The voltage-sensing domains in voltage-gated K+ channels each contain four transmembrane (TM) segments, termed S1 to S4. Previous scanning mutagenesis studies suggest that S1 and S2 are amphipathic membrane spanning α-helices that interface directly with the lipid membrane. In contrast, the secondary structure of and/or the environments surrounding S3 and S4 are more complex. For S3, although the NH2-terminal part displays significant helical character in both tryptophan- and alanine-scanning mutagenesis studies, the structure of the COOH-terminal portion of this TM is less clear. The COOH terminus of S3 is particularly interesting because this is where gating modifier toxins like Hanatoxin interact with different voltage-gated ion channels. To further examine the secondary structure of the COOH terminus of S3, we lysine-scanned this region in the drk1 K+ channel and examined the mutation-induced changes in channel gating and Hanatoxin binding affinity, looking for periodicity characteristic of an α-helix. Both the mutation-induced perturbation in the toxin–channel interaction and in gating support the presence of an α-helix of at least 10 residues in length in the COOH terminus of S3. Together with previous scanning mutagenesis studies, these results suggest that, in voltage-gated K+ channels, the entire S3 segment is helical, but that it can be divided into two parts. The NH2-terminal part of S3 interfaces with both lipid and protein, whereas the COOH-terminal part interfaces with water (where Hanatoxin binds) and possibly protein. A conserved proline residue is located near the boundary between the two parts of S3, arguing for the presence of a kink in this region. Several lines of evidence suggest that these structural features of S3 probably exist in all voltage-gated ion channels.
Voltage-gated K+ channels are tetrameric membrane proteins that contain six transmembrane (TM) segments per subunit (see Fig. 1 A). The first four TMs are thought to form the voltage-sensing domains, whereas the fifth through sixth TMs assemble to form the pore domain (Kubo et al. 1993; Doyle et al. 1998; Li-Smerin and Swartz 1998; Yellen 1998; Li-Smerin et al. 2000). Our understanding of the pore domain structure has been guided by the crystal structure of the bacterial KcsA K+ channel (Doyle et al. 1998; MacKinnon et al. 1998), a relatively simple K+ channel (Schrempf et al. 1995) that lacks the voltage-sensing domains found in voltage-gated K+ channels. Although the 3-D structure of the voltage-sensing domains has not yet been obtained, scanning mutagenesis studies have provided information about the likely secondary structure of the TMs contained therein (Monks et al. 1999; Hong and Miller 2000; Li-Smerin et al. 2000). These scanning approaches rely on a bimodal environment surrounding the presumed helix to detect underlying periodicity in secondary structure. The premise is that mutations will be tolerated at interfaces with solvent (either lipid or water) as long as the substituted side chain is solvent compatible, but that they will be poorly tolerated at protein-protein interfaces. Both tryptophan-scanning mutagenesis in the Shaker K+ channel (Monks et al. 1999; Hong and Miller 2000) and alanine-scanning mutagenesis in the drk1 K+ channel (Li-Smerin et al. 2000) suggest that at least portions of S1 through S4 adopt an α-helical secondary structure. The scans of S1 and S2 suggest that these segments are relatively simple amphipathic α-helices that span the full width of the membrane and make extensive contacts with the lipid membrane. In contrast, only portions of S3 and S4 display helical character. For example, the alanine-scan of S4 detected helical periodicity in only the COOH-terminal region (Li-Smerin et al. 2000). Since the entire S4 segment consists of a repeating triplet of a basic and two hydrophobic residues, it seems likely that the secondary structure of the full-length of S4 is helical and that the absence of helical character in the NH2-terminal portion results from complexities in the environment surrounding the segment.
The present study focuses on the complexities in the S3 segment. In both tryptophan- and alanine-scanning mutagenesis studies, when the entire S3 segment is considered as a single TM, there is no clear indication of helical structure (Hong and Miller 2000; Li-Smerin et al. 2000). However, both scanning studies found evidence for helical structure when restricting the analysis to only the NH2-terminal two thirds of S3, up to a conserved proline residue (Pro268 in drk1). In the alanine scan, which covers the entire region from the NH2 terminus of S1 through the COOH terminus of S4 (including linkers), an unanticipated peak in the α-periodicity index (α-PI) is observed in the region encompassing the COOH terminus of S3 and the S3–S4 linker. This is illustrated in Fig. 1 B, where a windowed α-periodicity analysis is shown for mutation-induced perturbations in gating for the region spanning from S1 to S4. Although the mutation-induced perturbations in gating were rather weak for the region in question (Li-Smerin et al. 2000), the results hint that the COOH-terminal portion of S3 may have a helical structure.
