Structural Conservation of the Pores of Calcium-activated and Voltage-gated Potassium Channels Determined by a Sea Anemone Toxin*

The structurally defined sea anemone peptide toxins ShK and BgK potently block the intermediate conductance, Ca2+-activated potassium channel IKCa1, a well recognized therapeutic target present in erythrocytes, human T-lymphocytes, and the colon. The well characterized voltage-gatedKv1.3 channel in human T-lymphocytes is also blocked by both peptides, although ShK has a ∼1,000-fold greater affinity forKv1.3 than IKCa1. To gain insight into the architecture of the toxin receptor in IKCa1, we used alanine-scanning in combination with mutant cycle analyses to map the ShK-IKCa1 interface, and compared it with the ShK-Kv1.3 interaction surface. ShK uses the same five core residues, all clustered around the critical Lys22, to interact with IKCa1 and Kv1.3, although it relies on a larger number of contacts to stabilize its weaker interactions with IKCa1 than with Kv1.3. The toxin binds to IKCa1 in a region corresponding to the external vestibule of Kv1.3, and the turret and outer pore of the structurally defined bacterial potassium channel, KcsA. Based on the NMR structure of ShK, we deduce the toxin receptor inIKCa1 to have x-y dimensions of ∼22 Å, a diameter of ∼31 Å, and a depth of ∼8 Å; we estimate that the ion selectivity lies ∼13 Å below the outer lip of the toxin receptor. These dimensions are in good agreement with those of the KcsA channel determined from its crystal structure, and the inferred structure of Kv1.3 based on mapping with scorpion toxins. Thus, these distantly related channels exhibit architectural similarities in the outer pore region. This information could facilitate development of specific and potent modulators of the therapeutically important IKCa1 channel.

The use of peptide toxins to obtain insight into the topology of the external vestibule of potassium channels has a long and successful history. Structurally defined peptides from scorpion venom and sea anemone have been used as molecular yardsticks to gauge the dimensions and shape of the external vestibules of the voltage-gated Shaker and Kv1.3 channels (17)(18)(19)(20)(21)(22)(23). The deduced dimensions are in good agreement with the recently published structure of the outer pore region of the bacterial potassium channel, KcsA, based on crystallographic data (24,25). Therefore, the use of peptide toxins as mapping tools, set in the context of the known KcsA crystal structure, can facilitate the architectural mapping of the outer pore regions of pharmacologically important mammalian potassium channels in the absence of direct structural data for these channels.
Although IKCa1 is only distantly related to the Shaker and Kv1.3 voltage-gated channels, it is potently blocked by some of the same scorpion and sea anemone peptides that inhibit these channels. It might therefore be feasible to use the peptidemapping approach to gain insight into the dimensions and shape of the toxin receptor on IKCa1 and to compare this topology with the external vestibules of Kv1. 3 and KcsA. For this purpose, we have used the structurally defined 35-amino acid peptide toxin, ShK, 1 from the sea anemone Stichodactyla helianthus. We used the alanine scanning method, coupled with mutant cycle analyses, to map the interactive surface between ShK and IKCa1 and compared this with the ShK: Kv1.3 interface. Our studies indicate that ShK binds to IKCa1 in an external vestibule that is architecturally similar to that of Shaker, Kv1.3, and KcsA, although ShK uses a significantly wider surface to interact with IKCa1 compared with its interaction with Kv1. 3. Such structural differences might be exploited to guide the design of novel peptides that specifically target IKCa1.

MATERIALS AND METHODS
Peptide Synthesis-Fmoc-amino acid derivatives were obtained from Bachem A.G. (CH-4416 Bubendorf, Switzerland). Solid-phase assembly was initiated with an Fmoc-Cys(Trt)-2-chlorotrityl resin to minimize potential racemization of the C-terminal Cys residue. Automated stepwise assembly was carried out entirely on an ABI-431A peptide synthesizer (Applied Biosystems, Foster City, CA). The ShK analogues were solubilized, oxidized, and purified by reverse phase-high pressure liquid chromatography using the method described previously (23,26), and high pressure liquid chromatography-pure fractions were pooled and lyophilized. The structure and purity of the peptides were confirmed by reverse phase-high pressure liquid chromatography, amino acid analysis, and electrospray ionization-mass spectroscopy analysis. Samples were weighed and adjusted to account for peptide content before bioassay.
