Voltage-gated K+ channels have distinct gates that regulate ion flux: the activation gate (A-gate) formed by the bundle crossing of the S6 transmembrane helices and the slow inactivation gate in the selectivity filter. These two gates are bidirectionally coupled. If coupling involves the rearrangement of the S6 transmembrane segment, then we predict state-dependent changes in the accessibility of S6 residues from the water-filled cavity of the channel with gating. To test this, we engineered cysteines, one at a time, at S6 positions A471, L472, and P473 in a T449A Shaker-IR background and determined the accessibility of these cysteines to cysteine-modifying reagents MTSET and MTSEA applied to the cytosolic surface of inside-out patches. We found that neither reagent modified either of the cysteines in the closed or the open state of the channels. On the contrary, A471C and P473C, but not L472C, were modified by MTSEA, but not by MTSET, if applied to inactivated channels with open A-gate (OI state). Our results, combined with earlier studies reporting reduced accessibility of residues I470C and V474C in the inactivated state, strongly suggest that the coupling between the A-gate and the slow inactivation gate is mediated by rearrangements in the S6 segment. The S6 rearrangements are consistent with a rigid rod-like rotation of S6 around its longitudinal axis upon inactivation. S6 rotation and changes in its environment are concomitant events in slow inactivation of Shaker KV channels.
Voltage-gated potassium channels (KV) play essential roles in a large variety of physiological processes: they contribute to the maintenance of the resting membrane potential of cells, influence the frequency and duration of action potentials in excitable tissues (Dodson and Forsythe, 2004), or regulate membrane potential-driven processes of non-excitable cells, such as antigen-dependent activation of T lymphocytes (Panyi et al., 2004; Varga et al., 2007). KV channels are tetramers formed by four subunits containing a K+-selective transmembrane pore. Each subunit consists of six transmembrane α-helical sequences (S1–S6) and intracellular N- and C-termini. S1–S4 helices form the voltage sensing domain (VSD) that moves from its resting position toward the extracellular compartment upon depolarization, as reported by gating current and voltage-clamp fluorometry measurements (Larsson et al., 1996; Baker et al., 1998; Bezanilla, 2000; Yellen, 2002). The pore domain (PD) is composed of the S5 and S6 helices and the pore loop connecting them contains the selectivity filter (SF) lined with carbonyl oxygen atoms that specifically interact with K+ ions allowing rapid and selective conduction (Fig. 1 A; Doyle et al., 1998; Bernèche and Roux, 2001, 2003; Bernèche and Roux, 2001; Morais-Cabral et al., 2001; Bernèche and Roux, 2003). The activation gate (A-gate or lower gate) that controls access to the pore (Liu et al., 1997; Perozo et al., 1999; del Camino and Yellen, 2001) is formed by the crossing of the S6 helices of each of the four subunits at the intracellular end of the pore (“bundle crossing”). The A-gate is functionally coupled to the VSD (Lu et al., 2002).
KV channels open in response to membrane depolarization allowing transmembrane K+ fluxes (Liu et al., 1997). During sustained depolarization, most KV channels enter a non-conducting inactivated state despite the continued presence of the activating stimulus. Inactivation proceeds by two distinct mechanisms, N- and C-type inactivation, which have been studied in detail and reviewed extensively (Kurata and Fedida, 2006). N-type inactivation is well described and occurs via the “ball-and-chain” mechanism (Hoshi et al., 1990; Zagotta et al., 1990), originally proposed for Na+ channels (Bezanilla and Armstrong, 1977). In the absence of N-type inactivation, KV channels are still able to inactivate through slow inactivation (Hoshi et al., 1991; Hoshi and Armstrong, 2013). The Shaker-IR channel lacking N-type inactivation is a dedicated model for examining the slow inactivation of KV channels (Hoshi et al., 1990, 1991).
Experimental, computational, and structural investigations suggest that slow inactivation is caused by molecular rearrangements at or near the SF region (here referred to as the inactivation gate or upper gate) that disrupt the interaction of the SF with K+ ions. This creates a non-conducting permeation pathway (Zhou and MacKinnon, 2004; Cordero-Morales et al., 2006a; Cordero-Morales et al., 2006b; Cuello et al., 2010a; Cuello et al., 2010b; Cuello et al., 2017; Li et al., 2018; Karbat et al., 2019; Szanto et al., 2021). Two opposite modalities, pore dilation versus constriction, have been proposed to constitute the mechanism (Loboda et al., 2001; Cuello et al., 2010b; Hoshi and Armstrong, 2013; Li et al., 2018; Reddi et al., 2022; Tan et al., 2022). Moreover, several pieces of evidence support the idea that an extensive network of H2O–H2O and H2O–protein hydrogen bond interactions behind SF are important for slow (“C-type”) inactivation in the bacterial KcsA K+ channel (Cordero-Morales et al., 2011; Ostmeyer et al., 2013; Pless et al., 2013; Cuello et al., 2017; Labro et al., 2018; Li et al., 2018) and in Shaker-like channels, as well (Karbat et al., 2019; Szanto et al., 2021).
According to the classic view, the combination of the A- and the inactivation-gates results in four composite gating states (Fig. 1 C). Left-to-right movement is the opening of the activation gate in response to membrane depolarization (C→O or CI→OI, where C and O represent the closed and open conformations of the activation gate, respectively, and I indicates a closed inactivation gate). Sustained depolarizations lead to slow inactivation (O→OI), and then, returning to sufficiently negative membrane potentials induces the closure of the A-gate (O→C and/or OI→CI). Using intramolecular Cd2+ bridges, we showed that locking the A-gate of the T449A/V476C Shaker-IR channel in the open state, the locked-open channel impedes recovery from inactivation, supporting that the OI→CI transition is a prerequisite for recovery from inactivation (Szanto et al., 2020).
