Molecular Determinants of Voltage-dependent Gating and Binding of Pore-blocking Drugs in Transmembrane Segment IIIS6 of the Na+ Channel α Subunit*

Mutations of amino acid residues in the inner two-thirds of the S6 segment in domain III of the rat brain type IIA Na+ channel (G1460A to I1473A) caused periodic positive and negative shifts in the voltage dependence of activation, consistent with an α-helix having one face on which mutations to alanine oppose activation. Mutations in the outer one-third of the IIIS6 segment all favored activation. Mutations in the inner half of IIIS6 had strong effects on the voltage dependence of inactivation from closed states without effect on open-state inactivation. Only three mutations had strong effects on block by local anesthetics and anticonvulsants. Mutations L1465A and I1469A decreased affinity of inactivated Na+ channels up to 8-fold for the anticonvulsant lamotrigine and its congeners 227c89, 4030w92, and 619c89 as well as for the local anesthetic etidocaine. N1466A decreased affinity of inactivated Na+ channels for the anticonvulsant 4030w92 and etidocaine by 3- and 8-fold, respectively, but had no effect on affinity of the other tested compounds. Leu-1465, Asn-1466, and Ile-1469 are located on one side of the IIIS6 helix, and mutation of each caused a positive shift in the voltage dependence of activation. Evidently, these amino acid residues face the lumen of the pore, contribute to formation of the high-affinity receptor site for pore-blocking drugs, and are involved in voltage-dependent activation and coupling to closed-state inactivation.

Clinically important drugs, including local anesthetics and some anticonvulsants and antiarrhythmics, exert their therapeutic effects by binding preferentially to the inactivated state and blocking voltage-gated Na ϩ channels. Local anesthetics and the related phenylalkylamine blockers of Ca 2ϩ channels have been characterized biophysically as pore blockers (8 -10) and therefore provide molecular probes for identification of pore-lining amino acid residues. Photoaffinity labeling studies of phenylalkylamine binding to L-type Ca 2ϩ channels initially implicated the IVS6 segment in forming the inner pore lining (11). Subsequent mutagenesis studies of Na ϩ channels and Ca 2ϩ channels (12,13) identified analogous sets of amino acid residues in the IVS6 segment that form the receptors sites for these drugs and therefore line the inner pores. Mutations F1764A and Y1771A in segment IVS6 reduced affinity of inactivated Na ϩ channels for the local anesthetic etidocaine by up to two orders of magnitude (12). Mutation of these homologous residues also substantially reduced block of inactivated Na ϩ channels by other local anesthetic, antiarrhythmic, and anticonvulsant drugs in brain type IIA, brain type III, and skeletal muscle Na ϩ channels (12, 14 -20).
X-ray crystallographic analysis of a K ϩ channel from Streptomyces lividans (KcsA) 1 shows directly that the channel pore is formed by transmembrane segments analogous to the S6 segments and S5-S6 pore loops of Na ϩ and Ca 2ϩ channels (21). A water-filled cavity on the intracellular side of the ion selectivity filter includes the amino acid residues analogous to Phe-1764 and Tyr-1771 of Na ϩ channels and likely corresponds to the binding site of pore-blocking drugs. In this structure, the S6-like segments cross to form an apparent barrier to ion movement at the intracellular end of the pore. Therefore, in addition to their role in forming the inner pore lining, the S6 segments may also serve as the activation gate and move during activation (22,23).
Because local anesthetics bind in a cavity in the pore formed by a symmetrical array of all four S6 segments, it is likely that the S6 segments in domains I, II, and III of Na ϩ channels also contribute to the receptor sites for local anesthetic, antiarrhythmic, and anticonvulsant drugs. Consistent with this hypothesis, site-directed mutagenesis of the skeletal muscle Na ϩ channel showed that replacement of Asn-434 and Leu-437 within transmembrane segment IS6 with lysine significantly reduced block by etidocaine, suggesting that the positive charge of the inserted lysine residue can disrupt normal drug * This work was supported by National Institutes of Health Grant NS15751 (to W. A. C.) and by research funds from Glaxo-Wellcome plc. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  binding (24). In this study, we have used alanine-scanning mutagenesis to investigate the role of amino acid residues in transmembrane segment IIIS6 in channel gating and block by anticonvulsant and local anesthetic drugs. Our results further define the pore-lining residues of the inner pore of Na ϩ channels and identify new components of the receptor site for local anesthetics and related pore-blocking drugs.

