Molecular determinants of high affinity binding of alpha-scorpion toxin and sea anemone toxin in the S3-S4 extracellular loop in domain IV of the Na+ channel alpha subunit.

α-Scorpion toxins and sea anemone toxins bind to a common extracellular site on the Na+ channel and inhibit fast inactivation. Basic amino acids of the toxins and domains I and IV of the Na+ channel α subunit have been previously implicated in toxin binding. To identify acidic residues required for toxin binding, extracellular acidic amino acids in domains I and IV of the type IIa Na+ channel α subunit were converted to neutral or basic amino acids using site-directed mutagenesis, and altered channels were transiently expressed in tsA-201 cells and tested for 125I-α-scorpion toxin binding. Conversion of Glu1613 at the extracellular end of transmembrane segment IVS3 to Arg or His blocked measurable α-scorpion toxin binding, but did not affect the level of expression or saxitoxin binding affinity. Conversion of individual residues in the IVS3-S4 extracellular loop to differently charged residues or to Ala identified seven additional residues whose mutation caused significant effects on binding of α-scorpion toxin or sea anemone toxin. Moreover, chimeric Na+ channels in which amino acid residues at the extracellular end of segment IVS3 of the α subunit of cardiac Na+ channels were substituted into the type IIa channel sequence had reduced affinity for α-scorpion toxin characteristic of cardiac Na+ channels. Electrophysiological analysis showed that E1613R has 62- and 82-fold lower affinities for α-scorpion and sea anemone toxins, respectively. Dissociation of α-scorpion toxin is substantially accelerated at all potentials compared to wild-type channels. α-Scorpion toxin binding to wild type and E1613R had similar voltage dependence, which was slightly more positive and steeper than the voltage dependence of steady-state inactivation. These results indicate that nonidentical amino acids of the IVS3-S4 loop participate in α-scorpion toxin and sea anemone toxin binding to overlapping sites and that neighboring amino acid residues in the IVS3 segment contribute to the difference in α-scorpion toxin binding affinity between cardiac and neuronal Na+ channels. The results also support the hypothesis that this region of the Na+ channel is important for coupling channel activation to fast inactivation.

␣-Scorpion toxins and sea anemone toxins bind to a common extracellular site on the Na ؉ channel and inhibit fast inactivation. Basic amino acids of the toxins and domains I and IV of the Na ؉ channel ␣ subunit have been previously implicated in toxin binding. To identify acidic residues required for toxin binding, extracellular acidic amino acids in domains I and IV of the type IIa Na ؉ channel ␣ subunit were converted to neutral or basic amino acids using site-directed mutagenesis, and altered channels were transiently expressed in tsA-201 cells and tested for 125 I-␣-scorpion toxin binding. Conversion of Glu 1613 at the extracellular end of transmembrane segment IVS3 to Arg or His blocked measurable ␣-scorpion toxin binding, but did not affect the level of expression or saxitoxin binding affinity. Conversion of individual residues in the IVS3-S4 extracellular loop to differently charged residues or to Ala identified seven additional residues whose mutation caused significant effects on binding of ␣-scorpion toxin or sea anemone toxin. Moreover, chimeric Na ؉ channels in which amino acid residues at the extracellular end of segment IVS3 of the ␣ subunit of cardiac Na ؉ channels were substituted into the type IIa channel sequence had reduced affinity for ␣-scorpion toxin characteristic of cardiac Na ؉ channels. Electrophysiological analysis showed that E1613R has 62-and 82-fold lower affinities for ␣-scorpion and sea anemone toxins, respectively. Dissociation of ␣-scorpion toxin is substantially accelerated at all potentials compared to wild-type channels. ␣-Scorpion toxin binding to wild type and E1613R had similar voltage dependence, which was slightly more positive and steeper than the voltage dependence of steady-state inactivation. These results indicate that nonidentical amino acids of the IVS3-S4 loop participate in ␣-scorpion toxin and sea anemone toxin binding to overlapping sites and that neighboring amino acid residues in the IVS3 segment contribute to the difference in ␣-scorpion toxin binding affinity between cardiac and neuronal Na ؉ channels. The results also support the hypothesis that this region of the Na ؉ channel is important for coupling channel activation to fast inactivation.
Voltage-gated Na ϩ channels are responsible for the conduc-tion of electrical impulses in most excitable tissues (1). The importance of their function is demonstrated by the effects of Na ϩ channel-specific neurotoxins that bind to at least six different receptor sites on the Na ϩ channel molecule and disrupt its normal behavior (reviewed in Refs. 2 and 3). These natural toxins are powerful tools for understanding and correlating ion channel structure and function, as exemplified by identification of molecular determinants for binding of the pore blocker tetrodotoxin, which has provided important information about the structure of the ion selectivity filter and pore (3,4). Similarly, the identification of molecular determinants for binding of toxins that modify activation or inactivation will likely provide important information about the mechanisms of channel gating.
Neuronal Na ϩ channels consist of a 260-kDa ␣ subunit with two auxiliary subunits, ␤1 and ␤2 (reviewed in Ref. 3). The ␣ subunit is independently functional when expressed in Xenopus oocytes or mammalian cells, and contains the ion pore and neurotoxin binding sites 1-3 and 5 (reviewed in Refs. 3 and 4). It contains four homologous domains (I-IV) that surround a central ion pore, and each domain contains six transmembrane segments (S1-S6) and a short membrane-penetrating segment (SS1-SS2) between segments S5 and S6. The four short SS1-SS2 segments form the ion selectivity filter and the tetrodotoxin receptor site, and the S4 transmembrane segments act as voltage sensors (3,4). The intracellular loop between domains III and IV acts as a fast inactivation gate, blocking the conduction pathway following channel activation (3,4).
Na ϩ channel-specific ␣-scorpion toxins and sea anemone toxins are distinct families of peptide toxins that share no sequence homology, but slow inactivation by binding to common or overlapping elements of neurotoxin receptor site 3 on the extracellular surface of the Na ϩ channel (reviewed in Ref. 2). The three-dimensional structures of ␣-scorpion and sea anemone toxins (5)(6)(7)(8)(9), the effects of peptide-specific antibodies and chemical modification (10 -14), and site-directed mutagenesis (15,16) indicate that conserved basic amino acid residues of these toxins are important for binding to the sodium channel.
The binding affinity of both of these classes of toxins is decreased by depolarization (17)(18)(19)(20)(21)(22). The voltage dependence of binding and the specific effect of these toxins on inactivation (19 -23) imply that membrane potential affects the structure of neurotoxin receptor site 3, that this region of the channel is important for the coupling of activation to inactivation, and that toxin binding to this site can slow or block a conformational change required for fast inactivation. Photoreactive derivatives of ␣-scorpion toxins covalently label both ␣ and ␤1 subunits (24 -26), but ␣ subunits expressed alone in Chinese hamster ovary cells retain high affinity binding of ␣-scorpion toxin (27). Two distinct regions of the Na ϩ channel ␣ subunit have been implicated in ␣-scorpion toxin binding by photoaffinity labeling of the S5-SS1 loop in domain I and inhibition of toxin binding by site-directed antipeptide antibodies directed against peptides of the S5-SS1 and SS2-S6 loops in domain I and the S5-SS1 loop in domain IV (28,29).
In this study, extracellular acidic amino acids in domains I and IV of the Na ϩ channel were converted to neutral or basic amino acids, and the resulting mutants were tested for ␣-scorpion toxin affinity by transient expression in mammalian cells. Glu 1613 in the extracellular loop between segments IVS3 and IVS4 was identified as a major determinant of LqTx 1 and ATX II binding. Mutation of additional residues within and adjacent to this loop identified residues important for either LqTx or ATX II binding, indicating that unique molecular determinants in this region form overlapping binding sites for ␣-scorpion toxins and sea anemone toxins and contribute to differences in binding of ␣-scorpion toxins between cardiac and neuronal Na ϩ channels.
