The signature sequence of voltage-gated potassium channels projects into the external vestibule.

A highly conserved motif, GYGD, contributes to the formation of the ion selectivity filter in voltage-gated K+ channels and is thought to interact with the scorpion toxin residue, Lys27. By probing the pore of the Kv1.3 channel with synthetic kaliotoxin-Lys27 mutants, each containing a non-natural lysine analog of a different length, and using mutant cycle analysis, we determined the spatial locations of Tyr400 and Asp402 in the GYGD motif, relative to His404 located at the base of the outer vestibule. Our data indicate that the terminal amines of the shorter Lys27 analogs lie close to His404 and to Asp402, while Lys27 itself interacts with Tyr400. Based on these data, we developed a molecular model of this region of the channel. The junction between the outer vestibule and the pore is defined by a ring (∼8-9-Å diameter) formed from alternating Asp402 and His404 residues. Tyr400 lies 4-6 Å deeper into the pore, and its interaction with kaliotoxin-Lys27 is in competition with K+ ions. Studies with dimeric Kv1.3 constructs suggest that two Tyr400 residues in the tetramer are sufficient to bind K+ ions. Thus, at least part of the K+ channel signature sequence extends into a shallow trough at the center of a wide external vestibule.

Voltage-gated K ϩ channels play diverse physiological roles in both excitable and non-excitable tissues and are attractive therapeutic targets for various diseases (1). These channels contain six putative transmembrane segments (S1 to S6) and a membrane-associated loop between S5 and S6 (P-region). The ion conduction pathway, bounded at its external and internal entrances by wide vestibules, is formed in large part by the P-region and the C-terminal half of S6 (2)(3)(4)(5)(6)(7)(8). The P-region of all cloned voltage-gated K ϩ channels contains a highly conserved G(Y/F)GD motif (1), which is thought to constitute an essential part of the K ϩ ion selectivity filter (9 -12).
The external vestibule is the receptor site for several scorpion peptide toxins that are potent blockers of K ϩ currents (7,(12)(13)(14)(15)(16). In an earlier study, we used four structurally related scorpion toxins as molecular probes to map the topology of this region in Kv1.3 (7), a voltage-gated K ϩ channel that plays a vital role in modulating lymphocyte activation (17)(18)(19). This was achieved by determining the three-dimensional structures of the toxins and by identifying multiple pairs of interacting toxin and Kv1.3 residues (7). Knowing the disposition of these interacting toxin residues from their NMR structures, we were able to deduce the architecture of a 30-Å-wide and ϳ8-Å-deep vestibule at the outer entrance to the Kv1.3 pore (7). The scorpion toxins interact with all four subunits in the Kv1.3 tetramer, and the conserved toxin residue, Lys 27 , protrudes into the pore, possibly interacting with residues in the signature sequence (7). By replacing Lys 27 in kaliotoxin (KTX) 1 with positively charged non-natural lysine analogs of varying chain lengths, we were able to show that the terminal amine at toxin position 27, positioned at distances anywhere from 2.2 to 7.7 Å from C ␣ , can interact suitably with the pore (7).
In the present study, we have used this series of KTX-Lys 27 mutants as a caliper to estimate the vertical distance between His 404 in the external vestibule, and Tyr 400 and Asp 402 , two residues contained within the signature sequence of Kv1.3. Using site-specific mutagenesis coupled with thermodynamic mutant cycle analysis (14,20), we assessed the interaction strength and estimated the distances between the terminal amines in each of these Lys 27 analogs and each of the three channel residues. Our data suggest that Asp 402 and Tyr 400 lie in a shallow depression at the center of a wide saucer-shaped outer vestibule. Based on these experimental data, we have developed a modified molecular model of the outer mouth of the Kv1.3 pore.

EXPERIMENTAL PROCEDURES
KTX Mutants-The five KTX-Lys 27 mutants have been described previously (7). Four of these contain positively charged non-natural lysine analogs of varying side chain lengths (DAP, 2.5 Å from C ␣ ; DAB, 3.8 Å; Orn, 5.0 Å; ThLys 7.7 Å); the neutral non-natural amino acid, Nle 27 (5.0 Å) replaced Lys 27 (6.3 Å) in the fifth mutant. Another KTX mutant used in this study, R24D, was also described earlier (7). Three batches of synthetic KTX (7) were obtained from Dr. James Boyd, Pfizer Central Research, Groton, CT. The first batch was used for all the His 404 mutant experiments; the second and third, slightly more potent batches were used for studies on the WT-D402N and WT-Y400V dimers and for the protonation experiments.