EXPERIMENTAL PROCEDURES
Mutagenesis of rIIA Channels-Mutations were prepared by a twostep polymerase chain reaction protocol using two mutagenic primers and two restriction site primers. The mutagenic fragment and the plasmid pCDM8rIIA were digested with BstEII and BlpI restriction endonucleases. The mutagenic fragment was then subcloned into the plasmid pCDM8rIIa via those restriction sites. The mutations were confirmed by restriction mapping and DNA sequence analysis.
Na ϩ Channel Expression in Xenopus Oocytes-Plasmids encoding wild-type and mutant Na ϩ channel ␣ subunits and wild-type ␤ 1 subunits were linearized, and RNA was transcribed as described previously (25). Xenopus laevis oocytes were harvested, maintained, and injected with RNA by standard methods as described previously (25).
Two-microelectrode Voltage-clamp Recordings from Oocytes-Na ϩ channel recordings were obtained from injected oocytes using a Dagan CA-1 voltage clamp (Dagan Corp.) as described previously (25). The bath was continuously perfused with Ringer solution containing (in mM): 115 NaCl, 2.5 KCl, 1.8 CaCl 2 , 10 HEPES, pH 7.2, adjusted with NaOH. Recording electrodes contained 3 M KCl and had resistances of Ͻ0.5 megohm. Stock solutions of lamotrigine and 619c89 (Glaxo Wellcome) were prepared in 25 mM HCl, 227c89 (Glaxo Wellcome) was prepared in water, and 4030w92 (Glaxo Wellcome) and etidocaine (Astra) were dissolved in dimethyl sulfoxide. All drugs stocks were then diluted to the desired concentration in the bath solution.
Analysis of Periodicity of Gating Perturbations-To evaluate the gating perturbations caused by IIIS6 mutants, we calculated the difference in Gibb's free energy between closed and open states at 0 mV (⌬G 0 ) according to ⌬G 0 ϭ ϪRTV 0.5 /k, where R is the gas constant, T is the absolute temperature in degrees K, V 0.5 is the half-maximal activation voltage, and k is a slope factor. V 0.5 and k values were determined from fits of activation curves to single Boltzmann functions. The difference in ⌬G 0 between wild-type and mutant channels was calculated according to ⌬G 0 ϭ ␦G mut 0 Ϫ ␦G wt 0 . To evaluate the periodicity of gating perturbations produced by IIIS6 mutants, we used Fourier transform methods (26 -29). The Fourier transform power spectrum (P()) was calculated according to the equation, , and is the angular frequency, n is the number of residues in a segment, V j is P⌬⌬G 0 P at a given position j, and ͗V͘ is the average value of ⌬⌬G 0 for the segment.

Effects of Mutations in Segment IIIS6 on Voltage-dependent
Activation of Na ϩ Channels-Alanine-scanning mutagenesis of amino acid residues in transmembrane segment IIIS6 of the ␣ subunit of the rat brain type IIA Na ϩ channel (rIIA) was used to investigate their functional role in activation, inactivation, and binding of pore-blocking drugs. Conversion of hydrophobic amino acid residues to alanine has only small effects on protein secondary structure but changes the size and chemical properties of the residue significantly (30,31). To identify functionally important residues in transmembrane segment IIIS6, we substituted alanine for individual native amino acid residues from Val-1454 to Ile-1473. We coexpressed the wild-type and mutant Na ϩ channel ␣ subunits and wild-type ␤ 1 subunits in Xenopus oocytes and measured Na ϩ current using a two-microelectrode voltage clamp.