The COOH-terminal region of S3 is of particular interest because it has been identified as an important region for interaction with various gating modifier toxins (Rogers et al. 1996; Swartz and MacKinnon 1997b; Li-Smerin and Swartz 1998, Li-Smerin and Swartz 2000; Winterfield and Swartz 2000). This is illustrated in Fig. 1 B for Hanatoxin, a 35–amino acid protein isolated from tarantula venom (Swartz and MacKinnon 1995) that inhibits the drk1 K+ channel by shifting activation to more depolarized voltages (Swartz and MacKinnon 1997a). The alanine scan of the region starting from the NH2 terminus of S1 to the COOH terminus of S4 in the drk1 K+ channel shows that the largest perturbations in toxin binding are located in the COOH terminus of S3 (Swartz and MacKinnon 1997b; Li-Smerin and Swartz 2000). In principle, the distribution of residues whose mutation alters toxin binding should also contain information about the secondary structure of the COOH terminus of S3. Although, the three positions where mutations had large effects on toxin binding (I273, F274, and E277) are distributed in a manner consistent with helical structure, clearly additional positions with large perturbations are needed to further examine this hint of secondary structure. The dimensions of Hanatoxin (∼20 × 25 Å) strongly suggest that many channel residues are located at the interface between toxin and channel (Takahashi et al. 2000). Perhaps, mutations more disruptive than substitutions to alanine may help to identify additional channel residues at the toxin–channel interface.
In this study, we systematically mutated 20 residues (one at a time) in the COOH terminus of S3 and the S3–S4 linker in the drk1 K+ channel to Lys to further evaluate the secondary structure of this region and to explicitly test the idea that an aqueous environment surrounds at least a portion of this region. The notion of water-exposed surfaces within S3 is based on the fact that Hanatoxin is a water-soluble toxin that presumably interacts with regions of the channel exposed to the extracellular aqueous environment. This is in contrast to S1, S2, and the NH2 terminus of S3, where the primary surrounding solvent appears to be lipid. We looked for helical periodicity in both the mutation-induced perturbation in Hanatoxin binding and in channel gating. As anticipated, the lysine scan resulted in larger perturbations at a greater number of positions, as compared with the previous alanine scan. Our results argue for α-helical secondary structure of the COOH terminus of S3, and suggest that one face of this helix is exposed to an aqueous environment where the toxin interfaces with the channel.
Materials And Methods
Mutagenesis and Channel Expression
Point mutations in drk1 K+ channels were introduced by generating mutant fragments by PCR and ligating into appropriately digested vectors. The construct used contained several previously introduced unique restriction sites (Swartz and MacKinnon 1997b). Mutations were confirmed by automated DNA sequencing. cDNAs encoding wild-type and mutant channels were linearized with NotI and transcribed with T7 RNA polymerase. Oocytes from Xenopus laevis frogs were removed surgically and incubated with agitation for 1–1.5 h in a solution containing (in mM): 82.5 NaCl, 2.5 KCl, 1 MgCl2, 5 HEPES, and 2 mg/ml collagenase (Worthington Biochemical Corp.), pH 7.6 with NaOH. Defolliculated Xenopus oocytes were injected with cRNA and incubated at 17°C in a solution containing (in mM): 96 NaCl, 2 KCl, 1 MgCl2, 1.8 CaCl2, 5 HEPES, and 50 μg/ml gentamicin (GIBCO BRL), pH 7.6 with NaOH for 1–4 d before recording.
Two-electrode voltage-clamp recordings were performed to study the toxin–channel interaction or the voltage-dependent gating behavior using an oocyte clamp (model OC-725C; Warner Instruments). Oocytes were perfused in a 160-μl recording chamber with a solution containing (in mM): 50 RbCl, 50 NaCl, 1 MgCl2, 0.3 CaCl2, and 5 HEPES, pH 7.6. Data were filtered at 2 kHz (eight-pole Bessel) and digitized at 10 kHz. Microelectrodes were filled with 3 M KCl and had resistances between 0.2 and 1.0 MΩ. All experiments were conducted at room temperature (∼22°C).
