Structural Differences of Bacterial and Mammalian K+Channels*

Using a peptide toxin, kaliotoxin (KTX), we gained new insight into the topology of the pore region of a voltage-gated potassium channel, mKv1.1. In order to find new interactions between mKv1.1 and KTX, we investigated the pH dependence of KTX block which was stronger at pH o 6.2 compared with pH o 7.4. Using site-directed mutagenesis on the channel and the toxin, we found that protonation of His34in KTX caused the pH o dependence of KTX block. Glu350 and Glu353 in mKv1.1, which interact with His34 in KTX, were calculated to be 4 and 7 Å away from His34/KTX, respectively. Docking of KTX into a homology model of mKv1.1 based on theKcsA crystal structure using this and other known interactions as constraints showed structural differences betweenmKv1.1 and KcsA within the turret (amino acids 348–357). To satisfy our data, we would have to modify theKcsA crystal structure for the mKv1.1 channel orienting Glu350 7 Å and Glu353 4 Å more toward the center of the pore compared with KcsA. This would place Glu350 15 Å and Glu353 11 Å away from the center of the pore instead of the distances for the equivalentKcsA residues with 22 Å for Gly53 and 15 Å for Gly56, respectively. Bacterial and mammalian potassium channels may have structural differences regarding the turret of the outer pore vestibule. This topological difference between both channel types may have substantial influence on structure-guided development of new drugs for mammalian potassium channels by rational drug design.

Voltage-gated potassium channels guide fundamental biological processes such as electrical signaling, osmotic balance and signal transduction (1). Peptide toxins from scorpions, snakes, and sea anemones, which inhibit ion conduction through potassium channels by binding within the outer pore region, have been used to characterize physiological significance, localization, and structural aspects of these channels (2). Charybdotoxin (CTX) 1 has been used to characterize the region of potassium channels bearing both the CTX receptor and the ion conduction pathway and to demonstrate the tetrameric stoichiometry of the channel (3,4). Estimations of the topology of the external vestibule of various potassium channels were successful because the structure of the peptide toxins, which was structurally defined by NMR studies, reports the complementary shape of the binding partner (5,6). Structural features of different channel types were announced by specific binding properties of peptide toxins (7). Our knowledge about structural characteristics of potassium channels was confirmed and extended through data from the bacterial Streptomyces KcsA channel based on crystallographic studies (8). Therefore, up to the availability of x-ray data for each potassium channel, further application of peptide toxins as molecular calipers to investigate the architecture of structurally non-defined targets appears to be reasonable.
Several toxins display a blocking affinity dependent on the extracellular pH (pH o ) but until now only a decreased block at acid pH o was known; both with peptide toxins (9 -11) and with other blockers working on potassium channels like TEA (12) but also working on sodium channels like tetrodotoxin and saxitoxin (13). The reduced CTX affinity of the F425H mutant Shaker channel and of the wt Kv1.3 channel (9, 11) as well as the lowered TEA affinity of mKv1.1 channels (12) at low pH o was caused by protonation of histidine residues within the outer vestibule of the respective channel proteins. Affinity of tityustoxin to the squid Kv1 channel SqKv1A, which contains a histidine at Shaker position 425, is also reduced at acid pH o (14). In contrast, in mKv1.1 and H404T mutant mKv1.3 channels, we observed a better KTX block at low pH o . In this paper we present evidence that protonation of a histidine in KTX, which interacts with negatively charged amino acids in the outer pore region of mKv1.1 and H404T mutant mKv1.3 channels, is responsible for the higher KTX affinity of these channels. Using electrostatic compliance we found distances of 4 and 7 Å between Glu 353 and Glu 350 in mKv1. 1 and His 34 in KTX, respectively. Our data define the position of the turret (amino acids 348 -357) in Kv1.1 and also Kv1.3 and imply structural differences between mammalian and bacterial K ϩ channels in that region. These results help refine our picture of the spatial arrangement of amino acid residues in the outer pore region of Kv1.1 and Kv1.3 that might facilitate therapeutic drug design.