Periodicity of Effects of Mutations on Voltage-dependent Activation-By analogy with the bacterial K ϩ channel KcsA, whose structure is known (21), S6 transmembrane segments are assumed to be ␣-helical in structure and to line the inner aspect of the ion-conducting pore. Therefore, one surface of each S6 helix is expected to face the lumen of the pore, whereas the other sides interact with adjacent transmembrane segments. S6 transmembrane segments are proposed to change A, conductance-voltage curves for wild-type, L1467A, F1468A, and I1469A. Peak current versus voltage relationships were measured using 30-ms test pulses to potentials from Ϫ50 mV to ϩ70 mV from a holding potential of Ϫ90 mV. Conductance was determined as I P /(V R Ϫ V), where I P is peak inward current, V R is the reversal potential, and V is the test pulse voltage. Normalized conductance was fit with a single Boltzmann relationship of the form: where V 0.5 is the half-maximal activation voltage and k is a slope factor. Mean V 0.5 and k values were determined for each mutant. The curves shown are plots of the Boltzmann relationship using these mean values. B, V 0.5 for activation of IIIS6 mutants compared with the wild-type Na ϩ channels. The histogram shows the differences between voltage for the half-maximal activation of the wild-type and mutant Na ϩ channels. Mean V 0.5 values were obtained from Boltzmann fits of normalized conductance versus voltage plots as described above. The asterisks indicate significant differences from wild-type as determined by t test (p Ͻ 0.01). conformation via rotation during channel activation (22,23,32,33), leading to changes in side-chain interactions with neighboring ␣-helices. Power spectral analysis of periodicity in the effects of mutations in successive positions along the ␣-helix can reveal position-dependent interactions with neighboring structural elements during activation gating (34 -36).
Changes in free energy of activation (⌬⌬G 0 ) for each mutation were calculated from the results of Fig. 1 and plotted in Fig. 2A. The ⌬⌬G 0 values parallel the changes in the voltage for half-maximal activation from which they were derived (Fig.  1B). Inspection of the ⌬⌬G 0 values reveals that they are consistently negative for the six most extracellular residues of the helix, indicating that these mutations favor activation. This pattern is interrupted at G1460A, mutation of which produces a strongly positive ⌬⌬G 0 value. Progressing inward through the IIIS6 segment, the ⌬⌬G 0 values assume a periodic pattern of positive and negative changes. These results suggest that the mutated residues make structurally or functionally important contacts with other transmembrane segments that differ depending on their location around the IIIS6 helix. If interactions of IIIS6 with different surrounding transmembrane segments have different effects on ⌬⌬G 0 for activation, a pattern of ⌬⌬G 0 values with ␣-helical periodicity of approximately 100°is expected. Power spectral analysis of the effects of mutations of the residues intracellular to G1460 that caused positive ⌬⌬G 0 values yields a peak at approximately 97° (Fig. 2B), consistent with an ␣-helical structure in which the native residues on one face of the helix make interactions that favor activation. Residues intracellular to G1460 with negative ⌬⌬G 0 values yielded a less well-defined peak of angular periodicity at 101°, suggesting that a different, broader face of the ␣-helix makes interactions that favor the resting state of the channel and oppose activation (data not shown). All mutations except I1458A and F1468A make relatively small (Ͻ1 kcal) perturbations in ⌬⌬G 0 , indicating that the change in energy of interaction for any single residue is relatively small.
Effects of Mutations in the IIIS6 Segment on Inactivation-Most mutations of amino acid residues in the cytoplasmic half of the IIIS6 segment also caused strong shifts in the voltage dependence of inactivation during 100-ms prepulses to the indicated membrane potentials (Fig. 3A). For example, the voltages for half-maximal inactivation of mutations L1465A and L1467A were shifted by Ϫ11 and Ϫ14 mV, respectively (Fig. 3A). Significant negative shifts were also caused by mutations F1463A, N1466A, F1468A, G1470A, V1471A, and I1472A (Fig. 3B). In contrast, mutations I1473A and I1469A caused significant positive shifts in the voltage dependence of  (34,35). B, angular frequency of positive shifts in ⌬⌬G 0 determined for mutations of residues in the intracellular portion of IIIS6 between F1462A and I1473A (26,36). Periodicity was evaluated using the Fourier transform power spectrum P() and plotted versus angular frequency.

