Research ArticleBIOPHYSICS

A unique mechanism of inactivation gating of the Kv channel family member Kv7.1 and its modulation by PIP2 and calmodulin

See allHide authors and affiliations

Science Advances  18 Dec 2020:
Vol. 6, no. 51, eabd6922
DOI: 10.1126/sciadv.abd6922

Abstract

Inactivation of voltage-gated K+ (Kv) channels mostly occurs by fast N-type or/and slow C-type mechanisms. Here, we characterized a unique mechanism of inactivation gating comprising two inactivation states in a member of the Kv channel superfamily, Kv7.1. Removal of external Ca2+ in wild-type Kv7.1 channels produced a large, voltage-dependent inactivation, which differed from N- or C-type mechanisms. Glu295 and Asp317 located, respectively, in the turret and pore entrance are involved in Ca2+ coordination, allowing Asp317 to form H-bonding with the pore helix Trp304, which stabilizes the selectivity filter and prevents inactivation. Phosphatidylinositol 4,5-bisphosphate (PIP2) and Ca2+-calmodulin prevented Kv7.1 inactivation triggered by Ca2+-free external solutions, where Ser182 at the S2-S3 linker relays the calmodulin signal from its inner boundary to the external pore to allow proper channel conduction. Thus, we revealed a unique mechanism of inactivation gating in Kv7.1, exquisitely controlled by external Ca2+ and allosterically coupled by internal PIP2 and Ca2+-calmodulin.

INTRODUCTION

Inactivation is an intrinsic property of numerous voltage-gated K+ channels (Kv), reflecting their crucial role in neuronal and cardiac excitability. Inactivation of most Shaker-like Kv channels occurs by N-type or/and C-type mechanisms. The fast N-type inactivation involves an intracellular peptide domain at the N terminus of α or β subunits, which occludes the channel inner pore and prevents ion permeation (15). The slow C-type inactivation is suggested to involve highly cooperative structural rearrangements in the outer pore, leading to a loss of K+ coordination sites in the selectivity filter (2, 3, 6, 7). The biophysical hallmarks of C-type inactivation include its inhibition by high external K+ or external tetraethylammonium (13, 8). However, the molecular mechanisms underlying these filter conformational changes remain controversial with studies, suggesting filter constriction and others proposing pore dilation (913).

It is still unclear whether other inactivation mechanisms different from N and C types operate in the Kv channel superfamily. Several groups including ours showed that inactivation of Kv7.1 channels (or KCNQ1) does not exhibit the hallmarks of N- or C-type mechanisms, notably the lack of effect of high external K+ (1418). The Kv7.1 channel is the only member out of five of the Kv7 subfamily of Kv channels that undergoes inactivation. Similar to all Kv channels, Kv7.1 is a tetramer, and each subunit features six transmembrane segments (S1-S6) containing a voltage-sensing domain (VSD) (S1-S4) and a pore module (S5-S6) (19, 20). Phosphatidylinositol 4,5-bisphosphate (PIP2) and calmodulin (CaM) are among the many signaling molecules that modulate Kv7.1 channels. While CaM interacts with coiled-coil helices A and B of the Kv7.1 proximal C terminus, PIP2 can bind to clusters of basic residues in S2-S3 and S4-S5 linkers, before helix A and helix B (2127).

Inactivation of wild-type (WT) Kv7.1 channels is not visible macroscopically but can be revealed by hooked tail currents upon repolarization, which reflects the fast recovery from an inactivation process (16, 18). Several mutants of Kv7.1 exhibit a clearly visible, voltage-dependent slow inactivation, in addition to hooked tail currents (14, 17, 28). However, the mechanisms underlying Kv7.1 inactivation gating remain obscure yet.

Here, we delineate a unique molecular mechanism of intrinsic inactivation gating in WT Kv7.1 channels. We show that removal of external Ca2+ produced a large, noticeable voltage-dependent inactivation, which does not share the attributes of N- or C-type mechanisms. Increasing external Ca2+ prevented Kv7.1 inactivation gating with a median inhibitory concentration (IC50) of 1.5 μM but did not affect the hooked tail currents, suggesting that Kv7.1 exhibits two discrete inactivation states, I1 and I2. Mutagenesis, molecular dynamics (MD) simulations, and electrophysiological studies suggest that two neighboring residues, Glu295 and Asp317 located, respectively, in the turret and filter external entrance of the channel pore together with water molecules are engaged in coordinating external Ca2+ ions. In the absence of external Ca2+, Asp317 side chain is quite flexible and unable to engage in H-bonding interaction with the pore helix residue W304, which destabilizes the selectivity filter and leads to inactivation I1. External acidification, which protonates E295 and D317 prevented Kv7.1 inactivation. Calcified CaM or PIP2 prevented inactivation gating, evoked in Ca2+-free external solutions, suggesting that PIP2 and Ca2+-CaM are not only important for VSD-pore coupling but are also crucial for preventing the activation-inactivation transition. Ser182 at the S2-S3 intracellular linker plays a pivotal role in transmitting the CaM signal to the outer pore to allow correct channel gating and pore conduction. Our results disclose a novel inactivation mechanism whereby external Ca2+ delicately controls intrinsic inactivation of a Kv channel that is allosterically modulated by PIP2 and calcified CaM at the inner face of the channel transmembrane core.

RESULTS

Removal of external Ca2+ elicits a large, voltage-dependent slow inactivation of WT Kv7.1 channels

Metal ions are known to affect ion channels either by blocking their permeation pathway or by modifying their gating properties (29). Here, we examined the effect of removing external divalent cations, Mg2+ and Ca2+, on Kv7.1 inactivation gating. We expressed WT Kv7.1 channels in Chinese hamster ovary (CHO) cells, where the same cell was successively exposed to two different external solutions, the first containing 1.8 mM Ca2+ and 1.2 mM Mg2+ and the second containing 2 mM EGTA and lacking Ca2+ and Mg2+ (0 Ca2+/0 Mg2+). The pipette internal solution contained nominal zero free Ca2+ [5 mM 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA), with no added Ca2+]. Upon depolarization, the percentage of visible inactivation was quantified from the ratio current of the lowest point after the decay of the transient component (dip) and the peak amplitude at +60 mV. In the presence of 1.8 mM Ca2+/1.2 mM Mg2+ solution, a small inactivation of 8% was detected, while in 0 Ca2+/0 Mg2+, the inactivation significantly increased to 62% (Fig. 1, A and C). The presence of 1.2 or 3 mM Mg2+ only (in 0 Ca2+) produced a significant inactivation of 41 and 35%, respectively (Fig. 1, B and C). In contrast, the presence of 3 mM Ca2+ (in 0 Mg2+) totally prevented the inactivation process (Fig. 1, B and C). These results suggest that the specific lack of external Ca2+ triggers the large visible inactivation of WT Kv7.1. Removal of external Ca2+ did not change the voltage dependence of channel activation in the presence of 1.2 mM Mg2+ (Fig. 1D). Addition of external Ca2+ suppressed in a concentration-dependent manner Kv7.1 inactivation with an IC50 of 1.5 μM Ca2+ (fig. S1, A and B). Since C-type inactivation is known to be prevented by high external potassium (13, 8), we examined whether high external potassium could prevent Kv7.1 inactivation. In the presence of 50 mM K+, the external solution containing 1.8 mM Ca2+/1.2 mM Mg2+ produced a 22% inactivation, while that containing 0 mM Ca2+/1.2 mM Mg2+ yielded a significantly larger inactivation of 55% (fig. S1, C and D). Thus, high external K+ does not preclude inactivation, suggesting that the Ca2+ binding site is not directly located in the permeation pathway and that Kv7.1 inactivation is different from the C-type mechanism. Divalent cations smaller than Ca2+ in their ionic radii such as Mg2+ and Mn2+ were ineffective in preventing Kv7.1 inactivation at similar concentrations. Higher concentrations of Mg2+ (3 mM) can somewhat reduce inactivation (Fig. 1C), suggesting some effectiveness but at much lower affinity compared to Ca2+. In contrast, Sr2+ and Cd2+ mimicked the effects of Ca2+ in preventing inactivation, due to their similar ionic radii and calculated hydration enthalpies (fig. S1, E and F).

