Ball-and-chain inactivation of a human large conductance calcium-activated potassium channel

Abstract

BK channels are large-conductance calcium (Ca²⁺)-activated potassium channels crucial for neuronal excitability, muscle contraction, and neurotransmitter release. The pore-forming (α) subunits co-assemble with auxiliary (β and γ) subunits that modulate their function. Previous studies demonstrated that the N-termini of β2-subunits can inactivate BK channels, but with no structural correlate. Here, we investigate BK β2-subunit inactivation using cryo-electron microscopy, electrophysiology and molecular dynamics simulations. We find that the β2 N-terminus occludes the pore only in the Ca²⁺-bound open state, via a ball-and-chain mechanism. The first three hydrophobic residues of β2 are crucial for occlusion, while the remainder of the N-terminus remains flexible. Neither the closed channel conformation obtained in the absence of Ca²⁺ nor an intermediate conformation found in the presence of Ca²⁺ show density for the N-terminus of the β2 subunit in their pore, likely due to narrower side access portals preventing their entry into the channel pore.

The accessory β2 subunit was proposed to inactivate BK channels via the ball-and-chain model, but structural evidence was missing. Here, using cryo-EM, the authors captured the occlusion of the BK channel pore by the N-terminus of the β2 subunit.

Introduction

Large conductance and calcium-activated potassium channels (also known as MaxiK, BK) are regulated by two physiological stimuli, intracellular calcium (Ca²⁺) and depolarizing voltages^1,2. BK channels have a particularly large conductance (~300 pS) and are therefore very effective in hyperpolarizing the membrane. They are ubiquitously expressed, and their physiological role is generally to dampen cell excitability by rapidly hyperpolarizing the cell membrane upon activation³. Activation by both cytosolic Ca²⁺ elevation and membrane depolarization makes these channels uniquely tailored to regulate voltage dependent Ca²⁺ channels in many cell types^4,5. Deficiencies in BK channels lead to diseases like hypertension, intellectual disability, autism, and epilepsy⁶.

BK is a tetrameric potassium (K⁺)-selective channel formed from four identical $\mathrm{\alpha }$ subunits encoded by the K_Ca1.1 (or Kcnma1, or Slo1) gene. Each $\mathrm{\alpha }$ subunit contains 7 transmembrane helices, including a voltage sensor domain (S1-S4 helices) and a pore domain (S5-S6 helices), followed by two RCK (regulator of K⁺ conductance) domains that form a large cytosolic gating ring. Several structures of BK channels in both closed and open states, and from different species have been reported^7–10. In mammals, Slo1 channels associate with auxiliary proteins (β and $\mathrm{\gamma }$) that confer tissue-specific gating and influence their voltage-dependent activation, calcium sensitivity and pharmacological properties^11,12. All four β subunits, encoded by the Kcnmb genes (Kcnmb1-4), display high sequence homology, with two transmembrane helices, cytosolic N and C termini, and a large cysteine-rich extracellular loop (Supplementary Fig. 1a). Cryo-electron microscopy (cryo-EM) structures of BK channels in complex with β4 subunits reveal their architecture and the 4α:4β stoichiometry and assembly⁹. BK channels in smooth muscle assemble with accessory β1 subunits, increasing their sensitivity to Ca²⁺ and modulating gating kinetics^13,14 while in the brain, these channels associate with the β4 subunit, regulating the kinetics of channel activation and deactivation¹⁵. β2, expressed in brain, pancreas and adrenal chromaffin cells, among others, confers rapid and complete inactivation of BK channels¹⁶. Some of the splice variants of the β3 subunits cause incomplete inactivation of BK channels¹⁷.

Inactivation is the process by which ion flux through channels is terminated despite continuous presence of the opening stimulus^18,19. In neurons, inactivation of voltage-gated Na⁺ and K⁺ (Kv) channels is crucial for action potential generation and regulation of firing frequency^18–20. Many K⁺ channels are believed to inactivate via the so-called ball-and-chain (or N-type) inactivation^21–23 where a N-terminal cytosolic “ball” domain, either intrinsic to the channel or belonging to an accessory channel subunit^16,24, is hypothesized to bind inside an open channel pore and plug the ion permeation pathway^21,22,25. The only structural correlate of ball-and-chain inactivation in a K⁺ channel comes from a prokaryotic homolog of BK channels, MthK, from M. thermoautotrophicum, where the first N-terminal 17 residues of the channel form the inactivating ball domain²⁶.

BK channel inactivation also plays essential roles in neuronal excitability. Specifically, it was shown to contribute to use-dependent spike broadening in the pyramidal cells of the hypothalamus and lateral amygdala neurons²⁷, potentially leading to enhanced transmitter release at the nerve terminal. In addition, the brain’s intrinsic clock circuit was proposed to use BK channel inactivation as a switch to set diurnal variation in suprachiasmatic nucleus excitability that underlies circadian rhythm in the hypothalamus²⁸. Inactivation of BK channels via β2 and β3 subunit-association was initially proposed to also occur via the ball-and-chain mechanism, mediated by their N-terminal segments, based on sequence similarity to other inactivating ball domains, sensitivity to trypsin, and sensitivity to permeant ion concentration^16,17,29. Indeed, the N-termini of the inactivating β subunits are necessary and sufficient for inactivation; N-terminal truncated constructs of inactivating β subunits produce non-inactivating BK channels and, conversely, grafting N-termini from inactivating β subunits to non-inactivating ones, confers inactivation onto BK channels. In addition, synthetic peptides that mimic N-terminal peptides, block non-inactivating BK channels, further supporting this mechanism^15,16. However, other findings later argued that the molecular mechanism for β subunit-induced inactivation in BK channels may be different from that observed in Kv channels. For example, unlike inactivation in Kv channels³⁰, BK channels do not reopen during recovery from inactivation, indicating that BK channels may be able to close with the inactivation domain still bound to the pore³¹. Also, while ball-and-chain inactivation in Kv channels is slowed by quaternary ammonium blockers that bind to the ion channel pore³², BK channel inactivation appears less sensitive to such blockers^16,31,33, suggesting a different mechanism from direct pore blockade. Thus, there are both differences and similarities between ball-and-chain inactivation in Kv channels and BK inactivation, and the details of the molecular mechanism are still unclear.

Here we investigate the inactivation of BK channels by solving cryo-EM structures of hSlo1 in complex with an inactivating β subunit (β2 N-terminus grafted on the non-inactivating β4 scaffold). Structures in the presence of Ca²⁺ reveal an open and inactivated channel where the N-terminus of one of the four β subunits plugs the open pore as in ball-and-chain type inactivation. Molecular dynamics simulations show that the first three large hydrophobic residues of the ball domain are crucial for inactivation by remaining stably bound to the pore, whereas the remainder of the domain is flexible, in agreement with previous studies²⁹. We also observe an additional, previously unreported conformation, where Ca²⁺ is bound to the gating ring but the pore is not as wide as in the open state, indicating that the inactivating β2 N-terminus stabilizes an on-path intermediate state. The β2 ball does not enter the intermediate state pore nor the closed channel pore, likely because access to the pore cavity is blocked by the side portals between the transmembrane region and the gating ring, which are narrower in both these states.

Results

β2_N-β4 subunit inactivates hSlo1 at depolarizing voltages

In order to structurally capture β2 subunit inactivation of hSlo1 BK channels, we made a construct where we appended the ball domain of β2 required for inactivation (the N-terminal first 45 residues^15,29, to the non-inactivating β4 subunit (Methods, Supplementary Fig. 1)). The β4 subunit is highly homologous to the β2 subunit (41.25% similarity, 26.67% identity) although it lacks an N-terminal inactivating domain. Unlike β2, β4 has been shown to purify and form a complex structurally with hSlo1, therefore we co-opted it as a well-expressing scaffold⁹. We named this hybrid subunit β2_N-β4. The β2_N-β4 subunit was successfully expressed in HEK293S GnTI^- cells together with hSlo1, and, importantly, conferred inactivation to hSlo1 currents (Fig. 1a, b). The inactivation characteristics induced by the β2_N-β4 subunit were comparable with those of the β2 subunit, with differences in the apparent voltage-dependence of the process, likely caused by the well-documented effects of the β4 subunit on BK channel gating^9,15. These differences were not probed further here, as we were only interested in conferring inactivation to BK channels via the β2 N-terminus. Such chimeric constructs between inactivating and non-inactivating β subunits were previously employed in functional assays to demonstrate the portability of the N-terminal inactivating domains, and the modularity of the gating effects¹⁵. We next purified the hSlo1/β2_N-β4 complexes from HEK293S GnTI^- cells and determined the cryo-EM structures in both the presence and absence of divalent ions (10 mM Ca²⁺ and 15 mM Mg²⁺) and in both detergent micelles, glyco-diosgenin (GDN), and MSP2N2 lipid nanodiscs (Supplementary Figs. 2–7 and Supplementary Table 1). All structures determined in this report revealed tetrameric hSlo1 channels with a full complement of β2_N-β4 subunits (4) peripherally associated with the α subunits as previously shown for hSlo1 in complex with β4⁹.