Mutagenesis and Assembly of cDNAs Encoding Na ϩ Channels with Site-directed Mutations-Three M13 constructs containing type IIa Na ϩ channel ␣ subunit cDNA sequence (30 -32) were used for mutagenesis: mp18KXNC, which contained a KpnI/XbaI fragment (nt 23-540); mp19XaI, which contained a XbaI/SmaI fragment (nt 541-1897); and, mp18RVNC, which contained an EcoRV fragment (nt 4279 -5997). Uracil-containing mutagenesis templates were prepared from each of these constructs, and oligonucleotide-directed mutagenesis was performed using the dut Ϫ ung Ϫ selection procedure (33). Mutations made in the above three M13 mutagenesis constructs were confirmed by sequencing (Sequenase, U. S. Biochemical Corp.), excised by restriction cutting at the sites used for subcloning into the M13 constructs, and isolated by low-melting point agarose gel electrophoresis and GeneClean. Fragments from mp18KXNC were subcloned into Kpn/XbaI cut pCDM8SalK-NC, and fragments from mp19XaI and mp18RVNC were subcloned into appropriately cut and purified pCDM8Sal-NC. All mutations were then confirmed in the final constructs by DNA sequencing and extensive restriction mapping. Approximately one cDNA clone in 20 was found in this way to have rearranged during bacterial growth and clonal amplification of the plasmids, probably due to recombination of the related cDNA sequences encoding the four homologous domains of the Na ϩ channel. To be certain that undetected rearrangements did not influence interpretation of our results, multiple independently iso-lated cDNA clones were analyzed for each mutant channel which had altered toxin binding properties.
For transient expression, pCDM8Sal-NC and pCDM8SalK-NC were created by digesting the rIIa ␣ subunit (31,32) with SalI and subcloning into two modified pCDM8 transient expression vectors. The vector pCDM8Sal was created by excision of the pCDM8 polylinker region with XbaI and addition of a SalI site to the polylinker, and was then used to make pCDM8SalK by removal of a KpnI site in pCDM8Sal through KpnI digestion, filling the single-strand overhang with T4 DNA polymerase, and ligating using standard protocols.
Construction of cDNAs Encoding Chimeras of Cardiac and Brain Sodium Channels-The chimeras in the IVS3-S4 segment of the Na ϩ channel (see Fig. 5) were constructed by site-directed mutagenesis as described above. An additional larger cardiac chimera in which the IVSS2-S6 loop was converted from brain to cardiac sequence (Table I) required an alternative strategy. The antisense oligonucleotide 5Ј-GAAAATCCCCACCGCGGGGTTCCCACAGCCAGCCGGCAGACG-TCGTGATTTGG-3Ј was annealed to rIIa nt 5124 -5262 in mp18RVNC to create a single-stranded loop-out of nt 5149 -5234. These nucleotides were then deleted by loop-out mutagenesis while creating unique AatII and SacII restriction sites at positions 5137 and 5249 through silent mutations. Rat cardiac RNA was prepared from 14-day rat heart tissue by homogenization with a Polytron (Brinkman) in acidic phenol according to the manufacturer's instructions (Trizol, Life Technologies, Inc./ BRL). Complementary DNA was prepared using Superscript reverse transcriptase (Life Technologies, Inc./BRL). The forward and reverse cardiac-specific (34) oligonucleotide primer/adapters 5Ј-CACGACGT-CAGCCGGCTGG-3Ј and 5Ј-CCACCGCGGGGCTCCCAC-3Ј, respectively, were used to amplify the analogous loop sequence from rat heart cDNA with the polymerase chain reaction according to manufacturer's instructions (AmpliTaq, Perkin Elmer). This product was purified with GeneClean (BIO-101), digested with AatII and SacII, and then subcloned into the correct reading frame of the above mutation. The junctions and the 219 base pairs of inserted rH1 sequence were then confirmed before subcloning into the full-length rIIA cDNA as above.
Transient Expression in tsA-201 Cells-Expression of Na ϩ channels for 125 I-LqTx binding was begun by plating 20,000 tsA-201 cells/cm 2 in a 150-mm tissue culture plate on the day before transfection. Cells were transfected with 55 g of pCDM8 vector containing the rIIa Na ϩ channel cDNA using calcium phosphate/DNA coprecipitation (35). Twenty to 24 h later, these cells were subcultured by rinsing twice with phosphate-buffered saline and treating with a solution containing 1 mg/ml trypsin and 2 mM EDTA for several minutes at room temperature. Due to the high residual calcium and high cell density, this solution was required to adequately remove cells and disperse them for secondary plating. The cells were then rinsed from the plate with culture medium, centrifuged, and thoroughly resuspended in 50 ml of supplemented media. In order for these cells to adhere and allow ␣-scorpion toxin binding to intact cells, 24-well tissue culture plates were treated for 5 min with 0.5 mg/ml poly-D-lysine, then aspirated and rinsed twice with phosphate-buffered saline. Treated plates were used on the day of preparation, as plates stored overnight gave unusually high background binding of 125 I-LqTx. Two ml of transfected cells were plated per well (ϳ500,000 cells/well), and the cells were returned to the incubator overnight for binding on the following day.
Na ϩ channels were transiently expressed for saxitoxin binding by transfecting in 100-or 150-mm plates as described above and maintained on the transfected plates until collection for binding 40 -48 h after transfection.
Transient expression for electrophysiological analysis of Na ϩ channels was done by calcium phosphate-mediated co-transfection of 200,000 tsA-201 cells in a 35-mm dish with a 10:1 molar ratio of Na ϩ channel expression plasmid and pEBO-pCD-leu2, a vector encoding the CD8 antigen, as described (35). Transfected cells were split 10 -20-fold on the day following transfection, then analyzed 40 -72 h after transfection. Cells expressing the CD8 antigen were identified by incubation with polystyrene microspheres precoated with anti-CD8 antibody (36), and used for electrophysiological recording.
125 I-LqTx Binding to Stably and Transiently Transfected Cells-125 I-LqTx binding to transfected cells was performed as described (27) with the addition of 10 g/ml gramicidin A to the binding solution to increase the membrane potential. Depending on expression efficiency, total 125 I-LqTx bound per well ranged from 0.2 to 6 fmol/well and nonspecific binding accounted for 0.05 to 0.08 fmol/well. Total cell protein ranged from 400 to 800 g/well using bovine serum albumin as a standard in a modified Lowry protein assay (37).
[ 3 H]Saxitoxin Binding to Intact Cells-Cell washes, preparation, and saxitoxin binding were all done on ice with 1XP solution containing no bovine serum albumin essentially as described (38). Transfected cells on 100-or 150-mm dishes were rinsed twice with 5 ml of 1XP (130 mM choline chloride, 50 mM Hepes-Tris (adjusted with Tris base to pH 7.4), 5.5 mM glucose, 0.8 mM MgSO 4 , and 5.4 mM KCl), scraped into 10 ml of 1XP, and placed in 15-ml polystyrene tubes (Falcon 2509). Cells were sedimented at 200 ϫ g for 5 min at room temperature in a clinical centrifuge, and the pellet was resuspended in 1 ml of 1XP per 8000 cm 2 of plate area. For a 250-l binding reaction, 200 l of cell suspension, 25 l of 1-100 nM [ 3 H]saxitoxin in 1XP, and 25 l of 1XP or 100 M TTX were combined in a 3-ml polypropylene tube (Falcon 2063) and incubated 1 h at 4°C. Reactions were then filtered over GF/C filters (Millipore) under vacuum, washed twice with ice-cold choline wash solution (163 mM choline chloride, 5 mM Hepes-Tris, 1.8 mM CaCl 2 , and 0.8 mM MgSO 4 , pH 7.4), and bound radioactivity was detected by liquid scintillation counting. For the wild-type channel, nonspecific binding accounted for 11-27% of the total binding.
Ligand Binding Analysis-Results of displacement or saturation binding experiments were analyzed using the iterative fitting programs EBDA and LIGAND (Elsevier Biosoft, UK). A Student's t test (paired or unpaired as appropriate) was used for statistical comparisons of toxin affinities, using p Ͻ 0.05 as the criterion of statistical significance.
Electrophysiological Analysis-Whole cell voltage clamp experiments were performed as described previously (39) using solutions that contained 90 mM CsF, 50 mM CsCl, 10 mM CsEGTA, 10 mM NaF, 2 mM MgCl 2 , 10 mM Hepes (pH 7.4) in the pipette and 70 mM NaCl, 70 mM N-methyl-D-glucamine, 5 mM CsCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, and 10 mM Hepes (pH 7.4) in the bath. Except where noted in the figure legends, LqTx was incubated with the cells for 30 min at 37°C before the beginning of the experiment. Data collection was initiated 10 min after breaking the cell membrane to obtain the whole cell voltage clamp configuration.

