Energy landscape of a Kv channel revealed by temperature steps while perturbing its electromechanical coupling

Abstract

Voltage-dependent potassium channels (Kv) play a crucial role in membrane repolarization during action potentials. They undergo voltage-dependent structural conformational transitions according to their distribution across their energy landscape. Understanding these transitions helps us comprehend their molecular function. Here, we used sudden and sustained temperature changes (Tstep) combined with different voltage protocols and mutations to dissect the energy landscape of the Shaker K⁺ channel. We used two mutations, ILT (V369I, I372L, and S376T) and I384N, that affect the coupling between the voltage sensor (VSD) and the pore domain (PD), to obtain the temperature dependence of VSD last transition and the intrinsic temperature dependence of the pore, respectively. Our findings support a loose or tight conformation of the electromechanical coupling. In the loose conformation, the movement of the VSD is necessary but not sufficient to efficiently propagate the electromechanical energy to open the pore. In contrast, this movement is effectively translated into pore opening in the tight conformation. Our results describe the energy landscape of the Shaker channel and how its temperature dependence can be modulated by affecting its electromechanical coupling.

Introduction

Voltage-gated potassium channels (Kv) regulate neuronal excitability and muscle contraction¹. These channels are tetramers with subunits containing six transmembrane segments (S1-S6). The S1-S4 segments comprise the voltage sensor domain (VSD), and the S5, P-loop, and S6 segments form the pore domain (PD)². The S4-S5 linker connects the VSD to the PD. Upon changes in membrane voltage, the S4-S5 linker plays a crucial role in the transduction of the movement of charged residues in S4 (gating charges) located in the VSD to the opening and closing of the PD gate^3,4. In Kv channels, the coupling between the VSD activation and PD opening is strict, meaning that all four VSDs need to activate before the channel can open, and for one of these VSDs to deactivate, the pore must close first⁵. Thus, the VSD acts directly to change the PD open probability (Po), and the value of Po is dependent on the coupling between VSD and PD. Therefore, the pore opening reflects the energetics of the PD, the VSD and the VSD-PD coupling. Temperature changes, which are often overlooked in the study of ion channel biophysics, offer a direct tool to study these energetic relationships.

The complexity of the Kv channel activation has resisted the inquiries of electrophysiologists and structural biologists. Although we have a wealth of information about the workings of voltage-dependent K⁺ channels, a detailed molecular picture of those changes is still lacking due to the large number of conformational changes that Kv channels undergo during activation. The importance of revealing the molecular details of Kv channel activation resides in these channels’ significance in the adequate physiological function of tissue, organs, and the nervous system⁶. We argue that the Kv activation energy landscape complexity can be dissected when the channel is under a voltage clamp, using temperature steps applied at different times during channel activation with simultaneous measurements of gating or ionic currents, which inform how the channel transits between closed states and the final transition leading to pore opening.

Temperature changes have been extensively used to study the properties of Kv channels^{7,8,9,10,11,12}. The Shaker K⁺ channel is a well-established prototypic Kv channel, and its biophysical characteristics were widely characterized using electrophysiological, spectroscopic, and, more recently, structural methods^13,14,15. Its temperature dependence has been explored in great detail, showing that the channel’s Po does not have a substantial temperature dependence^7,8. However, some transitions in the activation pathway have high enthalpic and positive entropic changes, indicating a significant temperature dependence⁷. Temperature modifies the energy landscape of ion channels by affecting their wells and peaks, the diffusion coefficients of their moving parts and their environment. Despite this extensive characterization, we still lack a detailed understanding of the energy landscape governing the complex process of voltage-dependent activation. A full description of the energy landscape of proteins allows us to comprehend the physical basis of their function. Temperature changes have been frequently used to understand the energy landscape of ion channels, but it usually has been done as a steady temperature change of the bath solution. In this condition, temperature affects the energy landscape without distinguishing close states defined by the voltage sensor activation and open states reflecting the pore gate activation. To dissect the energy landscape of Shaker, we used our recently developed framework for fast-temperature changes¹⁶ that are synchronized with different voltage protocols and mutations favoring the positions of the charges in particular states of the energetic landscape. Using this framework, we could probe state-dependent effects, minimize drift issues common with slower methods, and resolve fast-temperature-dependent transitions that would be masked otherwise.

We rationalized that the temperature dependence of the Po, or lack thereof, depends on the strict coupling between the VSD and PD, a hallmark of Shaker-like Kvs^14,17. Moreover, by altering the coupling between the VSD and the PD, we should be able to modify the activation process in a way that we can study the temperature dependence of isolated events along the channel activation pathway. To approach this experimentally, we used two mutants that affect the coupling between movements of the gating charges in the VSD with the opening of the pore in the PD, the S4 triple mutant V369I, I372L, and S376T (ILT) and the single mutant I384N, a residue in the S4-S5 linker. These mutations provide complementary tools for dissecting the activation energy landscape. ILT significantly separates the last VSD transition before opening^18,19, while I384N specifically disrupts the coupling between VSD activation and pore opening. Therefore, these mutants allow us to isolate and characterize the energetics of individual steps in the activation pathway²⁰.

Using our ability to control temperature and voltage, we determine their effects on the movement of the VSD (gating currents) and opening of the PD (ionic currents). In the WT channel, we observe the closure of the pore, which can be explained by the inward movement of the gating charges by a Tstep. Indeed, we observed an inward movement of the gating charges in response to a Tstep, revealing that temperature directly affects voltage sensor movement, causing a redistribution of voltage sensor states and affecting the voltage dependence of activation. Using the ILT mutant, we determine the thermodynamics of the last transition of the voltage sensor, the voltage-dependent transition that opens the pore. We find that temperature facilitates the last step of the VSD movement, increasing the Po of the ILT mutant. In contrast, temperature causes a substantial decrease in the Po of the I384N mutant in a process independent of the VSD movement, revealing an intrinsic temperature dependence of the pore. Our study and detailed analysis add to the increasing evidence that temperature modulates the gating of ion channels significantly and provides the temperature dependence of individual gating transitions, shedding light on the role of the S4 and S4-S5 linker in the temperature dependence of Kv channels. Moreover, our results show that the temperature dependence of a channel arises from the intrinsic properties of the VSD, the PD, and the coupling between them.

