Determinants of Apamin and d-Tubocurarine Block in SK Potassium Channels*

Small conductance calcium-activated potassium channels show a distinct pharmacology. Some, but not all, are blocked by the peptide toxin apamin, and apamin-sensitive channels are also blocked by d-tubocurarine. Cloned SK channels (small conductance calcium-activated potassium channel) recapitulate these properties. We have investigated the structural basis for these differences and found that two amino acid residues on either side of the deep pore are the primary determinants of sensitivity to apamin and differential block by d-tubocurarine. Therefore, the pharmacology of SK channels compared with other potassium channels correlates with structural differences in the outer pore region. However, introduction of a tyrosine residue in the position analogous to that which determines sensitivity to external tetraethylammonium for voltage-gated potassium channels endows SK channels with an equivalent tetraethylammonium sensitivity, indicating that the outer vestibules of the pores are similar. The pharmacology of channels formed in oocytes coinjected with SK1 and SK2 mRNAs, or with SK1-SK2 dimer mRNA, show that SK subunits may form heteromeric channels.

Small conductance calcium-activated potassium channels show a distinct pharmacology. Some, but not all, are blocked by the peptide toxin apamin, and apamin-sensitive channels are also blocked by d-tubocurarine. Cloned SK channels (small conductance calcium-activated potassium channel) recapitulate these properties. We have investigated the structural basis for these differences and found that two amino acid residues on either side of the deep pore are the primary determinants of sensitivity to apamin and differential block by d-tubocurarine. Therefore, the pharmacology of SK channels compared with other potassium channels correlates with structural differences in the outer pore region. However, introduction of a tyrosine residue in the position analogous to that which determines sensitivity to external tetraethylammonium for voltagegated potassium channels endows SK channels with an equivalent tetraethylammonium sensitivity, indicating that the outer vestibules of the pores are similar. The pharmacology of channels formed in oocytes coinjected with SK1 and SK2 mRNAs, or with SK1-SK2 dimer mRNA, show that SK subunits may form heteromeric channels.
Small conductance calcium-activated potassium channels (SK channels) 1 are responsible for the slow afterhyperpolarization (sAHP) following an action potential. With sustained stimulus, spike frequency adaptation occurs where the repetitive firing of action potentials is self-limiting, because the depth and time course of the sAHP are extended, preventing the membrane potential from reaching threshold (1)(2)(3)(4). The sAHP demonstrates a distinct pharmacology. In hippocampal interneurons, the sAHP is blocked by the peptide toxin, apamin (5,6), while in pyramidal cells, it is not affected (7). Apaminsensitive sAHPs (8,9) and SK channels (10) are also sensitive to the plant alkyloid, d-tubocurarine (dTC). Therefore, the structural relationship between the channels underlying apamin-sensitive and apamin-insensitive sAHPs was unclear.
Recently, the amino acid sequences of several SK channels have been described (11). Compared with other cloned K ϩ channels, and consistent with their distinct pharmacology and biophsyical properties, SK channels form a separate branch of the potassium channel family tree. They show no clear homology to other K ϩ channels except in part of the pore region.
Heterologously expressed SK2 channels are blocked by apamin with an IC 50 of 60 pM, while highly homologous SK1 channels are not blocked by 100 nM apamin. Also, SK2 channels are more sensitive to dTC than SK1 channels (11). In situ hybridization studies in rat brain showed a good correlation between the pattern of apamin-sensitive SK channel mRNAs, SK2 and SK3, and radiolabeled apamin binding sites, while SK1 mRNA was detected in cell types with apamin-insensitive AHPs (11)(12)(13)(14). Apamin-sensitive SK channels have been implicated in several important physiological processes. In the central nervous system, the sensory motor portion of the inferior colliculis is capable of seizure generating activity, which may be evoked by application of apamin (15). 2 Intracerebroventricular injections of apamin disturbed the circadian cycle and disrupted normal sleep patterns (16). Rats injected with apamin prior to but not following training showed accelerated acquisition rates and retention times of learned tasks (17), and increased levels of c-fos and c-jun mRNAs in the hippocampus (18). In the periphery, apamin application to guinea pig proximal colon blocked neurotensin-induced relaxation and resulted instead in contraction (19). Denervated skeletal muscle (20,21), or skeletal muscle from patients with myotonic dystrophy (22), contained apamin-sensitive SK channels and radiolabeled apamin binding sites, while normal adult skeletal muscle did not. Moreover, the hyperexcitability associated with both of these conditions may be ameliorated by direct application of apamin (21,23,24). These studies suggest that apamin-sensitive SK channels may be effective therapeutic targets.