Fig. 1 Removal of external Ca2+ elicits a large, voltage-dependent slow inactivation in WT Kv7.1 channels.

(A and B) Representative traces of WT Kv7.1 currents recorded from transfected CHO cells with pipette solution containing 5 mM BAPTA in nominally zero free Ca2+ concentration. The concentrations of Ca2+ and Mg2+ of the extracellular solutions are indicated in the panels. From a holding potential of −90 mV, cells were stepped for 3 s from −60 to +60 mV in 10 mV of increments and repolarized for 1.5 s to −60 mV. (C) The percentage of visible voltage-dependent inactivation was quantified from at +60 mV. Results are expressed as means ± SEM. In external solutions containing 1.8 mM Ca2+/1.2 mM Mg2+ (n = 10), 0 mM Ca2+/0 mM Mg2+ (n = 5), 0 mM Ca2+/1.2 mM Mg2+ (n = 13), 0 mM Ca2+/3 mM Mg2+ (n = 4), and 3 mM Ca2+/0 mM Mg2+ (n = 4), the inactivation was, respectively, 8 ± 3, 62 ± 7, 41 ± 4, 35 ± 3, and 0 ± 0% [one-way analysis of variance (ANOVA), F(4, 31) = 31.20, ****P < 0.0001; Dunnett’s multiple comparisons test, adjusted ****P < 0.0001 and **P = 0.0013, respectively]. (D) Normalized conductance-voltage relations of WT Kv7.1 currents in 1.8 mM Ca2+/1.2 mM Mg2+, V50 = −25.8 ± 1.5 mV, and in 0 mM Ca2+/1.2 mM Mg2+, V50 = −28.1 ± 2.5 mV; n = 10.

WT Kv7.1 channels populate two distinct inactivation states

In the presence of external Ca2+ (1.8 mM), inactivation of WT Kv7.1 is barely visible; however, these channels exhibit hooked tail currents, which reflects the fast recovery from an undetected inactivation process (16, 18). In various Kv7.1 mutants, an additional visible inactivation is observed (14, 17, 28). We and others suggested that these two types of inactivation are discrete processes, displaying distinct kinetics and pore properties (14, 15). Here, we show that these two distinct inactivation states are intrinsic to WT Kv7.1 channels and could be revealed in Ca2+-free external solutions. We compared the visible inactivation we called I1 with the undetected inactivation we called I2, whose fast recovery is seen by the hooked tail currents, and we examined their respective relaxation kinetics, their time constants of recovery, and their voltage-dependence (figs. S2 to S4). First, we noticed that the amplitude of I1 was unrelated to that of I2 (fig. S2A). The percentage of inactivation I2 was measured from the hooked tail currents. Channel availability is given by the ratio x/y, where x is the initial tail current amplitude and y the amplitude of the extrapolated single-exponential fit of the tail current (30). The percentage of inactivation I2 is given by 1-x/y (fig. S2, A and B). While in Ca2+-free external solutions, inactivation I1 was much larger than that in Ca2+-containing solution (41% versus 8%, respectively), inactivation I2 was very similar in either solution (fig. S2, A and C, and Fig. 1, A to C), suggesting that I1 and I2 reflect two discrete inactivation states.

The relaxation time course of inactivation I1 was recorded in Ca2+-free external solutions by stepping the membrane potential from −90 to +50 mV and was measured by one exponential fit of the decaying trace, which yielded a single exponential time constant of 45 ms (fig. S3, A and B). To measure the inactivation time constant of inactivation I2, we used a triple pulse protocol. A first depolarizing pulse of +50 mV opened the channels of which a yet undetected portion inactivated; then, a brief (15 ms) hyperpolarizing interpulse at −130 mV enabled fast recovery from channel inactivation; a third depolarizing pulse at +50 mV reopened and reinactivated the channel population that previously underwent inactivation, leading to an initial extra current. The fast relaxation measured at the third pulse corresponded to the kinetics of inactivation I2 with a single exponential time constant of 16 ms, which was significantly smaller than that of I1 (fig. S3B).

Then, the recovery time from the two types of inactivation was determined. The recovery from inactivation I1 was determined in Ca2+-free external solutions by a twin pulse protocol, where the cell was depolarized to +50 mV for 3 s to open and inactivate the channels, then repolarized to allow recovery at −60 mV with increasing time intervals, and restepped to +50 mV for 1.5 s to reopen and reinactivate Kv7.1. The recovery potential of −60 mV was chosen to compare the recovery time from inactivation I1 and I2. The recovery from inactivation I1 yielded a time constant of 244 ms (fig. S3, C and D). The recovery from inactivation I2 was determined by stepping the cell from −90 to +50 mV for 3 s and then repolarizing to −60 mV for 1.5 s. The recovery time from inactivation I2 was determined by the single exponential fit of the rising phase of the hooked tail currents at −60 mV and yielded time constants of 40 and 38 ms in Ca2+-containing and Ca2+-free external solutions, respectively, both values being significantly lower than that of recovery from inactivation I1 (244 ms) (fig. S3, E and F).

To measure the voltage dependence of inactivation I1, we used a classical steady-state inactivation protocol in Ca +-free external solutions. The cell membrane was stepped to inactivating prepulses (2 s) from −90 to +40 mV in 10 mV of increments, before a 150-ms test pulse to +40 mV was given (fig. S4A). Because inactivation I2 was not visible macroscopically, we used a triple pulse protocol to study its voltage dependency in Ca2+-containing external solutions. In this protocol, a 2-s conditioning prepulse to varying potentials (from −40 to +40 mV, in 10 mV of increments) was given, then a brief (15 ms) hyperpolarizing interpulse to −130 mV allowed fast recovery from inactivation before a test pulse of 150 ms to +40 mV was applied to reopen and reinactivate the channels (fig. S4B). The voltage dependence of inactivation I1 and I2 yielded significantly different values of V50 = −24.8 mV and V50 = −11.0 mV, respectively (fig. S4C). These results indicate that WT Kv7.1 channels exhibit two discrete inactivation states, I1 and I2, whose relaxation kinetics, time constants of recovery, and voltage dependencies are different.

To attest further the existence of two distinct inactivation states, we examined the impact of external Ca2+ removal on the pore helix mutant, Kv7.1 V307L, which was previously shown to abolish the hooked tail currents, thereby eliminating inactivation I2 (17). In agreement with this previous study (17), we found that Kv7.1 V307L exhibited slower activation kinetics compared to WT and did not show hooked tail currents, indicating that inactivation I2 was suppressed by the mutation (Fig. 2, A and B). When Kv7.1 V307L mutant was exposed to a Ca2+-free external solution, a large, voltage-dependent inactivation was noticeable (Fig. 2, B and C). These data clearly indicate that the two inactivation states of Kv7.1 can be easily discriminated by the mutation V307L, where inactivation I1 remains unaffected while inactivation I2 is abolished. Previous studies showed that inward tail currents of homomeric Kv7.1 channels are increased about threefold upon substitution of 100 mM K+ in the external solution with 100 mM Rb+, whereas heteromeric Kv7.1-KCNE1 channels have smaller inward Rb+ currents compared to K+ currents (31). It was found that the extent of inactivation and the ratio of Rb+/K+ inward currents were positively correlated, suggesting that inactivation of homomeric Kv7.1 channels reflects a fast flickery process of a late open state that depends on the occupancy of the pore by Rb+ (31, 32). We asked whether inactivation I1 triggered by removal of external Ca2+ exhibits a positive Rb+/K+ inward current ratio, similar to that found in homomeric Kv7.1channels with their related inactivation I2. To avoid a contamination of both inactivation I1 and I2, we probed the Rb+/K+ conductance ratio in the mutant V307L, which exhibit only inactivation I1 in Ca2+-free external solutions. The Rb+/K+ inward current ratio was measured in the same cell sequentially perfused with an external solutions containing either 130 mM K+ or 130 mM Rb+ in the absence of Ca2+ using a tail voltage protocol (Fig. 2, D and E). Results show that in contrast to inactivation I2 of homomeric Kv7.1, inactivation I1 of mutant V307L does not exhibit a threefold Rb+/K+ inward current ratio. Instead, at −100 mV, the Rb+/K+ current ratio is slightly lower than one (Fig. 2, D and E). In addition, the extent of inactivation I1 measured under these conditions at +60 mV was significantly larger in Rb+ than in K+ solutions (Fig. 2, D and F). Thus, while under conditions of inactivation I2, rubidium decreases the probability of the channel to be in the flicker-blocked state and thus increases the macroscopic inward tail currents, inactivation I1 triggered by removal of external Ca2+ does not appear to affect the Kv7.1 flicker pore properties.