Results

Tstep reveals opposite effects of heating in ILT and I384N

We measured the temperature dependence of ionic conduction using our recently developed framework to generate fast-temperature steps on the oocyte membrane (Fig. 1a, Methods)¹⁶. Since previous studies have indicated that the last concerted step in the opening of the channel has a high-temperature dependence^7,8, we used two mutants that modify the last steps of the activation pathway (ILT and I384N). Specifically, ILT affects the final transition of the VSD activation, and I384N disturbs the coupling by affecting the transition from closed to open after VSD activation. The ILT mutations are located next to the C-terminus of S4, while the I384N is located next to the N-terminus of the S4-S5 linker, where it interacts with the C-terminus of the S6 (Fig. 1b, c). Using the Tstep method, we can apply temperature steps at any time during the activation process, which allows us to obtain the temperature effects at a particular gating transition. After the ionic currents reached the steady state, we applied a temperature step in the middle of a voltage pulse protocol and analyzed the effect on ionic currents. In Shaker (WT) as soon as the temperature changes, we observe an increase in the ionic current which is more pronounced at depolarized potentials (~40% increase at +40 mV for 8 °C Tstep) (Fig. 1d). For the ILT mutant, we observed a more pronounced increase in the current (~100% increase at +180 mV for 5 °C Tstep) (Fig. 1e) while in the I384N mutant temperature induced a decrease in the current (~30% decrease at +150 mV for 7 °C Tstep) (Fig. 1f). When the time course of currents and temperature at different voltages were analyzed in the WT, we observed that at positive potentials (+40 mV) the current followed the time course of temperature, however at less depolarized voltages (<0 mV) it deviated and decreased after an initial increase (Fig. 1g). The rapid increase in the current upon application of the Tstep is consistent with an increase in the single-channel conductance (Δγ). In contrast, the decrease in current is related to a reduction of the Po, similar to previous observations in other channels¹⁶. For the ILT, we observed an increase in the current and Po (ΔP_o) at all the voltages tested (Fig. 1h). In contrast, for I384N, the ionic current initially increased due to the single-channel conductance and then decreased due to a change in Po for all voltages tested (Fig. 1i). To compare the effects of temperature between the WT and mutants, we calculated the temperature coefficient (Q₁₀) of the steady-state current after the Tstep (Eq. 3). The Q₁₀ for the WT saturates at 1.5 for voltages over 0 and decreases at voltages <0 mV due to the Po changes with temperature (Fig. 1j). The maximum Q₁₀ is close to the 1.44 value determined previously for the single-channel conductance⁷. The Q₁₀ of the current induced by the ILT mutant is significantly higher than that of the WT. For Tstep applied when the bath temperature was 5.6 °C, the Q₁₀ varies between 5–20 depending on the voltage. The results were more modest at higher bath temperatures (2–2.5 at 22 °C), likely due to the nonlinearity of Q₁₀ with temperature^21,22 (Fig. 1k). For I384N, when the bath temperature was 6 °C, the Q₁₀ varied between 0.8 and 1, whereas when the bath temperature was 17 °C, the Q₁₀ further decreased to 0.5–0.6 (Fig. 1l). These results demonstrate that mutations affecting distinct aspects of the VSD-PD coupling can drastically alter the channel’s temperature response and prompted us to examine whether these temperature-dependent effects are related to the VSD movement or the closure of the pore gate independently of VSD movement.

Fig. 1: Temperature dependence of ionic conduction in WT and mutant channels.

a Schematic representation of the experimental set-up used to apply fast-temperature steps to the oocyte membrane. A current-regulated homogenized laser light is used to illuminate the upper dome of the oocyte in the cut-open set-up, where the current recording occurs. The absorption of the visible light by melanin under the membrane of the oocyte generates a temperature change that is modulated by using the PWM of the laser light. b Structural location of the ILT mutations highlighting the location in the C-terminus of S4. c Structural location of the I384N mutation at the N-terminus of the S4-S5 linker, indicating its interaction with the S6 segment in the pore domain (PD), interacting pore subunit shown in blue, PDB: 7SIP¹⁵. d–f Current response to temperature steps application (red arrow) during a voltage pulse protocol (inset) for d Wild-type (WT) Shaker IR e ILT, and f I384N channels. The Tstep profile is shown on an expanded scale below the current traces. Scales for the current and temperature are shown for each trace family. The bath temperature is shown in blue, and the temperature reached during the Tstep is shown in red. g–i Comparison of the time course of normalized currents (black, I/I₀) and temperature (red) at different voltages for (g) WT, (h) ILT and (i) I384N channels. ∆γ and ∆P₀ indicate the change in single-channel conductance and the change in the open probability, respectively. j–l Plot of the Q₁₀ of ionic currents for the j WT, k ILT, and l I384N channels, colors indicate different initial bath temperatures. For (j), (k), and (l), N of at least 4 cells was measured for each bath temperature. (T: No Tstep, Tstep) indicates the value of bath temperature before the temperature step (No Tstep) and temperature reached during the temperature step (Tstep). Data are shown as mean ± SEM.

