A Critical Role for the S4-S5 Intracellular Loop in Domain IV of the Sodium Channel α-Subunit in Fast Inactivation*

Na+ channel fast inactivation is thought to involve the closure of an intracellular inactivation gate over the channel pore. Previous studies have implicated the intracellular loop connecting domains III and IV and a critical IFM motif within it as the inactivation gate, but amino acid residues at the intracellular mouth of the pore required for gate closure and binding have not been positively identified. The short intracellular loops connecting the S4 and S5 segments in each domain of the Na+ channel α-subunit are good candidates for this role in the Na+ channel inactivation process. In this study, we used scanning mutagenesis to examine the role of the IVS4-S5 region in fast inactivation. Mutations F1651A, near the middle of the loop, and L1660A and N1662A, near the COOH-terminal end, substantially disrupted Na+ channel fast inactivation. The mutant F1651A conducted Na+ currents that decayed very slowly, while L1660A and N1662A had large sustained Na+ currents at the end of 30-ms depolarizing pulses. Inactivation of macroscopic Na+currents was nearly abolished by the N1662A mutation and the combination of the F1651A/L1660A mutations. Single channel analysis revealed frequent reopenings for all three mutants during 40-ms depolarizing pulses, indicating a substantial impairment of the stability of the inactivated state compared with wild type (WT). The F1651A and N1662A mutants also had increased mean open times relative to WT, indicating a slowed rate of entry into the inactivated state. In addition to these effects on inactivation of open Na+channels, mutants F1651A, L1660A, and N1662A also impaired fast inactivation of closed Na+ channels, as assessed from measurements of the maximum open probability of single channels. The peptide KIFMK mimics the IFM motif of the inactivation gate and provides a test of the effect of mutations on the hydrophobic interaction of this motif with the inactivation gate receptor. KIFMK restores fast inactivation of open channels to the F1651A/L1660A mutant but does not restore fast inactivation of closed F1651A/L1660A channels, suggesting that these residues interact with the IFM motif during inactivation of closed channels. Our results implicate F1651, L1660, and N1662 of the IVS4-S5 loop in inactivation of both closed and open Na+ channels and suggest that the IFM motif of the inactivation gate interacts with F1651 and/or L1660 in the IVS4-S5 loop during inactivation of closed channels.

Na ؉ channel fast inactivation is thought to involve the closure of an intracellular inactivation gate over the channel pore. Previous studies have implicated the intracellular loop connecting domains III and IV and a critical IFM motif within it as the inactivation gate, but amino acid residues at the intracellular mouth of the pore required for gate closure and binding have not been positively identified. The short intracellular loops connecting the S4 and S5 segments in each domain of the Na ؉ channel ␣-subunit are good candidates for this role in the Na ؉ channel inactivation process. In this study, we used scanning mutagenesis to examine the role of the IVS4-S5 region in fast inactivation. Mutations F1651A, near the middle of the loop, and L1660A and N1662A, near the COOH-terminal end, substantially disrupted Na ؉ channel fast inactivation. The mutant F1651A conducted Na ؉ currents that decayed very slowly, while L1660A and N1662A had large sustained Na ؉ currents at the end of 30-ms depolarizing pulses. Inactivation of macroscopic Na ؉ currents was nearly abolished by the N1662A mutation and the combination of the F1651A/L1660A mutations. Single channel analysis revealed frequent reopenings for all three mutants during 40-ms depolarizing pulses, indicating a substantial impairment of the stability of the inactivated state compared with wild type (WT). The F1651A and N1662A mutants also had increased mean open times relative to WT, indicating a slowed rate of entry into the inactivated state. In addition to these effects on inactivation of open Na ؉ channels, mutants F1651A, L1660A, and N1662A also impaired fast inactivation of closed Na ؉ channels, as assessed from measurements of the maximum open probability of single channels. The peptide KIFMK mimics the IFM motif of the inactivation gate and provides a test of the effect of mutations on the hydrophobic interaction of this motif with the inactivation gate receptor. KIFMK restores fast inactivation of open channels to the F1651A/L1660A mutant but does not restore fast inactivation of closed F1651A/L1660A channels, suggesting that these residues interact with the IFM motif during inactivation of closed channels. Our results implicate F1651, L1660, and N1662 of the IVS4-S5 loop in inactivation of both closed and open Na ؉ channels and suggest that the IFM motif of the inactivation gate interacts with F1651 and/or L1660 in the IVS4-S5 loop during inactivation of closed channels.
In neurons and muscle cells, activation of voltage-gated Na ϩ channels leads to inward Na ϩ current which initiates the action potential (1). Within milliseconds after membrane depolarization, Na ϩ channels pass through a series of nonconducting closed states, enter an ion-conducting open state, and finally convert into a nonconducting inactivated state. Inactivated channels recover rapidly upon membrane repolarization and are thus available for reactivation by subsequent depolarizing stimuli. As inactivation exerts crucial control over Na ϩ channel activity, understanding the molecular basis of this process is an important step toward determining how Na ϩ channels function. Three subunits comprise the brain Na ϩ channel: ␣ of 260 kDa, ␤1 of 36 kDa, and ␤2 of 33 kDa (2). Expression of the ␣-subunit alone produces functional Na ϩ channels in Xenopus laevis oocytes (3,4). Coexpression of the ␤1and ␤2-subunits yields Na ϩ channels with kinetics and voltage dependence of inactivation more closely resembling those found in neurons (5,6). The ␣-subunit has four homologous domains (I-IV), each with six ␣-helical transmembrane domains (S1 through S6) (7,8). The S4 segments in each domain are thought to serve as voltage sensors, and the S5 and S6 segments and the short SS1-SS2 segments between them are thought to form the walls of the transmembrane pore (reviewed in Ref. 9).