Fig. 2 Effects of external Ca2+ removal on inactivation of the Kv7.1 pore helix mutant V307L.

(A) Mapping the pore helix residue V307 onto the human Kv7.1 cryo-EM structure (PDB: 6UZZ). SF, selectivity filter. (B) Representative traces of WT Kv7.1 and V307L mutant currents recorded using a similar protocol to that of Fig. 1A, with pipette solution containing 5 mM BAPTA in nominally zero free Ca2+ concentration. The extracellular solution contained either 1.2 mM Mg2+/1.8 mM Ca2+ or 1.2 mM Mg2+/0 mM Ca2+. Insets: Tail currents measured at −60 mV. (C) For WT Kv7.1, there is a significant increase in inactivation in Ca2+-free external solutions from 7 to 33% (n = 9; two-tailed paired t test; ***P < 0.0001, t = 8.060, df = 8); for the V307L mutant, there is also a significant increase in inactivation in Ca2+-free external solutions from 0 to 34% (n = 6; two-tailed paired t test; *P = 0.0188, t = 3.424, df = 5). (D) Representative traces of a CHO cell expressing V307L mutant currents and sequentially perfused with a Ca2+-free external solution containing either 130 mM K+ or 130 mM Rb+. From a holding potential of −90 mV, the membrane was stepped to +60 mV for 2 s and then the tail current was recorded for 1.5 s from +60 to −100 mV in 10 mV of decrements. (E) Normalized current-voltage relations are shown (n = 6). Current amplitude in Rb+ solution was pairwise normalized for each cell to that in K+ solution. (F) Quantification shows that inactivation I1 is significantly larger in Rb+ (35%) than in K+ (22%) solutions (n = 6; two-tailed paired t test; **P = 0.0021, t = 5.831, df = 5).

We elaborated a simple kinetic model that accounts for two distinct inactivation states in WT Kv7.1 (Fig. 3A). The model assumes two closed states (C1 and C2), two voltage-dependent open states (O1 and O2), and two voltage-dependent inactivation states I1 and I2, with a Ca2+-sensitive, slow inactivation state I1 and a Ca2+-insensitive, fast inactivation state I2. We generated a set of differential equations based on the gating scheme and fitted it to the observed time course of WT Kv7.1 currents in Ca2+-containing and Ca2+-free external solutions to obtain the transition rate constants and relative occupancies of each state. Simulating the gating behavior of WT Kv7.1 according to the scheme produced current kinetics that faithfully matched the experimental traces with coexistence of two distinct inactivation states (Fig. 3, B and C). In the kinetic model, the transition rate constants in Ca2+-free external solutions to the two different open states O1 (k2) and O2 (k4) are very different, with k2 being larger than k4, meaning that O1 is populated much faster than O2 (table S1). This feature together with the different rate constants of recovery from inactivation I1 (k-3) and I2 (k-5) explains the shape of the traces obtained in Ca2+-free external solutions, exhibiting a slow rising current that follows the transient peak current of inactivation I1 [Figs. 1 (A and B) and 3B and table S1].

Fig. 3 Kinetic model of Kv7.1 inactivation gating and sensitivity to external Ca2+ of mutants F351A and S338F.

(A) The model assumes two closed states (C1 and C2); two voltage-dependent open states (O1 and O2); a Ca2+-sensitive, slow, voltage-dependent inactivated state I1; and a Ca2+-insensitive, voltage-dependent fast inactivated state I2. For simplicity, we assumed that no openings occur when the cell is voltage clamped to −90 mV, and consequently, we fixed initial values of all states to zero, except C1. (B) Simulation of the gating behavior of WT Kv7.1 in Ca2 +-free external solutions according to the scheme in (A) generated current kinetics that closely matched the experimental traces. (C) Simulation of the gating behavior of WT Kv7.1 in Ca +-containing external solutions according to the scheme in (A) generated current kinetics that closely matched the experimental traces. (D) Mapping the S6 pore residues S338 and F351 onto the human Kv7.1 cryo-EM structure (PDB: 6UZZ). (E and G) Representative traces of F351A and S338F mutant currents, recorded using a similar protocol to that of Fig. 1A. As indicated, the extracellular solution contained either 1.2 mM Mg2+/1.8 mM Ca2+ or 1.2 mM Mg2+/0 mM Ca2+. (F) S338F inactivation I1 in the presence and absence of external Ca2+ is 30 and 60% in 5 μM free internal Ca2+ (Ca) and in 0 μM free internal Ca2+ (BAPTA), respectively (n = 5; two-tailed unpaired t test; **P = 0.0029, t = 4.220, df = 8; ****P < 0.0001).

Zaydman et al. (27) showed that the pore in Kv7.1 (KCNQ1) channels opens when the VSD activates to both intermediate and fully activated states, resulting in the intermediate open (IO) and activated open (AO) states, respectively (27). Hou et al. (15) found that when the VSD activates from the IO state to the AO state, the VSD-pore coupling has less efficacy in opening the pore, thereby producing inactivation. We suggest that the IO and AO states reported by the Cui’s group (15) may correspond to the O1 and O2 open states, which, as described in our study, can relax into I1 and I2 inactivation states, respectively. Hou et al. (15) identified two mutants in the S6 pore region (Fig. 3D), F351A and S338F, which suppress selectively the IO and AO states, respectively. If our assumption is correct, then inactivation I1 should be lost in mutant F351A and maintained in mutant S338F. Accordingly, we recorded F351A and S338F mutants in Ca2+-containing and Ca2+-free external solutions (Fig. 3, E to G). Results clearly indicate that mutant F351A failed to exhibit any inactivation in the presence or absence of external Ca2+ (Fig. 3E). In contrast, both in the presence and absence of external Ca2+, mutant S338F displayed a substantial voltage-dependent inactivation I1 of about 30 and 60%, respectively, in 5 μM free internal Ca2+ (Ca) and in 0 μM free internal Ca2+ (BAPTA) (Fig. 3, F and G). These data noticeably support the assumption that the IO and AO states may correspond to the O1 and O2 open states of this study.

As cardiac IKS channels result from the coassembly of Kv7.1 and KCNE1 subunits (33, 34), we coexpressed KCNE1 and Kv7.1 and examined the effect of Ca2+-free external solutions. Under these conditions, neither inactivation I1 nor inactivation I2 could be observed, suggesting that KCNE1 suppressed both types of inactivation in Ca2+-free external solution (fig. S5C). Nevertheless, a lower current amplitude could be observed (see Discussion). Multiple sequence alignment of the turret and pore regions of the Kv7 channel family shows that some differences exist in both the nature and length of the polypeptide between the different Kv7 members (fig. S5A). Kv7.2 homomeric channels displayed no inactivation in Ca2+-free external solution (fig. S5B). Similar results were obtained with Kv7.3, Kv7.4, and Kv7.5 homomeric channels, suggesting that inactivation gating is an exclusive property of Kv7.1 among the Kv7 channel family.