Temperature dependence of the voltage sensor

The effect of temperature on the VSD movement was determined by measuring the change in gating current promoted by a Tstep and, hence, the gating charge movement. To characterize the behavior of the voltage sensor, we used the V478W mutant, which is known to render a non-conducting channel²³. We favor the use of this mutation over other mutations to measure gating currents, such as W434F, because of the absence of residual ionic current in this mutant^24,25 and given that temperature changes induced ionic currents through the W434F pore (Supp Fig. 1). Similarly to experiments performed with ionic currents, we applied a Tstep in the middle of a voltage step after the gating currents had subsided and subtracted the linear optocapacitive component. We observe the development of inward gating currents from −120 to −30 mV in response to the Tstep (Fig. 2a, b). We can interpret the results on the gating currents as a change in the equilibrium distribution of the VSD through different conformational states with temperature. To verify this interpretation, we calculated the difference in the charge (ΔQ) displaced with and without Tstep at different voltages. The charge difference between the temperature of 16 °C and 22 °C shows an inverted bell-shape curve with a minimum of around −50 mV (Fig. 2c). This charge movement provides the underlying mechanism to explain the closure of the pore of the WT channels observed in the ionic currents (Fig. 1d, g). Since this only happens at specific voltage ranges, we only see the effect on the ionic currents at voltages where temperature displaces the VSD. Our analysis shows that Tstep drives the VSD into the resting state in a time-dependent manner. Thus, we can predict that applying a Tstep at different times before a voltage step would produce a time-dependent delay in the activation because deeper closed states become populated. This mechanism would manifest in the classical Cole-Moore shift, which describes the delay in the opening due to the transitions of the VSD along closed states before the pore can open²⁶. To test this, we applied temperature steps at different times before the onset of a depolarizing voltage pulse (Fig. 2d). The results reveal two distinct temperature effects. First, the kinetics of channel opening are faster in the presence of Tstep, reflecting the temperature dependence of transition rates between states. Second, and more importantly, the earlier the Tstep is applied before the test voltage step, the longer the delay in the activation onset. (Fig. 2e, f). This delay occurs because temperature drives the voltage sensors toward earlier closed states in the activation pathway, requiring channels to traverse more transitions before opening (Cole-Moore shift). As expected, the delay in the Cole-Moore observed in the ionic currents follows the time course of the charge movement with Tstep (Fig. 2e, inset). These findings demonstrate that temperature redistributes channels among closed states and accelerates the transitions between states. These experiments, which can only be observed by precisely controlling temperature and voltage, support our observation that the movement of the VSD by temperature produces the closure of the channel. Our gating current measurements reveal that temperature directly affects voltage sensor movement, causing a redistribution of voltage sensor states and affecting the activation’s voltage dependence. This redistribution can explain the closure of WT channels observed.

Fig. 2: VSD displacement in response to temperature produces a time-dependent delay in activation.

a Gating current measurements during a voltage and Tstep protocol. Tsteps were applied in the middle of a voltage step after the gating currents had subsided (indicated by the red arrow). The Tstep profile (red) is shown in an expanded scale above the current traces and voltage protocol. b Detail of temperature-dependent charge movement at different voltages. c Temperature-dependent charge movement vs. voltage (Q–V) curves for 4 different cells. d Ionic currents in response to a voltage step protocol, shown without a temperature step (No Tstep, gray) and with Tstep applied at 100 ms (yellow), 25 ms (green), and 2 ms (red) before the voltage pulse, arrows indicate the corresponding current traces. The inset depicts the voltage protocol; arrows in the voltage protocol indicate the time at which the Tstep is applied. The expanded Tstep trace is presented above the voltage protocol. e Delay in ionic current onset as a function of the time between Tstep and voltage step. Inset shows the time course of the charge movement calculated by the integral of the gating current in response to a similar Tstep in V478W at −60 mV (QTstep). f Relationship between delay in ionic current onset and voltage for different Tstep durations applied before the test 0 mV voltage step. Black squares indicate the delay in current onset in pulses without Tstep (No Tstep) (N = 4 independent measurements, Temperatures = 12 °C, before Tstep to 22 °C, after Tstep). Data are shown as mean ± SEM.

VSD displacement in response to a temperature step of ILT and I384N differs from the WT

Our data show that temperature affects the movement of the gating charges and pore closure. Shaker has its VSD strictly coupled to the PD. To understand the effects of temperature on the movement of the voltage sensor and the pore, we used mutations that stabilize the VSD at different positions. The ILT mutation separates the movement of the VSD, and the pore can only open when the last charge (R371) in the VSD passes the F290 residue in S2, which controls the energy barrier of the last gating transition^27,28. Therefore, using ILT and Tstep, we could probe the last transition in the VSD right before the pore opening. Temperature promoted two distinct effects on the ILT gating currents. We observed an inward current from −100 mV to −50 mV and an outward current at voltages higher than 40 mV (Fig. 3a, b). The displaced current shows a minimum at around −80 mV and a maximum at around +80 mV (Fig. 3e, g). This indicates that in the initial movement of the VSD, increasing temperature promotes a relative stabilization of the resting state, as it is in the WT channel. However, at more positive voltages (>40 mV), temperature favors the voltage sensor activation, which will favor the opening of the pore, producing a significant increase in the ionic current.

Fig. 3: Temperature-dependent VSD displacement in ILT and I384N.

Gating current measurements during a voltage and Tstep protocol for a ILT-V478W. Tstep were applied in the middle of a voltage step after gating currents had subsided (indicated by the red arrow) to analyze the effect of temperature on gating charge movement. The Tstep profile (red) is shown in an expanded scale above the current traces. b Detail of temperature-dependent charge movement on a at different voltages. c Gating current measurement during a voltage and Tstep protocol for a I384N-W434F. d Detail of temperature-dependent charge movement on c at different voltages. e, f Temperature-dependent charge movement vs. voltage curves for e ILT-V478W (n = 6 cells) and f I384N-W434F (n = 5 cells). g Fraction of the total gating charge moved at different voltages for WT = from 15.6 ± 0.3 °C to 22.8 ± 0.7 °C, ILT = from 22.2 ± 0.3 °C to 30.5 ± 0.9 °C, and I384N = from 15.4 ± 0.2 °C to 22.8 ± 1.1 °C (number of cells N for WT = 4, ILT = 6, I384N = 5). Data are shown as mean ± SEM.

