Role of the S3-S4 Linker in Shaker Potassium Channel Activation

The high sensitivity of voltage-gated potassium and sodium channels arises from a large charge displacement, ∼13 e₀ in magnitude, as a channel moves from its resting state to the open state (Schoppa et al., 1992; Hirschberg et al., 1995; Aggarwal and Mackinnon, 1996; Seoh et al., 1996). It is widely thought that this charge displacement involves the S4 region of the protein sequence, with contributions from other regions (Sigworth, 1994; Seoh et al., 1996). The S4 region of the Shaker potassium channel has seven basic residues which could contribute to the mobile charge, and recent experiments have shown changes in the accessibility of S4 residues upon voltage-dependent channel activation (Larsson et al., 1996; Yang et al., 1996). According to some theories, the S4 segment is an alpha helix that undergoes a large, helical screw movement normal to the lipid bilayer in order to transfer the necessary gating charge (Catterall, 1986; Durell et al. 1992). If indeed this is how the S4 region moves, then the S3-S4 linker, to which it is tethered at its NH₂-terminal end, must be flexible to support the large displacement.

There has been little study of the S3-S4 linker of voltage-gated channels, apart from work on L-type calcium channels by Nakai et al. (1994). These authors showed that exchange of the S3-S4 region in Domain 1 changed the channel from the skeletal muscle (slowly activating) to the cardiac (rapidly activating) kinetic behavior. A comparison of potassium channel S3-S4 linkers is shown in Fig. 1. The apparent length of the linker is variable between subfamilies, ranging from 25 residues for Drosophila Shaker to only 7 residues in Shaw. The members of the Shaker subfamily show a moderate degree of sequence identity, but within the Shab and Shal subfamilies there is a remarkable degree of identity between Drosophila and mammalian members, similar to the level of identity in the S4 regions. In the present study we investigate the role of the S3-S4 linker by asking how point mutations or major alterations in its sequence affect the voltage dependence and kinetics of activation gating in the Shaker K⁺ channel.

Our constructs were based on ShΔ, a Shaker 29-4 cDNA clone (Kamb et al., 1988) having a deletion (Δ2-30) to remove fast inactivation (Hoshi et al. 1990). To better allow comparison with other reports in the literature, we adopt the residue numbering of the Shaker B splice variant (Schwarz et al, 1988), which differs from 29-4 at the NH₂ terminus and at four residues in the COOH-terminal region. The 2.1 kb ShΔ cDNA was subcloned at the EcoRI and HindIII restriction sites of pOEV (Pfaff et al, 1990). The LP-USE technique (Deng and Nickoloff, 1992; Ray and Nickoloff, 1992) was used for site-directed mutagenesis. In this technique a mutagenic and a selection primer are joined using PCR to obtain a single long primer for subsequent annealing with the denatured parental DNA and elongation with T4 polymerase. The mutagenic primers (synthesized at the DNA Synthesis Laboratory, Yale University) were 24–60 nucleotides in length, while the 24-mer selection primer changed the unique HindIII site to a MluI site. A PCR product of ∼1.25 kb was generated using AmpliTaq DNA polymerase (Perkin Elmer Corp., Norwalk, CT) and separated from the unreacted primers by electrophoresis in low-melting 1% agarose gel. We used the Transformer Site-Directed Mutagenesis Kit (Clontech Laboratories, Palo Alto, CA) and the manufacturer's protocol to carry out the denaturation, annealing, elongation, ligation, and transformation steps into Escherichia coli strain BMH 71-18 mutS, which is defective in the mismatch repair function. Mutants were identified (blue colonies on X-gal/IPTG) by digestion with MluI restriction enzyme. The 1.25-kb PCR-generated segment in each positive mutant was completely sequenced (Sequenase Kit; U.S. Biochemical, Cleveland, OH). In a few cases, we obtained the desired mutant by direct colony transfer and hybridization with the respective mutagenic primer, using the ECL nonradioactive gene detection system (Amersham Corp., Arlington Heights, IL). The positive mutants were sequenced as stated above. The cDNAs were linearized with MluI (or EcoRV in some mutants obtained by the ECL system), and the capped, in vitro run-off T7 transcripts were synthesized and quantitated approximately by the intensity of ethidium bromide stained bands in 1% agarose (with 6% formaldehyde) gel. RNA was stored at −70°C.