MATERIALS AND METHODS
Cells-All experiments were carried out on single cells of a rat basophilic leukemia cell line, RBL cells (15). Cells were obtained from the American Type Culture Collection (Rockville, MD). The cells were maintained in a culture medium of Minimal Essential Medium with Earle's salts supplemented with 1 mM L-glutamine and 10% heatinactivated fetal calf serum in a humidified, 5% CO 2 incubator at 37°C. Cells were plated to grow non-confluently onto glass 1 day prior to use for injection and electrophysiological experiments (16).
Solutions-The experiments were done at room temperature (21-25°C). Cells measured in the whole cell configuration were normally bathed in mammalian Ringer's solution containing, in mM: 160 NaCl, 4.5 KCl, 2 CaCl 2, 1 MgCl 2, and 10 X (with X either Tris, HEPES, MES, or citrate), with an osmolarity of 290 -320 mOsm. The pH was adjusted to 5.5, 6.2 (X: citrate); to 6.2, 6.6, 6.8, 7.0 (X: MES), to 7.0, 7.4, 7.8 (X: HEPES), and to 7.8, 8.2 (X: Tris) with NaOH. No differences in current were seen comparing mammalian Ringer's solution containing either citrate or MES at pH o 6.2, either MES or HEPES at pH o 7.0, and either HEPES or Tris at pH o 7.8 as described earlier (12). A simple syringedriven perfusion system was used to exchange the bath solutions in the recording chamber. The internal pipette solution for the whole-cell recordings contained (in mM): 155 KF, 2 MgCl 2, 10 HEPES, 10 EGTA, adjusted to pH 7.2 with KOH, with an osmolarity of 290 -320 mOsm.
The lyophilized peptides were stored at Ϫ70°C. Stock solutions of 10 -100 M were made with mammalian Ringer's solution containing 0.1% bovine serum albumin. The final dilutions were prepared shortly before the experiment.
Electrophysiology-Experiments were carried out using the wholecell recording mode of the patch-clamp technique as described before (16). Electrodes were pulled from glass capillaries (Clark Electromedical Instruments, Reading, United Kingdom) in three stages, coated with Sylgard (Dow Corning, Seneffe, Belgium), and fire-polished to resistances measured in the bath of 2.5-4 M⍀. Membrane currents were recorded with an EPC-9 patch-clamp amplifier (HEKA Elektronik, Lambrecht, Germany) interfaced to a Macintosh computer running acquisition and analysis software (Pulse and PulseFit). Capacitative and leak currents were subtracted using the P/10 procedure. Series resistance compensation (Ͼ80%) was employed if the current exceeded 1 nA. The holding potential in all experiments was Ϫ80 mV.
Expression-pBSTA plasmids containing the entire coding sequence of the mKv1.1 wild type (wt) gene, and the pSP64T plasmids containing the sequences for the H404T mutant mKv1.3 channel (18) (a generous gift from Dr. K. George Chandy) were linearized with PstI and EcoRI, respectively and in vitro transcribed with the T7 (mKv1.1) and SP6 (mKv1.3 H404T) Cap-Scribe system (Roche Molecular Biochemicals). The resulting cRNA was phenol/chloroform-purified and could be stored at Ϫ75°C for several months.
Injection-The cRNA was diluted with a fluorescent FITC dye (0.5% FITC-dextran in 100 mM KCl) to a final concentration of 1 g/l. RBL cells were injected with the cRNA/FITC solution filled in injection capillaries (Femtotips ® ) using an Eppendorf microinjection system (Micromanipulator 5171 and Transjector 5246). In the visualized cells, specific currents could be measured 3-6 h after injection.
Thermodynamic Mutant Cycle Analysis-Thermodynamic mutant cycles assist in studying of coupling energies between pairs of amino acid in a protein-protein complex. The dimensionless ⍀ value, which indicates the interaction strength of a given channel-toxin pair, was calculated as shown before (9,19). The change in coupling energy, ⌬⌬G, for the channel-toxin pairs was calculated using the formula ⌬⌬G ϭ kT ln⍀ as described earlier (7). The distances between this pair of residues was estimated based on the studies of Schreiber and Fersht (20) and Hidalgo and MacKinnon (19), assuming that ⌬⌬G values Ն0.5 kcal mol Ϫ1 correspond to an inter-residue distance of Յ5 Å.