Functional crosstalk between these two gates in Shaker K+ channels is well-established: closure of the inactivation gate speeds opening and slows closing of the activation gate, i.e., stabilizing the gate in the open configuration (Panyi and Deutsch, 2006). Correlated movement of the activation gate at the bundle crossing (Fig. 1 A) with the movement of the slow inactivation gate has been observed directly by x-ray crystallography and EPR spectroscopy in the prokaryotic KcsA K+ channel (Cuello et al., 2010a). Furthermore, recent high-resolution cryo-EM of Shaker-IR and its non-conducting, permanently inactivated W434F mutant shows that slow inactivation involves an unanticipated conformational rearrangement of the external pore leading to rearrangements of the P-loop and dilation of the ion selectivity filter (Tan et al., 2022). However, these static cryo-EM screenshots do not necessarily reveal the molecular mechanism by which the gates are coupled. Structural rearrangements were reported in the selectivity filter region but surprisingly not in S5 or S6. Either no S5/S6 rearrangements occur, or the limited image resolution or the inappropriate time-window preclude the detection of coupled S5/S6 events. Therefore, the nature of coupled molecular movements upon slow inactivation requires direct functional assays.
To investigate the molecular mechanism of this coupling, we engineered cysteine residues at strategic positions in S6 and determined their state-dependent modification kinetics by methanethiosulfonate (MTS) reagents using patch-clamp techniques (Sheng et al., 2001; Liu et al., 2006; Sheets and Hanck, 2007; Börjesson and Elinder, 2011; Schmid and Grissmer, 2011). Covalent modification by these reagents renders the channel non-conducting, thereby allowing the reaction to be monitored by measuring the ionic current. This approach provides information about the state-dependent accessibility of cysteines (Liu et al., 1997). For example, the accessibility of cysteine residues at I470 and V474 in Shaker change upon inactivation: they are modified ∼20-fold faster in the open state than in the inactivated state (Panyi and Deutsch, 2006; Panyi and Deutsch, 2007), consistent with an increased accessibility in the open state. This suggests allosteric communication between the gates mediated by conformational changes along S6 (Yifrach and MacKinnon, 2002; Sadovsky and Yifrach, 2007). We hypothesize (1) that inactivation produces a conformational change in S6, possibly a rotation about its longitudinal axis, and (2) this rearrangement contributes to the allosteric communication between the gates. Since 471–473 side chains face away from the cavity in the open state (Fig. 1. A and B) and are thus inaccessible, an S6 rearrangement or rotation predicts a concomitant change in cysteine accessibility at A471, L472, or P473 positions upon slow inactivation.
We tested this hypothesis by mutating residues A471, L472, and P473 to cysteines, one at a time, and determined MTS accessibility in the open, closed, and slow-inactivated state of the channel. As expected, none of the cysteines in positions 471, 472, and 473 can be modified either in the closed or the open state. In contrast, inactivation increases the accessibility of residue A471C and P473C for MTSEA (ethylammonium methanethiosulfonate) but not for the larger MTSET (ethyltrimethylammonium methanethiosulfonate). This state-dependent modification of the A471C and P473C mutants is consistent with S6 rotating around its own longitudinal axis upon inactivation, thereby mediating the coupling between the activation and inactivation gates.
Materials and methods
Human embryonic kidney cells transformed with SV40 large T antigen (tsA201) were grown in DMEM-high glucose medium supplemented with 10% FBS, 2 mM l-glutamine, 100 U/ml penicillin-G, and 100 μg/ml streptomycin (Invitrogen) at 37°C in a 5% CO2 and 95% air humidified atmosphere. Cells were passaged twice per week following a 2–3 min incubation in PBS containing 0.2 g EDTA/liter (Invitrogen).
DNA clones and site-directed mutagenesis
Modified Shaker-IR channels in a GW1-CMV mammalian expression plasmid under the control of a highly expressing Kozak consensus promoter sequence (Kozak, 1991) were provided by R. Horn (Thomas Jefferson University, Philadelphia, PA; Ding and Horn, 2002). Mutations were introduced by PCR mutagenesis into Shaker-IR containing the deletion of residues 6 through 46 (Δ6–46) to eliminate the fast N-type inactivation (Hoshi et al., 1990). Moreover, all constructs include C301S and C308S point mutations to exclude possible modifications by MTS reagents of endogenous Shaker cysteines and accompanying functional effects. Point mutations expressing T449A/A471C, T449A/L472C, and T449A/P473C were introduced by using a site-directed mutagenesis kit (QuikChange; Stratagene). Calcium Phosphate Transfection Kit (Invitrogen) was used to co-transfect a GFP plasmid vector together with the Shaker-IR construct in a 1:10 molar ratio. As we could not detect any current from the homotetrameric T449A/P473C channels, a combination of homo- and heterotetramers were expressed by co-transfection of the T449 and T449A/P473C channel constructs at a molar ratio of 3:1. Heterotetramers are indicated as T449A/P473C//T449 throughout the text. Transfected cells were replated onto 35-mm polystyrene cell culture dishes (Cellstar, Greiner Bio-One) pretreated with poly-L-ornithine (Sigma-Aldrich) to improve cell adhesion for excising patches. Channels were expressed ∼24–48 h after transfection. The GFP-positive transfectants were identified by a Nikon Eclipse TS100 fluorescence microscope using bandpass filters of 455–495 and 515–555 nm for excitation and emission, respectively. More than 70% of the GFP-positive cells expressed the co-transfected ion channels, as well.