FIG. 3. Steady-state inactivation of the wild-type and mutant
Na ؉ channels. A, steady-state inactivation curves for wild-type, L1465A, L1467A, and I1469A. Inactivation curves were measured using 100-ms prepulses to the indicated potentials followed by a test pulse to 0 mV. Peak test pulse current was plotted as a function of prepulse potential, normalized, and fit with a Boltzmann function, , where V 0.5 is the membrane potential at the halfmaximal current and k is a slope factor. Mean V 0.5 and k values were determined for each mutant. Curves shown are plots of the Boltzmann function using these mean values. B, V 0.5 for inactivation of IIIS6 mutants compared with the wild-type Na ϩ channels. The histogram shows the differences between voltage for the half-maximal inactivation of the wild-type and mutant Na ϩ channels. V 0.5 values were obtained from Boltzmann fits of normalized current versus voltage plots. The asterisks indicate significant differences from wild-type as determined by t test (p Ͻ 0.01). inactivation (Fig. 3, A and B). Inactivation gating was unaffected for most of the mutations in the extracellular half of IIIS6. Only I1458A caused a significant negative shift (Fig. 3B). The two mutations giving positive shifts (I1469A and I1473A) are four residues apart, and mutations F1463A, L1467A, and V1471A, which define local peaks in the negative shifts of inactivation, are also four residues apart in the sequence. Therefore, the changes in voltage dependence of inactivation for mutations at the intracellular end of IIIS6 also conform approximately to an ␣-helical periodicity of 3.6 residues per 360°turn.
Effect of Mutations in Transmembrane Segment IIIS6 on Affinity of Inactivated Na ϩ Channels for Anticonvulsants and Local Anesthetics-The anticonvulsant lamotrigine, which is used for treatment of epilepsy and bipolar disorder, acts by blocking brain Na ϩ channels in a voltage-and frequency-dependent manner (37)(38)(39). Etidocaine, an effective local anesthetic that is used for regional anesthesia, also blocks brain Na ϩ channels with strong voltage and frequency dependence (40). The efficacy of these drugs as anticonvulsants and antiarrhythmics stems from their ability to selectively block Na ϩ channels during abnormal membrane depolarizations and rapid bursts of action potentials that characterize neuronal and cardiac pathologies (8,9,41). The selectivity of these drugs for Na ϩ channels in depolarized cells results from the preferential binding to the open and inactivated states that predominate at depolarized membrane potentials. This state-dependent drug action can be explained by an allosteric model in which a modulated drug receptor is in a low affinity conformation when the channel is in the resting state and converts to a high affinity conformation when the channel is inactivated by depolarization (8,9).
Etidocaine, lamotrigine, and related compounds have been previously used to identify amino acid residues involved in binding of pore-blocking drugs in transmembrane segment IVS6 (12,42). For primary screening of mutants in segment IIIS6 for their effect on drug binding to the inactivated state of the Na ϩ channel, block of inactivated Na ϩ channels by lamotrigine or compound 619c89, a tricyclic lamotrigine congener (43), was determined during a test pulse to 0 mV following a 15-s depolarization to a holding potential at which 70 -80% of the Na ϩ current was inactivated (Fig. 4A, control trace c). At such a depolarized holding potential (Ϫ50 mV), addition of 100 M lamotrigine (Fig. 4A, trace d) reduced the Na ϩ current by 75%. At the same concentration, no significant block of Na ϩ current was observed when the holding potential was Ϫ120 mV (Fig. 4A, traces a and b). From experiments like the one illustrated in Fig. 4A, we determined the dissociation constant for the inactivated state (K I ) for wild-type and mutant Na ϩ channels according to Kuo and Bean (44). Apparent affinities for lamotrigine are presented in Fig. 4B and for 619c89 in Fig. 4C. Wild-type Na ϩ channels were inhibited by lamotrigine and 619c89 with dissociation constants of 32 and 10 M, respectively, in agreement with previously reported results (42). L1465A and I1469A decreased the affinity of lamotrigine for the inactivated state of the channel, resulting in 8-and 3-fold increases in K I , respectively (Fig. 4B). The same mutations increase K I of 619c89 for the inactivated state of the channel approximately 3-fold (Fig. 4C). Both L1465A and I1469A caused positive shifts in voltage dependence of activation (Fig.  1B). To test whether these shifts in activation could account for the observed decrease in drug affinity by preventing full activation of the Na ϩ channels, we also examined inactivated-state affinity using strong depolarizations for which both wild-type and mutant channels were fully activated. We applied pulses to ϩ20 mV for 15 s followed by a short interpulse interval at Ϫ120 mV to allow recovery from fast inactivation but not from drug block, and then by 15-ms test pulse to ϩ20 mV. This protocol yielded dissociation constants for the drugs in agreement with those reported in Fig. 4B (data not shown).