Harvested oocytes from Xenopus laevis were defolliculated by incubating with collagenase (type 1a; Sigma Chemical Co., St. Louis, MO) in OR3 medium (Blumenthal and Kaczmarek, 1992), which consisted of 50% Leibovitz's L-15 medium (Gibco BRL, Grand Island, NY), 15 mM HEPES, 5 × 10⁴ U/liter Nystatin (Sigma Chemical Co.), 10 mg/liter Gentamycin (Sigma Chemical Co.), adjusted to pH 7.4. When the oocytes were well separated from each other (after ∼2 h at room temperature), they were repeatedly washed with the OR3 medium to remove collagenase. Stage V and VI oocytes were sorted and maintained in OR3 medium at 20°C, before and after RNA injection. 50–100 nl of cRNA solution was injected into the vegetal pole of each oocyte. Concentrations of injected cRNAs were varied to control the level of expression.

Oocytes were incubated for 2–6 d before recording currents from them at room temperature by two-microelectrode voltage-clamp (OC-725; Warner Instruments, New Haven, CT), using the Pulse software (HEKA-Electronic, Lambrecht, Germany) running on a Macintosh computer. Microelectrodes were filled with 1 M KCl and had resistances of 0.1 to 1 MΩ; a grounded shield between the electrodes reduced coupling capacitance. The standard bath solution was ND-96 (96 mM NaCl, 2 mM KCl, 1.8 mM MgCl₂, 1 mM CaCl₂, 5 mM HEPES, pH adjusted to 7.4). Bath electrodes were chlorided silver wires, except for experiments in which dithiothreitol was introduced into the bath, in which case a salt bridge was used. The current monitor signal was filtered with a 2 kHz Bessel filter and the linear leak and capacitative currents were subtracted by the P/4 method (Bezanilla and Armstrong, 1977) from a subtraction holding potential of −120 mV. The series resistance in the voltage-clamp system was estimated to be <200 Ω. For characterization of kinetics and activation, oocytes with currents at +70 mV in the range of 10–50 μA (mean 32 μA) were used.

Conductances were computed assuming a linear open-channel i-V relationship and a reversal potential of −80 mV, and normalized by the peak conductance (g_max) measured at +70 mV. Voltage dependence of activation was characterized by fitting to the fourth power of a Boltzmann function,

to obtain the steepness factor k_a and the midpoint voltage V_a.

Activation was characterized by a fit to the time course of the current elicited by depolarizations from a holding potential of −80 mV. The fit started near the time at which the current had reached 50% of maximum and ended at 50 ms. The fitting function was the sum of two exponential functions,

where I_t is the current at time t and I_o is the extrapolated value of the current at t = 0. The faster of the two time constants, τ_a1, always had the larger amplitude, >75% of the total amplitude at −30 mV and at 0 mV.

The time constants of current deactivation were obtained by fitting the currents following a 20-ms activating pulse to +40 mV, with a delay to allow instantaneous current changes to settle. The time course was fitted with the sum of two exponentials; the faster component τ_d1, which accounted for ∼90% of the amplitude (from −80 to −20 mV) is shown in Fig. 7 and Table I. Time constant values from multiple determinations are reported as geometric means and SEM values.