Solutions and perfusion system
All ionic current experiments were performed with excised inside-out patches. The standard intracellular (bath) solution (ICS) contained (in mM) 105 KF, 35 KCl, 10 EGTA, and 10 HEPES titrated to pH 7.36–7.38 with KOH for a final concentration of ∼160 mM K+ and an osmolarity of 285–295 mOsm, while the standard extracellular (pipette-filling) solution (ECS) was (in mM) 150 NaCl, 5 KCl, 1.5 CaCl2, 1 MgCl2, and 10 HEPES at pH 7.36–7.38 with NaOH and an osmolarity of ∼290 mOsm. For experiments using 50 mM K+, the composition of the 50 K+-ICS was (in mM) 50 KF, 55 NaF, 35 NaCl, 10 HEPES, and 10 EGTA. MTSET and MTSEA (Toronto Research Corp.) solutions were made fresh in ICS from 100 mM stocks in water stored at −80°C. MTS reagents were freshly diluted into the ICS from stocks and loaded immediately into the perfusion system ∼1 min before the start of MTS application. Freshly diluted MTS reagents were loaded into the perfusion apparatus every ∼10 min.
A Warner Instruments SF-77 A Perfusion Fast-Step system with three-barrel square glass (700 μM internal diameter) was used for rapid solution exchange. The patches were perfused with solutions at a rate of 0.5 ml/min. Fig. S1 shows the calibration of the solution exchange. The solution exchange is determined by a mechanical/electrical delay (d) followed by a rapid exponential decay characterized by an exchange time constant (τe). The medians of d and τe were 24.9 ± 0.7 and 10.9 ± 1.0 ms, respectively (n = 30). Based on the kinetics, complete solution exchange was achieved in ∼4 × τe = 40 ms.
Standard methods (Hamill et al., 1981) were used to record currents in inside-out patches. The direction of K+ currents and the voltage protocols are presented according to general conventions. Typical current amplitudes were 300–3,000 pA at +50 mV test potential, thereby allowing the recording of macroscopic currents. Micropipettes were pulled in four stages by using a Flaming Brown automatic pipette puller (Sutter Instruments) from Borosilicate Standard Wall with Filament aluminum-silicate glass (Harvard Apparatus Co). Tip diameters ranged from 0.5 to 1 μm and heat polish gave tip resistances of 2–8 MΩ. All measurements were carried out using an Axopatch 200B amplifier connected to a personal computer using Axon Digidata 1320 data acquisition hardware (Molecular Devices Inc.). In general, the holding potential was −120 mV. Records were discarded when the leak at the holding potential was >10% of the peak current at the test potential. Experiments were done at room temperature ranging between 20 and 24°C. Data were analyzed using the pClamp9 software package (Molecular Devices Inc.) and GraphPad Prism 8 (GraphPad). Before analysis, current traces were digitally filtered with a three-point boxcar smoothing filter. Reported errors are SEM.
To determine the essential biophysical properties of the investigated mutant channels, we studied the kinetics of activation and inactivation, the voltage dependence of steady-state activation and inactivation, and the kinetics of recovery from slow inactivation. To study activation kinetics, cells were depolarized to +50 mV for 15 ms from a holding potential of −120 mV every 15 s. K+ current traces were fitted with a single exponential function rising to the maximum according to the Hodgkin-Huxley n4 model where Ia is the amplitude of the activating curve component, τact is the activation time constant of the current, and C is the amplitude of the non-activating current component. The τact for a particular cell was determined as the average of τact values obtained for three to four depolarizing pulses repeated every 15 s. We recorded n = 22–24 cells of a given Shaker-IR construct, and the average of τact was used to demonstrate the activation kinetics. To study channel inactivation of T449A/A471C and T449A/L472C, 1-s-long depolarizing pulses to +50 mV were applied from a holding potential of −120 mV every 60 s. The rapid activation of the current is followed by a decay corresponding to slow inactivation. A single exponential function [I(t) = I0 × exp(−t/τi) + C], where I0 is the amplitude of the inactivating component of the current, τi is the inactivation time constant for the given Shaker-IR construct, and C is the steady-state current at the end of the depolarizing pulse, was fitted to the decaying part of the current traces to obtain the time constant (τi) characterizing the inactivation kinetics. In contrast, traces recorded from the T449A/P473C//T449 heterotetramers clearly showed a fast and a slow component of the decay, therefore, the falling phase of the currents was fit by the sum of two exponential terms, [I(t) = Afexp(−t/τi,f) + Asexp(−t/τi,s) + C] giving τi,f and τi,s. To determine the voltage dependence of steady-state activation of the current, the cells were held at −100 mV and depolarized to test potentials ranging from −80 to +70 mV in 10-mV increments every 60 s. Peak conductance (G(V)) at each test potential was calculated using G(V) = Ipeak/(Em − EK), where Ipeak is the peak current at a test potential of Em and EK is the reversal potential of K+ (−86.9 mV, based on the Nernst equation). The G(V) values were then normalized for the maximum conductance (G/G0) and plotted as a function of test potential along with the best-fit Boltzmann function where G/G0 is the normalized conductance, V is the test potential, V½ is the midpoint voltage, and s is the slope factor of the function. To describe the voltage dependence of steady-state inactivation, the fraction of noninactivated channels at each test potential was calculated as I/I−120, where I is the peak current evoked by the depolarization from a given prepulse potential and I−120 is the peak current evoked by identical depolarization from the holding potential of −120 mV. V1/2 and k were determined by fitting a Boltzmann function to the data points. To study the kinetics of recovery from inactivation, pairs of depolarizing pulses were applied from the holding potential of −120 to +50 mV for 400 ms. The duration of the first step was used to inactivate channels and measure the initial peak current amplitude (I1). After a recovery period, defined as the interpulse interval (ipi) at −120 mV, the second identical voltage step was applied and the peak amplitude of the recovered current (I2) was measured. The ipi at −120 mV varied between 1.8 and 60 s. The fractional recovery (FR) at a given ipi was calculated as FR = (I2 − Iss1)/(I1 − Iss1), where I2 and I1 are the peak currents of the second and first pulse, respectively, and Iss1 is the steady-state current at the end of the first depolarizing pulse. The FR versus ipi plot was fit with an exponential function rise to a maximum containing a single term, FR(t) = 1 – exp(−t/τrec) to give the time constant of recovery from inactivation, τrec. Prior to analysis, current traces were corrected for ohmic leak. Nonlinear least square fits were done using the Levenberg–Marquardt algorithm. The number of experiments on different cells (n) involved in the given analysis is shown in the text. Data are expressed as mean ± SEM.