Because mutations L1465A and I1469A had effects on the affinity of lamotrigine, we examined the affinities of these mutants for the structurally related compounds 4030w92 and 227c89 as well as the structurally unrelated local anesthetic etidocaine. L1465A, N1466A, and I1469A decreased the affinity of 4030w92 5-, 3-, and 5-fold, respectively (Fig. 5A). Block by compound 227c89 was disrupted in L1465A and I1469A mutants by about 3-fold (Fig. 5B). Larger disruptions in block by etidocaine were observed with mutations L1465A, N1466A, and I1469A resulting in 6-, 8-, and 7-fold increases in K I , respectively (Fig. 5C). Thus, mutations L1465A and I1469A decreased the affinity of all compounds tested substantially. In contrast, mutation N1466A caused the largest disruption of etidocaine binding but had no effect on binding of compound 227c89 or of lamotrigine and 619c89.
Effect of Mutations L1465A, N1466A, and I1469A on Block of FIG. 4. Affinity of inactivated wild-type and mutant Na ؉ channels for lamotrigine and 619c89. A, representative current traces for wild-type Na ϩ channels in the absence (traces a and c) or in the presence of 100 M lamotrigine (traces b and d). Traces a and b were measured during a test pulse to 0 mV following a 15-s step to Ϫ120 mV. Traces c and d were measured at 0 mV following a 15-s step to Ϫ50 mV. B and C, apparent dissociation constants for block of inactivated channels by lamotrigine (B) and 619c89 (C). For mutant Na ϩ channels, the depolarized holding potential was varied so that 70 -80% of the channels were inactivated after the 15-s conditioning prepulse. The dissociation constant for the inactivated state (K I ) for each mutant was calculated according to Kuo and Bean (44) where h is the fraction of inactivated channels, E max is the maximal block that is assumed to be the complete block of the current, and E is the amount of block at drug concentration of [D]. Error bars indicate S.E. The asterisks indicate significant differences from wild-type as determined by t test (p Ͻ 0.01).
Resting Na ϩ Channels by Lamotrigine and Etidocaine-To determine whether IIIS6 mutants affected block of resting Na ϩ channels at negative membrane potentials, we applied 15-ms test pulses from holding potentials of Ϫ80 to Ϫ120 mV in the absence and the presence of 500 M lamotrigine. Mutations L1465A, N1466A, and I1469A increased affinity of lamotrigine for resting channels approximately 2-fold (Fig. 6A). L1465A and N1466A also increased affinity of etidocaine for resting channels 3-fold (Fig. 6B). In contrast, I1469A did not affect the affinity of etidocaine for resting channels (Fig. 6B). No IIIS6 mutants caused significant decreases in resting affinity of lamotrigine or etidocaine (data not shown). Thus, the loss of drug binding affinity of these mutant channels is specific for the high affinity inactivated state.
Effect of Mutations F1462A, L1465A, N1466A, and I1469A on Frequency-dependent Block of Na ϩ Channels-Frequencydependent block of Na ϩ channels is observed for drugs that bind to the channel rapidly in the open state and then bind with high affinity to the inactivated state (8,9). Compounds 619c89 and etidocaine produce strong frequency-dependent block of the rat brain type IIA Na ϩ channels (Fig. 7) (12, 42). To determine whether IIIS6 mutants alter frequency-dependent block by compound 619c89, we applied 10-Hz trains of 20-ms pulses to 0 mV from a holding potential of Ϫ90 mV and recorded Na ϩ currents. Only I1469A significantly reduced frequency-dependent block by 619c89 compared with the wildtype (Fig. 7A). In contrast, mutations F1462A and N1466A substantially increased frequency-dependent block. Surprisingly, given the reduced affinity of these inactivated channels for block by 619c89, L1465A had no significant effect on fre-quency-dependent block. Mutations L1465A and I1469A significantly reduced use-dependent block by etidocaine during 2-Hz trains (Fig. 7B). In contrast, mutations N1466A and F1462A had little effect on frequency-dependent block by etidocaine (Fig. 7B).
