Open Structure of the Ca²⁺ Gating Ring in the High-Conductance Ca²⁺-Activated K⁺ Channel

Abstract

High conductance voltage-and Ca²⁺-activated K⁺ channels (Slo1 or BK channels) function in many physiological processes that link cell membrane voltage and intracellular Ca²⁺, including neuronal electrical activity, skeletal and smooth muscle contraction, and hair cell tuning^1–8. Like other voltage-dependent K⁺ (Kv) channels, BK channels open when the cell membrane depolarizes, but in contrast to other Kv channels they also open when intracellular Ca²⁺ levels rise. Channel opening by Ca²⁺ is conferred by a structure called the gating ring, located in the cytoplasm. Recent structural studies have defined the Ca²⁺-free, closed conformation of the gating ring, but the open conformation is not yet known⁹. Here we present the Ca²⁺-bound, open conformation of the gating ring. This structure shows how one layer of the gating ring, in response to the binding of Ca²⁺, opens like the petals of a flower. The magnitude of opening explains how Ca²⁺ binding can open the pore. These findings present amolecular basis of Ca²⁺ activation and suggest new possibilities for targeting the gating ring to treat diseases such as asthma and hypertension.

Regulators of K⁺ conductance (RCK) domains are ubiquitous among ion channels and transporters in prokaryotic cells^10–14. Eight RCK domains assemble in the cytoplasm to form a closed-ring structure that changes its diameter upon ligand binding, thus enabling the ‘gating ring’ to regulate allosterically the transmembrane component of the transport protein. In prokaryotic cells the gating ring most often consists of eight identical RCK domains, arranged as a tetrad of pairs, giving rise to a four-fold symmetric ring with identical top and bottom layers of RCK domains^12,15. Intracellular ligands such as Ca²⁺ or small organic molecules bind in a cleft between RCK pairs forming the top and bottom layers to affect the shape of the ring^12,15–17.

RCK domains are also found in higher eukaryotes in the Slo family of K⁺ channels^9,10,18,19. There, two non-identical RCK domains are encoded in the C-terminus of the K⁺ channel subunit. A tetrameric K⁺ channel thus provides eight RCK domains to make a gating ring, but in contrast to most prokaryotic gating rings with identical RCK domains, the eukaryotic gating ring has one kind of RCK domain (RCK1) forming its top layer and another kind (RCK2) forming its bottom layer (figure 1a). A structure of a Ca²⁺-activated (Slo1 or BK) K⁺ channel gating ring in its Ca²⁺-free, closed conformation was recently determined (PDB code 3NAF), as was the structure of an RCK1-RCK2 pair in the presence of Ca²⁺ (PDB code 3MT5), detailing the chemistry of the Ca²⁺ binding site^9,18. These studies showed that the BK gating ring indeed has distinct top and bottom layers, and also that a Ca²⁺ binding site known as the Ca²⁺ bowl is in a different location than the ligand binding site in prokaryotic gating rings. These unique structural properties suggest that in order to regulate conduction, the eukaryotic gating ring may undergo conformational changes in a different manner than its prokaryotic counterpart.

Figure 1. The BK channel and the gating ring.

a, Domain topology of the BK channel. Only two opposing subunits are shown for clarity. b, Crystal structure of the Ca²⁺-bound open gating ring with RCK1 in blue and RCK2 in red. Ca²⁺ ions are shown as yellow spheres. The N-termini of RCK1 that connect to the C-termini of the inner helices are indicated as green asterisks. c, Structure of the Ca²⁺-free closed gating ring with RCK1 in blue and RCK2 in red (PDB: 3NAF). d, The RCK1 N-terminal lobes (blue) and the Ca²⁺-bowls (red) from the Ca²⁺-bound open gating ring. The diagonal distance between the Cα atoms of the N-terminal residues (Lys 343) is indicated. e, The corresponding region to that shown in d from the closed gating ring (PDB: 3NAF).

We determined the crystal structure of the Ca²⁺-bound gating ring from the zebrafish BK channel at 3.6 Å resolution using an expression construct in which a loop consisting of residues 839–872 was deleted (figure 1b, Supplementary Table, and Supplementary Figures 1–2). This loop was disordered in both the Ca²⁺-free human BK gating ring (3NAF) and the Ca²⁺-bound RCK1-RCK2 pair (3MT5), and its deletion in both the human and zebrafish BK channels has no detectable affect on function^9,18 (Supplementary Figure 3). Therefore, the relative stability of conformational states of the BK gating ring appear to be unaffected by this loop. Initial phases were determined by molecular replacement using the RCK1-RCK2 Ca²⁺-bound structure as a search model¹⁸. The solution showed the presence of eight RCK1-RCK2 pairs(two gating rings) per asymmetric unit of the crystal, which allowed eight-fold noncrystallographic symmetry averaging of electron density maps and eight-fold restraints throughout model refinement. The first step of refinement required rigid body adjustment of the N-terminal lobe of RCK1. The final model was refined to working and free residuals (R_work and R_free) of 26.0% and 28.9%, respectively.