Treatment of the intracellular surface of Na ϩ channels with proteolytic enzymes prevents inactivation, implicating cytoplasmic components of the Na ϩ channel in the inactivation process (10,11). Site-directed antibodies against the intracellular loop connecting domains III and IV (L III-IV ) block inactivation (12,13), and expression of the ␣-subunit as two proteins cleaved in L III-IV greatly slows inactivation (14). Mutations of a cluster of three hydrophobic residues within this loop-Ile 1488 , Phe 1489 , and Met 1490 (IFM)-prevents fast inactivation (15) primarily by destabilizing the inactivated state of the channel (16). Small peptides containing the IFM sequence are sufficient to restore fast inactivation to Na ϩ channels with mutations in L III-IV , leading to the hypothesis that the IFM motif binds within the pore of the Na ϩ channel and blocks it during inactivation (17). This model is supported by recent results showing that a cysteine residue substituted in the IFM motif becomes inaccessible to reaction with cysteine-specific reagents during the inactivation process (18).
This hypothesis for inactivation implies the presence of amino acid residues in the intracellular mouth of the pore of the Na ϩ channel that are involved in conformational change(s) which couple activation to inactivation and bind the IFM motif in the inactivated state. For example, mutations of Phe 1764 and Val 1774 at the intracellular end of segment IVS6 strongly destabilize the inactivated state and increase the rate of recovery of the channel from inactivation (19). Although these studies define a region of the Na ϩ channel required for conformational coupling and formation of a stable inactivation gate receptor, it is unlikely that these residues are fully responsible for these functions.
* This work was supported by National Institutes of Health Research Grant NS15751, postdoctoral fellowships from the National Institutes of Health (to J. C. M. and D. S. R.), and the W. M. Keck Foundation. 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 U.S.C. Section 1734 solely to indicate this fact.
The S4-S5 intracellular loops in each Na ϩ channel domain are good candidates for a role in the fast inactivation process. They are located near the intracellular mouth of the pore and are thus ideally situated to form part of the receptor for the pore-blocking inactivation gate. In addition, they are directly connected to the S4 segments which serve as the voltage sensors of Na ϩ channels and move outward under the influence of the electric field to initiate channel activation (14,20,21). The S4-S5 intracellular loop has been implicated in inactivation of Shaker K ϩ channels (22,23), but its role in the inactivation of Na ϩ channels is unknown. In these experiments, we have undertaken a systematic analysis of the role of residues in the S4-S5 loop of domain IV (IVS4-S5) of the rat brain type IIA Na ϩ channel using a combination of deletion/insertion and scanning mutagenesis, expression in Xenopus oocytes, and analysis by whole cell and single-channel recording methods. Our results reveal several amino acid residues in IVS4-S5 which are required for fast Na ϩ channel inactivation. A preliminary report of these results has been published in abstract form (24).

EXPERIMENTAL PROCEDURES
All experimental procedures were performed as described (19). In brief, mutants were created using oligo-directed mutagenesis in an M13mp18 construct containing a portion of the ␣-subunit of the rat brain IIa Na ϩ channel that included the IVS4-S5 intracellular loop. The full-length, mutant Na ϩ channel was produced by subcloning the mutant cassette into the full-length ␣-subunit in one of the following vectors: ZemRVSP6 (25), pVA2580 (8), or CDM8 (26). Each vector contained a bacteriophage RNA polymerase promoter and the rest of the RIIa Na ϩ channel gene. mRNA was produced in vitro from these mutant constructs and then coinjected with ␤1-subunit mRNA into X. laevis oocytes for channel expression (19). 50 nl of a 2:1 (w/w) mixture of rat brain Na ϩ channel ␤1 RNA and ␣-subunit RNA containing 5-50 ng/l ␣-subunit RNA were injected. Injected oocytes were maintained for 2-5 days prior to electrophysiological recording.
Two-microelectrode voltage clamp recording was used to record whole cell currents. Macropatch and single channel data were collected from cell-attached patches except for the excised macropatch experiments with the KIFMK peptide. Pulses were applied and data acquired using a personal computer-based data acquisition system (Basic-Fastlab, Indec Systems, Sunnyvale, CA). Single channel openings were detected using standard half-amplitude threshold analysis after filtering the data at 2 kHz and omitting overlapping events in multichannel patches (27). Open times, closed times, and first latencies were analyzed using PSTAT software (Axon Instruments, Foster City, CA).
Conductance (g) was calculated from peak current (I) during depolarizations as I/(V-V rev ), where V was the test pulse potential and V rev was the reversal potential. Normalized conductance-voltage and inactivation curves were fit with a Boltzmann relationship, where V1 ⁄2 is the voltage of half-maximal conductance, g1 ⁄2 , or the voltage of half-maximal inactivation, h1 ⁄2 , and A is the fraction of g or h that varies with voltage.

