主题：Sodium

【关键词】  Inhibition

    ProTx-II, an inhibitory cysteine knot toxin from the tarantula Thrixopelma pruriens, inhibits voltage-gated sodium channels. Using the cut-open oocyte preparation for electrophysiological recording, we show here that ProTx-II impedes movement of the gating charges of the sodium channel voltage sensors and reduces maximum activation of sodium conductance. At a concentration of 1 µM, the toxin inhibits 65.3 ± 4.1% of the sodium conductance and 24.6 ± 6.8% of the gating current of brain Nav1.2a channels, with a specific effect on rapidly moving gating charge. Strong positive prepulses can reverse the inhibitory effect of ProTx-II, indicating voltage-dependent dissociation of the toxin. Voltage-dependent reversal of the ProTx-II effect is more rapid for cardiac Nav1.5 channels, suggesting subtype-specific action of this toxin. Voltage-dependent binding and block of gating current are hallmarks of gating modifier toxins, which act by binding to the extracellular end of the S4 voltage sensors of ion channels. The mutation L833C in the S3-S4 linker in domain II reduces affinity for ProTx-II, and mutation of the outermost two gating-charge-carrying arginine residues in the IIS4 voltage sensor to glutamine abolishes voltage-dependent reversal of toxin action and toxin block of gating current. Our results support a voltage-sensor-trapping model for ProTx-II action in which the bound toxin impedes the normal outward gating movement of the IIS4 transmembrane segment, traps the domain II voltage sensor module in its resting state, and thereby inhibits channel activation.

    Voltage-gated sodium channels are responsible for the increase in sodium permeability that initiates action potentials in electrically excitable cells and are the molecular targets for several groups of neurotoxins that bind to different receptor sites and alter voltage-dependent activation, conductance, and inactivation (Catterall, 1980; Cestèle and Catterall, 2000). Sodium channels are composed of one pore-forming  subunit of approximately 2000 amino acid residues associated with one or two smaller auxiliary subunits, β1to β4 (Catterall, 2000). The  subunit consists of four homologous domains (I-IV), each containing six transmembrane segments (S1-S6), and a reentrant pore loop (P) between S5 and S6 (Catterall, 2000). The S4 transmembrane segments are positively charged and serve as voltage sensors to initiate channel activation (Armstrong, 1981; Catterall, 1986; Stühmer et al., 1989; Yang and Horn, 1995; Chanda and Bezanilla, 2002). The "sliding helix" (Catterall, 1986) or "helical screw" (Guy and Seetharamulu, 1986) models for voltage sensing propose that the S4 segments, which have positively charged amino acids at intervals of three residues, transport gating charge outward to activate sodium channels in response to depolarization by moving along a spiral pathway through the protein structure (Yarov-Yarovoy et al., 2006). This transmembrane movement of the S4 gating segments is a molecular target for neurotoxin action via voltage-sensor trapping (Rogers et al., 1996; Cestèle et al., 1998, 2001, 2006).

    Scorpion venoms contain two groups of polypeptide toxins that alter sodium channel gating. The -scorpion toxins, as well as sea anemone toxins and some spider toxins, bind to neurotoxin receptor site 3 and slow or block inactivation (Catterall, 1977, 1979; Catterall and Beress, 1978; Nicholson et al., 1994). Amino acid residues that contribute to neurotoxin receptor site 3 are localized in the S3-S4 linker in domain IV (Rogers et al., 1996; Benzinger et al., 1998) and in the large extracellular loops in domains I and IV (Tejedor and Catterall, 1988; Thomsen and Catterall, 1989). Binding of toxins to IVS3-S4 is thought to slow inactivation by preventing the normal outward movement of the IVS4 transmembrane segment during channel gating (Rogers et al., 1996; Sheets et al., 1999). In contrast to these toxins that inhibit inactivation gating, β-scorpion toxins bind to neurotoxin receptor site 4 on sodium channels and enhance activation by shifting its voltage dependence to more negative potentials (Cahalan, 1975; Jover et al., 1980; Jaimovich et al., 1982). Our previous results implicate the extracellular loops S1-S2 and S3-S4 in domain II in formation of neurotoxin receptor site 4 (Cestèle et al., 1998). Moreover, a voltage sensor-trapping mechanism, in which the bound β-scorpion toxin holds the IIS4 segment in its outward, activated position was proposed to account for enhancement of activation (Cestèle et al., 1998, 2001, 2006). Voltage-sensor trapping by both - and β-scorpion toxins inhibits gating currents generated by the transmembrane movement of the S4 segments, providing a mechanistic signature for voltage-sensor trapping (Nonner, 1979; Meves et al., 1982).

    The polypeptide toxins from the tarantula Thrixopelma pruriens (Protoxins) are members of the inhibitory cysteine-knot family of protein toxins, consisting of 30 to 35 amino acid residues with three disulfide bridges (Norton and Pallaghy, 1998; Middleton et al., 2002; Priest et al., 2007). This family includes toxins that inhibit activation of sodium channels, such as ProTx-II, and potassium channels, such as hanatoxin, by interfering with the normal function of the voltage sensors (Swartz and MacKinnon, 1997; Middleton et al., 2002). In this study, we have probed the mechanism of ProTx-II action with combined measurements of its effects on sodium currents and gating currents conducted by NaV1.2 and NaV1.5 channels. Our results show that ProTx-II inhibits gating currents conducted during voltage-dependent activation of sodium channels. The inhibitory effects of the toxin can be reversed by strong, long-lasting positive voltage pulses, which drive the voltage sensor into its activated conformation. Voltage-dependent reversal of ProTx-II effects was more rapid for NaV1.5 channels, the primary sodium channel in the heart. Mutations in the S3-S4 linker in domain II reduce toxin affinity, and mutations in the IIS4 voltage sensor prevent voltage-dependent reversal of toxin action. Our results indicate that ProTx-II impedes the normal gating function of the IIS4 voltage sensor by a voltage-sensor trapping mechanism and can be knocked off its receptor site on the extracellular side of the voltage sensor by voltage-driven movement of the IIS4 voltage sensor into its activated conformation.

    Materials. ProTx-II from the venom of the tarantula T. pruriens was synthesized chemically (Middleton et al., 2002). Restriction endonucleases and other molecular biology reagents were purchased from New England Biolabs (Ipswich, MA) and Roche Applied Science (Indianapolis, IN). Bovine serum albumin was from Sigma (St. Louis, MO). pCDM8 vector and the MC1061 Escherichia coli bacterial strain were from Invitrogen (Carlsbad, CA). cDNAs encoding rat Nav1.2a -subunit (Auld et al., 1990) and rat Nav1.5 -subunit subcloned into pCDM8 vector (Rogers et al., 1996) were used for expression in Xenopus laevis oocytes. Point mutations L833C, L833N, E837Q, L838C, L840C, N842A, and G845N in the DIIS3-S4 linker as well as R850Q, R853Q and a double mutation RR850,853QQ in DIIS4 of Nav1.2a channel were produced in our lab previously (Cestèle et al., 1998, 2001; Sokolov et al., 2005). X. laevis frogs were purchased from Nasco (Fort Atkinson, WI).

    Expression in X. laevis Oocytes. pCDM8 plasmids encoding rat Nav1.5, Nav1.2a, and mutant Nav1.2a sodium channel subunits were linearized with XbaI or ClaI, respectively, and plasmids encoding β1 subunits were linearized with HindIII. Transcription was performed with T7 RNA polymerase (Ambion Inc., Austin, TX). Isolation, preparation, and maintenance of X. laevis oocytes were carried out as described previously (McPhee et al., 1995). Healthy stage V-VI oocytes selected manually were pressure-injected with 50 nl of a solution containing a 1:1 molar ratio of /β1 subunit RNA. Electrophysiological recordings were carried out 4 to 7 days after injection.

    Cut-Open Oocyte Voltage Clamp. Cut-open oocyte voltage-clamp experiments were performed as described by Stefani and Bezanilla (1998), except that access to the cytoplasm was obtained by rupturing the vegetal pole membrane of the oocyte. The oocyte capacitative transients were partially compensated with the voltage clamp amplifier (CA-1; Dagan Corporation). Online P/-4 leak subtraction was used. Microelectrodes were pulled from borosilicate glass capillary tubes 1.5 mm OD (A-M Systems, Carlsborg, WA) and had resistances of 250 to 350 k when filled with 3 M KCl. Extracellular solution contained 120 mM sodium methanesulfonate (MeSO3), 10 mM HEPES, 1.8 mM Ca-MeSO3, and 1% bovine serum albumin, pH 7.4. Internal solution consisted of 110 mM K-MeSO3, 10 mM Na-MeSO3, 10 mM EGTA, and 10 mM HEPES, pH 7.4. ProTx-II was prepared as a 100 µM stock in 120 mM Na-MeSO3, 10mM HEPES, and 0.2 mg/ml bovine serum albumin, aliquoted at 5 µM, and stored at -20°C. Aliquots containing toxin were thawed immediately before experiments and diluted in extracellular solution.

    The starting solution volume in the upper boat of the recording chamber (Dagan Corporation, Minneapolis, MN) was typically 150 µl. Volumes of 37.5 or 50 µl of solution containing 5 µM ProTx-II were typically added to the recording chamber, resulting in final concentrations of 1 to 1.25 µM, except where other concentrations are noted in the figure legends, and the cells were allowed to equilibrate for 3 to 5 min before recording. After each experiment, the recording and guard chambers were rigorously washed to remove all traces of ProTx-II in a multistep washing procedure using 75% EtOH and 95% MeOH.

    All experiments were performed at room temperature. Currents were filtered at 5 kHz with a low-pass Bessel filter, and then digitized at 20 kHz. Voltage commands were generated using Pulse 8.5 software (HEKA, Lambrecht/Pfalz, Germany) and ITC18 analog-to-digital interface (Instrutech, Port Washington, NY).

    Data Analysis. Data were analyzed with Igor Pro 4.0 WaveMetrics, (Lake Oswego, OR). Voltage-clamp protocols are described in the figure legends. Voltage dependences of activation and inactivation were fit by Boltzmann functions of the form Gmax/{1 + exp[(V - Va)/k]}, where Gmax is the maximum conductance, Va is the half-activation or inactivation potential, and k is a slope factor. Rates of inactivation in Fig. 1C were determined by single exponential fitting of current traces. Pooled data are reported as means ± S.E. Statistical comparisons were performed using Student's t test, with p < 0.05 as the criterion for significance.

