A hydrophobic element secures S4 voltage sensor in position in resting Shaker K+ channels


Corresponding author C.-C. Kuo: Department of Physiology, National Taiwan University College of Medicine, No. 1, Jen-Ai Road., 1st Section, Taipei, 100, Taiwan. Email: cckuo@ha.mc.ntu.edu.tw


The S4 transmembrane α-helix in voltage-gated channels contains several regularly spaced basic amino acid residues that could be protonated and moved across the membrane electric field in response to membrane potential changes. The translocation of the charge-carrying S4 transduces membrane voltage to gating conformational changes of the channel, but how it is positioned and moved with respect to membrane lipid remains controversial. We found that hydrophilic and especially arginine and lysine substitution for L361 at the external end of S4 causes a large negative shift with shallowed slope of both activation and inactivation curves in Shaker K+ channels. Also, the macroscopic kinetics of activation and inactivation become much faster and barely voltage dependent, especially in the L361R mutant channel. These steady-state and kinetic data suggest that the replacement of one single hydrophobic residue, leucine, with arginine may profoundly destabilize the resting conformation of S4, which therefore takes a partially extruded position (partly activated position) at resting potentials (e.g. −120 mV). Consistently, the L361R point mutation gives rise to an extracellularly exposed R365C that is readily modified by external hydrophilic sulfhydryl-specific agents in the resting channel. Moreover, the extruded S4 in the L361R mutant channel could be retracted by strong hyperpolarizing potentials (∼−180 mV), from which the mutant channel is gated with slower kinetics but evidently stronger voltage dependence. We conclude that hydrophobic interaction involving a highly conserved residue at the top of S4 is crucial for properly securing the gating voltage sensor in the resting position and thus appropriate gating control of the voltage-gated channels.

Voltage-gated ion channels are responsible for the generation and rapid propagation of electrical signals in nervous system and other excitable tissues. The voltage-sensing apparatus in the channel protein monitors changes in membrane potentials and enables an energetic coupling between membrane potential and channel activity. Among the large superfamily of voltage-gated channels, the voltage-gated K+ channel is prototypical in structure with four identical subunits assembled as a functional protein, and has been the most extensively studied. The fourth transmembrane segment (S4) in each subunit or domain of voltage-gated channels has been implicated as the major voltage sensor (Aggarwal & MacKinnon, 1996; Seoh et al. 1996; Bezanilla, 2000; Horn, 2000). S4 could be moved upon changes in membrane potential, maximally transferring three to four of its regularly spaced basic residues (chiefly arginines) across the membrane field (Larsson et al. 1996; Yang et al. 1996; Baker et al. 1998; Starace & Bezanilla, 2001). However, the molecular mechanism underlying S4 movement remains controversial.

Two different models were put forward for the molecular feature of S4 movement. The conventional proposal considered S4 as an α-helical transmembrane cylindrical peptide (Peled-Zehavi et al. 1996; Li-Smerin et al. 2000), in which the S4 charges are shielded from and thus stabilized in the lipid bilayer by the counter charges (the negatively charged amino acid residues) in the other part of the channel protein (Tiwari-Woodruff et al. 1997, 2000; Durell et al. 1998; Li-Smerin et al. 2000). S4 thus moves within a ‘gating canal’, an aqueous crevice lined by the hydrophilic amino acid residues (Starace & Bezanilla, 2004; Stanislav et al. 2005; Tombola et al. 2005). Since the membrane field may be just a thin ‘septum’ (no thicker than a few angstroms) in the canal, the actual physical movement of S4 is not necessarily large for the transfer of adequate gating charges across the membrane electric field (Goldstein, 1996; Larsson et al. 1996; Islas & Sigworth, 2001; Starace & Bezanilla, 2004). The limited transmembrane movement of S4 is further supported by more recent experimental results from different approaches (Chanda et al. 2005; Phillips et al. 2005; Posson et al. 2005). On the other hand, based on the crystallographic studies and analyses of bacterial voltage-gated K+ channel structures, Jiang et al. (2003a,b) proposed a ‘paddle model’ for voltage sensing. The paddle model argues that the S4 sensors are hydrophobic, cationic, and helix–turn–helix structures located on the channel's outer perimeter, and are stabilized in the lipid bilayer in a form of charged cargoes. The voltage sensor paddles are positioned near the intracellular surface in the lipid bilayer in the resting channel, and move a large distance (> 20 Å) across the cell membrane to the outside during channel activation. Except for the discrepancy in the moving scale of S4, the basic conceptual difference between the two models is the way how the charge-rich S4 can be stabilized in the membrane (field), presumably a highly hydrophobic structure.