The voltage was typically held at −110 to −80 mV where no steady-state inactivation could be detected. For most channels, inward tail current (carried by Rb+ to slow the deactivation) was elicited by repolarization to −50 mV, and the amplitude of the tail current was measured 2–3 ms after repolarization. For the mutant channels that had large leftward shifts in the voltage-activation relationship, more negative tail voltages were used. For some mutant channels with large rightward shifts in the voltage-activation relationship, 0 to +20 mV was used to elicit an outward tail current.
Analysis of Toxin–Channel Interaction
The concentration dependence for occupancy of the channel by Hanatoxin was examined as previously described (Swartz and MacKinnon 1997a; Li-Smerin and Swartz 2000). We used at least two concentrations of Hanatoxin (between 100 and 250 μM) to initially evaluate the binding affinity of Hanatoxin to each mutant channel. The equilibrium dissociation constant (Kd) for Hanatoxin was calculated according to:
This equation assumes that there are four independent and equivalent binding sites on the drk1 K+ channel for Hanatoxin (Swartz and MacKinnon 1997a). For the mutant channels that show altered gating properties, we adjusted all voltage protocols appropriately to ensure that the plateau phase in the I/I0-voltage relation was clearly defined. For mutant channels showing >10-fold change in toxin Kd in the initial screen, three to six toxin concentrations were tested, and the concentration dependence for toxin occupancy was examined as shown in Fig. 3 A. In these instances, the Kd for toxin binding was obtained by fitting a form of the above equation to the data. The change in the free energy of toxin binding (ΔΔGT) was calculated for each mutant channel based on ΔΔGT = −RTln(Kdmut/Kdwt).
Analysis of Channel Gating
To evaluate mutation-induced perturbations in channel gating, we regarded the channel as having two states, closed and open, with a single transition between them. Voltage-activation relations were fit with single Boltzmann functions (see Fig. 5) according to:
where I/Imax is the normalized tail current amplitude, z is the equivalent charge, V50 is the half-activation voltage, F is Faraday's constant, R is the gas constant, and T is temperature in Kelvin. The difference in Gibbs free energy between closed and open states at 0 mV (ΔG0) was calculated according to ΔG0 = 0.2389 zFV50. The change in ΔG0 caused by each mutation ΔΔG0) was calculated as ΔΔG0 = ΔG0mut − ΔG0wt.
Analysis of Periodicity
Fourier transform methods (Cornette et al. 1987; Komiya et al. 1988; Rees et al. 1989a,Rees et al. 1989b) were used to evaluate the periodicity of either ΔΔGT (toxin binding) or ΔΔG0 (channel gating) for mutations according to the following equation:
where P(ω) is the Fourier transform power spectrum as a function of angular frequency ω, n is the number of residues within a segment, Vj is ΔΔGT or |ΔΔG0| at a given position j, and <V> is the average value of ΔΔGT or |ΔΔG0| for the segment.
As defined by hydrophobicity analysis, the S3 segment in the drk1 K+ channel begins around F254 and ends near S278. The region we focused on in this study begins immediately after P268 in S3 and ends at V288, near the NH2 terminus of S4 (Fig. 1 B). Our hypothesis was that at least part of this region adopts an α-helical secondary structure, with one face of the helix interacting with water and constituting a major part of the receptor for gating modifier toxins. In contrast, a distinct face of the helix may interact in some way with other elements of the voltage-sensing domain. To test this hypothesis, we mutated each residue to Lys and examined both the effects on Hanatoxin binding affinity and on the gating behavior of the channel. We chose to mutate to Lys, a large and basic amino acid, with the hope that it would produce more robust perturbations in protein–protein interactions compared with the previous alanine scan. In addition, the hydrophilic nature of the Lys sidechain should discriminate between protein–protein and protein–solvent interfaces where the solvent is water. We anticipated that residues whose mutation altered toxin binding would show helical distributions and be well tolerated with respect to perturbations in gating because, in the absence of the toxin, they interact with water. The positions where mutations do not influence toxin binding might be predicted to strongly influence gating if they interact with other elements within the channel protein.