Fig. 2 summarizes the mutations that were made. The first five involved changes in the four contiguous acidic residues EEED at the NH₂-terminal end of the linker. Deletions of glutamate residues and charge reversals by replacement with lysines were performed. The second group of mutations involved deletions at the COOH-terminal end of the linker, where the AMS sequence is highly conserved within the Shaker subfamily. In the third group of mutations each of the proline residues in the S3-S4 region was changed to alanine, one at a time. The fourth group of mutations consisted of various insertions following P344. Insertions were a single cysteine residue or peptides 4–9 residues in length. The inserted sequences were influenza hemagglutinin (HA; Kast et al., 1995) and FLAG (Zhong et al., 1995) epitopes, and a Factor X_a cleavage-site sequence (Wearne, 1990). The fifth group were chimeric constructs, in which the Shaker linker was replaced with the shorter S3-S4 linkers from Shab, Shaw, and Shal genes.

The cRNAs bearing the various mutations were injected into Xenopus laevis oocytes, and the whole-oocyte K⁺ currents were recorded by two-microelectrode voltage-clamp. Robust currents from each of the 18 mutants implied that none of mutations in the S3-S4 linker affected the synthesis and assembly of the functional Shaker K⁺ channels in oocyte membranes.

Changes in channel activation due to the mutations were assayed using depolarizations and a deactivation protocol as shown in Fig. 3. Activation was characterized by the voltage dependence of steady-state conductance and an activation time constant. Starting at the time at which the current reaches 50% of its final value, the activation time course was fitted to the sum of two exponentials. The time constant τ_a1 of the faster exponential was taken to be a measure of the activation kinetics, since the slower exponential had a small and variable amplitude, always <25% of the total. Deactivation was similarly characterized by the faster time constant τ_d1 of a two-exponential fit to the decaying phase of the tail current. In representative mutants a rough “limiting slope” measurement of gating valence was also made, and estimates of reversal potential were made from tail currents from each mutant construct. The parameters obtained from each mutant are shown in Table I and are described in detail below.

The acidic residues E333, E334, E335, and D336 form the NH₂-terminal end of the linker in Shaker. As can be seen in Fig. 1, all Shaker-related mammalian K⁺ channels, except Kv1.4, have one or two acidic residues within these four positions. We tested the influence of these charged residues by either deleting them or by reversing the charge with lysine residues, either one at a time or simultaneously. The deletions had little effect on the voltage dependence of activation, but there was a trend toward faster activation at negative potentials with the single deletion ΔE (Fig. 4,A). The average shifts in the activation voltage V_a were between −7 and −11 mV (Table I), which is in the opposite direction from that expected if the neutralization of these acidic residues simply affects the surface charge at the extracellular surface.

The charge reversal E333K also resulted in a −7 mV shift in activation. The reversal of all four charges in the EEED → KKKK mutant produced a +5 mV shift in both steady-state activation and the time constant curves (Fig. 4 B); this small shift is in the direction expected for a surface charge effect.

The COOH-terminal ends of the linkers in Shaker and its mammalian homologues have four identical amino acid residues (QAMS), except for the presence of threonine in place of M356 in Kv 1.1 (Fig. 1). Deletion of M356 alone shifted the mean voltage of activation by −12 mV when compared to the wild-type channel. The ΔAMS mutant showed a mean activation voltage close to that of the wild type (Table I), although the slope of activation increased. The ΔA and ΔAMS mutants increased the activation time constant at depolarized voltages (Fig. 4,C). A comparison of activation time courses at −20 mV (Fig. 5) demonstrates the slowing of activation by these deletions.

The Drosophila Shaker has a proline-rich sequence (P-X-X-P-X-X-P) in the center of the S3-S4 linker, which is absent in the linkers of the Shaker-related mammalian channels and in the other K⁺ channel subfamilies. Proline is known to be a helix-breaker, and in multiples it appears predominantly in flexible regions of proteins (MacArthur and Thornton, 1991; von Heijne, 1991). We tested the effect of substituting each of the proline residues with alanine, one at time. The results showed that none of the substitutions resulted in an alteration in the channel activation voltage and little effect on kinetics (Fig. 6,A and Table I).