Electrostatic Compliance Measurements-For the estimation of the distances between the charges at KTX position His 34 and positions Glu 350 and Glu 353 in the mKv1.1 channel (equivalent to Asp 375 and Ser 378 in mKv1.3), we used the method of electrostatic compliance (5,9,12). Protonation of His 34 in KTX creates a higher effective KTX concentration at the receptor site of the channels compared with the bulk concentration. Calculation of the potential causing the higher effective KTX concentration and the appropriate distances between the interacting amino acids was performed as described by Bretschneider et al. (12).
Docking-KTX was docked into a homology model of mKv1.1 based on the x-ray data of the KcsA channel. To create a homology model of mKv1.1, we exchanged the KcsA amino acids with the equivalent mKv1.1 sequence. Using the KcsA structure as template (Protein Data Bank code 1BL8; Ref. 8), we calculated the mKv1.1 homology structure as described before (7). The protein backbone of the mKv1.1 homology model was identical to the KcsA backbone. Docking of KTX into the mKv1.1 homology model was performed using the interactions between mKv1.1 and KTX as described before (7,21). Additionally, the well characterized interaction between the terminal amino group of the central Lys 27 in KTX and the C ␤ atoms of Tyr 375 in the GYGD motive was exploited (6,21). The docking configuration and the distances between interacting amino acids were analyzed using RasMol 2.7.1.

RESULTS
To initially characterize the effect of extracellularly applied KTX on current through mKv1.1 channels, we measured whole cell currents in response to depolarizing steps from Ϫ80 to ϩ40 mV in the absence and presence of KTX (Fig. 1A, left). 60 nM KTX (Bachem) added to the external mammalian Ringer solution resulted in a peak current reduction of about 50%. To quantify the KTX affinity of mKv1.1, we measured, in additional experiments, the effect of different KTX concentrations on current through mKv1.1 channels and plotted the normalized peak currents against the applied KTX concentration (Fig.  1B, open triangles). A Hill equation was fitted to the data with a Hill coefficient of 1, indicating a 1:1 stoichiometry suggesting that one KTX molecule is sufficient to block one mKv1.1 channel. The fit gave a dissociation constant K d at pH o 7.4 of 62 nM (n ϭ 25), which is in agreement with earlier reports (22). The current reduction by KTX was fully reversible upon washout (data not shown). Application of 60 nM KTX (Bachem) to the bath solution at pH o 6.2 reduced the current much more than 50% (Fig. 1A, right). The determination of the KTX affinity of mKv1.1 at pH o 6.2 was made as for pH o 7.4 and showed a K d of 19 nM (n ϭ 28). Fig. 1B clearly demonstrates that the doseresponse curve for KTX at pH o 6.2 (open squares) to block currents through mKv1.1 channels was shifted toward lower concentrations. Therefore, lowering pH o from 7.4 to 6.2 increased the KTX affinity to mKv1.1 ϳ 3.3 times. The same relationship between K d pH o 7.4 and 6.2 was revealed with the recombinantly made KTX which exhibited an ϳ4 times higher affinity to mKv1.1 channels compared with KTX from Bachem ( Table I) probably because of higher purity of the recombinant KTX.
We wanted to know whether this observed higher KTX affinity of mKv1.1 at low pH o is caused by protonation of an amino acid in the channel protein or by protonation of a residue in KTX. To rule out effects of His 355 of the mKv1.1 channel (protonation of His 355 results in a decrease of the TEA affinity at low pH o (see Ref. 12)), we tested KTX on mKv1.1 mutant channels with a non-protonatable glycine instead of a histidine at position 355 (H355G/mKv1.1). KTX, however, seemed to bind to this H355G mutant mKv1.1 channel in an irreversible manner since currents could not be recovered after application of KTX even after wash-out for more than 1 h (data not shown).
To avoid possible problems with rundown versus block, we tested KTX on H404T mutant mKv1.3 channels. The H404T/ mKv1.3 channel also contains no protonatable histidine in the outer pore region ( Fig. 2A). Instead of the histidine, it has a glycine at the corresponding position 380; the histidine residue of the wt mKv1.3 channel at position 404 (equivalent to position 379 of mKv1.1) is changed to a threonine in this mutant of the channel ( Fig. 2A). We did identical experiments with KTX from Bachem for H404T mutant mKv1.3 channels as we did for mKv1.1 channels. KTX displayed an increased affinity to H404T mutant mKv1.3 channels at low pH o despite the absence of protonatable histidines in the outer pore region. The dose-response curve for pH o 6.2 was shifted toward lower concentrations compared with pH o 7.4 ( Fig. 1C) and revealed K d values of 0.24 nM (n ϭ 28) and 0.10 nM (n ϭ 28) for pH o 7.4 and 6.2, respectively.