For I384N, due to the lack of expression when we used the mutation V478W, we used the W434F mutation to record the movement of gating charges. The W434F mutant presents a small conductance at voltages larger than −30 mV, a voltage range where the pore starts to open; this prevented further analysis using only this mutation on the channel. However, we still observed the inward current in response to a temperature jump indicative of VSD movement into the resting state, like V478W (Suppl. Fig 1). We replaced the internal K⁺ from the oocytes, which allowed us to observe only gating currents in the I384N-W434F mutant (see “Methods”). The Tstep in this mutant induced an inward current at voltages of −110 mV to −30 mV (Fig. 3c, d). The displaced current shows a minimum of around −70 mV (Fig. 3f, g). At voltages larger than -30 mV, no charge displacement is observed, indicating that the decrease in the observed ionic currents is not associated with VSD movement. We can interpret the results on the gating currents as a change in the equilibrium distribution of different states of the VSD with temperature. When ΔQ is compared to the total gating charge, we find that Tstep can move up to 10 % of the total charge for WT, ILT and I384N (Fig. 3g). The distinct temperature-induced charge movement in ILT and I384N reveals how these mutations alter the energetics of voltage sensor. In I384N, like in the WT, temperature promotes the deactivation of the VSD, whereas in ILT, temperature promotes both the deactivation of early voltage sensor transitions and the activation of late transitions.

A thermodynamic mechanism for temperature dependence

To interpret the results shown in Figs. 1–3 and to obtain thermodynamic parameters, we calculated the enthalpic (ΔH) and entropic (ΔS) components for both the conductance vs. voltage (G–V) and charge vs. voltage (Q–V) curves, using Tstep protocols that reproduce bath temperature changes (Suppl. Figs. 2–4). The G–V curves of the WT, ILT and I384N shift approximately 0.9, −1.7, and 2.4 mV/°C, respectively (Fig. 4a–c and Table 1). Using the G–V measurements, we calculated the free energy of the opening at different temperatures and obtained the ΔH and ΔS components (Eq. 8). For WT and I384N, ΔS is negative, while for ILT, it is positive, as expected from their respective shifts observed in the G–V curves (Fig. 4d and Table 1).

Fig. 4: Pore opening and VSD movement thermodynamics in WT, ILT, and I384N channels.

Normalized G–V plots for WT (a), ILT (b), and I384N (c) channels under various temperatures indicated by different colors. The continuous line corresponds to a two-state model fitting. d Free energy (zFV—see “Methods”) vs. temperature for the G–V relations. The WT (gray) and I384N (blue) constructs show a negative ΔS (-slope), while ILT (red) exhibits a positive ΔS. e Q–V relationship for the V478W channel. f Free energy vs. temperature plot for the V478W Q–V. g Q–V relationship for the ILT-V478W channel. h Free energy vs. temperature plot for the ILT-V478W Q–V. i Q–V relationship for the I384N-W434F mutant. j Free energy vs. temperature plot for the I384N-W434F mutant. The free energy vs. temperature plot shows the energy for the first (ΔG₁, red) and second (ΔG₂, blue) component of the Q–V, the sum for the tetramer (4(ΔG₁ + ΔG₂), green) and using the V-median (gray). I384N only has one distinct component (ΔG₁, red). The continuous lines in the Q–V curves are illustrative to help show the shifts. Parameters used for fitting are found in Supplementary Table 1, Tables 1 and 2. For (a–c), (e), (g) and (i) N = 4−9 cells measured for each temperature, in E 3 cells measured with a bath temperature of 10 °C with different Tstep magnitudes. Data are shown as mean ± SEM.

Table 1 Thermodynamic parameters from G–V curves

The Q–V relationship for the WT (V478W) channel shows a displacement to the right with temperature (Fig. 4e). The difference between the Q–V curves for two temperatures is equivalent to the integral of the charge moved using the Tstep (Suppl. Fig. 5). To analyze the thermodynamics of the WT Q–V curves, we used two complementary approaches: (i) the Q–V was fitted using a three-state model where each transition has its own voltage dependence (Eq. 7); and (ii) using the V-median, a model-independent procedure for calculating energies²⁹. We observe a good agreement in the WT between both approaches when we account for the contribution of all 4 voltage sensors (Fig. 4f and Table 2). Unlike the V-median, the three-state model allows us to distinguish the temperature dependence of each of the two main VSD movement transitions. With the fit to the three-state model, we find that the first component has the largest shift with temperature. Using the V-median fit, we found a shift of about 1.1 mV per °C that closely matches the G–V shift. This is expected due to the strict coupling between the VSD and PD in the WT channel, implying that both steps of gating charge movement contribute to the shift of the G–V in the WT channel.

Table 2 Thermodynamic parameters from Q–V curves

The Q–V components are clearly distinct in the ILT case (ILT-V478W) (Fig. 4g). We observe that the first component of the Q–V (80–90% of total charge) has a right shift of 0.9 mV/°C, while the second component (10–20% of total charge) has a left shift of −2.6 mV/°C. Quantifying the thermodynamics of ILT VSD movement shows a negative ΔS for the first transition, compared to a positive ΔS for the second transition (Fig. 4h and Table 2). Using the V-median, we determine an overall shift of 0.46 m V/°C. This discrepancy can be explained by the fact that the V-median will be most affected by the first component of the Q–V, which carries most of the charge. Thus, in this case, it provides less thermodynamic information about the energetics of the gating process. Our results for the WT and ILT indicate that the temperature dependence of the G–V arises from the temperature dependence of the VSD movement, and in the case of ILT, it is dependent only on the second component. This is expected due to the separation of the Q–V components in ILT and the strict coupling between VSD activation and the opening of the channel, which has important implications for our understanding of channel function and temperature sensitivity.