Neither Shaker nor its higher homologues has a cysteine residue in the S3-S4 linker. Inserting a Cys after Pro 344 did not affect the voltage dependence of activation, but the insertion did slow the time course of activation at depolarized potentials. The mean τ_a1 at potentials between 0 and +40 mV was about twofold greater than that of the wild type (Fig. 6 B), but the deactivation time constant τ_d1 remained unchanged. Application of 1 mM dithiothreitol either before or during the recording did not change the amplitude or kinetics of the current.

The mutant channels with epitopes expressed very well and caused only small alterations in the channel's electrophysiological properties. Their activation voltages and kinetics were close to that of the wild type. However, both the HA and the Factor X_a mutants showed a slight reduction in the slope of activation (Fig. 6,B and Table I).

The relatively long Shaker linker (25 amino acids) was replaced with the much shorter linkers from Shab (9 amino acids), Shaw (7 amino acids), and Shal (7 amino acids). The Shab and Shaw linkers resulted in a three- to fourfold slowing of activation throughout the voltage range (Fig. 6 C), and produced −13 and +7 mV shifts, respectively. The Shal chimera showed a shift of +10 mV but little change in kinetics.

The kinetic effects of these linker substitutions can be seen in the activation and deactivation time courses of Fig. 7, A and B. Whereas the Shal linker results in time courses indistinguishable from wild type, channels with Shab and Shaw linkers activate and deactivate more slowly. The deactivation time constants (Fig. 7,C) are also about threefold larger in these mutants. These effects are in contrast to the effect of the Cys insertion, where the activation time course is slowed but deactivation is as rapid as wild type (Fig. 7, A and B).

The voltage sensitivity of channel opening, measured at negative potentials where the open probability is low, yields a lower-bound estimate on total channel gating charge (Almers 1978; Hirschberg et al., 1995). With the two-microelectrode voltage clamp we were able to make reliable measurements of channel activation only down to open probabilities p_o ∼ 10⁻². Nevertheless, we computed the apparent gating charge as a rough measure to check whether any of the various classes of mutations resulted in a large change in the limiting voltage sensitivity. Fig. 8,A shows activation data from a ShΔ oocyte; Fig. 8 B shows the apparent gating valence z_app computed according to:

and plotted as a function of the steady-state conductance G. For consistency we evaluated z_app in each oocyte at the potential where G is 1% of its maximum value. The values so obtained are much less than the expected asymptotic value of ∼13, but are expected to show any large changes in total gating charge or other effects on the gating mechanism. For all the mutants, these values were close to the value of 6.0 obtained for ShΔ (Table I).

To evaluate a possible change in the selectivity for K⁺ over Na⁺ ions, we estimated the reversal potentials in tail-current experiments. Because of the limited speed of the two-microelectrode voltage clamp, clear reversal of the current was not always seen. With the ND 96 bath solution containing 2 mM KCl and 96 mM NaCl, the ShΔ channel had an apparent reversal potential of −85 mV. In all but two of the mutants, reversal potentials were observed in the range −75 to −89 mV (means of 2–7 oocytes where current reversal could be recorded). For the ΔA and ΔE mutants the reversal was more difficult to detect, but nevertheless appeared to be in the same voltage range. The similarity of reversal potentials is consistent with there being no change in the selectivity due to mutations in the S3-S4 linker.

Our goal in the present study was to investigate the role of the S3-S4 linker region in the activation of the Shaker K⁺ channel. The results are interesting in two respects. First, the effects of mutations on the voltage dependence of activation were surprisingly small. Despite the proximity of this extracellularly disposed region to the S4 region, which has been implicated as an important part of the voltage sensor, large changes in charge and in the length of the S3-S4 linker produced only minor changes in the voltage dependence of activation. Voltage shifts were at most 13 mV in magnitude, and changes in other measures of voltage sensitivity were similarly small. Second, the main effect of some mutations, including the replacement of the 25-residue Shaker sequence with the 9 and 7-residue Shab and Shaw linkers, were changes in the speed of activation and deactivation. It is interesting that these kinetic effects occurred in the absence of large changes in the voltage dependence of activation.