Fig. 1. N-terminus of β2 subunit inactivates hSlo1 channels by plugging the open pore.

a Representative excised inside-out patch clamp current traces in response to voltage steps from −100 mV to +140 mV with a holding potential of −80 mV, each with a duration of 500 msec, from hSlo1, hSlo1 + β2, and hSlo1 + β2_N-β4 channels expressed in HEK293S GnTI^- cells. The bath solution contained 10 μM free Ca²⁺. Prior to the depolarizing voltage steps, a 50 ms, −180 mV pulse was applied. b Plot of the ratio of steady state and maximal currents for hSlo1, hSlo1 + β2, and hSlo1 + β2_N-β4 from currents as in a. Empty symbols with error bars are mean and standard errors for hSlo1 (n = 3), hSlo1 + β2 (n = 3), and hSlo1 + β2_N-β4 (n = 5), where n is the number of independent patches. Filled symbols represent individual measurements. c Cryo-EM density map of hSlo1 in complex with β2_N-β4 in the open state, class1, hSlo1 channel in gray, β4 in blue, β2 inactivating ball and chain in orange, Ca²⁺ bound to the RCK in yellow, Mg²⁺ bound to the RCK in pink and lipids in pale yellow. Models for the transmembrane regions for 2 opposing hSlo1 subunits (gating rings removed) in gray with the β2_N-β4 subunit (β4 in blue and β2 in orange) to better display the ball density (orange) beneath the selectivity filter for class 1 (d), class 2 (e), class 3 (f). Black dashed line represents missing connecting regions (e.g. from residue 13 to 34 in d–f). Modeled residues of β2 N-terminus from 34 to 43 are in orange. Inset in d, shows the enlarged cryo-EM density of the ball with modeled residues F2-Y13. g Overlay of all densities hypothesized to belong to β2 from all three classes (d–f).

N-terminus of β2 subunit occludes the open pore of BK channels

Unlike the previous BK channel (with and without the β4 subunit) structures obtained in the presence of Ca²⁺ and Mg²⁺, where only one open state conformation was determined, we obtained two distinct conformations from the divalent-bound hSlo1/β2_N-β4 complexes in both detergent micelles and in MSP2N2 lipid nanodiscs containing POPC:POPE:POPS (5:3:2) (Supplementary Figs. 2, 3, 5 and 6).

In detergent, in the presence of Ca²⁺ and Mg²⁺, initial data processing yielded two classes, each containing about half of the particles (Supplementary Figs. 2 and 3). One class with 46% of particles has hSlo1 in an open conformation with the S6 helix bending at the G310 hinge and widening the cavity to the previously reported size of intracellular gate in the open state and with the gating ring in a divalent-bound and activated state⁷. This structure is identical to that already reported for the divalent-bound hSlo1/β4 complex⁹ (Cα r.m.s.d. of ~1 Å compared with PDB 6V22) with one major addition: a very strong, bulky density is observed lodged inside the open pore cavity of the channel plugging the pore and preventing K⁺ flux indicative of ball-and-chain inactivation (Fig. 1c, d). We assigned the most N-terminal region of the β2 subunit to this density (residues 2–13) and we call it the “ball”, while the remainder ~20 unresolved amino acid residues connecting the ball to the first transmembrane domain of the β subunit is the flexible, disordered “chain” domain (dotted line in Fig. 1d). This conformation is an open-inactivated state.

The second class with 54% of particles was an intermediate state (Supplementary Fig. 8a, b). Despite Ca²⁺ being bound to the RCK domains, the gating ring remained in a resting, apo-like conformation⁷ (Cα root-mean-square-displacement, r.m.s.d., of 1.8 Å with the Ca²⁺-free hSlo1/β2_N-β4 individual gating ring). As a consequence, the Mg²⁺ binding site is not formed and Mg²⁺ is not bound despite being present in the buffer (Supplementary Fig. 8c). Although their overall structure is similar, the gating ring of the Ca²⁺-bound intermediate state is noticeably rotated compared to the Ca²⁺-free gating ring, when the structures are aligned at the transmembrane domain (the gating ring Cα r.m.s.d. ~16 Å). The S6 pore lining helices were straighter than in the open state, narrowing the intracellular gate, similar to the previously-reported divalent-free structure of BK channels⁷. Interestingly, the S6 helix is unwound in the middle of the membrane, at the level of the hinge glycine, suggesting that this conformation is on-path between closed and open channel states (Supplementary Fig. 8b). The gate opening is slightly more dilated in this intermediate structure than in the previously reported detergent closed state⁹ and the selectivity filter has 4 ion densities compared to only one in the closed state (Supplementary Fig. 8f, h). It is worth noting that the closed conformation of BK channels has a large pore cavity and intracellular gate, with an opening of ~10 Å in diameter, which is wide enough to allow passage of hydrated K⁺ ions⁹. Such a conformation is usually considered open in other channels³⁴. Despite the relatively large cavity and pore entrance, no density for the β2 ball is observed in the pore of this intermediate state (Supplementary Fig. 8b).

In MSP2N2 lipid nanodiscs containing POPC:POPE:POPS (5:3:2), the same 2 conformations were obtained as described for the detergent-solubilized complexes, with 49% of particles in open-inactivated state and 51% in intermediate state (Supplementary Figs. 5 and 6). One difference was that the density in the open pore that we had assigned to the β2 ball is smaller than in the open-inactivated detergent conformation (only 7 residues assigned), although still very strong and bulky (Supplementary Fig. 8g). No ball-like density was observed in the pore of the intermediate state (Supplementary Fig. 8d, e).

β2 ball enters the hSlo1 open pore via side access portals

The divalent-bound hSlo1/β2_N-β4 open-inactivated conformation in detergent micelles has the above-described elongated density inside the pore starting right beneath the selectivity filter and extending down close to Glu324 on the S6 pore-lining helix, but the chain connecting it to the transmembrane scaffold was not resolved (Fig. 1d, dashed line). We assigned the β2 ball to this density because exhaustive functional studies have shown that the N-terminal region of β2 is responsible for inactivating BK channels^15,29. However, since we could not unambiguously fit the residues to the rather featureless density nor identify the remaining loops of the chain (~20 unresolved residues indicated by the dashed line, Fig. 1d), we performed further data processing on the open state class.

After symmetry expansion, 3D classification with focused mask and 3D refinement in C1, we identified 3 classes, all identical except with different densities inside the pore (Supplementary Fig. 2). Class 1 was identical to the original open-inactivated structure discussed above, had the most continuous density inside the pore with a local resolution of ~2.6 Å to which we assigned the first 13 amino acid residues of the β2 ball (Fig. 1d). This class also had density for 9 residues extending from the TM1 of β2_N-β4 subunit from all the four protomers moving away from the RCK and bending upwards towards the lipid bilayer (Fig. 1d). Classes 2 and 3 had similar bulky, globular densities in the pore cavity under the selectivity filter assignable to the inactivating ball, although a bit smaller than in Class 1 (Fig. 1e, f). In addition, both classes 2 and 3 displayed densities extending from the end of the TM1 of one of the 4 β subunits towards a crevice that opens up between the transmembrane domains and the gating ring, called side access portal, through which the N-terminal β2 ball could sneak inside the channel pore^34,35 (Fig. 1c, arrow). We did not model any residues into these densities. However, these fragmented densities not only highlight the pathway by which the ball domain may access the pore but they may also represent alternative binding sites for the ball domain on its way to plugging the pore, as previously proposed^36,37 (Fig. 1e–g).

Additional processing, symmetry expansion, and 3D classification of the divalent-bound complex in nanodiscs did not lead to multiple classes, as in the detergent case (Supplementary Fig. 5). For most analyses to follow (unless stated otherwise), we will use the structure of the divalent-bound channel complex in detergent (class 1) since more of the inactivating ball was resolved.