It is interesting that, despite of the potential importance of hydrophobic interactions in the placement of S4 in the membrane, the role of the hydrophobic residues in S4 has been much less extensively studied than the charged (basic) residues. Li-Smerin et al. (2000) did an alanine scanning of the residues between the N-terminal of S1 and C-terminal of S4. They found that the ‘high-impact’ (|ΔΔG°| between the open and closed states ≥ 1 kcal mol−1) and ‘low impact’ (|ΔΔG°| < 1 kcal mol−1) residues on S4 roughly follows the pattern of a repeating triad, and thus proposed an α-helical secondary structure for different transmembrane segments S1–S4. There are other scattered studies involving point mutations in and around S4 (e.g. Larsson et al. 1996; Schönherr et al. 2002; Gandhi et al. 2003; Ahern & Horn, 2004). But these studies are focused on the topographic interactions between S4 and the other parts of the channel protein (e.g. the ‘gating canal’ or the pore domain) in different gating states or on the gating movement and charge-carrying capability of S4. The biophysical contribution and molecular mechanisms underlying the role of the S4 hydrophobic residues in K+ channel gating have remained largely uncharacterized. Here we report a very ‘high impact’ hydrophobic residue (L361), which is conserved at a location just external to the outermost basic residue of S4 in the voltage-gated channels from a wide range of species and plays a crucial role in securing S4 in position in the resting Shaker K+ channel. Hydrophilic substitutions for this residue lead to profound destabilization of S4 in its ‘normal’ resting position. S4 is even partly extruded in the resting L361R mutant channel, leading to drastically reduced voltage dependence of channel gating. Moreover, this displaced S4 could be ‘pulled back’ by strong hyperpolarizing potentials to a more retracted position, from which S4 is moved outward to drive a partially restored gating behaviour of the mutant channel upon subsequent membrane depolarization.


Molecular biology and expression of Shaker K+ channels

The Shaker GH4 K+ channel cDNA is essentially the same as Shaker H4 channels reported by Kamb et al. (1987) except for the addition of three ‘silent’ restriction enzyme sites and a Xenopusα-globin untranslated sequence. Point mutations done with cDNA template and the QuikChange mutagenesis kit (Stratagene, LA Jolla, CA, USA) in both the original (fast inactivation-preserved) WT channels and the N-terminus (6–46th residues) deleted WT channels (the fast inactivation-removed or IR channels). The mutations were verified by DNA sequencing and by biophysical examination of at least two independent clones. The full-length cRNA transcript was synthesized from the ShGH4 K+ channel cDNA containing the L361 mutation using the T7 mMESSAGE mMACHINE transcription kit (Ambion, Austin, TX, USA). Mature female Xenopus laevis frogs were maintained and handled under the supervision of National Taiwan University College of Medicine and College of Public Health Institutional Animal Care and Use Committee (IACUC). For oocyte isolation, the animals were anaesthetized in cold water containing 1.7 g l−1 of tricaine (ethyl 3-aminobenzoate methanesulphonic acid, Sigma), and the ovarian sacs were removed by sterile surgical procedures. The frogs were allowed to recover in a water tank at room temperature immediately after the surgery. The animals were killed following the final collection of oocytes by decapitation after anaesthetization. Isolated and defolliculated oocytes (stage V–VI) were injected with the cRNA transcript and maintained at 18°C for 1–7 days before electrophysiological studies.

Electrophysiological recordings and data analyses

Macroscopic K+ currents in oocytes were recorded by a standard two-microelectrode voltage-clamp method. The oocyte was put in a chamber continuously perfused with modified ND-96 solution with 4 mm K+ (mm: 96 NaCl, 4 KCl, 1 MgCl2, 0.3 CaCl2, 5 Hepes, pH 7.6) or with 50 mm K+ when indicated (mm: 50 NaCl, 50 KCl, 1 MgCl2, 0.3 CaCl2, 5 Hepes, pH 7.6). 2-aminoethyl-methanethiosulphonate (MTSEA) (Toronto Research Chemicals, North York, Ontario, Canada) was stocked at −70°C and dissolved in water to make 100 mm aqueous stock solution that was freshly prepared daily, stored at −20°C, and diluted immediately before use. Both voltage-sensing and current-passing electrodes were filled with 3 m KCl and had serial resistance of 0.1–0.4 MΩ. Membrane potential was controlled by a two-electrode voltage-clamp amplifier with a virtual ground circuit (model OC-725C, Warner Instrument, Hamden, CT, USA). Data were recorded at room temperature of ∼25°C and digitized at 20–100 μs interval using a Digidata-1200 analog/digital interface and the pCLAMP software (Axon Instruments, Union City, CA, USA). All statistics are given as mean ± standard error of mean. To plot the normalized conductance–voltage curves (activation curves), the oocyte expressing the fast inactivation-removed (IR) variant of the wild-type or the L361F, E, Q, K, or R mutant Shaker K+ channel was subjected to a series of test pulses increased by a 10 mV step to elicit K+ currents. The holding potential was −120 mV for the wild-type and the L361F, E, and Q mutant channels. For the L361K and L361R mutant channels that are significantly activated at very negative potentials, the holding potential was set at −140 mV and the external K+ concentration was increased to 50 mm to avoid inaccurate measurement of the currents around the reversal potentials. The increase of the current amplitude in each 10 mV voltage step is normalized to the maximal increase to give the normalized conductance, which is then plotted against the test-pulse voltage to obtain the activation curve. We found no change in the activation curves of the wild-type as well as the L361F, E, and Q mutant channels with a holding potential of −140 mV and 50 mm external K+ (data not shown). For measurement of the inactivation curves, the fast inactivation-preserved clones of wild-type and mutant channels were expressed. The current was measured at +10 mV after a 100 ms prepulse at different voltages from a holding potential of −120 mV or −140 mV. At +10 mV, the elicited ‘net’ current is defined as the difference between the sustained and the peak current. The net current amplitude is then normalized to the maximal net amplitude and plotted against the prepulse voltage to obtain the inactivation curve. In the L361R mutant channel, the elicited current after a relatively depolarized prepulse (e.g. −110 mV) could show a very fast activation speed which makes it difficult to identify the current peak. In this case, the current peak is arbitrarily defined by the current amplitude at the same time point as that of a discernible current peak that is evoked with a slightly more negative prepulse.


