Generating a high affinity scorpion toxin receptor in KcsA-Kv1.3 chimeric potassium channels.

The crystal structure of the bacterial K(+) channel, KcsA (Doyle, D. A., Morais, C. J., Pfuetzner, R. A., Kuo, A., Gulbis, J. M., Cohen, S. L., Chait, B. T., and MacKinnon, R. (1998) Science 280, 69-77), and subsequent mutagenesis have revealed a high structural conservation from bacteria to human (MacKinnon, R., Cohen, S. L., Kuo, A., Lee, A., and Chait, B. T. (1998) Science 280, 106-109). We have explored this conservation by swapping subregions of the M1-M2 linker of KcsA with those of the S5-S6 linker of the human Kv-channel Kv1.3. The chimeric K(+) channel constructs were expressed in Escherichia coli, and their multimeric state was analyzed after purification. We used two scorpion toxins, kaliotoxin and hongotoxin 1, which bind specifically to Kv1.3, to analyze the pharmacological properties of the KcsA-Kv1.3 chimeras. The results demonstrate that the high affinity scorpion toxin receptor of Kv1.3 could be transferred to KcsA. Our biochemical studies with purified KcsA-Kv1.3 chimeras provide direct chemical evidence that a tetrameric channel structure is necessary for forming a functional scorpion toxin receptor. We have obtained KcsA-Kv1.3 chimeras with kaliotoxin affinities (IC(50) values of approximately 4 pm) like native Kv1.3 channels. Furthermore, we show that a subregion of the S5-S6 linker may be an important determinant of the pharmacological profile of K(+) channels. Using available structural information on KcsA and kaliotoxin, we have developed a structural model for the complex between KcsA-Kv1.3 chimeras and kaliotoxin to aid future pharmacological studies of K(+) channels.

During the last decade, important attention has focused upon T-cell K ϩ channels as potential pharmaceutical targets for modulating immune system function. In particular, the blockade of the voltage-gated K ϩ channel encoded by Kv1. 3 inhibits T-cell activation, lymphokine secretion, and cell proliferation (3). Kv1.3 is very scorpion toxin-sensitive (4). These toxins constitute useful molecular tools to study physiological and structural properties of voltage-gated potassium (Kv) channels (5,6). They form a class of basic peptides containing 30 to 40 amino acid residues highly reticulated with three or four disulfide bridges (6). They are structurally related, sharing a characteristic backbone fold called the cysteine-stabilized ␣/␤ motif (6 -8). The toxins bind with a 1:1 stoichiometry and may inhibit Kv channel activity by plugging the external pore entryway (2,6,9). Amino acid residues between hydrophobic transmembrane segments S5 and S6 (the S5-S6 linker region with the P domain) form the receptor site for scorpion toxin as well as the outer entrance to the Kv channel pore and a substantial part of the ion conduction pathway (2, 9 -11). Based on the hypothesis that complementary surfaces of scorpion toxin and Kv channel interact, the known three-dimensional structure of scorpion toxins has been exploited to study the topology of the external mouth of Kv channels, e.g. Shaker and Kv1.3 (11)(12)(13). Thermodynamic mutant cycle analysis has been used to identify specific amino acid residues in the S5-S6 linker region as part of the scorpion-toxin receptor site (11)(12)(13). The strength of electrostatic and hydrophobic interactions between potentially interacting amino acid residues of Kv1.3 channel and scorpion toxin, e.g. kaliotoxin (KTX) 1 (9), was shown to influence the affinity of the scorpion toxin to its receptor.
Recent studies have demonstrated a remarkable structural conservation between the pore structures of a prokaryotic K ϩ channel from Streptomyces lividans (14), KcsA, and eukaryotic Kv channels (2). The determination of the three-dimensional structure of KcsA by x-ray analysis has provided the first molecular description of an ion-selective channel (1). Although KcsA subunits contain only two transmembrane segments (M1 and M2) and not six like Kv-channel subunits, KcsA and Kv channels are believed to share essentially the same pore structure. KcsA does not normally bind scorpion toxins. However, mutation of three KcsA channel residues (Q58A, T61S, R64D) in the KcsA pore region sufficed to generate a competent, low affinity agitoxin 2 (AgTX2) binding site with an equilibrium dissociation constant (K d ) of about 0.6 M (2). This value differed by 6 orders of magnitude with the one reported for AgTX2 binding to Kv1.3 (ϳ0.3 pM) (15). Obviously, some energetically coupled residue pairs were absent at the recognition interface needed for a high affinity toxin binding site.