Cysteine modification assay
Inside-out patches were depolarized from a holding potential of −120 to +50 mV for the time required to achieve the open or fully inactivated state of the channels. Pulse protocols were repeated three to four times in the absence of the MTS reagents to verify the stability of the peak currents. Patches showing >10% rundown in peak current were discarded. MTSEA (2 mM) or MTSET (0.2 mM) were applied for the indicated duration (L) either at the hyperpolarized holding potential (e.g., −120 mV) or during depolarization, depending on the state of the channel being probed. Voltage protocols used for state-dependent accessibility assays were repeated 8–12 times in the presence of the modifying agents to monitor their effect. The normalized current was calculated as I(t)/I0, where I0 is the peak current at +50 mV prior to application of the MTS reagent and I(t) is the peak current at cumulative MTS modification time t. A single exponential function was fit to the normalized current-cumulative MTS modification time graph: I(t) = I0 × exp(−t/τmod) + C, where τmod is the modification time constant and C is the steady-state current at the equilibrium current reduction. The modification rate constant (kmod) was determined as 1/(τmod × [MTS]), where [MTS] is the molar concentration of the MTS reagent. Nonlinear least square fits were done using the Levenberg–Marquardt algorithm. Data are expressed as mean ± SEM.
Online supplemental material
Fig. S1 shows the characterization of the solution exchange kinetics of the perfusion system. Fig. S2 shows the biophysical characterization of the T449A/L472C Shaker-IR channel K+ currents. Fig. S3 shows the biophysical characterization of the heteromeric T449A/P473C//T449 Shaker-IR channel K+ currents. Fig. S4 shows the extended application (300 ms) of MTSEA to the T449A/A471C Shaker-IR channel in the open state.
Biophysical characterization of the mutant channels
To study the accessibility of S6 residues 471–473, we introduced these cysteines, one at a time, into a well-established Shaker-IR background (Yellen, 2002). This background construct has an N-terminal deletion (Δ6–46) to eliminate fast N-type inactivation and C301S and C308S to obviate modification of native cysteines. In addition, we replaced the native T449 at the external mouth of the channel with alanine to produce a slow inactivation rate compatible with the planned experiments. Wild-type Shaker-IR inactivates slowly (τi ∼1,000 ms), whereas the T449A mutant inactivates considerably faster (τi ∼100 ms; López-Barneo et al., 1993; Szanto et al., 2021), which is ideal for cysteine modification measurements (see below; Panyi and Deutsch, 2006; Panyi and Deutsch, 2007). Thus, we engineered T449A/A471C, T449A/L472C, and T449A/P473C mutants in the Shaker-IR background. The first two mutants expressed K+ currents in tsA201 cells (Fig. 2 A and Fig. S2 A), whereas the T449A/P473C mutant did not, similar to other Shaker-IR channels mutated at residue P473 (Hackos et al., 2002) or KV1.2 channels mutated at the analogous position (Choe and Grabe, 2009). However, functional heterotetrameric channels were produced by coexpression of T449A/P473C and the T449 WT Shaker-IR background that has a proline at 473, referred to as T449A/P473C//T449 (see below, Figs. 3, 4, 5, and 6; and Fig. S3). Although the random subunit stoichiometry results in a heterogeneous channel population, this construct enabled us to perform the accessibility experiments.