    Fig. 1. Effects of ProTx-II on properties of ionic currents through Nav1.2a  subunits coexpressed with β1 subunits in X. laevis oocytes. A and C, currents were elicited using 10-ms depolarizations to potentials from -60 mV to +75 mV in 5-mV steps from a holding potential of -100 mV. A, families of ionic currents before (top) and 5 min after bath application of 1 µM ProTx-II (bottom). For potentials between -60 to -20 mV, traces are shown every 5 mV. At more positive potentials, the interval is 10 mV. B, concentration-response relationship for ProTx-II block of sodium current at -10 mV. The data were fit with a one-site binding isotherm with an EC50 of 540 nM (n  3 for each concentration). C, normalized conductance-voltage relationship for Nav1.2a/β1 channel before (, V = -26.7 mV, k = 5.9 mV, n = 15) and after application of 1 µM ProTx-II (, V = -25.7 mV, k = 6.9 mV, normalized Gmax = 0.36 ± 0.03, n = 12). Error bars are smaller than the symbols. D, voltage dependence of the time constant of current decay () during the 10-ms depolarizations to the indicated potentials for Nav1.2a/β1 in control (, n = 14) and with 1 µM ProTx-II (, n = 12). E, kinetics of recovery from fast inactivation induced by a 20-ms conditioning pulse to -10 mV in control (,  = 2.2 ± 0.3 ms, n = 5) and in the presence of 1 µM ProTx-II (,  = 2.5 ± 0.3 ms, n = 7). A 5-ms test pulse to -10 mV was applied after a recovery interval of variable duration at the holding potential of -100 mV. Fractional recovery was measured as the peak inward current during this test pulse normalized to peak inward current during the conditioning pulse. F, voltage dependence of steady-state inactivation in control (, V = -48.8 ± 0.3 mV, n = 6) and with 1 µM ProTx-II (, V = -47.4 ± 0.4 mV, n = 9). Conditioning pulses of 200 ms to the indicated potentials were followed by 5-ms test pulses to -10 mV.

    Inhibition of Sodium Currents by ProTx-II. We expressed brain Nav1.2a channels in X. laevis oocytes together with auxiliary β1 subunits and studied effects of ProTx-II on ionic currents and gating currents with the cut-open voltage clamp technique (Stefani and Bezanilla, 1998). Ionic currents were recorded in 120 mM Na+ control solution before and 5 min after adding 1 µM ProTx-II to the recording chamber (Fig. 1). ProTx-II (1 µM) inhibited 65.3 ± 4.1% of ionic current in Nav1.2 channels at a test pulse potential of -10 mV (Fig. 1A). Previous results (Middleton et al., 2002) showed that the effect of ProTx-II as a function of concentration can be fit by a single binding site model. Plotting the effect of ProTx-II on sodium current versus concentration yielded an inhibition curve that was fit by a single binding site model with an EC50 of 540 nM and inhibition of essentially 100% of the sodium current at saturation (Fig. 1B). The EC50 of ProTx-II for sodium channels expressed in X. laevis oocytes is significantly higher than previously observed in mammalian cells (Middleton et al., 2002).

    Fig. 2. Inhibition gating currents of Nav1.2a channels by ProTx-II. A, representative of Qon and Qoff gating currents in control (left) and after addition of 1 µM ProTx-II to bath (right) were recorded in the presence of 1 µM tetrodotoxin to block ionic currents through the central pore of Nav1.2a. Depolarizations of 10 ms to potentials from -90 mV to +100 mV were applied from a holding potential of -100 mV. For this representative experiment, gating charge was reduced by 27% at + 100 mV. B, mean normalized voltage dependence of Qon gating charge movement in control (, n = 7) and 5 min after application of 1 µM ProTx-II (, n = 5). Gating currents for each experiment were normalized to Qon at +50 mV in the absence of toxin. C, mean values for percentage block by 1 µM ProTx-II of ionic current (at -10 mV, 65.3 ± 4.1%, n = 13) and gating charge Qon (at +100 mV, 24.6 ± 6.4%, n = 5).

    Depolarization to a range of membrane potentials revealed that ProTx-II inhibits activation of sodium channels across the full range of test potentials with little change in the voltage-dependence of activation (Fig. 1C). This is the behavior expected for a toxin that traps a voltage sensor in its resting conformation if sodium channels with toxin bound do not activate during the test pulse because, in this case, only unbound channels can activate. Consistent with the evidence that sodium channels with ProTx-II bound do not activate, there were no detectable effects on fast inactivation in the presence of ProTx-II. The rate of entry into the inactivated state (Fig. 1D), the rate of recovery from inactivation (Fig. 1E), and the voltage dependence of fast inactivation (Fig. 1F) were all unaffected by ProTx-II.

    Effects of ProTx-II on Gating Currents. Gating currents of Nav1.2a channels were recorded in the presence of 1 µM tetrodotoxin to block ionic conductance through the central pore of the channel. Families of Qon and Qoff gating currents were recorded in control solution and after 1 µM ProTx-II was applied (Fig. 2). ProTx-II substantially reduced the amplitude of gating currents (Fig. 2A). Integration of Qon currents revealed the voltage dependence of gating charge movement across the membrane during sodium channel activation (Fig. 2B). ProTx-II inhibited the gating charge movement across a wide range of membrane potentials. Inhibition of gating charge movement was 24.6 ± 6.4% at +100 mV, the most positive membrane potential studied. On average, 65% reduction of sodium current is caused by 24.6% reduction in gating current (Fig. 2C). By extrapolation, assuming the same concentration-response relationship for gating and ionic current (Fig. 1B), a concentration of ProTx-II that inhibits 100% of sodium channels would block 38% of gating charge movement. Evidently, ProTx-II blocks an essential component of gating charge movement, which greatly reduces activation of NaV1.2 channels.

    Fig. 3. Selective inhibition of a fast component of gating charge Qon by ProTx-II. A, kinetics of Qon movement showed in expanded scale in response to a depolarization to +50 mV from a holding potential of -100 mV in control (CON) and with 1 µM ProTx-II (PRO). B, subtraction of gating charge movement kinetics without and with ProTx-II (shaded area) reveals that ProTx-II blocks only the fast component of gating charge Qon. C, fraction of ON gating current, IQon, blocked by ProTx-II as measured at the peak of IQon (Peak Q, n = 10) or at a time when IQon in the absence of toxin had decayed to 20% of its peak value (Late Q) (n = 9).

    Fig. 4. Reversal of ProTx-II inhibition by strong depolarization. A, pulse protocol for toxin reversal experiments includes a conditioning depolarization to +100 mV of varying duration ranging from 10 to 630 ms, a 20-ms return to the holding potential of -100 mV to allow recovery from fast inactivation, and a test pulse (-10 mV for ionic and +50 mV for gating currents). B, ionic (left) and gating (right) currents recorded in the presence of 1 µM ProTx-II increase in amplitude as the duration of the conditioning depolarization increases (arrows indicate current after the shortest (10 ms) conditioning pulse). C, kinetics of ionic (,  = 0.39 ± 0.08 s, n = 4) and gating (,  = 0.36 ± 0.06 s, n = 4) current increase as a result of prolonged conditioning at +100 mV. To isolate the time course of voltage-dependent relief of toxin block of ionic and gating current, the data were normalized as follows. The steady-state level of ionic or gating current recorded during test pulses to -10 or +50 mV, respectively, in the presence of ProTx-II, were set to 0. The amount of current achieved after a 630 ms long depolarization to +100 mV was set to 1. Normalized I as a function of time (t) was plotted as [I(t) - I(0)]/[(I(630 ms) - I(0)].

    To further define the component of gating charge movement that is blocked by ProTx-II, we measured the kinetics of movement of the Qon gating charge by recording gating currents at high resolution in the absence and presence of 1 µM ProTx-II (Fig. 3A). We found that the toxin blocked only the fast component of Qon movement (Fig. 3B). The shaded area represents the difference between Qon in control and in the presence of toxin. Mean peak Qon current was reduced by 42% at +50 mV, but no significant reduction was observed in Qon current measured late in the pulse when only 20% of the peak Qon current remained (Fig. 3A). Preferential block of a rapid component of Qon current was observed across a broad range of potentials (Fig. 3C), consistent with the conclusion that the voltage sensor(s) affected by ProTx-II remain the same at all test voltages.

    Voltage-Dependent Reversal of ProTx-II Inhibition. Binding of -scorpion toxins is reversed by strong depolarization, indicating that voltage sensor activation can reverse toxin binding (Catterall, 1977; Rogers et al., 1996). We investigated whether the effects of ProTx-II on ionic currents and gating currents are also reversed by strong depolarization. In the first type of experiment, a strong conditioning depolarization (+100 mV) of increasing duration was followed by 20 ms at the holding potential (-100 mV) to allow recovery from fast inactivation and then by a test pulse to either -10 mV to measure ionic currents or +50 mV to measure gating currents (Fig. 4). The conditioning pulses reversed ProTx-II inhibition of both ionic and gating currents (Fig. 4B). The amplitudes of both ionic currents and gating currents increased because of the depolarizing pulses, and reversal of ProTx-II inhibition followed an exponential time course with similar kinetics for both ionic and gating currents (Fig. 4C). However, even the strongest conditioning depolarizations gave incomplete relief of inhibition of ionic current, reaching 60 to 65% of control at +100 mV and 630 ms.

    Applying conditioning pulses of different magnitude ranging from +40 to +100 mV allowed us to measure the kinetics of reversal of protoxin inhibition at these voltages (Fig. 5). Reversal of ProTx-II inhibition was fastest and most prominent at +100 mV (+100mV = 260 ms, 130% maximum increase in ionic current amplitude) and was very slow at +40 mV (+40mV = 1.5 s, 25% maximum increase in amplitude). These results indicate that strong depolarizations are required to overcome the energy of ProTx-II interaction with the resting state of the voltage sensor, drive it into the activated conformation, and cause dissociation of ProTx-II.