Effects of Mutations F1462A, L1465A, N1466A, and I1469A on Recovery of Na ϩ Channels from Inactivated-state Block-Recovery of drug-blocked inactivated Na ϩ channels to the resting state was studied for the IIIS6 mutants that affected voltage-and/or frequency-dependent block of Na ϩ channels by 619c89 and etidocaine. We measured the rate of recovery by applying a 500-ms conditioning prepulse to 0 mV to produce drug block of inactivated channels followed by a recovery interval of variable duration and a test pulse to 0 mV. In control conditions, recovery after depolarization follows a double-exponential time course (Fig. 8A). In the presence of 50 M 619c89, the fast time constant reflects recovery from inactivation of the small fraction of channels that was not blocked during the conditioning prepulse. The slow time constant reflects slow dissociation of the drug from the channels that were blocked during the conditioning prepulse ( drug ). For drug concentrations in which a large fraction of channels was blocked during the depolarization, only a single drug-induced recovery component was observed despite the two exponential components observed in control. Slow time constants ( drug ) induced by 619c89 are presented in Fig. 8C. For wild-type channels drug was 384 Ϯ 15 ms. drug for F1462A was not significantly different (379 Ϯ 27 ms) (Fig. 8, A and C). In contrast, I1469A recovered with drug of 187 Ϯ 6 ms (Fig. 8, B and C), 2-fold faster than the wild-type recovery rate. Recovery from block of both L1465A and N1466A was about 2-fold slower than for wild-type channels with drug values of 710 Ϯ 61 ms and 820 Ϯ 30 ms, respectively. The slower dissociation of 619c89 from L1465A would counteract the effect of reduction in binding affinity by that mutation, and the slower dissociation from N1466A would enhance frequency-dependent block.
Wild-type Na ϩ channels recovered from block by etidocaine with drug of 2.3 Ϯ 0.2 s, 6-fold slower than for compound 619c89. F1462A and I1469A recovered with drug values of 1.7 Ϯ 0.3 s and 2.0 Ϯ 0.3 s, respectively, which were not significantly different from the wild-type channels. L1465A and N1466A recovered with drug values of 0.76 Ϯ 0.05 s and 1.05 Ϯ 0.05 s, respectively (Fig. 8D), which were approximately 3-and 2-fold faster than the recovery of wild-type channels.
We also examined the rate of onset of block by 619c89 and etidocaine of F1462A, L1465A, N1466A, and I1469A during depolarizations. Only F1462A affected onset block by 619c89 significantly. Onset of block was 2.7-fold faster than with the wild-type channel (data not shown). This is consistent with the faster and greater development of use-dependent block during a train by 619c89 in this mutant channel (Fig. 7A). Only L1465A had a significant effect on the rate of onset of block by etidocaine, which was 3-fold slower compared with the wildtype channel (data not shown).  (Figs. 1B and 2A). The positive shift of the activation curve of these mutants indicates that the native residues at these positions make interactions that stabilize the open channel or destabilize closed channels. These amino acid residues spread across a 100°section of the circumference of the IIIS6 ␣-helix. They are proposed to face the lumen of the pore in the activated and inactivated states of the channel, because Leu-1465 and Ile-1469 are required for high affinity binding of the pore-blockers lamotrigine and etidocaine (Fig. 9).
Thr-1464, Leu-1467, Phe-1468, Gly-1470, Val-1471, and Ile-1472 are spread across 200°of the opposite side of the circumference of IIIS6 segment (Fig. 9). Point mutations of Thr-1464, Leu-1467, Phe-1468, Val-1471, and Ile-1472 to alanine induced negative shifts of the voltage for half-maximal activation (Figs.  1B and 2A). These residues are proposed to face away from the pore. The negative shifts caused by these mutations indicate that the native residues at these positions contact other protein components of the channel and stabilize the closed state and/or destabilize the open state.
This pattern of shifts in the voltage dependence of activation Recovery was measured using a 500-ms conditioning pulse to 0 mV followed by a recovery interval of the indicated duration (1.5-4000 ms) at Ϫ90 mV followed by a test pulse to 0 mV. The peak test pulse current was divided by the peak conditioning pulse current and plotted against the recovery interval. The curves are least-squares fits of a two-exponential function to the data. The asterisks indicate significant differences from wild-type as determined by t test (p Ͻ 0.01).
is consistent with an ␣-helical conformation of the IIIS6 segment and with a rotational motion of this helix during activation. Rotational movement of these mostly hydrophobic residues (i.e., Phe-1462, Leu-1465, Asn-1466, Ile-1469, and Ile-1473) to face the lumen of the pore during activation would break interactions with surrounding transmembrane segments in domain III and allow new interactions of these amino acid residues with the adjacent S6 segments from domains II and IV. These new interactions would be destabilized by alanine substitutions, resulting in impaired channel activation. Rotational motion of an S6-like segment has been detected with EPR probes in the KcsA channel (22), and movements of the S6 segment of K ϩ channels have been proposed to open the activation gate (22,23,32,33,35,45,46). Our results provide support for rotational movement of the IIIS6 segment of Na ϩ channels during activation.