We have generated a high affinity toxin binding site on KcsA using an alternative approach. Various parts of the S5-S6 linker region of Kv1.3 were transferred to KcsA to generate a panel of KcsA-Kv1.3 chimeras. They were expressed in Escherichia coli and, when possible, were solubilized, purified, and assayed for toxin binding. The results demonstrated that KcsA-Kv1.3 channels could be obtained that bound KTX with a K d (ϳ4 pM) similar to the one reported for native Kv1.3 (ϳ3 pM) (15). Also, the results showed that certain Kv1.3 amino acid residues in the S5-S6 linker region were not suitable for transfer to KcsA, possibly interfering with a correct subunit fold and/or subunit assembly. We have used our results in combination with the known structures of KcsA (1) and KTX (9) to construct a three-dimensional structural model for the KTX-KcsA channel complex. The chimeras developed here can also be exploited in putting forward an understanding of potassium channel pharmacology and in developing promising therapeutic agents.

EXPERIMENTAL PROCEDURES
Materials-For the production of recombinant proteins E. coli strains were grown in super broth (25 g of bacteriological peptone, 15 g of yeast extract (Life Technologies, Inc.), and 5 g of NaCl per liter). Antibiotics were from Sigma. Ampicillin and kanamycin were added to final concentrations of 100 g/ml and 25 g/ml, respectively. Isopropyl-1-thio-␤-D-galactopyranoside was from Roth. n-Decyl-␤-D-maltopyranoside (16) was purchased from Calbiochem. Prestained molecular weight markers were from Sigma. All other reagents were from Fluka or Merck. For polymerase chain reactions, Pfu DNA polymerase (Stratagene) was used in a Peltier Thermal Cycler PTC-200. Eam1104I was supplied by Stratagene; other restriction enzymes and T4 DNA ligase were from MBI Fermentas. E. coli XL1-Blue (Stratagene) served as host for the propagation of pQE-32 (Qiagen) and constructs; E. coli M15 pREP4 (Qiagen) was used for the production of recombinant proteins.
Construction of KcsA-Kv1.3 Chimeras-The KcsA K ϩ channel gene (GenBank accession number Z37969) was amplified from S. lividans strain 66 genomic DNA (DSMZ 46482) using SKC1F and SKC1R primers, as described by Schrempf et al. (14). The polymerase chain reaction product was cloned in pT7Blue (Novagen), and the 546-base pair SphI-SalI fragment encoding KcsA was subcloned into pQE32 (Qiagen). The kcsA-pQE32 construct and Kv1.3 cDNA from human T lymphocytes (GenBank accession number L23499) in pRcCMV (17) were used as templates in polymerase chain reactions to construct the chimeras Chi I to Chi VIII. Chi I was constructed according to the Seamless method (Stratagene). Chimeras II to VIII were prepared by overlap extension using the polymerase chain reaction (18) or by using the Quickchange site-directed mutagenesis method (Stratagene). At the N terminus and at the C terminus, each chimera contained a His tag and a streptavidin affinity tag, respectively (19). The His tag allowed purification using Ni 2ϩ -chelating column. Recombinant clones were analyzed using standard methods (20). Sequences of all constructs were verified by sequencing both strands.
Induction and Purification of KcsA and KcsA-Kv1. 3 Chimeras-E. coli M15 pREP4 cells were transformed with constructs as described previously (20). Transformed cells were plated on a LB plate containing appropriate antibiotics, and a single colony was picked for overnight pre-culture. The pre-culture was diluted into 500 -1000 ml of super broth (containing antibiotics) to obtain a final absorbance of 0.2 at 600 nm and grown at 30°C to mid-logarithmic phase. The expression of recombinant proteins was induced with 0.5 mM isopropyl-1-thio-␤-Dgalactopyranoside for 2 h. The cells were harvested by centrifugation and washed once in buffer containing 150 mM KCl, 50 mM MOPS (pH 7.0), resuspended in the same buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 M pepstatin A, 2 M leupeptin), and disrupted by a French press. Unbroken cells and inclusion bodies were removed by centrifugation (12,000 ϫ g, 15 min), and membranes were extracted from the supernatant by high speed centrifugation (110,000 ϫ g, 45 min, 4°C). The membranes were resuspended in a buffer containing 100 mM NaP i and 150 mM KCl (pH 7.0), and proteins were extracted with 40 mM n-decyl-␤-D-maltopyranoside (Calbiochem) for 30 min at 4°C using a rotary mixer. This mixture was centrifuged for 45 min at 85,000 ϫ g, and imidazole was added to the supernatant to a final concentration of 30 mM for affinity purification on Ni 2ϩnitrilotriacetic acid agarose, as described previously (21). Protein concentrations were determined by spectrophotometry and the BCA test (Pierce). Proteins were analyzed by SDS-PAGE (22).