In order to find out whether voltage-dependent binding of KTX could effect the pH o dependence of KTX block, we investigated voltage dependence of KTX block on current through H404T mutant mKv1.3 channels (Fig. 3). Crest et al. (23) and Mourre et al. (24) reported that the KTX block of calcium-dependent potassium channels with large conductance (MaxiK channels) and of voltage-dependent potassium channels is voltage-independent at physiological pH o , but there were no studies testing a possible voltage dependence of KTX block at acid  Fig. 3 (A and C) to conductances by dividing the peak currents through the driving force for [K ϩ ] (E Ϫ E K ). At pH o 7.4 application of 50 pM recombinant KTX reduced maximum conductance to 68% of the values obtained in the absence of KTX without changing activation of the channels. The suppression of the conductance was identical at membrane potentials more positive than 0 mV up to ϩ100 mV, indicating that the KTX block on H404T mutant mKv1.3 channels was voltage-independent in that potential range at physiological pH o . At pH o 6.2, application of 50 pM recombinant KTX reduced maximum conductance g K(max) significantly higher compared with pH o 7.4 to only 40% of the control value. The KTX block was again voltage-independent; membrane potentials more positive than ϩ20 mV did not modify the blocking strength of KTX (Fig. 3D). Since the KTX effect seemed to be independent of the applied voltage for depolarizations between 0 and ϩ100 mV at both pH o values, it seemed sufficient for later experiments to evaluate the KTX block at one potential, in our case ϩ40 mV, using the ratios of the peak currents (see also Fig. 1C (Table I).
To find out further indications for our hypothesis that protonation of an amino acid in KTX is responsible for the higher affinity at pH o 6.2, we investigated the pH o dependence of block by other peptide toxins like CTX and NTX on current through the H404T mutant mKv1.3 channels. CTX blocked the current through H404T/mKv1.3 with equal strength at pH o 7.4 as well as at pH o 6.2, which can be seen by the identical K d values shown in Table I (Table I). We therefore concluded that the increase in affinity by lowering pH o is specific for KTX since CTX and NTX did not show this behavior.
By comparing the amino acid sequence of the toxins, we found that KTX possesses a unique histidine at position 34 not present in CTX and NTX (Fig. 2B). In order to find out whether His 34 might be the protonatable amino acid in KTX, we used KTX 3 which has an Asp at position 34 whose protonation should not vary between pH o 7.4 and 6.2. KTX 3 exhibited nearly the same affinity to H404T mutant mKv1.3 channels at pH o 7.4 and 6.2 with K d values of 150 Ϯ 15 pM (n ϭ 6) and 170 Ϯ 25 pM (n ϭ 6) at pH o 7.4 and 6.2, respectively ( Table I). The loss of the pH o -dependent KTX effect in KTX 3 indicated that His 34 of KTX might be responsible for this effect because of the lack of a histidine residue at position 34 in KTX 3 .
Additional experiments with KTX-Lys 27 mutants on H404T mutant mKv1.3 channels were made to exclude Lys 27 of KTX as the protonatable amino acid of KTX. The KTX mutants K27Dap, K27Dab, and K27ThLys contained non-natural positively charged lysine analogs of varying side chain lengths (Dap, 2.5 Å from C ␣ ; Dab, 3.8 Å; ThLys, 7.7 Å; natural lysine, 6.3 Å), whereas in K27A and K27Nle the neutral alanine and the non-natural neutral norleucine replaced lysine (6,9). If Lys 27 was somehow involved in the pH o dependence of KTX block, we would expect a pH o -independent block with the noncharged amino acids at position 27 and perhaps an altered pH o dependence with the shorter and longer non-natural lysine analogs Dap, Dab, and ThLys. Table I shows that all KTX-Lys 27 mutants bound in a pH o -dependent manner to H404T mutant mKv1.3 channels with a higher affinity at low pH o , revealing that Lys 27 is not responsible for pH o dependence of KTX block.