␣-Scorpion Toxin Binding to Stably and Transiently Ex-
pressed Na ϩ Channels-Previous experiments have shown that mammalian cells stably expressing only the ␣ subunit of the type IIa Na ϩ channel bind LqTx with a 2-to 3-fold lower affinity than found with rat brain synaptosomes or neuroblastoma cells in culture (27). These non-neuronal cells have a more positive resting membrane potential than neuroblastoma cells. 2 As LqTx binding is inhibited by depolarization, gramicidin was used to increase the resting membrane potential and binding affinity. In the Na ϩ -free, choline-substituted isotonic solution used for binding assays, gramicidin allows K ϩ efflux without choline influx (40) to increase the negative membrane potential. Addition of gramicidin increased the receptor occupancy at 0.2 nM 125 I-LqTx by reducing the apparent K D from 5.4 Ϯ 3.1 nM to 2.0 Ϯ 0.6 nM in CNaIIa-1 cells (mean Ϯ S.D.; Fig.  1A). This reduction in apparent K D is consistent with a coincident increase in membrane potential and binding affinity, and results in an LqTx affinity identical to that seen in neuroblastoma cells (17).
For analysis of mutations, Na ϩ channels were expressed in the tsA-201 subclone of human embryonic kidney cells (HEK 293) (35). Fig. 1B illustrates LqTx binding to cells transfected with an expression vector containing no Na ϩ channel sequence or containing the rIIa ␣ subunit cDNA. Cells transfected with the empty vector exhibited no specific binding of 125 I-LqTx, whereas cells expressing wild-type or modified channels specifically bound 0.7-1.25 fmol of 125 I-LqTx per well in the presence of 0.5 nM labeled toxin and gramicidin. A Scatchard transformation of these results yielded K D values for LqTx binding to wild-type or selected mutant channels between 2.6 and 3.9 nM and B max values of 28 -50 pM, corresponding to 5.6 -10 fmol/culture well (Fig. 1B, inset). In this transient expression system, wild-type Na ϩ channels and those containing sitedirected mutations are expressed with varying efficiency depending on the mutation, plasmid preparation, and prepara-tion of the calcium phosphate precipitate. Analysis of LqTx binding to the wild-type Na ϩ channel at levels ranging from 0.1 to 6 fmol of specific 125 I-LqTx binding per well (0.2-10 fmol/mg of cell protein) indicated no effect of expression level on LqTx affinity.
Depolarization with a high extracellular K ϩ concentration would be expected to decrease binding affinity by decreasing the K ϩ equilibrium potential and the resting membrane potential. To test this, cells transiently expressing the wild-type ␣ subunit were tested in an isotonic binding solution containing gramicidin and either 5.4 or 60 mM extracellular K ϩ (Fig. 1B). Depolarization with high extracellular K ϩ increased the apparent K D of wild-type Na ϩ channels expressed in tsA-201 cells from 3.2 Ϯ 0.2 to 8.3 Ϯ 0.7 nM.
Charge Mutations in the ␣ Subunit of the rIIa Na ϩ Channel-Extracellular loops in domains I and IV of the Na ϩ channel ␣ subunit have been implicated in ␣-scorpion toxin binding (28,29), and basic amino acids of ␣-scorpion and sea anemone toxins are thought to be important for interaction with the receptor site (10 -16). In order to test the role of acidic amino acids in the extracellular loops of these two domains in toxin binding, conserved extracellular acidic residues were altered individually or in clusters to neutral or positively charged residues using oligonucleotide-directed site-specific mutagenesis (Table I). The extracellular acidic residues Asp 384 and Asp 1717 in domains I and IV were not tested because they have been shown to affect ion conductance as well as the binding of saxitoxin and tetrodotoxin (4). Since the receptor sites for saxitoxin and ␣-scorpion toxin do not interact (17), and ␣-scorpion toxins do not affect ion conductance (41,42), these residues were considered unlikely to participate in ␣-scorpion toxin binding.
Mutant Na ϩ channels were expressed transiently in tsA-201 cells and tested for displacement of 0.5 nM 125 I-LqTx with unlabeled LqTx. Surprisingly, all of these altered channels appeared identical to wild type with the exception of E1613R and E1613H, which showed no specific binding of 125 I-LqTx (Table I). Receptor occupancy for 0.5 nM 125 I-LqTx binding to the wild-type channel (K D ϭ 2.8 nM) is less than 0.2, and a 25-fold reduction in affinity would decrease receptor occupancy to undetectable levels in our experiments. Thus, E1613R and E1613H may be poorly expressed, or they may have greater than 25-fold reduction in affinity.
[ 3 H]Saxitoxin Binding to Mutants at Glu 1613 -To test the level of expression of these Glu 1613 mutations, tsA-201 cells were transiently transfected with control vectors or with E1613R or E1613H and assayed for [ 3 H]saxitoxin binding. Cells were prepared for saxitoxin binding in an isotonic solution with gentle homogenization to assess the level of cell surface expression. Cells transfected with an empty expression vector showed no specific saxitoxin binding, whereas those transfected with the wild-type Na ϩ channel construct expressed 145 Ϯ 48 fmol of specific saxitoxin receptor sites/mg of cell protein when assayed in the presence of 5 nM [ 3 H]saxitoxin (n ϭ 5). Cells expressing E1613R or E1613H showed a level of saxitoxin binding very similar to that of wild type ( Fig. 2A), indicating that these channels reach the cell surface and that the mutation of Glu 1613 does not significantly disrupt saxitoxin binding. These results indicate that loss of ␣-scorpion toxin binding in E1613R channels is caused by reduced affinity for ␣-scorpion toxin.
To further test the specificity of the E1613R mutation, the saxitoxin binding affinities were determined for transiently expressed wild-type and E1613R channels. Scatchard analysis of these data indicates that the K D for saxitoxin binding to E1613R is 0.30 Ϯ 0.03 nM (mean Ϯ S.E.; n ϭ 3), essentially identical to that for the wild-type channel (0.28 Ϯ .02 nM, n ϭ 3, Fig. 2B). These values for saxitoxin affinity are in close agreement with [ 3 H]saxitoxin binding to mammalian cells stably expressing the rIIa ␣ subunit in the absence and presence of ␤ 1 subunits (38). Thus, this mutation is specific in that it does not disrupt channel expression or saxitoxin binding affinity.
Glu 1613 is conserved in cloned rat brain Na ϩ channels (30 -32) and is an Asp in the cardiac Na ϩ channel (34) as well as in the skeletal muscle, Drosophila para, eel electroplax, and squid Na ϩ channels. Most alignments (e.g. Refs. 30, 32, and 34) predicted Glu 1613 to be within the IVS3 transmembrane segment, but two analyses predicted that this residue is located at the extracellular end of the IVS3 transmembrane segment (43,44). We found that analysis of the rIIa sequence from Thr 1591 to The indicated extracellular acidic residues in domains I and IV were changed to neutral, basic, or alanine residues to test their importance in LqTx binding. Each was transiently expressed in tsA-201 cells, assayed for displacement of 0.5 nM 125 I-LqTx with unlabeled LqTx, and the results were fit with the iterative computer programs EBDA and LIGAND (Elsevier-Biosoft). ⌬Q is the change in net charge as a result of the mutations. n is the number of independent experiments for which toxin affinity was determined. Comparison of mean affinities with the wild-type Na ϩ channel revealed no significant differences in affinity for any constructs showing measurable specific binding (Student's t test, p Ͼ 0.05). NB for E1613R and E1613H indicates no measurable specific 125 I-LqTx binding.

Mutant
Loop Arg 1626 by the Predict Protein program (45) also predicted Glu 1613 to be on the extracellular surface. Importance of Charged Amino Acids in the IVS3-S4 Extracellular Loop for ␣-Scorpion Toxin and ATX II Binding-To test the specificity of interaction between Glu 1613 and ␣-scorpion toxin, Glu 1613 was changed to either Asp or Gln and LqTx affinity was determined. Although neither of these mutations caused a statistically significant change in LqTx binding affinity (p Ͼ 0.05), the E1613Q mutation may slightly decrease LqTx affinity (p Ͻ 0.1 for n ϭ 3, Fig. 3A). In addition, the role of the basic residue at the extracellular end of the IVS4 segment in ␣-scorpion toxin binding was tested by changing Arg 1626 to a neutral or acidic residue. These changes had no significant effect on LqTx affinity (Fig. 3A).
Since ␣-scorpion toxins and sea anemone toxins share a common binding site (18), the ATX II affinity of these mutant channels was also tested by competition for 125 I-LqTx binding with unlabeled ATX II (Fig. 4, open squares). The K D of type IIa Na ϩ channels for ATX II was 76 Ϯ 6 nM. E1613D significantly increased ATX affinity (13 Ϯ 3 nM; p Ͻ 0.05) while E1613Q caused a smaller increase in affinity (Fig. 3B). Substitutions of Arg and His at position 1613 could not be tested because these mutants do not bind LqTx detectably.