Western Blotting-SDS-PAGE was performed on a 15% gel. Proteins were transferred onto nitrocellulose membrane (Schleicher & Schuell) by electrotransfer. Immunodetection was carried out using the polyclonal anti-Arg-Gly-Ser-(His) 4 antibodies (Qiagen) as described previously (21). For binding assays, the membranes were blocked 1 h at room temperature in 5% nonfat milk in phosphate-buffered saline and rinsed briefly with phosphate-buffered saline before incubation with 0.2 nM 125 I-KTX in the binding buffer B (20 mM Tris/HCl (pH 7.4)), 0.1% bovine serum albumin, 50 mM NaCl). After a 1-h incubation, the membranes were briefly washed three times with binding buffer B containing 1% bovine serum albumin. Nonspecific binding was assayed in the presence of 100 nM KTX. Binding was detected after overnight exposure in densitometer cassette using PhosphorImager.
Binding Studies-125 I-KTX and 125 I-HgTX 1 -A19Y/Y37F were prepared as described previously (23,24). In each experiment, recombinant proteins were incubated at room temperature in binding buffer B containing 2 mM n-decyl-␤-D-maltopyranoside. When using membrane preparations of E. coli, the incubations were performed in binding buffer B. At the end of the incubation period, the reaction was stopped by dilution with 4 ml of wash buffer (20 mM Tris/HCl (pH 7.4), 150 mM NaCl, 0.2% bovine serum), and the solution was filtered through a GF/C glass fiber filter. The filters were washed twice with 4 ml of the wash buffer. Filter-retained radioactivity was determined by ␥ counting. Nonspecific binding was defined in the presence of 100 nM unlabeled toxin. In each experiment, triplicate assays were carried out, and the data were averaged.
For competition experiments, recombinant channels were incubated for 1 h at room temperature in binding buffer with radiolabeled toxin (30 -100 pM) and a series of cold toxin concentrations in a total volume of 250 l. Equation 1, where B min is the minimum binding, B max is the maximum binding, and IC 50 is the inhibitory concentration at 50% inhibition, was fit to the data. For saturation binding, the incubation time was 2 h. Equation 2, where K d is the equilibrium dissociation constant, was fit to the data. Association data were analyzed as a single exponential according to Equation 3 below.
where k is the pseudo-first-order rate constant of the association process. k is related to the second-order rate constant of the association process (k on ), the labeled toxin concentration, and the first order rate constant of the dissociation process (k off ) according to Equation 4, as follows.
Dissociation data were analyzed as a one-phase exponential decay according to Equation 5, as follows.
where B ϱ refers to the new binding equilibrium reached after the addition of a large excess of unlabeled toxin. The half-life value is, All the data were subjected to nonlinear regression analysis using PRISM (GraphPad), assuming the presence of a single class of binding sites.
Protein Model-Our modeling is based on the crystal structure of KcsA solved at 3.2 Å resolution (1) (Protein Data Bank, accession number 1BL8). To create a model of a chimeric channel KcsA-Kv1.3, the following mutations were introduced: R52A, G53D, A54D, G56T, A57S, Q58G, L59F, I60S, T61S, Y62I, R64D, respective to the KcsA amino acid sequence number. A model of the chimeric channel KcsA-Kv1.3, Chi VIII, was generated by the mutations L81M and Y82H to Chi IV. Both models can be viewed on the Internet (information is available online at the website of the AFMB laboratory).