First, we characterized the electrophysiological properties of the mutant channels in inside-out patches to optimize our pulse sequence voltage protocols and the timing of the application of MTS reagents. Figs. 2, S2, and S3 show the essential gating parameters of all Shaker-IR constructs including the voltage dependence of activation (peak conductance–voltage, or G–V, curve), the voltage dependence of steady-state inactivation, the kinetics of activation, slow inactivation, and recovery from inactivation. The G–V curve was calculated from the current–voltage relationship, which was obtained with a series of 1-s-long depolarizing pulses ranging from −80 mV to +70 mV every 60 s (Fig. 2 A, Fig. S2 A, and Fig. S3 A), and the voltage dependence of the conductance was determined (Fig. 2 B) by fitting a Boltzmann function to the normalized conductance (G/G0)-test potential relationships to give the midpoint voltage (V1/2) and slope factor (s). Although the mutations produce a range of midpoint voltages (see Table 1 for V1/2 and s values, similar to other Shaker mutations that change V1/2, s, or both; Hackos et al., 2002; Soler-Llavina et al., 2006), each channel was fully activated by depolarizing pulses to +50 mV. The activation time constants of the currents are similar for all constructs (0.3–0.7 ms) in response to a depolarization to +50 mV. Pulses of 5 ms duration were sufficient to activate all channels. Time constants of slow inactivation (Fig. 2 C, Fig. S2 C, and Fig. S3 C) were obtained by fitting decaying exponential functions to the current traces recorded during long (1.0–3.0 s) depolarizing pulses. Traces from T449A/A471C and T449A/L472C were well fit by single exponential functions, but traces recorded from the T449A/P473C//T449 heterotetramers contained a fast and a slow component of the decay (see Table 1 for the time constants of inactivation, τi). Nevertheless, during depolarizing pulses of a 3-s duration, the current amplitudes decayed to <15% of the peak values, consistent with the majority (>85%) of the channels having entered the slow inactivated state. To assess the voltage dependence of steady-state inactivation (h∞), we applied 3-s-prepulse potentials from −110 to −20 mV in 10-mV increments before stepping the test potential to +50 mV for 5 ms. The fraction of noninactivated channels at each voltage was calculated as I/I−120 and plotted as a function of prepulse potential (see Materials and methods). The voltage dependence of steady-state inactivation was characterized by V1/2 and a slope factor determined from Boltzmann fits for I/I−120–prepulse potential relationships (Fig. 2 D, Fig. S2 D, and Fig. S3 D; see Table 1 for V1/2 and s values). The kinetics of recovery from inactivation were studied using pairs of depolarizing pulses delivered from the holding potential of −120 to +50 mV for 400 ms and varying the ipi at −120 mV. The FR versus ipi plot was fit with a single-exponential function regardless of the channel mutant to give the time constant of recovery from inactivation (τrec; Fig. 2 F, Fig. S2 F, and Fig. S3 F; see Table 1 for the time constants). Recovery time constants for the T449A/A471C, T449A/L472C, and T449A/P473C//T449 channels were 9.4 ± 0.8 (n = 5), 4.4 ± 0.7 (n = 5), and 1.8 ± 0.2 (n = 5) s, respectively. Full recovery from inactivation was achieved by holding the patch at −120 mV for a duration of at least five times longer than the time constant for recovery from inactivation for a given construct.
The gating parameters (Table 1) were used to define the voltage range (more than +50 mV) where currents quickly and fully activate, and slow inactivation can be studied in isolation. Moreover, all of these data allowed us to define a sufficiently negative holding potential (−120 mV) and interepisode time (>45 s) that ensured full recovery of the channels from inactivation and that all channels are in the closed state at the holding potential. Altogether, the kinetic parameters of the mutant channels, i.e., an inactivation time constant of a few tens of milliseconds and a recovery time constant of a few seconds, rendered them ideal for MTS modification measurements, whose aim is to determine whether inactivation alters the solvent accessibility of S6 residues, thereby implicating a mechanism by which the gates are coupled.
Cysteine accessibility in the closed and open states of the channels
A previous cysteine scan of the S6 transmembrane segment between the selectivity filter and the bundle crossing (positions 470–477) showed that these residues are not accessible to MTS reagents when the activation gate is closed, but residues at positions 470 and 474–477 were readily modified in the open state (Liu et al., 1997). However, no significant modification of residues 471–473 was detected in the open state. We first confirmed these findings in our system before examining the accessibility of these residues with MTS reagents in the slow-inactivated state. The accessibility of the introduced cysteine residues was tested using the neutral/partially positively charged MTSEA (3.6 Å) and the larger and permanently positively charged MTSET (5.8 Å). These differences in size allowed us to explore the spatial alterations in the molecular microenvironment of S6 at the water-filled cavity upon inactivation. The closed state is characterized by an open inactivation gate and closed activation gate, as shown by pictograms in Fig. 3. The heterotetrameric T449A/P473C//T449 channels are indicated by a combination of a white- and grey-filled rectangle in the pictograms throughout the text. The accessibility of the closed state was studied as follows: the channels were exposed to 2 mM MTSEA or 0.2 mM MTSET for 500 ms (L = 500 ms, see Materials and methods) while being held closed at −120 mV. After a complete wash-out of the reagents by perfusing the patch with MTS-free ICS for 600 ms (lag time), the effect of the MTS application on the peak K+ currents was assessed by brief depolarizing pulses to +50 mV. A step to +50 mV ensures that all channels open quickly since the opening probability is maximal at this voltage (Fig. 2 B; Fig. S2 B and Fig. S3 B). These steps were repeated every 15 s. As shown in Fig. 3, neither MTSEA nor MTSET caused a significant current reduction in any of the mutant channels confirming the inaccessibility of these pore-lining residues in a channel with a closed activation gate. The normalized currents after 3.5 s cumulative perfusion with MTSEA and MTSET were 0.935 ± 0.002 (n = 5) and 0.924 ± 0.043 (n = 4) for T449A/A471C; 0.954 ± 0.014 (n = 4) and 0.947 ± 0.025 (n = 4) for T449A/L472C; and 0.939 ± 0.024 (n = 5) and 0.947 ± 0.036 (n = 4) for T449A/P473C//T449, respectively.