Glu 1616 was tested in a similar manner (Fig. 3B). E1616Q (K D ϭ 163 Ϯ 5 nM) and E1616R (K D ϭ 189 Ϯ 15 nM) both had significantly decreased affinity for ATX II. Mutations of Arg 1626 at the extracellular end of IVS4 which neutralize or reverse the charge at this position appeared to increase ATX affinity slightly (Fig. 3B), although not to a statistically significant degree (p Ͼ 0.05). In general, conversion of Glu 1613 , Glu 1616 , or Arg 1626 to more negative or shorter side chains increased ATX II affinity, and conversion to more positive side chains decreased affinity. .NC, or three constructs containing site-directed mutations that showed no specific 125 I-LqTx binding. Each construct was tested two to four times in a total of five experiments. To correct for different expression efficiencies between experiments, binding results from each experiment were normalized to the total [ 3 H]saxitoxin binding of cells expressing the wild-type rIIa Na ϩ channel in that experiment. The normalized results were averaged between experiments and plotted (mean Ϯ S.D.). B, saturation binding of 0.1 to 11 nM [ 3 H]saxitoxin to tsA-201 cells transiently expressing wild type (q, solid line) or E1613R (Ⅺ, dotted line) Na ϩ channels was carried out for 1 h at 4°C in the absence or presence of 500 nM TTX. Results were analyzed with LI-GAND/EBDA and plotted. Apparent affinities from this Scatchard plot were 0.28 nM for the wild-type channel and 0.29 nM for the E1613R channel.
FIG. 3. Apparent LqTx and ATX II affinities of type IIa Na ؉ channels containing charge and alanine-scanning mutations in the IVS3-S4 extracellular loop. A and C, Na ϩ channels containing charge or alanine scanning mutations in the IVS3-S4 loop (residues 1613 to 1626) were tested for displacement of bound 125 I-LqTx with unlabeled LqTx and fit with LIGAND/EBDA. Results of 3-18 determinations with each mutation are presented as box plots, where the box represents a 95% confidence interval of the median, the solid and dashed lines mark the median and mean of the data, whiskers show the main body of the data, and outliers are plotted individually. B and D, these altered Na ϩ channels were also tested for the ability of ATX II to displace bound 125 I-LqTx, and IC 50 values were converted to K I values using the calculated ␣-ScTx affinities and the equation K I ϭ (IC 50 /(1 ϩ [ 125 I-LqTx]/K D , LqTx). Results are presented as box plots, with asterisks indicating K I values significantly different from the K I of wild-type channels (p Ͻ 0.05).

Alanine-scanning Mutagenesis of the IVS3-S4 Extracellular
Loop-Having found several charged residues in the IVS3-S4 loop that were important for ␣-scorpion toxin and/or sea anemone toxin binding, the effects of mutation of residues Leu 1614 through Phe 1625 were then tested by conversion of each to Ala. Two of these amino acid changes, K1617A and P1622A, were found to significantly increase ␣-scorpion toxin affinity (p Ͻ 0.05, Fig. 3C). The increase in affinity with these individual mutations suggests that the positive charge and conformation of this loop may hinder ␣-scorpion toxin binding in wild type.
The importance of individual residues in this extracellular loop for ATX II binding was assessed by competition with 125 I-LqTx binding as well. Replacement of five consecutive residues, Leu 1614 through Tyr 1618 , individually with Ala resulted in one mutant with lower ATX II affinity (E1616A), and two with higher ATX II affinities (L1614A and K1617A, Fig. 4). As summarized in Fig. 3D, Ala replacements of Leu 1614 , Lys 1617 , and Ser 1621 all significantly increased ATX II binding affinity (p Ͻ 0.05), whereas substitution of residues Glu 1616 , Val 1620 , and Leu 1624 significantly decreased affinity (p Ͻ 0.05). The changes in affinity for Leu 1614 , Lys 1617 , Ser 1621 , and E1616A are also consistent with the generalization that shorter side chain length and more negative charge lead to higher ATX affinity. The reduction in affinity with the V1620A and L1624A mutations may indicate that side chain length or hydrophobicity is important for their interaction with ATX II, or that the folding of this loop is disrupted with these mutations.
Surprisingly, replacement of aromatic residues Tyr 1618 and Phe 1619 with Ala did not affect toxin affinity or expression, suggesting that these residues may be oriented toward the channel protein or membrane and are not critical for toxin binding. It was also surprising that P1622A had normal ex-pression and high affinity binding, as an Ala at this position would be expected to relieve some conformational constraints and alter a bend in this loop. Mutant F1625A did not express sufficiently well to detect either ␣-scorpion toxin binding or saxitoxin binding ( Fig. 2A), indicating that this residue is probably important in channel assembly or folding.
Amino Acid Residues which Cause Differences in ␣-Scorpion Toxin Binding to Brain and Cardiac Na ϩ Channels-Cardiac Na ϩ channels bind ␣-scorpion toxin with lower affinity than brain Na ϩ channels (46). Rat brain type IIa Na ϩ channels expressed in tsA-201 cells have a K D for LqTx binding of 2-5 nM (Figs. 1 and 5), whereas rat cardiac rH1 Na ϩ channels have a K D of 18 nM (Fig. 5, B and C). These two Na ϩ channel ␣ subunits have several differences in amino acid sequence in transmembrane segment IVS3 and in the IVS3-S4 extracellular loop (Fig. 5A). We constructed chimeric Na ϩ channels in which the amino acid residues in the IVS3-IVS4 segment of the rat brain type IIa ␣ subunit were converted to those in the cardiac isoform (Fig. 5A). Conversion of two extracellular residues (Chim1) or four extracellular residues (Chim2) did not significantly affect LqTx affinity (Fig. 5, B and C). The lack of significant effects of these extracellular mutations on ␣-scorpion toxin affinity is consistent with the results of alaninescanning mutants (Fig. 3). In contrast, conversion of seven amino acid residues in this region, including three residues predicted to be in transmembrane segment IVS3, reduced LqTx affinity to a level similar to that of the cardiac Na ϩ channel (Chim3, K D ϭ 24 nM, Fig. 5, B and C). This result suggests that residues at the extracellular end of the IVS3 transmembrane segment confer isoform-specific LqTx binding properties on the ␣-scorpion toxin receptor site. We also analyzed the binding of LqTx to chimeras in which each of the other 15 extracellular loops of the rIIa ␣ subunit had been individually replaced with the corresponding amino acids from the rH1 ␣ subunit. 2 All of these chimeras had K D values for LqTx which were identical to rIIa channels, indicating that the amino acid residues near the extracellular end of the IVS3 transmembrane segment may be primarily responsible for differences in LqTx binding between cardiac and neuronal Na ϩ channels.
Electrophysiological Properties of Na ϩ Channels Containing Mutations of Glu 1613 -To determine the toxin affinity and electrophysiological properties of the E1613R and E1613H mutants, wild-type and mutant channels were transiently expressed in tsA-201 cells and analyzed by whole cell voltage clamp. The Na ϩ currents elicited by depolarization of cells expressing wild-type, E1613R (Fig. 6), or E1613H (not shown) appeared identical in time course without LqTx present. The voltage eliciting half-maximal activation (V1 ⁄2 ) from a holding potential of Ϫ140 mV was Ϫ19.3 Ϯ 3 mV for wild-type rIIa channels (n ϭ 3), Ϫ20.9 Ϯ 2.3 mV for E1613R (n ϭ 5), and Ϫ30.5 Ϯ 3.7 mV for E1613H (n ϭ 3) (data not shown). The V1 ⁄2 for steady-state inactivation was Ϫ59.0 Ϯ 2.9 mV (n ϭ 5) for wild-type channels, Ϫ61.4 Ϯ 2.7 mV for E1613R (n ϭ 5), and Ϫ74.0 Ϯ 0.4 mV for E1613H (n ϭ 3) (data not shown). Thus, the voltage dependence of channel activation and inactivation was very similar for wild type and E1613R, but both of these parameters were shifted in the hyperpolarized direction for E1613H.
Affinity of Na ϩ Channels Containing Mutations at Glu 1613 for LqTx and ATX II-In the absence of toxin, Na ϩ conductance through wild-type Na ϩ channels decays to approximately 5% of the peak within 2 ms. In contrast, after addition of a saturating concentration of LqTx (20 nM) up to 70% of the current remains after 2 ms (Fig. 6A). Mutant E1613R was more than 10-fold less sensitive to ␣-scorpion toxin, since 200 nM toxin caused less slowing of inactivation for the mutant than 20 nM did for wild-type (Fig. 6B).