Expression and Electrophysiological Analysis-The human wild-type IKCa1 and IKCa1-Lys 239 constructs were linearized with NotI, the Kv1.3-Val 404 mutant with EcoRI, and these constructs were transcribed in vitro (20,28). The cRNA along with a marker dye was injected into rat basophilic leukemia cells as described previously (13,28). After 2-6 h, dye-containing cells with specific currents could be characterized using the patch-clamp method. Cell lines stably expressing Kv1.3-wild type (27) were trypsinized and plated onto glass coverslips at least 3 h before measurement. All cells were measured in the whole-cell configuration and bathed in mammalian Ringer solution with 0.1% bovine serum albumin (Sigma) containing (in mM): 160 NaCl, 4.5 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 HEPES, adjusted to pH 7.4 with NaOH, with an osmolarity of 290 -320 mOsm. In the K ϩ -Ringer solution, NaCl was replaced with KCl. A simple syringe-driven perfusion system was used to exchange the bath solutions in the recording chamber. The internal pipette solution for the Kv1. Free Energy Difference of Binding [⌬F]-The free energy difference of binding was calculated as ⌬F ϭ RT ln(K d ShK-analogue/K d ShK-wild type), where R ϭ 1.987 cal/mol and T ϭ 295 o K (30). ⌬F (in kcal mol Ϫ1 ) is a measure of the difference in free energy between the interaction of an ShK analogue with the channel, compared with that of wild-type ShK (26).
Double Mutant Cycle Analysis-This method evaluates the strength of the interaction between any pair of channel and toxin residues. For each mutant cycle, we measured the potency (K d ) of ShK and its analogues on IKCa1 and its mutants. The change in coupling energy, ⌬⌬G, for a given pair of ShK-IKCa1 residues and their mutants was calculated using the formula ⌬⌬G ϭ kT ln ⍀, as described earlier (20,23). Based on the studies of Schreiber and Fersht (29) and Hidalgo and MacKinnon (19), ⌬⌬G values Ն 0.5 kcal mol Ϫ1 indicate that a particular pair of ShK and IKCa1 residues are likely to lie within 5 Å of each other.

RESULTS
The Sea Anemone Toxins, ShK and BgK, Block IKCa1 and Kv1.3 Channels-We compared the toxin sensitivities of the two K ϩ channels, present in activated human T lymphocytes, IKCa1 and Kv1.3. Two potent peptide inhibitors, ShK and BgK, from the sea anemones S. helianthus and B. granulifera, were chosen for analysis (30,31). These peptides share 31% se-quence identity (Fig. 1A). Both contain six conserved cysteines that form three disulfide bonds, a feature common to many channel-blocking peptide inhibitors from scorpion venom as well as many defensins (32). However, the structures of ShK and BgK are significantly different from that of the scorpion toxins and defensins (23,33,34).
The cloned IKCa1 and Kv1.3 genes were expressed in mammalian cells, and representative currents are shown in Fig. 1B. IKCa1 currents were elicited by 1 M calcium in the pipette following break-in, whereas depolarizing pulses were used to generate Kv1.3 currents. ShK and BgK, when applied externally, block the IKCa1 channel in the low nanomolar range with K d values of 30 Ϯ 7 nM and 172 Ϯ 43 nM, respectively (  (Fig. 1B, top panel), indicating that this peptide has an exquisite ability to discriminate between the two T lymphocyte K ϩ channels.
Determining the IKCa1 Channel-binding Surface of ShK-To elucidate the molecular basis for the ability of ShK to discriminate between IKCa1 and Kv1.3, we used alanine-scanning mutagenesis to identify ShK residues essential for binding to both these channels. A series of monosubstituted peptide analogues, in which each residue is substituted with alanine, was tested for their ability to block the IKCa1 and Kv1.3 channels. The only exception is His 19 , which was replaced by a lysine. Structurally critical ShK residues that were not substituted included the six half-cysteine residues, as well as Asp 5 , Ala 14 , and Gly 33 . K d values measured in this "Ala-scan" were used to calculate the free energy difference of binding, ⌬F (see "Materials and Methods"). The greater the change in free energy (⌬F), the greater the influence of a particular ShK residue for channel binding.