To further substantiate that protonation of His 34 in KTX is responsible for the pH o dependence of KTX block on current through mKv1.1 and H404T mutant mKv1.3 channels, we made mutants of this peptide toxin by recombinant methods (17). His 34 of KTX seemed to be a good candidate for the protonatable amino acid in KTX, as indicated by the investigations with H404T/mKv1.1 channels and CTX, NTX, and KTX 3 . Therefore, we investigated the KTX mutants H34A and H34K. In addition, we also replaced Arg 31 and Lys 32 in KTX with neutral amino acids to examine whether these two positions in KTX were also involved since earlier work indicated that these two positions participate in important interactions (9,27). We therefore tested the pH o dependence of block of all four KTX mutants with H404T mutant mKv1.3 channels (Table I) Table I). Additional replacement of Arg 31 with alanine in the H34A/R31A mutant KTX exhibited no further effect (Table I). Replacement of His 34 in KTX by lysine led to the same effect as substitution by alanine concerning the pH o -independent block but exhibited a higher affinity to H404T/mKv1.3 than H34A and wt KTX (Fig. 4A (bottom) and Table I). Therefore H34A, H34K, and R31A/H34A mutants of KTX blocked currents through H404T mutant mKv1.3 channels in a pH o -independent manner, whereas KTX-K32A block is influenced by pH o . These results suggest that only the histidine residue at position 34 of KTX is the protonatable amino acid of KTX that causes pH o dependence of block while Arg 31 and Lys 32 of KTX do not seem to play a role in this effect.
If protonation of His 34 in KTX is the reason for the better block of mKv1.1 and H404T mutant mKv1.3 channels with lowering pH o , H34A and H34K mutants of KTX should (a) have a pH o -independent block as shown before (Fig. 4A and Table I) and (b) represent the fully protonated (H34K) and the fully unprotonated (H34A) form of wt KTX responsible for the pH o dependence of block. Fig. 4B shows the titration of wt KTX and the mutants H34K and H34A from pH o 5.5 to pH o 7.8 examined on H404T mutant mKv1.3 channels. KTX-H34K exhibited the same K d of ϳ16 pM in the investigated pH o range, and KTX-H34A had identical K d values of ϳ100 pM at 6 different pH o values between pH o 5.5 and 7.4. Linear regressions through the data points of both KTX mutants had slopes of zero, indicating that these KTX mutants cannot be titrated in the pH o range used. This also suggests that histidine at position 34 of KTX is protonatable. The affinity of wt KTX to H404T mutant mKv1.3 channels decreased with increasing pH o . The fit through the data points using the Hill equation gave a K d(min) of 18 pM at pH o 5.5 and a K d(max) of 60 pM at pH o 7.8 and a pK a of 6.5, which represents the pH of the half-maximum protonation. The Hill coefficient of 1.7 suggests that more than one negatively charged amino acids of the channel protein could sense His 34 of KTX. The K d(min) of 18 pM wt KTX at pH o 5.5 is the expected value with good agreement with the K d of KTX-H34K, which stands for the fully protonated form of wt KTX. In contrast, KTX-H34A, which represents the unprotonated wt KTX, had a lower affinity to H404T/mKv1.3 channels with 100 pM compared with K d(max) of wt KTX with 60 pM at pH o 7.8. The reason for this difference might be a non-ionic interaction between the unprotonated His 34 of wt KTX and an amino acid of the channel that is abolished by the substitution with alanine. A more conservative replacement with glutamine or asparagine might solve this question (28,29). Since the data confirmed our predictions apart from this difference between K d(max) of wt KTX and K d of KTX-H34A, we concluded that indeed protonation of His 34 in KTX caused the pH o dependence of block.