The Shaker K+ channel with L361R or L361K point mutation is activated at extremely negative membrane potentials

We first made positive-charge (arginine and lysine) substitutions for L361, a hydrophobic residue right next and external to the outmost basic residue (R362) of S4 in Shaker K+ channels. Figure 1 shows representative currents through the wild-type and the L361 mutant channels. In sharp contrast to the wild-type channel, both of the L361R and L361K mutant channels are activated (and inactivated) at much more hyperpolarized membrane potentials. Moreover, the macroscopic kinetics of activation and inactivation at equivalent test-pulse potentials are evidently faster with less voltage dependence in the mutant channels than in the wild-type channels.

Figure 1.

Representative currents of the wild-type and the L361R and L361K mutant channels
The currents through both the fast inactivation-preserved (left column) and the fast inactivation-removed (IR, right column) wild-type and mutant channels are elicited by a series of consecutive step depolarization increased by 10 mV from a holding potential of −120 mV. One of the test-pulse voltages is indicated beside the elicited current sweep to show the very different voltage dependence of gating among these channels. The time scale shown in the top left panel applies to all panels in this figure. The time to current peak as well as the current decay time constant is evidently shorter and less voltage dependent in the L361R and L361K mutant channels than in the wild-type channels. Note that the time to current peak is almost unchanged with different test pulses in the mutant channels (left column). The IR variant of the wild-type channel shows clearly a delay of macroscopic activation especially at more negative test pulses, whereas the activation kinetics of the IR-mutant channels are too fast to have a discernible delay.

The steady-state activation and inactivation curves are shallower and profoundly negatively shifted with arginine or lysine substitution for L361

Consistent with the observation in Fig. 1, the steady-state activation curves show a very large negative shift on the voltage axis in the L361R and L361K mutant channels (Fig. 2A). In addition, the curves in the mutant channels have significantly shallower slope. There is little difference between the L361R and L361K mutants in terms of either midpoint (Vhalf) or slope of the activation curves (see a summary of the fitting results in Fig. 2C and D). According to a simplified gating model, the negative shift of the activation curve could result from an increased forward rate and/or a decreased backward rate of activation, i.e. destabilization of the resting state (presumably destabilization of S4 in its resting position in this case) and/or stabilization of the activated state (stabilization of S4 in its activated position) of the channel. It is more likely that these L361 mutations significantly destabilize the resting conformation (e.g. the resting position of S4) of the channel because of the much faster activation kinetics (see Fig. 1 and below). Because channel activation and fast inactivation are coupled yet distinct gating processes, the finding that the changes in the inactivation curves well parallel those in the activation curves in both the L361R and L361K mutant channels also supports that the mutations chiefly change the resting conformation of the channel rather than happen to stabilize different gates to the same degree (Fig. 2BD and also see below). It is thus plausible that the S4 voltage sensors are ‘freed’ by the mutation and have an increased tendency to be moved externally to an intermediate position at a membrane potential as negative as −120 mV, where most of the channels remain closed. The increased tendency of S4 to stay in intermediate stops would also be consistent with the shallower slope of the activation (and inactivation) curve.

Figure 2.

The activation and inactivation curves of the wild-type and L361R and L361K mutant channels
The experimental protocols and analysis are described in Methods (n= 4–5). A, the averaged activation curves (normalized conductance–voltage curves). The lines are fits of a Boltzmann function 1/(1 + exp((VhalfV)/k)), where V is the test-pulse voltage in mV. B, the averaged inactivation curves. The lines are fits of a Boltzmann function of 1/(1 + exp((VVhalf)/k)), where V is the prepulse voltage in mV. The apparent fractions of available channels are ‘paradoxically’ decreased at prepulses more negative than ∼−120 mV in the L361R mutant channels and are indicative of slowed activation rates ascribable to different S4 positions at those negative potentials (see text and Fig. 5 for detail). For simplicity, we omit the paradoxical phase here and fit the other points with a Boltzmann function. C, the Vhalf values of the Boltzmann fitting curves in A and B. D, the k (slope factor) values of the Boltzmann fitting curves in A and B.

The activation and inactivation curves are similarly altered with different hydrophilic substitutions for L361

To further examine the effects of replacement of L361, we documented the activation and inactivation curves in the L361E, L361Q and L361F mutant channels (replacement of L361 with an acidic, a neutral hydrophilic and a neutral hydrophobic residue, respectively). The substituting moieties with comparable size to leucine were chosen to minimize the steric factor. Phenylalanine is also an alternative hydrophobic residue at the equivalent position in some of voltage-gated ion channels such as the human Kv2.1 K+ channel, the bacterial KvAP K+ channel, and the domain IV of most cloned Na+ channels. Similar to those found in the L361R and L361K mutations, both activation and inactivation curves of the L361E and the L361Q mutant channels are negatively shifted on the voltage axis and have shallower slope compared to those of the wild-type channel, although the extent of the Vhalf shift is smaller than that in the L361R and L361K mutants (Fig. 3). On the other hand, the L361F mutant channel shows very similar activation and inactivation curves to those of the wild-type channel. These findings are consistent with the idea that the L361 residue plays a crucial role in stabilization of S4 in the resting position via hydrophobic interaction with its surroundings. Moreover, the activation curve and the inactivation curve always have very similar Vhalf changes in all of the tested L361 mutant channels (Figs 2C and 3C), supporting the view that L361 mutations chiefly affect a process common to both activation and inactivation (e.g. S4 movement).