A different picture emerges for I384N (I384N-W434F), where the Q–V shows only one component with a 0.31 mV/°C shift (Fig. 4i and Table 2). The V-median, in this case, gives 0.43 mV/°C shift. The thermodynamic parameters for I384N Q–V show a reduced ΔS compared to the WT (Fig. 4j and Table 2). The difference in the thermodynamic parameters between a two-state model fitting and the V-median procedure likely arises due to the underestimation of the energetics when fitting a multistep process with a simple two-step model³⁰. In this case, the V-median is expected to represent a better estimation of the thermodynamic parameters of the VSD movement. As indicated, the Q–V develops between −120 and −20 mV, voltages at which there is little or no opening of the channel (pore starts opening usually at −30 mV for I384N), demonstrating that unlike in the WT channel and ILT, in this mutant, the shift of the G–V does not reflect the shift of the Q–V with temperature. Since this channel is not strictly VSD-PD coupled, this indicates that we are observing the temperature dependence of the pore itself.

Our thermodynamic analysis of the G–V and Q–V relationship reveals thus how specific transitions in the activation pathway contribute to temperature sensitivity. The strict coupling in WT channels results in minimal temperature effects, while ILT’s separated voltage sensor transitions and I384N reduced coupling produce distinct temperature-dependent behaviors. In the ILT, the voltage sensor still drives the pore to open, while in the I384N, the voltage sensor influences the opening, but it is not efficient in transferring the movements of VSD into the pore opening.

Discussion

Here, we used a newly developed technique to probe the temperature effects on the gating mechanisms of a Kv channel and two temperature-sensitive mutants. We have observed two significant temperature effects in the Shaker channel: (i) the closure of the pore due to the inward movement of the gating charges by a Tstep; (ii) mutations in the S4 and S4-S5 linker modulate the temperature dependence of the channel allowing us to determine the thermodynamics properties of specific transitions in the activation energy landscape. The effects of Tstep observed in the WT channel are in line with previous reports⁷. The Tstep technique has the advantage over bath temperature changes in that it allows us to apply a fast-temperature change at any given time during a voltage protocol, making it possible to reveal the characteristics of the energy landscape during channel gating. An example of this is the unprecedented observation of the movement of the gating charges by a temperature pulse, which led to the prediction of the temperature and time dependence of the Cole-Moore shift. This example demonstrates the potential of Tstep technique to dissect the energy landscape of ion channel gating mechanisms.

Where does the temperature dependence of I384N and ILT come from?

From the previous kinetic analysis of the WT channel, we know that the last transition before channel opening has a shallow voltage dependence and large negative ΔH and ΔS components (−30 kcal/mol and −100 cal/K mol, respectively)⁸. This significant and negative enthalpic and entropic change would lead to a large rightward shift in the G–V curve with temperature, as observed in the I384N mutant. Since the I384N mutant effectively uncouples the movement of the VSD from the PD, this mutation reveals the temperature dependence of the pore itself. Our data further shows that the pore opening is intrinsically affected by temperature. The WT channel does not change its G–V like I384N because of the strict coupling between VSD and PD. In this scenario, the individual energetic contribution of each step is combined to produce the temperature behavior of the G–V. Therefore, through detailed kinetics analysis of the WT channel or uncoupling the VSD-to-PD, we can observe this intrinsic property of the channel. In Shaker, the opening of the PD in one subunit is affected by the state of the PD gate in the adjacent subunit, giving rise to a highly cooperative opening of the PD. This highly cooperative opening requires the activation of all the VSD. Thus, the PD acts like a load on the VSD due to the strict coupling between VSD activation and PD opening. The uncoupling found in I384N suggests that the VSD in this mutant can move without the extra load of the PD, which would explain why we observe lower absolute values for ΔH and ΔS and why we only observe one component in the Q–V.

Our mechanistic interpretation of the observed behavior of the I384N mutant is that the S4-S5 linker has a loose conformation⁴, where movements of the VSD are not efficiently coupled to the S6 gate opening, and temperature promotes the closure of the S6 gate. Conversely, the ILT mutation had the opposite effect with temperature, increasing the amount of ionic current up to three-fold with a 5 °C Tstep. The ILT mutation introduces a voltage-dependent step that is also highly temperature-dependent and drives the concerted opening of the channel¹⁹. It is unclear what step in the gating transition is affected by ILT. Originally, it was proposed to be the concerted opening of the pore, but later work using tetrameric concatemers of the channels suggested that is the previous step, the last transition of the voltage sensor^18,19. The ILT shares a similar effect as the F290A mutation²⁸. Residue F290 determines the translocation of the last gating charge of the VSD (R371) necessary for the opening of the PD^27,28. The fact that mutations in F290 and ILT have additive effects in the G–V and the last component of the Q–V supports the notion that ILT also affects the translocation of the last charge rather than the concerted opening step. By measuring the gating and ionic currents at high depolarized potentials, we observed a left shift in the G–V curve consistent with the shift in the second component of the Q–V. These shifts arise from the positive ΔS observed for the last transition in ILT (Table 2, Eq. 8). We believe this positive entropic change occurs at least in part due to conformational entropy of the VSD. We envision that in the intermediate state of the VSD in ILT (before the last gating charge moves), the S4 segment charges are unlikely to undergo a rotational conformation by the constraints imposed by the hydrophobic plug^30,31. When the VSD moves from the intermediate to active conformation, the S4 charges are exposed to the solvent and thus have more conformational freedom, producing a net increase in the number of possible microstates and increasing entropy. This increase in the number of possible microstates is not observed on WT because the last gating charge movement is lumped with the rest of the gating process, and we cannot distinguish the movement of the individual gating charges in that situation. Therefore, in ILT, the net change in entropy of the VSD is the primary driver of the Q–V and G–V shifts.