In the open state (state O in Fig. 1 C), both the activation gate and the inactivation gate are in their conducting conformation (see cartoons in Fig. 4). The inside-out patches were perfused with the MTS-free, KF-based ICS solution during the interpulse holding potential of −120 mV. Perfusion with MTS reagents started 200 ms prior to the depolarizing pulse that ensures solution exchange considering the delays in the perfusion system (Fig. S1) and the MTS reagent (2 mM MTSEA or 0.2 mM MTSET) was present during the entire duration (5 or 10 ms) the channels were held open. It was necessary to confine the duration of the depolarizing pulses to 5–10 ms to isolate the open conformation of both gates and avoid the entry of channels into the inactivated state. The voltage and MTS application protocol was repeated every 15 s. The peak currents for each pulse (I(t)) were normalized to the peak current of the last pulse in the absence of the MTS reagent (I0) and plotted as a function of the cumulative modification time as shown in Fig. 4. Our results confirmed previous findings that cysteines at positions 471–473 are not modified by MTS reagents even in the open state of the channel when the A-gate is open. The normalized currents after 12 consecutive depolarizing pulses in MTSEA and MTSET were 0.982 ± 0.019 (n = 5) and 0.936 ± 0.06 (n = 6) for T449A/A471C; 0.936 ± 0.034 (n = 6) and 0.947 ± 0.057 (n = 7) for T449A/L472C; and 0.949 ± 0.041 (n = 6) and 0.951 ± 0.012 (n = 3) for T449A/P473C//T449, respectively. Even when the cumulative MTSEA exposure of the T449A/A471C open state is as long as 300 ms, there is no reduction in current (the normalized current after 60 consecutive depolarizing pulses in MTSEA was 0.935 ± 0.018 [n = 6]; Fig. S4), thereby suggesting a lower probability that slow open-state modification will contaminate our measurements of modification of the inactivated state.
Cysteine accessibility in the inactivated channel
I470C and V474C (see Panyi and Deutsch, 2006), but not 471–473C, are accessible to MTS reagents in the open state. This may be explained by their position on the S6 helix where side chains of 471–473 face away from the aqueous cavity pathway in the open state (Fig. 1, A and B). However, cysteines in these positions may rotate to become solvent accessible during the conformational rearrangements associated with slow inactivation. We tested this hypothesis by applying the MTS reagents during a period when the majority of the channels are in the inactivated state (Figs. 5 and 6). The A471C and L472C mutants with fast inactivation were completely inactivated by depolarizing pulses of 1-s duration. Application of the MTS reagents was started 300 ms after the start of the depolarization and lasted for L = 500 ms. Fig. 5, A and B, shows that intracellular application of 2 mM MTSEA to inactivated A471C channels irreversibly reduced the current with a time constant of 1.24 ± 0.12 s (n = 13). The modification rate constant (kmod) is calculated as 1/(τmod × [MTSEA]), where [MTSEA] is the molar concentration of MTSEA. The average of 13 replicate experiments yields a kmod for the inactivated state of 485 ± 82 M−1s−1. The modification rate of T449A/A471C channels in the OI state by MTSEA (485 M−1s−1) is slightly slower than that of the T449A/I470C construct (660 M−1s−1; Panyi and Deutsch, 2007) but substantially slower than the modification of V474C (∼18,000 M−1s−1; Panyi and Deutsch, 2007). However, MTSET at 0.2 mM caused no current reduction under the same conditions (Fig. 5, C and D). L472C was not affected by either reagent (Fig. 6, A and B).
Due to the slow biphasic inactivation of T449A/P473C//T449, we used longer depolarizing pulses of 3-s duration and started MTS application 2 s after the start of the depolarizing pulse, and the reagents were applied for an extended duration of L = 800 ms. In the case of mutant P473C, current from heterotetramer channels comprised of T449A/P473C and T449/P473 subunits was reduced by MTSEA, but not MTSET (Fig. 6, C–F). The kinetics and the modification rate of P473C in the OI state could not be mathematically determined since the normalized peak current versus cumulative modification time relationship does not follow a single exponential, but rather, a sigmoidal curve. These complex kinetics may originate from the weight of each channel type having 1, 2, or 3 of the T449A/P473C subunits present in the heterotetramer and consequently, an undetermined number of cysteines to be modified to result in current loss (Yellen et al., 1994). This uncertainty is also reflected in the larger error bars in the normalized current-cumulative modification time graph in Fig. 6 D. In contrast, the homotetrameric wild-type (T449/P473) channel was not modified by MTSEA (Fig. 6 D, empty circles), indicating specific interaction of the reagent with the engineered P473 cysteine in the heterotetramer. Regardless, qualitatively, the P473C residues are modified by MTSEA in the OI state over a comparable cumulative modification time range (∼8 s) as A471C (∼4 s).
In this study, we investigated the state-dependent accessibility of cysteine residues in the S6 helix of the voltage-gated Shaker K+ channel to gain insight into the molecular mechanism of coupling between the A-gate and the slow-inactivation gate of the channel (Panyi and Deutsch, 2006, 2007). The communication between the two gates is bidirectional: the closed slow-inactivation gate influences the rate of A-gate opening and closure (Panyi and Deutsch, 2006) whereas A-gate opening precedes the closure of the inactivation gate. Moreover, recovery from inactivation requires A-gate closure (Szanto et al., 2020). The molecular rearrangements in the selectivity filter associated with slow inactivation and the consequent changes in the K+ occupancy of the K+ coordination sites controlling slow inactivation have already been proposed by x-ray crystallography and MD simulations in KcsA (Zhou and MacKinnon, 2004; Cuello et al., 2010b; Cuello et al., 2017; Li et al., 2018; Renart et al., 2019) and KV1.2 (Reddi et al., 2022), and more recently in Shaker and KV1.3 using cryo–EM (Ong et al., 2022; Tan et al., 2022). Despite these characterizations of the inactivated state of Shaker, several questions remain unanswered.