Molecular dynamics simulations reveal interactions between the β2 ball and the pore

Assignment of the first 12 residues of the β2_N-β4 ball at the pore cavity was challenging, because the corresponding density had poor structural features (Figs. 1d and 2). A bulky but featureless density at the region could indicate that the ball is flexible and does not have a single dominant binding mode. This possibility is supported by structural investigations of soluble inactivating ball domains of several K⁺ channels, including the BK channel β2 subunit, which revealed a disordered, flexible structure in solution^36,38. Therefore, we built three different models for the peptide in three different binding modes where the r.m.s.d. of alpha carbons of the first three ball residues (FIW) in between the three models are 3.1~4.3 Å. (Fig. 2a–d, Supplementary Fig. 9b, c). No major clash with surrounding residues was found, as the clash score generated by Phenix real-space refinement was ~9 in all three models. We performed all-atom molecular dynamics (MD) simulation to examine flexibility of the ball, where three independent replicas were started for each of the three different ball models. The FIW residues rearrange slightly during the 100 ns equilibration runs, when the r.m.s.d. of alpha carbons of the FIW against their initial positions (t = 0) vary from 2 to 5 Å (Supplementary Fig. 9d–f). The root-mean-square-fluctuations (r.m.s.f.) of alpha carbons of the FIW residues against their average positions during the 400 ns run were below 4 Å, indicating that the three residues were still bound at the pore cavity but with some flexibility in the backbone. On the other hand, the rest of the N-terminal residues in the loop, from R8 to Y22, were considerably more dynamic than the first three (FIW) and did not make any stable contacts with surrounding residues in the pore cavity (r.m.s.f. values are greater than 6 Å, Fig. 2b–d). FIW made non-unique hydrophobic interactions mostly with T287, L312, and A316 of the pore-lining residues (Fig. 2e, f), allowing FIW to adopt multiple binding modes as hypothesized. Thus, the featureless ball density in our map could represent the ensemble of different modes.

Fig. 2. Flexible N-terminal β2 ball peptide makes non-unique hydrophobic interactions with the pore cavity.

a Cryo-EM density mesh (gray) of the ball peptide in conformation 1 (orange, same model as in Fig. 1d). b r.m.s.f. of alpha carbons of individual residues of the β2_N-β4 subunit in the simulations against conformation 1 in a. The coordinates are taken at t = 400 ns in replica #1. The first three N-terminal residues (FIW) bind to the bottom of the selectivity filter (green), then each residue of the β2_N-β4 is colored in scale by its r.m.s.f. value varying from 3 to 8 Å (bar). The inset shows the zoomed in view of the β2 ball beneath the selectivity filter. Two alternative conformations of the β2 ball (pink conformation 2 in c and blue conformation 3 in d) modeled in the density beneath selectivity filter (gray mesh) of Class 1 were used as starting models for additional simulations identical to those detailed in b, but with only the zoomed-in detail of the resulting ball peptide r.m.s.f. shown in boxes as the inset in b. e Surface representation of the ball peptide in conformation 1 and the channel pore area (gray) highlighting the ball-interacting residues from the pore lining helices. f Per-residue contact surface area between individual hSlo1 residues and FIW of β2 subunit is calculated using difference of solvent accessible surface area (SASA) of the hSlo1 and β2_N-β4 subunit complex between the presence and absence of β-subunit. Bars are the averages of n = 3 independent replicas, and error bars are standard deviations. The residues where average contact surface area is less than 1 Ų are excluded.

When FIW was substituted to GGG in our simulations starting from individual initial model, GGG escaped from the pore cavity within ~200 nanoseconds (Supplementary Fig. 9g, h). The simulation results agree with previous reports²⁹, showing that a peptide containing FIW as the first three residues followed by a chain of at least ~13 residues is sufficient for inactivation, but the first three residues must be bulky and hydrophobic.

β2 chain binds to the gating ring of the Ca²⁺-free BK channel

So far, we found that in the presence of the activating stimulus, Ca²⁺, one of the four β2 balls of the β2_N-β4 subunit enters the open pore of the BK channel, likely via side access portals, to bind to the pore and occlude ion conduction. We did not observe any density for the other three β2 ball domains. To determine the conformation of the inactivating β2_N-β4 subunit when the channel is closed and find where the inactivation ball may bind when not in the channel pore, we analyzed the structures of the hSlo1/β2_N-β4 complexes with cryo-EM in the absence of Ca²⁺ in both detergent and nanodiscs (Supplementary Figs. 4 and 7). We obtained a density map of 3.4 Å resolution (Fig. 3a) of the closed channel conformation whose overall architecture and structure is the same as that previously reported for hSlo1/β4 in the absence of Ca^{2+ 7,9}. There is no density for the β2 ball in the closed BK channel pore. However, we were able to identify density for 9 additional residues (36-TVTALKAG-43) for the β2 chain, continuing from the end of the TM1 helix of the β2_N-β4 subunit, which interact with the N- terminal region of the RCK1 (339-VSGR-342) domain (yellow region in Fig. 3a, b). Tiny changes in the RCK1 domain (in the loop that contacts the 9 residues of the visible β2 chain) were observed as a consequence of this interaction (an overall r.m.s.d. of ~1 Å with the previously reported gating-ring structures in the absence of Ca^2+ 7,9).

Fig. 3. N-terminal chain domain of the inactivating β2 subunit binds to the gating ring and not the pore of the Ca²⁺-free BK channel.

a, b Divalent-free, closed state structure of hSlo1 + β2_N-β4 in detergent (GDN). c, d Divalent-free, closed state structure of hSlo1 + β2_N-β4 in MSP2N2 nanodiscs. Segmented cryo-EM density map of hSlo1 channel in gray, β4 in blue, β2 inactivating chain in orange (a, c). The transparent detergent micelle and nanodisc and orange β2 densities are superimposed at a higher threshold for visibility. Modeled hSlo1 channel monomer (gray), β4 in blue, β2 inactivating chain in orange, RCK region interacting with β2 inactivating chain in green (b, d). Insets of b and d highlight the residues of β2 in orange interacting with the RCK residues in green. Shown are overlays of the gating ring conformers of the Ca²⁺-free (e) and Ca²⁺-bound (f) channels indicated in the figure panels in surface representation, viewed from the extracellular side (Ca²⁺-free: PDB 6V35, 9CZJ, and 9CZK; Ca²⁺-bound: PDB 6V22, 9CZQ, and 9CZM). hSlo1 + β4 is in gray, hSlo1 + β2_N-β4_N- (detergent) is in pink, and hSlo1 + β2_N-β4_N (nanodisc) is in cyan.

The Ca²⁺-free hSlo1/β2_N-β4 complex in nanodiscs (Supplementary Fig. 7) yielded a density map at 3.6 Å resolution with a conformation different from the Ca²⁺-free detergent complex (compare Fig. 3c, d with Fig. 3a, b, r.m.s.d. of ~13 Å). In fact, the closed Ca²⁺-free nanodisc structure is more similar to the intermediate state in the presence of Ca²⁺ (r.m.s.d. of ~2 Å), with a slightly wider pore than the closed Ca²⁺-free detergent structure (Fig. 4a, b, Supplementary Fig. 8f) and a gating ring rotated further relative to the transmembrane domain compared to the open state (Fig. 4c, Supplementary Fig. 8i). No density for the β2 ball was observed in this Ca²⁺-free conformation, with a presumably closed pore. However, in the Ca²⁺-free nanodisc structure, density for a longer β2 chain (~ 20 residues) extending from the end of the TM1 of the β2_N-β4 subunit (24-KIRDHDLLDKRKTVTALKAG-43, we modeled this region as a polyalanine chain due to low resolution), which interacts with the cytosolic gating ring (Fig. 3c, d). Because of the large gating ring rotation with respect to the transmembrane domain in this conformation compared to the closed state in detergent, the longer β2 chain interacts with a different region of the gating ring than in the detergent structure, near the second Ca²⁺ binding site in RCK2 (compare Fig. 3a, b with Fig. 3c, d). The different conformation observed for the Ca²⁺-free channel in nanodiscs could be related to the presence of lipids in the sample, suggesting that lipids influence BK channel gating, as previously proposed^39,40.