This helical pattern of effects on activation for the inner two-thirds of the IIIS6 segment contrasts with our results for the extracellular one-third of the IIIS6 segment located on the N-terminal side of Gly-1460. In this outer part of the segment, reduction of size and hydrophobicity of residues by substitution of alanine caused negative shifts of activation at all positions except Ile-1457, indicating that interactions of the native residues at all these positions favor the closed state of the channel. By analogy with the structure of the KcsA channel (21), it is thought that this outer part of IIIS6 interacts with the pore loop that forms the narrow part of the conduction pathway rather than with the water-filled lumen of the pore. Therefore, this outer part of the transmembrane segment has all of its residues in contact with neighboring peptide segments rather than water. The different effects of mutations in the intracellular and extracellular parts of IIIS6 on the activation process may result from their different interactions; i.e. peptide segments on all sides for the extracellular part compared with the interaction of one face of the inner segment with water in the lumen of the pore.
It is of interest that Gly-1460 defines the boundary between these two parts of the IIIS6 segment. Glycine residues have no side chain and therefore provide points of flexibility where ␣-helical segments can kink or unwind. Gly-1460 may provide flexibility to allow partially or completely independent movements of the inner and outer parts of the S6 segment during activation gating. For example, the inner part of the helix may rotate to open the activation gate, while the outer part remains comparatively immobile and retains its association with the pore loop region.
Effects of Mutations in the Inner Part of the IIIS6 Segment on the Voltage Dependence of Closed-state Inactivation-Mutations of amino acid residues in the cytoplasmic half of the IIIS6 segment also strongly affected inactivation gating. Most residues in this region caused negative shifts in the voltage dependence of inactivation when mutated to alanine (e.g. Phe-1463, Leu-1465, Asn-1466, Leu-1467, Phe-1468, Gly-1470, Val-1471, and Ile-1472). In contrast, alanine substitutions of residues Ile-1469 and Ile-1473, which are proposed to face the pore lumen ( Fig. 9), induced strong positive shifts in the voltage for half-maximal inactivation. The voltage dependence of inactivation during long prepulses to membrane potentials more negative than Ϫ40 mV measures primarily the voltage dependence of inactivation from closed states, because channel openings are very rare at such negative potentials. In contrast to several mutations in the IVS6 segment (25,47), no alanine substitution mutations in IIIS6 disrupted inactivation from open states, because none slowed the rate of decay of the Na ϩ current. Thus, residues in the cytoplasmic half of IIIS6 appear to play an important role in conformational changes leading to channel inactivation from closed states but not from open states.
Mutations L1465A, N1466A, and G1470A produced opposite shifts of the voltage dependence of activation and inactivation, shifting activation to more positive potentials and inactivation to more negative potentials. These mutations therefore impair the coupling of channel activation to channel opening, allowing channels to activate and undergo closed-state inactivation, but not to open. Evidently, these amino acid residues participate both in closed-state inactivation and in the coupling of activation to channel opening.
A Receptor Site for Local Anesthetic and Anticonvulsant Drugs in Transmembrane Segments IIIS6 and IVS6 of the Na ϩ Channel-Our present results for transmembrane segment IIIS6 and similar data for transmembrane segment IVS6 (42) define components of the receptor site for the anticonvulsant lamotrigine and related drugs and for the local anesthetic etidocaine. In the IIIS6 segment, mutation of amino acid residues Leu-1465 and Ile-1469, which are located on the same side of the ␣-helix (Fig. 9), decreased the inactivated-state block by lamotrigine and related compounds and by etidocaine. Mutation N1466A also reduced inactivated-state block by compound 4070w92 and had a particularly strong effect on block by etidocaine. Local anesthetics like etidocaine are thought to bind in the channel pore based on biophysical studies (8). Thus, we propose that residues Leu-1465, Asn-1466, and Ile-1469 define a pore-facing surface of the IIIS6 ␣-helix and form part of receptor sites for anticonvulsant and local anesthetic drugs. Similarly, previous studies indicate that residues Phe-1764, Tyr-1771, and Ile-1760 of segment IVS6 face the pore and form part of the receptor site for anticonvulsants and local anesthet- ics (12,14,42). Evidently, this site for binding of pore-blocking drugs involves one face of at least two of the four symmetrically located S6 segments that form the inner pore. Further work will be required to assess the possible roles of amino acid residues in the S6 segments of domains I and II in the receptor site for anticonvulsant and local anesthetic drugs.
