end of the S1–S2 helix hairpin forms similar interactions with helices S5 and P1 from domain III. It appears that this interaction is stabilized by a cholesterol derivative, cholesteryl hemisuccinate (CHS), which is clearly resolved in both our structure of Ca_V2.2 and the structure of Ca_V3.1 (Figure S3F), consistent with earlier reports suggesting that cholesterol is important to regulating activity of Ca_V channels (Purcell et al., 2011; Xia et al., 2008).

Compared with S4 in the activated VSD_II from Ca_V1.1, the S4 helix in the resting state undergoes a sliding movement by two helical turns, ∼13 Å, toward intracellular side (Figure 3H). The displacement of S4 is comparable with previous observations in structures of TPC1 channel and Na_VAb channel (Figures S3G and S3H) (Guo et al., 2016; Kintzer and Stroud, 2016; Wisedchaisri et al., 2019). With the exception of R5 (K590) of Ca_V2.2, all gating charges are facing to one side, and no rotation about the helix axis is observed, which support a sliding-helix model of voltage-dependent gating (Wisedchaisri et al., 2019; Zhang et al., 2018). As a consequence of the S4 sliding inward, the S4–S5 linker is tilted by 11° toward the intracellular side, and this movement generates more contacts between the S4–S5 linker and S6 helix and thus stabilizes the inner gate in its closed state.

An extra density was identified laying above the S4–S5 linker of domain II in the cryo-EM map, which presumably represents a lipid with two hydrophobic tails and a “palm”-like head group. Its head group is located at the intracellular cavity of the VSD_II and flanked by S3 and S4 helices (Figure 3I). Its hydrophobic tails project toward the extracellular side of the membrane (Figure S3I). This lipid appears to act as a plug and would block the S4 helix from sliding upward, reminiscent of the PIP₂that attenuates opening of the K_V1.2 channel by interacting with the S4–S5 linker (Rodriguez-Menchaca et al., 2012a). We attempted to fit a PIP₂ into the density. Whereas its hydrophobic tails and inositol ring agree well with the density, the two phosphate groups attached to the inositol ring could not be well resolved (Figure S3I). Considering that PIP₂ is predominant in the inner leaflet of the plasma membrane and reported to positively shift the activation curve of Ca_V2.2, which are consistent with our structural observations, we postulate that this lipid molecule is a PIP₂. In a closer look at the PIP₂ lipid binding site, several positively charged residues from helices S0, S4, and S4–S5 linker provide a highly positively charged environment for PIP₂ binding (Figures S3J and S3K). Similar local environment is not observed in the other three VSDs (Figure S3J), potentially providing a mechanism by which PIP₂ selectively binds to VSD_II. In addition, the inositol ring of PIP₂ is positioned proximal to the AID helix of the β1 subunit, and this observation is consistent with previous investigations indicating that regulation of the Ca_V channel by PIP₂ depends on type of β subunit present (Suh et al., 2012).

CSI mechanism of the Ca_V2.2 channels

The N-type Ca_V channel bears a preferential voltage-dependent CSI (Jones et al., 1999; Patil et al., 1998), featuring Ca²⁺ insensitive U-shaped inactivation curve; substantial inactivation accumulated during interpulse in a two-pulse protocol and cumulative inactivation during voltage-clamped AP trains (Patil et al., 1998). In our structure, the intracellular gate is comparable with that in structures of the Ca_V1.1 and Ca_V3.1, and it appears to be stabilized at a closed state (Figure S2G). Strikingly, an additional α helix was unexpectedly revealed in our structure lying underneath the intracellular gate and flanked by the four S6-helix bundle, forming a 36° angle to the membrane plane (Figure 4A). This structural element has not been observed previously in Ca_V channels. The high-resolution cryo-EM map allowed us to align the amino acid sequence with the model with high confidence. This helix is composed of amino acid residues A764 to S783, which are located in the D_II-IIIlinker (Figure 4B). The side chain of W768 points upward into the intracellular gate, forms extensive hydrophobic interactions with residues from S6 helices (Figures 4B and 4C), and thus stabilizes the intracellular gate in its closed state. Thus, we term this six-turn helix as the W-helix. In addition to W768 serving as a plug inserting into and blocking the intracellular gate, the other residues from the W-helix extensively interact with the S6 helices, including hydrogen bond and electrostatic interactions, to further strengthen the association of the W-helix with the intracellular gate (Figures 4B and 4C). We speculate that the newly discovered W-helix functions as a blocking lid, similar to the “hinged lid” model for fast sodium channel inactivation, and is important for the stability of the intracellular gate in the closed/inactivated state of Ca_V2.2.