The zebrafish channel is 93% identical in amino acid sequence to the human BK channel, for which the Ca²⁺-free, presumably closed conformation is known (figure 1c and Supplementary sequence alignment)⁹. The high degree of identity between zebrafish and human BK channels justifies comparison of the two structures to assess potential conformational changes underlying Ca²⁺-mediated gating. The Ca²⁺-bound and Ca²⁺-free gating rings are shown (figure 1b,c). In these representations RCK1, which forms the layer of the gating ring closest to the membrane, is colored blue and RCK2 is colored red. In both gating rings a green asterisk marks the N-terminal residue of RCK1, Lys 343. In a full-length BK channel the N-terminal residue of RCK1 is connected by a 16 amino acid linker to the C-terminus of the transmembrane inner helix, which forms the pore’s gate²⁰.

The two gating ring structures, Ca²⁺-bound and Ca²⁺-free, differ in two main respects. First, in the Ca²⁺-bound structure electron density attributable to Ca²⁺ is present in the Ca²⁺ bowl at the ‘assembly interface’, and second, the RCK1 layer in the Ca²⁺-bound gating ring is expanded from a diameter of 81 Å to 93 Å, measured at the position of Lys 343 (figure 1d,e and Supplementary Figure 2 c,d). An objective comparison of conformational differences between the Ca²⁺-bound and Ca²⁺-free gating rings is achieved by superimposing all Cα atoms of the tetrameric gating rings. Such a superposition shows that the RCK2 components of the ring undergo very little change, and only modest conformational differences occur at the assembly interface, except locally near the Ca²⁺ bowl (figure 2c, Supplementary Figure 4). The Ca²⁺ bowl structure is nearly identical to that of the Ca²⁺-bound RCK1-RCK2 pair¹⁸ (Supplementary Figure 2e). The major conformational change involves the N-terminal lobe of RCK1, which changes its angle with respect to RCK2 (figure 2a,b, Supplementary Figure 5). Viewing an interpolative movie in which the Ca²⁺-free gating ring is morphed to the Ca²⁺-bound gating ring, it appears that Ca²⁺ binding causes the N-terminal lobes of the RCK1 units to open up on the membrane-facing surface of the gating ring akin to petals opening on a flower (Supplementary Movie 1, part 1).

Figure 2. The flexible and the assembly interfaces.

a, The flexible interface from the open gating ring structure with RCK1 in blue and RCK2 in red. The Ca²⁺ ions are shown as yellow spheres. The N-and C-terminal lobes of RCK1 and RCK2 are labeled. Large disordered regionsare indicated by dashed lines. b, The flexible interface from the closed gating ring with RCK1 in blue and RCK2 in red (PDB: 3NAF). c, Superposition of the assembly interfaces from the open and the closed gating ring structures.

We next analyzed whether these conformational changes are compatible with prior structural data on the opening and closing of K⁺ channels. We started with the open prokaryotic K⁺ channel MthK and replaced its gating ring with the Ca²⁺-bound BK channel gating ring¹⁵. This step involved alignment of the four-fold axis followed by minimization of the distance between the BK gating ring N-terminus and the corresponding position of the MthK gating ring through least-squares minimization of residues 343 of BK with the corresponding residues, 116, of MthK. The resulting model of an open BK channel gating ring with an open K⁺ pore is shown (figure 3a). The distance between the C-terminal residue of the inner helix of the pore and the N-terminal residue of RCK1, which was not a constraint in making the model, is 33 Å. In a second step we replaced the open MthK pore with the closed KcsA pore by aligning the invariant selectivity filter, while also replacing the Ca²⁺-bound BK channel gating ring with the Ca²⁺-free BK channel gating ring by aligning the nearly invariant RCK2 layer^9,21. The resulting model of a Ca²⁺-free BK gating ring with a closed K⁺ pore is shown (figure 3b). In this closed model the distance between the C-terminal residue of the inner helix and the N-terminal residue of RCK1, which again was not a constraint in making the model, is 32 Å. In this analysis the similar pore-to-gating ring distance (~32 Å) in both the open and closed models suggests that the magnitude of conformational change observed within the gating ring is compatible with the magnitude of change known to occur in the pore of other kinds of K⁺ channels (Supplementary Movie 1 part 2). Whether or not the pore of the BK channel undergoes conformational changes similar to or different than other K⁺ channels remains to be determined^22–26.

