Structure of human Ca_v2.2 channel blocked by the pain killer ziconotide

Abstract

The neuronal-type (N-type) voltage-gated calcium (Ca_v) channels, which are designated Ca_v2.2, have an important role in the release of neurotransmitters^1–3. Ziconotide is a Ca_v2.2-specific peptide pore blocker that has been clinically used for treating intractable pain^4–6. Here we present cryo-electron microscopy structures of human Ca_v2.2 (comprising the core α1 and the ancillary α2δ−1 and β3 subunits) in the presence or absence of ziconotide. Ziconotide is thoroughly coordinated by helices P1 and P2, which support the selectivity filter, and the extracellular loops (ECLs) in repeats II, III and IV of α1. To accommodate ziconotide, the ECL of repeat III and α2δ−1 have to tilt upward concertedly. Three of the voltage-sensing domains (VSDs) are in a depolarized state, whereas the VSD of repeat II exhibits a down conformation that is stabilized by Ca_v2-unique intracellular segments and a phosphatidylinositol 4,5-bisphosphate molecule. Our studies reveal the molecular basis for Ca_v2.2-specific pore blocking by ziconotide and establish the framework for investigating electromechanical coupling in Ca_v channels.

Ca_v2.2 channels are also known as the N-type Ca_v channels for their role in transmitter release in the central and peripheral nervous systems. Combining different splice forms of the voltage-dependent Ca²⁺ conducting core subunit α1 with auxiliary subunits—including the extracellular α2δ and the cytosolic β subunits—yields various Ca_v2.2 heteromers that exhibit distinct membrane distribution and biophysical properties^1,2,7,8. Ca_v2.2 channels in the primary afferent terminals are involved in pain signalling. Suppression of Ca_v2.2 activity thus represents a strategy for the development of analgesic agents^9–11.

Ca_v2.2 can be potently and selectively inhibited by a number of peptide toxins, some of which have been exploited for pharmacological applications^3,12–14. Among these, the ω-conotoxin MVIIA was developed to produce ziconotide (also known as SNX-111 and Prialt), which has been approved by the US Food and Drug Administration for the treatment of severe pain^4–6. However, its intrathecal application and adverse effects have limited the broad use of ziconotide^15,16. High-resolution structures of ziconotide-bound Ca_v2.2 are required to elucidate the molecular basis for the subtype-specific inhibition and to facilitate drug discovery.

Cryo-electron microscopy (cryo-EM) structures of rabbit Ca_v1.1 and human Ca_v3.1 alone and in complex with various modulators have previously been published^17–21. Despite substantial sequence variations, the structures of Ca_v1.1 and Ca_v3.1 are similar in their transmembrane regions, which comprise four homologous repeats (designated I, II, III and IV; throughout, subscript Roman numerals indicate location on the respective homologous repeat) each. The transmembrane segments 5 and 6 (S5 and S6) from the four repeats enclose the central pore domain that is surrounded by the four VSDs, each of which is constituted by S1 to S4 (Supplementary Fig. 1). In contrast to the conserved transmembrane domain, the long linkers between the repeats (the I–II linker and the II–III linker) and the C-terminal segments vary substantially both in length and in the primary sequence. These cytosolic segments remain unresolved in the structures of Ca_v and closely related Na_v channels.

The gating of the pore domain is coupled to the movements of VSDs in response to membrane potential changes, in a mechanism that is known as the electromechanical coupling. Despite minor deviations, all of the previously reported structures of Ca_v1.1 and Ca_v3.1 channels exhibit similar ‘inactivated’ states, in which the pore domain is closed and all four VSDs are in the ‘up’ conformations^18–21. Investigating electromechanical coupling necessitates structural resolution of the channels in distinct conformations.

Cryo-EM analysis of the Ca_v2.2 complex

Details of protein production and cryo-EM analysis of human Ca_v2.2 in the absence (apo) and presence of purchased ziconotide (determined at 3.1 and 3.0 Å resolution, respectively (Extended Data Fig. 1)) are provided in the Methods. The 3D electron microscopy reconstructions of apo Ca_v2.2 and Ca_v2.2–ziconotide are nearly identical, and show only limited structural deviations. Unless indicated otherwise, the structural analysis discussed below is based on Ca_v2.2–ziconotide because of its higher resolution.

We could reliably assign side chains to α1, α2δ−1, ziconotide and the α1-interacting guanylate kinase domain of β3 (Extended Data Figs. 2–4, Extended Data Table 1, Supplementary Video 1). The distal domain of β3 and the C-terminal domain of α1 were docked as rigid bodies. In addition to the protein components, one Ca²⁺ ion is coordinated by the signature EEEE motif²² (Glu314, Glu663, Glu1365 and Glu1655) at the outer site of the selectivity filter (Extended Data Fig. 3c). Sixteen sugar moieties were assigned to eight glycosylated Asn residues (seven on α2δ−1 and one on α1). Among the resolved cholesterol, phospholipid and detergent molecules that surround α1, a phosphatidylinositol 4,5-bisphosphate (PIP₂) molecule binds to the interface of VSD_II and the pore domain in the inner leaflet (Fig. 1a, Extended Data Fig. 4).