To determine the K D for LqTx, transiently expressed wildtype or mutant Na ϩ channels were incubated with varying concentrations of LqTx for 30 min at 37°C. Cells were then voltage clamped at a holding potential of Ϫ140 mV, and currents were elicited by a depolarizing pulse to Ϫ10 mV. The fraction of conductance remaining 2 ms after the peak is proportional to the number of channels modified by ␣-scorpion toxin, and this fraction can be used to estimate receptor occupancy and toxin affinity according to the formula, where F G is the fraction of Na ϩ current remaining 2 ms after the beginning of the pulse, and F G is the maximum fraction of current 2 ms after the beginning of the pulse in the presence of a saturating concentration of ␣-scorpion toxin (1 M, Ref. 47). From this analysis, the K D for LqTx binding was 1.7 nM for the wild-type channel, 14 nM for E1613H, and 106 nM for E1613R channels (Table II).
Similar experiments were carried out for ATX II binding to mutant E1613R. Addition of 1 nM ATX II slowed the inactivation of a significant fraction of wild-type Na ϩ channels, and 10 nM slowed inactivation of most wild-type Na ϩ channels (Fig.  6C). Averaged results indicated a K D value at Ϫ140 mV of 3.3 nM (Table II). A concentration of 100 nM ATX II was much less effective in slowing inactivation of E1613R than 10 nM ATX II was in slowing wild type (Fig. 6D), and averaged results indicated a K D value of 270 nM for E1613R (Table II). Thus, the E1613R mutation reduces the affinity for LqTx 62-fold and the affinity for ATX II 82-fold. These results are consistent with the conclusion that Glu 1613 is an integral component of overlapping binding sites for both ␣-scorpion toxins and sea anemone toxins.
Electrophysiological Determination of the Kinetics of LqTx Binding to Wild-type, E1613R, and E1613H Channels-Binding of ␣-scorpion toxins is reversible, and the difference in toxin FIG. 5. Analysis of chimeric brain/cardiac Na ؉ channels for isoform-specific LqTx binding. A, alignment of the IVS3-S4 sequence of brain (rIIa) and cardiac (rH1) Na ϩ channels and three chimeric brain channels containing the indicated site-directed changes to substitute cardiac sequence. Lowercase letters indicate positions with nonconservative amino acid changes. B, displacement of bound 125 I-LqTx with unlabeled LqTx was measured with tsA-201 cells transiently expressing wild-type rIIa (q), rH1 (f), Chim1 (å), Chim2 (ç), and Chim3 (ࡗ) channels. Curves were normalized and fit as described and the apparent affinities from these fits were 3.6 nM for rIIa, 12.8 nM for rH1, 4.6 nM for Chim1, 4.2 nM for Chim2, and 23.1 nM for Chim3. C, summary of affinities from three to five separate experiments. Paired t tests were used to assess significant differences from rIIa affinity (p Ͻ 0.05, *).

FIG. 6. Electrophysiological effect of LqTx and ATX II on wildtype and E1613R Na ؉ channels transiently expressed in tsA-201 cells.
Cells expressing the rIIa wild-type Na ϩ channel ␣ subunit (A and C) or rIIa-E1613R (B and D) were held at Ϫ140 mV and Na ϩ currents were elicited with a 25-ms step to Ϫ10 mV. LqTx or ATX II was added in bath solution with 10-fold higher toxin concentration, allowed to equilibrate for 5-10 min, and Na ϩ currents were recorded as described under "Experimental Procedures." Na ϩ currents elicited in the presence of the indicated final toxin concentrations are shown. affinity at equilibrium is a reflection of differences in the kinetic rate constants for toxin association and/or dissociation. Previous biochemical and electrophysiological work has shown that ␣-scorpion toxin binding is inhibited by depolarization due to acceleration of toxin dissociation (17)(18)(19)(20)(21)(22). In order to compare the rates of toxin dissociation from these channels, strong depolarizing pulses of increasing duration were used to follow the time course of toxin dissociation from wild-type, E1613R, and E1613H channels. tsA-201 cells transiently expressing each channel were incubated in 100 nM LqTx for 30 min at 37°C, voltage clamped to Ϫ120 mV to stimulate maximum toxin binding, and then depolarized to ϩ100 mV for intervals from 9 to 89 ms to induce toxin dissociation, repolarized briefly to Ϫ120 mV to reverse channel inactivation, and depolarized to 0 mV for measurement of Na ϩ current (Fig. 7A, inset). This protocol was repeated with 10-ms increments in the duration of the interval at ϩ100 mV to yield the series of traces in Fig. 7A. Progressively faster Na ϩ channel inactivation was observed with each 10-ms increment at ϩ100 mV, until essentially complete dissociation of the toxin and loss of toxin effect was reached during 89-ms prepulses to ϩ100 mV for wild type (Fig. 7A).
The time course of toxin dissociation was determined for wild-type and mutant Na ϩ channels at a range of membrane potentials using the protocol described in Fig. 7A. By plotting the ratio of current at 2 ms after the peak to the current at the peak as a function of dissociation time during the positive prepulses, the loss of toxin action with increasing prepulse duration was determined and fit with a single exponential equation (Fig. 7, B-D). For wild-type, the time constant of toxin dissociation at ϩ100 mV was 37.2 ms (Fig. 7B), but this time constant was 3.7 ms for the E1613R channel (Fig. 7C) and 4.0 ms for the E1613H channel (Fig. 7D). Comparison of the time constants of toxin dissociation at different potentials and toxin concentrations showed that the rates of toxin dissociation from all three channels were voltage-dependent, with more rapid dissociation during stronger depolarizations, and were concentration-independent (Fig. 7, B-D). The wild type channel has the slowest dissociation at all membrane potentials, and the E1613R mutant has the fastest (Fig. 7, legend). These rapid dissociation rates may reflect the rate and voltage dependence of the change of channel state induced in the toxin-channel complex by the voltage change from Ϫ90 mV to ϩ100 mV as well as the rate of dissociation of the toxin-channel complex itself.
The rates of toxin association were assessed for the wild-type and mutant channels using a 200-ms prepulse to ϩ100 mV to cause toxin dissociation followed by progressively longer hyperpolarizing prepulses to follow the time course of toxin re-binding and action. Fig. 8A shows the stimulus protocol and the cumulative slowing of inactivation as toxin reassociates with the wild-type channel. The first Na ϩ current recorded 102 ms after repolarization to Ϫ120 mV was rapidly inactivating; more slowly inactivating currents were recorded following progressively longer hyperpolarizing prepulses as toxin binding approached equilibrium. The time courses of toxin association at several different potentials were determined for wild-type and mutant channels using this protocol. The association kinetics were fit with a single exponential time constant, which was 548 FIG. 7. Determination of the voltage-dependent dissociation rates of LqTx. A, cells transiently expressing the wild-type Na ϩ channel were incubated in 100 nM LqTx for 10 min at a holding potential of Ϫ120 mV to allow binding. The rate of toxin dissociation at ϩ100 mV was determined with the illustrated pulse paradigm by stepping to a conditioning pulse of ϩ100 mV for 9 to 89 ms (cond), returning to Ϫ120 mV for 20 ms to recover from fast inactivation, then eliciting Na ϩ current with a 10-ms test pulse to 0 mV (test). Traces show the acceleration of inactivation due to toxin dissociation with longer conditioning pulses. A new steady-state was reached after 79 -89-ms conditioning pulses. By using conditioning pulses to other potentials, the time course of dissociation at other potentials was also determined. B, kinetics of dissociation of 2 nM LqTx from cells expressing the wild-type Na ϩ channel using the same protocol at conditioning potentials of ϩ60 mV (å), ϩ80 mV (f), and ϩ100 mV (q). The ratio of current 2 ms after the depolarizing step relative to peak current was plotted in order to normalize for the change in peak Na ϩ current resulting from slow inactivation and to determine the extent of inactivation removal by LqTx. Toxin dissociation was fit with a single exponential function with the following time constants: ϩ100 mV ϭ 37.2 ms, ϩ80 mV ϭ 76.4 ms, and ϩ60 mV ϭ 209.2 ms. C, the time course of dissociation of 100 nM ␣-ScTx from cells expressing the E1613R was determined at ϩ100 mV (q, ϭ 3.7 ms), ϩ80 (ç, ϭ 5.5 ms), ϩ60 (f, ϭ 12.4 ms), ϩ40 (ࡗ, ϭ 16.9 ms), ϩ20 (å, ϭ 26.4 ms), and 0 (q, ϭ 47.1 ms). D, time courses of dissociation of 100 nM ␣-ScTx from E1613H at ϩ100 mV (q, ϭ 4.0 ms), ϩ80 mV (f, ϭ 6.3 ms), ϩ60 mV (å, ϭ 9.5 ms), ϩ40 mV (ç, ϭ 19.1 ms), and ϩ20 mV (ࡗ, ϭ 40.5 ms).