To investigate the functional role of the W-helix and to validate our hypothesis, we designed two mutants, including a deletion of the W-helix (Ca_V2.2^ΔW-Helix) and a point mutation, W768Q (Ca_V2.2^W768Q), and we performed whole-cell patch-clamp recording experiments with Ba²⁺ as charge carrier. The voltage activation curves of these two mutants were indistinguishable from the WT Ca_V2.2 complex, suggesting the W-helix does not participate in channel activation. In sharp contrast, the steady-state inactivation curves of these two variants exhibit a positive shift of ∼24 mV (Figures 4D, S4A, and S4B), indicating the W-helix and W768 are critical for the CSI. To investigate recovery from CSI, the membrane potential of pre-pulse was held at –40 mV, where the channel has not opened yet. In this protocol, recovery of the Ca_V2.2^ΔW-Helix variant is significantly faster than that of the WT complex (Figures 4E and S4C), which further supports the conclusion that the W-helix plays a vital role in CSI of the Ca_V2.2 channel. These results are also in line with a previous functional characterization showing that the domain II–III linker is crucial for CSI of Ca_V2.2 (Kaneko et al., 2002; Thaler et al., 2004).

To further validate our finding in a more physiologically relevant condition, we repetitively activated the Ca_V2.2 channels with AP trains that were recorded in hippocampal CA1 neurons in whole-cell current-clamp configuration. As observed in previous reports (Patil et al., 1998; Thaler et al., 2004), we found that WT Ca_V2.2 channels exhibited cumulative inactivation in response to AP trains (Figures 4F and S4D). Interestingly, Ca_V2.2^ΔW-Helix variant inactivated more slowly, suggesting that W-helix-regulated CSI contributes to Ca_V2.2 channel inactivation during the AP trains and may play an important role in short-term synaptic plasticity (Catterall et al., 2013). Collectively, our results show that the W-helix is essential for CSI of the Ca_V2.2 channel. Interestingly, the W-helix is conserved only in the Ca_V2 subfamily (Figure S4E), suggesting that Ca_V2.1 and Ca_V2.3 channels may have similar inactivation mechanisms.