Figure 3.

The activation and inactivation curves of the wild-type and the L361F, E and Q mutant channels
The experimental protocols and analyses are described in Methods and are similar to those in Fig. 2 (n= 4–5). The data of the wild-type channel are from Fig. 2. A, the averaged activation curves (normalized conductance–voltage curves). The lines are fits of a Boltzmann function 1/(1 + exp((VhalfV)/k)), where V is the test-pulse voltage in mV. B, the averaged inactivation curves. The lines are fits of a Boltzmann function 1/(1 + exp((VVhalf)/k)), where V is the prepulse voltage in mV. C, the Vhalf values of the Boltzmann fitting curves in A and B. D, the k-values of the Boltzmann fitting curves in A and B.

The L361R and L361K mutant channels are activated and inactivated with much faster kinetics but less voltage dependence from a holding potential of −120 mV

Figure 4A and B examines the macroscopic activation and inactivation kinetics and their voltage dependence in the wild-type and mutant channels with a holding potential at −120 mV. The activation kinetics in the L361R and L361K mutant channels are so fast that an accurate measurement of the activation time constant is not feasible, and thus the 90% current rising time is measured instead. In the wild-type channel, the kinetics of both activation and inactivation show evident voltage dependence when examined at more negative test pulses, and reach an asymptotical or saturating value with little voltage dependence at more positive voltages. The saturation can be viewed as if the voltage-dependent S4 movement is made so fast by the strong positive test pulses that the rate-limiting steps of the sequential activation and inactivation reactions shift from the early voltage-dependent steps (e.g. S4 movement) to the late voltage-independent steps (e.g. gate movement itself). The L361F mutation causes little changes in both activation and inactivation kinetics, indicating that this mutant channel behaves exactly like the wild-type channel in terms of S4 movement. In the L361K and especially the L361R mutant channels, however, the macroscopic activation and inactivation kinetics are very fast but show much less voltage dependence. Nevertheless, the gating kinetics in all of these mutant channels reach the same or very similar asymptotical or saturating value as the wild-type channel at more positive test voltages (Fig. 4A and B). These kinetics data are consistent with the dramatic Vhalf and slope changes found in the steady-state activation and inactivation curves (Fig. 2), suggesting that the S4 voltage sensor is seriously destabilized and has a strong tendency to leave its resting position in the L361R and L361K mutant channels. The voltage-independent step (e.g. the gate movement itself), on the other hand, is not affected by the mutations. To have a more quantitative view of these kinetics changes, we further analysed the voltage dependence of macroscopic inactivation time constant (instead of the activation kinetics, which are too fast to be accurately fitted for time constants) to estimate the equivalent gating charges (measured at the most negative test-pulse potentials where the voltage-dependent transitions would be the most rate-limiting; lower panel in Fig. 4B). The obtained equivalent charges are much smaller in the L361K and especially in the L361R mutants than in the wild-type channel.

Figure 4.

Macroscopic activation and inactivation kinetics at different test-pulse voltages with a holding potential of −120 mV
A, macroscopic activation kinetics. The oocytes (n= 4–6) expressing the IR variants of different mutant channels are stepped to the indicated test-pulse voltages from a holding potential of −120 mV to elicit K+ currents. Because the activation kinetics in the positive-charge mutant channels are so fast that a reliable exponential fit of the activation phase is difficult (e.g. see representative currents in Fig. 1), we instead measured the 90% rising time of the macroscopic currents (the time from the depolarization step to the generation of 90% maximal current). The L361R and L361K mutant channels are activated with much faster but less voltage-dependent kinetics. B, macroscopic inactivation kinetics. The experiments were performed in the inactivation-preserved channels with the same protocol as that of A. The monoexponential time constant (τ) of macroscopic current decay (n= 3–6) is plotted against the test pulse voltage (upper panel). The lower panel is a replot of the logarithm of the inverse of inactivation τ against test pulse voltages. The data at the three most negative test pulses are fitted by a regression line (dotted line) with a slope of zF/RT, where z denotes the apparent equivalent gating charges involved in the inactivation process, and is 1.6, 2.2, 0.70 and 0.14 for the wild-type, L361F, L361K and L361R channels, respectively. The data at more negative test-pulse voltage ranges are selected for the fitting procedures to avoid contamination of the voltage-independent processes.

The gating kinetics get slower with steeper voltage dependence when more negative holding potentials applied to the L361R mutant channel