TABLE II Effects of mutations of Glu 1613 on binding of LqTx and ATX II
The indicated amino acids were substituted for Glu 1613 by oligonucleotide-directed mutagenesis, the resulting mutant Na ϩ channels were expressed in tsA-201 cells, and the K D values for binding of LqTx and ATX II were measured as described in Fig. 6  Ϯ 208 ms for wild-type at Ϫ100 mV (n ϭ 4, Fig. 8B), 429 Ϯ 49 ms for E1613R (n ϭ 3, Fig. 8C), and 365 Ϯ 204 ms for the E1613H (n ϭ 5, data not shown). By comparing the time constants measured at a range of voltages (Fig. 8, legend), the rate of toxin association at negative potentials was found to be voltage-independent, and similar for wild-type and mutant channels. Fig. 9 compares the voltage dependence of time constants for LqTx association and dissociation. The time constants for association of LqTx between Ϫ120 and Ϫ60 mV appeared unaffected by voltage and were similar for both of these channels (Fig. 9A) and for E1613H (not shown). In contrast, the dissociation rates were much faster for the E1613R channels and were strongly voltage-dependent, with more depolarized potentials significantly accelerating dissociation (Fig. 9B). A semilog plot of the voltage dependence of toxin dissociation from E1613R channels was linear between 0 and ϩ100 mV and indicated an e-fold increase in toxin dissociation rate for every 36.4 mV of depolarization. Similar experiments with the wild-type channel were more difficult as the dissociation rate was much slower and could only be measured above ϩ60 mV. However, using the dissociation rates at ϩ60, ϩ80, and ϩ100 mV (Fig.  9B), the toxin dissociation rate for wild-type channels was found to increase e-fold for every 22-25 mV. Although the narrow voltage range examined and the much slower dissociation rate for the wild-type channel make the comparison difficult, it appears that the voltage dependence of the rate of toxin dissociation is less steep for mutant E1613R than for wild type in this positive voltage range.
Voltage Dependence of Toxin Binding at Equilibrium-Previous models describing the voltage dependence of ␣-scorpion toxin binding have proposed two receptor states with high and low toxin affinities. These states are in reversible equilibrium described by a voltage-dependent, allosteric equilibrium constant (17,19,22). A negative membrane potential favors the high affinity conformation, depolarization favors the low affinity conformation, and intermediate voltages reveal a distribution of receptor affinities as described by the Monod-Wyman-Changeux model for allosteric modulation of oxygen binding to hemoglobin (17,19,48). Mutations of Glu 1613 may cause a reduction in binding affinity by directly disrupting binding, or by indirectly shifting the voltage-dependent allosteric equilibrium constant. By measuring the extent of toxin association at equilibrium at a range of negative membrane potentials, the K D values for LqTx binding at voltages between Ϫ120 mV and 0 mV for wild-type and E1613R channels were determined (Fig.  10, A and B, circles). With both channels, the voltage-dependent affinity change was sigmoidal, with the conversion between high and low affinity states occurring between Ϫ80 and Ϫ40 mV. Fitting with the Boltzmann equation yielded a V1 ⁄2 value for the affinity change of Ϫ55 Ϯ 5.9 mV (n ϭ 3) and a slope factor of e-fold per 4.6 Ϯ 1.7 mV for wild-type channels and a V1 ⁄2 of Ϫ51.6 Ϯ 1.8 mV (n ϭ 3) and slope factor of e-fold per 4.2 Ϯ 0.4 mV for E1613R channels (Fig. 10, legend). Similar results were obtained with mutant E1613R when the extent of toxin dissociation at equilibrium was measured (Fig. 10B, squares). At potentials more negative than Ϫ100 mV, the K D values for wild-type and E1613R channels were 1.7 and 106 nM, respectively. At ϩ40 to ϩ100 mV, the K D values were 26 and 800 nM, respectively. These results indicate that, although the binding affinities of these channels differ by over 60-fold at Ϫ100 mV, the voltage dependence of LqTx binding to these channels at equilibrium is similar.
By combining the equations K D ϭ k Ϫ1 /k 1 and ϭ (k 1 ϩ k Ϫ1 ) Ϫ1 , the K D values determined at each potential, and the association and dissociation time constants determined above, the corresponding dissociation (k Ϫ1 ) or association (k 1 ) rate constants could be calculated using the equations and At negative potentials, the K D and association time constants for LqTx binding to wild-type and E1613R channels were approximately constant between Ϫ120 and Ϫ80 mV (Figs. 9 and 10), indicating that the dissociation rate did not change substantially over this voltage range. Using the assoc rates from several experiments at Ϫ100 mV and K D values at Ϫ100 mV of FIG. 8. Determination of rates of LqTx association to transiently expressed Na ؉ channels at different membrane potentials. A, cells transiently expressing the wild-type Na ϩ channel were incubated in 100 nM LqTx for 10 min at a holding potential of Ϫ120 mV to establish binding. A 200-ms depolarizing step to ϩ100 was then used to dissociate all of the bound toxin and the cell was returned to Ϫ120 mV for increasing durations (102, 202, 402, 602, 1102, 2102, 3102, or 4102 ms, see diagram) before testing the extent of inactivation removal with a 10-ms test pulse to 0 mV. The pulse following 102 ms at Ϫ120 mV elicited a rapidly inactivating current. Subsequent cycles with longer re-binding periods at Ϫ120 mV elicited Na ϩ currents with successively slower inactivation kinetics until equilibrium was reached. Re-binding periods at different potentials (e.g. Ϫ100 and Ϫ80 mV) were used to assess the voltage dependence on LqTx association. B, kinetics of association of 100 nM LqTx to cells expressing the wild-type Na ϩ channel using the same protocol at conditioning potentials of Ϫ80 mV (f) and Ϫ100 mV (q). The ratio of current 2 ms after the depolarizing step relative to peak current was plotted in order to normalize for the change in peak current and to determine the extent of inactivation removal by LqTx. The association time courses were fit by single exponential time constants of Ϫ80 mV ϭ 639.6 ms and Ϫ100 mV ϭ 550.5 ms. C, association rates for LqTx binding to cells transiently expressing E1613R at Ϫ60 mV (ç, ϭ 695.8 ms), Ϫ70 mV (⌬, ϭ 520.8 ms), Ϫ80 mV (å, ϭ 388.1 ms), Ϫ90 mV (Ⅺ, ϭ 506.1 ms), Ϫ100 mV (f, ϭ 440.9 ms), Ϫ110 mV (E, ϭ 463.4 ms), and Ϫ120 mV (q, ϭ 485.5 ms). The LqTx occupancy at Ϫ40 mV (É) was too low to fit. 1.7 and 106 nM for the wild-type and E1613R channels, the association rate constants were calculated to be 1.8 ϫ 10 7 and 2.2 ϫ 10 7 M Ϫ1 s Ϫ1 , respectively. Using the equation k Ϫ1 ϭ K D ⅐ k 1 and the K D values determined between Ϫ120 and Ϫ80 mV (Fig. 9), the calculated k Ϫ1 values in this voltage range were 3.1 ϫ 10 Ϫ2 s Ϫ1 and 2.3 s Ϫ1 for the wild-type and E1613R channels, respectively.

DISCUSSION
High Affinity ␣-Scorpion Toxin Binding to Transiently Expressed Na ϩ Channels-Voltage-gated Na ϩ channels can be functionally expressed in a variety of nonexcitable cells. Our results show that LqTx binding with the high affinity characteristic of neurons can be observed for Na ϩ channels expressed in tsA-201 cells if the membrane potential of the cells is hyperpolarized by incubation in Na ϩ -free medium containing gramicidin. Thus, high affinity, voltage-dependent binding of ␣-scorpion toxin requires only the ␣ subunit of the Na ϩ channel and a sufficiently negative membrane potential. Previous labeling of Na ϩ channels with photoreactive derivatives of ␣-scorpion toxin have shown specific incorporation into both the ␣ and ␤1 subunits of the Na ϩ channel (24 -26), and different photoreactive derivatives label the two subunits in different ratios (26). In light of our present results, photolabeling of the ␤1 subunit likely represents covalent attachment of photoreactive LqTx to an area of the ␤1 subunit which is near, but on the periphery of, the ␣-scorpion toxin receptor site on the ␣ subunit.