Figure 3. A Ca²⁺-gating model for the BK channel.

a, Views of the BK channel model with an open gating ring and an open pore. Structure of the Ca²⁺-bound BK gating ring is docked onto the open MthK channel (PDB: 1LNQ) by aligning the gating rings. Yellow spheres represent Ca²⁺ ions. The N-terminal residues of RCK1 and the C-terminal residues of the inner helices are shown as black spheres. b, Views of the BK channel model with a Ca²⁺-free gating ring and a closed pore. Here, the Ca²⁺-bound gating ring in a is replaced by the Ca²⁺-free gating ring (PDB: 3NAF) and the open MthK pore in a is replaced by the closed KcsA pore (PDB: 1K4C). The innerhelix from KcsA is truncated to match the length in MthK. See Supplementary Movie 1.

Another feature of the model in figure 3 is noteworthy. In the published structure of the Ca²⁺-free gating ring, part of the linker extending from the gating ring centrally towards the pore was present in the crystal structure and visible on the surface of the gating ring⁹. The linker was held by a four-helix bundle fused to the N-terminus to facilitate crystallization. Although the linker position in the structure might be affected by this attachment, we note that it runs a course almost coincident with the dashed lines in our closed model (figure 3b). The agreement between the observed linker position in the closed gating ring crystal structure and its inferred position in the closed conformation of our model is an interesting correlation because the radial direction of the linkers in the model (dashed lines) would naturally allow the expansion of the gating ring to exert an opening force on the pore’s inner helices. The idea that the gating ring exerts a direct force on the pore has been proposed on the basis of mutagenesis studies showing that the linker length is critical to efficient Ca²⁺-mediated gating^19,20.

The structural differences between the BK gating ring and the MthK gating ring underlie distinct ligand-induced conformational changes. A higher degree of molecular symmetry in the MthK gating ring stems from its containing eight identical RCK domains^12,15. Consequently the MthK gating ring changes equally on both layers so that Ca²⁺ binding causes a shrinking of its height and an expansion of its diameter (Supplementary Movie 2)¹². By contrast, in the BK gating ring Ca²⁺ binding causes only the layer facing the membrane to undergo a significant conformational change (Supplementary Movie 1). This comparison presents a fascinating example in which the evolution of molecular structure has given rise to new or modified mechanical properties within a class of molecules. Precisely how the free energy of Ca²⁺ binding in either the BK or MthK gating rings is transduced into mechanical work remains an important outstanding question that will require further experimental and theoretical work to understand at a deep level.

The data and analysis presented here define endpoints in the Ca²⁺-induced gating conformational transition of the BK gating ring. Given the large conformational change that occurs upon Ca²⁺ binding it should in principle now be possible to identify small molecules that stabilize one conformation or the other. Such small molecules promise to have important and possibly beneficial physiological effects because BK channels regulate smooth muscle tone in pulmonary airways and vascular beds^27–30.

Methods Summary

The cytoplasmic domain (CTD) construct of the zebrafish BK channel including residues 341–1060 with a loop deletion (residues 839–872) was expressed and purified following the same method as described¹⁸. However, the size-exclusion chromatography was carried out in buffer containing 20 mM Hepes pH 8.0, 150 mM NaCl, and 20 mM dithiothreitol (DTT). The protein was concentrated to about 6 mg/ml and 10 mM CaCl₂ was added before crystallization experiments. Crystals were grown at 20°C using hanging-drop vapor diffusion by mixing equal volumes of protein and a reservoir solution containing 50 mM 2-(N-morpholino) ethanesulfonic acid (MES), 4% (w:v) PEG 4000, and 100 mM potassium sodium tartrate (pH 6.3). The crystals belong to space group P2₁2₁2₁ with cell dimensions of a = 137.65 Å, b = 210.82 Å, c = 238.76 Å and α = β = γ = 90°. Each asymmetric unit contains eight protein subunits forming two gating rings. The structure was determined by molecular replacement using the monomeric Ca²⁺-bound human BK CTD (PDB code: 3MT5) as the search model. The final model was refined to 3.6 Å resolution with R_work = 0.260 and R_free = 0.289. Electrophysiology recordings were conducted using patch-clamp on Xenopus oocytes expressing the wild-type and mutant channels.