We investigated the molecular basis for the different pharmacology among cloned SK channels. The results show that two amino acids residing on opposite sides of the deep pore determine the relative sensitivities to both apamin and dTC. Despite the structural and pharmacological differences between SK channels and other potassium channels, introduction of a tyrosine residue in SK1 at the position analogous to that which determines TEA (tetraethylammonium) sensitivity in voltage-gated K ϩ channels (25, 26) endows equivalent TEA sensitivity, suggesting that the architecture of the outer vestibules of these two classes is similar. In addition, expression of SK1-SK2 dimers results in channels which have intermediate apamin and dTC sensitivities, indicating that more than one subunit mediates toxin binding and that SK subunits may coassemble into heteromeric channels.

EXPERIMENTAL PROCEDURES
Molecular Biology-All channel subunits were subcloned into the oocyte expression vector pBF, 3 which provides 5Ј-and 3Ј-untranslated regions from the Xenopus ␤-globin gene flanking a polylinker containing multiple restriction sites. In vitro mRNAs were generated using SP6 polymerase (Life Technologies, Inc.); following synthesis, mRNAs were evaluated spectrophotometrically and by ethidium bromide staining after agarose gel electrophoresis. Chimeras were constructed either by the use of common restriction endonuclease sites or by a method described by Horton et al. (27) in which the chimeric junctions were generated by overlap extension of PCR primers that encoded the desired sequence (chimera 1: 1-338 of SK1 and 350 -581 of SK2; chimera 2: 1-396 of SK1 and 407-581 of SK2; chimera 3: 1-407 of SK2 and 396 -562 of SK1; chimera 4: 1-337 of SK2 and 326 -562 of SK1). Sitedirected mutations were created either by overlap PCR (see above) or by the altered sites method (Promega) as described previously (28). Dimer construction also used the overlap PCR method, introducing a stretch of 10 glutamine residues between the C terminus of SK1 and the N terminus of SK2, as described previously (29). All PCRs were performed using Vent polymerase (New England Biolabs), and the nucleotide sequences of all mutations, chimeric junctions, and PCR products were verified before use (Sequenase, U. S. Biochemical Corp.). Oligonucleotides were from Genosys, and sequence comparisons were performed with the GAP program from the Genetics Computer Group suite.
Electrophysiology-Xenopus care and handling were in accordance with the highest standards of institutional guidelines. Frogs underwent no more than two surgeries, separated by at least 3 weeks, and surgeries were performed using well established techniques. Frogs were anesthetized with an aerated solution of 3-aminobenzoic acid ethyl ester. Oocytes were studied 2-14 days after injection with 0.2-2 ng of mRNA.
Inside-out macropatches were excised into an intracellular solution containing 116 mM potassium gluconate, 4 mM KCl, 10 mM HEPES (pH 7.2, adjusted with KOH) supplemented with CaCl 2 to give free calcium concentration of 5 M; the proportion of calcium binding to gluconate was determined by a computer program (CaBuf) assuming a stability constant for Ca 2ϩ gluconate of 15.9 M Ϫ1 (30). Electrodes were pulled from thin-walled, filamented borosilicate glass (World Precision Instruments) and filled with 116 mM potassium gluconate, 4 mM KCl, 10 mM HEPES (pH 7.2). For experiments with apamin, the toxin was added to the patch pipette solution, because bath application did not yield satisfactorily reproducible results. Patch formation and excision were performed in ϳ20 s, and the initial ramp reflected control currents before block. For outside-out patches, the solutions were reversed, with calcium-containing solution in the bath and calcium-free solution in the patch pipette. dTC and TEA were added to the bath solution, as described in the text. Electrode resistance was typically 2-5 M⍀. Membrane patches were voltage-clamped using an Axopatch 1B (Axon Instruments) or an EPC-9 (HEKA Electronics) amplifier. The data were low-pass Bessel-filtered at 1 kHz and acquired using Pulse software 3 B. Fakler, unpublished data.

FIG. 1.
A, schematic representation of the chimeric subunits used to localize the region determining apamin sensitivity. Solid regions represent SK1 and hatched regions represent SK2 amino acid sequences. Channels formed following expression of chimeric subunits were tested for sensitivity to 100 nM apamin as indicated in the schematic. B, comparisons of the pore regions from representative members of K ϩ channel subfamilies. Kir2.1 is a voltage-independent inward rectifier K ϩ channel with two transmembrane domains (46). TWIK is a voltageindependent K ϩ channel with four transmembrane domains (2 ϩ 2) and two pore sequences (47). TOK-1 is a yeast K ϩ channel with eight transmembrane domains (6 ϩ 2) and two pores (48). mSlo is a voltageand calcium-gated K ϩ channel with six transmembrane domains (49); HERG (50,51), mEAG (52), Kv1.1 (37,53), and Kv1.2 (38) are voltagegated K ϩ channels with six transmembrane domains. SK1 and SK2 are voltage-independent, calcium-activated K ϩ channels with six transmembrane domains (11). Gaps in the sequences were introduced to optimize the alignments. For SK2 and SK3, only the residues different from SK1 are presented.