    Fig. 5. Voltage dependence of reversal of protoxin-II inhibition. A, pulse protocol for estimation of voltage-dependent kinetics of reversal of toxin inhibition of ionic current. Conditioning pulses of increasing duration (10 to 630 ms) were applied at various amplitudes (+40 to +100 mV) followed by a 20-ms repolarization to the holding potential of -100 mV and test pulse to -10 mV. Representative test-pulse traces of ionic current recorded after conditioning pulses of 10, 30, 50, 90, 170, 310, and 630 ms to the indicated potentials are shown. Arrows denote ionic current after the 10-ms conditioning pulse. B, voltage dependence of reversal at +100 mV (,  = 0.39 ± 0.08 s, n = 5), +80 mV (,  = 0.89 ± 0.23 s, n = 4), +60 mV (,  = 1.13 ± 0.13s, n = 4) and + 40 mV (,  = 1.5 ± 0.1 s, n = 4). Peak test pulse ionic current (Itest pulse n)was measured and normalized to that recorded after the 10-ms (Itest pulse 1) conditioning depolarization. Normalized current [100 x (Itest pulse n)/(Itest pulse 1)] is plotted versus conditioning pulse duration. C, voltage dependence of activation of toxinmodified channels. Peak test pulse currents after 630-ms prepulses to various potentials from experiments in B were measured and normalized to the peak test pulse current achieved at the most positive potential (150 mV). These normalized test pulse currents after 630 ms prepulses to the indicated potentials are plotted as a function of prepulse potential () and fit with a Boltzmann relationship with Vhalf = 64 mV and k = 17.4 mV. The normalized activation curve for control peak sodium current from Fig. 1B is replotted for comparison ().

    The energy required to cause dissociation of the toxin at positive potentials, and thereby permit activation of the channel, shifts the effective activation curve toward more positive potentials for toxin-bound channels, as observed in the voltage dependence of recovery of the sodium current during long depolarizations in the presence of toxin. Plotting the normalized percentage increase in test pulse current after 630-ms conditioning depolarizations from the experiments of Fig. 5B versus the conditioning pulse potential results in an isochronal activation curve for toxin-bound channels, which is strongly shifted toward more positive potentials relative to activation of toxin-free channels (Fig. 5C). After a 630-ms conditioning depolarization to +20 mV, there is no relief of toxin block, suggesting that toxin-bound channels cannot activate at this potential. Conversely, after 630 ms at +100 mV, the current increases 2.33-fold. Further relief is not achieved with increased depolarization. Thus, toxin-bound channels can undergo voltage-dependent activation, but that activation is greatly slowed and positively shifted because of the energy required to dissociate the toxin from the channel before it can activate.

    Protoxin Action on Cardiac NaV1.5 Channels. ProTx-II inhibits sodium currents conducted by NaV1.5 channels to a similar extent as NaV1.2 channels (Middleton et al., 2002; Fig. 6A). However, the reversal of inhibition during strong depolarizations is faster than for NaV1.2 channels (Fig. 6B). This difference in the rate of reversal of ProTx-II inhibition is observed across a wide range of prepulse potentials (Fig. 6C). These results indicate that the dissociation rate for ProTx-II from the activated state of NaV1.5 channels is approximately 2.5-fold faster than for NaV1.2 channels, suggesting subtype-specific effects of ProTx-II on sodium channel gating.

    Fig. 6. Voltage-dependent kinetics of reversal of ProTx-II inhibition for Nav1.2a and Nav1.5 channels. A, currents through Nav1.2a/β1 and Nav1.5/β1 channels before (CON) and 5 min after application of 1 µM ProTx-II to the external solution (PRO). B, normalized kinetics of toxin reversal as a result of +100-mV conditioning depolarization for representative cells expressing either Nav1.2a () or Nav1.5 (). Same pulse protocol as in Fig. 5 was used. C, the voltage-dependent kinetics of ProTx-II reversal by conditioning depolarization for Nav1.2a (filled bars, n  5) and Nav1.5 (open bars, n  5).

    Role of the IIS4 Voltage Sensor in Protoxin Binding and Action. Our findings that ProTx-II prevents a portion of gating charge movement and that its inhibition of sodium channels is reversed by positive prepulses suggest that ProTx-II is a gating modifier toxin that prevents channel activation via a voltage sensor-trapping mechanism. Previous studies and our results presented so far are most consistent with ProTx-II selectively inhibiting movement of one of the four voltage sensors of sodium channels. Inhibition of sodium currents by ProTx-II is fit by a single binding isotherm, suggesting interaction with a single site (Fig. 1B) (Middleton et al., 2002). A rapidly moving component of gating current is selectively inhibited (Fig. 3), suggesting that a kinetically distinct voltage sensor movement is blocked. As the voltage sensor in domain IV moves slowly (Chanda and Bezanilla, 2002), primarily the voltage sensors in domains I, II, and III contribute to the measured gating charge movement. Only 38% of gating charge movement is blocked when channel activation is completely inhibited, consistent with complete inhibition of movement of one of the voltage sensors in domains I, II, or III by ProTx-II. Gating modifier toxins bind to the S3-S4 linkers of sodium channels (Rogers et al., 1996). The β-scorpion toxin Css-IV binds to IIS3-S4 (Cestèle et al., 1998), whereas the -scorpion toxin LqTx binds to IVS3-S4 (Rogers et al., 1996). Even though ProTx-II stabilizes sodium channels in their closed state and β-scorpion toxins stabilize sodium channels in their open state, they both affect channel activation via a voltage-sensor trapping mechanism. Because β-scorpion toxins bind to IIS3-S4, we investigated this region as a potential receptor site for ProTx-II.

    Fig. 7. Effects of point mutations in the IIS3-S4 linker on inhibition by protoxin-II. A, a wash-in assay was used to estimate sensitivity of mutant channels to ProTx-II. Central pore current was recorded during test pulses to -10 mV applied every 10 s in control conditions (CON) and for 3 min after 1 µM ProTx-II was added to recording chamber. Percentage of peak current inhibition by the end of 3 min is taken as a measure of toxin sensitivity. Representative currents for the wild type Nav1.2a (left) and mutant L833C channel (right) are shown. B, percentage of peak current inhibited by 1 µM ProTx-II for wild type and various point mutants in the IIS3-S4 linker (n  4). The inset indicates the amino acid sequence of NaV1.2a between amino acids 831 and 860 with the mutated amino acids indicated in bold.

    We analyzed the effects of ProTx-II on mutant Nav1.2 channels with substitutions for selected amino acid residues in the IIS3-S4 loop (Fig. 7). Mutant L833C was significantly resistant to ProTx-II inhibition (27 ± 7% block compared with 65 ± 4% for wild-type, p < 0.05), suggesting that this amino acid residue may be involved in protoxin binding. The voltage-dependence of activation of L833C mutant channels was similar to that of WT channels (WT, V = -26.7 mV;L833C, V = -23.8, n = 12). The L833C mutation was surprisingly specific, because mutating the same amino acid to Asn in L833N had no effect on ProTx-II action. In contrast, mutations of the other amino acid residues in the IIS3-S4 loop had lesser or no effect.

    If ProTx-II modifies voltage-dependent conformational changes by binding to the outer end of the IIS4 voltage sensor, neutralizing the gating charges in the outer end of the IIS4 segment may affect the toxin-channel interaction considerably. Therefore, we tested whether the double mutant RR850,853QQ, in which the two outer gating charges in IIS4 are neutralized, interacts differently with ProTx-II. Previous studies have shown that these mutations do not substantially alter activation and inactivation of sodium channels (Cestèle et al., 2001; Sokolov et al., 2005). Sodium currents conducted by the double mutant RR850,853QQ were almost as sensitive to 1.25 µM ProTx-II as wild-type Nav1.2 channels (Fig. 8A; mean values, 44 ± 5.7%, n = 8 for RR850,853QQ versus 63 ± 5.7%, n = 7 for WT). These results indicate that there was no marked reduction in the affinity of the toxin for the resting channel and that the toxin was still able to trap the mutant IIS4 voltage sensor in its resting state.

    Fig. 8. Altered effects of ProTx-II on ionic and gating current for mutant RR850,853QQ. A, wash-in of 1.25 µM ProTx-II in wild-type Nav1.2a (left) and mutant channel (right). Arrows denote current amplitude 3 min after addition of toxin to the bath. B, loss of effect of ProTx-II on gating current in RR850,853QQ channels. Left, gating current transients during depolarizations to +100 mV for WT and RR850,853QQ channels in the absence (CON) and presence (PRO) of ProTx-II. Right, mean block of peak gating current at +100 mV by 1.25 µM ProTx-II in WT and RR850,853QQ channels (n  3). C, pulse protocol for conditioning-dependent toxin reversal and representative records in the presence of 1.25 µM ProTx-II for Nav1.2a and RR850,853QQ. Arrows denote current in the absence of conditioning. The slight decrease in current in RR850,853QQ is due mainly to slow inactivation of channels during conditioning pulses. D, kinetics of conditioning-dependent current relief normalized to the current in the absence of conditioning for Nav1.2a (control, , n = 4; 1.25 µM ProTx-II, , n = 5) and RR850,853QQ (control, , n = 6; 1.25 µM ProTx-II, , n = 6).

    In contrast to the lack of major effects of these mutations on the function of NaV1.2 channels and the inhibition of ionic current by ProTx-II, the block of gating current by ProTx-II was nearly completely lost in the RR850,853QQ mutant (Fig. 8B). Moreover, voltage-dependent reversal of inhibition during conditioning depolarizations was almost completely abolished in the double mutant RR850,853QQ channel (Fig. 8, C and D). No significant voltage-dependent reversal of inhibition was detected for RR850,853QQ with conditioning depolarizations as strong as +150 mV (data not shown). These results indicate that the gating movement of Arg850 and Arg853 at the outer end of the IIS4 segment is responsible for voltage-dependent reversal of protoxin inhibition and support the conclusion that ProTx-II interacts specifically with the IIS4 voltage sensor to inhibit gating current.