However, this hypothesis does not fully explain the observed increase in the ILT ionic current when the temperature increases. A clear example of this effect is when a Tstep is applied during a pulse to 180 mV, a voltage where the Q–V and G–V are saturated, but we still observed a three-fold increase in the current with a 5 °C step (Fig. 1h). Since ILT single-channel conductance is the same as the WT¹⁶, it is unlikely that changes in the single-channel conductance can account for these effects. As mentioned above, it is possible that ILT also affects the last concerted step. At low temperatures, the S4-S5 linker in the ILT mutation also has a loose conformation where there is a fraction of the channels that do not open in response to the movement of the voltage sensor, and the S4-S5 linker is not able to propagate the electromechanical energy to the S6 gate. At high temperatures, the S4-S5 linker adopts a tighter conformation, propagating the motion of the VSD more efficiently to the S6 gate, opening a significant fraction of channels. This mechanism explains the increase in the Po and the increase in the rates of activation observed in the ILT mutant when Tstep is applied. The process of loose or tight coupling can be related to the interaction between the S4-S5 helix and the S6 C-terminus. Given that residues in the ILT mutant are also interacting with residues in S5 that are implicated in a VSD-PD noncanonical coupling (not mediated by interactions between the S4-S5 linker and the S6 C-terminus) may offer an alternative explanation for the loose or tight mechanism^31,32,33. However, the contribution of the noncanonical mechanism to the ILT phenotype remains untested.

To illustrate how the proposed mechanism reproduces the observed effects of temperature and the energy landscape of Shaker, we built a kinetic model using the thermodynamic insight we have obtained (Supplementary Fig. 6 and Supplementary Table 2). This model is based on the Zagotta, Hoshi, Aldrich model¹⁷, where the voltage sensor moves in two steps (resting to intermediate and intermediate to active), and every voltage sensor must activate before the channel can open (Fig. 5a). Unlike the original model, we add a transition that represents the opening of the channel due to the opening of the bundle crossing gate (transition C15 to O—Supplementary Fig. 6a). The transitions between these states are defined by the rates describing each mutant’s energy landscape (Supplementary Fig 6B–D and Supplementary Tables 2, 3). The voltage sensor transitions carry most of the charge movement, while the last opening transition has low voltage dependence. In the case of the WT, temperature mainly affects the resting to the intermediate transition of the VSD by favoring the resting state and the last transition favoring the closed state, consistent with our experimental results and previous findings⁸. The equilibrium of the last transition is displaced towards the closed states, but it does not produce a considerable change in the open probabilities by itself due to the strong coupling with the VSD. In response to a simulated Tstep, this model produces an inward current due to the VSD movement, which leads to the closure of the channel. At larger voltages where the Q–V saturates, no such effect is observed; only an increase in current due to the increase in single-channel conductance is observed (Fig. 5b). To reproduce the ILT phenotype, namely the split in the Q–V curves, we assume that the movement of most of the charge is carried in the resting to intermediate transition while a small fraction is accounted for in the intermediate to active transition (Supplementary Table 2). This accounts for the inferred effect of ILT in the translocation of the last gating charge. The ILT mutant exhibits a significantly larger difference in free energy in the intermediate to active transition with an enhanced temperature dependence compared to WT. Additionally, temperature favors the open state with temperature, reflecting the tightening of the coupling mechanism between the VSD and PD. In response to a Tstep, this model produces a biphasic effect on the gating currents and channel opening (Fig. 5c). On the other hand, in I384N, only one transition of the VSD was observed, so we included most of the ΔH and ΔS in the first transition of the VSD. In this case, the opening of the channel exhibits a shallow voltage dependence, and the ionic current movement is displaced compared to the VSD movement. The large negative ΔS would produce a large rightward shift in the voltage dependence with temperature, promoting closure. In response to a Tstep, this model produces an inward current due to the VSD movement, like the WT, and channel closing at all voltages tested (up to +150 mV) (Fig. 5d). Our model can provide a visual and thermodynamic representation which aids in understanding the experimental observations and proposed gating mechanisms for the WT and mutant channels (Figs. 1, 2, 3, and 4). These models reproduce the observed properties of the Q–V, G–V and temperature displacement of the gating currents for the WT, ILT and I384N (Fig. 5e–g). Overall, this modeling effort further demonstrates that changes in the energetics of VSD-PD coupling can affect the temperature dependence of a prototypical Kv.

Fig. 5: Simulating the effects of temperature steps on the gating transitions of Shaker WT, I384N, and ILT.

a Schematic representation of different conformational states of the channel, from left to right: Resting (VSD down, pore closed), Intermediate (VSD intermediate, pore closed), Active (VSD up, pore closed), and Open (VSD up and pore open). The color of the transmembrane helix denotes the movement of the S4 segment in the resting (purple), intermediate (yellow) and active state (red). Created in BioRender. Pinto, B. (2025) https://BioRender.com/p64k114. b, c, d are the ionic and gating current simulations when Tstep is applied in the middle of a voltage protocol. The model used and its equations are shown in Supplementary Fig. 6. The parameters used for simulation are shown in the Supplementary Tables 2 and 3. (e-g) Simulated Q–V (red), G–V (blue) and the fraction of the charge moved (gray) for the WT (e), ILT (f), and I384N (g) mutants at 290 (filled symbols) and 300 K (empty symbols). The voltage protocols used are the same as presented in Fig. 1.

Our mechanistic interpretations suggest that the S4-S5 linker can adopt either a loose or tight conformation, depending on the mutations and temperature conditions. In the loose conformation, the movement of the VSD is necessary but not sufficient to efficiently propagate the electromechanical energy to the S6 gate, leading to a reduced open probability. Conversely, in the tight conformation, the S4-S5 linker more effectively couples the VSD motion to the S6 gate, opening the PD. This functional switching is related to the coupling mechanism and can be tuned by interactions between the intracellular part of the S4, S4-S5 linker and the C-terminus of the S6 segment. Our results show that temperature dependence can arise by changes in the coupling between VSD and PD, likely by modifying the interactions between S4, S4-S5 linker and S6.