Biophysical Characterization-Kv1.3 cRNA (WT, mutant or dimer) were transcribed in vitro using a kit purchased from Ambion Inc. (Austin, TX) and injected into oocytes (Xenopus laevis purchased from Nasco, Fort Atkinson, WI) as described (21). K ϩ currents were measured at room temperature using the two-electrode voltage-clamp tech-nique (21) and data analyzed using pClamp software (version 5.5.1, Axon Instruments, Burlingame, CA). Whole oocytes were held at Ϫ100 mV and depolarized to ϩ40 mV over 500 ms; time between pulses was 30 s. Capacitative and leak currents were subtracted prior to analysis using the P/4 procedure. The oocyte bathing solution (ND96) contained (in mM): 96 NaCl, 2 KCl, 1.8 CaCl 2 , 1 MgCl 2 , 5 HEPES, 0.1% bovine serum albumin, pH 7.6. For K ϩ competition experiments, NaCl was replaced with KCl. Stock solutions of toxins were made up in ND96 containing 0.1% bovine serum albumin, pH 7.6, and stored at Ϫ20°C. Protonation experiments were performed at pH 6.8 (control) and pH 6.0 (protonated form) based on earlier studies (7).
Scorpion toxins bind to Kv1.3 with 1:1 stoichiometry (7). The dissociation constant (K d ) for toxin block was calculated as described earlier (7) using the formula: Thermodynamic Mutant Cycle Analysis-Mutant cycle analysis is a simple and powerful tool to study the strength of interaction between any pair of residues in proteins (14,20,22). For each mutant cycle, we measured the potency (K d values) of KTX and each of its mutants on WT-Kv1.3 and each of the channel mutants. The change in coupling energy, ⌬⌬G (20), for a given pair of interacting toxin-channel residues and their mutants was calculated using the formula, ⌬⌬G ϭ kTln⍀, where ⍀ (14) is a dimensionless value given by the formula: ). An ⍀ value of 1 indicates no interaction, while values deviating from 1 suggest stronger energetic contacts (14).
The distance between each pair of residues was estimated using the criteria of Schreiber and Fersht (20) who demonstrated a strong correlation between ⌬⌬G values obtained by mutant cycle analysis and inter-residue distances obtained from the crystal structures of Barnase and Barstar. They showed that a ⌬⌬G value of Ն0.5 kcal mol Ϫ1 corresponded to an inter-residue distance of Յ5 Å (2 error), and this value declined as ⌬⌬G increased. Note that although high ⌬⌬G values indicate tight interactions, failure to detect an energetically coupled interaction does not exclude the possibility of close proximity, since residues that are physically close may be energetically "silent." This is especially true when random scanning mutagenesis is used to identify coupling (12,22).
In all our experiments we have assumed that KTX and all its mutants sit in the vestibule with a similar geometry, with the side chains of residues at position 27 interacting with the pore in their fully extended conformation. For the purpose of comparison, the toxins containing positively charged residues at position 27 (DAP, DAB, Orn, Lys, ThLys) were treated as "WT" in the mutant cycle analyses and evaluated against a single mutant toxin containing the neutral residue, Nle.
Modeling the Kv1.3 Pore-To create a model of the Kv1.3 pore, Phe 425 , Lys 427 , Thr 449 , Gly 452 , Phe 453 , and Trp 454 in the model of the Shaker channel (3) were mutated to Gly 380 , Asn 382 , His 404 , Thr 407 , Ile 408 , and Gly 409 , respectively. Assuming 4-fold symmetry, the distances we experimentally estimated between residues Gly 380 , Asp 386 , His 404 , Asp 402 , and Tyr 400 (Ref. 7 and this paper) were included as constraints during optimization of the model (for distances and to correct for any bad Van der Waals contacts) with SYBYL 6.2 (Tripos Inc., St. Louis, MO).

RESULTS AND DISCUSSION
Terminal Amines of Shorter KTX-Lys 27 Analogs Are Close to His 404 in Kv1.3-His 404 , located at the entrance to the pore (7), forms the external binding site for tetraethylammonium (23,24). Using mutant cycle analyses, we estimated the strength of interaction between His 404 and each of terminal amines at KTX position 27. Examples of such experiments for a short (DAP 27 ) and long (Lys 27 ) residue are shown in Fig. 1A (K d values in Table I). Replacing His 404 with the hydrophobic valine (H404V), strongly perturbed the interaction of the short analog with the pore (⍀ ϭ 150), but not that of the longer analog (⍀ ϭ 3.4; Fig. 1A). These results suggest that His 404 and KTX-DAP lie close to one another.