To determine which amino acid(s) of the channel protein interacted with the protonatable His 34 of KTX, we tested some mutants of the mKv1.1 and the H404T/mKv1.3 channels, respectively. Six negatively charged residues in the mKv1.1 pore region could be candidates for the interaction with His 34 in KTX: Glu 348 , Glu 350 , Glu 351 , Glu 343 , Asp 361 , and Asp 377 (Fig.  2A). The mKv1.3 channel contains all of these negatively charged amino acids except one at position 378, which corresponds to position 353 of mKv1.1 with an uncharged serine instead of the glutamate in mKv1.1. We did not investigate the Asp 377 of mKv1.1 (Asp 402 in mKv1.3) of the GYGD motif, since we had found that the shorter lysine analogs Dap and Dab at position 27 of KTX that were shown to interact with Asp 402 in mKv1.3 (6) still blocked in a pH o -dependent manner (Table I). In addition, from the KcsA crystal structure data, we concluded that the negative charges at positions 348 -353 of mKv1.1 would be the best candidates for the interacting amino acids. We expected that replacing negatively charged amino acids in  (Table I). Nevertheless, the ratio of (K d pH o 7.4/K d pH o 6.2) was reduced to 2.4, the same ratio as for H404T mutant mKv1.3 channels. Therefore, we concluded that (a) the lack of a negative charge at position 378 in mKv1.3 (equivalent to 353 in mKv1.1) reduced the (K d pH o 7.4/K d pH o 6.2) ratio, and (b) that Glu 353 in mKv1.1 is one of the negatively charged residues in mKv1.1, which interacts with the protonatable amino acid in KTX. This result was confirmed by the introduction of a negatively charged amino acid at mKv1.3 position 378 (P377A/S378E/H404T and S378E/H404T mutant mKv1.3 channels) exhibiting the stronger improvement of KTX block with lowering pH o like mKv1.1 channels (Table I). Therefore, protonation of His 355 in mKv1.1 (12) seemed not to cause the stronger pH o effect of mKv1.1 compared with H404T/mKv1.3.
As additional candidate for an interacting amino acid of the channel, we investigated the D361N mutant mKv1.1 channel lacking the negative charge at position 361 (equivalent to position 386 in mKv1. 3). As can be seen from The knowledge about the effective KTX concentration enabled us to determine the potential , which caused the higher effective KTX concentration at low pH o . for wt mKv1.1 was Ϫ30.4 mV, for E353S/mKv1.1 Ϫ22 mV, and for the calculated mutant E350A/mKv1.1 Ϫ8.6 mV. These values for were correct if only one Glu 350 and Glu 353 , respectively, of the channel would sense protonation of His 34 in KTX. Because it was not possible to determine the number of interacting channel subunits without investigating heterotetrameric channels, we calculated the potentials for one, two, and four interacting channel subunits by dividing the above values by these numbers of subunits. Because of the distance dependence of electrostatic interactions, we were able to determine the distance between interacting amino acids of mKv1.1 and KTX. For this calculation the degree of charge that was caused by protonation of His 34 has to be known and was calculated with 67% protonation at pH o 6.2 (12). We estimated distances of 4, 6 -6.5, or 8.5 Å between His 34 /KTX and one, two, or four interacting Glu 350 / mKv1.1 and distances of 6.5-7, 9.5, or 12.5 to 14 Å between His 34 /KTX and one, two, or four interacting Glu 353 /mKv1.1.
Corresponding calculations were performed for interacting residues in H404T/mKv1.3 and His 34 in KTX. We estimated the same distances for His 34 /KTX and Asp 375 /mKv1.3 as for His 34 / KTX and Glu 350 /mKv1.1. The distance between His 34 /KTX and S378E/mKv1.3 also seemed to be the same as for His 34 /KTX and Glu 353 /mKv1.1.
Furthermore, we calculated the distance between the only interacting residue in H404T/mKv1.3 channels, Asp 375 , and His 34 /KTX by electrostatic compliance using KTX 3 and KTX-H34K. By application of the known charge difference of two between Asp 34 in KTX 3 and Lys 34 in H34K (Fig. 2) and the K d values (Table I), we calculated a local potential 89 of Ϫ55.9 mV. Therefore, we found distances between Asp 375 /mKv1.3 and His 34 /KTX of 4 -4.5 Å for the interaction with one channel subunit, of 6.5-7 Å for the interaction with two subunits, and of 9 Å for the interaction with all four channel subunits, respectively. The agreement of this electrostatic compliance calculation using the known charge difference between H34K and KTX 3 at position 34 with the electrostatic compliance using the degree of protonation at His 34 /KTX enabled us to confirm the above determined pK a of His 34 /KTX.