The fact that increasing temperature promoted more opening in ILT and closure of I384N makes it tempting to draw parallels with the thermosensitive channels (Thermo-TRP) and consider the ILT mutant channel like TRPV1 (heat receptor) and I384N like TRPM8 (cold receptor). However, the Shaker channel has a strict coupling of the voltage sensor with the pore opening, which contrasts with thermo-TRP channels, where the coupling of the voltage and temperature sensors to the pore is allosteric³⁴. This lack of strict coupling means that in TRP channels, unlike in Kv channels, the PD can open even if all the VSD are resting and that the temperature sensor promotes opening independently of the state of the VSD³⁵. This mechanistic distinction made by a strict or allosteric coupled voltage sensor to the pore is further underscored by the physical limitations given by the voltage dependence of G–V in Kv channels. Following Eq. 8, the large value of apparent z in the G–V of Kv channels constrains the shift in the V_1/2 with different temperatures compared with TRPs, which contains a small apparent z for their G–V³⁶. This can explain the larger shifts in G–V with temperature present in ILT and I384N, where the z is about 1–1.5, compared to the 2.5 from WT (Supplementary Table 1). However, it does not tell us the directionality. The opposite shifts with T in ILT and I384N are most likely given by the specific characteristic of the uncoupling promoted in each mutation.

Our study has provided molecular cues for transforming a Kv channel temperature dependence by introducing mutations that can change the magnitude of the VSD-pore coupling at different positions along the Shaker channel activation energy landscape. Importantly, it has allowed a dissection of the energy landscape of the voltage sensor movement and its coupling to the opening of the pore, which will enable us to gain insights into their molecular function.

Methods

Channels expression in Xenopus oocytes

Xenopus laevis ovaries were purchased from Xenopus 1 (Dexter, Michigan). The follicular membrane was digested by 2 mg/ml collagenase supplemented with bovine serum albumin (BSA) 1 mg/ml. Oocytes were kept at 12 or 18 °C in SOS solution containing (in mM) 96 NaCl, 2 KCl, 1 MgCl₂, 1.8 CaCl₂, 10 HEPES, pH 7.4 (NaOH) supplemented with gentamicin (50 µg/ml). After 6–24 h of harvesting, they were injected with 5–50 ng of cRNA diluted in 50 nl of RNAse-free water and incubated for 1–4 days before recording. We used clones from the Shaker zH4 K⁺ channel with removed N-type inactivation (IR, ∆6–46) in the pBSTA vector³⁷. Mutations were performed using Quick-change site-directed mutagenesis, and cRNA was transcribed from linearized cDNA, using mmessage mmachine T7 Transcription kit from Invitrogen (Waltham, Massachusetts). cDNAs were sequenced to attest to the correct sequence.

Electrophysiology

Ionic and gating currents were recorded from oocytes using the cut-open voltage-clamp method³⁸. Voltage-sensing pipettes were pulled using a horizontal puller (P-87 Model, Sutter Instruments, Novato, CA), and the resistance ranged between 0.2–0.5 MΩ. Data were filtered online at 20–50 kHz using a built-in low-pass four-pole Bessel filter in the voltage-clamp amplifier (CA-1B, Dagan Corporation, Minneapolis, MN, USA) sampled at 1 MHz, digitized at 16-bits and digitally filtered at Nyquist frequency (USB-1604; Measurement Computing, Norton, MA). The voltage command and the current elicited were filtered using the same frequency. An in-house software was used to acquire (GPatch64MC) and analyze (Analysis) the data. The chamber temperature was measured by a thermocouple and controlled through a negative feedback loop using a Peltier cooler. Transient capacitive currents were subtracted from the recorded currents by a dedicated circuit. We used an offline linear subtraction procedure to remove the optocapacitive and leak effects of Tstep using voltages at which there is no ionic current or gating charge movement¹⁶. For ionic current measurements, the external solution was composed of (mM): KOH 12, CaOH₂ 2, HEPES 10, EDTA 0.1, N-methyl-D-glucamine (NMDG) 108 and the internal solution of (mM): KOH 120, EGTA 2, HEPES 10. For gating current measurements, the external solution was composed of (mM): NMDG 120, CaOH₂ 2, HEPES 10 EDTA 0.1 and the internal solution of: NMDG 120, EGTA 2, HEPES 10. In all cases, all the solutions were adjusted to pH 7.4 with methanosulfonic acid. For I384N-W434F mutant gating currents measurements using Tstep, oocytes were incubated for 30–60 min in an internal solution prior to the experiment to dialyze the internal potassium. All chemicals used were purchased from Sigma-Aldrich (St. Louis, MO).

Voltage protocols were adapted for each variant to accommodate their gating characteristics and to minimize artifacts. For instance, the ILT mutant was held at 0 mV during recordings to prevent contamination of the ionic current recordings with gating currents. This holding potential also allowed us to clearly resolve the two principal components of gating charge movement. Slight voltage protocol adjustments were made for I384N. Previous studies have shown that depolarized holding potentials affect the voltage dependence of the WT Shaker channel by entry into a relaxed state, however these effects are not observed in ILT (Supplementary Fig. 7) or I384N^19,20,39. Thus, our voltage protocol adjustments preserve their gating properties, allowing us to compare results across different variants.

Tstep set-up

After absorbing photons, non-fluorescent materials dissipate the energy through non-radiative processes, including vibrational relaxation, where it is converted into thermal energy⁴⁰, which can deliver sudden temperature changes in biological preparations^41,42. In our Tstep set-up, the heating mechanism relies on light absorption by endogenous melanin in Xenopus oocytes. When the laser light is absorbed by melanin granules, which are in close proximity to the membrane⁴³, the absorbed energy is rapidly converted to heat, causing a localized temperature increase in the membrane region. By applying a pulse width modulated laser, we can obtain precise temporal and spatial control of membrane temperature. A 3.5 W 447 nm diode laser (Osram PLPT9 450D_E A01) was placed on top of the recording chamber and aligned with the oocyte dome. The beam was collimated using an aspheric lens (A230TM-A, Thorlabs Inc. Newton, New Jersey) followed by a pair of microlens arrays (MLA300-14AR-M, Thorlabs) to ensure homogenization of the beam⁴⁴, and focused into the oocyte dome using another aspheric lens (ACL2520U-A, Thorlabs). An in-house current modulated power supply was used to achieve rapid turn-on of the laser. The laser was triggered by an arbitrary wave generator (4075B, B&K Precision Corp., Yorba Linda, CA). Due to the different content of melanin in oocytes, the Tstep pulse protocols were not the same for all the experiments performed.