Perturbation of the Toxin–Channel Interaction with Lysine Substitutions
First, we tested whether Lys substitutions at each of the 20 residues perturb Hanatoxin binding to the channel. Wild-type and mutant channels were expressed in Xenopus oocytes and the inhibition by Hanatoxin was examined using two-electrode voltage-clamp recording techniques. We measured the fractional inhibition of tail currents carried by 50 mM Rb+ at various toxin concentrations. Fig. 2 shows the inhibitory effect of 1 μM Hanatoxin on the wild-type (left) and two exemplary mutant channels, T276K (middle) and F274K (right). Hanatoxin greatly inhibited the wild-type channel, but only moderately inhibited T276K, and was almost without effect on F274K, as shown by the families of current records (Fig. 2 A) and the tail current voltage-activation relations (Fig. 2 B). To quantitatively evaluate the concentration dependence for toxin occupancy of the channel, we calculated the fraction of uninhibited current for various strength depolarizations (Fig. 2 C). The fraction of uninhibited current (I/I0) measured in the plateau phase at negative voltages (In/In0), where toxin-bound channels do not open, approximates the fraction of unbound channels and, therefore, was used to determine the Kd for toxin binding (see materials and methods). Fig. 3 A shows the concentration dependence for fractional inhibition of the wild-type and seven mutant channels (I273K, F274K, L275K, T276K, S278K, V282K, and F285K). Fitting of a four site model to the data yielded Kd values of (in μM): 0.21 for the wild type; 0.76 for I273K; 2.1 for T276K; 2.2 for V282K; 2.4 for S278K; 6.2 for L275K; 20.8 for F285K; and 41.5 for F274K. Fig. 3 B summarizes the changes in the Kd values and in the free energy of toxin binding caused by Lys substitutions for all 20 residues. When compared with the previous results with Ala substitutions (Li-Smerin and Swartz 2000), mutations to Lys identified five additional positions where mutations produce large effects on toxin binding. In addition to the previously described mutations (F274K and E277K), substitution with Lys at L275, T276, S278, V282, and F285 resulted in ΔΔGT values >1 kcal mol−1. In contrast, Lys substitution at I273 caused a smaller change than did mutation to Ala (0.8 kcal mol−1 for Lys vs. 1.6 kcal mol−1 for Ala). These results are consistent with previous work identifying the COOH terminus of S3 as a major component of the Hanatoxin receptor and importantly identify additional residues whose mutation influences toxin binding.
Having these newly identified positions in hand, we examined whether the distribution of toxin binding perturbations was consistent with the α-helical structure using Fourier transform methods and both helical wheel and net diagrams. For an ideal amphipathic α-helix, Fourier transform analysis should reveal a predominant peak with an angular frequency of 100° (360°/3.6 residues per turn of the helix). The helical periodicity can be evaluated quantitatively by integrating the power spectrum and calculating the α-PI (see materials and methods), with values ≥2 indicating α-helical structure (Rees et al. 1989b). The results of Fourier transform analysis of ΔΔGT are presented in Fig. 4. The power spectrum of a stretch of 18 residues beginning with V271 and ending with V288 (Fig. 4 B) exhibits a rather complex profile with multiple peaks at different angular frequencies. Although there is a significant peak at 93°, the α-PI value for this spectrum is only 1.4. In contrast, the power spectrum for a shorter stretch (eight residues long), beginning with T272 and ending with N279, reveals a primary peak at 111°, with a rather strong α-PI value of 2.7 (Fig. 4 B, right). When the mutation-induced perturbations are plotted on helical wheel or net diagrams (Fig. 4 C), it is striking that with only one exception the “hot residues” (Fig. 4 C, ΔΔGT > 1 kcal mol−1, black and gray shading) are located on one side of the helical diagram. To quantitatively evaluate the clustering of ΔΔGT values, we plotted individual ΔΔGT values as vectors along with the sum vector. The sum vector (5.5 kcal mol−1) is larger than any individual ΔΔGT value and, therefore, highlights a well-defined face where mutations have the largest effects on toxin binding. These results with perturbations in toxin binding argue that the COOH terminus of S3 adopts an α-helical secondary structure and define a face of the helix that is exposed to an aqueous environment, at least in the conformation of the channel protein to which the toxin binds.