Fig. 1 |. Specific pore blockade of Ca_v2.2 by ziconotide.

a, Overall structure of the Ca_v2.2–ziconotide complex at an averaged resolution of 3.0 Å. CTD, C-terminal domain; Zi, ziconotide. The resolved lipid, cholesterol and cholesterol hemisuccinate molecules are shown as black sticks. Sugar moieties are omitted for visual clarity. All structure figures were prepared in PyMol with the same colour scheme. b, Ziconotide is caged by the ECLs. The sequence of ziconotide is shown above, with the three disulfide bonds indicated. The surface electrostatic potential, shown in semi-transparent presentation, was calculated in PyMol. c, Specific coordination of ziconotide by α1 of Ca_v2.2. The residues that are not conserved in Ca_v channels are labelled blue (in all Figures). The bound Ca²⁺ is shown as green sphere, and the EEEE motif (Glu314, Glu663, Glu1365 and Glu1655) that determines Ca²⁺ selectivity is shown as thin sticks. Letters in parentheses denote backbone groups, N for amide and O for carbonyl oxygen (in all Figures) d, ECL_III and α2δ−1 concertedly move upward to accommodate ziconotide. When the structures of the Ca_v2.2–ziconotide (coloured according to domain) and apo Ca_v2.2 (blue) are superimposed relative to α1, the only deviation (indicated by orange arrows) occurs in ECL_III and α2δ−1. Inset, upward shift of ECL_III to avoid clash with ziconotide. The distances between the indicated residues are in Å.

The overall structures of apo Ca_v2.2 and rat Ca_v1.1 can be superimposed with a root mean square deviation of 1.10 Å over 1,728 Cα atoms. The α2δ−1 subunit and its interface with the extracellular segments of α1 remain identical, despite minor structural differences in ECL_I and ECL_III (Extended Data Fig. 5). On the cytosolic side, the helix after S6_I bends to be parallel with the membrane plane and interacts with β3 through its α1-interacting domain (AID) motif²³. We refer to this transverse helix as the AID helix. These common structures—which have been illustrated in detail for Ca_v1.1¹⁸—are not discussed further here: here we focus on ziconotide recognition and repeat II, which exhibits several features that are distinct from known structures of eukaryotic Ca_v and Na_v channels, for detailed analysis.

Pore blockade of Ca_v2.2 by ziconotide

Ziconotide has previously been reported to specifically block Ca_v2.2 with a half-maximal inhibitory concentration of 0.7–1.8 nM^24,25. Our structure reveals that the blocker, which contains three disulfide bonds, is nestled in the electronegative cavity that surrounds the entrance to the selectivity filter, and that specific recognition is mediated mainly by charged and polar residues on the P1 and P2 helices and the ECLs in repeats II, III and IV. Although ECL_I is also in the vicinity of ziconotide, none of the repeat I residues is directly engaged in ziconotide coordination (Fig. 1b, c, Extended Data Fig. 6a).

The binding pose of ziconotide is reminiscent of that of μ-conotoxin KIIIA (KIIIA) in its complex structure with Na_v1.2, in which Lys7 of KIIIA points to the outer mouth of the selectivity filter²⁶ (Extended Data Fig. 6b). Ziconotide does not possess an equivalent residue to directly seal the entrance to the vestibule of the selectivity filter. Instead, it blocks ion entrance by neutralizing the outer electronegativity and sterically hindering the ion access path to the entrance of the selectivity filter (Fig. 1b). To neutralize the acidic residues, ziconotide engages Arg10 and Tyr13 to bind to Asp664, which marks the beginning of P2_II. Ser19 interacts with Glu1659, which is on the first helical turn of P2_IV. Thr17 interacts with Asp1345 on ECLIII, and Arg21 and Lys4 interact with Asp1628 and Asp1629, respectively, on ECLIV (Fig. 1c).

Our structure-guided sequence comparison showed that four of the eight ziconotide-coordinating residues in Ca_v2.2 (Thr643, Asp1345, Lys1372 and Asp1629) are not conserved in other Ca_v channels (Fig. 1c, Supplementary Fig. 1), which explains the subtype specificity of pore blockade by ziconotide. Consistent with our structure, Y13A or R10A substitutions have previously been shown to substantially reduce pore blocking by ziconotide, and Y13R abolished its activity ^24,27,28.

The only structural change upon ziconotide binding is a slight tilt of α2δ−1 and ECL_III of α1 as a rigid body. The upward movement of ECL_III is necessary to avoid a clash with ziconotide (Fig. 1d, Extended Data Fig. 6c, d). The α2δ−1 subunit has previously been shown to reduce ziconotide affinity²⁹, which may result from the energy penalty for lifting ECL_III and α2δ−1 to accommodate ziconotide.

Cytosolic segments unique to Ca_v2

On the cytosolic side, S6_II is extraordinarily long and contains seven additional helical turns (residues 714 to 739), which we designate the cytosolic segment of S6_II (S6_IIC). The ensuing sequence (residues 740 to 786) forms two additional helices (which we designate cytosolic helix (CH)1_II and CH2_II) that fold back towards the membrane. CH1_II is anti-parallel with S6_IIC, whereas CH2_II is perpendicular to it (Fig. 2a). S6_IIC is not conserved among the three Ca_v families and the sequence corresponding to CH1_II and CH2_II is missing in Ca_v1 and Ca_v3 families, which affords an explanation for the lack of these structural elements in Ca_v1.1 and Ca_v3.1^18,20 (Supplementary Fig. 1).