Deletion and Insertion Mutations in Segment IVS4-S5 Affect
Fast Inactivation-To determine whether the IVS4-S5 region of the ␣-subunit has a role in the fast inactivation of rat brain type IIA Na ϩ channels, we examined the effects of one deletion and one insertion mutation in this region (Fig. 1). WT 1 or mutant ␣-subunit RNA was co-expressed with ␤1-subunit RNA in X. laevis oocytes, and Na ϩ currents were analyzed via twomicroelectrode voltage clamp recording as described under "Experimental Procedures." Deletion ⌬4Na45, which deletes the Leu, Met, Met, and Ser residues in positions 1653-1656, and insertion I4Na45, which inserts four alanine residues between Met 1654 and Met 1655 , both disrupt fast inactivation. Upon depolarization, the WT Na ϩ channels rapidly activate and then inactivate within a few milliseconds (Fig. 1). In contrast, I4Na45 channels inactivate very slowly, and both these and ⌬4Na45 channels inactivate incompletely during the pulse, leaving a sustained current at the end of the 30-ms depolarization. These results implicate IVS4-S5 in the inactivation process.
Mutations of a Heptad Leucine Repeat Slow Inactivation-The S4-S5 loop contains four Leu and Ile residues spaced at seven-residue intervals (8)  Such heptad repeats are often involved in protein-protein interactions including leucine zippers (28,29). A similar repeat in the Shaker K ϩ channel S4-S5 loop plays a role in voltage-dependent gating (30). The double mutations L1639A/I1646A and L1653A/L1660A each disrupt half of this heptad leucine repeat. As observed for the deletion mutation ⌬4Na45, both L1639A/ I1646A and L1653A/L1660A channels inactivate incompletely during the pulse, and a substantial sustained current remains at the end of the 30-ms depolarization (Fig. 1). Evidently, the integrity of the leucine heptad motif is essential for normal fast inactivation.
Scanning Mutagenesis of Segment IVS4-S5-To examine the role of individual amino acid residues in the IVS4-S5 segment in fast inactivation, we substituted each amino acid in this region (Scheme 1) with alanine except for the charged residues in this region which were mutated to glutamine. Substitution of alanine residues is expected to alter the chemical characteristics of each residue without causing substantial conformational change because alanine is found with high frequency in both ␣-helices and ␤-sheets and has limited hydrophobic and hydrophilic character (31)(32)(33)(34). While most of these mutations had no effect on Na ϩ channel function, four of them caused slow and/or incomplete inactivation during depolarizing test pulses to Ϫ10 mV in two-microelectrode voltage clamp. L1639A had slowed inactivation, L1660A had incomplete inactivation, and F1651A had both dramatically slowed and incomplete inactivation ( Fig. 2A). No inactivation could be observed during 30-ms test pulses with the N1662A mutant ( Fig. 2A). The double mutant combination of F1651A and L1660A also com- 1 The abbreviation used is: WT, wild type.
FIG. 1. Effects of deletion, insertion, and heptad repeat mutations in IVS4-S5 on fast inactivation. Two-microelectrode voltage clamp recordings of currents due to expression of WT, ⌬4Na45, L1639A/ I1646A, I1653A/L1660A, and I4Na45 Na ϩ channel ␣-subunits in combination with ␤1-subunits in Xenopus oocytes. Currents were evoked by 30 ms pulses to Ϫ10 mV from a holding potential of Ϫ90 mV. Normalized currents are shown. Oocytes were injected with RNA encoding the appropriate ␣-subunit construct together with a 2-fold excess of RNA encoding the rat brain ␤1-subunit (5).
pletely prevented inactivation ( Fig. 2A). These results indicate that the transition between the open state and the inactivated state and/or the stability of the inactivated state is impaired by these mutations.
For the single mutants which retained significant inactivation, we also measured the voltage dependence of inactivation during 100-ms prepulses to a range of negative membrane potentials (Fig. 2B). This protocol measures primarily inactivation from closed states which are predominant at potentials more negative than Ϫ20 mV. For WT channels, 100-ms prepulses to potentials more positive than Ϫ30 mV cause nearly complete inactivation, leaving only 3 Ϯ 0.3% noninactivating current. The L1639A mutant also is nearly completely inactivated at positive voltages, although its voltage dependence of inactivation is shifted to more positive potentials (Fig. 2B). In contrast, the steady state inactivation curves for F1651A and L1660A approach non-zero asymptotes of 18 Ϯ 1% and 29 Ϯ 4%, respectively (Fig. 2B), indicating that inactivation is incomplete at positive potentials for these mutants as expected from their large sustained currents at the end of depolarizing test pulses ( Fig. 2A). These inactivation curves also reveal that the voltage dependence of inactivation is shifted to more positive membrane potentials for F1651A and L1660A (Fig. 2B). Consistent with the positively shifted inactivation curves, recovery from inactivation at negative membrane potentials was accelerated in these mutants (Fig. 2C). Thus, these mutations have important effects on inactivation from both open and closed states.
Although these selected mutations had large effects on Na ϩ channel inactivation, most of the mutations of amino acid residues in IVS4-S5 had little or no effect. The relative effects of all of the scanning mutations on the fraction of noninactivating Na ϩ current at the end of the test pulse are illustrated in Fig.  3A. Of 26 single residue mutations analyzed, only 3 caused a significant increase in the fraction of noninactivating current.
Evidently, the effects of the mutations at positions 1651, 1660, and 1662 on the fraction of noninactivating current are highly specific, suggesting that these residues have an important role in determining the stability of the inactivated state.