Methods

Cloning, expression, and purification

A synthetic gene encoding the zebrafish BK channel (GI: 189526846) was purchased from BioBasic Inc. and served as the template for subcloning. The amino acid sequence is 93% identical to that of the human BK channel. Based on previous structural and functional studies of the human BK channel¹⁸, a cytoplasmic domain (CTD) construct including residues 341–1060 with a loop deletion (residues 839–872) was expressed and purified following the same method as described¹⁸. However, the size-exclusion chromatography was carried out in 20 mM Hepes pH 8.0, 150 mM NaCl, and 20 mM dithiothreitol (DTT). The protein was concentrated to about 6 mg/ml and 10 mM CaCl₂ was added to achieve the ligand-bound open conformation before crystallization experiments.

Crystallization and structure determination

Crystals of the zebrafish BK CTD were grown at 20°C using hanging-drop vapor diffusion by mixing equal volumes of protein and a reservoir solution containing 50 mM 2-(N-morpholino)ethanesulfonic acid (MES), 4% (w:v) PEG 4000, and 100 mM potassium sodium tartrate (pH 6.3). The crystals grew to a maximum size of about 0.2 x 0.2 x 0.3 mm³ within three days. Before being flash frozen in liquid nitrogen, the crystals were briefly transferred to a cryoprotectant solution containing 50 mM MES, 6% (w:v) PEG 4000, 100 mM potassium sodium tartrate, 150 mM NaCl, 10 mM CaCl₂, and 30% ethylene glycol (pH 6.3). The crystals belong to space group P2₁2₁2₁ with cell dimensions of a = 137.65 Å, b = 210.82 Å, c = 238.76 Å and α = β = γ = 90°. Each asymmetric unit contains eight protein subunits.

Diffraction data were measured at beamline X29 from the National Synchrotron Light Source and were processed with the HKL2000 program suite³¹. The scaled data set was anisotropically corrected to resolution limits of 4.0, 3.6, and 3.6 Å along the reciprocal cell directions a*, b*, and c* respectively, using the diffraction anisotropy server at the University of California, Los Angeles³². An isotropic B factor of −29.98 Ų was applied to restore the magnitude of the high-resolution reflections diminished by anisotropic scaling. The structure was determined by molecular replacement using the monomeric Ca²⁺-bound human BK CTD (PDB code: 3MT5) as the search model¹⁸. Initial phases were improved by eight-fold noncrystallographic symmetry (NCS) averaging carried out with DM from the CCP4 program suite³³. Iterative model building was carried out in COOT³⁴. Rounds of refinement were performed with CNS³⁵ and REFMAC³⁶ and strong eightfold noncrystallographic symmetry restraints were maintained throughout refinement. The final model was refined to 3.6 Å resolution with R_work = 0.260 and R_free = 0.289. A few disordered regions were not modeled due to weak electron density and the final refined model includes residues 343–571, 581–614, 688–810, 817–837, 875–949,953–1024, and 1029–1059. Residues for which side chain density was poorly defined were modeled as alanines. The majority (95.9%) of the residues lie in the most favored region in a Ramachandran plot, with the remaining 4.1% in the additionally allowed regions. Eight Ca²⁺ ions were modeled in the Ca²⁺ bowls from the two gating rings in the crystal. Data collection and structure refinement statistics are shown in Supplementary Table 1. All structural illustrations were prepared with PYMOL (www.pymol.org). The multiple-chain morphing script for CNS³⁷ was used to generate coordinates for the supplementary movies.

Electrophysiology

Channel expression in Xenopus laevis oocytes and electrophysiological recordings in inside-out patches were carried out as previously described¹⁸. Wild-type and deletion zebrafish BK constructs were cloned into the pGEM-HE vector³⁸. Plasmids were linearized with PciI and capped cRNA was produced by in vitro transcription from the T7 promoter using the AmpliCap-Max T7 kit (Epicentre Biotechnologies). Xenopus oocytes were injected with ~10ng of cRNA and recordings were made 48–60 hours later. For recording, inside-out patches were excised from freshly devitellinized oocytes using fire-polished glass pipettes with a typical resistance of 1–2 MΩ. In Ca²⁺-free conditions the bath solution contained 140 mMK-MeSO₃, 20mMHepes, 2mMKCl, 5 mM EGTA (final pH: 7.0). For recordings in 10 μM free Ca²⁺, the same solution was supplemented with 4.83 mM CaCl₂, adjusting the pH to 7.0 (we assumed a K_d of 376 nM for the EGTA-Ca²⁺ complex at pH 7.0). The pipette solution was the same as the Ca²⁺-free bath solution, but supplemented with 2mM MgCl₂. During recording, the membrane voltage was held at −60 mV and currents were elicited by pulsing the voltage from −120 to +140 mV in 10 mV increments (length of pulse: 20 ms). For conductance versus voltage plots, measurements were taken from tail currents at −60 mV,700 μs after the voltage step.