Coordination of Ca2+ in the external pore region prevents inactivation I1

From the recent cryo–electron microscopy (cryo-EM) structures of the frog and human Kv7.1 (25, 26), one could notice that the extracellular loop connecting the S5 segment and the pore helix forms a negatively charged cap surrounding the channel entrance, which includes residues D286, E290, S291, E295, and D317. After careful inspection of the human Kv7.1 structure (26), we identified within the same subunit the residue E295 in the turret as being in atomic proximity to D317, located at the external pore entrance, where the main chain carbonyl of D317 is at 2.9 Å distance from the carboxylate of E295 (Fig. 4A). We assumed that at pH 7.3 and in the absence of external Ca2+, residues D317 and E295 are hydrated, deprotonated, and, thus, negatively charged. This renders that any interaction between them is highly improbable because of electrostatic repulsions. We postulated that under these conditions, D317 side chain would be quite flexible and unable to engage in H-bonding with the pore helix residue W304, which, as previously shown in other Kv channels (3537), destabilizes the selectivity filter and leads to inactivation (Fig. 4A). Assuming a minimum coordination number of 6, we hypothesized that for each subunit, one Ca2+ ion may be coordinated by the main chain carbonyl of D317, the carboxylate of E295 (likely monodentate coordination), and four oxygens from water molecules, thereby allowing H-bonding between D317 and W304, which could prevent inactivation. If this assumption is correct, then neutralizing E295 will mimic the Ca2+-bound state of WT Kv7.1, in which the negative charge at E295 is shielded by external Ca2+, thereby preventing inactivation I1. Therefore, we generated the neutralizing mutants E295A and E295Q and the charge reversal mutant E295R and exposed them to Ca2+-containing and Ca2+-free external solutions. In line with our assumption, we found that mutants E295A, E295Q, and E295R exhibited significantly smaller inactivation I1 than WT Kv7.1 in Ca2+-free external solutions (Fig. 4, B and D). Expectedly, when we neutralized the residue D317 by generating the mutant D317N, no ionic currents could be recorded, because, as shown for Shaker K+ channels (37), D317N (D447 in Shaker) cannot anymore form H-bonding with W304 (W434 in Shaker) and is unable therefore to stabilize the selectivity filter, which adopts a nonconducting inactivated conformation (Fig. 4, A and C). Similarly, mutants D317S and D317H were not functional.

Fig. 4 Mutants in the turret region of Kv7.1 prevent inactivation I1.

(A) Mapping residues E295, D317, and W304 in the turret, external pore entrance, and pore helix residue, respectively, onto the human KV7.1 cryo-EM structure (PDB: 6UZZ). (B) Representative traces of WT Kv7.1, E295A, and E295Q mutant currents recorded using a similar protocol to that of Fig. 1A, with pipette solution containing 5 mM BAPTA in nominally zero free Ca2+ concentration. The extracellular solution contained either 1.2 mM Mg2+/1.8 mM Ca2+ or 1.2 mM Mg2+/0 mM Ca2+. (C) Mutant D317N at the external pore entrance recorded as above in an extracellular solution containing 1.2 mM Mg2+/1.8 mM Ca2+ does not yield functional currents. (D) Removal of external Ca2+ produced a significant smaller inactivation I1 in mutants E295A (12%; n = 12), E295Q (10%; n = 5), and E295R (13%; n = 6) than WT (48%; n = 12) [one-way ANOVA and Dunnett’s multiple comparisons test; ****P < 0.0001, t = 3.925 F(7, 71)].

To validate our experimental results and assumptions, we performed MD simulations for the human Kv7.1 (KCNQ1) cryo-EM structure without PIP2 [Protein Data Bank (PDB) code: 6UZZ], which was determined at a resolution of 3.1 Å by Sun and MacKinnon (26). The structures were simulated in the presence of a calcium ion in each subunit, as well as in its absence. For each system, three independent 50 ns of MD simulations with different random seeds were performed (see the Supplementary Materials and movies S1 to S3). In almost all cases, the average distance between the Ca2+ ion and the carboxylate of E295 was 2.3 Å (fig. S6A) and remained stable in each subunit of the tetramer. The coordination of Ca2+ ion by the carbonyl main chain of D317 occurred at high frequency (~80% of the simulation time) at least in two subunits, and the average distance was 2.4 Å. Consequently, there was a significant higher probability of maintaining a distance of less than 3 Å between the carboxylate of E295 and the main chain carbonyl of D317 in the presence of external Ca2+ than in its absence (fig. S6B). A significant higher probability for a distance of less than 3 Å arose between the D317 carboxylate and the hydrogen of the indole nitrogen of W304 in the presence of external Ca2+ than in its absence, allowing H-bond formation (fig. S6C).

External acidification mimics the effects of external Ca2+ and prevents inactivation I1 in Ca2+-free external solutions

We reasoned that protonating the acidic residues E295 and D317 by external acidification will mimic the Ca2+-bound state of WT Kv7.1 and prevent inactivation I1. In accordance with our assumption, exposing WT Kv7.1 channels to a Ca2+-free external solution at pH 5.5 significantly reduced inactivation I1 (fig. S7, A and B). At this pH, a non-negligible fraction of E295 and D317 side chains should be protonated (propka-PyPI server; http://propka.org/). E295 has a pKa (where Ka is the acid dissociation constant) of 6.2, and D317 has a pKa of 4.5. Therefore, at pH 5.5, E295 will be mostly (~80%) protonated, while D317 will be weakly protonated (~10%). In addition to its attenuation of inactivation I1, external acidification significantly slowed down the activation kinetics of WT Kv7.1 in both Ca2+-containing and Ca2+-free external solutions (fig. S7, C and D). In line with our structural assumptions, these results indicate that similar to the neutralization (E295A and E295Q), protonation of E295 mimics the effects of external Ca2+ and prevents inactivation I1 in Ca2+-free external solutions.

Internal Ca2+-CaM prevents inactivation I1 of WT Kv7.1 in Ca2+-free external solutions

We and others previously showed that CaM is essential for Kv7.1 assembly and gating (28, 38). Furthermore, the recent cryo-EM structure of Kv7.1 showed that in the PIP2-free state, the third EF hand of CaM contacts the intracellular S2-S3 linker of the VSD, suggesting a direct impact of CaM on Kv7.1 channel gating (25, 26). Hence, we asked whether CaM could affect WT Kv7.1 inactivation gating in Ca2+-free external solution. We recorded WT Kv7.1 current from transfected CHO cells in Ca2+-containing and Ca2+-free external solutions. Using 5 mM BAPTA, as a fast chelator of free Ca2+, the pipette solution contained purified WT CaM either in the presence of Ca2+ (Ca2+-CaM) (3 μM CaM and 5 μM free Ca2+) or in its absence (BAPTA-CaM) (3 μM CaM and zero free Ca2+). The introduction of BAPTA-CaM or Ca2+-CaM in the pipette solution was designed to study the impact of CaM on channel gating and exclude any potential effects on channel trafficking. BAPTA-CaM (Apo-CaM) in the pipette solution retained inactivation I1 and I2 of WT Kv7.1 in Ca2+-free external solutions (Fig. 5, A and B). In contrast, Ca2+-CaM virtually abolished inactivation I1 but maintained inactivation I2 with the presence of hooked tail currents. Similar outcomes were obtained when instead of adding BAPTA-CaM or Ca2+-CaM in the pipette solution, WT Kv7.1 was cotransfected with the CaM1,2,3,4 mutant (a Ca2+-insensitive CaM, mutated at the four EF hands) or WT CaM, respectively (Fig. 5, A and B). These results explain why in the Xenopus oocyte expression system, inactivation I1 of WT Kv7.1 is barely visible under Ca2+-free external solutions, while hooked tail currents (inactivation I2) are readily noticed. The oocyte contains plenty of Ca2+-CaM. One needs to inject BAPTA to see clearly the inactivation I1 of WT Kv7.1 (38).

Fig. 5 Internal calcified CaM and PIP2 prevent inactivation I1 of WT Kv7.1 in Ca2+-free external solutions.