Fig. 4. Side portals as access points for the β2 ball into the BK channel pore.

a Overlay of the pore cavity structures (two opposing subunits shown for clarity) of closed (green), intermediate (purple), and open (pink) states to show the ball (orange, surface representation of conformation 1) may be able fit in all states, depending on its binding mode. b Pore radii vs. distance along the pore of closed, intermediate, and open states. c Overlay of gating rings of closed, intermediate, and open states, aligned on the TM domains highlighting the rotation relative to the TMs upon changing to intermediate or open states. Left, side view. Right, top view. d Views of the side portals (outlined in magenta) between the S6 helix and the gating ring in Ca²⁺-free closed, intermediate, and Ca²⁺-bound open BK channel structural models shown in surface representation. The β4 subunit is in blue and the visible regions of the N-terminus of β2 are in orange. The unresolved parts of the β2 chain and ball are rendered as dashed lines and orange sphere, respectively, only for the purposes of illustrating our hypothesis. The portal widens upon Ca²⁺-binding and pore opening due to the outward movements of both the gating ring and the S6 helix and allows entry of the ball in open but not in the closed or intermediate states. All structures are from detergent samples.

Our cryo-EM structural data suggests that the gating ring conformer is dependent on its interaction with the β2 N-terminus. The Ca²⁺-bound hSlo1 open and open-inactivated conformations are nearly identical across various conditions (with or without β subunits), with an average gating ring Cα r.m.s.d. of ~1 Å, whereas the Ca²⁺-free hSlo1 structures exhibit distinct heterogeneity in the gating ring (Fig. 3e, f). Here, we term all the Ca²⁺-free RCK conformations as the resting gating ring. Since no density was observed for the β2 ball domains in these conformations, we concluded that the ball does not bind anywhere else on the channel other than the open pore, similar to what was reported for MthK channels²⁶.

β2 ball cannot access the channel pore in the closed and intermediate states

We found it intriguing that the β2 ball was not observed to bind to either the closed or the intermediate BK channel pore conformations, which, although narrower than the open state, could be sufficiently wide (~10 Å diameter) to accommodate a β2 ball (for reference, we superimposed the model of the ball in the overlay of the open, intermediate and closed pores in Fig. 4a, b). We hypothesized that either: 1) the β2 ball assumes a (partially) folded state when bound to the pore, making it too large for the closed and intermediate states, or 2) the accessibility of the β2 ball to the inside of the pore cavity is impaired in the closed and intermediate states, despite the observed rotation of the gating ring (Fig. 4c). We investigated the latter by examining the possible points of access of the β2 ball to the inside of the channel pore. The β2 N-terminus comprising the ball and the chain is ~45 residues long and extends from the end of the TM1 of the β2 subunit (as shown in Fig. 4d by the orange ball cartoon extending via the dashed line chain). The chain is not long enough to wrap over and under the gating ring allowing the ball to reach the pore from below it. The only other possibility, also supported by the densities we identified above (Fig. 1e–g), is for the ball to enter the pore via the side access portals (Fig. 1c). These portals are narrower in the closed and intermediate states than in the open conformation (Fig. 4d). The side portals narrow by ~6 Å in the closed state as a consequence of the movement of the pore lining helices and the gating ring upon Ca²⁺ unbinding from the latter. The S6 pore-lining helices bend inwardly about ~10 Å when the channel closes (Fig. 4a, b) and the gating ring changes its conformation from activated to resting and rotates with respect to the transmembrane domain (Fig. 4c and Supplementary Movie 1). Such opening is not sufficient for ball domain entry, while the ~12 Å wider portal in the open state is.

Discussion

Ball-and-chain inactivation in voltage-gated ion channels was first reported more than 50 years ago and, over the years, major insights into how the process works at the molecular level have been gained through the efforts of many groups, and its generality to many types of ion channels emerged^{19,21,23–26,29,30,32,37,41–45}. The current model for ball-and-chain inactivation is the following: 1) the cytosolic N-terminus from either the channel itself or from an accessory subunit that co-assembles with the channel is necessary and sufficient for inactivation; 2) the inactivating N-terminal region is composed of a ball domain that plugs the channel pore and a flexible chain domain that connects the ball to the transmembrane region; 3) the ball is composed of ~20 residues with the first 10 residues predominantly hydrophobic and the second half mostly basic; 4) the amino acid composition and the length of the flexible chain is variable. Structural evidence, however, was missing.

In this study, we captured the key structural states involved in the β2-subunit-induced inactivation of human BK channels (Fig. 5). We showed that the N-terminus of the inactivating β2 subunit binds deep into the cavity of the Ca²⁺-bound BK channel hSlo1, to plug the open pore in a ball-and-chain-like manner. The binding is driven largely by hydrophobic interactions with the channel pore cavity, thus the orientation of the bound ball to the pore is not unique. While the first three N-terminal bulky hydrophobic residues of the ball are relatively rigidly bound underneath the selectivity filter, the remainder of the ball domain and the chain are flexible. In the absence of Ca²⁺, the N-terminus of the β2 subunit is no longer observed inside the channel pore cavity, but instead, makes interactions with the RCK domains of the cytosolic gating ring. We propose that when Ca²⁺ binds, the large conformational changes in the gating ring, including its rotation with respect to the transmembrane domain, lead to not only the dissociation of the ball domain from them but also to a large increase in the size of the side access portals into the pore, allowing the ball to sneak its way into the pore to plug it. Almost all aspects of this process were previously proposed or predicted based on clever functional experiments, and our data now provides the needed structural framework.

Fig. 5. Key states in the process of BK channel inactivation by the β2 subunit.

a In the absence of Ca²⁺, the β2 chain binds to the resting gating ring and the channel is closed. Upon Ca²⁺-binding to the gating ring, both an intermediate (b) and an open (c) state form. b In the intermediate state, the gating ring (Ca²⁺ is bound to both RCK sites although the gating ring is still in the resting conformation) is slightly rotated with respect to the transmembrane domain, the S6 helix becomes partially unwound around the glycine hinge and slightly more dilated than in the closed state, and the β2 chain binds at a different location on the gating ring. c In the open state, the gating ring is activated and the S6 helices bend at the hinge glycine to open the pore wider. We did not observe an open, not inactivated state in our study. d When the pore is open, the β2 ball enters the pore to inactivate the channel and may also bind at an alternate site that does not (completely) occlude the pore. This state is drawn slightly transparent to convey its uncertainty. e The β2 ball binds to the pore and plugs the permeation pathway to inactivate the channel. The model shown highlights the main conformational states structurally identified here and it is not meant to account for all previously reported functional data.

The structural correlates of the ball-and-chain inactivation induced by the N-terminus of the β2 subunit of the mammalian BK channel are the same with the only other published structural report on this type of inactivation, previously observed for a prokaryotic Ca²⁺-activated K⁺ channel homolog of BK channels, MthK²⁶. Although for MthK, inactivation is caused by the N-terminus of the channel itself, both channels appear to inactivate via the N-terminus (from either channel or accessory subunit) binding deep inside the pore cavity to sterically plug ion permeation. Ball-and-chain inactivation has been proposed to be a general mechanism of terminating flux through potassium channel pores⁴¹ and we provide here structural evidence for the functional conservation of ball-and-chain inactivation across different taxonomic domains (bacterial to mammalian) and across different types of inactivation (intrinsic²¹ vs. an accessory subunit^11,24). Accessory subunits associated with inactivation are diverse and not necessarily part of the same family of proteins. The β subunits of BK channels are transmembrane proteins¹¹, while β subunits of Kv channels are cytosolic²⁴. However, they all share a 20-residue long half-hydrophobic and half-hydrophilic inactivating N-terminus⁴¹. This suggests evolutionary conservation of this astonishingly simply designed mechanism of terminating ion conduction, and also suggests that other channels, such as Kv channels, use the same structural framework for inactivation.

We found that the N-terminus of β2 subunit is mostly flexible and binds degenerately to inactivate BK channels, in agreement with previous NMR studies of isolated inactivating domains showing substantial conformational heterogeneity^{36,38,46–48}. The disorder of the ball-and-chain domains, especially when not within the confines of the pore cavity, is also supported by our failure to detect density for most of it when not in the pore, except for a few residues in the chain that interact with the resting gating ring. This disorder may be necessary to facilitate the movement of the peptide to access the pore cavity through the narrow portals between the gating ring and the S6 transmembrane helix. We also found that the first three bulky hydrophobic residues bind the most rigidly within the pore cavity, in complete agreement with a mutational study²⁹ where the authors reported the same first three residues to be necessary and sufficient for binding, while the rest of the N-terminus can tolerate drastic changes to residue types and domain length while still maintaining its inactivating nature.