For example, how are filter conformational changes coupled to structural changes in the activation gate? What is the molecular mechanism of the coupling between the activation and inactivation gate? To address these issues and the status of the S6 helix upon slow inactivation, we used state-dependent modification assays as a tool, similar to that first used by Yellen and co-workers, to locate the activation gate in the lower portion of S6 (Liu et al., 1997; del Camino and Yellen, 2001). We found that slow inactivation increases the accessibility of S6 residues that are buried in the open state of Shaker-IR channels. This is consistent with the idea that communication of the gates is mediated by rearrangements in the S6 segment, as discussed below.
We have studied the kinetics of slow inactivation in three Shaker-IR mutants having inactivation time constants ranging from ∼15 to ∼600 ms. The current activates quickly (within 1 ms) in all three mutants (471–473) and the inactivation kinetics are fast enough to conduct the experiments on a reasonable time scale. All three mutants contain an alanine mutation at position T449 that has already been confirmed to modify rather than disrupt slow inactivation (López-Barneo et al., 1993; Cordero-Morales et al., 2006b; Hoshi and Armstrong, 2013; Cuello et al., 2017; Szanto et al., 2021). These indicate that the mutants utilized in our study are suitable to study slow inactivation. The mutants modify the inactivation rate rather than directly alter the conformational changes in the selectivity filter that underlie slow inactivation (Tan et al., 2022).
The state-dependent cysteine accessibility technique is an excellent probe to study conformational changes occurring within the K+ channel pore, such as those that are associated with inactivation (Yellen et al., 1994; Liu et al., 1996; Panyi and Deutsch, 2006, 2007) and those underlying the reduced intracellular TEA+ affinity of the inactivated Shaker-IR channels (Panyi and Deutsch, 2007). This method was also used to demonstrate that several pore-lining residues in S6 are rapidly modified by MTS reagents faster than when the channel is inactivated or closed (Liu et al., 1997; Panyi and Deutsch, 2006).
The results displayed in Fig. 3 constitute evidence that when the activation gate is closed, no modification of cysteine residues (471–473) occurs as the hydrophilic MTS molecules are unable to penetrate the membrane. Therefore, the closed activation gate prevents modification of cysteine side chains. In contrast, when the activation gate is open (O state, Fig. 1 C), S6 residues, namely I470C and V474C, are modified by MTS reagents (Liu et al., 1997; Panyi and Deutsch, 2006; Panyi and Deutsch, 2007). Residues 471–473C, however, are inaccessible to cysteine-modifying reagents in the open state of the channels. These findings suggest that only S6 residues facing the water-filled cavity in the open state are available for MTS modification.
In contrast, the accessibility of selected cysteine residues in the OI state of the channels, where the A-gate is still open but the slow inactivation gate is closed (Fig. 1 C), is clearly demonstrated by the progressive decrease of the peak currents when MTSEA was repeatedly applied to the T449A/A471C and T449A/P473C//T449 channels (Figs. 5 and 6). When heteromeric channels were exposed to MTSEA, some of the channels were still in the O state (Fig. 6 C). As P473C in the open state cannot be modified by MTSEA, we argue that the fraction of channels in the open state does not compromise our conclusion that the loss of the current is a consequence of channel modification in the OI state. However, cell-to-cell variability of the inactivation kinetics and the fraction of channels in the MTSEA-modifiable OI state may contribute to the relatively large errors and the complexity of the curve in Fig. 6 D.
The slower modification rate for T449A/A471C and T449A/P473C//T449 compared to T449A/I470C and T449A/V474C may reflect the limited access of 471C and 473C side chains for MTSEA as they face away from the central cavity, but slow inactivation exposes them sufficiently for MTS modification. Similar experiments using 0.2 mM MTSET produced only a negligible current reduction in the inactivated state of the channels. This may be due to the larger size of MTSET (5.8 Å) compared with MTSEA (3.6 Å) rather than the charge or charge delocalization. Moreover, the A471C residues on each subunit are relatively far away from each other as intracellular Cd2+ is unable to form crossbridges and block ion conduction (data not shown), a clear contrast to cysteines at positions V474 and I470 where intersubunit Cd2+ crossbridges form in both the open and OI states. Consistent with these data, residue L472C, which faces away from the water-filled cavity (see helical wheel, Fig. 1 B), is completely inaccessible to MTS reagents regardless of the state of the gates. The current is not modified by either MTSEA or MTSET in either the O or OI gating states.
A rigid-body rotation of S6 may predict a simultaneous reduction in the accessibility of V474C and I470C and an increase in the accessibility of A471C, whereby A471C is repositioned with respect to the aqueous cavity during a counter-clockwise turn of S6 (see Fig. 1 B). Consequently, the rotation of S6 upon inactivation displaces L472C toward A471C’s position in the open state. Thus, L472C is still inaccessible to MTS reagents in the OI state. This rotation scenario would put P473C into an even less accessible position. A plausible explanation for the accessibility of P473C in our study (Fig. 6, C and D) may be that the symmetry of the channel is broken in the P473C heterotetramer (as suggested by the biphasic inactivation kinetics in Fig. 6, C–E). This asymmetry may result in an unexpected exposure of P473C to MTSEA. To our knowledge, our experiments are the first to address the rearrangements of S6 in the inactivated state by systematically substituting amino acids for cysteines along S6 and determining their accessibility to MTS reagents in O vs. OI gating states of the channel. These novel data, combined with previous experiments on the state-dependent accessibility of V474C and I470C, are consistent with the rotation of the S6 helix along its longitudinal axis concomitant with channel entry into the slow inactivated state. Whether this S6 motion is necessary and sufficient to mediate the crosstalk between activation and inactivation gates remains to be elucidated; however, our results provide a mechanistic background for such coupling. Thermodynamic mutant cycle analyses demonstrated that gating-sensitive residues that are as much as 15 Å apart in S6 are coupled to each other (Yifrach and MacKinnon, 2002; Sadovsky and Yifrach, 2007). These residues cluster mainly in the bundle-crossing region of the activation gate and the upstream region where S6 contacts the pore helix. While the underlying interactions responsible for similar coupling during slow inactivation gating in Shaker are not known, one possibility is S6 rearrangement, including a rigid-body rotation of S6, which may be dynamically short-lived in the gating transition and therefore not structurally detectable (Tan et al., 2022). Recent cryo-EM structures of Shaker corresponding to the O state (Shaker-IR) and OI state (W434F mutant of Shaker-IR, which renders the channel permanently slow inactivated; Perozo et al., 1993; Yang et al., 1997) show that the S6 helices in these two structures are overlapping. The rotation of the S6 helix along its longitudinal axis, as proposed by us, can be reconciled with the cryo-EM data if we assume that the S6 rotates transiently during the transition between the O and OI states.