Because L361R and L361K mutations drastically destabilize S4 in its ‘normal’ resting position, the accelerated gating rates with decreased overall voltage dependence may simply result from an acceleration of the original rate-limiting step (and thus a shift of the rate-limiting step) of S4 movement. On the other hand, the destabilized S4 may already take a partly extruded position in the resting channel, contributing to the profound decrease in voltage dependence of gating kinetics. In the latter case, application of a strong negative holding potential may ‘pull’ S4 back to the normal retracted position and restore part of the voltage dependence of channel gating at subsequent depolarizing pulses, whereas a more negative holding potential should have no effect on the gating behaviour if S4 is not pre-extruded. Figure 5 examines the inactivation curve of the L361R mutant channel over a wider range of prepulses. The apparent fraction of channel availability seems ‘paradoxically’ decreased with prepulses more negative than ∼−120 mV. This paradoxical decrease of peak current amplitude well parallels the prepulse-dependent lengthening of the activation time (time to 90% peak current, inset of Fig. 5), and both reach an asymptotical value at the same prepulse voltage range (i.e. ∼−160 to −180 mV). The decreased macroscopic peak currents thus most likely does not reflect a true change in channel availability but a result of slowed channel activation caused by strong negative prepulses. Moreover, the macroscopic inactivation in the L361R (but not in the wild-type and, surprisingly, also not in the L361K mutant channels) also shows evidently slower kinetics and increased voltage dependence with more hyperpolarizing preceding potentials (Fig. 6AC). Figure 6D summarizes the voltage dependence (in the form of equivalent charge z) of the inactivation kinetics with different holding voltages in the wild-type, L361K, and L361R channels. With a holding potential at −180 mV, the inactivation kinetics show evident test-pulse voltage dependence in all three channels. However, when the holding potential goes more positive, the test-pulse voltage dependence (the equivalent charge z) decreases from a saturating value of ∼0.4 (at ∼−180 mV) to ∼0.1 (at −120 mV) in the L361R channel. By contrast, the L361K mutant channel has a rather constant z-value of ∼0.5 (roughly similar to the saturating value of ∼0.4 in the L361R channel), and the wild-type channel shows a constant and significantly larger z-value of ∼1.2. These data indicate that S4 still stays in a most retracted (or ‘normal’ resting) position over a wide range of holding potentials when residue 361 is a leucine or lysine (though S4 may have been greatly destabilized at the position in the resting L361K mutant channel), and thus the inactivation kinetics would remain unchanged with different holding potentials. With the L361R mutation, however, the S4 is not only unstable in this most retracted resting position, but also is moved to take a partially extruded position at −120 mV and would require a much stronger negative membrane potential to pull it back to the retracted position. The saturation of the holding-potential effects implies that the backward S4 movement with a L361R mutation is not unlimited but has a final destination. The midpoint of the Boltzmann fit to the L361R data in Fig. 6D indicates that half occupancy of this intermediate partially extruded position of S4 is achieved at a membrane potential of ∼−135 mV. The k value (∼5.6) of the Boltzmann fit, on the other hand, implies a total charge transfer of at least ∼4 (≥ 1 charge per subunit) between the most retracted and the partially extruded positions of S4 in the L361R mutant channel. (A Boltzmann fit in this case may not be very inappropriate considering the likelihood that there is no intervening stops between the most retracted and the intermediate positions of S4 in the L361R mutant channel. However, the k value from the fit could still be a low estimate of the total charge transfer because of different combinations of the four S4s distributed between the most retracted and the intermediate positions, i.e. not a true two-state system.) The increase of z values from ∼0.1 to ∼0.4 along with the slower inactivation kinetics with more negative holding potentials suggests a change of the rate-limiting ratchet step of S4 movement in the L361R mutant channel. The early outward S4 movement from the most-retracted position to the intermediate (partially extruded) position seems to be much slower and more voltage dependent compared to the late transitions of S4 during the L361R channel activation. L361 thus probably stays in a strictly hydrophobic environment with the most retracted S4, and then travels through a relatively confined pathway in its early than late steps of outward movement. The larger z values of the L361K mutant and especially the wild-type channels, on the other hand, may imply that the S4 movement here is less ratchet-like than the L361R channels from a functional point of view (see Discussion).

Figure 5.

Paradoxical changes in L361R channel availability with very negative prepulse potentials
The oocyte was subjected to a series of 100 ms prepulses from −180 to −70 mV and then stepped to +10 mV to elicit macroscopic K+ currents. The interpulse interval is 3 s, at −120 mV. Some of the prepulse voltages are indicated beside the sample currents (upper right panel). The averaged inactivation curves (n= 4) are obtained as those described in Methods and Fig. 2B. The dotted line is the fit of a Boltzmann function with Vhalf values of −104.7 mV and k values of 5.7. The inset shows the prepulse potential-dependent changes in activation kinetics. The 90% rising time is measured as that described in Fig. 4A and plotted against the prepulse voltage. There is close correlation between apparent channel availability and macroscopic activation rate over the prepulse voltages from −180 to −120 mV (e.g. both reaching an asymptotical value at potentials more negative than −160 mV).

Figure 6.

The test-pulse voltage dependence of the inactivation kinetics with different holding potentials (Vh) or prepulses
The oocytes (n= 3–6) expressing the wild-type (A), the L361K mutant (B) or the L361R mutant Shaker K+ channels (C) were voltage clamped and stepped to different test pulses (the transverse axis) to measure the inactivation time constant (τ) of the macroscopic current. The same experiments were repeated with different holding potentials from −120 to −160 mV (at an interpulse interval of 1 s). Because it is difficult to keep the oocyte at a long holding potential at −180 mV, a 100 ms preceding pulse at −180 mV is instead given before each test pulse that elicits current. In this case, the potential at the interpulse interval was set at −120 mV. The length of the prepulses has been varied to ascertain that 100 ms is long enough to drive the voltage sensors to a steady-state distribution (data not shown), and thus a 100 ms prepulse is practically equivalent to Vh in this regard. D, the logarithm of the inverse of inactivation τ at the four most negative test pulses in A–C is fitted by a regression line with a slope of zF/RT, where z denotes the equivalent gating charges involved in the inactivation process. We chose to fit the more negative test-pulse voltage ranges, where the movement of the voltage sensor is more likely to be rate-limiting in the overall inactivation process than the closure of the inactivation gate itself (or other voltage-independent processes). The estimated ‘z’ is then plotted against the holding (prepulse) voltages to reveal the holding-voltage effect on the amount of the gating charges involved in the subsequent gating upon depolarization. The data points of the L361R mutant channels are fitted with a Boltzmann function of 0.29/(1 + exp((V− (−140.4))/5.6)) + 0.1 (the continuous line).