In the representative example shown in Fig. 1B (left), an alanine substitution at ShK position 20 significantly reduces the affinity of the toxin for IKCa1 (K d ϭ 2,450 nM) compared with that of the wild-type toxin (K d ϭ 30 nM). The change in free energy of binding, caused by this substitution, is 2.5 kcal mol Ϫ1 . Using this approach, we determined ⌬F for each alanine substitution. The ShK residues most critical for IKCa1 binding, with ⌬F values greater than or equal to 2.5, are shown in red in Fig Fig. 2A). Fig. 3 (left) shows the positions of the ShK residues (bottom and side view) that contribute to its interaction with IKCa1, color coordinated with the histograms in Fig. 2A. All the highly critical residues (⌬F Ն 2.5) cluster together on one surface of the ShK peptide (red), and Met 21 (orange) lies immediately adjacent. Residues with moderate influence (yellow and blue) form the margins of the ShK:IKCa1-binding surface. In general, nonessential residues cluster together on the opposite surface of the peptide (Fig. 3, left). Thus, the surface of ShK that binds IKCa1 extends in its greatest distance from Arg 1 on one side to Phe 15 on the other (Fig. 3, left, bottom view). Lys 22 lies at the lowest point in the channel-binding surface (Fig. 3,  left, side view).
Comparison of the ShK Toxin-binding Surfaces in IKCa1 and Kv1.3-All ShK analogues were tested on the Kv1.3 channel to define the ShK-binding surface for this structurally well defined channel (20,21,23).  (Fig. 2B). The only exception is ShK-Arg 29 , which seems to be slightly more important for binding to Kv1.3 than for IKCa1 (⌬F ϭ 0.88 and 0.56, respectively). Our results thus suggest that ShK uses a larger number of residues to stabilize its interaction with IKCa1 than with Kv1.3, although it uses the same core domain of five clustered residues for binding to both channels.
ShK-Lys 22 protrudes into the Kv1.3 pore and lies in close proximity to Tyr 400 and Asp 402 in the selectivity filter, and is critical for the interaction of the toxin with this channel (23). To determine the contribution of Lys 22 to the interaction of the toxin with IKCa1, we compared the affinity of IKCa1 and Kv1.3 for four ShK-Lys 22 analogues (Fig. 2B); we also examined the effect of these analogues on the Kv1.3-His 404 3 Val 404 mutant, and these data are presented in a subsequent section. In two analogues, the positively charged lysine is replaced with the shorter, positively charged amino acids, ornithine (Orn) and diaminopropionic acid (Dap), whereas the other two analogues have the neutral residues norleucine (Nle) and alanine (Ala) at position 22. These residues differ in their side chain lengths (Dap, 2.5 Å; Ala, 2 Å; Orn, 5.0 Å; Nle, 5.0 Å; lysine, 6.3Å). Our results show that Ala 22  alter the affinity of ShK for Kv1.3 (Fig. 1B, bottom left and Fig.  2B). The longer ShK-Orn 22 substitution also significantly disrupts the affinity of the toxin for IKCa1 (⌬F ϭ 1.7), although to a lesser extent than Dap 22 , possibly because it is better anchored in the channel pore, whereas this analogue blocks Kv1.3 with potency equivalent to wild-type ShK (Fig. 2B). Fig. 3 highlights the residues that ShK uses to interact with IKCa1 (left) compared with those for Kv1.3 (right). The residues are color-coordinated with the histogram in Fig. 2, A and B. ShK residues required for binding to both channels are clustered together on one surface (colored), whereas white residues not required for the interaction are located on the opposite toxin surface. ShK uses eight essential residues, Arg 11 23 , and Arg 24 ). Alanine substitutions at any of these five critical positions in the binding core domain severely disrupt the ShK-channel interaction in both channels (⌬F Ͼ 1.5, Fig. 3, red and orange). ShK residues that surround this critical core domain (ShKpositions 1, 6, 7, 11, 13, 15, 21, 26, and 27) exhibit a lower influence on binding to Kv1.3 compared with IKCa1. Thus, the overall binding surface of ShK for IKCa1 is larger and contains more essential interacting residues than its binding surface for Kv1.3 (Fig. 3). Nevertheless, the affinity of the toxin for Kv1.3 is significantly greater than for IKCa1 (Fig. 1B), suggesting that the picomolar affinity of ShK for Kv1.3 is dependent on a few very tight toxin-channel contacts, whereas its ϳ1,000-fold lower nanomolar affinity for IKCa1 is because of a greater number of weaker interactions.