Although the mutations in IVS4-S5 caused large and specific effects on the rate and extent of inactivation, their effects on the voltage dependence of activation and inactivation were smaller and less specific (Fig. 3, B and C). Of the IVS4-S5 mutations discussed above, only N1662A caused a large (ϩ12 mV) shift in the potential for half-maximal activation. As illustrated in Fig. 2B, the L1639A, F1651A, and L1660A mutants each had small positive shifts in the half-maximal steady state inactivation curves of 8, 9, and 5 mV, respectively. Other mutations in the IVS4-S5 loop affected the voltage dependence of steady state activation or inactivation without producing sustained or slowly decaying Na ϩ currents (Fig. 3, B and C). The mutant with the most dramatic effect (S1656A) had a large negative shift of 14 mV in both the steady state activation and inactivation curves, but no effect on the rate or extent of inactivation during test pulses. The results show that the effects on the kinetics and extent of fast inactivation observed for L1639A, F1651A, L1660A, and N1662A are not secondary to changes in the voltage dependence of channel gating.
In contrast to the mutants that disrupted inactivation, mutation of 3 residues at the COOH terminus of this region enhanced closed state inactivation as revealed by large negative shifts in steady state inactivation curves (I1663A, G1664A, and L1666A; Fig. 3C). Mutant G1664A differed from the other two mutants in that activation (Fig. 3B) and inactivation ( Fig.  3C) are shifted to similar extents compared with WT and recovery from inactivation ( Fig. 2C) was indistinguishable from WT. Because inactivation is coupled to activation, the negative shifts in h1 ⁄2 for G1664A and S1656A are probably secondary to negative shifts in activation. In contrast, the shifts of inactivation curves for mutants I1663A and L1666A occurred without Currents were elicited by test pulses to 0 mV following 100-ms conditioning pulses to various potentials from a holding potential of Ϫ90 mV. Peak test pulse current from each experiment was plotted as a function of prepulse potential, normalized and fit with a Boltzmann equation as described under "Experimental Procedures." Mean V1 ⁄2 and k were determined for each mutant construct. The curves shown are plots of the Boltzmann equation using these mean values. C, recovery from inactivation at Ϫ90 mV. Recovery from inactivation was studied using two 30-ms long depolarizations separated by a recovery interval of variable duration. Fractional recovery for each recovery time was determined as (peak current during the second test pulse)/(peak current during the first test pulse). concomitant shifts in the voltage dependence of activation (Fig.  3B), suggesting a specific effect on inactivation. In addition, recovery from inactivation was greatly slowed in these two mutants (Fig. 2C). Thus, inactivation of closed Na ϩ channels is stabilized in mutants I1663A and L1666A in the carboxylterminal portion of the IVS4-S5 loop.
Analysis of Mutants by Macropatch Recording-To examine the inactivation properties of selected IVS4-S5 mutants in more detail, we recorded macroscopic currents from cell-attached macropatches. Recordings from macropatches give bet-ter time resolution and voltage control than two-microelectrode voltage clamp recordings. Fig. 4A shows typical records obtained during depolarizations to ϩ20 mV. At this voltage, activation is fast relative to inactivation, so the decay of currents reflects mainly the rate of channel inactivation from the open state (35,36). The mutants L1639A, F1651A, and L1660A all inactivate more slowly than WT, and all cause a detectable level of noninactivating Na ϩ current at the end of the 11-ms test pulse (Fig. 4A). The alterations of the Na ϩ currents conducted by mutant and WT channels are similar in the macropatch and two-microelectrode voltage clamp measurements, except that a small noninactivating current is observed in the macropatch recordings for L1639A. We could detect no N1662A current in macropatches, consistent with the relatively poor expression of this mutant observed in two-microelectrode voltage clamp recordings.
Comparison of the Na ϩ current records during test pulses to potentials from Ϫ40 to ϩ40 mV showed substantial impairment of inactivation of L1639A, F1651A, and L1660A at a wide range of potentials (Fig. 4, C-E) compared with WT (Fig. 4B). At potentials more positive than 0 mV at which most Na ϩ chan- FIG. 3. Summary of effects of IVS4-S5 Na ؉ channel point mutations on the steady state fraction of noninactivating current and the voltage dependence of activation and inactivation. A, fraction of current that fails to inactivate for each mutant construct. Currents were elicited by pulses to Ϫ5 mV as described in the legend to Fig. 1. The fraction of noninactivating current was determined as the current 15 ms after the beginning of the pulse divided by the peak inward current. Mean Ϯ S.E. is plotted for each IVS4-S5 mutant. Data are from three to eight experiments for each mutant. B, shift of activation curves for mutants relative to WT. Currents were elicited by 30-ms long, two-microelectrode voltage clamp pulses to a range of test potentials from a holding potential of Ϫ90 mV. Peak inward current was measured, converted to conductance and fit with a Boltzmann relationship as described under "Experimental Procedures." The mean WT g1 ⁄2 value (Ϫ18.6 Ϯ 4.6 mV) was subtracted from the mean g1 ⁄2 value for each mutant, and the difference Ϯ S.E. was plotted. C, shift of steady state inactivation curves for mutants relative to WT. Inactivation curves were recorded and fit with a Boltzmann relationship as described for Fig. 2B. The average WT h1 ⁄2 value (Ϫ47.4 Ϯ 4.2 mV) was subtracted from the mean h1 ⁄2 value for each mutant, and the difference Ϯ S.E. was plotted.