FIG. 2. SK channels rundown in excised patches.
A, representative currents evoked by voltage ramp commands for SK2, recorded in an inside-out patch. Currents recorded at 0, 5, 10, and 19 min after excision into 5 M intenal Ca 2ϩ are shown. B, same as in A except that 200 pM apamin was included in the patch pipette. C, SK2 current amplitudes, measured at Ϫ100 mV every minute, were plotted as a function of time after excision. Data from four patches from the same oocyte are shown. Open circles represent the patch shown in A, above. D, same oocyte as for C except that 200 pM apamin was included in the patch pipette. Open circles represent the patch shown in B, above. E, superimposed normalized currents as a function of time after excision from C and D, above. F, averaged normalized currents as a function of time after excision from the four patches with 200 pM apamin (open symbols) and four patches without apamin (closed symbols) presented in E. Error bars are S.D.. Data were fit with a single exponential, yielding time constants of 2.2 min for rundown of SK2 and 6.2 min for block and rundown with apamin in the pipette solution.
(HEKA Electronik). Analysis was performed using Pulse, Kaleidagraph (Abelbeck), or IGOR (Wavemetrics) software. All experiments were performed at room temperature from a holding potential of 0 mV. 2.5-s voltage ramps from Ϫ100 to either 60 or 100 mV were acquired at a sampling frequency of 500 Hz. Values are expressed as mean Ϯ S.D. Statistical differences were determined using an unpaired t test; p values Ͻ0.05 were considered significant. Apamin was from Calbiochem, and d-tubocurarine was from Research Biochemicals International.

Determinant of Apamin and d-Tubocurarine Sensitivity-To
initially localize the sites of interactions between apamin or d-tubocurarine and SK channels, chimeric subunits between sensitive and insensitive SK subunits were constructed. The results presented in Fig. 1A show that channels containing the pore region of SK1 were not sensitive to 100 nM apamin, while those containing the pore region from SK2 were blocked.
The pore sequences of cloned potassium channels contain a consensus motif centered around GYG, which endows the characteristic selectivity for potassium ions (31). Fig. 1B shows an alignment of three SK subunit pore sequences with representative sequences from other K ϩ channel subfamilies. The sequences show progressively less homology with distance from the GYG motif, particularly in the N-terminal domain, suggesting that these differences might endow the outer vestibules with the different pharmacologies. The pore regions of the apamin-sensitive channel, SK2, and the apamin-insensitive channel, SK1, contain three different residues, all on the periphery of the deep pore region. To examine the residues mediating the differential pharmacology among SK channel types, site-directed mutants were constructed and expressed.
Determining the blocking potency of apamin was complicated by the slow blocking rate at concentrations near the K 0.5 and by SK current rundown. Also, apamin application to out-side-out patches did not yield reproducible results; block was not readily reversed, and significant amounts of apamin were retained in the bath chamber, even after extensive solution exchange. In contrast, apamin added to the patch pipette solution in the inside-out configuration yielded consistent and reproducible results. Therefore, apamin block was examined in inside-out patches (Fig. 2). Representative current traces from an inside-out patch, evoked by voltage ramp commands either without or with 200 pM apamin in the patch pipette, are shown in A and B, respectively. Currents were evoked every minute after excision, and the current amplitudes at Ϫ100 mV for each time point from four individual patches excised from the same oocyte are plotted in Fig. 2, C (without apamin) and D (with apamin). The first trace was recorded within 20 s of gigaohm seal formation and rapid patch excision. The data were normalized by comparing the current at Ϫ100 mV for a given time point with the initial current (Fig. 2E); the averaged data from those patches without or with apamin are shown in Fig. 2F. Patches recorded in the absence of apamin show that even though the absolute amplitudes of the initial and final currents varied between patches, due at least in part to differences in patch pipette resistance, the currents randown with a similar time course and were stable by 20 min. In the presence of apamin the current also reached steady state by 20 min after patch excision. For each oocyte, the relative decrease of current amplitudes at Ϫ100 mV with apamin in the patch pipette was determined in four patches and was corrected for rundown by comparison with four patches from the same oocyte recorded without apamin in the pipette. This analysis yielded a ratio of 0.18 for 200 pM apamin block of SK2 channels, equivalent to an 82% block, consistent with our previously published data (11). None of the mutations described below, or application of either toxin, obviously affected rundown. The far right column shows the effect of 100 nM apamin applied to SK1 channels. The relative current at Ϫ100 mV recorded 19 min after inside-out patch excision was determined from the average of four patches with and without apamin in the patch pipette (external) solution (see Fig. 2). The number of oocytes in each group is given in parentheses. Error bars are S.D. B, representative ramp currents from one oocyte for each of the channels shown in the order presented in B, with (AP) or without apamin in the patch pipette (external) solution. In all cases, the reduced current trace was recorded in the presence of apamin (200 pM, except for SK1, which was 100 nM).