A more extensive mutagenesis study supported this conclusion (Fig. 1B (Fig. 1B). These results suggest that His 404 might hydrogen-bond with the terminal amines of the shorter KTX-Lys 27 analogs, since this interaction is maintained by polar, but not hydrophobic, substitutions.
If the short analogs are close to His 404 as these data suggest, protonation of His 404 , brought about by changing the external pH from 6.8 to 6.0 (7), would be expected to reduce their potency via electrostatic repulsion of the terminal amine at position 27; such an effect should be less pronounced for the longer toxin-lysine 27 analogs that do not interact energetically with His 404 . To test this idea, we performed mutant cycle experiments, measuring the K d values for each of the KTX-Lys 27 analogs on the WT-Kv1.3 channel at either pH 6.8 or 6.0. As predicted, protonation of His 404 disrupted its interaction with the shorter analogs to a greater extent than the longer analogs (Fig. 1B, Table I).
The ⌬⌬G values for each of the Kv1.3-H404V-KTX-X27Nle mutant cycles (where X ϭ DAP, DAB, Orn, Lys, or ThLys), derived from the ⍀ values in Fig. 1, A and B, were plotted against the length of the side chain at toxin residue 27 (Fig.  1C). Based on the criterion of Schreiber and Fersht (20), we assume that a coupling energy value of Ն0.6 kcal mol Ϫ1 corresponds to an inter-residue distance of Յ5 Å for a given pair of interacting toxin-channel residues. The ⌬⌬G values were highest for the shorter analogs, and these values diminished as the length of the side chain at toxin-position 27 increased (Fig. 1C). This molecular yardstick places His 404 , within 5 Å of the terminal amines in the shorter KTX-Lys 27 analogs.
Asp 402 Is Close to the External Vestibule-The highly conserved aspartate in the signature sequence of Kv channels, Asp 402 in Kv1.3, has long been an attractive candidate for a salt-bridge interaction with Lys 27 . Unfortunately, it has not been possible to test this idea, since mutations of Asp 402 result in non-functional channels. We circumvented this problem by using a dimeric construct consisting of one wild-type and one mutant (D402N) subunit, which yielded a functional channel in which the overall negative charge at position 402 is halved.
The WT-D402N dimer ( Fig. 2A, center) inactivates more rapidly than the WT Kv1.3 channel ( Fig. 2A, left). Dose-response curves demonstrate that all five toxins containing positively charged residues at position 27 blocked the WT Kv1.3 channel with almost equal potency (Fig. 2B, left; Table I), while the WT-D402N dimer was significantly less sensitive to DAP 27 and DAB 27 than to the longer analogs (Fig. 2B, middle; Table I). The neutral KTX-Nle 27 toxin mutant blocked WT Kv1.3 and the dimer with substantially lower potency than the positively charged KTX-Lys 27 forms (Table I) Tyr 400 Forms an Energetic Contact with KTX-Lys 27 , and This Interaction Is Dependent on the K ϩ Ion Concentration in the Pore-Tyr 400 in the signature sequence is highly conserved among all cloned K ϩ channels, regardless of their gating behavior, and has been suggested to play a role in K ϩ ion selectivity (reviewed in Refs. 1, 6, 8, and 9). Mutant cycle experiments were used to assess the interaction between each of the KTX-Lys 27 analogs and Tyr 400 in the K ϩ channel signature sequence. For this purpose, we generated a Kv1.3 dimer containing one WT subunit and a second Y400V domain; the same substitution at the homologous position in Shaker (Y445V) has been reported to produce a K ϩ -selective channel (10). The WT-Y400V dimer inactivated rapidly ( Fig. 2A, right) and was completely blocked, in a dose-dependent manner, by all five toxins containing positively charged residues at position 27 (Fig. 2B,  right). Mutant cycle analysis demonstrated that Tyr 400 lies within 5 Å of the terminal amine in KTX-Lys 27 (Fig. 1C), a result consistent with a recent report that placed Lys 27 of agitoxin-2 in close proximity to Tyr 445 in Shaker (12). The decreased blocking potency of KTX (Fig. 2B, right; Table I) on the dimer (which contains only two tyrosines) also suggests that optimal binding of KTX requires more than two Tyr 400 residues in the tetramer.