RESULTS
Given the reported structural conservation between the pores of bacterial KcsA channels (2) and eukaryotic Kv channels, we explored possibilities of transferring the scorpion toxin receptor site of human (h)Kv1.3 channels to the toxin-insensitive KcsA channels. Accordingly, we replaced parts of the KcsA M1-M2 linker region by a homologous sequence of the Kv1.3 S5-S6 linker region, which most likely contains the complete scorpion toxin receptor. We divided the S5-S6 linker from hKv1.3 into three subregions (Table I) similar to the ones that were previously transferred independently from Kv1.3 to Kv2.1 subunits (25). First, we constructed KcsA-Kv1.3 chimera Chi I, displaying the complete Kv1.3 S5-S6 linker, and Chi II, containing subregions I and III of the linker (see Table I). Isopropyl-1-thio-␤-D-galactopyranoside induction of KcsA-Kv1.3 protein expression in E. coli transformed with Chi I or Chi II DNA constructs produced an immediate arrest of bacterial growth followed by cell lysis (data not shown). For that reason, no recombinant Chi I or Chi II protein could be obtained for analysis. Then we transferred only subregion I from Kv1.3 to KcsA (chimera Chi III). This time, we achieved chimeric KcsA-Kv1.3 protein expression. The yield of purified Chi III protein, however, was too low (0.03 mg of protein/liter of bacterial culture) for detailed biochemical and pharmacological studies. Next, we expressed KcsA-Kv1.3 chimeras Chi IV and Chi V containing shortened versions of subregion I (Table I). Now, the yields of purified Chi IV and Chi V proteins (1 to 1.5 mg of purified protein/liter of bacterial culture) were comparable with the one of KcsA protein (Table I). Thus, we assayed primarily Chi IV and Chi V preparations for their ability to bind K ϩ channel toxins. In filter binding assays we showed that Chi IV and Chi V were competent for binding 125 I-KTX and 125 I-HgTX 1 -A19Y/Y37F, respectively (Fig. 1, A and B). Similar data were obtained with Chi VI (Fig. 1A), which contained, in addition to subregion I sequences, Met-81 of subregion III (Table I).
In contrast, the iodinated scorpion toxins did not bind to wildtype KcsA in agreement with previous data obtained with AgTX2 (2).
The mobility of solubilized and purified Chi IV and Chi V proteins in SDS-PAGE correlated with a molecular mass of 65 kDa ( Fig. 2A). This indicated a tetrameric structure for purified Chi IV and Chi V. After boiling in SDS-PAGE sample buffer, Chi IV and Chi V migrated in SDS-PAGE as monomers ( Fig.  2A). A similar behavior has been observed with KcsA ( Fig. 2A) (21), indicating that the tetrameric channel structure had been disrupted by the heat treatment. The SDS-PAGE results showed that it was possible to prepare Chi IV and Chi V, respectively, either in tetrameric or in monomeric form. Chi IV tetramers and monomers were blotted onto nitrocellulose membranes and tested for their 125 I-KTX binding activity. The results showed that 125 I-KTX bound to Chi IV tetramers but not to Chi IV monomers (Fig. 2B). It demonstrated that the dissociation of Chi IV tetramers to monomers had also disrupted the toxin receptor site in Chi IV, providing direct biochemical evidence for a tetrameric structure of the scorpion toxin receptor.
The selectivity of Chi IV and Chi V against other K ϩ channel toxins like apamin, PO5, and ␣-dendrotoxin (␣-DTX) was investigated. Apamin as well as PO5 bind with high affinity to SK channels (26). ␣-DTX is inactive on Kv1.3 but binds with high affinity to the closely related Kv1.1 and Kv1.2 channels (4). 0.1 M concentrations of PO5, apamin or ␣-DTX had no effect on 125 I-KTX binding to Chi IV (n ϭ 3). Apamin and PO5 interfered only at 1 M concentration to some extent with 125 I-KTX binding (n ϭ 3) (Fig. 3A). Similar results were obtained using Chi V (not shown). Collectively, the results sug- gested that the Kv1.3 subregion I sequences sufficed to transfer a selective scorpion toxin receptor site to KcsA. This observation was further corroborated after transferring subregions I from the S5-S6 linker region of KTX-insensitive Kv1.4 or Kv1.5 to KcsA (see Fig. 3B). The Kv1.4-KcsA (Chi IX) and Kv1.5-KcsA (Chi X) chimeras were expressed, purified in tetrameric form (data not shown), and tested for 125 I-KTX binding in filter binding assays (Fig. 3C). The results demonstrated that chimeras Chi IX and Chi X, like native Kv1.4 and Kv1.5 channels, did not interact with 125 I-KTX to a significant extent (4).