β subunit-induced inactivation in BK channels has been proposed to occur in two steps, where the ball domain can also bind to a different site proposed to be located near the pore, a so-called “pre-inactivated” conformation in which the ions are still able to permeate the pore, followed by a final step to occlude the pore^36,37,49. This is in contrast with the canonical one-step ball-and-chain inactivation observed for Shaker, and other Kv channels^21,22,30,50, although other studies support a two-step inactivation in Kv channels as well^34,43. Our structural data displays a strong density in the pore cavity, bound right under the selectivity filter and on the four-fold axis of the channel. However, some classes were found where additional density was observed along a putative pathway for N-terminus β2 entry into the pore through the side portals. Such additional density could not be unambiguously fit with specific residues of the N-terminus and is thus inconclusive, but it could represent either transiently captured parts of the chain, or part of the ball captured in such a pre-inactivated binding site.

Similar to previously-reported BK channel structures^7,9, our Ca²⁺-free closed BK channel complex structure, displays a resting gating ring and a pore cavity that is narrower than that observed in the open state, as a consequence of the straightening of the S6 pore-lining helices. This conformational change reduces the size of the opening of the intracellular entryway, although not sufficiently to sterically prohibit hydrated ion passage or even accommodate a peptide such as the ball domain in an extended configuration⁵¹. Nevertheless, we did not find any ball domain density in the pore of either the closed or the intermediate channel conformation. Although previous reports^16,31,52 suggested that closed BK channels can be inactivated by the β2 subunit, in apparent conflict with our findings, they also indicate that the voltage sensors may need to be in a permissive state for inactivation to occur, perhaps explaining the absence of ball from the pore in these states. It is possible that these studies may not be directly comparable and are complicated by differences in construct design.

The structures of the closed and intermediate states of the BK channel complexes presented here showed that part of the β2 subunit chain interacts with the gating ring in the resting state. Upon Ca²⁺ binding, the gating ring also undergoes a rigid body rotation relative to the transmembrane domain when the channel opens. This rotation, together with the associated conformational changes may be obstructed when the β2 subunit is present and interacts with the gating ring, and may contribute to the slower BK channel activation kinetics observed in the presence of β2^15,16. Slower activation was also reported with β1 and β4 subunits¹⁵, which could be occurring by a similar mechanism. Furthermore, we observed in our data an intermediate state, unlike previous reports with either BK channels alone or in complex with the β4 subunit, where only an open state was found. This intermediate state has a resting gating ring and a narrow intracellular gate opening similar to the closed state. However, the gating ring is rotated with respect to the transmembrane domain, the S6 helix is partly unwound near the hinging area responsible for channel opening, and the selectivity filter has 4 ions in the filter, like in the open state. These features are consistent with this conformation being an on-path, intermediate state that is more stable in the BK/β2 channel complex. The increased stability of a likely non-conductive intermediate on-path to activation may also contribute to the slower kinetics for activation and deactivation reported for the BK channels when in complex with β2^15,16.

Previous studies reported that although the most important features of the β subunit-induced inactivation of BK channels are shared with those described for the ball-and-chain inactivation of Shaker and other Kv channels, there are some that are not. For example, in Shaker channels, the ball was proposed to dissociate first from the open pore and pass through an open state before the channel can close, as evidenced by the presence of large tail currents upon membrane hyperpolarization^30,36. In BK/β2, no tail currents are observed upon hyperpolarization and the hypothesis was raised that BK channels can close with the ball bound^29,36. Although we did not detect any β2 ball domain inside the closed channel pore, this does not preclude the possibility that other closed states may exist that were not captured. Alternatively, the lack of tail currents in BK/β2 may simply reflect slower dissociation of β2 from the BK channel pore.

Inactivation in BK channels is important in many physiological processes, such as use-dependent spike broadening and daytime excitability in the circadian clock in neurons^27,28. This study provides structural insights into this fundamental channel gating property, helping to pave the way towards modulating such processes with targeted pharmacological interventions.

Methods

Constructs

Human Slo1 (hSlo1) used in this study was a generous gift from Roderick Mackinnon. hSlo1, with a deletion of 57 residues from the C-terminus, 1–1056, in a modified pEG BacMam vector⁵³ so that the resulting protein has green fluorescent protein (GFP) and a 1D4 antibody recognition sequence on the C-terminus, separated by a PreScission protease cleavage site.

The full length human β4 (residues 1–210) subunit of hSlo1 subcloned into a similarly modified pEG BacMam vector with mCherry and a deca-histidine affinity tag (mCherry-His10) replacing the GFP-1D4 fragment was also a gift from Roderick Mackinnon.

N-terminal 45 residues (MFIWTSGRTSSSYRHDEKRNIYQKIRDHDLLDKRKTVTALKAGED) of the synthetic gene fragment (Addgene plasmid #113558) encoding the β2 inactivating subunit of hSlo1 (residues 1–235), was cloned into a modified β4 subunit (with a deletion of the first 15 residues from the N-terminus) pEG BacMam vector with mCherry and a deca-histidine affinity tag (mCherry-His10) resulting in the β2_N-β4 chimera referred here as the β2_N-β4 subunit. The chimera was confirmed by sequencing (Genewiz).

Electrophysiology

HEK293S GnTI^- cells were cultured at 37 °C and 5% CO₂ in Dulbecco’s Modified Eagle Medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco), in the absence of antibiotics. Cells were plated on 35 mm tissue culture treated dishes and when confluency reached 50%, they were transfected with 250–350 ng of hSlo1 with 1 μl of Lipofectamine 3000 (Invitrogen) and 0.8 μl of P3000 solution (Invitrogen). For co-transfections, hSlo1 was combined 1:1 with 250–350 ng of β2 or β2_N-β4. The cell medium was changed to fresh DMEM 3-4 hrs post-transfection to remove the DNA-lipofectamine complexes.

Pipets were pulled from thick-walled borosilicate glass (Sutter Instrument, BF150-86-7.5) with a p97 puller (Sutter Instrument) to a resistance of 2–4 MΩ. Currents were acquired 24 hrs post-transfection using an EPC10 amplifier (HEKA Elektronik), filtered at 2.9 kHz, and digitized with the built-in LIH 8 + 8 converter (Instrutech Corp). The bath solution (140 mM KCl, 20 mM KOH, 10 mM HEPES, 5 mM HEDTA, 10 μM free Ca²⁺, pH 7.0) and the pipet solution (140 mM KCl, 20 mM KOH, 10 mM HEPES, 2 mM MgCl₂, pH 7.2) had a measured osmolarity of 281–294 mOsm/kg H₂O (Fiske MicroOsometer 210, Advanced Instruments). A free metal chelator calculator (Webmaxc Extended, https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/webmaxc/webmaxcE.htm) was used to determine the amount of CaCl₂ to add in the bath solution to obtain 10 μM free Ca²⁺.

Inside-out patches were exposed to 500 msec depolarizing voltage steps ranging from −100 mV to +140 mV at 20 mV intervals, with a holding potential of −80 mV. To synchronize the hSlo1 channels, prior to eliciting the voltage steps, a −180 mV pulse for a duration of 50 msec was used. Recordings were performed at room temperature using Patchmaster Next v1.3 (HEKA Elektronik) and analysis was performed in Igor Pro 8 (Wavemetrics) and MATLAB R2020b (Mathworks). For reproducibility, currents from 3 to 5 cells were measured. The ratio of steady state current to maximal current from each experiment was plotted as a function of voltage and the curves were fitted with sigmoidal functions: y = base + (max/(1 + exp((xhalf − x)/rate))) for hSlo1 + β2 and hSlo1 + β2_N-β4, or a linear function for hSlo1 alone (where x is voltage, and y is the ratio of steady state current and maximum current).