An examination of the linker sequence within the members of the four different K⁺ subfamilies reveal some notable features (see Fig. 1). First, in Shaker and its mammalian homologues the linker shows greater variation in length and composition than it does in the members of the Shab, Shaw and Shal subfamilies. The Drosophila Shaker linker is the longest (25 amino acid residues) and except for a few identical amino acids or conservative substitutions at the NH₂- and the COOH-terminal ends, it is quite different than the mammalian linkers, which are much shorter and have a preponderance of glycine and glutamine residues.

The shorter (7 to 9 residue) linkers in the Shab, Shaw, and Shal subfamilies are strikingly well conserved across the different members. The Drosophila Shal linker is essentially identical to its rat, mouse, and human counterparts. The linkers within the mammalian Shab-related sequences are also identical to each other, and the Drosophila Shab linker is related to these in having five identical residues out of the total nine and has one conservative substitution. A potential N-glycosylation consensus site in this linker is conserved across all Shab members and species. The linkers within the members of the Shaw-like mammalian subfamily too are nearly identical; however the Drosophila linker matches poorly, having only one identical residue and three conservative substitutions out of a total of seven residues. Evidently, the S3-S4 linkers do not represent a collection of random sequences, but form groups of related sequences, which are characteristically specific for each of the four subfamilies; some even have conserved gly-cosylation sites.

We created several different types of mutations in the Shaker linker and performed an initial characterization of their effects on channel activation, using the two-microelectrode voltage clamp technique. The results showed that these mutations had at most moderate effects on the voltage dependence of activation. Evidently, the linker does not form a part of the voltage-sensor of the channel, despite its proximity to S4. The small effects from the deletion or charge reversal of the residues most distant in the sequence from S4, the acidic residues E333, E334, E335, and D336, suggest that they are physically distant from the voltage-sensing moieties of the channel. The substitutions of each of the proline residues with an alanine also had no effect on the voltage dependence or the kinetics, even though the proline triplet repeat would be expected to contribute to the flexibility of the linker. Elongating the linker by adding several different types of short peptide epitopes also did not affect the voltage dependence of activation or the kinetics significantly. The FLAG peptide is very hydrophilic in its composition (7 of its 8 residues are either acidic or basic), while the HA peptide would increase the hydrophobicity in the linker. The results with these peptides suggest that the polarity of the linker's environment does not change in the course of channel gating. It will be interesting to test for a change in the accessibility of the linker in situ from the external surface either by using the monoclonal antibody against the HA peptide, as has already been done in the case of a FLAG epitope by Shih and Goldin (1995), or by enzymatic cleavage of the linker by enterokinase and Factor X_a protease.

Certain of the mutations affected the time course of activation of the macroscopic ionic currents. Deletion of one or more of the residues at the conserved COOH-terminal end of the linker, introduction of a single Cys residue, or swapping the Shaker linker with the short linker segment from either the Shab or the Shaw subfamily of voltage-gated K⁺ channels slowed the rates of activation at all depolarizing potentials. Recent studies have shown that the position M356 can be accessed from the outside at both resting and depolarized membrane potentials (Larsson et al. 1996; Mannuzzu et al. 1996). Our results show that deletion of the adjacent A355 alone or with two neighboring residues (ΔAMS) increase the activation time constant with only small effects on the voltage dependence, whereas deletion of M356 alone slightly decreases the time constant of activation and shifts the activation voltage. Thus the residues which are closest to S4 seem to be important in affecting the rate of one or more transitions leading to the opening of the channel; however the surprising effect of the Cys insertion in the middle of the linker suggests that other interactions may also be present.