The binding of 125 I-KTX to chimeric KcsA-Kv1.3 proteins was quantitatively assessed in competition experiments with KTX, such as those shown in Fig. 4A for Chi IV. The KTX concentrations, with which half-maximal binding of 125 I-KTX was obtained, were determined for Chi III to Chi VI (Table II). Apparently, KTX had the highest affinity for Chi III and Chi IV (IC 50 ϳ 0.2 nM). Chi VI, containing an additional modification (Met-81) in subregion III, showed a 6-fold reduced KTX binding affinity in comparison to Chi IV. The results showed a saturable specific binding of 125 I-KTX to Chi IV (and Chi V) under equilibrium conditions (Fig. 4B). A single isotherm was fit to the data, with K d ϭ 0.37 nM and B max ϭ 12.2 fmol/ng of purified Chi IV protein and with K d ϭ 2.25 nM and B max ϭ 13.1 fmol/ng purified Chi V protein (Table II). The B max values were in the expected range for active purified KcsA-Kv1.3 chimeras, which theoretically should be 12.3 fmol/ng. This indicated that the KTX binding activity of the purified KcsA-Kv1.3 chimeras was close to 100%. Also, the fitted equilibrium dissociation constants (K d ) were in good agreement with the IC 50 values ob-tained in the competition experiments (Table II). In additional competition experiments with 125 I-KTX, we have assayed the binding of HgTX 1 and AgTX2 to Chi IV. The toxin concentrations, with which half-maximal binding of 125 I-KTX was obtained (IC 50 ), were 2.2 nM (n ϭ 3) for HgTX 1 and 6.4 nM (n ϭ 3) for AgTX2 (Fig. 3A).
For Chi IV (and Chi V), we complemented the competition and saturation binding experiments with a kinetic analysis of 125 I-KTX binding (Fig. 4, C and D). A plot of the association rate versus 125 I-KTX concentration can be fit to a pseudo-firstorder reaction (Fig. 4C, inset), from which a rate constant k on of 1.8 ϫ 10 7 M Ϫ1 s Ϫ1 was calculated. The addition of an excess of unlabeled KTX led to dissociation of bound ligand. The time course closely followed mono-exponential kinetics (Fig. 4D), indicative of a first-order reaction. The data could be fit to a first-order rate constant of 0.93 ms Ϫ1 , independent of toxin concentration (t1 ⁄2 ϭ 745 s). The k off value is also in the range of previously reported dissociation rates of other iodinated toxins, such as margatoxin and HgTX 1 (15,24). K d calculated from the ratio of k off /k on is 0.052 nM, in agreement with the one determined under equilibrium conditions. The binding results with Chi IV showed that subregion I was sufficient to generate on KcsA a selective toxin receptor site with a high affinity for KTX (Table II).
KTX binds to the toxin receptor site on Kv1.3 channels with an IC 50 of 3.1 pM (15). The value is ϳ60-fold smaller than the K d value for binding KTX to Chi IV. We reasoned, therefore, that additional amino acid residues of the S5-S6 linker region should contribute to a Kv1.3-like KTX receptor site. Accordingly, we produced further Kv1.3-KcsA chimeras, e.g. Chi VII and Chi VIII, which contained additional amino acid residues from Kv1.3 subregion III (Table I). Crude membrane preparations of Chi VII and Chi VIII bound 125 I-KTX (Fig. 5A) with a very high affinity. The competition binding experiments with KTX (Fig. 5B) revealed IC 50 values of 3.8 and 4.2 pM. They were nearly identical to the one for binding KTX to Kv1.3 channels (Table II; Ref. 15). In controls, we showed that 125 I-KTX did not bind to crude membrane preparations of KcsA (Fig. 5A). Collectively, the results showed that Chi VII and Chi VIII displayed the most competent KTX receptor site. However, Chi VII and Chi VIII were expressed in E. coli to a relatively low yield (Table I). Also, they could not be purified in a stable tetrameric form ( Fig. 2A). Thus, solubilized Chi VII and Chi VIII were not suited for toxin binding studies in solution (Fig. 5A).