Sample preparation

hSlo1 was co-expressed with the β2_N-β4 subunit in HEK293S GnTI^- cells using the BacMam method⁵³. The expression and purification protocols were followed as mentioned in Tao et al., elife, 2019 with minor changes as mentioned below⁹. The bacmids carrying hSlo1 or β2_N-β4 subunit were generated by transforming E. coli DH10Bac cells with the corresponding pEG BacMam construct according to the manufacturer’s instructions (Bac-to-Bac; Invitrogen). Baculoviruses were produced by transfecting Spodoptera frugiperda Sf9 cells in ESF921 media with 3 μg of bacmid using Cellfectin II (Invitrogen) following manufacturer’s protocol. Baculoviruses, after two rounds of amplification in Sf9 cells using established protocols⁵³, were used for cell transduction. 2 liters of suspension culture of HEK293S GnTl^- cells was grown in FreeStyle 293 media with 2% heat-inactivated fetal bovine serum (FBS) and 1% PenStrep at 37 °C to a density of ~3 ×10⁶ cells/ml. For co-expression of hSlo1 and β2_N-β4 subunit, HEK293S GnTI^-suspension cell culture was infected with 10% (v:v) hSlo1 plus 10% (v:v) of β2_N-β4 subunit P3 baculoviruses to initiate the transduction. After 20 hr, 10 mM sodium butyrate was added and the temperature was changed to 30 °C. Cells were harvested by centrifugation at 5000 × g for 10 min at 4 °C, ~44 hr after the temperature switch.

For purification of the hSlo1 complexed with β2_N-β4 subunit, cells were gently disrupted by stirring in a hypotonic solution containing 10 mM Tris-HCl pH 8.0, 3 mM dithiothreitol (DTT), 1 mM EDTA supplemented with protease inhibitors including 0.1 mg/ml pepstatin A, 1 mg/ml leupeptin, Roche cocktail protease inhibitor, DNase and 1 mM phenylmethysulfonyl fluoride (PMSF). Cell lysate was then centrifuged for 45 min at 40,000 × g and pellet was homogenized in a buffer containing 20 mM Tris-HCl pH 8.0, 320 mM KCl, 5 mM EDTA, 15 mM MgCl₂ supplemented with DNase and protease inhibitors including 0.1 mg/ml pepstatin A, 1 mg/ml leupeptin, Roche cocktail protease inhibitor and 0.2 mM PMSF. The lysate was extracted with 1% GDN (Sigma) for 2 hr with stirring and then centrifuged for 60 min at 40,000 × g. Supernatant was added to GFP nanobody-conjugated affinity resin column (CNBr-activated Sepharose 4B resin from GE Healthcare) pre-equilibrated with wash buffer (20 mM Tris-HCl pH 8.0, 450 mM KCl, 5 mM EDTA, 15 mM MgCl₂, 0.02% GDN, 0.1 mg/ml 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE): 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC): 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphate (POPA) 5:5:1 (w:w:w), Roche cocktail protease inhibitor and PMSF).

The suspension was passed through the affinity column twice to ensure complete binding. The GFP-nanobody affinity column was washed with 50 column volumes of wash buffer in batch mode to ensure removal of nonspecific proteins. The protein was then digested on resin with PreScission protease (~20:1 w:w ratio) overnight with gentle rocking. Flow-through was then collected the following day, concentrated and further purified on a Superose-6 size exclusion column in 20 mM Tris-HCl pH 8.0, 450 mM KCl, 5 mM EDTA, 15 mM MgCl₂, 0.02% GDN and 0.05 mg/ml POPE:POPC:POPA 5:5:1 (w:w:w). All purification procedures were carried out either on ice or at 4 °C. The peak fractions corresponding to the tetrameric complex of hSlo1 and β2_N-β4 subunit was concentrated to about 12 mg/ml using 100,000 MWCO Amicon concentrators (Millipore) and used for preparation of cryo-EM sample grids. A total of 2.4 mg of the hSlo1 complex was obtained from a 2 L culture of ~10⁶ cells/ml.

Nanodisc reconstitution

The concentrated hSlo1/β2_N-β4 complex was reconstituted into nanodiscs with MSP2N2 and POPC:POPE:POPS (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine) (5:3:2) lipids in a ratio of 1:2:85 (Protein:MSP:Lipids). Lipid mixture composed of POPC:POPE:POPS (5:3:2) were prepared from chloroform stock solutions (Avanti Polar Lipids). Lipids were dried under constant nitrogen stream, rinsed with pentane to remove residual chloroform and dried again under nitrogen stream to form a lipid film. The lipid film was then dissolved in 20 mM Tris-HCl pH 8.0, 150 mM KCl, 2% w/v CHAPS by sonication in water bath to achieve a final concentration of 20 mM.

The concentrated hSlo1/β2_N-β4 complex was mixed with MSP2N2 (expressed from the plasmid #29520 from Addgene) and the lipid mixture in the molar ratio 1:2:85 (hSlo1 (β2_N-β4): MSP2N2: Lipids). After incubating for 1 hr at room temperature, bio-beads SM-2 (Bio Rad, 20 mg per 100 μl mixture) were added to remove the detergent and initiate reconstitution. The bio-beads were removed after 1 hr of gentle shaking at room temperature and the same amount of bio beads were added again to the solution for an additional 12 hr of gentle shaking at room temperature. Supernatant was collected and filtered through a 0.22 μm Spin-X centrifugation filter (Costar) before applying the sample to a Superose-6 size exclusion column in 20 mM Tris-HCl pH 8.0, 150 mM KCl at 4 °C. The nanodisc corresponding peak was collected, concentrated to about 10 mg/ml 100,000 MWCO Amicon concentrator (Millipore) and used for preparing cryo-EM grids.

Cryo-EM grid preparation and data collection

The nanodisc samples (~10 mg/ml) were supplemented with 3 mM Fos-choline-8, Fluorinated (Fos8-F, Anatrace) right before freezing. For the Ca²⁺-bound hSlo1/β2_N-β4 complex, 10 mM Ca²⁺ was added to the protein sample and incubated for 2 hr on ice before freezing on the grid. For the Ca²⁺-free hSlo1/β2_N-β4 complex, 5 mM EDTA was added to the protein sample right before freezing. 3.5 μl of purified protein sample was pipetted onto glow discharged UltraAufoil Au 300 mesh, R 1.2/1.3 holey carbon grids (Quantifoil). Grids were blotted at 15^oC for 4 s with a blotting force of 1 and humidity of 100% after a wait time of 10 s and flash frozen in liquid-nitrogen-cooled liquid ethane using a FEI Vitrobot Mark IV(FEI). Grids were then transferred to a FEI Titan Krios electron microscope operating at an acceleration voltage of 300 keV. Images were recorded on a Gatan K3 Summit detector (Gatan) set to super-resolution/counting mode.

Images of Ca²⁺-bound hSlo1/β2_N-β4 subunit complex in GDN detergent micelle and nanodisc were recorded with an energy filter of 20 eV at a super resolution pixel size of 0.4125 Å/pixel, defocus range of 0.6–3.0 µm, total dose is 48.45 e⁻/Ų. Images of Ca²⁺-free hSlo1/β2_N-β4 subunit complex in detergent were collected with an energy filter of 20 eV in counting mode at a pixel size of 0.83 Å/pixel, defocus range of 0.6–3.0 µm, total dose is 48.45 e⁻/Ų. Images of Ca²⁺-free (closed) hSlo1/β2_N-β4 subunit complex in nanodisc were recorded with an energy filter of 20 eV in super-resolution mode at a pixel size of 0.5413 Å/pixel, defocus range of 0.7–2.4 µm for a total accumulated dose of 47.08 e⁻/Ų.

Image processing

For all the samples, Ca²⁺-bound and Ca²⁺-free hSlo1/β2_N-β4 subunit complex both in detergent and nanodiscs, the processing was done in Relion 4.0.0⁵⁴.

For Ca²⁺-bound hSlo1/β2_N-β4 subunit complex in GDN detergent micelle dataset (Supplementary Figs. 2 and 3), 8478 movies were collected in super-resolution mode, dose weighted and binned by 3 using MotionCor2⁵⁵ in Relion 4.0.0⁵⁴. The contrast transfer function (CTF) on the resulting dose-weighted micrographs were determined by CTFFIND4⁵⁶. 1.45 million particles were picked by Cryolo 1.8.0⁵⁷ and extracted with a box size of 256 pixels. After 2 rounds of 2D classification, 0.55 million good particles were selected for ab initio model for the initial map reference. These particles were subjected to 3D classification with alignment, without symmetry into 8 classes, using the initial map lowpass-filtered to 60 Å as a reference. The classes without density in transmembrane or cytosolic region were discarded. 6 classes with 0.45 million particles were selected and the orientation and translational parameters for these selected particles were refined with the auto-refine algorithm of Relion without imposing symmetry, using the initial map lowpass-filtered to 40 Å as a reference and a mask surrounding the protein and excluding the detergent micelle. The resulting refined particle images were subjected to 3D classification without image alignment. Classes without density in transmembrane or cytosolic region were discarded.