Estimation of the distance between Asp 375 /mKv1.3 and His 34 /KTX was also carried out by mutant cycle analysis. The KTX block of the channel mutants D375A/H404T and D375A/ H404T was tested at pH o 7.4 and 6.2 (Table I)  To refine the docking conformation of mKv1.1 with KTX, we confirmed the known interacting pair of Gly 380 /Kv1.3 and Arg 31 /KTX for the mKv1.1 channel (9). We investigated wt and H355K mutant mKv1.1 channels with the KTX mutants H34A versus R31A/H34A. wt mKv1.1 exhibited an ϳ10 times higher affinity to H34A compared with R31A/H34A with K d values of 38 Ϯ 4 nM and 410 Ϯ 40 nM, respectively. In contrast, R31A/ H34A blocked current through H355K mutant mKv1.1 channels only 2 times more weakly than the H34A mutant KTX with K d values of 1120 Ϯ 180 nM and 620 Ϯ 80 nM for R31A/ H34A and H34A block, respectively. Application of mutant cycle analysis revealed a ⌬⌬G value of 1.05 kcal mol Ϫ1 , which indicates a proximity of Յ5 Å between the interacting residues His 355 in mKv1. 1  mKv1.1 and mKv1.3 and maybe other mammalian voltagegated potassium channels seem to be structurally different from KcsA in the turret. DISCUSSION In this report we demonstrate that the pH o dependence of KTX block on current through mKv1.1 and H404T mutant mKv1.3 channels is caused by protonation of the histidine at KTX position 34. We could define two interacting positions of the channel proteins: 350/353 in mKv1.1 and 375/378 in H404T mutant mKv1.3 channels, respectively. Furthermore, we could confirm an equivalent interaction between Arg 31 in KTX and His 355 in mKv1.1 known for Kv1.3 and KTX (9). Comparing experimentally derived and docking distances between interacting residues of mKv1.1 and KTX indicated that the turret of mKv1.1 is oriented more toward the center of the pore compared with KcsA whose structure was used as template for the mKv1.1 homology model. In summary, we present new pairs of interacting residues between KTX and the potassium channels mKv1.1 and mKv1.3 to increase the knowledge about the outer pore region of these channels, which might be advantageous in designing novel drugs of higher affinity and specificity. This is the first report of a peptide toxin binding to potassium channels whose affinity to the channel's receptor was increased in low pH o . Deutsch et al. (10) reported about an enhanced CTX binding to the voltage-gated K ϩ channel in human T lymphocytes with increasing pH o . The reduced CTX and KTX block of wt Kv1.3 in acid pH o was shown to be a consequence of protonation of His 404 in wt Kv1.3 channels, which repelled the positively charged toxins (9). Replacing His 404 in Kv1.3 with non-protonatable threonine caused a pH o -independent CTX block and an increased KTX block corresponding to our results (9). The diminished CTX block of F425H mutant Shaker channels at low pH o was also a result of protonation of F425H of this Shaker mutant (11). Diminished sensitivity of tityustoxin to the squid potassium channel SqKv1A was also attributed to the protonation of a histidine at position 351 (equivalent to Shaker position 425) of the channel (14).
Other blocking agents like TEA also exhibited a pH o dependence of block. Diminished TEA block on current through mKv1.1 channels was caused by protonation of His 355 in mKv1.1, which induced an electrostatic repulsion of the positively charged TEA (12). Weaker sensitivity of saxitoxin and tetrodotoxin to block voltage-gated sodium channel at low pH o was also discussed to be a result of protonation of the receptor (1, 13), but there is also evidence that the decreased block with increasing [H ϩ ] o or increasing concentration of di-or trivalent cations is caused by neutralization of negative surface charges lowering the saxitoxin and tetrodotoxin concentration at its receptor site (30).