Capacitance-based temperature measurement (CTM)

We measured the temperature due to Tstep, as reported previously¹⁶. Briefly, we applied a sinusoidal voltage wave during the Tstep. We calculated the complex voltage and current signal after subtraction of the optocapacitive current using the Hilbert transform. We calculate the impedance by dividing the complex voltage and current. The impedance of the system is given by:

$$Z=\frac{\widetilde{V}}{\widetilde{I}}{=R}_{s}+\frac{{R}_{m}}{1+{(\omega {C}_{m}{R}_{m})}^{2}} - {j} \frac{{{R}_{m}}^{2}\omega {C}_{m}}{1+{(\omega {C}_{m}{R}_{m})}^{2}}$$

(1)

where Rs, Cm, Rm, ω and j are the series resistance, membrane capacitance, membrane resistance, angular frequency of the sinusoidal wave and the imaginary unit, respectively. When ${(\omega {C}_{m}{R}_{m})}^{2}$ > >1, we get:

$${Imag}\left(Z\right)\approx -\frac{{{R}_{m}}^{2}\omega {C}_{m}}{{\left(\omega {C}_{m}{R}_{m}\right)}^{2}}=-\frac{1}{\omega {C}_{m}}$$

(2)

Thus, we can obtain the membrane capacitance time course using the time course of the imaginary part of the impedance. To ensure that the assumption of Eq. 2 is valid, we worked at voltages where there is no conductance for the conductive channels or no gating movement for the non-conductive channels.

We used an average of 100 traces and filtered the results using a 10 KHz offline Bessel-like filter to obtain the capacitance time course. A linear fitting that converts capacitance to temperature was used to get the temperature¹⁶. The relationship between membrane capacitance and temperature remains linear within our experimental temperature range (5–30 °C) and with a slope of 1.07% per degree Celsius, allowing us to calculate the temperature change from capacitance changes.

Data analysis

The temperature-dependent (Q₁₀) factor for the ionic currents was obtained by:

$${Q}_{10}={\left(\frac{{I}_{{Tstep}}}{{I}_{{NoTstep}}}\right)}^{10/\Delta T}$$

(3)

where I_Tstep, I_NoTstep, and ΔT correspond to the current at the end of the Tstep, the current right before the Tstep is applied and the temperature difference of the Tstep, respectively.

The G–V curves were measured from the tail currents after a voltage protocol and fitted using a two-state model given by the equation:

$$G\left(V\right)=\frac{1}{1+\exp \left(-\frac{{zF}}{{RT}}\left(V-{V}_{1/2}\right)\right)}$$

(4)

where z is the apparent charge expressed in units of elementary charge (${e}_{0}$), V is the voltage and ${V}_{1/2}$ is the voltage of half-maximal conductance. R, T, and F have their usual meanings. For WT, the ionic current data was taken when the current reached a steady-state level and then converted to conductance using the following relationship:

$$G\left(V\right)=\frac{I}{V-{Vrev}}$$

(5)

where, I is the ionic current at steady-state, V is the membrane voltage, and Vrev is the Nernst potential for the conducting ion. The G–V was then fitted using Eq. 4.

The charge was obtained by integrating the off-gating currents, which were plotted against the voltage to obtain the Q–V curves. For the analysis of the Q–V curves, we used three different approaches:

1- A two-state model fitting equivalent to the one in Eq. 4 was used to fit the I384N-W434F mutant and the individual components of the ILT Q–V. We used a global fitting for the pairs Q–Vs obtained for different temperatures in the same cell, where we kept the same z for all temperatures.

$$Q\left(V\right)=\frac{1}{1+\exp \left(-\frac{{zF}}{{RT}}\left(V-{V}_{1/2}\right)\right)}$$

(6)

2- A three-state model fitting is given by the following equation⁴⁵:

$$Q\left(V\right)=N\frac{{z}_{2}+{z}_{1}\left(1+\left(\exp \frac{{z}_{2}F}{{RT}}\left({V}_{2}-V\right)\right)\right)}{1+\exp \left(\frac{{z}_{2}F}{{RT}}\left({V}_{2}-V\right)\right)\left(1+\exp \left(\frac{{z}_{1}F}{{RT}}\left({V}_{1}-V\right)\right)\right)}$$

(7)

where N, z₁, z₂, V₁, and V₂ are the number of channels, the charges associated and equilibrium voltages for the first and second transition respectively. For the fitting at different temperatures is assumed that N, z₁, and z₂ remain constant for all temperatures.

3- Using the V-median of the Q–V²⁹. For the V-median, the number of charges per channel (z) was considered the same for all channels: 13.6 elementary charges^14,46,47

To obtain the ΔS and ΔH of the G–V and Q–V curves, we used the following linear relationship⁴⁸:

$${z}_{i}F{V}_{i}=\Delta H-T\Delta S$$

(8)

where the z_i and V_i represent the relative charges and are obtained from the fitting of the G–V and Q–V (for different transitions) curves.

The statistical significance of the difference in linear fittings was calculated using the extra sum of squares F-test in GraphPad Prism 5. The F-test was employed to assess differences between fittings as it evaluates whether the slopes of the lines significantly differ by comparing the variance explained by group differences to the variance within groups.

To determine the delay related to the Cole-Moore shift, we first fitted a two-exponential function (Eq. 9) to the ionic currents.

$$f\left(t\right)={A}_{1}{e}^{(-\frac{t}{{\tau }_{1}})}+{A}_{2}{e}^{(-\frac{t}{{\tau }_{2}})}+c$$

(9)

Where ${A}_{1}$, ${A}_{2}$, ${\tau }_{1}$, ${\tau }_{2}$, and c are, respectively, the amplitude of the first component, the amplitude of the second component, time constant of first component, time constant of second component and a constant. After fitting this equation to the ionic currents, we extrapolated it to the point where the current was zero, giving the delay for the Cole-Moore shift. Supplementary Fig. 8 shows the procedure in detail.

Reporting summary

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