Earlier work on the Shaker channel suggested that Lys 27 in agitoxin-2 interacted with a residue at or close to a K ϩ ion binding site within the pore (12); occupation of this site by a K ϩ ion destabilized the toxin interaction with the channel via electrostatic repulsion of Lys 27 (12). Similarly, K ϩ competi-tively inhibited the interaction between Lys 27 in charbydotoxin and residues in the pore of the "maxi" calcium-activated K ϩ channel (25,26). Since Kv1.3-Tyr 400 is close to KTX-Lys 27 , it is a good candidate for such a K ϩ ion binding site. We examined the effect of changing the external K ϩ concentration from 2 to 100 mM on the blocking potency of KTX-Lys 27 or KTX-Nle 27 on either the WT channel or the WT-Y400V dimer. If Tyr 400 is a K ϩ binding site, its interaction with KTX-Lys 27 , but not KTX-Nle 27 , should be disrupted as the K ϩ ion concentration in the pore increases. Results with the WT-Y400V dimer will depend on whether K ϩ binds to all four residues in the tetramer or whether two will suffice. For example, if two tyrosines are sufficient for binding K ϩ , then increasing the K ϩ concentration in the pore should competitively inhibit KTX-Lys 27 binding to the dimer and consequently produce a high ⌬⌬G value by mutant cycle analysis. Alternatively, if four tyrosines participate in K ϩ binding, then halving the number (as in our dimer), would diminish the channel's affinity for K ϩ ions and reduce the ability of external K ϩ to compete with KTX-Lys 27 for Tyr 400 . The mutant cycle experiments presented in Fig. 2C favor the first alternative.
As shown in Fig. 2C (left), increasing external K ϩ concentration reduced by 6-fold, the potency of KTX-Lys 27 on the WT Kv1.3 channel, and this effect was even more pronounced (38-   (Fig. 2C, right) was not dependent on the concentration of K ϩ ions in the pore (⌬⌬G ϭ 0.1 kcal mol Ϫ1 ). These results support the notion that two Tyr 400 residues in the tetramer are sufficient to bind K ϩ ions.
Modifying the Model of the Kv1.3 Pore-In our earlier model of Kv1.3 (Fig. 3 of Ref. 7) the entrance to the ion conduction pathway was formed by an alternating ring of Met 403 and His 404 residues. The GYGD motif was positioned vertically below this ring, lining the cylinder constituting the ion conduction pathway, with Asp 402 and Tyr 400 located 4 -5 Å and 10 -11 Å below His 404 , respectively. We have modified this model based on the additional structural constraints imposed by our new experimental data. In the present model, Asp 402 has been pulled closer to the external vestibule so that the carboxyl group lies in the same horizontal plane as His 404 , resulting in a ring (8 -9-Å diameter) of four alternating His 404 residues and Asp 402 residues (Fig. 3A). The sulfur atom of Met 403 is located 3 Å farther away from the central axis in this modified model, while the four Tyr 400 residues lie 4 -6 Å below both Asp 402 and His 404 with their aromatic side chains protruding into the pore (Fig. 3B). In this model, the external vestibule has a shallow saucer shape with the selectivity filter forming a ϳ5-Å central depression.
We docked KTX (Fig. 3) into the modified vestibule to con-firm the feasibility of the experimentally determined interactions and to predict new ones. Docking was achieved by guiding the KTX-Lys 27 residue into the center of the pore and then rotating the toxin around the central axis of the pore until Arg 24 was aligned with Asp 386 . This docking configuration (Fig. 3A) does not violate any of the previously reported toxin-channel interactions, including Arg 24 -Asp 386 , Asn 30 -Asp 386 , Ser 11 -His 404 , Phe 25 -His 404 , Met 29 -His 404 , and Thr 36 -His 404 (7,12,14,16). In summary, using a series of KTX-Lys 27 mutants as molecular yardsticks, we have determined the spatial location of two residues in the K ϩ channel signature motif (Asp 402 and Tyr 400 ), relative to a third residue, His 404 , in the external vestibule of Kv1.3 (7). A molecular model based on these and earlier (7,12,14,16) experimental data reveals the existence of a 30-Å wide outer vestibule, at the center of which lies a ring, consisting of four alternating Asp 402 residues and His 404 residues, which defines the boundary between the external vestibule and the pore. Positioned 4 -6 Å deeper into the pore is Tyr 400 , which interacts with KTX-Lys 27 and binds K ϩ ions, possibly via cation/pi-electron interactions (27). Our model has considerable experimental support, both from our results (Ref. 7 and this paper) and those of others (12,14,16). It provides a structural basis for understanding the biophysical properties involving the external vestibule and pore, including ion permeation and C-type inactivation. The model might also facilitate the design of novel non-peptide blockers of Kv1.3 for use as immunosuppressants; these agents could mimic the interactive surface of the toxins, in addition to taking advantage of the 4-fold symmetry of the channel.