Fig. 2 |. Cytosolic segments unique to Ca_v2 in the II–III linker.

a, An extended cytosolic segment of S6_II is followed by CH1 and CH2. The sequence for S6_IIC (residues 714–742) is not conserved among the Ca_v family and the ensuing helices are missing in Ca_v1 and Ca_v3 channels. Supplementary Fig. 1 shows the sequence alignment. b, β3 is sandwiched between the AID helix and S6_IIC. Inset, detailed interactions between α1 and β3. Polar interactions are indicated by red dashed lines. c, CH2_II simultaneously interacts with all four S6 segments below the intracellular gate. A tilted side view of the S6 tetrahelical bundle is shown. d, S6_III is three helical turns shorter in Ca_v2.2 than in Ca_v1.1. Unwinding the last three turns of S6_III yields space to accommodate CH2_II. Ca_v1.1 (Protein Data Bank (PDB) code 5GJW) is coloured in wheat. e, CH2_II tightens the closed intracellular gate. Several layers of hydrophobic residues on the S6 tetrahelical bundle intertwine to close the intracellular gate. Trp768, a Ca_v2-specific residue, interacts with several gating residues.

The U-turn between S6_IIC and CH1_II mediates weak interaction with β3, in addition to the well-characterized primary interface between the AID and the guanylate kinase domain of β3^30–32 (Fig. 2b). Glu740, which is conserved in Ca_v2 family only and marks the end of S6_II, may be hydrogen-bonded to His348 on the α9 helix of β3. Glu743, which is positioned at the tip of the U-turn, may interact with the amide of Glu111 in the SH3 domain of β3 (Fig. 2b inset). With this additional interface, β3 is sandwiched between AID and S6_II.

CH1_II has only limited contact with S6_IIC, whereas CH2_II interacts with all four S6 segments (Fig. 2c). As the sequence corresponding to CH2_II is not found in Ca_v1 and Ca_v3 channels, this structural feature is probably unique to Ca_v2 channels (Supplementary Fig. 1). To accommodate CH2_II, the last three helical turns in S6_III are unwound, resulting in a shorter S6_III than that in Ca_v1.1. The N-terminal residues of CH2_II are placed immediately beneath the closed intracellular gate (Fig. 2d, e)

The intracellular gate of Ca_v2.2 comprises a large number of residues that intertwine to form a hydrophobic structural core. At the bottom of this exceptionally thick hydrophobic gate is Trp768 on CH2_II, which is surrounded by Ala360, Ala713, Ile1417 and Phe1711 from the four S6 segments (Fig. 2e insets). By pulling all four S6 segments together and directly engaging Trp768 to secure the gate, CH2_II may facilitate the closing of the pore domain.

VSD_II in a ‘down’ conformation

The four VSDs of Ca_v2.2 carry different numbers of gating charge residues: five in VSD_I (R1 to R4, and K5; to make the numbering of the gating charge residues consistent, we designate the residue on the first helical turn of the S4 segment as 1), VSD_II (R2, R3, R4, K5 and K6) and VSD_IV (R2, R3, R4, K5 and R6), and six in VSD_III (K1, R2, R3, R4, K5 and K6) (Fig. 3a). K5 in VSD_I, K5 and K6 in VSD_III and K5 and R6 in VSD_IV are below, and the other gating charge residues are above, the occluding Phe on S2 (a state that we define as up). By contrast, in VSD_II R3, R4, K5 and K6 are all below the occluding Phe525, and only R2 is above it (a state that we define as down) (Fig. 3a). When the four VSDs are superimposed, S4_II slides down by about 11–12 Å from the other S4 segments, and the S1, S2 and S3 segments are relatively well superimposed (Fig. 3b).

Fig. 3 |. VSD_II in a down state.

a, Structure of the four VSDs. In each VSD, the gating charge residues (gating charge numbering in parentheses) on S4 and the surrounding residues that may facilitate gating charge transfer are shown as sticks. An1, An2, conserved acidic or polar residues on S2. The gating charge residues above and below the occluding Phe on S2 are labelled dark cyan and brown, respectively. b, Distinct conformation of VSD_II. When the four VSDs are superimposed, S4_II slides down by about 12 Å compared to the S4 segments in other VSDs. c, Conformational changes of the α1 subunit in Ca_v1.1 and Ca_v2.2. A cytosolic view of the superimposed α1 structures is shown; the CTD is omitted for clarity. The major structural shifts from Ca_v1.1 to Ca_v2.2 are indicated by blue arrows. Right, structural comparison of the diagonal repeats II and IV. d, Coupled shifts of VSD_II and S6_II. Left, the downward motion of S4_II is accompanied by the rotation of VSD_II and the displacement of S4–5_II and the nearby S6_IIC. Right, pronounced shifts of VSD_II between Ca_v1.1 and Ca_v2.2 in the context of the overall α1 structure. The distances shown on the left and right indicate the linear displacement of the Cα atoms and side chains of the corresponding gating charge residues, respectively. Supplementary Video 2 shows the conformational shifts.