Chimeric channels (ShΔ/Shab and ShΔ/Shaw) produced a larger slowing of channel activation (three- to fourfold increase in τ_a1), whereas the ShΔ/Shal chimera had little kinetic effect. Deactivation kinetics were also slower for the channels with the Shab and Shaw linkers, although the Shal linker, being like Shaw only seven residues in length, yielded normal deactivation kinetics. It is interesting that the changes in kinetics were accompanied by only small shifts in voltage dependence. This is in contrast to the effects of most mutations that affect Shaker activation, where kinetic changes are typically accompanied by large voltage shifts. The origin of the kinetic effects of the mutations studied here is unclear. One explanation is that the mutations may affect the stability of intermediate conformations that are visited during the activation process. The long delay in the time course of activation of the channels implies that there are many such intermediate states (Zagotta et al., 1994). The interactions that would allow the mutations to affect the stability of intermediates are unknown, but it should be kept in mind that two- to fourfold changes in rates represent small free-energy differences, and would reflect rather weak interactions.

Wei et al. (1990) showed that the four voltage-gated K⁺ subfamilies in Drosophila differed in the rates of activation of their macroscopic currents, in the order Shaker > Shal > Shaw > Shab and spanning an overall range of 16-fold. Though on a reduced scale, our chimeras follow the same rank order of changes in activation rates (Fig. 6 C). It appears that in the chimeric channels the S3-S4 linker partially confers the parental channel type's kinetics on Shaker. This suggests a possible role of the linker in establishing the time scale of activation of the channels. A similar conclusion has been reached from studies of L-type Ca²⁺ channel chimeras (Nakai et al, 1994), where substitution of the S3-S4 linker of Domain 1 conferred the kinetic phenotypes of skeletal muscle and cardiac Ca²⁺ channels.

Residues in the S4 region appear to constitute most of the gating charge in Shaker potassium channels (Aggarwal and MacKinnon, 1996; Seoh et al., 1996). Accompanying the voltage-dependent gating charge movement are conformational changes that dramatically alter the accessibility of S4 residues to the inner and outer membrane surfaces in Shaker channels (Larsson et al., 1996) and in Domain 4 of sodium channels (Yang et al., 1996). This apparent translocation of several residues from the inner to the outer membrane surface upon depolarization can explain much of the observed charge movement, ∼4 e₀ per subunit in magnitude. A popular model for the underlying conformational change is a helical-screw motion of an alpha-helical S4 region (Catterall, 1986; Durell and Guy, 1992). In this model some 20 Å of displacement would be required to translocate four charges, requiring that the S3-S4 linker be highly flexible. The Shaker S3-S4 linker, being relatively long and having multiple proline residues, seems well suited for this function. However, the linkers in the other potassium channel subfamilies are much shorter; in mammalian sodium channels the linkers are also very short, only 4–9 residues in length (Goldin, 1995). Do the shorter linkers preclude a large S4 translation? Molecular modelling by Dr. T.B. Woolf (personal communication) shows that the 7-residue Shab linker is sufficiently long to substitute for the Shaker linker the model of Durell and Guy in its “closed” and “open” states. Even with the Shab linker, no change in secondary structure of S3 or S4 is required between these states.

The large movement of the S3-S4 linker in the helical-screw model nevertheless would predict that alterations in the linker sequence would result in substantial differences in the free-energy difference ΔG for channel opening. The largest voltage shift (13 mV) induced by our mutations corresponds to a ΔΔG of at most 1.2 kcal/mole. This value is calculated from the electrostatic energy shift qΔV, where q is the total gating charge of 4.3 e₀ per subunit. The small ΔΔG suggests that the linker undergoes little if any conformational change in the process of gating. This is consistent with alternative views of S4 motion, in which it undergoes state-dependent movements or secondary structure changes that leave the S3-S4 region unaffected. A scheme Involving secondary structure changes in S4 was proposed by Guy and Conti (1990) to allow charge movement without requiring linker motion. A more recent proposal of a helix-to-loop rearrangement of S4 (Aggarwal and MacKinnon, 1996), which could also leave the S3-S4 linker largely stationary, becomes plausible in view of the very large peptide translocations that occur in voltage-dependent gating of colicin IA (Slatin et al., 1994; Qiu et al., 1996).

We thank Dr. Tom Woolf (Johns Hopkins University) for his insights from molecular modelling.

This work was supported by National Institutes of Health grant NS21501.