    Protoxin Blocks Gating Currents. A signature of gating modifier toxins acting on sodium channels is block of gating charge movement (Nonner, 1979; Meves et al., 1982). This effect probably reflects binding to a specific conformation of the voltage sensor and stabilizing that conformation according to a voltage-sensor trapping mechanism (Rogers et al., 1996; Cestèle et al., 1998, 2001, 2006). Our results show that ProTx-II shares this signature effect with other gating modifier toxins. By binding to a single site, ProTx-II reduces ON gating charge movement by 24.6 ± 6.4% at 1 µM and by approximately 38% when extrapolated to saturation. The toxin specifically reduces a rapidly moving component of gating charge, suggesting that it selectively impairs the movement of one or more voltage sensors that contribute to the rapid component of gating charge movement.

    Voltage-Dependent Reversal of ProTx-II Action. A second hallmark of gating modifier toxins is voltage-dependent enhancement or reversal of toxin action, first demonstrated for reversal of the binding and action of -scorpion toxins (Catterall, 1977; Rogers et al., 1996). This effect is thought to represent the voltage-driven outward movement of the IVS4 voltage sensor pushing the bound toxin off its binding site. As for -scorpion toxins, our results show that reversal of ProTx-II action is dramatically accelerated by strong positive prepulses. These results are most consistent with a voltage-sensor trapping model in which ProTx-II binds to one of the four S4 voltage sensors in its resting conformation and holds it in this inward position. Strong positive prepulses provide sufficient electrical energy to force outward movement of the voltage sensor into its activated conformation and are able to overcome the chemical energy of ProTx-II binding and push it off its receptor site. We observe this effect as depolarization-dependent reversal of toxin action.

    Protoxin Acts on the IIS4 Voltage Sensor. ProTx-II inhibits sodium channels with a concentration dependence that is fit by a single binding site model (Middleton et al., 2002; Fig. 1B). Because the voltage sensor in domain IV moves slowly (Chanda and Bezanilla, 2002), the voltage sensors in domains I, II, and III are probably primarily responsible for the gating charge movement measured in our experiments. Our findings suggest that approximately 38% of gating current would be blocked by ProTx-II at toxin concentrations at which all sodium channels are inhibited. They also show that a rapidly activating component of gating current is preferentially blocked. These results are consistent with the hypothesis that binding of a single ProTx-II molecule inhibits the function of a single voltage sensor.

    Three independent lines of investigation implicate the IIS4 voltage sensor in ProTx-II action. First, we found that the mutation L833C in the IIS3-S4 linker reduces the affinity for ProTx-II action, consistent with involvement of this amino acid residue in toxin binding. Second, we found that neutralization of the first two gating-charge-carrying arginine residues in IIS4, Arg850 and Arg853, completely prevents the effects of ProTx-II on gating current, indicating that these residues are required for the charge movement blocked by ProTx-II. Neutralization of these gating charges may inhibit movement of the entire IIS4 segment, including the potential gating charges carried by Arg856 and Lys859, and it may further impede the movement of gating charges in other S4 segments through the allosteric interactions among voltage sensors described previously (Chanda et al., 2004). Third, the loss of depolarization-dependent reversal of block by ProTx-II also requires Arg850 and Arg853 at the outer end of segment IIS4. These results suggest that Arg850 and Arg853 interact either sterically or electrostatically with ProTx-II as they move outward during activation and thereby push the toxin off its receptor site. All together, these three lines of evidence provide strong support for the conclusion that ProTx-II inhibits sodium channels by inhibiting the normal activation of the voltage sensor in domain II.

    ProTx-II and β-scorpion toxins both affect sodium channel activation but have opposing effects on that process. According to our hypothesis, ProTx-II impedes gating charge movement and channel activation by trapping the IIS4 voltage sensor in its resting conformation. β-Scorpion toxins facilitate channel activation by trapping the IIS4 voltage sensor in its activated conformation. The specific effects of both β-scorpion toxin and ProTx-II on the IIS4 voltage sensor suggest that this voltage sensor may have a privileged role in the actions of gating modifier toxins that affect sodium channel activation, whether they enhance or inhibit it. Further studies are required to completely map the amino acid residues that form the receptor site for ProTx-II on NaV1.2 channels.

    In a recent study of site-directed mutants of NaV1.5 channels, Smith et al. (2007) analyzed a large number of amino acid residues for involvement in the action of ProTx-II and found no major effects. They did not analyze the equivalent of mutant L833C, but they found that different mutations of residues in the IIS3-S4 loop had no effect on ProTx-II action, in agreement with our work. Based on this extensive analysis, it seems that ProTx-II has a unique site of action compared with other gating modifier toxins. It will be of great interest to define its receptor site and determine how it impedes movement of the IIS4 voltage sensor.

    Subtype-Specific Reversal of ProTx-II Action. Brain and cardiac Na+ channels display similar sensitivity for inhibition of channel activation by ProTx-II at hyperpolarized potentials where the channels are in their resting conformations (Fig. 1) (see also Middleton et al., 2002). However, the rates for voltage-dependent reversal of protoxin-II inhibition differ substantially between Nav1.2 and Nav1.5. The cardiac Nav1.5 isoform has faster and more prominent relief of inhibition at conditioning potentials ranging from +40 mV to +100 mV. These results indicate that at least part of the ProTx-II receptor site differs between brain and cardiac sodium channels and that the release of ProTx-II after voltage-dependent activation is more rapid for the Nav1.5 channel. This suggests that ProTx-II has a lower affinity for the activated state of the Nav1.5 channel than for that of Nav1.2. Determination of the amino acid residues responsible for the difference in ProTx-II action between NaV1.2 and NaV1.5 channels may give insight into the molecular basis for the substantially different voltage dependence of activation of these channel subtypes.

    Comparison of Voltage-Sensor Trapping by ProTx-II and Hanatoxin. Hanatoxin is another gating modifier toxin that inhibits KV2.1 channels (Swartz and MacKinnon, 1995). This channel is composed of four independent, identical subunits that form a noncovalently associated tetramer, in contrast to sodium channels whose four nonidentical but homologous domains are covalently linked in a single polypeptide chain. Hanatoxin blocks gating charge movement more completely than ProTx-II (Lee et al., 2003), as expected because hanatoxin binds to all four subunits of KV2.1 channels and prevents gating movement of all four S4 segments. Like Protoxin-II, hanatoxin binding to only one of the four homologous subunits/domains is sufficient to impair channel opening (Swartz and MacKinnon, 1997; Lee et al., 2003), and hanatoxin affinity for KV2.1 is reduced after prolonged depolarization (Phillips et al., 2005). Hanatoxin-blocked channels can open during long depolarizing pulses with toxin bound, resulting in accelerated deactivation and in strongly shifted voltage dependence of activation (Swartz and MacKinnon, 1997; Phillips et al., 2005) that resembles the strong positive shift of the voltage dependence of activation observed for ProTx-II-blocked sodium channels (Fig. 5). Thus, hanatoxin and ProTx-II have similar mechanisms of action that depend on trapping a voltage sensor in its resting conformation. Differences in the details of their mechanisms of action probably reflect the ability of KV2.1 channels to bind four hanatoxin molecules to their four voltage sensor domains.

【参考文献】
  Armstrong CM (1981) Sodium channels and gating currents. Physiol Rev 61: 644-682.[Free Full Text]

Auld VJ, Goldin AL, Krafte DS, Catterall WA, Lester HA, Davidson N, and Dunn RJ (1990) A neutral amino acid change in segment IIS4 dramatically alters the gating properties of the voltage-dependent sodium channel. Proc Natl Acad Sci U S A 87: 323-327.[Abstract/Free Full Text]

Benzinger GR, Kyle JW, Blumenthal KM, and Hanck DA (1998) A specific interaction between the cardiac sodium channel and site-3 toxin anthopleurin B. J Biol Chem 273: 80-84.[Abstract/Free Full Text]

Cahalan MD (1975) Modification of sodium channel gating in frog myelinated nerve fibres by Centruroides sculpturatus scorpion venom. J Physiol 244: 511-534.[Abstract/Free Full Text]

Catterall WA (1977) Membrane potential-dependent binding of scorpion toxin to the action potential Na+ ionophore. Studies with a toxin derivative prepared by lactoperoxidase-catalyzed iodination. J Biol Chem 252: 8660-8668.[Free Full Text]

Catterall WA (1979) Binding of scorpion toxin to receptor sites associated with sodium channels in frog muscle. Correlation of voltage-dependent binding with activation. J Gen Physiol 74: 375-391.[Abstract/Free Full Text]

Catterall WA (1980) Neurotoxins that act on voltage-sensitive sodium channels in excitable membranes. Annu Rev Pharmacol Toxicol 20: 15-43.[CrossRef][Medline]

Catterall WA (1986) Molecular properties of voltage-sensitive sodium channels. Annu Rev Biochem 55: 953-985.[CrossRef][Medline]

Catterall WA (2000) From ionic currents to molecular mechanisms: The structure and function of voltage-gated sodium channels. Neuron 26: 13-25.[CrossRef][Medline]

Catterall WA and Beress L (1978) Sea anemone toxin and scorpion toxin share a common receptor site associated with the action potential sodium ionophore. J Biol Chem 253: 7393-7396.[Abstract/Free Full Text]

Cestèle S and Catterall WA (2000) Molecular mechanisms of neurotoxin action on voltage-gated sodium channels. Biochimie 82: 883-892.[Medline]

Cestèle S, Qu Y, Rogers JC, Rochat H, Scheuer T, and Catterall WA (1998) Voltage sensor-trapping: enhanced activation of sodium channels by beta-scorpion toxin bound to the S3-S4 loop in domain II. Neuron 21: 919-931.[CrossRef][Medline]

Cestèle S, Scheuer T, Mantegazza M, Rochat H, and Catterall WA (2001) Neutralization of gating charges in domain II of the sodium channel subunit enhances voltage sensor trapping by β-scorpion toxins. J Gen Physiol 118: 291-302.[Abstract/Free Full Text]

Cestèle S, Yarov-Yarovoy V, Qu Y, Sampieri F, Scheuer T, and Catterall WA (2006) Structure and function of the voltage sensor of sodium channels probed by a β-scorpion toxin. J Biol Chem 281: 21,332-21,344.[Abstract/Free Full Text]

Chanda B, Asamoah OK, and Bezanilla F (2004) Coupling interactions between voltage sensors of the sodium channel as revealed by site-specific measurements. J Gen Physiol 123: 217-230.[Abstract/Free Full Text]