In previous work (42), we found that the pattern of effects of mutations of Ile-1760, Phe-1764, and Tyr-1771 in transmembrane segment IVS6 differed for lamotrigine and its three congeners, suggesting specific interactions of the different chemical moieties of these drugs with individual amino acid residues in the receptor site. In contrast, in the present experiments we found a more similar pattern of effects of the mutations in transmembrane segment IIIS6 on affinity for lamotrigine, its three congeners, and etidocaine. The affinity of inactivated Na ϩ channels for all of these drugs was substantially reduced by mutation of Leu-1465 and Ile-1469. Only mutation N1466A revealed drug-specific differences. The affinity for etidocaine was decreased 8-fold, the affinity for 4030w92 was reduced 3-fold, and the affinity for the other drugs was unaffected. Therefore, differential interactions with amino acid residues in transmembrane segment IVS6 may be primarily responsible for mediating drug-specific effects on Na ϩ channels, but the interactions of amino acid residues in transmembrane segment IIIS6 with these drugs are less likely to contribute to drug-specific effects within this family of compounds.
State-dependent Effects of Mutations in the IIIS6 Segment-In contrast to their inhibitory effects on inactivatedstate block, mutations L1465A, N1466A, and I1469A increased resting-state block by lamotrigine. L1465A and N1466A also increased resting-state block by etidocaine, but I1469A had no effect. We suggest that, in the resting conformation of the channel, each of these mutations that substitute alanine for a larger hydrophobic residue creates additional space for the drug molecule to reach its binding site in the resting state of the channel and thereby enhances resting-state block.
Frequency-dependent block involves drug entry and binding to the open state of the channel followed by stabilization of the bound drug during inactivation. Mutations L1465A and I1469A reduced frequency-dependent block by etidocaine, and I1469A reduced frequency-dependent block by 619c89. Measurement of on-and off-rates of 619c89 and etidocaine for L1465A, N1466A, and I1469A revealed different effects of these mutations on the kinetics of interaction of these drugs with their receptor site. None of these mutations had any effect on the rate of onset of channel block by 619c89. In contrast, L1465A decreased the onset rate of etidocaine block, whereas N1466A and I1469A had no effect. The rate of recovery from 619c89-bound channels at the resting membrane potential was faster for I1469A and slower for L1465A. In contrast, L1465A induced faster recovery of etidocaine-bound channels and I1469A had no significant effect. Thus, both faster recovery from block and reduced inactivated-state affinity contribute to the reduction in frequencydependent block by I1469A, whereas reduced inactivated-state affinity is the primary effect for L1465A. Increased use-dependent block of N1466A by 619c89 is due to slower recovery rate from block by this compound. Reduced affinity for inactivated channels, slower onset rate, and faster recovery rate all contribute to the large decrease in frequency-dependent block of L1465A by etidocaine, whereas the smaller effect of I1469A on frequency-dependent block is due entirely to reduced affinity for inactivated-state block. The complex effects of these mutations on the kinetics of drug binding and dissociation suggest that interactions of these amino acid residues with different chemical moieties of the drug molecules can specifically affect the kinetics of these processes.
Overall, our results show that the effects of these mutations are greatest for high affinity drug binding to the inactivated state of Na ϩ channels and are smaller and variable for low affinity drug binding to resting and open states. We conclude that the high affinities of local anesthetic and anticonvulsant drugs for inactivated Na ϩ channels depend on interactions with Leu-1465 and Ile-1469 in the IIIS6 segment and Ile-1760, Phe-1764, and Tyr-1771 in the IVS6 segments. Gating movements of the IIIS6 and IVS6 segments may allow access of these drugs to their receptor site, which becomes available as the channel opens and increases in affinity as it inactivates.