In VSD_I, VSD_III and VSD_IV, the one or two extracellular helical turns of S4 are in the form of an α-helix, and the remaining segments conform to a 3₁₀ helix. In VSD_II, the entire S4 is a 3₁₀ helix, and R2, R3, R4 and K5 all line the same side. K6 (Lys590), which is at the tip of a sharp and short turn that immediately precedes the S4–5_II segment, is on the opposite side to the other four gating charge residues and its side chain projects into the cytosol (Fig. 3a, b). Compared to the deactivated structures of toxin-bound VSD_IV in the Na_vPaS-1.7 chimera³³ and rat Na_v1.5³⁴, S4_II in Ca_v2.2 is lower by one more helical turn (Extended Data Fig. 7a).

Structural comparison of the α1 subunits in Ca_v2.2 and Ca_v1.1 shows that the pore domain, VSD_I and VSD_IV remain nearly identical. Major shifts occur in repeats II and III (Fig. 3c, Extended Data Fig. 7b). VSD_III undergoes a slight clockwise rigid-body rotation from Ca_v1.1 to Ca_v2.2 in the intracellular view. The motion of VSD_II involves both domain rotation and intradomain rearrangement (Fig. 3c, d, Extended Data Fig. 7c, d, Supplementary Video 2). The S1, S2 and S3 segments in VSD_II of Ca_v1.1 and Ca_v2.2 remain nearly identical in domain comparison, whereas S4 is dislodged by about 12 Å towards the intracellular side from Ca_v1.1 to Ca_v2.2 through a spiral sliding, leading to the transfer of two gating charge residues (R3 and R4) across the occluding Phe on S2 (Extended Data Fig. 7d). The most marked shift occurs to K6, the amine group of which is displaced by 25 Å through both secondary structure transition and side-chain swing (Fig. 3d, Supplementary Video 2).

Accompanying the rotation of VSD_II and the marked shift of S4_II from Ca_v1.1 to Ca_v2.2, the ensuing N terminus of S4–5_II is pushed down by about 3 Å. The motion of S4–5_II drives the adjacent S6_IIC to shift accordingly, establishing a potential transmission path from VSD_II motion to pore gating (Fig. 3d, Extended Data Fig. 8, Supplementary Video 2). As only VSD_II is in the down conformation and the other VSDs are up, we refrain from assigning a functional state to the present Ca_v2.2 structure. As, to our knowledge, a down structure of a VSD has not previously been captured in the absence of mutagenesis or modulators, we attempt to identify the structural elements that stabilize the deactivated state of VSD_II.

Stabilization of the down VSD_II by PIP₂

The intracellular segments that are unique to Ca_v2 contribute to the stabilization of the down conformation of VSD_II. S6_IIC, which interacts extensively with S4–5_II, contacts the AID helix and CH2_II through several polar residues (Fig. 4a). The AID helix binds S0, S2 and S3 in VSD_II, preventing rotation of these segments. Simultaneous association of S0_II, S2–3_II, S4–5_II and S6_IIC with the straight AID helix may restrain VSD_II from rotation, therefore stabilizing the present conformation. More than half of the residues in the extensive interaction network are Ca_v2-specific, which provides a potential explanation for the distinct VSD_II conformations in Ca_v2.2 (Fig. 4a, b).

Fig. 4 |. The down conformation of VSD_II is stabilized by several intracellular segments and a bound PIP₂.

a, Cytosolic segments that are unique to Ca_v2 stabilize the down VSD_II. Inset, a network of extensive polar interactions among the AID helix, S4–5_II, S6_II and CH2_II. b, The AID helix is an organizing centre for segments within and near VSD_II. The straight AID helix, in addition to mediating the channel modulation by the β subunits, may serve as a lever that couples the motion of VSD_II to S6_I and S6_II. c, The bound PIP₂ favours a down conformation of VSD_II. Left, VSD_II and the ensuing S4–5_II in the up conformation as in Ca_v1.1 (wheat) would clash with PIP₂ in its current binding pose. Right, a cytosolic view of PIP₂ coordination by polar residues in Ca_v2.2 (coloured according to domain).

The bound PIP₂ also appears to favour a down VSD_II, as its binding pose is incompatible with the up conformation of Ca_v1.1 VSD_II. The head group of PIP₂ wedges into a cytosolic cavity in VSD_II through the interface of S3 and S4, and the tails are coordinated by hydrophobic residues on S3 toS6 in repeat II and S5 and S6 in repeat III (Fig. 4c, Extended Data Fig. 9a, b). The PIP₂-binding site is reminiscent of that in KCNQ1³⁵, but is at a higher position (Extended Data Fig. 9c). The 5-phosphate group of PIP₂ is coordinated by R4 (Arg584) and K5 (Lys587) on S4_II. The head group is anchored by Ser544 (which marks the beginning of S3_II), the amide group of Phe546 and Arg596 on S4–5_II (Fig. 4c, Extended Data Fig. 9a). When VSD_II moves upward, its interface with the neighbouring S5_III is also rearranged (Extended Data Fig. 9d). Consequently, PIP₂, the displacement of which may be restrained by the hydrophobic membrane, can no longer retain these polar and hydrophobic interactions, but clashes with the lifted S4–5_II (Fig. 4c, Extended Data Fig. 9a). Therefore, the bound PIP₂ may stabilize the down conformation of VSD_II.