Three classes with 134,000 particles had good density in TM and RCK regions but no density inside the pore beneath the selectivity filter. The particles in these classes which had no difference between them even after local refinement, were pooled and the orientation and translational parameters for these particles were unbinned with a box size of 384 pixels subjected to 3D refinement, CTF refined and polished with Bayesian polishing to get a map with a resolution of 3.2 Å. These classes are referred to as Ca²⁺-bound intermediate state because the RCK1 and RCK2 domains although bound with Ca²⁺ were still in the resting state and the S6 helix was similar to that in the detergent closed state (Supplementary Figs. 2 and 3).

One class with 133,000 particles had good density in TM and RCK regions and had an additional elongated density inside the pore beneath the selectivity filter. This class was further subjected to 3D classification without alignment using the initial map lowpass-filtered to 40 Å as a reference and with a mask surrounding the transmembrane domain to further separate the classes with particles having density in the pore region. 4 classes with 88,000 particles were selected which had elongated density in the pore along with density in the transmembrane helices. 3D refinement without symmetry followed by CTF refinement and Bayesian polishing per particle per micrograph, were performed, resulting in a map with a resolution of 2.7 Å. Particles were unbinned with a box size of 384 pixels and the map was 3D-refined in C4 symmetry with the initial map lowpass-filtered to 20 Å as reference to generate a map with resolution of 2.4 Å. These particles were further subjected to symmetry expansion. The symmetry-expanded star file was 3D classified in 6 classes without alignment, without symmetry, with the C1 map as the initial reference map lowpass-filtered to 20 Å, with a focused mask around the TM1 of one of the protomers of β2_N-β4 which included space needed for a loop that would extend from the TM1 up to the pore beneath the selectivity filter. Duplicate removal of the symmetry expanded particles was performed for each final class. However, since some particles may be the same in different classes, the final number of particles will be larger than what was in the beginning, which is what we define as the symmetry-expanded number of particles. Subsequently, orientation and translational parameters of the particles from the remaining classes were individually refined with local refinement without imposing symmetry, with the C1 map as the initial reference map lowpass-filtered to 10 Å and with a mask to exclude the detergent micelle. Class 1 had well-resolved elongated density (ball) inside the pore whereas classes 2 and 3 did not have as elongated a density in the pore as in Class 1 but they had additional density we hypothesized belongs to the inactivating loop extending from TM1 and moving up towards the adjacent protomer as well as additional unfeatured fragmented density towards the inside of the channel pore cavity, suggestive of metastable regions of the inactivating loop as it interacts with the channel, potentially indicating one pathway by which the ball domain accesses the pore. The orientation and translational parameters of the particles from these 3 classes were refined separately without imposing symmetry using the auto-refine algorithm of Relion resulting in final maps that achieved resolution of 2.9 Å, 2.93 Å and 3.01 Å respectively as assessed by Fourier shell correlation (FSC) using the 0.143 cut-off criterion (Supplementary Figs. 2 and 3).

The Ca²⁺-bound hSlo1/β2_N-β4 subunit complex in nanodisc was processed in a similar manner as the complex in detergent (Supplementary Figs. 5 and 6). For both open-inactivated structures (detergent and nanodisc), the particle image number reflects the symmetry expanded particle number.

For Ca²⁺-free hSlo1/β2_N-β4 subunit complex in detergent (Supplementary Fig. 4), 10,178 movies were collected in counting mode, dose-weighted and binned by 2 using MotionCor2⁵⁵. The CTF was determined with CTFFIND4⁵⁶. 2.8 million particles were picked by Cryolo 1.8.0⁵⁷ and extracted with a box size of 300 pixels. After 4 rounds of 2D classification, 93,000 particles were selected for ab initio model for the initial reference map. These 93,000 particles selected were subjected to 3D classification, with alignment, without symmetry into 8 classes, using the initial map lowpass filtered to 60 Å as a reference. The classes without density in transmembrane or cytosolic region were discarded. 4 classes with 66,000 particles were selected and the orientation and translational parameters for these selected particles were refined with the auto-refine algorithm of Relion without imposing symmetry and using the initial map lowpass-filtered to 40 Å as the reference. The refined particle images were further 3D classified without alignment into 6 classes using the same reference initial map. Only 1 good class with 40,000 particles had density in both the TM and RCK regions. This class was selected, 3D refined, and used for template-based picking by Autopick of Relion⁵⁴. The selected particles were extracted and subjected to 2D-classification to get additional classes in different orientations. This drastically increased the number of particles in good classes (around 250,000 particles). The orientation and translational parameters for these particles were refined using auto-refine algorithm of Relion imposing C4 symmetry and using the initial map lowpass-filtered to 40 Å as the reference map and with a mask around protein to remove the detergent micelle. The resulting refined particle images with increased density in TM region and a resolution of up to 4.0 Å was further subjected to 3D classification without alignment in 6 classes. From here the particles in classes with good density in TM region and the inactivating loop of the β2_N-β4 chimera was selected. The orientation and translational parameters for the selected particle images were refined using the auto-refine algorithm of Relion and a map of resolution 3.85 Å was obtained. To improve the resolution, the particles were CTF refined and polished with Bayesian polishing to get a map with a resolution of 3.43 Å. This map was further subjected to 3D classification without alignment with a mask around TM region of the protein only to separate the classes with a good density of the inactivating loop. Three classes with 74,000 particles with a clear density in the inactivating loop and well resolved TM region without much tilt in the TM regions were selected and their orientation and translational parameters were refined imposing C4 symmetry with the auto-refine algorithm of Relion resulting in a 3.40 Å resolution map as assessed by the 0.143 FSC cut-off criterion. All masks for post processing excluded the detergent micelle (Supplementary Fig. 4).

The Ca²⁺-free hSlo1/β2_N-β4 subunit complex in nanodisc was processed in a similar way as for the Ca²⁺-free hSlo1/β2_N-β4 subunit complex in detergent (Supplementary Fig. 7).

Model building and refinement

The model for the open-inactivated Ca²⁺-bound hSlo1/β2_N-β4 complex in detergent was built by docking the PDB 6V22 for open state hSlo1/β4 complex⁹ in the density map using UCSF ChimeraX (version 1.3)⁵⁸ and then manually rebuilt in Coot (version 0.8.9.2)⁵⁹ to fit the density. For the inactivating ball and chain, additional residues of the β2_N-β4 from TM1 and also inside the pore beneath the selectivity filter were manually built in Coot⁵⁹ following the backbone density and assigning the larger side chains of amino acid residues first. This model built in Coot was subjected to real space refinement with minimization global, occupancy and morphing parameters while applying secondary structure constrains and non-crystallographic symmetry in Phenix (version 1.20.1)⁶⁰. Ca²⁺ and Mg²⁺ were placed in RCK1 and RCK2 in their respective densities. Rotamer outliers were fixed and validation of the final structures were done by Molprobity and EMRinger⁶¹. The final model after a few iterations of real space refinement and manual rebuilding has good geometry and contains amino acids from the inactivating loop of β2.

The model for open-inactivated, Ca²⁺-bound hSlo1/β2_N-β4 complex in nanodisc was built by docking the structure we built for the open-inactivated, Ca²⁺-bound hSlo1/β2_N-β4 complex in detergent micelle (solved above), in the density map of the open-inactivated complex in nanodisc. The nanodisc open state structure was very similar to the detergent open state structure.

Model building for the Ca²⁺-free hSlo1/β2_N-β4 complex in detergent micelle was done by docking the PDB 6V35 for closed state hSlo1/β4 complex⁹ in the density map for the Ca²⁺-free hSlo1/β2_N-β4 complex using UCSF ChimeraX (version 1.3)⁵⁸ and manually rebuilt in Coot (version 0.8.9.2)⁵⁹ to fit the density^58,59. For the inactivating loop, additional residues of the β2_N-β4 subunit were manually built in Coot (version 0.8.9.2)⁵⁹ following the backbone density and assigning resolved residues sidechains. The hSlo1/β2_N-β4 complex model was subjected to real space refinement in Phenix (version 1.20.1)⁶⁰ with minimization global, occupancy and morphing parameters while applying secondary structure constrains and non-crystallographic symmetry. Rotamer outliers were fixed and validation of the final structures were done with Molprobity and EMRinger⁶¹. The final model after a few iterations of real space refinement and manual rebuilding has good geometry and contains 9 amino acids from the inactivating chain of β2 subunit.