Chanda B and Bezanilla F (2002) Tracking voltage-dependent conformational changes in skeletal muscle sodium channel during activation. J Gen Physiol 120: 629-645.[Abstract/Free Full Text]

Guy HR and Seetharamulu P (1986) Molecular model of the action potential sodium channel. Proc Natl Acad Sci U S A 83: 508-512.[Abstract/Free Full Text]

Jaimovich E, Ildefonse M, Barhanin J, Rougier O, and Lazdunski M (1982) Centruroides toxin, a selective blocker of surface Na+ channels in skeletal muscle: voltage-clamp analysis and biochemical characterization of the receptor. Proc Natl Acad Sci U S A 79: 3896-3900.[Abstract/Free Full Text]

Jover E, Couraud F, and Rochat F (1980) Two types of scorpion neurotoxins characterized by their binding to two separate receptor sites on rat brain synaptosomes. Biochem Biophys Res Commun 95: 1607-1614.[CrossRef][Medline]

Lee HC, Wang JM, and Swartz KJ (2003) Interaction between extracellular Hanatoxin and the resting conformation of the voltage-sensor paddle in Kv channels. Neuron 40: 527-536.[CrossRef][Medline]

McPhee JC, Ragsdale DS, Scheuer T, and Catterall WA (1995) A critical role for transmembrane segment IVS6 of the sodium channel subunit in fast inactivation. J Biol Chem 270: 12,025-12,034.[Abstract/Free Full Text]

Meves H, Rubly N, and Watt DD (1982) Effect of toxins isolated from the venom of the scorpion Centruroides sculpturatus on the Na currents of the node of Ranvier. Pflugers Arch 393: 56-62.[CrossRef][Medline]

Middleton RE, Warren VA, Kraus RL, Hwang JC, Liu CJ, Dai G, Brochu RM, Kohler MG, Gao YD, Garsky VM, et al. (2002) Two tarantula peptides inhibit activation of multiple sodium channels. Biochemistry 41: 14,734-14,747.[CrossRef][Medline]

Nicholson GM, Willow M, Howden MEH, and Narahashi T (1994) Modification of sodium channel gating and kinetics by versutoxin from the Australian funnel-web spider Hadronyche versuta. Pflugers Arch 428: 400-409.

Nonner W (1979) Effects of Leiurus scorpion venom on the "gating" current in myelinated nerve. Adv Cytopharmacol 3: 345-352.[Medline]

Norton RS and Pallaghy PK (1998) The cystine knot structure of ion channel toxins and related polypeptides. Toxicon 36: 1573-1583.[Medline]

Phillips LR, Milescu M, Li-Smerin Y, Mindell JA, Kim JI, and Swartz KJ (2005) Voltage-sensor activation with a tarantula toxin as cargo. Nature 436: 857-860.[CrossRef][Medline]

Priest BT, Blumenthal KM, Smith JJ, Warren VA, and Smith MM (2007) ProTx-I and ProTx-II: gating modifiers of voltage-gated sodium channels. Toxicon 49: 194-201.[Medline]

Rogers JC, Qu Y, Tanada TN, Scheuer T, and Catterall WA (1996) Molecular determinants of high affinity binding of -scorpion toxin and sea anemone toxin in the S3-S4 extracellular loop in domain IV of the Na+ channel subunit. J Biol Chem 271: 15,950-15,962.[Abstract/Free Full Text]

Sheets MF, Kyle JW, Kallen RG, and Hanck DA (1999) The Na channel voltage sensor associated with inactivation is localized to the external charged residues of domain IV, S4. Biophys J 77: 747-757.[Medline]

Smith JJ, Cummins TR, Alphy S, and Blumenthal KM (2007) Molecular interactions of the gating modifier toxin ProTx-II with NaV1.5: implied existence of a novel toxin binding site coupled to activation. J Biol Chem 282: 12,687-12,697.[Abstract/Free Full Text]

Sokolov S, Scheuer T, and Catterall WA (2005) Ion permeation through a voltage-sensitive gating pore in brain sodium channels having voltage sensor mutations. Neuron 47: 183-189.[CrossRef][Medline]

Stefani E and Bezanilla F (1998) Cut-open oocyte voltage-clamp technique. Methods Enzymol 293: 300-318.[Medline]

Stühmer W, Conti F, Suzuki H, Wang X, Noda M, Yahadi N, Kubo H, and Numa S (1989) Structural parts involved in activation and inactivation of the sodium channel. Nature 339: 597-603.[CrossRef][Medline]

Swartz KJ and MacKinnon R (1995) An inhibitor of the Kv2.1 potassium channel isolated from the venom of a Chilean tarantula. Neuron 15: 941-949.[CrossRef][Medline]

Swartz KJ and MacKinnon R (1997) Hanatoxin modifies the gating of a voltage-dependent K+ channel through multiple binding sites. Neuron 18: 665-673.[CrossRef][Medline]

Tejedor FJ and Catterall WA (1988) A site of covalent attachment of -scorpion toxin derivatives in domain I of the sodium channel subunit. Proc Natl Acad Sci U S A 85: 8742-8746.[Abstract/Free Full Text]

Thomsen WJ and Catterall WA (1989) Localization of the receptor site for -scorpion toxins by antibody mapping: implications for sodium channel topology. Proc Natl Acad Sci U S A 86: 10,161-10,165.[Abstract/Free Full Text]

Yang N and Horn R (1995) Evidence for voltage-dependent S4 movement in sodium channel. Neuron 15: 213-218.[CrossRef][Medline]

Yarov-Yarovoy V, Baker D, and Catterall WA (2006) Voltage sensor conformations in the open and closed states in ROSETTA structural models of K+ channels. Proc Natl Acad Sci U S A 103: 7292-7297.[Abstract/Free Full Text]

作者单位：Department of Pharmacology, University of Washington, Seattle, Washington (S.S., T.S., W.A.C.) and Merck Research Laboratories, West Point, Pennsylvania (R.L.K.)

【关键词】  effect

    The effect of monoamine uptake inhibitor-type antidepressants on sodium channels of hippocampal neurons was investigated. Members of the tricyclic group of antidepressants are known to modify multiple targets, including sodium channels, whereas selective serotonin-reuptake inhibitors (SSRIs) are regarded as highly selective compounds, and their effect on sodium channels was not investigated in detail. In this study, a representative member of each group was chosen: the tricyclic antidepressant desipramine and the SSRI fluoxetine. The drugs were roughly equipotent use-dependent inhibitors of sodium channels, with IC50 values 100 µMat -150 mV holding potential, and 1 µMat -60 mV. We suggest that therapeutic concentrations of antidepressants affect neuronal information processing partly by direct, activity-dependent inhibition of sodium channels. As for the mechanism of inhibition, use-dependent inhibition by antidepressants was believed to be due to a preferential affinity to the fast-inactivated state. Using a voltage and perfusion protocol by which relative affinities to fast-versus slow-inactivated states could be assessed, we challenged this view and found that the affinity of both drugs to slowinactivated state(s) was higher. We propose a different mechanism of action for these antidepressants, in which slow rather than fast inactivation plays the dominant role. This mechanism is similar but not equivalent with the novel mechanism of usedependent sodium channel inhibition previously described by our group (Neuroscience 125:1019-1028, 2004; Neuroreport 14:1945-1949, 2003). Our results suggest that different drugs can produce use-dependent sodium channel inhibition by different mechanisms.It has become increasingly evident that the monoamine hypothesis cannot fully explain the pathophysiological mechanism of depression and the action of antidepressants (Shytle et al., 2002; Castren, 2005). Although the uptake blockertype antidepressants inhibit monoamine transporters in the low nanomolar range (Torres et al., 2003), their therapeutic effect appears only at a much higher plasma and brain concentration (Muscettola et al., 1978; Bolo et al., 2000). In therapeutic (i.e., low micromolar) concentrations, however, these drugs affect other protein targets as well; most importantly, they inhibit several types of ion channels (Sernagor et al., 1989; Hennings et al., 1999; Deak et al., 2000; Pacher et al., 2000; Yang and Kuo, 2002; Eisensamer et al., 2003; Gumilar et al., 2003; Choi et al., 2004). Thus, the activity of neurons is modified by antidepressants in a complex way, by elevating monoamine levels and by directly modulating ion channels.

    Neuronal sodium channels are ubiquitous and are crucial in dendritic integration, action potential initiation, and conduction. The potency of tricyclic antidepressants as sodium channel inhibitors was shown by the low micromolar IC50 values in vitro (e.g., Bou-Abboud and Nattel, 1998; Nicholson et al., 2002; Pancrazio et al., 1998), and by their in vivo efficiency against neuropathic pain (Namaka et al., 2004).

    Binding of antidepressants to sodium channels has been demonstrated (McNeal et al., 1985; Nicholson et al., 2002). Binding of [3H]batrachotoxin was inhibited by imipramine and amitriptyline, whereas [3H]saxitoxin binding was not altered (Nicholson et al., 2002). The exact location of the binding site remains unresolved; it may be different from the local anesthetic binding site (Barber et al., 1991) or may overlap with it (Wang et al., 2004).

    The main properties of sodium channel inhibition by antidepressants (use dependence, voltage dependence, and a hyperpolarizing shift of the inactivation curve) are shared by a variety of drugs (Deffois et al., 1996; Kuo et al., 2000), most notably local anesthetics and certain anticonvulsants. Because of the shared properties of inhibition, it has been assumed that the underlying mechanism must be the same (Ogata and Narahashi, 1989; Kuo and Bean, 1994; Nau et al., 2000; Yang and Kuo, 2002; Wang et al., 2004). However, as we have shown (Mike et al., 2003, 2004), use-dependent inhibition can also be explained by an alternative mechanism, in which the slow-inactivated state has the highest affinity to the drug, and the onset rate of inhibition is limited not only by slow association to inactivated channels (a binding reaction) but also by slow inactivation (a gating transition). Our aim, therefore, was to clarify the mechanism of sodium channel inhibition by the tricyclic antidepressant agent desipramine and to extend the investigation to a chemically unrelated antidepressant, the SSRI fluoxetine, because the interaction of this latter compound with sodium channels has not been studied in detail previously.