FIG. 4. Voltage dependence of inactivated state stability. A, normalized currents evoked by depolarizations to ϩ20 mV from a holding potential of Ϫ140 mV in cell-attached macropatches on X. laevis oocytes expressing WT, L1639A, F1651A, and L1660A mutants. Each trace is an average of 10 -15 sweeps. B-E, normalized currents from cell-attached macropatches on oocytes elicited by 11-ms pulses to Ϫ40, Ϫ20, 0, 20, and 40 mV from cells expressing WT, L1639A, F1651A, and L1660A, respectively. F, the fraction of noninactivating current in cellattached macropatches on oocytes expressing F1651A and L1660A was determined at each potential and plotted as a function of test pulse voltage. This fraction was defined as the ratio of sustained current at 10 ms after the beginning of the depolarizing pulse to peak current. The data are from three experiments each. nels are fully activated, the impairment of inactivation increased with depolarization, as illustrated by the increased fraction of noninactivating current at the end of the test pulse (Fig. 4F). These results imply that the stability of the inactivated state is progressively impaired with increasing depolarization in these mutant channels.
Alterations in Gating of Single Mutant Na ϩ Channels-Each of the IVS4-S5 mutants having impaired inactivation was analyzed by single-channel recording methods as outlined under "Experimental Procedures." Na ϩ channel gating can be described simply by the reaction pathway illustrated in Scheme 2 (16,35,37,38). Upon depolarization, Na ϩ channels undergo voltage-dependent transitions through multiple closed states (states C n through C 0 ), open (state O), and then inactivate (state I). The inactivated state can be reached from the open state and also from one or more closed states. For WT Na ϩ channels, inactivation is essentially irreversible at depolarized potentials, indicating that the O-I transition strongly favors inactivation at those potentials.
In response to 40-ms depolarizations to Ϫ20 mV, WT Na ϩ channels generally opened once or twice at the beginning of the depolarization and then inactivated and did not reopen (Fig.  5A). Most depolarizations of mutant L1639A channels showed similar behavior to WT (Fig. 5B, top trace). However, occasionally one channel in a multichannel patch generated sustained bursts of openings and had delayed inactivation (Fig. 5B, bottom trace). In contrast, nearly all F1651A channels had long openings and an increased frequency of reopenings relative to WT (Fig. 5C). L1660A channels had short openings like WT, but had a high frequency of reopenings throughout the depolarizing pulse (Fig. 5D). The ensemble currents for these mutants recreated the slow decay and significant level of noninactivating current observed in macropatch recordings (Fig. 5, B, C, and D, bottom traces).
Patches containing the N1662A mutant channel were rare due to its low expression at the whole cell level; therefore, we used large patches suitable for macropatch recording to obtain single channel data. The openings of the N1662A mutant channels are long, and reopenings occur at a high frequency for this mutant relative to WT (Fig. 5E). The ensemble average currents show no appreciable inactivation during the test pulses, as observed in two-microelectrode voltage clamp recordings of this mutant channel.
Total channel activity can be estimated as the integrated probability (P o ) of opening for the duration of the 40-ms depolarizing sweeps. The value of P o for the WT channels is low, Ͻ0.5%. For each of the three IVS4-S5 mutants with substantial noninactivating current, the P o was higher than WT: F1651A, 11.2%; L1660A, 8.4%; and N1662A, 57.0%. Thus, this quantitative measure of increased channel activity also shows the substantial impairment of inactivation in these mutants.
No significant effects on the single channel conductance were observed for any of the IVS4-S5 mutants. The WT single channel current at Ϫ20 mV was 1.05 Ϯ 0.005 pA. The single channel currents for the F1651A, L1660A, and N1662A mutants were: 1.1 Ϯ 0.06 pA, 1.03 Ϯ 0.04 pA, and 1.06 Ϯ 0.06 pA, respectively.
Decreased Inactivation from Closed States in Single Channel Recordings-Many depolarizations of single Na ϩ channels produce "null" sweeps containing no single channel activity (35)(36)(37). This behavior results from direct inactivation of the single channel from a closed state (Scheme 2) (19,35). The number of nulls decreased for each of the three IVS4-S5 mutants which caused substantial noninactivating current. For WT channels, approximately 35% of the recorded sweeps were nulls. The number of nulls observed for the F1651A, L1660A, and N1662A mutants were 3.6, 1.0, and less than 1.0%, respectively. The decrease in the number of nulls for these mutants correlates with their increased frequency of reopenings during a depolarizing pulse (see below). Thus, the loss of nulls is caused by a combination of slow entry into the inactivated state from closed states (decreased rate constant e in Scheme 2) and frequent return from the destabilized inactivated state (increased rate constants f and h in Scheme 2).
Latency to First Activation-The latency before the first single channel opening following depolarization reflects the dwell time in the multiple closed states in the activation pathway prior to opening (C n through C 0 , Scheme 2) (35). Fig. 6 shows the cumulative first latency distributions for WT, F1651A, L1660A, and N1662A channels at a test potential of Ϫ20 mV, corrected for the number of channels in the patch (39). We were unable to obtain patches with small numbers of L1639A channels and, thus, could not obtain an adequate estimate of first latency or closed times for this mutant. For WT, single channels open rapidly, but the maximum probability of opening is only about 0.65 because 35% of the depolarizations fail to elicit channel opening, resulting in null sweeps. As noted above, these nulls represent transitions directly from the closed states in the activation pathway to the inactivated state during depolarization (35, 36) (Scheme 2).