The model for the Ca²⁺-free hSlo1/β2_N-β4 complex in nanodisc was built by docking the model of Ca²⁺-free complex in detergent (solved above) into the density map for the complex in nanodisc using UCSF ChimeraX (version 1.3)⁵⁸. The additional residues of the inactivating loop were manually built in using Coot (version 0.8.9.2)⁵⁹ and refined for several iterations imposing secondary structure restrains and non-crystallographic symmetry in Phenix (version 1.20.1)⁶⁰. The final model after a few iterations of real space refinement and manual rebuilding has good geometry and contains 20 amino acids from the inactivating loop of β2 binding at the RCK.

The Ca²⁺-bound intermediate structure in detergent micelle and nanodisc were determined by docking in the Ca²⁺-free hSlo1/β2_N-β4 complex in nanodisc (solved above) using UCSF ChimeraX (version 1.3)⁵⁸ and then by manually rebuilding the model using Coot (version 0.8.9.2)⁵⁹ and refining using Phenix (version 1.20.1) real space refinement⁶⁰ to get the final models for Ca²⁺-bound intermediate state of hSlo1/β2_N-β4 complex in the detergent micelle and nanodisc.

Molecular dynamics simulation

The cryo-EM coordinates of the three alternate conformations of the N-terminus of the inactivating ball in the pore modeled in the density beneath the selectivity filter, in the open-inactivated state, conformation #1 (Fig. 2a), #2 (Fig. 2c) and #3 (Fig. 2d) were used as initial inputs for MD simulations. Each of the conformations consisted of four protomers of each of hSlo1 and β2_N-β4 subunit, with the N-terminus of one of β-subunit protomers blocking the pore of hSlo1. 12 N-terminal β-subunit residues (F2-Y13) were modeled in different conformations in #1-#3. 17 missing residues after H15 (D16-E32) were built as random coil using modeller version 10.4⁶². Truncated sidechains of other residues were rebuilt with the psfgen tool in VMD software version 1.9.3⁶³. Disulfide bonds were formed between C84-C178, C98-C149, C102-C106, and C114-C143 in β-subunits, as they were defined in the cryo-EM models. The simulation system was constructed using membrane builder tool of the CHARMM-GUI website (http://www.charmm-gui.org/)⁶⁴, where the hSlo1/β2_N-β4 subunit complex was embedded in a lipid membrane consisting of 410 POPC, 246 POPE, and 164 POPS molecules in both sides of the bilayers, which are roughly in 5:3:2 ratio, to mimic the lipid composition of the nanodisc. The protein-lipid complex is solvated with ~141,000 water molecules, and ~460 K⁺ and ~250 Cl^- ions were added in the solvent space to mirror the physiological ionic strength (100 mM). Four K⁺ in the selectivity filter, and one Mg²⁺ and two Ca²⁺ at the binding sites of an hSlo1 tetramer in the cryo-EM coordinates were included in the simulation system (Supplementary Fig. 9a). The system contains ~600,000 atoms in total. The simulation box was set to be orthorhombic with periodic boundaries applied at x-y-z axes and dimensions of 180 Å × 180 Å × 190 Å. CHARMM36 force field⁶⁵ was employed for the protein, lipids, and ions, and the TIP3P model⁶⁶ was used for waters. pKa values calculated by PropKa version 3.1⁶⁷ of all acidic and basic residues were smaller (or larger) than the system pH, which was set to be 7, therefore the protonation state of all acidic and basic residues were set to be their default state. All equilibration and production simulations were performed and analyzed with Gromacs package version 2022.3⁶⁸. Initial energy minimization and equilibration steps were performed following the CHARMM-GUI setup. Three replicas were generated for each conformation (#1, #2 and #3) by assigning initial velocities at 300 K using different random seed at the beginning of the equilibration step. The position restraints on protein and lipid were gradually released during 100 ns equilibration run, followed by 400 ns production run for each replica. For the GGG mutant, starting from t = 0 of each conformation of wild-type (WT), F2, I3, and W4 of β2_N-β4 subunit blocking the selectivity pore, were substituted by Gly residue, then each conformation was re-equilibrated following the same protocol as in the WT simulation. One replica of 400 ns production run was generated for the GGG mutant. All other simulation setup details were taken from our previous work⁶⁹. The initial and final coordinates of all systems listed in Supplementary Table 2 are available at 10.5281/zenodo.13258001.

Contact surface area was defined as the difference between the sum of the solvent accessible surface area (SASA) of the isolated protein and the N-terminus ball of the β2_N-β4 subunit following the method used in a previous work⁷⁰. Per-residue contact surface area was calculated between all heavy atoms of FIW of β2_N-β4 subunit and individual pore lining residues of hSlo1.

Reporting summary

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

Author contributions

S.A. and C.N. designed the study. S.A. prepared cryo-EM samples, collected and analyzed cryo-EM data. E.K. performed the electrophysiology experiments and analyzed the data, and generated the morph. A.S. performed initial electrophysiology experiments. S.L. and A.A. performed and analyzed the MD simulations. S.A., E.K. and C.N. assembled the manuscript and wrote the paper with input from all authors.

Peer review

Peer review information

Nature Communications thanks Show-Ling Shyng who co-reviewed with Camden Driggers; and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Data supporting the findings of this manuscript are available from the corresponding authors upon request. The maps of all the calcium-free and calcium-bound hSlo1 + β2_N-β4 complex in detergent micelle and in nanodiscs have been deposited with Electron Microscopy Data Bank (EMDB) with accession codes. Atomic coordinates for all the hSlo1 complex structures have been deposited with the Protein Data Bank (PDB). The atomic coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 9CZJ (hSlo1 + β2_N-β4 complex structure without calcium in detergent); 9CZK (hSlo1 + β2_N-β4 complex structure without calcium in nanodisc); 9CZQ (open inactivated hSlo1 + β2_N-β4 complex structure in the presence of calcium in detergent); 9CZM (open hSlo1 + β2_N-β4 complex structure in the presence of calcium in nanodisc); 9CZH (intermediate hSlo1 + β2_N-β4 complex structure in the presence of calcium in detergent); 9CZO (intermediate hSlo1 + β2_N-β4 complex structure in the presence of calcium in nanodisc); 9D18 (conformation 2 of the calcium bound inactivating domain in open inactivated state of the hSlo1 + β2_N-β4 complex in detergent); and 9D19 (conformation 3 of the calcium bound inactivating domain in open inactivated state of the hSlo1 + β2_N-β4 complex in detergent). The cryo-EM maps have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes EMD-46418 (hSlo1 + β2_N-β4 complex structure without calcium in detergent); EMD-46419 (hSlo1 + β2_N-β4 complex structure without calcium in nanodisc); EMD-46425 (open inactivated hSlo1 + β2_N-β4 complex structure in the presence of calcium in detergent); EMD-46421 (open inactivated hSlo1 + β2_N-β4 complex structure in the presence of calcium in nanodisc); EMD-46416 (intermediate hSlo1 + β2_N-β4 complex structure in the presence of calcium in detergent); EMD-46423 (intermediate hSlo1 + β2_N-β4 complex structure in the presence of calcium in nanodisc); EMD-46467 (conformation 2 of the calcium bound inactivating domain in open inactivated state of the hSlo1 + β2_N-β4 complex in detergent); and EMD-46468 (conformation 3 of the calcium bound inactivating domain in open inactivated state of the hSlo1 + β2_N-β4 complex in detergent). This study refers to the previously published models of hSlo1 + β4 complex PDB 6V22 (open inactivated hSlo1 + β4 complex structure in the presence of calcium in detergent); and PDB 6V35 (Ca²⁺ free hSlo1 + β4 complex structure in the absence of calcium in detergent). For the MD simulations, the initial and final coordinates of all systems listed in Supplementary Table 2 are available at [10.5281/zenodo.13258001]. The source data for Figs. 1b and  2f are provided as a Source Data file. Raw traces of patch clamp recordings are available from the corresponding author upon request. Sample raw data traces are provided in the corresponding figure and electrophysiology data are not deposited due to their large and complex size and requiring the use of proprietary scientific software. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-56844-4.