    All experimental procedures were approved by the local ethics committee and were in accordance with National Institutes of Health guidelines. Pregnant rats (17-18-day gestation) were anesthetized with a mixture of ketamine (50 mg/ml) and xylazine (10 mg/ml). The uterus was dissected out, placed in a laminar airflow box, and kept sterile. Individual fetuses were isolated; their whole brains were put into ice-cold minimal essential medium, and kept there during further dissection. Hippocampi of four to six fetuses were dissected out, incubated in 0.25% trypsin for 10 min, mechanically dissociated in minimal essential medium containing 10% fetal bovine serum, and plated at a density of 150 to 300 x 103 per 35-mm Petri dish (precoated with poly-L-lysine, 2 µg/ml). At 24 h after plating, the medium was replaced with B27-supplemented Neurobasal medium (Invitrogen, Carlsbad, CA), containing 25 µM 2-mercaptoethanol, 0.5 mM glutamine, and 25 µM glutamate. Half of the medium was changed twice a week thereafter to the same medium (i.e., Neurobasal + B27) without glutamate. Electrophysiological experiments were performed on neurons cultured for 7 to 24 days. Chemicals used for culture, and experiments were obtained from Sigma, unless otherwise mentioned.

    Transmembrane currents were recorded by whole-cell or outsideout patch configurations of the standard patch-clamp technique (Hamill et al., 1981) using an Axopatch 200B amplifier and the pClamp software (Molecular Devices, Sunnyvale, CA). Borosilicate glass patch pipettes (1.4-3.8 M) were coated with Sylgard (Dow-Corning, Midland, MI) to minimize capacitance. Series resistance was in the range of 3.5-9 M (recordings with series resistance values exceeding9M were excluded from analysis), and was compensated to 60-80%. The same relatively large-diameter pipettes were used for outside-out macropatches. Experiments were performed at room temperature (22°C). Pipettes were filled with an intracellular solution of the following composition: 70 mM CsCl, 70 mM CsF, 10 mM NaCl, 10 mM HEPES, and 10 mM Cs-EGTA; the pH was adjusted to 7.3 with CsOH. The composition of the external solution was 150 mM NaCl, 5 mM KCl, 1.4 mM CaCl2, 10 mM glucose, and 5 mM HEPES; pH was adjusted to 7.3 with NaOH. Because decreased external sodium concentration was shown to modify the rates and equilibria of both fast and slow inactivation, we avoided partial substitution of external sodium ions. Instead, all major results found in whole-cell experiments were repeated in outside-out patches at only one of the drug concentrations (30 µM) to confirm that drug-induced changes are not affected by errors of inadequate voltage control. Currents were low-pass-filtered at 10 kHz and sampled at a rate of 100 kHz.

    Drug application was performed using an improved version (Mike et al., 2004) of the dual U-tube method (Mike et al., 2000). Solution exchange time constants were in the 1-20-ms range and were not dependent on the duration of drug application.

    Subtraction of leak and capacitive artifacts was performed off line, using a standard P/n protocol. Stability of passive properties was monitored by comparing successive responses with test voltage pulses applied at the end of each voltage protocol. Data in which significant shift of passive properties was recorded were not used for analysis.

    Curve fitting was performed by the Solver function of Microsoft Excel. Time constants , or 1 and 2 were extracted from monoexponential or biexponential equations: I(t) = (Imax - Imin) x exp(-t/) + Imin and I(t) = (Imax - Imin) x [A1 x exp(-t/1) + A2 x exp(-t/2)] + Imin, where A1 and A2 are the contribution of components to the amplitude. Concentration-inhibition curves were fit to the Hill equation: I = Icontrol/[1 + ([D]/IC50)nH], where [D] is the drug concentration, IC50 is the concentration that causes 50% inhibition, and nH is the Hill coefficient.

    Statistical significance was determined using unpaired Student's t test or analysis of variance followed by Tukey-Kramer multiple comparisons test; p < 0.05 was considered significant. Results are presented as mean ± S.E.M. (unless otherwise noted) and the number of cells tested (n).

    Concentration-Inhibition Relationship at Different Holding Potentials. Both fluoxetine and desipramine inhibited sodium currents in a concentration-dependent manner. The potency of antidepressants was dependent on the holding potential but also influenced by other parameters of the stimulation protocol, such as the frequency of depolarizations or pulse duration (due to the use-dependent nature of inhibition).

    Onset of drug action was monitored by 10-Hz trains, each consisting of 10 depolarizations. The holding potential was varied during the experiment according to the pattern illustrated in Fig. 1A, bottom. It was of the following values: -150, -120, -90, or -60 mV. For all holding potentials, sodium currents were evoked by 10-ms pulses to -20 mV. Interpulse interval was thus 90 ms, whereas intertrain interval was 9 s. Increasing the intertrain interval to 19 s had no effect on the extent or the time course of inhibition by the drugs (data not shown), which argues against the possibility that use-dependence is caused by state-dependent access to the binding site.

    Fig. 1. Example for the holding potential-dependent inhibition of sodium channels caused by the antidepressants. The effect of 30 µM fluoxetine. A, peak amplitudes of currents evoked by 10-Hz trains of depolarizations delivered every 10 s during the experiment, whereas holding potentials were varied as shown at the bottom of the figure. Letters (from a to l) mark amplitudes for which the corresponding currents are shown in B. Amplitudes are expressed as relative to the amplitude of the first evoked current. B, individual currents evoked by trains of depolarizations. Letters (from a to l) indicate their position during the experiment as seen in A. Insets show peak amplitudes of individual trains on an expanded time scale. Top, control; Bottom, in the presence of 30 µM fluoxetine. Scale bars, 1 nA, 1 ms.

    From -150 to -90 mV, the amplitude and kinetics of single depolarization-evoked currents were similar. Amplitudes of currents evoked from holding potentials -120, -90, and -60 mV were 100 ± 0.9, 91 ± 1.5, and 35 ± 3.1% of the amplitudes evoled from -150 mV, respectively (calculated from the first depolarization of the last trains at each holding potential). Currents evoked from -120 and -150 mV were of the same amplitude; the small (9%) decrease observed with currents evoked from -90 mV was already significant (p < 0.01), whereas at -60 mV, already substantial inactivation was present. The relative amplitude of the currents evoked by the second, third, etc., depolarization differed more, depending on the holding potential, because of the voltage dependence of the recovery rate from inactivation. For holding potentials -150, -120, -90, and -60 mV, the tenth/first amplitude ratio of last trains was 0.98 ± 0.002, 0.94 ± 0.007, 0.88 ± 0.011, and 0.82 ± 0.017, respectively (n = 41; all differences between pairs of groups were significant, except -90 versus -60 mV). Because at this high number of cells (n) the S.E.M. value may be misleading; the 90% confidence interval values (supposing normal distribution) are more informative: 0.96-1.01, 0.87-1.02, 0.76-1.00, and 0.64-1.01, for -150, -120, -90, and -60 mV holding potentials, respectively. The protocol used for studying the holding potential dependence can be seen in Fig. 1A. For illustration, a cell with a larger-than-average decay within trains was chosen so that alteration of peak amplitudes within trains (Fig. 1B) could be better seen. Drug application was started at -120 mV holding potential; before and after that, the membrane was kept at different potentials for 40 s (or 50 s in the case of -60 mV). Washout was monitored using the same protocol (Peak amplitudes from a whole experiment, including both onset and washout, are shown in Fig. 3B).

    Fig. 3. Experiments on the accessibility of the binding site. A, the effect of changing the pH of the extracellular fluid on sodium channel inhibition by fluoxetine. Left, examples for plots of peak amplitude values as a function of time, obtained using the 10-Hz protocol (the holding potential was -120 mV). Right, mean ± S.E.M. values for n = 4 to 6 individual cells. Top, pH was lowered to 6.0 before fluoxetine perfusion. Middle, control (pH = 7.3). Bottom, pH was increased to 8.5 before fluoxetine perfusion. B, example for the lack of inhibition by intracellularly applied 100 µM fluoxetine. Top and bottom, plots of peak amplitudes obtained by the 10 Hz protocol as described at Fig. 1 (only this time both onset and offset are shown; furthermore, -150-mV holding potential was not investigated-see bars below figures indicating the holding potential). Top, ("Standard IC"), the intracellular solution contained no fluoxetine. Bottom, ("100 µM Fluoxetine IC"), 100 µM fluoxetine was dissolved in the intracellular solution. No significant difference in the extent of inhibition was observed.

    At strongly hyperpolarized membrane potential values (-150 and -120 mV), the inhibition had a tonic and a phasic component. Development of the tonic component is best seen at the amplitudes of currents evoked by the first pulses of each train at -120 mV before and during drug application (Fig. 1A, see the highest points of each group from f to g).

    The tonic component at strongly hyperpolarized membrane potential values (-150 and -120 mV) probably reflects association to resting state (note that a significant inhibition is already present at the very first evoked current after start of fluoxetine perfusion); therefore, we will call this component "resting state inhibition" (RSI). Because RSI was not usedependent (not dependent on intertrain interval, as we have discussed), it was most likely due not to stabilization of an inactivated conformation but possibly to steric occlusion of the conduction pathway. Phasic inhibition at strongly hyperpolarized potentials (Fig. 1A, compare f and g or b and h; the inhibition within a train was augmented in the presence of the drugs), on the other hand, was probably due to the increased affinity of antidepressants to inactivated states (it was caused either by an increased association of the drug during the 10-ms depolarized periods or by stabilization of the drug-bound inactivated state, which hindered recovery during the interpulse periods). This component, which developed because of the preferential affinity toward inactivated states, will be called "state-dependent inhibition" (SDI). At -90 mV, the tonic component of drug-induced inhibition was not entirely attributable to RSI; rather, it is a mixture of RSI and SDI. Whenever the holding potential allowed a fraction of the ion channel population to inactivate, the component of SDI developed, and it became more dominant with both increased depolarization and increased concentration of the drugs. In Fig. 1, SDI is reflected both by the enlarged phasic inhibition within a train (f versus g, b versus h, or c versus i) and by the enlarged tonic component at -90 and -60 mV holding potentials (e.g., see the first amplitudes at -90 mV in the presence of fluoxetine). Under physiological conditions (-60 mV), SDI was the major component of antidepressant-induced inhibition. The exact mechanism of this component is the main subject of this study.