Discussion

The cryo-EM structures of the human Ca_v2.2 complex reveal the molecular basis for the specific pore blockade of Ca_v2.2 by ziconotide and provide a mechanistic interpretation for lowered ziconotide affinity in the presence of α2δ−1 (Fig. 1d). The down conformation of VSD_II, which is stabilized by previously unresolved cytosolic segments and a PIP₂ molecule, provides important insights into the electromechanical coupling in Ca_v channels.

The AID helix, which follows S6_I through a tight turn and interacts with S6_II, is sandwiched by VSD_II and β3 (Fig. 4b). It may thus serve as a lever to couple the conformational changes of VSD_II with S6_I, S6_II and the intracellular segments. The β subunits may also modulate channel activity in part through the interaction with AID.

PIP₂ has previously been shown to both cause a right shift of the activation voltage and reduce the current rundown of Ca_v2.1³⁶. The PIP₂-binding site observed in this study may account for the voltage-dependent inhibitory modulation as it stabilizes the down conformation of VSD_II. It remains to be investigated whether a separate PIP₂-binding site is responsible for the voltage-independent rundown reduction, a mechanism that may involve G_βγ proteins³⁷. Structural elucidation of representative members in the three Ca_v families lay the foundation for future structural and mechanistic investigation of the electromechanical coupling and the regulation of Ca_v channels by a variety of modulators^3,8.

METHODS

No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.

Transient expression of the human Ca_v2.2 complex in HEK293F cells

Codon-optimized cDNAs of CACNA1B for full-length Ca_v2.2 α1 (2,339 residues, Uniprot Q00975–1), CACNA2D1 for α2δ−1 (1,103 residues, Uniprot P54289–1) and CACNB3 for β3 (484 residues, Uniprot P54284–1) were synthesized (BGI Geneland Scientific) and subcloned into the pCAG vector. An amino-terminal Flag tag and a carboxy-terminal His₁₀ tag were added to each subunit in Ca_v2.2–ziconotide and untagged α2δ−1 was used for the apo Ca_v2.2 complex. DNA sequences were examined in SnapGene. HEK293F suspension cells (Thermo Fisher Scientific, R79007) were cultured in Freestyle 293 medium (Thermo Fisher Scientific) at 37 °C, supplied with 5% CO₂ under 60% humidity. When cell density reached 1.5–2.0 × 10⁶ cells per ml, a mixture of expression plasmids and polyethylenimine (Polysciences) was added into cell culture to initiate the transient expression of human Ca_v2.2 following a standard transfection protocol. No further authentication was performed for the commercially available cell line. Mycoplasma contamination was not tested.

Protein purification of human Ca_v2.2 and complex preparation with ziconotide

Approximately 72 h after transfection, the HEK293F cells were collected by centrifugation at 3,600g for 10 min and resuspended in the lysis buffer containing 25 mM HEPES (pH 7.4), 150 mM NaCl and the protease inhibitor cocktail containing 2.6 μg ml⁻¹ aprotinin and 1.4 μg ml⁻¹ pepstatin. The suspension was supplemented with glyco-diosgenin (GDN) (Anatrace) to a final concentration of 1% (w/v), n-dodecyl-β-d-maltopyranoside (DDM, Anatrace) to 0.2% (w/v), and cholesteryl hemisuccinate Tris salt (CHS) (Anatrace) to 0.04% (w/v). After incubation at 4 °C overnight, the mixture was centrifuged at 35,000g for 30 min, and the supernatant was applied to anti-Flag M2 affinity resin (Sigma). The resin was rinsed with wash buffer (buffer W) containing 25 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM CaCl₂ and 0.01% GDN. Eluted with buffer W plus 0.2 mg ml⁻¹ Flag peptide (synthesized by GenScript), the eluent was concentrated using a 100-kDa cut-off Amicon (Millipore) and further purified through size-exclusion chromatography (Superose 6 10/300 GL, GE Healthcare) that was preequilibrated in buffer W. The peak fractions were pooled and concentrated to a concentration of about 20 mg ml⁻¹ with α2δ−1 in excess. For structural determination of Ca_v2.2–ziconotide, purchased ziconotide (Alomone labs) was added to the concentrated protein solution at a final concentration of 2 mg ml⁻¹. The mixture was incubated at 4 °C for 30 min before making cryo-grids.