(A) Representative traces of WT Kv7.1 in 1.8 mM Ca2+/1.2 mM Mg2+ and 0 mM Ca2+/1.2 mM Mg2+ external solutions. (B) Removal of external Ca2+ produced a significant smaller inactivation I1 as indicated (2.9% versus 39.4%, respectively; n = 5, two-tailed unpaired t test; ***P < 0.0001, t = 6.996, df = 8). A significant smaller inactivation I1 was obtained following removal of external Ca2+ upon cotransfection with WT CaM than with CaM1,2,3,4 mutant (1.3% versus 22.2%, respectively; n = 7 to 11, two-tailed unpaired t test; *P = 0.0131, t = 2.789, df = 16, **P = 0.0042, t = 3.684 df = 10). (C) Representative traces of WT Kv7.1 in 0 mM Ca2+/1.2 mM Mg2+ external solutions (left red trace) and WT Kv7.1 cotransfected with the PIP4,5-kinase in 1.8 mM Ca2+/1.2 mM Mg2+ (black trace) and 0 mM Ca2+/1.2 mM Mg2+ external solutions (red trace). (D) Increase in cellular PIP2 by cotransfection with the PIP4,5-kinase significantly reduced the inactivation I1 of WT Kv7.1 in Ca2+-free external solutions (2.9% versus 35.1%, respectively; n = 5 to 10, two-tailed unpaired t test; ***P = 0.0002, t = 5.030, df = 13). (E) Representative current traces before (black trace) and after (red trace) activation of DrVSP (Danio rerio voltage-sensitive phosphatase) in cells transfected with DrVSP (top) or with empty vector (bottom). (F) PIP2 depletion by DrVSP significantly increased inactivation I1 of WT Kv7.1 from 45 to 63% (n = 5; two-tailed paired t test; **P = 0.0069, t = 5.115, df = 4).

PIP2 prevents inactivation I1 of WT Kv7.1 in Ca2+-free external solutions

On the basis of our recent findings that PIP2 and CaM can compete for a similar binding site (27, 39) and prior knowledge that PIP2 is necessary for reducing Kv7.1 spontaneous rundown (21, 27, 40), we examined the impact of PIP2 on Kv7.1 inactivation gating. The experiments were conducted with a pipette solution containing 5 mM BAPTA in zero free Ca2+ to measure the percentage of inactivation I1. CHO cells were cotransfected with WT Kv7.1 and PIP4,5-kinase that elevates PIP2 levels by producing PIP2 from phosphatidylinositol 4-phosphate (PI4P). When cells were exposed to Ca2+-free external solutions, elevated PIP2 levels strongly attenuated inactivation I1 while preserving inactivation I2, as reflected by the presence of hooked tail currents (Fig. 5, C and D). Conversely, we checked whether PIP2 depletion enhances Kv7.1 inactivation I1. Cells were cotransfected with DrVSP (Danio rerio voltage-sensitive phosphatase), which depletes cellular PIP2 at +90 mV. Kv7.1 currents were recorded in Ca2+-free external solutions, and the extent of inactivation I1 was quantified before and after activation of DrVSP, as previously described (39). A triple pulse protocol was applied where the membrane was first stepped from −90 to +30 mV. Then, PIP2 depletion was induced by activating DrVSP using a 200-ms step depolarization from −90 to +90 mV 100 times. To measure the consequences of PIP2 depletion induced by DrVSP activation on inactivation I1, a 3-s step depolarization from −90 to +30 mV was reapplied. Results show that depletion of PIP2 by DrVSP significantly increased inactivation I1 of WT Kv7.1 from 45 to 63% (Fig. 5, E and F). Together, these data indicate that similar to Ca2+-CaM, PIP2 prevents Kv7.1 inactivation I1.

Disrupting the interaction between CaM and the S2-S3 linker triggers inactivation I1

The structure of the Kv7.1-CaM complex (PDB code: 6UZZ) reveals that CaM contacts Kv7.1 in the VSD and the proximal C terminus at helix A and helix B (26). More specifically, in the CaM C lobe, residues Y100 and N138 are at atomic distances of residue S182 in the intracellular linker S2-S3 of the VSD (Fig. 6A). Mutational disruption of the CaM and S2-S3 linker interaction by the Kv7.1 mutant S182A triggered by itself a large, visible inactivation I1 even in Ca2+-containing external solution (5 mM BAPTA in zero free Ca2+ internal solution) (Fig. 6, B and C; I1 of 43% for S182S versus 9% for WT). In contrast to WT Kv7.1, introduction of Ca2+-CaM to the pipette solution could not completely prevent inactivation I1 of mutant S182A in Ca2+-containing external solution (Fig. 6, B and C; I1 of 21% for S182A versus 0% for WT). Cotransfecting CHO cells with mutant S182A and PIP4,5-kinase completely prevented inactivation I1 and preserved inactivation I2 in Ca2+-containing external solution (5 mM BAPTA in zero free Ca2+ internal solution), suggesting that PIP2 acts via a different residue from S182 to protect Kv7.1 from inactivation I1 (Fig. 6, C and D).

Fig. 6 Disrupting the interaction between CaM and the S2-S3 linker triggers inactivation I1.

(A) Mapping residues S182 in the intracellular S2-S3 linker, onto the human KV7.1 cryo-EM structure (PDB: 6UZZ); shown in atomic proximity, residues Y100 and N138 of CaM. (B) Representative traces of WT Kv7.1 and the Kv7.1 S2-S3 linker mutant S182A in 1.8 mM Ca2+/1.2 mM Mg2+ external solutions. (C) A significant larger inactivation I1 was observed with the mutant S182A, as compared to WT in 1.8 mM Ca2+/1.2 mM Mg2+ external solutions, when the pipette solution contained either Ca2+-CaM [21.7% versus 0.4%, respectively; n = 5 to 17, one-way ANOVA, F(5, 51) = 15.84; ****P < 0.0001; Tukey’s multiple comparisons test; adjusted *P = 0.024] or BAPTA-CaM [43.1% versus 9.2%, respectively; n = 5 to 17, one-way ANOVA, F(5, 51) = 15.84; ****P < 0.0001; Tukey’s multiple comparisons test, adjusted ****P < 0.0001]. The mutant S182A exhibited a larger inactivation when the pipette solution contained BAPTA-CaM (nominally zero free Ca2+) than Ca2+-CaM (nominally 5 μM free Ca2+) [43.1% versus 21.7%, respectively; n = 17, one-way ANOVA, F(5, 51) = 15.84; ****P < 0.0001; Tukey’s multiple comparisons test; adjusted ***P = 0.0002]. Cotransfection with PIP4,5-kinase completely prevented inactivation I1 displayed by mutant S182A in 1.8 mM Ca2+/1.2 mM Mg2+ external solutions, when the pipette solution contained 5 mM BAPTA (nominally zero free Ca2+) [0% versus 28. 3%, respectively; n = 5 to 8, one-way ANOVA, F(5, 51) = 15.84; ****P < 0.0001; Tukey’s multiple comparisons test; adjusted **P = 0.0044]. (D) Representative traces of the Kv7.1 S2-S3 linker mutant S182A in 1.8 mM Ca2+/1.2 mM Mg2+ external solutions, where the pipette solution contained 5 mM BAPTA in nominally zero free Ca2+ concentration and no added CaCl2 (BAPTA) in the absence (red trace) or presence of cotransfected PIP4,5-kinase (black trace). (E) Molecular model of Kv7.1 the activation-inactivation O1-I1 transition (see Discussion).