Scorpion toxins like KTX have a positively charged lysine residue (Lys 27 in KTX) that is critical for their pore-occluding activity (27)(28)(29)(30). The ⑀ amino group of this lysine occupies the external entrance to the deep pore resembling a tethered K ϩ ion (27). It has been shown that a high occupancy of K ϩ binding sites in the pore may repel scorpion toxins from binding to Kv channels. In agreement with this observation was that an increase in K ϩ ions readily interfered with 125 I-KTX binding to Chi IV (Fig. 6A). Half-maximal blockade of 125 I-KTX binding to Chi IV was obtained at ϳ55 mM K ϩ . By contrast, much higher concentrations of Na ϩ were required to weaken 125 I-KTX binding to Chi IV (Fig. 6B).

DISCUSSION
In this paper, we used KTX to study the scorpion toxin receptor site of Kv1.3 channels. Previous structure-function studies (9 -11, 27-30) used KTX or AgTX2 to characterize the scorpion toxin receptor site of Kv1.3 and other Kv channels related to the superfamily of Shaker K ϩ channels (31). AgTX2 and KTX are closely related scorpion toxins differing by only three amino acids. Therefore, it is likely that the interactions of AgTX2 and KTX with K ϩ channels are qualitatively similar, but quantitative differences may exist depending on the spe- cific pair of toxin and K ϩ channel (9,25). It has been shown that the AgTX2 receptor site can be transferred from the mammalian Kv1.3 channel to another, AgTX2-insensitive Kv channel (Kv2.1) by transferring the S5-S6 linker of Kv1.3 (25). We now show that the scorpion toxin receptor site of Kv1.3 can be transferred to KcsA channels. Although the complete S5-S6 linker of Kv1.3 could not be transferred, the transfer of specific parts of the S5-S6 linker was sufficient to generate KcsA-Kv1.3 chimeras with a scorpion toxin receptor site of very high affinity. In agreement with numerous indirect structure-function studies, the biochemical studies with purified Kv1.3-KcsA chimeras showed that KTX binding only took place when the purified protein resembled tetramers. Monomeric chimeras had no binding activity. These results provide direct biochemical evidence that a tetrameric channel structure is necessary for forming a functional scorpion-toxin receptor site.
Previously, mutational analyses have shown that modifications of several residues of the S5-S6 linker region of Kv channels may influence scorpion toxin binding. Our pharmacological studies showed that Chi IV and Chi V had a high affinity for scorpion toxins and were insensitive to other toxins like apamin, PO5, and ␣-DTX. This suggests that the specificity of the scorpion toxin receptor may depend on the nature of particular subregion I residues. Subregion I sequences are highly variable among Kv channels. Therefore, it is likely that differences in the subregion I sequences may be correlated with the different toxin sensitivities of the various Kv channels. This suggests that subregion I constitutes an important molecular determinant of the pharmacological profile of K ϩ channels. Three specific modifications of subregion I of KcsA-Q58A, T61S, and R64D produced a low affinity AgTX2 receptor site to which AgTX2 bound with a K d of 0.6 M (2). Despite different experimental conditions, it was apparent from our results that the binding affinities of scorpion toxins to KcsA-Kv1.3 chimeras were considerably higher. 125 I-KTX bound to Chi IV with a K d value of 0.3 nM, and AgTX2 displayed in competition experiments an IC 50 -value of 6.4 nM. The affinity of AgTX2 to Chi IV is similar to the one for Shaker K ϩ channels (K d ϭ 1 nM) (29). It has been shown that the presence of G58 (see Table I) at the equivalent position in Kv1.3 (9) or Kv2.1-Kv1.3 chimeras (25) is an important determinant of both AgTX2 and KTX affinity. It is likely that Gly-58 represents an important interaction site between KcsA-Kv1.3 chimeras and scorpion toxins like AgTX2 and KTX. This may also explain the considerably higher affinity of AgTX2 for Chi IV than for KcsAQ58A/T61S/R64D. A visual inspection of docking KTX to the KcsA-Kv1.3 interaction surface supports our proposition (see below).