Perturbation of Voltage-Dependent Gating with Lysine Substitutions
We next asked whether the effects of Lys substitutions on channel gating display helical distributions, as previously shown for this region with alanine-scanning mutagenesis (Li-Smerin et al. 2000). Voltage-activation relations were obtained for the wild-type and the 20 Lys mutants described above using two-electrode voltage-clamping recording techniques. The amplitude of tail currents carried by 50 mM Rb+ was measured for various strength depolarizations and single Boltzmann functions fit to the resulting relations (see material and methods). The parameters derived from these fits were then used to evaluate the difference in free energy between the closed and open states of the channel at 0 mV (ΔG0) for the wild-type and each mutant channel. The mutation-induced perturbation in gating was calculated as the difference in ΔG0 between the wild-type and the mutant channels (ΔΔG0).
Fig. 5 A shows families of current records for the wild-type (middle) and two exemplary mutant channels, F285K (left) and F274K (right). Compared with the wild-type channel, activation of F285K was shifted to more negative voltages, whereas activation of F274K was shifted to more positive voltages. Voltage-activation relations for the same channels, along with single Boltzmann fits, are presented in Fig. 5 B. Mutation-induced perturbation energies (ΔΔG0) are shown for all 20 mutants in Fig. 5 C. The smallest perturbations observed were 0.2 kcal mol−1 for S281K and N287K, and the largest was 2.4 kcal mol−1 for L275K. In general, the perturbations observed for mutations to Lys were significantly larger than those previously observed for mutations to Ala. For the alanine scan there were 6/20 residues with ΔΔG0 ≥ 1 kcal mol−1 (Li-Smerin et al. 2000), whereas for the lysine-scan, 10/20 residues had ΔΔG0 ≥ 1 kcal mol−1. The gating properties derived from Boltzmann fits of the voltage-activation relations along with perturbation energies for all 20 mutant channels are summarized in Table.
We analyzed the periodicity of gating perturbations with the same approach used above for the toxin–channel interaction. The power spectrum of an 18-residue long stretch (V271-V288) shows a predominant peak at 103° angular frequency and a reasonably strong α-PI value of 1.9 (Fig. 6 B, left). Similar to the results for the toxin–channel interaction, we see multiple peaks in this spectrum of gating perturbations. However, in this case, the peak in the helical frequency is more prominent than the corresponding peak in the toxin–channel interaction perturbations. When the region analyzed was restricted to only 10 residues (overlapping with the short eight-residue stretch used for toxin perturbation analysis), we again obtained a very strong helical profile, as shown in Fig. 6 B (right). We see a primary peak at 106° angular frequency and a very large α-PI value of 3.1. Fig. 6 C shows the distribution of the gating perturbation on a helical wheel (left) or in a net diagram (right). Residues whose mutation resulted in ΔΔG0 values ≥1 kcal mol−1 (shaded letters) tend to cluster together, although they occupy a large portion of the wheel compared with the perturbations of the toxin–channel interaction. Individual |ΔΔG0| values were plotted as vectors in Fig. 6 C. The sum vector had a value (4.6 kcal mol−1) larger than any individual |ΔΔG0| value. Taken together, the results obtained from perturbations in both the toxin–channel interaction and in gating provide compelling evidence for the presence of an α-helix in the COOH terminus of S3.
The goal of this paper was to examine the secondary structure of the COOH terminus of S3 and the S3–S4 linker, a region of voltage-gated ion channels previously shown to be an important component of the receptor for gating modifier toxins (Rogers et al. 1996; Swartz and MacKinnon 1997b; Li-Smerin and Swartz 1998, Li-Smerin and Swartz 2000; Winterfield and Swartz 2000). We lysine-scanned 20 residues located in this region of the drk1 K+ channel and then examined the changes in Hanatoxin binding affinity and channel gating, looking for periodicity characteristic of an α-helix. The strongest indication for helical secondary structure came from analyzing the perturbations in toxin binding and gating for a stretch of 8–10 residues in length that is located within the COOH terminus of S3, as defined by hydrophobicity analysis (Fig. 4 and Fig. 6). The analysis of gating perturbations for longer stretches (extended towards S4) was also consistent with a helical structure (Fig. 6), however, the equivalent analysis of perturbations in toxin binding was less clear (Fig. 4). Our results strongly support the existence of an α-helix in the COOH terminus of S3 that is at least ∼10 residues long.