Ca2+-CaM and PIP2 impede inactivation I1 of the Kv7.1 mutants, S338F and G246W

To get some mechanistic insight about the molecular mechanisms underlying the activation-inactivation O1-I1 transition, we investigated whether inactivation I1 of Kv7.1 mutants located in functionally important regions could be precluded by Ca2+-CaM or/and PIP2, in a similar way to that observed for WT Kv7.1 in Ca2+-free external solutions or to that of mutant S182A in the S2-S3 linker that contacts CaM. We examined two interesting mutants, S338F and G246W, which are located, respectively, in the S6 pore region and in the S4-S5 intracellular linker. S338F was previously found to eliminate the AO state or O2, thereby suppressing its relaxation into inactivation I2 (15). Furthermore, as mentioned above, mutant S338F displayed a significant voltage-dependent inactivation I1 both in the presence and absence of external Ca2+ (Fig. 3, F and G, and fig.S8B). Similarly, mutant G246W exhibited a large, voltage-dependent inactivation I1 both in the presence and absence of external Ca2+. G246W is located in the S4-S5 linker, a crucial region for the PIP2-dependent VSD-pore coupling (fig. S8C) (27). Cotransfection of mutant S338F with PIP4,5-kinase or CaM significantly reduced inactivation I1 from 59 to 4 and 25%, respectively (fig. S8, B and D). Likewise, cotransfection of mutant G246W with PIP4,5-kinase or CaM significantly hampered inactivation I1 from 58 to 32 and 25%, respectively (fig. S8, C and E). These data indicate that the role of Ca2+-CaM and PIP2 in preventing inactivation I1 is not restricted to WT Kv7.1 in Ca2+-free external solutions but is of more general importance and is effective in Kv7.1 mutants that undergo inactivation and are located in functionally essential regions of the channel.

DISCUSSION

In this study, we revealed a unique mechanism of intrinsic inactivation gating in WT Kv7.1 channels, whose characteristics are different from those of N and C types, as found in many members of the Kv channel superfamily (2, 3). WT Kv7.1 channels inherently populate two distinct open states O1 and O2, which respectively relax into two discrete inactivation states, a slow, voltage-dependent, Ca2+-sensitive inactivation I1 and a fast, voltage-dependent, Ca2+-insensitive inactivation I2 (Fig. 3). This assumption is supported by the following data. First, amplitude of I1 was independent to that of I2. While in Ca2+-free external solutions, inactivation I1 was much larger than that in Ca2+-containing solutions; inactivation I2 was very similar in either solution (fig. S2). Second, for inactivation I1 and inactivation I2, the relaxation kinetics, the time constants of recovery, and the voltage dependence were very different (figs. S3 and S4). Third, while for the O2-I2 transition, Rb+ decreased the probability of the channel to be in the flicker-blocked state, thereby increasing the Rb+/K+ inward current ratio; for the O1-I1 transition triggered by removal of external Ca2+, Rb+ did not affect the Kv7.1 flicker pore properties (Fig. 2D). Fourth, I1 and I2 inactivation states can be easily discriminated by the mutation V307L, where in Ca2+-free external solutions inactivation I1 occurs and remains unaffected while inactivation I2 is abolished (Fig. 2). Fifth, Ca2+-CaM and PIP2 prevent inactivation I1 but preserve inactivation I2 (Fig. 5). Sixth, although our simple kinetic model may not account for all the details of Kv7.1 channel gating, its simulation reasonably matched the experimental traces (Fig. 3). Our data considerably support the assumption that the IO and AO states previously described by Zaydman et al. (27) may correspond to the O1 and O2 open states of this study. However, more work will be necessary to properly integrate the state diagram of the Cui’s group (15, 27) and that of the present work.

To our knowledge, this is the first case where external Ca2+ is found to prevent with high affinity (IC50 of 1.5 μM) Kv channel inactivation. In Shaker K+ channels, extracellular addition of 10 to 40 mM Ca2+ was shown to enhance C-type inactivation by increasing the rate of inactivation during depolarization and markedly promoting inactivation and/or suppressing recovery from inactivation at negative voltages (41). This effect of external Ca2+ on C-type inactivation was antagonized by millimolar concentrations of external K+, suggesting that Ca2+ competes with K+ at the channel’s outer mouth and prevents K+ from interacting with the filter’s outer carbonyl ring (41). A similar effect was observed with external La3+. External Ca2+ has also a rapid blocking action on Shaker K+ channels and leads to an acceleration of the channel gate closure possibly by changing the ion occupancy state of the channel pore (42). These features are very different from those of Kv7.1 inactivation gating, where external Ca2+ does not promote but prevents inactivation and external K+ does not preclude inactivation triggered by Ca2+-free external solutions. This suggests that in contrast to Shaker K+ channels, the Ca2+ binding site in Kv7.1 does not overlap with a K+ binding site in the permeation pathway.

In the cryo-EM structure of the human Kv7.1 (26), we noticed that within the same subunit, residue E295 in the turret is in atomic vicinity with D317 located near the outer pore entrance, where the main chain carbonyl of D317 is at 2.9-Å distance from the carboxylate of E295. At neutral pH and in the absence of external Ca2+, we assume that the flexibility of the side chains of both D317 and E295 will increase, since they are deprotonated and negatively charged, which precludes the D317 carboxylate to stably interact via H-bonding with the hydrogen of the indole nitrogen of W304 in the pore helix. In Shaker-like and KcsA K+ channels, the lack of H-bonding between these two conserved residues was previously shown to destabilize the selectivity filter and lead to inactivation (3537, 43). This aspartate residue D317 (D447 in Shaker and D80 in KcsA) has a pivotal role in ion conduction for all Kv channels. Behind the filter, these two residues (D447 and W434 in Shaker; D317 and W304 in hKv7.1; D80 and W67 in KcsA) were shown to play a crucial role in stabilizing the filter in a conducting conformation, thanks to the H-bond formed between the hydrogen of the indole nitrogen and the carboxylate moiety of the Trp and Asp residues, respectively (3537, 43). Strengthening the H-bond by incorporating fluorinated Trp slows down C-type inactivation, while weakening the H-bond by incorporating 2-amino-3-indol-1-yl-propionic acid, a synthetic isosteric Trp derivative, speeds up inactivation (37). It was suggested that abolishing this stabilizing H-bond leads W434 (W304 in hKv7.1) to rotate, pushing its ring on that of Y445 (Y315 in hKv7.1), thereby causing a rotation of Y445 and movement of its carbonyl away from the pore axis (11, 41).

Our assumption that external Ca2+ acts to screen the negative charge of E295, thereby providing appropriate H-bonding between D317 and W304 and preventing inactivation I1, is supported by two crucial data. First, neutralizing the turret residue E295 with mutants E295A or E295Q mimics the Ca2+-bound state of WT Kv7.1 by significantly reducing inactivation I1 in Ca2+-free external solutions (Fig. 4). Second, in line with our suggestion that protonating acidic residues E295 and D317 by external acidification will mimic the Ca2+ bound state of WT Kv7.1, we showed that exposing WT Kv7.1 channels to a Ca2+-free external solution at pH 5.5 markedly decreases inactivation I1 (fig. S7). MD simulations showed a significant higher probability for a distance of less than 3 Å between the carboxylate of D317 and the hydrogen of the indole nitrogen of W304, in the presence of external Ca2+ than in its absence (fig. S6). Regarding residue 317, as expected, its neutralization with the mutant D317N yielded no ionic currents. In Shaker K+ channel, the homologous charge-neutralizing mutation D447N or the pore helix mutant W434F leads to a complete loss of ionic currents, resulting only in gating currents by trapping the selectivity filter in a nonconducting conformation (44, 45). Despite the clear differences existing between C-type inactivation of Shaker-like Kv channels and inactivation I1 of Kv7.1 such as the role of external K+ and of Ca2+ coordination in the turret region, it is interesting to notice that important features are also shared and converged to the selectivity filter. Notably, the two residues Asp and Trp and their H-bonding network behind the filter may be considered as an end point for filter inactivation in Kv channels, regardless of the different trajectories that converge to it. In line with this idea, a recent study on the Ca2+-dependent inactivation (CDI) of voltage-gated Ca2+ channels (Cav) showed that although CDI initiates within the cytoplasmic Ca2+-CaM sensor, a conserved selectivity filter domain II (DII) aspartate is essential for CDI, suggesting that the Cav selectivity filter is the CDI end point (46). This study even suggests that the selectivity filter inactivation is an ancient process arising from the shared voltage-gated ion channel pore architecture (46).