Interactions of ShK with Residues in the External Vestibule of IKCa1-Peptides from scorpion and snake venom, as well as from sea anemone, bind to residues in the external vestibule of the ion conduction pathway of eukaryotic voltage-gated potassium channels and occlude their pores (17)(18)(19)(20)(21)(22)(23)34). This vestibule corresponds to the outer pore and turret region in the structurally defined bacterial K ϩ channel, KcsA (24), and mutations in the turret region of KcsA greatly enhance the ability of this channel to bind peptide inhibitors (25). We therefore undertook a series of mutational analyses to define the interactions of ShK with residues in the external vestibule of IKCa1.
Charge-reversal Mutation at IKCa1-Asp 239 Dramatically Reduces Sensitivity to ShK-We first aligned the turret and pore regions of KcsA, Kv1.3, and IKCa1 (Fig. 4A). All three channels are remarkably similar in the pore region with absolute conservation of the GYGD motif (Fig. 4A). Interestingly, IKCa1 contains an aspartate (Asp 239 ) at the position homologous to Asp 386 in the turret of Kv1.3 that is essential for the interaction of Kv1.3 with ShK and various scorpion toxins (20,23). A charge-reversal mutation involving Asp 386 in Kv1.3 (Asp 3 Lys) substantially reduces the channel sensitivity to ShK (⌬F ϭ 1.9), kaliotoxin (⌬F Ͼ 2.5), and charybdotoxin (⌬F Ͼ 2.5) (20,23). The converse charge-reversal mutation at the homologous position in KcsA (Arg 64 3 Asp 64 ) enhances this channels sensitivity for agitoxin-2 (25). We therefore replaced IKCa1 Asp 239 with the positively charged Lys, and examined the sensitivity of the mutant channel to ShK. The IKCa1-Lys 239 mutant was expressed in rat basophilic leukemia cells, and representative currents, elicited by 1 M calcium in the pipette solution, are shown in Fig. 4B. The mutant channel is ϳ18-fold less sensitive to ShK (K d ϭ 548 Ϯ 37 nM) than wild-type IKCa1, suggesting that the Asp 3 Lys charge-reversal mutation at position 239 in IKCa1 has the same consequences on ShK binding as the identical mutation at the homologous position in Kv1.3 (23). The change in free energy for this channel mutant (⌬F ϭ 1.7 kcal mol Ϫ1 ) is equivalent to severe Ala substitutions (⌬F Ͼ1.5 kcal mol Ϫ1 ) in ShK (compare with Fig. 2A) (23), we used mutant cycle analysis to determine whether this peptide residue is close to Asp 239 in IKCa1. Two additional ShK residues were evaluated for their ability to interact with Asp 239 . ShK-Arg 11 is ϳ21 Å from Arg 29 on the channel-binding surface, whereas ShK-Lys 9 is on the nonbinding surface of ShK (Fig. 3). Fig. 4C shows the three mutant cycles that were used to analyze the ShK interactions with IKCa1-Asp 239 . ShK-Arg 29 and ShK-Arg 11 couple tightly with IKCa1-Asp 239 (⌬⌬G values ϭ 0.79 and 1.56 kcal mol Ϫ1 , respectively), suggesting that these toxin residues are within ϳ5 Å of Asp 239 in different IKCa1 subunits, Arg 11 being closer to Asp 239 than Arg 29 . As expected, because of its position on the opposite site of the interactive toxin surface, ShK-Lys 9 does not couple with IKCa1-Asp 239 (⌬⌬G ϭ 0.06 kcal mol Ϫ1 ). Because Arg 11 and Arg 29 lie ϳ21 Å apart and are oriented at an angle of ϳ105 o from the center of the toxin, they most likely interact with Asp 239 residues in adjacent subunits (Fig. 5, left), as has been reported for Kv1. 3 (23) rather than with Asp 239 in opposite IKCa1 subunits.