Cryo-EM sample preparation and data collection

Aliquots of 3.5 μl concentrated apo Ca_v2.2 or Ca_v2.2–ziconotide were loaded onto glow-discharged holey carbon grids (Quantifoil Cu R1.2/1.3, 300 mesh for Ca_v2.2–ziconotide, Quantifoil Au R1.2/1.3, 300 mesh for the apo channel), which were blotted for 6 s and plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (Thermo Fisher) at 8 °C with 100% humidity. Grids were transferred to a Titan Krios electron microscope (Thermo Fisher) operating at 300 kV and equipped with a Gatan Gif Quantum energy filter (slit width 20 eV) and spherical aberration (Cs) image corrector. Micrographs were recorded using a K2 Summit counting camera (Gatan) in super-resolution mode with a nominal magnification of 105,000×, resulting in a calibrated pixel size of 0.557 Å. Each stack of 32 frames was exposed for 5.6 s, with an exposure time of 0.175 s per frame. The total dose for each stack was about 50 e⁻ per Ų. SerialEM was used for fully automated data collection³⁸. All 32 frames in each stack were aligned, summed and dose-weighted using MotionCorr2³⁹ and twofold-binned to a pixel size of 1.114 Å per pixel. The defocus values were set from −1.9 to −2.1 μm and estimated by Gctf⁴⁰.

Image processing

A total of 3,384 and 1,317 cryo-EM micrographs were collected for Ca_v2.2–ziconotide and the apo channel, respectively, and 1,909,156 (Ca_v2.2–ziconotide) and 928,665 (apo Ca_v2.2) particles were auto-picked by RELION-3.0. Particle picking was performed using 2D classes of rabbit Ca_v1.1 (EMD-22426) in the side and tilted views as reference. All subsequent 2D and 3D classification and refinement was performed with RELION-3.0⁴¹. Two rounds of reference-free 2D classification were performed to remove ice spots, contaminants and aggregates, yielding 1,101,746 and 807,595 particles for Ca_v2.2–ziconotide and apo Ca_v2.2, respectively. The particles were processed with a global search with K = 1 to determine the initial orientation alignment parameters using bin2 particles. The electron microscopy map of rabbit Ca_v1.1 (EMD-22426), low-pass-filtered to 20 Å, was used as an initial reference²¹. The output of the 35–40 iterations was further applied to local angular search 3D classification with four classes. A total of 305,200 and 159,079 particles (for Ca_v2.2–ziconotide and apo Ca_v2.2, respectively) were selected by combining the good classes of the local angular search 3D classification. The particles were then re-extracted using a box size of 280 and pixel size of 1.114 Å. These particles yielded reconstructions at 3.2 Å and 3.3 Å (for Ca_v2.2–ziconotide and apo Ca_v2.2, respectively) after 3D auto-refinement with an adapted mask. Multi-reference 3D classification using bin1 particles after Bayesian polishing resulted in a reconstruction at 3.0 Å from 170,839 particles for Ca_v2.2–ziconotide and 3.1 Å from 56,616 particles for the apo channel. Focused refinement was performed using a core mask for β3 and α1 S6_II.

All 2D classification, 3D classification and 3D auto-refinement was performed with RELION 3.0. Resolutions were estimated using the gold-standard Fourier shell correlation 0.143 criterion with high-resolution noise substitution^42,43.

Model building and refinement

Ca_v2.2, together with the other two Ca_v2 subtypes (Ca_v2.1 and Ca_v2.3) and the four Ca_v1 subtypes (Ca_v1.1 to Ca_v1.4), are high-voltage-activated, in contrast to the low-voltage-gated Ca_v3.1, Ca_v3.2 and Ca_v3.3 channels⁴⁴ (Supplementary Fig. 1). Model building of Ca_v2.2 was thus based on the structure of the Ca_v1.1 channel complex. The starting model of Ca_v2.2 α1 subunit and α2δ−1 were built in SWISS-MODEL⁴⁵ based on the structure of rabbit Ca_v1.1 (PDB code 5GJW), and that of β3 was based on the crystal structure of rat β3 (PDB code 1VYU). The starting models of Ca_v2.2 and ω-conotoxin MVIIA (PDB code 1OMG) were then manually docked into the 3.0 Å toxin-bound electron microscopy map in Chimera⁴⁶. The model was manually adjusted in COOT⁴⁷, followed by refinement against the corresponding maps by phenix.real_space_refine program in PHENIX⁴⁸ with secondary structure and geometry restraints. For model building of the apo channel, coordinates for Ca_v2.2 α1, α2δ−1 and β3 were docked into the 3.1 Å apo Ca_v2.2 map separately and manually adjusted in COOT.

The excellent map quality supports reliable assignment of 2,622 side chains in both structures. Similar to the structures of other eukaryotic Ca_v and Na_v channels^{18,20,49–56}, most of the cytosolic segments in the α1 subunit are invisible, including residues 1–85 in the N terminus, residues 407–463 in the I–II linker, residues 787–1138 in the II–III linker, and residues 1838–2339 in the C terminus.

Statistics of the map reconstruction and model refinement can be found in Extended Data Table 1. All structure figures were prepared in PyMol⁵⁷.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Extended Data

Extended Data Fig. 1 |. Cryo-EM analysis of the human Ca_v2.2 complex alone and in the presence of ziconotide.

a, Representative electron micrograph and 30 classes of 2D class averages for Ca_v2.2–ziconotide. The green circles indicate particles in distinct orientations. The box size for the 2D averages is 312 Å. Scale bar, 50 nm. Left, a half view of one micrograph out of 3,384 in total for Ca_v2.2–ziconotide. b, Workflow for electron microscopy data processing (Methods). c, The gold-standard Fourier shell correlation (FSC) curves for the 3D reconstructions. The graph was prepared in GraphPad Prism. d, FSC curves of the refined model versus the summed map that it was refined against (black); of the model versus the first half map (red); and of the model versus the second half map (green). Z-complex, Ca_v2.2–ziconotide.