R365C becomes accessible to extracellular MTSEA with a double mutation at L361R

We have argued that the S4 in the resting L361R mutant channel is partially extruded, and it may be desirable to measure the gating currents at this point for an additional support of the argument. However, because of the huge negative shift of the gating curve in the voltage axis, we need to go to ∼−200 mV or even more negative voltages to obtain accurate template currents for subtraction of the linear capacitance currents. It is technically very difficult to have small-enough leak currents at those negative voltages to document reliable template currents. We therefore tried another, and perhaps even more straightforward, experimental approach. The transmembrane movement and positioning of S4 had been assessed by the voltage-dependent changes in the accessibility of extracellular or intracellular hydrophilic sulfhydryl-specific probes to different S4 residues (e.g. Bezanilla, 2000; Gandhi & Isacoff, 2002; Jiang et al. 2003a). We made cysteine substitutions for the top three arginines, R362, R365 and R368, of S4 and examined their accessibility to external hydrophilic MTSEA in the resting state (MTSET was also tested, but the effect was not evident in all cases under our experimental conditions). Figure 7A shows that only the R362C channel can be slightly modified by MTSEA, and both R365C and R368C do not respond to this external chemical reagent at −100 mV. The findings are consistent with a slightly externally exposed R362 and buried or internally exposed R365 and R368 in the resting channel (Larsson et al. 1996; Baker et al. 1998). When a second mutation of L361R is added, however, the accessibility of R362C is significantly increased, and most dramatically, R365C becomes highly accessible to external MTSEA at −100 mV (Fig. 7B and C), as if the buried R365C becomes exposed to the external aqueous environment because of the extrusion of S4 made by L361R in the resting channel. R368C still has no response to MTSEA with the L361R double mutation. The external exposure of the partly buried R362 and the completely buried or internally located R365 (but not R368) is very much consistent with the finding of ≥ 1 charge transfer per subunit between the partly extruded and the most retracted position of S4 in the L361R mutant channel (Fig. 6D), and is also reminiscent of the previous finding that the top two arginines (R362 and R365) are exposed to the external aqueous environment in the activated channel (probed also by MTS modification method, Larsson et al. 1996; Baker et al. 1998). It is likely that, at a membrane potential of ∼−100 to −120 mV, S4 in the L361R mutant channel has been moved close to the activated position (with the top two arginines exposed and accessible to the external MTS reagents). The macroscopic activation and inactivation kinetics of the L361R channel therefore would stay close to the saturating value of the wild-type channel with much less voltage dependence of gating when the currents are elicited from a holding potential of −120 mV (Fig. 4A and B). The voltage dependence is steeper and the midpoint is leftward shifted in the activation curve of the MTSEA-modified R365C plus L361R double mutant channel compared to that of the unmodified channel (Fig. 7C), as if the gating charges are restored and the extruded position of S4 is stabilized by the addition of a positively charged adduct back to the position of R365. These results thus confirm a partly extruded S4 in the resting L361R mutant channel.

Figure 7.

The changes in the accessibility of S4 basic residues to extracellular MTSEA in the resting channels by L361R mutation
IR variants of the indicated mutants are expressed. The currents are transiently elicited and recorded upon a repeated 40–50 ms step depolarization at +40 mV with a holding potential at −100 mV for 3 s between the steps. The freshly prepared 1 mm MTSEA solution is externally perfused until a steady-state effect is reached. The drug perfusion is followed by a thorough washing with ND-96 to ensure an irreversible modification. The representative currents and the activation curves before and after MTSEA modification are then superimposed for comparison. The dotted lines indicate zero current level. A, MTSEA (grey lines) slightly increases the macroscopic current of the R362C mutant, but does not affect the R365C and R368C mutant currents. B, with a concomitant mutation of L361R, both R362C and R365C become readily accessible to MTSEA (grey lines), leading to an increase of the macroscopic current. The R368C plus L361R double mutant channel still does not respond to external MTSEA. C, the activation curves in the R365C plus L361R double mutant channel. The negative shift of Vhalf (−44.4 mV vs. −0.3 mV) and the decrease in k value (4.4 vs. 8.3) are manifest after MTSEA modification.