ShK-Lys 22 Protrudes into the IKCa1 Pore and This Interaction Is Dependent on the K ϩ Ion Concentration in the Pore-Because Lys 22 is located at the lowest and central point in the channel-binding surface of ShK (Fig. 3), there is a good likelihood that it lies close to or within the pore of IKCa1, as has been reported for Kv1.3 (23). To test this idea, we examined whether the terminal amine of ShK-Lys 22 lies close to a potassium-binding site in the ion selectively filter. Earlier studies have shown that Lys 22 in ShK, and the homologous Lys 27 in the scorpion toxins, agitoxin-2 and kaliotoxin, lie in close proximity to a potassium-binding site in the ion selectivity filter of the Kv1. 3 and Shaker channels (21,22). Occupancy of this site by a K ϩ ion appears to destabilize the interaction of the native toxins with these channels via electrostatic repulsion of Lys 27 , because it has little effect on toxin analogues containing neutral substitutions at position 27 (21,22). Similarly, potassium ions were shown to inhibit the interaction between Lys 27 (but not Asn 27 ) in charybdotoxin and residues in the pore of the large conductance, calcium-activated K ϩ channel, BK (35). We have used a similar approach to determine whether ShK-Lys 22 lies in the vicinity of a potassium-binding site in the IKCa1 pore. We compared the effect of changing the external K ϩ concentration from 4.5 to 164.5 mM on the affinity of the IKCa1 channel for ShK-Lys 22 and ShK-Ala 22 . Consistent with earlier reports, increasing external [K ϩ ] also reduces the potency of ShK-Lys 22 on the IKCa1 channel (7.7-fold), whereas block by ShK-Ala 22 is minimally affected (1.6-fold). We assessed the strength of this interaction using mutant cycle analysis (Fig.  4D). The ⌬⌬G value for this cycle (0.91 kcal mol Ϫ1 ) indicates that ShK-Lys 22 lies within 5 Å of a K ϩ -binding site located in the IKCa1 pore, as has been reported for Kv1.3 (23).
Since ShK-Lys 22 protrudes into the pores of both IKCa1 and Kv1.3, why is the IKCa1 pore more sensitive to ShK-Lys 22 substitutions than the Kv1.3 pore, especially those involving shorter positive charged residues (Orn 22 and Dap 22 )? A comparison of the sequences of the outer pore regions of the two channels suggests an explanation (Fig. 4A). Kv1.3 contains a histidine (His 404 ) at the entrance to its pore that is positioned just above the selectivity filter, whereas IKCa1 contains a hydrophobic residue (Val 257 ) at the corresponding position. Might the difference in the nature of the residue at the pore entrance account for the differential sensitivity to Shk-Dap 22 and Shk-Orn 22 ? To test this possibility, we replaced Kv1.3-His 404 with valine, a mutation that makes the outer pore of anemone S. helianthus, uses to interact with two distinct potassium channels present in activated human T-lymphocytes: the intermediate conductance, Ca 2ϩ -activated K ϩ channel IKCa1, and the voltage-gated K ϩ channel Kv1. 3. Although using the same core domain, the IKCa1-binding surface of ShK is more extensive than its Kv1.3-binding surface, and yet ShK blocks Kv1.3 with ϳ1,000-fold greater potency than IKCa1. A few tight Kv1.3-ShK contacts appear to underlie the picomolar affinity of ShK for Kv1.3, whereas the IKCa1-ShK interface is stabilized by a greater number of weaker interactions.