In addition to subregion I, we found that the modification Y82H in subregion III of KcsA-Kv1.3 chimeras dramatically increased KTX affinity. The IC 50 (K d ) value for binding 125 I-KTX to Chi IV and to Chi VIII were 370 pM and 4 pM, respectively (see Table II). In contrast, a similar modification of Tyr to His at the equivalent position in the S5-S6 linker of Kv1.3 channels was not very significant for KTX affinity and even slightly decreased the KTX affinity of the modified Kv1.3 channels (9). The discrepancy suggests that the vestibules of KcsA-Kv1.3 and of Kv1.3 channels have some different properties. Visual inspection of the KcsA crystal structure showed that the side chains of Tyr-78 and Tyr-82 from adjacent subunits are so close to each other to allow for an intersubunit stacking interaction. Apparently, this interaction contributes to the stability of the tetrameric KcsA structure. In Chi VII and Chi VIII, Tyr-82 had been replaced by histidine. It is likely that the Tyr-78/His-82 intersubunit stacking interaction is energetically weaker, leading to a less compact channel structure. This may result in a better accessibility of KTX to the scorpion toxin receptor site. This hypothesis may also explain why stable Chi VII and Chi VIII tetramers could not be purified in solubilized form like Chi IV and Chi V.
A topological model has recently been proposed for the KTX binding site of Kv1.3. The model was developed using the NMR structure of KTX as a caliper in combination with molecular biological data on important pairs of amino acid side chains participating in Kv1.3 channel-KTX interaction (9,30). More recently, AgTX2 was docked onto the KcsAQ58A/T61S/R64D channel structure using energetic data borrowed from Shaker K ϩ channel studies (2). We used the available information in combination with our binding data to assist the docking of KTX to KcsA-Kv1.3 chimeras. Details of the structural model are available on the Internet. The distance between KTX residues Arg-24 and Arg-31 (33Å) is complementary to the one between two Asp-64 residues located at opposite subunits at the rim of the K ϩ channel vestibule. The equivalent residues in Shaker channels (Asp-431) and in Kv1.3 (Asp-386) have been found to electrostatically interact with toxin residues Arg-24 and Arg-31, respectively (11). KTX was docked to Chi VIII by guiding KTX-Lys-27 into the center of the channel pore bringing KTX-Arg-24 and -Arg-31 close to two Asp-64 channel residues by rotating the toxin around the central pore axis. As for the AgTX2/KcsAQ58A/T61S/R64D pair (2), KTX seems to fit per-  fectly into the vestibule of the KcsA-Kv1.3 chimeras. Strong contacts have been proposed for the AgTX2/KcsAQ58A/T61S/ R64D off-center residue pairs (AgTX2-Gly-10, Ala-58), (AgTX2-Arg-24, Asp-64), (AgTX2-Phe-25, Leu-81). Our modeling of the KcsA-Kv1.3 interface suggests additional contacts between KTX and the KcsA-Kv1.3 vestibule. Most importantly, Gly-58 seems to be in close contact to KTX-Phe-25 and Gly-58 of the opposite subunit to KTX-Arg-31. In agreement with the experimental binding data, this suggests that Gly-58 is an important residue in Kv channels for obtaining high affinity scorpion toxin receptor sites. Also, KTX-Gly-10 seems to be near Pro-55, Thr-56, Ser-57, and KTX-Arg-31 near Ser-61, i.e. near subregion I residues that may contribute to the specificity of the interaction between toxin and toxin receptor.
In conclusion, the scanning of the M1-M2 linker of KcsA has allowed definition of sequences that can be exchanged with equivalent ones of Kv1.3. Recombinant chimeric KcsA-Kv1.3 channels were expressed and purified from E. coli as tetramers. They displayed a high affinity for scorpion toxins. The results demonstrated that the complete scorpion toxin receptor site could be transferred from Kv1.3 to KcsA. The data strongly support the previous notion that the outer pore structures of K ϩ channels may have been conserved from bacteria to human. The combination of structural modeling and functional data for the interaction of toxins with the K ϩ channel vestibule can be exploited to advance our knowledge in K ϩ channel pharmacology. Finally, the possibility of transferring the linker regions of other Kv channels, e.g. Kv1.4 and Kv1.5, should enable us to study the structure of the pore region of those channels in detail as well as the structure of other toxin receptor sites.