Extended Data Fig. 2 |. Cryo-EM structure of the human Ca_v2.2 complex bound with ziconotide.

a, Heat map for local resolution of the complex. The resolution map was calculated in Relion 3.0 and prepared in Chimera. Top inset, bound ziconotide (labelled as Zi) is well-resolved. Bottom inset, resolution of the β3 subunit, after a focused refinement, allows for reliable model building using the crystal structure of rat β3 (PDB code 1VYU) as a template. b, Overall structure of the complex at an averaged resolution of 3.0 Å. Left, the complex comprises the α1 core subunit (silver), the extracellular α2δ−1 subunit (light pink for α2 and green for δ) and the cytosolic β3 subunit (wheat). The peptide pore blocker ziconotide is coloured brown. The resolved lipid, cholesterol and CHS molecules are shown as black sticks. The bound PIP₂ is shown as black ball-and-sticks. Sugar moieties are shown as thin sticks. Right, surface presentation of the structure. The four repeats are coloured grey, cyan, yellow, and pale green. The III–IV linker and the CTD are coloured orange and pale purple, respectively.

Extended Data Fig. 3 |. Electron microscopy densities for representative segments of Ca_v2.2–ziconotide.

a, Electron microscopy maps for representative segments in α1 and β3. The densities for the β3 segments are from focused refinement, and the others are from the overall map. All the densities shown are contoured at 4σ. b, The electron microscopy map for ziconotide. c, Electron microscopy densities for the bound Ca²⁺ ion and surrounding residues in the selectivity filter. The maps were prepared in PyMol.

Extended Data Fig. 4 |. Lipids resolved in the structures.

a, A PIP₂ molecule binds to VSD_II in both structures. All the densities shown are contoured at 3σ. b, The densities for the resolved cholesterol (Cho) and CHS molecules in Ca_v2.2–ziconotide. c, Lipids resolved in the structure of Ca_v2.2–ziconotide. The α1 subunit are shown in two opposite side views. The numbering for cholesterol and CHS is consistent with that in b. Two phospholipids are also resolved and assigned as phosphatidylethanolamine (PE).

Extended Data Fig. 5 |. Structural comparison of the Ca_v1.1 and Ca_v2.2 channel complexes.

a, Superimposition of the overall structures of human Ca_v2.2 (apo) and rabbit Ca_v1.1 (PDB code 5GJW). For visual clarity, the γ subunit in the endogenous Ca_v1.1 complex is not shown. The conformational shift of VSD_II from Ca_v1.1 (wheat) to Ca_v2.2 (blue) is indicated by the blue arrow. b, Identical structures of the α2δ−1 subunit in the two channel complexes. A detailed structural description of the α2δ−1 subunit can be found in a previous publication¹⁸. c, Structural differences of the ECLs between Ca_v1.1 and Ca_v2.2. An extracellular view of the superimposed α1 subunits in the two channels is shown.

Extended Data Fig. 6 |. Conformational shifts of Ca_v2.2 upon ziconotide binding.

a, ECL_I does not participate in ziconotide coordination. An extracellular view perpendicular to that in Fig. 1c is shown. b, Slightly different mode of action of KIIIA for Na_v1.2²⁶. Lys7 in KIIIA directly blocks the outer mouth of the selectivity filter vestibule of Na_v1.2 (PDB code 6J8E), in a manner similar to a cork. Ziconotide lacks an equivalent basic residue. c, Relative shift of α2δ−1 between apo (blue) and ziconotide-bound Ca_v2.2 (domain-coloured) when the two structures are superimposed relative to the α1 subunit. The rest of the complex remains identical except for ECL_III. d, Concerted motion of α2δ−1 and ECL_III of α1. The α2δ−1 subunit in the two structures can be superimposed with a root mean square deviation of 0.28 Å over 847 Cα atoms, indicating nearly identical conformations. When the two structures are superimposed relative to α2δ−1, the entire α1 undergoes a relative shift—except for ECL_III, which stays as a rigid body with α2δ−1.

Extended Data Fig. 7 |. Conformational changes of VSD_II and VSD_III between Ca_v1.1 and Ca_v2.2.

a, Structural comparison of Ca_v2.2 VSD_II with other VSDs that exhibit down conformations. To make the nomenclature consistent, we define the gating charge residue on the first helical turn of the S4 segment as R1. The PDB accession codes are 6NT4 for VSD_IV in the chimeric Na_vPaS-1.7, 7K18 for toxin-bound VSD_IV in rat Na_v1.5, 4G80 for the antibody-locked VSD of a voltage-sensitive phosphatase, and 6UQF for the VSD of HCN1 in hyperpolarized conformation. b, Structural comparison of Ca_v1.1 and Ca_v2.2 shows a slight rotation of VSD_III around the pore domain. The superimposed structures of the diagonal repeats I and III of Ca_v1.1 (wheat) and Ca_v2.2 (domain-coloured) are shown. c, VSD_III remains nearly rigid in these two structures. When the structures of VSD_III in the two channels are individually superimposed, the S4 segment and the gating charge residues align well. d, Marked shift of S4_II between Ca_v1.1 and Ca_v2.2 when the two structures are compared relative to VSD_II. S4_II undergoes a combination of spiral sliding and secondary structural transition. S1, S2 and S3 remain nearly unchanged in these two VSD_II structures, which suggests a concerted rotation of the other three segments pivoting around S4.