Because ShK interacts with IKCa1 and Kv1.3 using the same core domain involving His 19 , Ser 20 , Lys 22 , Tyr 23 , and Arg 24 (Fig. 3), this toxin is likely to sit in both channels with a similar geometry. Several lines of evidence support this idea. First, charge-reversal mutations at homologous positions in IKCa1 (Asp 239 3 Lys 239 ) and Kv1.3 (Asp 386 3 Lys 386 ) reduce the sensitivity of these channels to ShK block in a similar fashion. Second, ShK-Arg 29 shows energetic coupling with Asp 239 in IKCa1, as well as with the homologous Kv1.3-Asp 386 (Fig. 4C) (23). Third, ShK-Arg 11 and ShK-Arg 29 appear to couple with residues in adjacent subunits of IKCa1 as has been reported for Kv1.3 (23). Fourth, the critical ShK-Lys 22 lies close to a potassium-binding site in the selectivity filter of the Kv1.3 pore (23), and mutant cycle analyses suggests the same is true for IKCa1 (Fig. 4D). Last, replacement of the critical ShK-Lys 22 with bulky neutral residues (Nle and Ala) substantially reduces the affinity of the toxin for both channels, possibly because such residues are not tolerated in the channel pore. These results indicate that ShK binds to IKCa1 in a region corresponding to the external vestibules of Kv1.3 and uses a comparable docking geometry. Such a docking configuration would place Arg 1 and Phe 15 , the two residues at opposite margins of the IKCa1binding surface (Fig. 3A), in close proximity to channel residues in opposite subunits in the IKCa1 tetramer (Fig. 5).
Knowing the NMR structure of ShK (33) and its approximate docking configuration in IKCa1, we used this peptide as a structural template to estimate the dimensions of the toxin receptor in the external vestibule of IKCa1. We have obtained two independent estimates of the diameter of the toxin receptor in the IKCa1 external vestibule. First, we estimate the width of the IKCa1 toxin receptor to be ϳ31 Å (Fig. 5) based on the width of the IKCa1-binding surface of ShK (distance from Arg 1 to Phe 15 ). Second, the x-y dimension of the toxin receptor is estimated to be ϳ22 Å (Fig. 5, left), based on the distance between the two toxin residues, ShK-Arg 11 and ShK-Arg 29 (Fig. 5,right), that interact with Asp 239 residues in adjacent IKCa1 subunits. This value implies (by Pythagorean triangulation) a distance of ϳ31 Å between Asp 239 residues in opposite subunits (Fig. 5). We estimate that the toxin receptor is approximately ϳ8 -9 Å deep, based on the vertical distance between one horizontal line joining Arg 1 and Phe 15 , and a second horizontal line connecting the terminal amines of Arg 11 and Arg 29 (Fig. 5,  right). The selectivity filter is estimated to lie ϳ12-13 Å below the outer edge of the toxin receptor in the vestibule based on the vertical distance between the terminal amine of Lys 22 and the horizontal line joining Arg 1 and Phe 15 (Fig. 5, right). Our deduced dimensions of the toxin receptor in the IKCa1 vestibule are in good agreement with those obtained by crystallography for the KcsA vestibule (24,25), and by toxin-mapping for Kv1.3 (20,21,23) and Shaker (17)(18)(19)22).
Thus, the IKCa1 external vestibule appears to be topologically similar to those of the distantly related Kv1.3, Shaker, and KcsA channels, a result that provides support for the development of homology models of the IKCa1 outer pore based on the KcsA crystal structure. The experimentally determined differences in the ShK-IKCa1 and ShK-Kv1.3 binding interfaces raises the possibility of designing novel peptides and small molecule inhibitors that selectively target the IKCa1 Dimensions of the external vestibule of IKCa1 were determined using the NMR-derived ShK structure as a molecular caliper. Left, schematic representation of the external vestibule with Asp 239 residues highlighted (derived from the KcsA structure) with ShK (gray) docked. ShK-Arg 11 and Arg 29 are ϳ21 Å apart, interacting with Asp 239 residues in adjacent channel subunits. ShK-Lys 22 is shown at the geometric center of the tetramer, protruding into the pore. In this geometry Arg 1 and Phe 15 (at opposing ends of the ShK:IKCa1-binding surface) are located at opposite channel subunits. The diameter of the tetramer based on the distance between ShK-Arg 1 to Phe 15 and the Asp 239 x-y triangulation is approximately 31 Å. Right, schematic side view of the ShK toxin (compare Fig. 3, bottom) with the channel-binding surface highlighted. The distances between key toxin residues are indicated and correspond approximately to dimensions of the IKCa1 external vestibule (see "Discussion"). channel. Such reagents might be useful in elucidating the role of the IKCa1 channel in diverse cell types and may also have therapeutic value.