Recognition and specificity of ziconotide to the Ca_V2.2 channel

The ziconotide derived from ω-conotoxin MVIIA is able to potently inhibit the Ca_V2.2 with high selectivity and used to treat severe chronic pain. To gain insight into how the ziconotide blocks the Ca_V2.2 complex, we determined the ziconotide-bound Ca_V2.2 complex at 3.0-Å resolution (Ca_V2.2^MVIIA) (Figures S5A and S5D). The ziconotide is positioned right above the selectivity filter and directly blocks the outer entry of calcium, in line with previous speculation that the ziconotide acts as a channel blocker (Figure 5A). The ziconotide is composed of 25 amino acids with six cysteine residues. A serial of structural analysis by NMR spectroscopy revealed that the ziconotide is compact folded, primarily defined by three disulfide bonds, including C1-C16, C8-C20, and C15-C25 (Kohno et al., 1995), which also present in ziconotide from the structure of Cav2.2^MVIIA complex (Figure 5B). Interestingly, upon association with the Ca_V2.2 channel, the main chain and side chains of the ziconotide undergoes substantial conformational changes compared with the NMR structure (PDB ID: 1OMG), suggested by RMSD of ∼1.4 Å for 100 main-chain atom pairs and ∼2.8 Å for 178 side-chain included atom pairs (Figure S6A). We speculate that such conformational change is essential for ziconotide to associate with Ca_V2.2 and thus block channel activity. Taking a close look at the contacts between the ziconotide and Ca_V2.2, we found interactions are formed in electrostatic and shape complementary manners. The structural elements of P1–P2 helices and ECLs play important roles to stabilize the ziconotide (Figures 5C–5E). In particular, the R10 of the ziconotide (R10^Z) is posed proximal to the selectivity-filter ring of glutamate and is important for the blockage of the ion pathway (Figure 5C). The residues K4^Z, R21^Z, and R24^Z assume extensive electrostatic interactions with D1627 and D1629 from ECL4. In an earlier report, substitution of K2^Z or Y13^Z by Ala results in a remarkable reduction in activity (Kim et al., 1995). We found that the K2^Z contacts with ECL2 and ECL3 by electrostatic interaction with E640, E1314, and E1330 in our structure. The hydroxyl group of Y13^Z engages in hydrogen bonds with Q658 and D664, and the benzene ring of Y13^Z, together with L11^Z and M12^Z, forms extensive van der Waals interaction with surrounding side chain and backbone from the Ca_V2.2 channel (Figure 5E).

We also compared structure of the Cav2.2^MVIIA complex with apo-form Ca_V2.2 complex. It turns out the Ca_V2.2 undergoes only subtle conformational change upon the ziconotide association. The majority of the model, including VSDs and helices S5–S6, are fairly superimposable between these two structures. The ECLs, including ECL1, ECL2, and ECL3, surrounding the ziconotide are slightly expanded to enlarge the ziconotide binding cavity (Figures S6B and S6C). Consequently, the α2δ1 exhibits a little tilt relative to the remaining complex, because the ECL2 and ECL3 are directly packed with the α2δ1 subunit (Figure S6D). To assess the binding specificity of the ziconotide, the structures of L-type Ca_V1.1 and T-type Ca_V3.1 are superimposed on the structure of Cav2.2^MVIIA complex. We found the ECL1 of Ca_V1.1 and Ca_V3.1 is longer than that in Ca_V2.2, pointing to the ziconotide binding cavity and resulting in direct steric clashes with the ziconotide in Ca_V2.2 (Figures S6E and S6F), suggesting the ECLs are critical for selectivity of the ziconotide.

Blockage of Ca_V2.2 by PD173212

The PD173212 is a potent and selective channel blocker of Ca_V2.2, consisting of phenyl, tyrosinamide, and 4-tertbutylbenzyl groups, which has been reported to be efficacious in the audiogenic seizure mouse model (Hu et al., 1999) (Figure 6A). We determined the structure of the Ca_V2.2 in complex with the PD173212 at 3.0-Å resolution (Figures S5B and S5E), which clearly shows two connected strip-shaped densities located in the central cavity underneath the selectivity filter and nearly parallel to the membrane plane, which are compatible with the PD173212 structure (Figures 6B–6D). Taking a close look at the PD173212 binding site, the interaction is primarily mediated by extensive hydrophobic interaction and van der Waals interactions. The tert-butyl and isobutyl groups of tyrosinamide project into the central cavity and are stabilized by interactions with hydrophobic residues from surrounding helices, such as F345, L352, V1412, L1700, F1703, etc. These two groups are positioned right underneath the selectivity filter, implying their important role in blocking ion influx. The 4-tertbutylbenzyl group occupies the D_III-D_IVfenestration, and its benzene ring is stabilized by forming a π-π interaction with the surrounding aromatic residues (Figure 6E). The phenyl group also contributes to stabilize the C-terminal 4-tertbutylbenzyl group by a π-π stacking. Upon PD173212 association, the conformation of the pore module and majority of side chains involved in the ligand binding are unaltered, except the three aromatic residues, including Y1289, F1411, and F1693, which are around the D_III-D_IV fenestration site (Figure 6F). The side chains of all three residues undergo discernable conformational change and thus contribute to stabilize the phenyl group of the PD173212. Strikingly, two of the three conformationally changed residues, Y1289 and F1693, are substituted by T935-M1366 and C1396-L1813 in Ca_V1.1 and Ca_V3.1, respectively (Figures S7C and S7D). Considering the side fenestration is critical for association of pore blocker with channel (Gamal El-Din et al., 2018), these substitutions suggest a possible mechanism through which PD173212 could selectively inhibit Ca_V2.2. In contrast with PD173212, the pore blockers diltiazemand verapamil for Ca_V1.1 bind in the middle of the central cavity, the pore blocker Z944 for Ca_V3.1 extends into the D_II-D_III fenestration, and the binding position of PD173212 extends toward the D_III-D_IV fenestration (Figure S7B), providing a site for rational drug design to minimize off-target side effects.