Glu 1613  A 200-ms pulse to ϩ100 mV was applied to cause complete toxin dissociation followed by repolarization for 100 ms at Ϫ120 mV to recover from fast inactivation and a 10-ms test pulse to 0 mV to assess toxin dissociation. For determination of steady-state binding at each potential, a 2000-ms pulse to the indicated binding potential (Ϫ100 to 0 mV) was used to allow LqTx to bind the channel, followed by a 100-ms period at Ϫ120 to recover from fast inactivation and a 10-ms test pulse to 0 mV. This was repeated 11 or 16 times in order to reach steady-state at each binding potential. K D was calculated with the equation K D ϭ [LqTx]/(F G /F G Ϫ 1) as described in the text and plotted as a function of the voltage at which binding was monitored. The values of K D were fit with a Boltzmann equation (solid line) of the form K D ϭ 1/(1-exp((V Ϫ V 1/2 /k), where V 1/2 is the potential giving half-maximal affinity and k is the slope factor. For this cell the V 1/2 was Ϫ55 mV and the slope factor was 3.8 mV. The mean values were V 1/2 ϭ Ϫ55 Ϯ 5.9 mV and k ϭ 4.6 Ϯ 1.7 mV (n ϭ 3). B, binding of LqTx to cells expressing E1613R was assessed using steady-state binding with both the on-rate (Fig. 8C, E) and off-rate (Fig. 7C, Ⅺ) protocols. Steady-state binding following dissociation was monitored at positive potentials with E1613R because of the rapid dissociation, low affinity, and large amount of toxin required for association at these potentials. A Boltzmann fit (solid line) gave values for this cell of V 1/2 of Ϫ50.8 mV and k ϭ 4.6 mV. The mean values were V 1/2 ϭ Ϫ51.6 Ϯ 1.8 mV and k ϭ 4.2 Ϯ 0.4 mV (n ϭ 3).
ing a tight interaction with these residues. 3) Replacement of individual residues within this loop with Ala identifies a third charged residue (Lys 1617 ) and six uncharged residues that significantly affect LqTx and/or ATX II affinity. It is likely that Glu 1613 and the IVS3-S4 loop constitute an important component of neurotoxin receptor site 3, the ␣-scorpion toxin and sea anemone toxin receptor site.
Previous photoaffinity labeling studies led to covalent incorporation of photoreactive ␣-scorpion toxin derivatives into the IS5-S6 loop of the ␣ subunit (28), and anti-peptide antibodies directed against sequences in the IS5-SS1, ISS2-S6, and IVS5-SS1 segments of the ␣ subunit reduce ␣-scorpion toxin binding (29). While it was suggested that the IVS3-S4 loop may play a role in ␣-scorpion toxin binding (29), this extracellular loop was not experimentally identified as part of the ␣-scorpion toxin receptor site in previous studies, perhaps because the sitedirected antibodies and the photoreactive moiety on the ␣-scorpion toxin derivatives could not interact with this short loop. Our present results show that negatively charged amino acid residues in the IS5-S6, IVS5-SS1, and IVSS2-S6 loops are not required for ␣-scorpion toxin binding, but that both negatively charged and neutral amino acid residues in the IVS3-S4 loop are required for high affinity binding. Further mutagenesis studies will be required to identify the individual amino acid residues in the IS5-S6, IVS5-SS1, and IVSS2-S6 loops which also participate in toxin binding.
Acidic amino acids of the Na ϩ channel have been proposed to be important for ␣-scorpion toxin and anemone toxin binding based on structural information about these toxins (10 -16). The identification of Glu 1613 as an important determinant of ␣-scorpion toxin binding and Glu 1613 and Glu 1616 as determinants of anemone toxin binding is consistent with an electrostatic interaction between acidic residues of the Na ϩ channel and basic residues of ␣-scorpion toxins or sea anemone toxins. These mutations also demonstrate that this loop contains more determinants for binding of ATX II than LqTx, and therefore that the receptor sites for ␣-scorpion toxin and anemone toxins are overlapping but not identical. Previous studies with anemone toxins have implicated the conserved Arg 14 residue in anemone toxin binding (14). Recent mutagenesis studies of the Anthopleurin anemone toxins found that double mutants which neutralize both Arg 12 and Arg 14 in ApB have the most dramatic reduction in affinity among the double mutants, while mutations of residues Arg 12 and Lys 49 alter the cardiac selectivity of ApA (16). Based on the residues of the Na ϩ channel IVS3-S4 loop identified in toxin binding, residues 12 and 14 of ApB may interact with Glu 1613 and Glu 1616 of the Na ϩ channel ␣ subunit, whereas Lys 49 may interact with residues outside of this loop to contribute to cardiac selectivity. Sea anemone toxins are different from ␣-scorpion toxin in many respects, including smaller size, unique sequence, distinct three-dimensional structure, generally lower affinities for neuronal channels and higher affinities for cardiac Na ϩ channels, and weaker inhibition of binding with depolarization (2, 7-9, 13, 14, 18, 22). Although we did not identify acidic binding determinants in regions previously implicated in ␣-scorpion toxin binding (IS5-S6 and IVS5-S6; Refs. 28 and 29), non-acidic residues in these other regions may contribute to ␣-scorpion toxin binding by providing unique determinants that are involved in interactions with LqTx but not ATX II. This is consistent with the biochemical evidence for the involvement of the IS5-S6 and IVS5-S6 loops in toxin binding (28,29), and with the previously suggested concept of multiple attachments points for ␣-scorpion toxin binding (10,11,14). Our results indicate that ATX II and LqTx bind to overlapping, but not identical, determinants in the IVS3-S4 loop which form part of neurotoxin receptor site 3.
Amino Acid Residues in Segment IVS3 Cause Differences in ␣-Scorpion Toxin Binding between Brain and Cardiac Na ϩ Channels-Na ϩ channels in neurons have a significantly higher affinity for LqTx than Na ϩ channels in cardiac cells (46). Consistent with this, rat brain type IIa Na ϩ channels have 4 -10-fold higher affinity than rat cardiac H1 channels when expressed in tsA-201 cells. However, the difference in binding affinity between these two cloned and expressed channels is not as great as the difference observed between neuronal and cardiac cells in cell culture (65-fold, Ref. 46). This may reflect differences in the membrane potential in different cell populations, in the voltage dependence of toxin binding between the two Na ϩ channel isoforms, in the assay methods used in the different studies (toxin-stimulated ion flux versus toxin binding), or in the channel processing and second messenger modulation between different cell types. Nevertheless, our results suggest that the difference in K D for ␣-scorpion toxin between the rIIa and rH1 ␣ subunit isoforms expressed in parallel in tsA-201 cells is due, at least in part, to amino acid sequence differences in the IVS3 transmembrane segment. These amino acid residues near the extracellular end of the IVS3 transmembrane segment may interact directly with the bound LqTx polypeptide themselves, or they may influence the position of the extracellular end of the IVS3 segment containing Glu 1613 or the conformation of the IVS3-S4 loop, which our results suggest are sites of direct toxin interaction.
Kinetics of ␣-Scorpion Toxin Binding-In our experiments, the kinetics of ␣-scorpion toxin binding have been determined over a wider range of voltages (Ϫ120 to ϩ100 mV) than in previous studies in order to determine rate constants and equilibrium dissociation constants. The association rate constants determined in the present experiments at Ϫ100 mV for wild type and E1613R (1.82 ϫ 10 7 and 2.20 ϫ 10 7 M Ϫ1 s Ϫ1 , respectively) were faster than the k 1 value of 1.5 ϫ 10 5 M Ϫ1 s Ϫ1 determined electrophysiologically at Ϫ100 mV with a scorpion toxin of lower affinity (␣LqIIa, Ref. 22), but quite similar to the k 1 value of the higher affinity AahII scorpion toxin (k 1 ϭ 1.5 ϫ 10 7 M Ϫ1 s Ϫ1 ) determined biochemically (49). By using the K D values determined at Ϫ100 mV, the corresponding calculated k Ϫ1 values of 3.09 ϫ 10 Ϫ2 s Ϫ1 and 2.33 s Ϫ1 for the wild-type and E1613R channels, respectively, are faster than the k Ϫ1 values determined for ␣LqIIa or AaH II (both 1.6 ϫ 10 Ϫ3 s Ϫ1 , Refs. 22 and 49). Thus, the association rate constants determined at negative potentials are consistent with previous work using other ␣-scorpion toxins, the calculated dissociation rate constants are somewhat faster than for those other toxins, and the difference in affinity of the wild-type and E1613R channels at negative potentials is entirely due to the difference in the dissociation rate.