The reduced blocking strength of the above toxins was caused by protonation of the receptor or by changing the surface potentials. None of these examples revealed altered affinity because of protonation of the blocker itself. Therefore, our observations regarding the pH o dependence of KTX block are different in two respects. First, the KTX block of mKv1.1 and H404T mutant mKv1.3 channels is increased with decreasing pH o . Second, this higher affinity is the result of protonation of an amino acid in the peptide toxin, whereas protonation of the KTX receptor seemed not to play a role in causing this effect.
pH o independence of CTX and KTX 3 block on current through H404T/mKv1.3 indicated that protonation of His 34 / KTX could cause the pH o dependence of KTX block. Our hypothesis was proven by the pH o -independent KTX-H34A and KTX-H34K block. Titration of wt and mutant KTX revealed that H34K behaved like the fully protonated wt KTX. On the other hand, H34A, which should behave like fully unproto-nated wt KTX, certainly blocked H404T/mKv1.3 in a pH oindependent manner, however, with a lower affinity compared with wt KTX at pH o 7.8 with a degree of protonation less than 5%. The difference in affinity between wt KTX at high pH o and H34A might be caused by an interaction between His 34 /KTX and residues of the channel via the nitrogen atoms, which is destroyed by the substitution with alanine. More conservative exchanges of His 34 with glutamine or asparagine might have the same blocking strength as unprotonated wt KTX, as shown for several histidine interactions in other proteins (28,29).
Titration of wt KTX also indicated a pK a of 6.5 for His 34 /KTX. Gairi et al. (27) determined the pK a of His 34 in unbound KTX by NMR investigations with a value of about 5.2. These two observations seem to be in contrast; however, we investigated the pK a of KTX that was interacting with H404T/mKv1.3 channels and not while unbound in solution. The stronger interaction of protonated His 34 /KTX with negatively charged residues of the H404T channel might facilitate the protonation of His 34 and therefore raise its pK a . Similar observations with lowered pK a values (20) and elevated pK a values (31) due to unfavorable and advantageous interactions with the receptor or ligand have been reported previously.
Replacing negatively charged residues in the outer pore vestibule of mKv1.1 and H404T/mKv1.3 or vice versa revealed Glu 350 and Glu 353 in mKv1.1 and Asp 375 in H404T to sense protonation of His 34 in KTX. Since mKv1.1 and mKv1.3 exhibited the same interacting positions and the same strength of interaction, we could transfer results from one channel to the other. Using electrostatic compliance we were able to calculate the distances between the interacting residues of the KTX and both channels (5,9). Application of two mutant cycles refined the docking configuration of KTX and suggested that His 34 / KTX interacted with only one subunit of the channel.
We used these experimental data for a docking of KTX into a mKv1.1 homology model based on the KcsA coordinates (8). The average sequence identity of about 30% between KcsA and mKv1.1 is sufficient to generate homology models using the KcsA structure as template (32). The docking of KTX into the mKv1.1 homology model built on the basis of the KcsA structure exhibited a distance of about 5 Å between His 355 /mKv1.1 and Arg 31 /KTX, which agrees with the experimentally derived distance. The pore diameter of about 30 Å at position 355/ mKv1.1 corresponds to the equivalent positions in Kv1.3 and Shaker channels (9,19). In contrast, the docking distances at mKv1.1 positions 350 and 353 within the turret disagreed from the experimentally obtained distances, suggesting shorter pore diameter at these positions for mKv1.1 compared with KcsA. To get an agreement between the experimental and the docking data, we would have to modulate the KcsA structure within the turret (amino acids 348 -357) guiding Glu 350 7 Å and Glu 353 4 -4.5 Å more toward the center of the pore. Therefore, the shape of the mKv1.1 turret seemed to be distinct from KcsA, particularly in the region near the S5 segment but not in the region close to the pore helix. The spatial arrangement of the turret in mKv1.3 and mKv1.1 seems to be very similar since we found identical interacting positions and the same distances between the interacting positions. Evidence for topological differences between KcsA and voltage-gated potassium channels within the turret is also indicated by distinct toxin affinities of both channel types (33,34). Only the replacement of the complete outer pore region of KcsA with the equivalent region of a mammalian voltage-gated potassium channel could create a similar toxin affinity in KcsA (35).
In conclusion, using KTX as a molecular caliper, we were able to characterize topological features within the turret of the voltage-gated potassium channels mKv1.1 and mKv1. 3. The turret of both channels seems to be structurally different compared with the bacterial KcsA channel. Since the receptor for many potassium channel modulators is located in the outer pore region, this information could aid the rational drug design and therefore accelerate the generation of new and improved drugs working on potassium channel.