How could this large inactivation I1 triggered by Ca2+-free external solutions be physiologically relevant since physiological solutions contain calcium concentrations that would prevent it? Under what pathological conditions one would see this type of inactivation in a native cell? The inactivation I1 seen in WT Kv7.1 under Ca2+-free conditions actually reveals a unique intrinsic inactivation gating that is shared and observed recurrently in many human long QT mutations in Ca2+-containing solutions. The best example is provided in this study, where the long QT mutant S338F exhibits a voltage-dependent inactivation I1 (Fig. 3). Likewise, we previously characterized long QT mutations impairing CaM binding to the Kv7.1 proximal C terminus (helix A), which lead to channel inactivation I1 (28). For some long QT inactivating mutants such as R366P, KCNE1 is even unable to suppress the large inactivation I1 (28). Most long QT inactivating mutants coexpressed with KCNE1 exhibit much lower current densities, which are not necessarily accounted for by a right shift in the voltage dependence of channel activation. Instead, we speculate that part of the loss of the current amplitude may arise from the entry of the mutant channel into inactivation I1 that could not be visible macroscopically because it is faster than the slow activation kinetics in the presence of KCNE1, suggesting a similar behavior to that seen for Herg channels.

A unique feature of WT Kv7.1 inactivation gating is the ability of Ca2+-CaM or PIP2 to prevent inactivation I1 in Ca2+-free external solutions. Unlike subtypes of voltage-gated Ca2+ and Na+ channels, where CaM supports a CDI process (47, 48), Ca2+-CaM and PIP2 can preclude the entry of Kv7.1 to inactivation state I1. Our experimental data clearly indicate that Ca2+-CaM prevents inactivation I1, leaving inactivation I2 unaffected (Fig. 5). Our data indicate that increasing the levels of PIP2 can, similar to Ca2+-CaM, preclude Kv7.1 inactivation I1, in line with our recent study showing a dynamic interaction between Ca2+-CaM and PIP2 in Kv7.1 channel gating (Fig. 5) (39). PIP2 is essential for maintaining Kv7.1 channel activity (21, 27). PIP2 regulates Kv7.1 channel function by providing efficient coupling between the VSD motion and the pore opening, thereby stabilizing the channel open state conformation (27). The present study suggests that PIP2 and Ca2+-CaM are not only important for VSD-pore coupling but are also crucial for preventing the activation-inactivation O1-I1 transition.

How can intracellular Ca2+-CaM or PIP2 prevent inactivation I1 in Kv7.1, a process that involves the outer pore region and converges to the selectivity filter? The recent cryo-EM structure of Kv7.1 showed that in a PIP2-free state, the C lobe of CaM, more specifically the third EF hand, contacts the intracellular S2-S3 linker of the VSD (25, 26). In this configuration, the linker connecting S6 to helix A adopts a flexible loop structure, which bends the proximal C terminus and features a closed pore, despite a fully activated VSD conformation. In the PIP2-bound state, the linker connecting S6 to helix A undergoes a structural rearrangement from a loop to helix, which straightens the proximal C terminus. This large rearrangement causes the CaM C lobe to release its interaction with the S2-S3 linker of the VSD and features a dilated pore with a fully activated VSD conformation (25, 26). Since Kv7.1 inactivation I1 is voltage dependent and likely involves the VSD motion, we suggest that with an IO state VSD conformation, Ca2+-CaM prevents inactivation I1 thanks to its interaction with the S2-S3 intracellular linker and with helices A and B of the proximal C terminus. This latter region acts as a gating module to couple the CaM signal via the S6 pore domain to the turret and converges to the selectivity filter as an end point of inactivation I1. In support of this assumption, we showed that disrupting proper interaction between CaM and the S2-S3 linker triggers inactivation I1 as observed in mutant S182A (Fig. 6). Likewise, disrupting proper CaM interaction with helix A at the proximal C terminus also triggers large inactivation and it is not by chance that all long QT mutations clustered in this region produced large inactivation I1 (28). Similarly, the S6 pore mutant S338F, which locates to a region that may relay the CaM signal to the outer pore also undergoes inactivation I1. In line with this notion, a recent study showed that in Cav channels, S6 is likely to be critical for coupling the Ca2+-CaM signal from its intracellular boundary to the selectivity filter gate as end point for CDI (46).

We propose a tentative model that reflects the O1-I1 gating transition. In the O1 open state, the VSD may adopt an IO conformation, with the CaM C lobe contacting the S2-S3 VSD linker. In this O1 state, PIP2 may interact with different regions including the S2-S3 and S4-S5 intracellular linkers, as well as downstream S6 at the proximal C terminus, not only to couple the VSD for pore opening (O1) but also to prevent the activation-inactivation O1-I1 transition to occur (Fig. 6E). Of course, PIP2 is also crucial for the transition O1-O2. The relaxation to inactivation I1 of WT Kv7.1 in Ca2+-free external solutions and of many mutants in normal Ca2+ solutions may involve weakened CaM or/and PIP2 interactions with the channel, which spoils correct coupling to the pore. Under these conditions, substantial addition of Ca2+-CaM or PIP2 can recover the O1 open state from the nonconducting inactivation I1 state (Fig. 6E). Our results clearly indicate that the proximal C-terminal bent structure involving the CaM interaction with the S2-S3 linker of the VSD takes an integral part in Kv7.1 gating and does not necessarily reflect a closed channel conformation. However, much work is necessary to dissect the conformational trajectory involved in this allosteric mechanism.

In summary, we revealed a unique mechanism of inactivation gating in a member of the Kv channel family, Kv7.1, whose characteristics are different from N or C types. This mechanism involves coordination of external Ca2+ ions in the turret region that specifically prevents intrinsic inactivation of WT Kv7.1 and is allosterically modulated by PIP2 and calcified CaM at the inner face of the channel transmembrane domain.

MATERIALS AND METHODS

Constructs

Human Kv7.1, KCNE1, Kv7.2, Kv7.3, and Kv7.4 as well as WT CaM and mutant CaM1,2,3,4 were cloned into the pcDNA3 vector to allow eukaryotic expression. The mutations in Kv7.1 were introduced using the Polymerase Chain Reaction-based QuikChange site-directed mutagenesis (Stratagene) and were verified by full sequencing of the entire plasmid vector.

Cell culture and transfection

CHO cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 2 mM glutamine, 10% fetal calf serum, and antibiotics. Briefly, 40,000 cells seeded on poly-d-lysine–coated glass coverslips (13 mm in diameter) in a 24-multiwell plate were transfected with pIRES-CD8 (0.3 μg) as a marker for transfection and with 0.5 μg of WT, mutant Kv7.1, 1 μg of KCNE1, 0.5 μg of Kv7.2, Kv7.3, or Kv7.4. Transfection was performed using 3.6 μl of X-tremeGENE 9 (Roche) according to the manufacturer’s protocol. For electrophysiology, transfected cells were visualized approximately 40 hours after transfection, using the anti-CD8 antibody–coated beads, method as described previously (39).