Extended Data Fig. 8 |. A closed pore domain with one small fenestration.

a, The pore domain is in a closed conformation. Four perpendicular side views of the pore domain are shown. S4–5_II is pushed downward as a result of the sliding of S4_II. b, Side walls that involve S6_II are sealed without fenestration. Side views of the pore domain surface are shown. There is only one fenestration on the interface of repeats III and IV.

Extended Data Fig. 9 |. A PIP₂ molecule may help to stabilize the down conformation of Ca_v2.2 VSD_II.

a, The binding pose for PIP₂ is incompatible with an up VSD_II. Left, coordination of the head group of PIP₂ by Ca_v2.2. Side local view of VSD_II is shown. Right, in an up state of VSD_II (as in Ca_v1.1), R4 and K5 can no longer interact with the PIP₂ head group, and S4–5_II directly clashes with PIP₂. Structures of Ca_v1.1 and Ca_v2.2 are superimposed relative to the α1 subunit and Ca_v2.2 is omitted to highlight the relative position of PIP₂ to Ca_v1.1. b, The hydrophobic tails of PIP₂ interact extensively with multiple segments in repeats II and III. Hydrophobic residues on segments S3 to S6 in repeat II and S5 and S6 in repeat III contact the two tails of PIP₂. c, The PIP₂ molecule in the K_v channel KCNQ1 is bound at a similar, but lower, position. The PDB code for the KCNQ1 structure is 6V01. d, Rearrangement of the interface of VSD_II and pore domain between Ca_v2.2 and Ca_v1.1. Ca_v1.1 is coloured with the same scheme as for Ca_v2.2. Alternative sets of hydrophobic residues between the gating charge residues on S4_II are used for interacting with S5_III in Ca_v1.1 and Ca_v2.2 as a result of the rotation of S4_II. The sequence numbers for corresponding VSD_II residues in these two channels differ by 50, and those for S5_III residues differ by 354. As labelled in the parentheses, Val1298 and Phe1292 on the S5_III segment of Ca_v2.2 are at loci corresponding to Cys944 and Leu938 in Ca_v1.1, respectively.

Extended Data Table 1 |.

Statistics for data collection and structural refinement

	Cav2.2-ziconotide (EMDB-23867) (PDB 7MIX)	Cav2.2-apo (EMDB-23868) (PDB 7MIY)
Data collection and processing
Magnification	105,000	105,000
Voltage (kV)	300	300
Electron dose (e-/Å2)	50	50
Defocus range (μm)	−2.1~−1.9	−2.1~−1.9
Pixel size (Å)	1.114	1.114
Symmetry	C1	C1
Initial particle images (no.)	1,909,156	928,665
Final particle images (no.)	170,839	56,616
Map resolution (Å)	3.0	3.1
FSC threshold	0.143	0.143
Map resolution range (Å)	2.8-31.2	2.9-50
Refinement
Initial model used (PDB code)	5GJW, 1VYU, 1OMG	Cav2.2-ziconotide
Model resolution (Å)	3.3	3.4
FSC threshold	0.5	0.5
Map sharpening B factor (Å2)	−79	−59
Model composition
Non-hydrogen atoms	21654	21485
Protein residues	2623	2598
Ligands	28	28
B factors (Å2)
Protein	58.27	78.42
Ligand	69.59	62.16
R.m.s deviations
Bond lengths (Å)	0.004	0.003
Bond angles (°)	0.759	0.727
Validation
MolProbity score	2.00	1.95
Clashscore	9.72	9.68
Poor rotamers (%)	1.47	1.31
Ramachandran plot
Favored (%)	94.76	94.99
Allowed (%)	5.12	4.86
Disallowed (%)	0.12	0.16

Data availability

The atomic coordinates and electron microscopy maps for Ca_v2.2 in complex with ziconotide and alone have been deposited in the PDB with the accession codes 7MIX (with ziconotide) and 7MIY (without ziconotide) and in the Electron Microscopy Data Bank with the codes EMD-23867 (with ziconotide) and EMD-23868 (without ziconotide), respectively. The atomic coordinates of the proteins for structural comparison in this study can be found in the PDB: rabbit Ca_v1.1 (5GJW) (https://www.rcsb.org/structure/5GJW), toxin-bound human Na_v1.2 (6J8E) (https://www.rcsb.org/structure/6J8E), toxin-bound Na_vPaS-1.7 chimera (6NT4) (https://www.rcsb.org/structure/6NT4), toxin-bound rat Na_v1.5 (7K18) (https://www.rcsb.org/structure/7K18), Ci-VSP (4G80) (https://www.rcsb.org/structure/4G80), HCN1 (6UQF) (https://www.rcsb.org/structure/6UQF) and KCNQ1 (6V01) (https://www.rcsb.org/structure/6V01). Expression plasmids for the Ca_v2.2 subunits are available from the corresponding author upon reasonable request.