    Fig. 2. Concentration - inhibition curves for fluoxetine (left) and desipramine (right) at holding potentials -150, -120, -90, and -60 mV. Each data point is an average of four to six individual measurements. Thin lines show curves obtained by fitting the Hill equation (see Materials and Methods) to data.

    SDI can be due either to state-dependent affinity or to state-dependent accessibility of the binding site. The former hypothesis, called the "modulated receptor model" (Hille, 1977), supposes that association is possible to all conformational states of the channel, but with a different affinity. The latter hypothesis, the "guarded receptor model" (Starmer et al., 1984), assumes that only certain states are accessible for drug binding.

    Because the onset of SDI did not depend on previous activation, and the extent of SDI was dependent on the frequency and not the number of pulses (data not shown), we found that the modulated receptor model was more adequate in explaining our experimental findings than the guarded receptor model. Therefore, throughout the description of our results, we will propose hypotheses within this conceptual framework.

    Repeated trains of depolarizations from holding potentials -150, -120, and -90 mV evoked similar successive groups of currents, indicating that at these holding potentials, the equilibrium was reached within 9 s (i.e., during the time between two trains). At -60 mV holding potential, however, the amplitude was progressively decreasing, and equilibrium was not reached within 50 s (in 98 of 98 cells). The time constants of both fast inactivation and recovery from fast inactivation were in the range of 0.2 to 20 ms (see below), so equilibrium between resting and fast-inactivated states should have been reached within a few hundred milliseconds. This indeed was indicated by the fact that the exponential fit of the decrease of amplitude within the train (Fig. 1B, insets) gave time constants in the range of 136-204 ms. The fact that the equilibrium was not reached as fast as expected (based on the time constants of fast inactivation and recovery from fast inactivation) revealed that under control conditions at -60 mV, slow inactivation did take part in determining the steady-state conformational equilibrium, and thus the availability of ion channels for activation.

    In the presence of both antidepressants, prolonged development of equilibrium was observed already at more negative membrane potentials: at -90 mV at a 3 µM concentration of either drug, it was observable in roughly half of the experiments (five of nine cells), and, at higher concentrations, in all (34 of 34) cells. At -120 mV, it was apparent in all cells at 100 µM concentration, and in four of nine cells at 30 µM, which may indicate that antidepressants promoted the appearance of slow inactivation.

    Summarizing the points illustrated by Fig. 1, the antidepressants caused a low potency inhibition (RSI) at resting conformation of the ion channels, most probably by steric occlusion of the pore. Upon depolarization, a high-potency SDI developed. Properties of state dependence in the case of these drugs suggest that it can be explained by the modulated receptor model rather than the guarded receptor model. Under control conditions, occurrence of slow inactivation is observed only at -60 mV holding potential, whereas in the presence of the antidepressants, slow inactivation seems to be promoted, being present at more negative holding potentials as well.

    Figure 1 suggests that slow inactivation seems to occur at more negative potentials in the presence of the drugs (e.g., at -120 and -90 mV in the presence of 30 µM fluoxetine in Fig. 1). Therefore, we have to ask whether this indicates accelerated entry into and slower recovery from this state. The traditional view explains this phenomenon solely by slow and membrane potential-dependent association of the drugs. In subsequent sections of this article, we challenge this oversimplified explanation and investigate whether preferential affinity to the slow-inactivated state was an essential element of sodium channel inhibition by antidepressants.

    Concentration-inhibition curves are shown in Fig. 2. The extent of drug-induced inhibition was calculated from the amplitude of the current evoked by the first depolarization of the last train at each holding potential (in the example shown in Fig. 1, these relative amplitudes are shown by the points right below letters b, c, d, and f in control and h, i, j and l in the presence of fluoxetine). Note that the curves constructed this way mostly reflect RSI at holding potentials -150 and -120 mV but predominantly reflect SDI at -90 and especially -60 mV. The potency of both drugs was thus profoundly dependent on the holding potential. The IC50 values for fluoxetine were 107.9 µM(-150 mV holding potential, nH = 1.19), 74.1 (-120 mV, nH = 1.39), 23.9 (-90 mV, nH = 1.45), and 1.11 µM(-60 mV, nH = 0.90). For desipramine the following IC50 values were found: 83.4 (-150 mV, nH = 1.11), 56.7 (-120 mV, nH = 1.40), 33.3 (-90 mV, nH = 1.21), and 1.68 µM(-60 mV, nH = 0.76). The inhibitory effect of antidepressants was substantially influenced by the membrane potential, especially between -90 and -60 mV, which means, on the one hand, that the drugs were very potent around the resting membrane potential and, on the other hand, that the inhibition was most voltage-sensitive in this membrane potential range.

    Accessibility of the Binding Site. Inhibition was more potent at depolarized membrane potentials. However, because drugs were applied from the extracellular side, this seems to be in contradiction with the fact that the majority of both drugs are in their positively charged forms at pH 7.3 and would only be consistent with the idea that the drugs entered a binding site from the intracellular side (after permeating the membrane). To investigate this possibility, we studied the accessibility of the antidepressant binding site from the extracellular and intracellular sides.

    One possible approach to study the accessibility is manipulation of the extracellular pH. Fluoxetine and desipramine both contain a tertiary amine nitrogen that is positively charged in neutral pH (pKa values for fluoxetine and desipramine are 10.3 and 9.9, respectively). Alkalinization of the extracellular fluid increases the ratio of the neutral form of fluoxetine (from 0.1% at pH 7.3 to 1.55% at pH 8.5), thus helping to overcome a supposed hydrophobic barrier. Acidification, on the other hand, practically eliminates the neutral form (0.005% at pH 6.0). We compared the inhibition caused by 30 µM fluoxetine at three different pH values: 6.0, 7.3, and 8.5 (Fig. 3A). Acidification of the external fluid in itself caused a small inhibition of sodium currents that was probably due to channel blocking properties of hydrogen ions (Woodhull, 1973). Control amplitude of currents (relative to the amplitude measured at pH 7.3) was 0.87 ± 0.02 at pH 6.0 but 1.00 ± 0.01 at pH 8.3. The tenth/first amplitude ratio was not changed significantly. These results indicate that the gating of channels in itself was not affected noticeably. (A significant effect on the recovery rate from either the fast- or slow-inactivated states, as well as on the onset rate of slow inactivation, would have been obviously visible when the 10 Hz protocol was used; a change in the inactivation-recovery from inactivation equilibrium would have altered the ratio of amplitudes.) In the presence of 30 µM fluoxetine, the amplitude of the first current of the train (which reflects RSI) was 78.0 ± 3.0% of the control amplitude at pH 7.3; the inhibition significantly increased at pH 8.5 (26.10 ± 12.4% of the control) and decreased at pH 6.0 (90.2 ± 0.6% of the control) (n = 3to6; p < 0.01 and p < 0.05 for pH 8.5 and 6.0, respectively). Ratios of tenth/first amplitudes changed similarly, indicating that SDI became larger at pH 8.5 and smaller at pH 6.0; tenth/first amplitude ratios at pH 7.3, 8.5, and 6.0 were 0.70 ± 0.02, 0.28 ± 0.08, and 0.94 ± 0.01, respectively. These results suggest either that the active species is neutrally charged and thus may have significant hydrophobic interactions with its binding site or that fluoxetine molecules have to overcome a hydrophobic barrier on their path to the binding site.

    Fig. 4. Effect of fluoxetine and desipramine on steady-state sodium channel inactivation. A, voltage protocol used for the study. Prepulse duration was 0.4, 2, or 8 s. B, example for the prepulse duration-dependent shift of inactivation curves in the presence of 30 µM fluoxetine (left) and 30 µM desipramine (right). Current amplitudes are expressed as relative to the amplitude evoked from a -120-mV holding potential. Whereas in the control (open symbols) no significant difference was observed depending on the time provided for equilibration, in the presence of antidepressants (filled symbols), the shift was larger with longer time for equilibration (n = 4 for each data point). C, summary of results regarding the concentration-dependent and prepulse duration-dependent shift of inactivation curves. V values were calculated for each individual cell. Each column is the average of measurements from n = 4to6 cells. Significant differences (compared with control obtained using the same prepulse duration) are marked by asterisks. The level of significance is not indicated; it ranged from p < 0.05 to 0.001. Significant difference between results with prepulse durations 0.4 and 8 s was found at all concentrations, as indicated by signs.

    For potassium channels, it has been proposed that fluoxetine acts as an open channel blocker from the inside of the cell (Choi et al., 2004). To test this possibility, we studied the effect of high concentration of fluoxetine applied intracellularly. When 100 µM drug was dissolved in the pipette solution and recording of sodium currents was started within 5 s after break-in (i.e., reaching whole-cell configuration), no significant decrease of sodium currents was observed during diffusion of the pipette solution into the cell; peak amplitude at 4 min was 100.6 ± 0.8% of the peak amplitude at <5s(n = 8). Under control conditions (no fluoxetine in the pipette), the same percentage was 100.5 ± 0.1% (n = 8). In contrast, when 1 mM QX314 (the lidocaine derivative sodium channel inhibitor, which is known to have an intracellular binding site) was included in the pipette, sodium currents decreased to 35.4 ± 6.7% of the amplitude evoked by the first depolarization within a few minutes after break-in [the time constant of decay was 36.0 ± 6.4s(n = 5)]. Furthermore, when (during the continuous intracellular presence of 100 µM fluoxetine) 10 µM fluoxetine was applied extracellularly, the extent of inhibition (92.7 ± 1.5% of control at -120 mV, n = 6) was not significantly different from the results obtained with standard pipette solution (94.2 ± 2.9% of control, n = 4) and the onset kinetics was apparently identical. Figure 3B shows an example for the effect of 10 µM fluoxetine using the protocol described above, with either control or 100 µM fluoxetinecontaining pipette solution.

    The fact that intracellular application of fluoxetine was ineffective shows first of all that voltage-dependent association of the positively charged form is not a significant source of voltage-dependent inhibition. Our experiments suggest that 1) the majority of drug molecules are either neutral at their binding sites or have to overcome a hydrophobic barrier to reach it, and 2) the binding site seems to be more easily accessed from the extracellular side. As for the reason for the slow onset of inhibition, slow accumulation within the cell (implying intracellular accessibility) can be excluded.