What can we say about the environment surrounding this helix? The results from the perturbations in the toxin–channel interaction define a face of a helix in the COOH terminus of S3 that interacts with the toxin. The effects of Hanatoxin on gating currents in the drk1 K+ channel suggest that the toxin stabilizes the resting conformation of the voltage-sensing domains (Swartz 1999). This observation suggests that the toxin-interacting face of the helix in the COOH terminus of S3 is probably exposed to an aqueous environment when the voltage-sensing domains are in a resting conformation that predominates at negative membrane voltages. One of the motivations for substituting with Lys was to explicitly identify water-exposed surfaces within the protein with the assumption that hydrophilic substitutions would be well tolerated at protein–water interfaces. Therefore, we anticipated that the mutations causing little or no perturbation in gating (well tolerated) would cluster together with large perturbations in toxin binding. The helical wheel diagrams in Fig. 7 A show the distributions for the two types of perturbations. Although the distribution in the perturbations in toxin binding and gating are partially distinct, as revealed by ∼50° difference between the sum vectors, there is also substantial overlap. One possible explanation for this overlap between the two types of perturbations is that Lys substitutions may perturb gating both by local disruption of protein–protein interactions and by more distant electrostatic mechanisms. However, the contribution of electrostatic effects cannot entirely explain the discrepancy because the alanine scan shows very similar shifts in the direction of the sum vectors (∼60°) for the perturbations in toxin binding and gating (Fig. 7 B). An alternate explanation is that the region in question undergoes a structural rearrangement during activation, and that this movement causes a change in the solvent exposure of some residues. The voltage-dependent changes in conformation and/or environments surrounding fluorophores introduced at positions near S4 have been previously studied rather extensively (Mannuzzu et al. 1996; Cha and Bezanilla 1997, Cha and Bezanilla 1998; Cha et al. 1998,Cha et al. 1999; Gandhi et al. 2000). However, the position closest to the COOH terminus of S3 where fluorophores have been introduced is S346 in the Shaker K+ channel. Based on our sequence alignments, this position is in a region of the S3–S4 linker that is not present in the drk1 K+ channel (see last paragraph of discussion and Fig. 8). Thus, as far as we are aware, no studies have addressed whether the COOH terminus of S3 moves during activation.
Perhaps one of the most interesting aspects of the COOH terminus of S3 is that it forms a canonical binding site for gating modifier toxins in all types of voltage-gated cation channels (Rogers et al. 1996; Swartz and MacKinnon 1997b; Li-Smerin and Swartz 1998, Li-Smerin and Swartz 2000; Winterfield and Swartz 2000). Can we extrapolate from the present work in K+ channels to voltage-gated Ca2+ and Na+ channels? Examination of the S3 sequences from different voltage-gated channels provides some insight (Fig. 8). First, the residues in the NH2 terminus of S3 that are most conserved among K+ channels are also well conserved in the Ca2+ and Na+ channels (Fig. 8, bold letters). Second, although the sequence of the COOH terminus of S3 is less conserved between voltage-gated channels, it is a common target for gating modifier toxins, some of which have been shown to cross-react between the three voltage-gated ion channel types (Chuang et al. 1998; Li-Smerin and Swartz 1998). Third, Glu residues (Fig. 8, highlighted in red) that are critical determinants of gating modifier toxin receptors in voltage-gated Ca2+ channels (Winterfield and Swartz 2000), Na+ channels (Rogers et al. 1996) and K+ channels are located at equivalent positions in S3 (Fig. 8). All of these points suggest that the 3-D structure of this region has been conserved between voltage-gated K+, Ca2+, and Na+ channels. Thus, it seems likely that the COOH terminus of S3 is helical and has significant water exposure in all voltage-gated channels.