The amino acid residues responsible for the differential sensitivity were identified by examining SK1 subunits with one, two, or three of the different residues in the pore converted to those found in SK2 (see Fig. 1B). Substitution of SK1 His 357 by Asn (H357N) endowed the channels with partial apamin sensitivity, being 21 Ϯ 2% blocked by 200 pM and 52 Ϯ 8% by 2 nM. Similar results were obtained by substituting E330D (19 Ϯ 1% block by 200 pM; 25 Ϯ 7% by 2 nM). These mutations were additive, as the double mutant, E330D,H357N, was 49 Ϯ 11% blocked by 200 pM apamin, almost completely converting the insensitive SK1 channel to that of SK2. SK1 channels in which Lys 328 was changed to Gln were not blocked by 100 nM apamin, and the triple mutant, K328Q,E330D,H357N, was not more sensitive than the double mutant (p ϭ 0.21), being 60 Ϯ 7% blocked by 200 pM (Fig. 3, A and B). SK3 subunits contain one of the two determinants of apamin sensitivity, Asp 315 , but His, as found in SK1, at the other position. As predicted from the results presented above, SK3 channels are less sensitive than SK2 channels, being 50% blocked by 2 nM apamin (not shown).
As for native SK channels (10), apamin-sensitive SK2 channels are blocked by dTC (IC 50 ϭ 5.4 M, n ϭ 4), while apamininsensitive SK1 channels are less sensitive (IC 50 ϭ 354.3 M, n ϭ 4). In contrast to apamin, block by dTC developed rapidly, was reproducible, and the toxin readily washed out from outside-out patches. Therefore, structural determinants for dTC sensitivity were examined using outside-out patches with 5 M Ca 2ϩ in the internal (patch pipette) solution; patches were allowed to stabilize for 20 min prior to addition of dTC. As for apamin, the effects of the mutations SK1 E330D (IC 50 ϭ 62.6 M, n ϭ 7) and H357N (IC 50 ϭ 11.1 M, n ϭ 4) were additive; the double mutant shifted dTC sensitivity to that of SK2 channels (IC 50 ϭ 6.3 M, n ϭ 4; Fig. 4). Changing Lys 328 to Gln did not significantly alter dTC sensitivity.
Sensitivity to External TEA-TEA, a small quaternary amine with 4-fold symmetry, has been employed as a molecular ruler, probing the pore dimensions of cloned K ϩ channels (32)(33)(34)(35). Like their native counterparts, cloned potassium channels show different sensitivities to external TEA. Among the voltage-gated K ϩ channels, Kv1.1 channels are most sensitive (IC 50 ϭ 0.3 mM (36, 37), while closely related Kv1.2 channels are only weakly blocked (IC 50 ϭ 150 mM). Mutation of Val 381 in Kv1.2 to Tyr, the residue present in Kv1.1 shifts the sensitivity almost to that of Kv1.1 (25,26). The structural differences between the pores of SK channels and voltage-gated K ϩ channels is reflected by their distinct pharmacologies. However, the C-terminal domain of SK channel pores is identical to that of Kv1.2 over seven residues, including the major determinant of sensitivity to TEA (38). To investigate whether the architecture of the outer pore of SK channels is similar to voltage-gated potassium channels, block by external TEA was examined. SK1 channels were blocked with an IC 50 ϭ 14.6 mM (n ϭ 4). Changing Val 355 to Tyr, the amino acid in Kv1.1, which mediates high TEA sensitivity, increased sensitivity to external TEA to that seen for Kv1.1 channels (IC 50 ϭ 0.3 mM, n ϭ 4; Fig. 5). This result shows that the general architecture of the external vestibule of SK channel pores is similar to that of voltage-gated K ϩ channels.