Most F1651A channels open rapidly like WT, but the open probability reached after this early rapid phase is nearly 0.85 (Fig. 6). The high open probability suggests that these channels may have slowed inactivation from closed states resulting in a higher proportion of the channels being available for opening. A small slow phase of activation is also apparent and is consistent with a low frequency of channel opening after first inactivating from the closed state in agreement with the observation of increased channel reopenings for this mutant.
The cumulative first latency distribution for mutant L1660A rises rapidly to an open probability of 0.6 similar to WT channels and then continues to increase slowly to an open probability approaching 1.0 (Fig. 6). The fast component represents channels that open normally, indicating that channel activation is not affected by this mutation. The second, delayed component represents channels that inactivate normally from closed states, but unlike WT, leave the destabilized inactivated state and then open.
For mutant N1662A, the first latency distribution rises rapidly to 1.0, but at a slightly slower rate than for WT. Thus, N1662A channels open rapidly like WT but rarely inactivate irreversibly from the closed state. As was suggested for F1651A, these characteristics are likely to result from a combination of slow entry into the inactivated state and frequent reopening if they do enter the destabilized inactivated state. The slightly slower opening rate agrees well with the observed positive shift of the voltage dependence of N1662A macroscopic activation. The first latency curves for F1651A, L1660A, and N1662A all contain a rapid opening rate similar to WT channels and open to an equal or greater open probability than WT channels, suggesting that these mutations have little effect on the activation process and predominantly affect the inactivation process.
Mean Open Times-At depolarized potentials, the mean open time depends primarily on the rate of entry of single Na ϩ channels into the inactivated state from the open state, because the inactivation rate is much faster than the closing rate (g Ͼ Ͼ d in Scheme 2) and the channel leaves the open state primarily via entry into the inactivated state rather than via return to the closed state (11,16,35). Consequently, if mutant channels enter the inactivated state more slowly from the open state (reduced g in Scheme 2), they will have an increased mean open time. Open time data for WT, L1639A, and L1660A could be fit with a single exponential (Fig. 7). The mean time constants for these Na ϩ channels were short, 0.33 Ϯ 0.27 ms (n ϭ 5), 0.39 Ϯ 0.09 ms (n ϭ 3), and 0.35 Ϯ 0.03 ms (n ϭ 4) (Fig. 7, A and C). For each of these channels, therefore, the rate of entry into the inactivated state from the open state is similar to WT. In contrast, the open time histogram for F1651A is fit with a single, longer exponential time constant. The mean time constant was 0.90 Ϯ 0.23 ms (n ϭ 3), significantly longer than that for the WT channel (Fig. 7B). Thus, this mutant has a decreased rate of entry into the inactivated state relative to WT.
The distribution of open times of N1662A could be best fit with 2 exponentials. The data in Fig. 7D  Closed Times-Analysis of closed time distributions for single channels provides information about the closed states which they enter between openings. For WT channels the inactivation rate is rapid, and return to closed states (C n through C 0 , Scheme 2) after the first opening is rare. For mutants with slowed entry into the inactivated state, return to the last closed state becomes more probable and these closures are expected to add additional components to the distribution of closed times. for L1660A, and 1.4 ms for N1662A) (Fig. 8, A-C). In each case, this longer closed time is longer than the first latency, and therefore cannot represent return to a closed state in the activation pathway (40). Thus, the long closed times must represent the dwell times of these channels in the destabilized inactivated state. These closed times provide a direct measure of the increased rate of return from the inactivated state for these mutants. The long closed times due to closure to the inactivated state for these three mutants are voltage-dependent and become shorter at more depolarized voltages (Fig. 8D). Evidently, the inactivated state is less stable at more depolarized voltages, as was also suggested by the increased noninactivating Na ϩ currents at more depolarized potentials (Fig. 4F).

Effects of the Inactivation Gate Peptide KIFMK on Decay of Na ϩ Currents Through the F1651A/L1660A Double Mutant-
The intracellular loop between domains III and IV forms the Na ϩ channel inactivation gate (12)(13)(14). Mutagenesis studies have identified four residues that are critical components of the gate, Ile 1488 , Phe 1489 , and Met 1490 (IFM), and Thr 1491 . Mutation of Phe 1489 to Gln is sufficient to nearly completely block fast inactivation (15). The IFM residues are thought to bind to a hydrophobic site in or near the mouth of the Na ϩ channel pore (16,17). One possible interpretation of the effects of the F1651A and L1660A mutations is that the native residues at these positions form the hydrophobic binding site for the IFM motif so that mutations of these amino acids disrupt the binding site.
To test this hypothesis, we examined the effects of application of a peptide, acetyl-KIFMK-amide (KIFMK), to the intracellular surface of excised membrane patches from oocytes expressing mutant Na ϩ channels. Mutations of an inactivation gate receptor would be expected to affect block by this peptide.
Application of this peptide to the intracellular surface of patches containing Na ϩ channels with a mutation in the critical Phe of the intrinsic IFM motif (F1489Q) reduces the peak Na ϩ current and produces current decay during the pulse that resembles fast inactivation from closed and open states, respectively (19,41) (Fig. 9A). The time course of the Na ϩ current in the presence of KIFMK can be fit to a model in which inactivation by binding KIFMK occurs in parallel with inactivation by the intrinsic inactivation gate as illustrated in Scheme 3.