Blockage of Ca_V2.2 by Ca_V2.2 blocker 1

A novel and potent pyrazolyltetrahydropyran N-type calcium channel blocker, termed as Ca_V2.2 blocker 1, has been identified recently, and oral administration of Ca_V2.2 blocker 1 in a rat model showed promising efficacy in reducing inflammatory and neuropathic pain (Wall et al., 2018). The Ca_V2.2 blocker 1 consisted of 4-ketopyran, pyrazole, and two modified benzene groups (Figure 7A). Our cryo-EM map clearly defined the binding site of the Ca_V2.2 blocker 1 within the Ca_V2.2 channel (Figures 7B, S5C, and S5F). Its 4-ketopyran group points to the central cavity stabilized by direct contacts with adjacent hydrophobic residues and thus blocks the ion pathway. One of the benzene groups is extended to the D_III-D_IVfenestration and stabilized by extensive hydrophobic interactions (Figure 7C). Addition of the methyl group enlarges the size of the benzene group, consequently resulting in close pack with residues from S5_III (Figures 7D and 7E), supporting that this methyl group is critical for potency of the Ca_V2.2 blocker 1 (Wall et al., 2018). Interestingly, although the chemical structures of the PD173212 and Ca_V2.2 blocker 1 are different, they share an overlapped binding site close to the D_III-D_IVfenestration, suggesting this fenestration site could be a hotspot for future drug designment (Figure S7E). Similar to the PD173212, association of the Ca_V2.2 blocker 1 does not induce significant conformational change of the Ca_V2.2 channel, and only the side chain of the F1693 exhibits dramatic conformational change (Figure 7F), suggesting that F1693 acts as a “door,” a critical residue for recognizing and stabilizing N-type Ca_V channel blocker.

Concluding remarks

Here, we elucidated the structures of the human Ca_V2.2 complex in apo, ziconotide-bound, PD173212-bound, and Ca_V2.2 blocker 1-bound states. These structures show the VSD_II is trapped at resting state by a PIP₂ molecule, featuring only one gating charge located at the extracellular cavity. A W-helix is determined underneath the closed intracellular gate formed by S6 helices, with the W768 inserting into and stabilizing the closed gate. Our electrophysiological experiments demonstrated that the W768 and W-helix are fundamentally important for the CSI of Ca_V2.2. Moreover, the structure of the ziconotide-bound complex illustrated that ziconotide is located above the selectivity filter and acts as channel blocker to inhibit Ca_V2.2. By comparing with structures of the Ca_V1.1 and Ca_V3.1, we conclude that the ECLs are structural determinants for the ziconotide to specifically bind with the Ca_V2.2 channel. Furthermore, the complex structures of Ca_V2.2 with small molecular ligands PD173212 and Ca_V2.2 blocker 1 reveal the D_III-D_IV fenestration site is critical for pore blocker binding and could be a crucial position for further designment of a Ca_V2.2-specific drug to treat chronic pain.