Previous scanning mutagenesis results uncovered evidence for a helical structure in the NH2-terminal two thirds of S3 and suggest that, unlike the COOH-terminal part, it is surrounded by lipid and protein (Hong and Miller 2000; Li-Smerin et al. 2000). Thus, while most of S3 seems to adopt an α-helical structure, the environments surrounding the NH2- and COOH-terminal portions are quite different. It is interesting that there is a conserved Pro residue located near the junction between the two parts of S3 in voltage-gated K+ channels (Fig. 1 B and 8). The helix destabilizing effects of Pro residues are well established (O'Neil and DeGrado 1990; MacArthur and Thornton 1991; Blaber et al. 1993; Monne et al. 1999). Thus, we speculate that S3 is composed of two α-helices (termed S3N and S3C in Fig. 8) separated by a short nonhelical stretch or kink near the conserved Pro residue. It is noteworthy that the Pro residue that is conserved in K+ channels is conspicuously absent in the voltage-gated Ca2+ or Na+ channels. Instead, these channels most commonly have Gly or Ser residues at positions that are equivalent or adjacent to that of the Pro in K+ channels (Fig. 8, boxed positions). Like Pro, the helix destabilizing effects of Gly are well known (O'Neil and DeGrado 1990; MacArthur and Thornton 1991; Blaber et al. 1993; Monne et al. 1999). Although the helical propensity of the Ser residue in soluble proteins is not particularly low (Blaber et al. 1993), in membrane proteins Ser has a tendency to hydrogen bond with backbone carbonyls, thereby destabilizing the helix (Gray and Matthews 1984). We examined the crystal structures of six integral membrane proteins and tabulated the residues typically found at kinks in TM helices (Table). Gly was the most common residue found (33%) at the 36 kinks examined, followed by Pro and Ser, each of which was observed at a frequency of 22%. These observations fit nicely with the idea that there is a kink that separates the two helical parts of S3 not only in voltage-gated K+ channels, but also in voltage-gated Na+ and Ca2+ channels.
Several studies point to the existence of a water-filled vestibule surrounding the S4 helix (Larsson et al. 1996; Mannuzzu et al. 1996; Yang et al. 1996; Cha and Bezanilla 1997, Cha and Bezanilla 1998; Sorensen et al. 2000). In the Shaker K+ channel, the S3–S4 linker has been proposed to contribute to forming this aqueous vestibule (Sorensen et al. 2000). Our results in the drk1 K+ channel suggest that the helix in S3C extends from around V271 to near K280 (Fig. 8, underlined sequence). This leaves only a relatively short S3–S4 linker in this K+ channel before the presumed beginning of the S4 segment. It is interesting that, according to our designations, the S3–S4 linker is virtually absent in some K+ channels (e.g., Kv4 family) and several repeats of both voltage-gated Ca2+ and Na+ channels. In contrast, the Shaker K+ channel stands out among the voltage-gated ion channels as having an unusually long S3–S4 linker. Importantly, the region within the S3–S4 linker of the Shaker K+ channel where both thiosulfonates and fluorescence probes have been introduced is not present in most voltage-gated ion channels. Thus, whereas the S3–S4 linker may contribute to a vestibule in Shaker K+ channels, it seems unlikely that it makes a fundamental contribution to the aqueous vestibule surrounding the S4 helix in most voltage-gated ion channels. In contrast, S3C displays significant sequence conservation, is at least partially exposed to water, and is the target of gating modifier toxins. In addition, deletions in the S3–S4 linker of Shaker that do not extend into the region corresponding to S3C generally cause only subtle changes in gating behavior (Mathur et al. 1997; Gonzalez et al. 2000). Although deletions that extend only a few residues into S3C do result in functional channels, they display more substantial gating perturbations (Gonzalez et al. 2000). Thus, this region of S3 may serve an important role as a component of the aqueous vestibule or crevice surrounding the S4 helix.
We thank David Hackos for providing the Fourier transformation software and critical reading of the manuscript, and J. Nagle and D. Kauffman from the NINDS DNA sequencing facility for DNA sequencing.
Address correspondence to Kenton J. Swartz, Molecular Physiology and Biophysics Unit, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Building 36, Room 2C19, 36 Convent Drive, MSC 4066, Bethesda, MD 20892. Fax: (301) 435-5666; E-mail: firstname.lastname@example.org
Abbreviations used in this paper: α-PI, α-periodicity index; TM, transmembrane.