As for F1489Q, Na ϩ currents in excised patches containing the double mutant F1651A/L1660A have impaired inactivation in the absence of peptide, and KIFMK increases both the rate and extent of inactivation (Fig. 9B). However, the same concentration of peptide had different effects on this mutant compared with F1489Q. There was little reduction of the peak current (Fig. 9B). Because the reduction in peak current reflects inactivation of closed Na ϩ channels, this effect of the F1651A/L1660A mutation suggests that binding of KIFMK to the closed states of this mutant channel is substantially impaired. Fit of the data to a model based on Scheme 3 revealed a 5.7-fold reduction in the maximum value of rate constant i for KIFMK-mediated inactivation from the closed state. This impairment of binding of the IFM motif may contribute to the loss of closed state inactivation of the F1651A and L1660A mutants revealed by the increase in their maximum probability of open-  9. Effect of KIFMK peptide on Na ؉ currents in excised, inside-out macropatches from oocytes expressing the F1489Q and F1651A/L1660A mutants. Currents elicited by depolarizing pulses to Ϫ20 mV from a holding potential of Ϫ140 mV in the absence and presence of 50 M KIFMK peptide. The traces shown are representative of 2 patches for each mutant construct. The smooth lines are fits of Scheme 3 to the data as described (16). For peptide block of F1489Q, the channel makes the C to I-KIFMK transition (Scheme 3) 28% of the time that it arrives in state C 0 , whereas for F1651A/L1660A it makes that transition a maximum of 5% of the time. The rate constants for the fit to F1489Q alone were: a, 12,000 s Ϫ1 ; b, 100; c, 10 ing (Fig. 6). In contrast to the difference in the reduction of peak Na ϩ currents between the inactivation gate mutant F1489Q and mutant F1651A/L1660A, there is not a marked difference in the effects of KIFMK on the decay of the Na ϩ current during the test pulse (Fig. 9). For both mutants, the peptide increases the rate of current decay markedly and decreases the level of current remaining at the end of the test pulse. Fits of these results to a model based on Scheme 3 showed that the rate constants k and l, governing the transition from the open state to the inactivated state, were similar for both mutants. Therefore, the double mutation F1651A/ L1660A has a specific effect on KIFMK binding to closed channels and little or no effect on KIFMK binding to open Na ϩ channels.

DISCUSSION
Mutations of F1651, L1660, and N1662 in the IVS4-S5 Loop Impair Fast Inactivation of Open Na ϩ Channels-The mutations F1651A, L1660A, and N1662A all slow the rate of decay of Na ϩ currents. Analysis of single channel currents shows that all three mutations impair the stability of the inactivated state, equivalent to increasing rate constant, h, for exit from the inactivated state in Scheme 2. In addition, F1651A and N1662A also slow entry into the inactivated state from the open state, equivalent to decreasing rate constant g in Scheme 2. Thus, these amino acid residues are necessary for rapid transition from open to inactivated Na ϩ channels and for the irreversibility of that transition at depolarized membrane potentials.
Mutations of F1651, L1660, and N1662 in the IVS4-S5 Loop Impair Fast Inactivation of Closed Na ϩ Channels-Two lines of evidence support the conclusion that these mutations in the IVS4-S5 loop also impair fast inactivation from closed states. First, these mutations all reduce the frequency of null channel traces in single channel analysis, consistent with slowing the entry of channels from the closed state into the inactivated state (rate constant e, Scheme 2) and increasing the frequency of opening from the inactivated state (rate constants f and h). Second, F1651A and L1660A both shift the voltage dependence of steady state inactivation to more positive membrane potentials, consistent with an increased energy requirement for the transition from the closed states to the inactivated state. Evidently, these amino acid residues in the IVS4-S5 loop are involved in both of the dual pathways to the inactivated state from the closed and open states as illustrated in Scheme 2.
Molecular Basis for Impairment of Fast Inactivation by Mutations in the IVS4-S5 Loop-Our results with the KIFMK peptide give additional insight into the molecular basis for the effects of these mutations on inactivation from open and closed states. Previous results have shown that the IFM motif enters into a hydrophobic interaction with a putative receptor in or near the intracellular mouth of the pore during the inactivation process (16). Disruption of this interaction by substitution of hydrophilic amino acid residues for Phe 1489 in the inactivation gate greatly impairs the stability of the inactivated state and slows entry into the inactivated state. Therefore, it is expected that mutation of hydrophobic amino acid residues in the inactivation gate receptor to less hydrophobic ones would also slow entry into and destabilize the inactivated state by impairing the hydrophobic interaction with the IFM motif. The phenotypes of mutants F1651A, L1660A, and N1662A are consistent with this type of mechanism. They each slow entry into and/or reduce the stability of the inactivated state.
This mechanism can be tested specifically with the KIFMK peptide since it is thought to bind to the inactivation gate receptor but it is not coupled physically to the gating processes of the channel protein. Surprisingly, our results show that amino acid residues Phe 1651 and/or Leu 1660 are essential for rapid interaction of the KIFMK peptide with closed channels but not with open channels. We cannot assess the effects of these two mutations individually because they do not prevent inactivation completely enough when present singly to allow clear effects of the KIFMK peptide to be observed. Nevertheless, the finding that the region of the IVS4-S5 loop defined by these two residues may interact with the IFM motif during inactivation of closed Na ϩ channels is an important advance. In our previous studies of candidate amino acid residues for interaction with the IFM motif, we showed that those mutants were blocked normally by KIFMK, implying that those residues do not interact directly with the IFM motif (19). Thus, Phe 1651 and Leu 1660 are the first residues implicated in the receptor for the IFM motif using this criterion.