SK Subunits Form Heteromeric Channels-Several highly homologous SK subunits have been cloned. In situ hybridization in rat brain and Northern blot analysis of peripheral tissues indicates widespread, but distinctive, expression patterns, and in several instances more than one mRNA is detected (11). 4 To investigate whether SK subunits may coassemble to form heteromeric channels, approximately equal amounts of SK1 and SK2 mRNAs were coinjected. Currents from inside-out patches recorded with 10 nM apamin in the patch pipette were blocked by 86 Ϯ 5% (n ϭ 5; not shown), indicating that few homomeric SK1 channels were formed. To obtain a fixed stoichiometry, SK1-SK2 dimers were constructed. Injection of dimer mRNA into Xenopus oocytes resulted in functional SK channel activity; voltage ramp commands in the presence of 5 M cytoplasmic Ca 2ϩ resulted in robust, inwardly rectifying currents. The currents were reduced by apamin or dTC in a manner consistent with hetero-meric channels, being reduced 33 Ϯ 11% by 1 nM and 84 Ϯ 3% by 100 nM apamin (n ϭ 3) and with an IC 50 ϭ 22.3 M (n ϭ 7) for dTC (Fig. 6). For coinjected or dimer mRNA-injected oocytes, the lack of significant apamin-insensitive currents indicates that most, if not all, channels formed were heteromeric assemblies of SK1 and SK2 subunits. These results also show that more than one subunit mediates sensitivity to apamin and dTC. DISCUSSION The results presented here demonstrate that two residues residing on opposite sides of the outer vestibule of the SK pore determine sensitivity to the bee venom peptide toxin, apamin, and the plant alkyloid, d-tubocurarine. No other class of K ϩ channels is blocked by these drugs, and among cloned K ϩ channels the residues that endow sensitivity are present at those positions only in SK2 and SK3. Autoradiographic studies and in situ hybridization in rat brain show a correlation between the expression pattern of rSK2 and rSK3 mRNAs and 125 I-apamin binding sites (11,13,14). Taken together, these results suggest that SK channels may be the sole class of apamin receptors in brain.
The pore regions of all cloned K ϩ channels contains clear primary sequence conservation. However, as reflected by their different pharmacologies, the overall structure of the outer vestibules may be quite different. For example, attempts to endow the pore regions of inward rectifier K ϩ channels with sensitivity to external TEA resulted in nonfunctional channels (39). However, introducing a tyrosine residue at the position that mediates external TEA sensitivity in voltage-gated K ϩ channels endows SK channels with TEA sensitivity equivalent to that for Kv1.1, demonstrating that the overall architecture of the outer vestibule is similar for the two channel types.
Apamin is an 18-amino acid peptide with two internal disulfide bridges that hold the peptide in a tight, pear-shaped tertiary conformation (40,41) similar to that proposed for several other larger peptide ion channel blockers such as ␣-dendrotoxin and ␤-bungarotoxin (42). Structure-activity studies showed that one of the two adjacent Arg (arginine) residues (Arg 13 , Arg 14 ) and Gln 17 are crucial for activity, likely through electrostatic as well as hydrophobic interactions (5,41,43,44). The results presented here are consistent with a model in which Asp 341 on the SK2 channel interacts with one of the Arg residues on the toxin, and Asn 368 on the channel interacts with Gln 17 on the toxin.
For both blockers, the effects of the two channel residues are additive, and channels formed by expression of SK1-SK2 dimers have intermediate sensitivities between SK1E330D and SK1H357N. Channels formed from dimers are expected to contain two SK2D341 and two SK2N368 residues, while homomeric SK2 channels will contain four of each, suggesting that maximal sensitivity requires interactions between the toxin and at least two channel subunits. Several brain regions and peripheral tissues express more than one SK channel mRNA (11) 4 , and the results obtained from coexpression of SK1 and SK2 and expression of the SK1-SK2 dimer indicate that SK subunits may form heteromeric channels, giving rise to structural and functional diversity among this class of K ϩ channels. The cloned SK subunits form channels with similar calcium sensitivities and conduction properties. However, the intracellular N and C termini demonstrate considerable sequence divergence, and these domains may mediate specific functions in response to intracellular signals, such as regulation by cyclic AMP-dependent protein kinase (12). The structural differences in the outer vestibule may reflect differential regulation by endogenous extracellular ligands. An endogenous peptide with structural and functional 4 T. M. Ishii, J. Maylie, and J. P. Adelman, unpublished observation. similarities to apamin in brain has been reported (45), and apamin application has profound physiological effects, presumably through its interactions with SK channels. Therefore, understanding the molecular determinants of apamin binding may provide a framework for the design of novel therapeutic agents affecting seizures (15), circadian cycle (16), learning disorders (17), intestinal motility (19), and myotonic dystrophy (23).