Electrophysiology

Recordings were performed using the whole-cell configuration of the patch-clamp technique. Signals were amplified using an Axopatch 700B patch-clamp amplifier (Axon Instruments), sampled at 5 kHz, and filtered at 2.4 kHz via a four-pole Bessel low-pass filter. Data were acquired using pClamp 10.5 software in conjunction with a Digidata 1440A interface. The patch pipettes were pulled from borosilicate glass (Harvard Apparatus) with a resistance of 3 to 5 megohms. The intracellular pipette solution contained 130 mM KCl, 5 mM K2-ATP, 10 mM Hepes (pH 7.3) (adjusted with KOH), and 5 mM K4BAPTA (no added CaCl2, nominally 0 free Ca2+ concentration). When CaM was added to the pipette solution, the BAPTA-CaM solution contained 5 mM K4BAPTA and 3 μM purified CaM and no added CaCl2 (nominally 0 free Ca2+ concentration), while Ca2+-CaM solution contained 5 mM K4BAPTA and 3 μM purified CaM and added CaCl2 [for nominally 5 μM free Ca2+ concentration, as calculated by MAXCHELATOR (WEBMAXC STANDARD) software; https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/webmaxc/webmaxcS.htm], with sucrose added to adjust osmolarity to 290 mosmol. The external Ca2+-containing solution was 140 mM NaCl, 4 mM KCl, 1.8 mM CaCl2, 1.2 mM MgCl2, 11 mM glucose, and 5.5 mM Hepes, adjusted with NaOH to pH 7.3 (310 mOsM). The Ca2+-free external solution contained 140 mM NaCl, 4 mM KCl, 1.2 mM MgCl2, 2 mM EGTA with no added CaCl2 (nominally 0 free Ca2+ concentration), 11 mM glucose, and 5.5 mM Hepes, adjusted with NaOH to pH 7.3 (310 mOsM). The Mg2+-free external solution was as above but with no added MgCl2. The series resistances were compensated (75 to 90%) and periodically monitored. Electrophysiological data analysis was performed using the Clampfit program (pClamp 10.5; Axon Instruments), Microsoft Excel (Microsoft, Redmond, WA), and Prism 5.0 (GraphPad Software Inc., San Diego, CA). Leak subtraction was performed offline, using the Clampfit program of the pClamp 10.5 software. Chord conductance (G) was calculated by using the following equation: G = I/(VVrev), where I corresponds to the current amplitude measured at the end of the pulse, and Vrev is the calculated reversal potential assumed to be −90 mV in CHO cells. G was estimated at various test voltages (V) and then normalized to a maximal conductance value, Gmax. Activation curves were fitted by one Boltzmann distribution: G/Gmax = 1/{1 + exp[(V50V)/s]}, where V50 is the voltage at which the current is half-activated and s is the slope factor. All data were expressed as means ± SEM. Statistically significant differences were assessed by paired, unpaired two-tailed Student’s t test or one-way analysis of variance (ANOVA), as indicated in figure legends.

Kinetic model

We elaborated a simple kinetic model that accounts for two distinct inactivation states in WT Kv7.1 (fig. S5). The model assumes two closed states (C1 and C2), two voltage-dependent open states (O1 and O2), and two voltage-dependent inactivation states I1 and I2, with a Ca2+-sensitive, slow inactivation state I1 and a Ca2+-insensitive, fast inactivation state I2. The procedure was similar to that described by Gibor et al. (14). The conductance time course for each recording was calculated by dividing the current amplitude by the driving force for K+, assuming a reversal potential of −90 mV in CHO cells under the experimental settings. We generated a set of differential equations based on the kinetic scheme (fig. S4C) and fitted it to the observed time course of WT Kv7.1 activation and inactivation to obtain the transition rate constants and relative occupancies of each state. For simplicity, we assumed that no openings occur when the cell is voltage-clamped to −90 mV (the holding potential), and consequently, we fixed the initial values of all states to zero, except C1. According to the scheme, the open channel probability can be calculated from Eq. 1P0=O1+O2/C1+C2+O1+O2+I1+I2(1)and the relative state occupancy (ϕ) is defined as Eq. 2ϕ=SiΣinS(2)We selected five representative cells for which the voltage steps from −90 to +60 mV and then from +60 to −60 mV were available for the whole range of external Ca2+ concentrations (nominally 0 nM, 26 nM, 78 nM, 12.5 μM, and 1.8 mM Ca2+). The hooked tail currents that were obtained for the voltage step from +60 to −60 mV were analyzed as previously described (16) to obtain the fast inactivation parameters. Subsequently, conductance time courses were averaged for each Ca2+ concentration. We fitted both conductance waveforms (−90 to +60 and to −60 mV) with differential equations generated in Berkeley Madonna for Windows and numerically solved using fourth-order Runge-Kutta integration method. In addition, since the fast inactivation parameters (I2) were Ca2+ independent, we fixed these parameters in all fitting procedures for the whole range of Ca2+ concentrations. To test the validity of the obtained rate constants, we simulated the entire voltage protocol for two Ca2+ concentrations (0 and 1.8 mM). For these kinds of simulations, we picked up the rate constant set for 0 and 1.8 mM Ca2+ concentrations and used them for simulation of the current for 3 s; subsequently, we adjusted the driving force and switched to rate constants obtained from analysis of the voltage step from −90 to −60 mV (figs. S5 and S6). Time course fitting and simulations were done with Berkeley Madonna software (version 8.0.2 for Windows; Kagi Shareware, Perth, Western Australia).

Computational methods

System preparation. MD simulations were performed for the human KCNQ1 cryo-EM structure (PDB code: 6UZZ) in its closed state, which was determined at a resolution of 3.1 Å by Sun and MacKinnon (26). Before simulation, the CaM chains were removed from the structure. The structures were simulated in the presence of a calcium ion in each chain (overall, four calcium ions), as well as in the absence of calcium ions. Since the cryo-EM map was recorded without calcium ions, the ion was manually placed next to Glu295 residue in each of the four chains. Next, each structure (with and without calcium ions) was prepared using the Prepare Protein protocol, as implemented in Maestro version 11.2 (Schrödinger). The transmembrane helices of the prepared structures were inserted into an explicit 150 Å × 150 Å membrane patch composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipids. The initial positioning of the helices within the membrane was determined using the Orientations of Proteins in Membranes (OPM) server. Then, the systems were solvated in TIP3P water molecules using the Solvate plugin in Visual Molecular Dynamics (VMD) to form 150 Å × 150 Å × 176 Å simulation box. Sodium and chloride ions were added to the water phase using the Ionize plugin in VMD, to neutralize the system and to obtain a salt concentration of 0.15 M.

MD simulations. MD simulations were performed with NAMD 2.12 with CHARMM36 force field including the φ/ψ cross-term map (CMAP) correction with the modified parameters for ions (49). The simulations were conducted under a Langevin temperature and pressure control, using periodic boundary conditions with particle mesh Ewald electrostatics with 12-Å cutoff for long-range interactions. First, the simulated systems were energy minimized with the conjugate gradient minimization algorithm to remove van der Waals clashes within the systems. Following the minimization, the systems were equilibrated as follows: (i) the lipid membrane tails were allowed to melt for 10 ns, while all other components of the system were fixed to introduce the suitable disorder of a fluid-like bilayer; (ii) the entire system was allowed to equilibrate for 10 ns while constraining the positions of the protein atoms; and (iii) the entire system was allowed to equilibrate for 10 ns with no constraints. Last, the production simulations were carried out for 50 ns with a constant temperature of 300 K and constant pressure of 1 atm (under Nosé-Hoover Langevin Piston coupling algorithm) and with an integration time step of 2 fs. Each of the systems was simulated three times, each time starting from a different random seed.

Data analysis. The resulting trajectories were visually inspected using VMD 1.9.3 software. The stability of the resulting trajectories and the average mobility of all residues were tested on the basis of the root mean square deviation (RMSD) of the backbone atoms of the protein from the initial structure and on the root mean square fluctuations (RMSF), respectively. RMSD and RMSF were calculated using the rms and rmsf utilities of the GROMACS 5.0.6 package, respectively. Distances between residues were calculated using the distance utility in GROMACS 5.0.6.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/51/eabd6922/DC1

https://creativecommons.org/licenses/by-nc/4.0/

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We thank A. V. Tzingounis (Department of Physiology and Neurobiology, University of Connecticut, Storrs, CT) for gift of the PIP4,5-kinase plasmid. Funding: This work was supported by a grant from the Israel Science Foundation (ISF 1365/17) to B.A., who holds the Andy Libach Professorial Chair in clinical pharmacology and toxicology. J.A.H. was supported by an Israel Science Foundation grant 1500/16. Author contributions: A.P., Y.P., and B.A. designed the research; W.S.T., M.L., A.P., L.S., A.Y., D.Y., and G.L. performed research; G.L. and J.A.H. contributed new reagents/analytic tools; and B.A. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article