The gating diagrams of Schemes 2 and 3 include only a single inactivated state, suggesting that the IFM motif should have the same molecular interactions in inactivation of closed or open Na ϩ channels. However, this gating scheme is a simplification, and it is clear that a complete scheme for inactivation must include formation of multiple closed/inactivated states as well as an open/inactivated state (42). Most likely, there is an inactivated state associated with each of the C n through C 0 closed states, as presented in previous complete gating models for Na ϩ channels (42,43). In the context of these more complete gating models, our results indicate that rapid entry into the inactivated state C 0 I is likely to require hydrophobic interactions of the IFM motif of the inactivation gate with Phe 1651 or Leu 1660 in IVS4-S5. In contrast, these interactions are less important for rapid entry into the inactivated state I from the open state O. Thus, the IFM motif may form a series of interactions with different amino acid residues in a complex inactivation gate receptor region, depending on the activation gating status of the intracellular end of the pore of the Na ϩ channel.
Additional Effects of Mutations at the Ends of the IVS4-S5 Loop-The effects of the L1639A mutation, located at the NH 2terminal boundary of IVS4-S5, are different from the other mutations located in the central segment of this intracellular loop. At the macroscopic level, the decay of the Na ϩ current through the L1639A mutant channel is slowed, suggesting an effect on inactivation. Although the overall behavior of single L1639A channels is not qualitatively different from WT, the channels undergo occasional transient failures of inactivation that are observed as prolonged openings or bursts of openings. These bursts of activity are responsible for the sustained current observed in macropatch recordings. These high open probability events ended before the end of the test pulses, explaining the minimal increase in the noninactivating current at the end of depolarizations. The L1639A mutation may cause a conformational change in IVS4-S5 that slows the development of inactivation, perhaps by slowing formation of the inactivation gate receptor. The position of this residue at the inner end of segment IVS4 suggests that it may participate in coupling of voltage sensor movements to the inactivation process and that this coupling may occasionally fail in the mutant channel.
Two mutations at the COOH-terminal end of IVS4-S5, I1663A and L1666A, increase the stability of the inactivated state. This result suggests that the natural Ile and Leu residues at these positions actually destabilize the inactivated state. Thus, the WT residues in IVS4-S5 contribute both stabilizing and destabilizing interactions within the overall energetics of inactivation. The sum of these interactions will set the voltage dependence of inactivation, the rate of inactivation during depolarizations, and the rate of recovery from inactivation between depolarizations.
Comparison with Results of IVS4-S5 Mutants from Other Na ϩ Channels-Recent studies of mutations made in the IVS4-S5 loops of human heart (hH1), rat skeletal muscle (rSkM1), and human skeletal muscle (hSkM1) Na ϩ channels have yielded some similar and some contrasting results to those presented here for the rat brain Na ϩ channel (rIIA). Similar to the data presented for the rat brain F1651A mutant, mutations in the corresponding residue in human (F1473S) and rat (F1466C) SkM1 also produce substantial slowing of Na ϩ current decay for these channels (44,45). Apparently, a role for this Phe residue in inactivation is conserved in different Na ϩ channels. In contrast, mutation of Leu 1660 to Ala in type IIA channels produced dramatic effects on inactivation, but the similar mutations in human SkM1 (L1482A) or rat SkM1 (L1475C) produced little or no change in steady state inactivation or recovery from inactivation (44,45). A double mutation of M1651/M1652 of the human heart Na ϩ channel to either QQ or AA caused substantial slowing of Na ϩ current decay, a positive shift of steady state inactivation, and an increased rate of recovery from inactivation. In contrast, the single mutations of each Met in rIIA (M1654A and M1655A) or the rSkm1 (M1469C, M1470C) had little or no effect on inactivation (44 -46). Apparently, species or tissue differences in the role of specific amino acid residues in inactivation exist, but the importance of Phe 1651 and the IVS4-S5 loop in the inactivation process is conserved across species and tissue boundaries.
Possible Function of IVS4-S5 in Inactivation-The characteristics of the F1651A, L1660A, and N1662A mutants show that the IVS4-S5 region is important for the inactivation of brain Na ϩ channels. What is the exact role of these residues? This region of the Shaker K ϩ channel is also important in inactivation and has been suggested to form the receptor for the NH 2 -terminal inactivation particle of K ϩ channels (22,23). In agreement with this suggestion, the IVS4-S5 region is strategically located near the intracellular mouth of the pore and is proximal to a voltage sensor of the channel. The phenotype of the F1651A, L1660A, and N1662A mutants makes them potentially good candidates to form part of the receptor for the Na ϩ channel inactivation gate, since each mutation decreases the stability of the inactivated state by increasing the rate of reversal of inactivation. The effect of the L1639, F1651A, and N1662A mutations to slow entry into the inactivated state may indicate that these residues participate in a conformational change which is required for formation of an effective inactivation gate receptor. Results with the KIFMK peptide implicate Phe 1651 or Leu 1660 in interaction with the IFM motif of the inactivation gate during inactivation of closed Na ϩ channels but these residues are not as important for fast inactivation of open Na ϩ channels. Thus, these residues may form part of a multifaceted inactivation gate receptor region which interacts with the inactivation gate differently depending on the functional state of the channel gating machinery. Further definition of these state-dependent events may elucidate molecular interactions important for gating transitions among closed, open, and inactivated states of Na ϩ channels.