L361 plays a critical role in stabilization of S4 in the appropriate resting position

The outermost charged residue in the S4 segment in voltage-gated channels (e.g. R362 in the Shaker K+ channel) serves as an important gating charge that could be located within the membrane field in the resting channel and emerge to the extracellular space upon channel activation (Chahine et al. 1994; Yang & Horn, 1995; Yang et al. 1997). L361 is a hydrophobic residue just external to R362 in the Shaker K+ channel. This hydrophobic residue is highly conserved in mammalian voltage-gated channels as well as the voltage-gated channels from lower organisms. However, the microenvironment and the functional role of this hydrophobic residue remain essentially unexplored. Our findings indicate that the L361R mutation greatly elevates the free energy level of S4 at its original resting position. S4 thus has been partially extruded at a resting membrane potential of −120 mV, and needs stronger hyperpolarizing potentials to drive it back to the original resting position. This profound destabilization and mobilization of S4 in the resting channel is never found with the other manipulations on the S3–S4 linker residues (Larsson et al. 1996; Mathur et al. 1997; Gonzalez et al. 2000, 2001). When nearly the entire S3–S4 linker is deleted except L361 (i.e. deletion of residues 330–360), the activation curve is only rightward shifted by ∼20 mV (Gonzalez et al. 2000, 2001), which is a much smaller shift and opposite in direction compared to that caused by a point mutation at L361. Moreover, the activation kinetics are significantly slowed rather than accelerated by the deletion of the S3–S4 linker. The great elevation of the free energy level in the resting state by a point mutation at L361 suggests that the S4 helix in the resting channel is probably ‘anchored’ by a very narrowed region at its amino terminus (at or close to L361, the pivotal residue at the junction of S4 and S3–S4 linker) rather than by the much larger neighbouring S3–S4 linker. In this regard, we have also done the arginine scanning on the other hydrophobic residues in the vicinity of L361 (from A355 to L375). None shows a shift in the gating voltage range comparable to that of L361, although L358R and I364R also have a smaller-scale negative shift (roughly 20–30 mV, data not shown). These three residues (L358, L361, and I364) in a ‘row’ separated by the same pitch of three as the arginines in the S4 α-helix may mark an imperative short ‘side’ of the S4 helix. This ‘side’ could be responsible for the proper positioning of the voltage sensor in the resting channel, with its effect centred and emphasized in L361.

L361R mutation may cause at least ∼1 kcal mol−1 free energy change in a flexible hydrophobic locus

The change of the relative tendency to stay in the resting and activated (or inactivated) states at 0 mV by a mutation (i.e. ΔΔG°) can be derived by subtraction of the free energy difference (ΔG°) between the two gating states in the wild-type channel from that in the mutant channels. Because channel activation or inactivation probably involves intermediate gating states (see below), the Vhalf and equivalent gating charge (z, being equal to 25 divided by the slope factor k in mV) of the activation and inactivation curves could not be as rigorously interpreted as that in a truly two-state system (e.g. the equivalent gating charges obtained from the fits of Boltzmann function tend to be an underestimate of the total gating charges transferred between the fully resting and activated states of the channel). Nevertheless, one may still get a rough estimate of ΔΔG° by simplistically assuming the product of Vhalf and zF as ΔG°, keeping the limitations of this approach in mind. The L361R mutation in this regard causes a ΔΔG° that destabilizes the resting state (relative to the activated or inactivated states) by at least ∼4 kcal mol−1, or ∼1 kcal mol−1 per subunit (Fig. 2). A rough estimate of the ΔΔG° caused by L361R mutation in the resting channel can also be derived by the findings that ∼40 mV more negative potential (e.g. −120 to −160 mV) and the transfer of ≥∼1 charge (Fig. 6D) or ∼1.5 charges (R365 and part of R362, Fig. 7) across the membrane field would drive S4 back to a fully retracted resting position. The calculated result, ∼40–60 meV or ∼1–1.4 kcal mol−1 per subunit, is very much consistent with the foregoing estimate from steady-state gating curves and is a much larger energy term than the cases of the high-impact residues reported by Li-Smerin et al. (2000). In view of the extremely fast activation kinetics at −120 mV that can be ‘normalized’ by even more negative holding potentials, the major part of the ∼1 kcal mol−1ΔG° change in the L361R channel presumably is ascribable to the destabilization of S4 in the resting (retracted) position rather than to the stabilization of S4 in the activated or inactivated (extruded) position. Because it has been reported that movement of a charged arginine residue in a short peptide into the lipid bilayer could cost ∼0.7 kcal mol−1 higher energy than movement of a glycine residue (White & Wimley, 1999), the energy contributed by the interaction between just one altered critical residue and its surroundings could be compatible with the profound change in channel gating in our case. Replacement of L361 with other hydrophilic residues (e.g. L361K, Q, or E) may also cause substantial free energy increase in the resting state. By contrast, replacement of L361 with a hydrophobic residue of different side chain geometry (e.g. L361F) causes little gating changes. These findings imply that hydrophobic interaction is critical for the stabilization of L361 and thus S4 in the retracted resting position. Because effective hydrophobic interaction usually requires close proximity of the binding counterparts (to have adequate London dispersion forces by the attraction between the induced dipole-induced dipole and the entropy gain due to freeing the water molecules from the hydrophobic surfaces, Zimmerman & Feldman, 1989), L361 in the resting S4 probably resides in an absolutely hydrophobic but flexible environment, so that leucine is energetically replaceable by the structurally different phenylalanine. In this regard, L361 residue (the top of S4) is likely surrounded and coordinated by the hydrocarbon parts of the membrane lipid in the resting state. This proposal is reminiscent of a recent structural study which locates the lipid-exposed S4 domain (e.g. the first two arginines) in the activated Shaker K+ channels at the lipid–protein interface close to the negatively charged phosphodiester groups (Long et al. 2005; Schimidt et al. 2006; see also Cuello et al. 2004; Jiang et al. 2004). On the other hand, however, this proposal may also be compatible with the more conventional ‘gating-canal’ model, where S4 is mostly shielded from the membrane lipid by the other parts of the K+ channel protein, with the first arginine (R362) positioned very close to the pore domain of the other subunit (Gandhi et al. 2003; Laine et al. 2003). In this latter case, it is still possible for L361, which is located 100 deg away from R362 in the α-helix, to face a biochemically absolutely hydrophobic but structurally flexible environment such as membrane lipid. At any rate, the short ‘row’ of hydrophobic residues centred at L361 on one side of S4 α-helix may well be responsible for the stabilization and properly positioning of the charge-rich voltage sensor into the cell membrane, and consequently for appropriate gating control. In view of the vital stabilization effect of L361 on S4, it is notable that this hydrophobic residue right external to the first arginine is highly conserved in essentially all of the voltage-dependent K+, Na+ and Ca2+ channels, and that the hydrophilic substitutions at this position in the fourth domain of the Na+ channel have been shown to largely facilitate channel inactivation (Yang & Kuo, 2003).