Voltage Dependence of ␣-Scorpion Toxin Binding at Equilibrium-The proximity of the IVS3-S4 loop to the voltage-sensing IVS4 transmembrane segment provides a potential molecular basis for understanding the voltage dependence of toxin binding and the coupling of activation to inactivation. The voltagedependence of equilibrium binding is well described by a Boltzmann distribution (19 -21, Fig. 10), and the voltage-dependent changes in affinity are observed in the same range (Ϫ80 to Ϫ40 mV) over which voltage-dependent gating transitions occur within the channel. The midpoint for the voltage-dependent change in affinity for ␣-scorpion toxin is essentially identical for wild-type and mutant channels. The steepness of the change in ␣-scorpion toxin binding affinity (k ϭ 4.2-4.6 mV) is also similar between wild-type and mutant channels, but is steeper than previously reported for other ␣-scorpion toxins binding to Na ϩ channels in amphibian node of Ranvier (k ϭ 6.0 -9.8 mV, Ref. 20; k ϭ 9.5 mV, Ref. 21; k ϭ 11.5, Ref. 23). Wang and Strichartz (23) noted that both the range and steepness of voltage dependence were dependent on the toxin studied, which may explain most of the quantitative differences in voltage dependence in different studies. The similarity in voltage dependence of LqTx binding to wild-type and E1613R channels indicates that the 30 -60-fold change in affinity as a result of the E1613R mutation does not strongly affect the voltagedependent transition between channel states in the voltage range from Ϫ80 to Ϫ40 mV.
In our experiments, the voltage dependence of toxin binding to transfected rIIa channels was significantly more negative than the voltage dependence of activation, and steeper and slightly more positive than the voltage dependence of steadystate inactivation. In comparison to our results, the voltage dependence of ␣-scorpion toxin binding to neuroblastoma cells and frog sartorius muscle measured in equilibrium binding experiments was much more positive than steady-state inactivation and correlated approximately with the voltage dependence of activation (17,19), while the voltage dependence of toxin binding to Na ϩ channels measured electrophysiologically in frog node of Ranvier was 20 mV more positive than steadystate inactivation, and was positioned between the voltage dependence of activation and inactivation (20,21). These differences in voltage dependence of toxin binding relative to channel gating may result from inherent isoform-or speciesspecific differences in gating among different Na ϩ channels.
The process of inactivation is thought to be intrinsically voltage-independent and to acquire its voltage dependence from coupling to activation (50). Multiple voltage-dependent transitions between closed states occur during the activation process (50). These voltage-dependent transitions among closed states along the activation pathway are likely to be responsible for voltage-dependent coupling of activation to steady-state inactivation and for voltage-dependent changes in affinity for ␣-scorpion toxins and sea anemone toxins. The voltage dependence of transitions among closed states leading to activation falls between that of activation and steady-state inactivation (50). The slowing of inactivation and the reduction in the steepness of voltage-dependence of steady-state inactivation caused by ␣-scorpion toxins and sea anemone toxins (2,(17)(18)(19)(20)(21)(22)(23) suggest that the toxin receptor site undergoes a conformational change that is required for fast inactivation, that bound toxin slows this conformational change and thereby slows the inactivation process, and that toxin binding is destabilized as a result of conformational changes which lead ultimately to inactivation. Therefore, our results implicate the S3-S4 loop in domain IV of the ␣ subunit in coupling of activation to inactivation.
Voltage Dependence of Dissociation of ␣-Scorpion Toxin in the Positive Potential Range-The K D for ␣-scorpion toxin remained constant from Ϫ40 mV up to at least ϩ100 mV. In contrast, dissociation rates for ␣-scorpion toxin increased steadily between 0 and ϩ100 mV. The rapid rates of dissociation in this voltage range may reflect the rate of change of state of the toxin-channel complex from a high affinity conformation at negative membrane potentials to a low affinity conformation at positive potentials as well as toxin unbinding itself. Voltagedependent dissociation is likely to be driven by the voltage-dependent conformational change to the low affinity state. The voltage dependence of LqTx dissociation can be fit with a single exponential equation (21,22,51) and the voltage dependence of dissociation from wild-type and E1613R channels (e-fold/22-25 mV and e-fold/36 mV, respectively) is similar to previous reports (e-fold/25 mV, Ref. 21; e-fold/32 mV, Ref. 51). The significance of the apparent difference in steepness of the voltage dependence of toxin dissociation at positive membrane poten-tials between wild-type and E1613R channels (Fig. 10) is not clear, as the voltage-dependence of equilibrium binding in the negative voltage range where channel gating occurs is similar between the two channels (Fig. 9). These voltage-dependent changes in dissociation kinetics at positive potentials may reflect voltage-dependent transitions in the toxin-channel complex which are outside the normal voltage range of channel gating and may differ between wild-type and mutant channels.
A Model for Binding and Action of ␣-Scorpion Toxins and Sea Anemone Toxins-Na ϩ channel gating currents are the most direct measurement of voltage-dependent conformational changes in ion channels and may represent outward movement of the S4 segments (3). Both ␣-scorpion toxin and sea anemone toxins reduce the effective gating current, abolish the later, slower components of the gating current, and slow or block the immobilization of gating charge that occurs following normal channel activation (52,53). The S4 segment in the domain IV of the Na ϩ channel has been shown to move toward the extracellular space during depolarization (54). Toxin bound across the IVS3-S4 extracellular loop may slow this movement or subsequent conformational changes that are necessary for fast inactivation. The S4 segments of the Electrophorus electroplax, squid optic lobe, and all sequenced vertebrate Na ϩ channels contain 4, 5, 6, and 8 positive charges in domains I, II, III, and IV, respectively (e.g. Refs. 30, 32, and 34). Therefore, in response to depolarization the IVS4 segment may move further, at a different rate, or in a different voltage range than the other voltage sensors. Translocation of IVS4 may be required for the inactivation gate to close, and ␣-scorpion toxin or sea anemone toxin bound at the extracellular end of IVS4 may slow or block this translocation, preventing inactivation and gating charge immobilization. In fact, several mutations of the skeletal muscle Na ϩ channel responsible for paramyotonia congenita also lie within this loop at positions corresponding to Leu 1614 and Arg 1624 (55). Genetic defects at these positions slow inactivation (55), disrupt coupling of activation to inactivation (54), and FIG. 11. A model for the ␣-ScTx/ATX receptor. A cross-sectional view of the Na ϩ channel depicting the juxtaposition of domains I and IV on one side of the ion pore with ␣-ScTx or ATX bound. Also shown are the regions previously implicated in ␣-ScTx binding (lightly shaded segments), sites of glycosylation (), the intracellular III-IV loop which acts as the inactivation gate (IFM) and contains a known site of phosphorylation (P), and the IVS3-S4 loop closely interacting with bound toxin (dark shaded segment). Outward movement of the IVS4 segment with depolarization is inhibited by bound toxin and accelerates the off-rate of bound toxin. Slowing of this translocation or of a subsequent conformational change in the IVS3-S4 loop is proposed to slow the rate of fast inactivation. result in muscular dysfunction. Together with this previous work, our results therefore identify a short extracellular loop of the Na ϩ channel molecule that is critical for the binding of ␣-scorpion and sea anemone toxins and is also important for the coupling of channel activation to fast inactivation. Fig. 11 illustrates a model of the ␣-scorpion toxin receptor site based on the LqTx and ATX II binding determinants reported here, and two regions previously implicated in ␣-scorpion toxin binding (28,20). In this model, ␣-scorpion toxins are proposed to bind across the IVS3-S4 loop through electrostatic interactions with Glu 1613 and additional, unidentified contacts in the IS5-S6 and IVS5-S6 loops. The voltage dependence of ␣-scorpion toxin and anemone toxin binding is due to steric interactions as the IVS4 segment moves outward in response to depolarization and encounters bound toxin. Torsion between the binding of ␣-scorpion toxin to Glu 1613 and other determinants outside of the IVS3-S4 loop results in an increased rate of toxin dissociation. The binding of sea anemone toxins differs in that they bind intimately with several residues of the IVS3-S4 loop through electrostatic, hydrogen-bonding, or van der Waals interactions. Sea anemone toxin binding is less voltage-dependent (18,23), perhaps because it has fewer binding contacts outside of the IVS3-S4 loop so is subjected to less steric or torsional distortion when the channel is depolarized. Both of these toxins may slow inactivation by slowing or preventing the resulting conformational changes in the IVS3-S4 loop upon translocation of the IVS4 segment.