There is precedence in the literature for rigid-body rotation of α-helices in several transmembrane proteins, which indicates that this molecular motion is energetically plausible. For example, F α-helices of bacteriorhodopsin undergo helical rigid-body movements during the photocycle (Xiao et al., 2000). Moreover, the rotation of transmembrane helices relative to each other is a concept that appears to supersede the more traditional idea of ligand-induced dimerization of growth factor receptors (Fleishman et al., 2002; Zakany et al., 2020; Kovacs et al., 2022). EPR spectroscopy and a computational strategy also predicted rigid-body movements, including rotation in three dimensions, for the KcsA channel during activation gating (Sompornpisut et al., 2001).
However, movements of side chains independent of the rigid-body motion of α-helical structures were observed when comparing the open and closed structural models of KcsA (Cuello et al., 2010a, 2010b). An alternative explanation of our data, where S6 does not rotate but rather the microenvironment around the helix is altered with consequent changes in S6 cysteine accessibilities, is thus possible. Such a scenario, e.g., global changes in the water-filled cavity upon inactivation, might account for a change in TEA+ affinity of the inactivated state (Panyi and Deutsch, 2007) and the blocking of KV1.3 by small-molecule inhibitors that bind specifically to the slow-inactivated state (Hanner et al., 2001). Alterations in the microenvironment of S6 residues due to their proximity to the S6 PVPV bend (del Camino et al., 2000; Webster et al., 2004) may also explain different accessibilities. Moreover, it is also known that there are many channels that manifest 310-helices and π-bulges during gating transitions, including KV channels (Villalba-Galea et al., 2008; Infield et al., 2018), calcium-activated Cl− channels (Paulino et al., 2017), and cyclic-nucleotide-gated channels (Clayton et al., 2008). Another intriguing possibility is that S6 may not be a canonical rigid α-helix in the OI state (but might be rigid both in O and C or CI states), which allows more flexible backbone or side-chain movements during slow inactivation. Moreover, the reactivity of certain residues to MTSEA depends on cysteine ionization, the dielectric of the microenvironment, and proximal charged residues (Linsdell, 2015). All of these determinants can be affected by side-chain rotamerizations and/or altered intersubunit interactions that are induced in different gating states.
Regardless of the coupling mechanism between the two gates, the crosstalk between them will surely affect cell excitability because inactivation governs the amount of available K+ channels in cells contributing critically to the resting membrane potential and the shape and frequency of the action potential, all of which influence other membrane potential–dependent processes (e.g., neurotransmitter release; Dodson and Forsythe, 2004) and thus, are associated with several neurological and psychiatric disorders (Adelman et al., 1995; Hübner and Jentsch, 2002; Yellen, 2002; Kurata and Fedida, 2006). On the other hand, state-dependent structural changes bear relevance on drug design, as numerous small molecule channel blockers are known to bind to the central cavity of voltage-gated potassium channels, preferentially in the slow-inactivated state. Due to these interactions, understanding the gating transitions of KV ion channels may also aid in the design of more efficient drug molecules that bind to ion channels in a state-dependent manner. This can be important, e.g., in the development of anti-arrhythmic drugs with higher specificity and more favorable state-dependent properties.
In summary, our results support the idea that the rotation of S6 may mediate the communication between the activation gate and the inactivation gate controlling slow inactivation in KV channels and provide fine details on the contribution of residues lining the cavity to the slow-inactivation gating of Shaker.
The data are available from the corresponding author upon reasonable request by e-mail communication.
Christopher J. Lingle served as editor.
The authors thank Cecilia Nagy and Adrienn Bagosi for their expert technical assistance.
This work was supported by the Hungarian Academy of Sciences projects KTIA_NAP_13-2-2015-0009 and KTIA_NAP_13-2-2017-0013 (to Z. Varga); National Research Development and Innovation Office, Hungary, grants OTKA K132906 (to Z. Varga) and OTKA K143071 (to G. Panyi); Ministry of Human Capacities, Hungary, grant EFOP-3.6.2-16-2017-00006 (to G. Panyi); and Ministry of Finance, Hungary, grant GINOP-2.3.2-15-2016-00044 (to G. Panyi); and National Institutes of Health R01 GM052302 (to C. Deutsch).
Author contributions: T.G. Szanto: Conceptualization, Investigation, Formal analysis, and Writing—Original Draft. F. Papp: Investigation and Formal analysis. F. Zakany: Investigation and Formal analysis. Z. Varga: Conceptualization, Writing—Review & Editing. C. Deutsch: Conceptualization, Writing—Review & Editing, Methodology. G. Panyi: Conceptualization, Writing—Original Draft, Writing—Review & Editing, Funding acquisition, and Methodology.
The authors declare no competing interests exist.