The S4 in the L361R mutant channel is extruded by a major proportion of its ‘normal’ pathway at −120 mV

We have argued that at −120 mV S4 in the L361R mutant channels already assumes a partially extruded (intermediate) position, whereas S4 in the wild-type channel would tend to stay in the fully retracted position. Because the S4 helix presumably is moved as a whole, the other S4 residues should not be surrounded by energetically very unfavourable microenvironments in the intermediate position (so that a partially extruded S4 is not more unfavourable than a most retracted S4 where the L361R residue stays with a hydrophobic environment in the L361R mutant channel). This is consistent with the proposal that such microenvironmental interactions for the ‘deeper’ residues in S4 are similar in the resting and the activated channels (Gandhi et al. 2000). Similarly, it has been reported that activation of the Na+ channel exposes the top arginines of S4, but not the hydrophobic residues lying in between, to the external hydrophilic solution (Yang et al. 1997). The partial extrusion of S4 in the resting L361R channel therefore very likely takes the same path as that S4 ‘normally’ goes through. In this case, the decrease of equivalent gating charge by 75% (approximately from 0.4 to 0.1) in Fig. 6D would then suggest that S4 has gone through a large portion of its ‘normal’ outward journey in the L361R channel at −120 mV. The large portion of movement is also supported by the findings in Fig. 7, which demonstrate that S4 in the resting L361R mutant channel (at −100 mV) takes a very similar extruded position to that of the S4 in the activated wild-type channel. The similar constant z-value in the L361K channel (∼0.5) to the saturating z-value in the L361R channel (∼0.4, Fig. 6D) suggests that the S4 in the L361K channel would probably rather stay in the retracted position than take the intermediate position at −120 mV, and thus the intermediate position is more ‘selective’ for arginine than lysine. This is reminiscent of the fact that arginines rather than lysines are used as charge carriers in the S4 voltage sensor. In this regard, the much larger z-value of ∼1.2 in the wild-type channel (Fig. 6D) would then imply an even less ratchet-like S4 movement (i.e. with fewer intermediate stops). The transfer of 3–4 gating charges in each subunit probably can be functionally or operationally viewed as a one-step reaction at a moderately depolarizing potential, well consistent with the mechanistic rationales underlying the perfect description of K+ channel activation by a n4 curve (Hodgkin & Huxley, 1952).

Opening of the Shaker K+ channel needs a complete outward translocation of S4

When the voltage sensors are driven back to a more retracted position by an extreme negative holding potential, it is notable that the L361R channel shows much slower kinetics of subsequent activation (and inactivation) than the wild-type channel. The inner (or early) part of the pathway that L361 (one side of the top of S4) travels through during channel activation thus seems to be narrow enough to delay the transition of the larger side-chain of arginine (but not leucine). This is consistent with previous reports that voltage-sensor translocation could be slowed by the other bulky attachments on S4 (e.g. methanethiosulphonate–ethyltrimethylammonium or biotin–avidin, see Jiang et al. 2003b; Yang et al. 1996). These findings together suggest that the initial part of the outward journey of S4 is probably through a rather confined pass, which may involve a protein ‘canal’ or lipid–protein interface (Cuello et al. 2004; Jiang et al. 2004; Long et al. 2005). Moreover, the early movement of S4 from the most-retracted position to the partially extruded position should be much slower than the late transitions of S4 during the L361R channel activation. Thus the activation speed is greatly accelerated (even faster than the wild-type case) when the L361R channel is activated from −120 mV (Figs 1 and 4). The confined path along with the partially protruded S4 that skips the confined path and thus leads to accelerated gating kinetics may also be part of the mechanisms underlying the intriguing fast and slow gating modal changes in the EAG K+ channel (see Schönherr et al. 2002). On the other hand, a cooperative movement (or intersubunit interaction) of the four S4 segments in a channel may further contribute to the greatly accelerated activation speed in the L361R channel, where all four S4s may have been largely extruded at a resting potential. It is then surprising that, with a large portion of the designated S4 movement finished at −120 mV, the L361R channel still remains closed. This would strongly suggest that S4 should be fully extruded before the channel gate can be opened, an idea well consistent with the rationales underlying the description of deactivation kinetics of K+ currents typically by a monoexponential decay (Hodgkin & Huxley, 1952).



This work was supported by grant NSC-94–2320-B-002–004 from the National Science Council, Taiwan, grant NHRI-EX96-9606NI from the National Institute of Health, Taiwan, and the Topnotch Stroke Center at Taipei Medical University, Taiwan.