    Time-Dependent Shift of Inactivation Curves. A characteristically slow development of equilibrium was also observed while we measured inactivation curves. Three separate protocols with prepulse durations 0.4, 2, and 8 s were used (Fig. 4A). In control, inactivation curves obtained with the three protocols differed only slightly: half-inactivation voltages (V) were -64.02 ± 1.22, -64.31 ± 1.16, and -64.82 ± 1.14 mV (n = 20), respectively. A time-dependent leftward shift of the inactivation curve during recording was observed even without drug application (less than 5 mV). Because of this, and because of the cell-to-cell variability of V in control, drug induced changes were expressed by the shift of V (V), calculated from control values, both before and after drug application, as V = (V control + V washout)/2 - V drug. Using 8-s prepulses, the shifts caused by 3, 10, 30, and 100 µM fluoxetine were -2.82 ± 0.79, -8.44 ± 0.76, -21.50 ± 1.68, and -43.24 ± 1.78, respectively. In the case of desipramine, the corresponding values were: -3.57 ± 1.34, -7.66 ± 0.53, -14.49 ± 3.57, and -38.67 ± 1.46 for 3, 10, 30, and 100 µM, respectively (n = 3 to 5; V values significantly different from zero are marked by asterisks in Fig. 4C) In contrast to control, in the presence of antidepressants, the hyperpolarizing shift markedly depended on prepulse duration. Figure 4B shows an example for the prepulse duration-dependent shift of inactivation curves in the presence of 30 µM fluoxetine and desipramine. Averaged concentration- and prepulse durationdependence is illustrated in Fig. 4C. The difference between the shift at 0.4 s versus 8 s was significant at 3 µM fluoxetine and 30 µM desipramine (paired t test; significant differences are indicated by signs in Fig. 4C). Data presented in Fig. 4 were obtained by whole-cell measurements. Excision of the membrane patch in itself caused a substantial (12.6 ± 0.43 mV, n = 8) leftward shift in the steady-state inactivation curve (the reasons and mechanism of which were not addressed in this study). Nevertheless, the degree of druginduced shift was similar in outside-out patches. In the presence of 30 µM fluoxetine (n = 10), it was 6.18 ± 1.63, 8.27 ± 2.19, and 10.67 ± 3.23 mV; 30 µM desipramine caused 5.00 ± 2.16, 4.50 ± 2.99, and 8.25 ± 4.48 mV (n = 4) shift of V, for 0.4-, 2-, and 8-s prepulse durations, respectively.

    Shift of the inactivation curve can be caused by stabilizing either the fast- or the slow-inactivated state. The slow onset of the shift could be explained in both cases: by slow gating (if the slow-inactivated state is stabilized) or by slow association (if the fast-inactivated state is stabilized).

    Gating Transition Rates in the Presence of Antidepressants. Both fast and slow inactivated states can be stabilized by either accelerating transition into the relevant inactivated state or reducing the rate of recovery from it. We studied, therefore, the effect of antidepressants on the relevant gating transition rates.

    During the discussion of drug effects on gating transition rates, however, we need to bear in mind that affinity is assumed to be state-dependent. Therefore, if we do experiments in the continuous presence of the drugs, we will see not only the effect of voltage on gating but also the combined effects of voltage on gating and binding. The interpretation of the results, therefore, is not straightforward, as we will discuss.

    Fast-Inactivated State. In previous studies with the tricyclic antidepressant imipramine, high concentrations of the drug (30-300 µM) were shown to accelerate the decay phase of sodium currents (Yang and Kuo, 2002), whereas low (but already effective) concentrations (2-5 µM) have no effect on it (Ogata and Narahashi, 1989). Possible mechanisms for acceleration of the decay phase are rapid association during activation, which results in either steric occlusion of the conduction pathway (channel block), or acceleration of inactivation. (Detectable association must happen within the time window in which a significant fraction of the channels is open, and it must either block open channels, or promote inactivation. This mechanism requires a high association rate: 1 x 107 M-1s-1. Alternatively, acceleration of the decay phase may happen with a mechanism that does not require this high association rate, if we assume that drugbound channels can conduct. Because the drug associates even to resting channels (although with a lower affinity), in the presence of the drug, a fraction of the channels will be drug-bound even at hyperpolarized membrane potential. When the process of activation and inactivation is started by depolarization, drug-bound channels already inactivate with the increased rate inflicted upon them by the drug molecule.

    We studied the decay phase of evoked sodium currents in both whole-cell and outside-out patch measurements. Activation time and decay kinetics were significantly faster in outside-out patches, whereas the voltage-dependence of activation did not differ significantly. In neither of the two configurations did any of the antidepressants significantly alter the decay time constant of sodium currents at any concentration (1-100 µM). Because outside-out patch data reflect channel kinetics more faithfully, we illustrate antidepressant effect using results obtained in this configuration (Fig. 5A). Depolarization-evoked currents are illustrated on the left in Fig. 5A. The first 1.5 ms of decay was sufficiently well fit by a monoexponential equation; time constants varied between 0.28 and 0.67 ms (0.42 ± 0.048 ms) in control and were not significantly changed in the presence of up to 30 µM fluoxetine (Flx/Ctr = 0.976 ± 0.045 paired t test, p = 0.934, n = 6) or 30 µM desipramine (DMI/Ctr = 1.003 ± 0.098 paired t test, p = 0.979, n = 4) (Fig. 5A).

    Fig. 5. Modification of gating kinetics by antidepressants. A, decay rate (which reflects the onset of fast inactivation) is not significantly affected by antidepressants. Left, example for a depolarization-evoked sodium current in the absence and presence of 30 µM fluoxetine. The current evoked in the presence of fluoxetine was scaled up to match the amplitude of control current. Inset shows currents with the original amplitudes. Right, average of decay time constants in the absence (Ctr) and presence (Flx and DMI) of the drugs. Single exponentials were fit to the decay phase of currents. B to D, different aspects of gating kinetics was studied using double-pulse protocols. Control data, as well as results obtained in the presence of 3 and 30 µM concentrations of either fluoxetine (left) or desipramine (right) are shown. Data points are mean of three to seven individual measurements; error bars show S.E.M. Dashed lines show biexponential curves fit to data. The double-pulse voltage protocol used in that particular series of experiments is illustrated in boxes. Amplitudes were normalized to the amplitude of current evoked by the first pulse. B, recovery from fast inactivation is not affected significantly by the drugs. Interpulse intervals were varied between 0.25 and 512 ms. C, onset of slow inactivation is accelerated in the presence of fluoxetine and desipramine. First pulse duration was varied between 0.25 and 4096 ms. D, rate of recovery from slow inactivation is reduced in the presence of fluoxetine and desipramine. Interpulse intervals were varied between 0.2 ms and 10 s.

    We also measured the recovery from fast inactivation. A standard double-pulse protocol was used. Activation and inactivation were evoked by the first 10-ms depolarization to -20 mV (this duration is enough to produce >99% inactivation). After an interpulse interval, the extent of recovery was tested by a second identical depolarization. The length of the interpulse interval was varied between 0.25 and 512 ms; the cells were kept at -150-mV holding potential during this time. The ratio of the second (test pulse-evoked) and the first (control pulse-evoked) current amplitudes was calculated and plotted as a function of interpulse interval duration. The average of recovery curves (n = 13) were fit by a biexponential function. The fast time constant, which contributed 89.4% of the total amplitude, was 1.16 ms; the slow time constant was 24.2 ms. There was no significant change in the fast time constant of recovery at any of the concentrations from 1 to 30 µM (Fig. 5B). Values were in the range of 1.00 to 1.22 ms, contributing 85.7 to 92.0% of the total amplitude. Slow time constants showed a concentration-dependent tendency to increase (up to 100 ms), but this did not cause significant change in the overall shape of recovery curves, due to this component's small contribution to the amplitude. (Current amplitudes shown in Fig. 5, B-D were normalized to the current evoked by the first pulse in the presence of the drug. Therefore, the figure does not show the reduction of amplitude, which was significant in the presence of 10 and 30 µM concentrations of either of the drugs.) In summary, neither the onset of fast inactivation nor the recovery from fast inactivation was significantly affected by antidepressants.

    Slow-Inactivated State. To study slow inactivation, a similar double-pulse protocol was used, consisting of two depolarizations (to -20 mV) from a hyperpolarized (-150 mV) holding potential (Fig. 5C, inset). The current evoked by the first depolarization served as a control; the duration of this depolarization-which was intended to evoke slow inactivation-was varied between 0.25 and 4096 ms. Before the second depolarization, the membrane was held for 10 ms at hyperpolarized holding potential (-150 mV); this duration was found to be enough for nearly full (85-95%) recovery from fast inactivation. The extent of slow inactivation was estimated by the ratio of the currents evoked by the second and first depolarizations. Data points were fit by biexponential equations. The fast time constant was 9.83 ms in control; it was responsible for only 8.0% of the decrease in amplitude. This component probably included residual fast inactivation (the fraction of fast-inactivated channels that could not recover within 10 ms) and intermediate forms of inactivation. [A form of inactivation with an intermediate onset and offset kinetics has been shown to be responsible for a small fraction of inactivation in skeletal muscle sodium channels (Kambouris et al., 1998). In addition, a rapid onset-slow offset form of inactivation has been observed in pyramidal cells (Mickus et al., 1999).] The slow time constant, which probably corresponded to slow inactivation, was 2347 ms in control; considerable (>25%) slow inactivation took place only during depolarizations longer than 1 s (Fig. 5C). In the presence of 3 µM fluoxetine and 10 µM desipramine, the time-dependent decrease in availability was significantly accelerated, as shown by a marked and concentration-dependent decrease in the slow time constant (paired t test, based on fits to individual measurements). Slow time constants of biexponential equations fit to mean data were 2347.3, 1493.7, 647.4, and 418.2 ms in the presence of 1, 3, 10, and 30 µM fluoxetine, respectively, and 2817.5, 2941.6, 2226.2, and 1234.2 in the presence of 1, 3, 10, and 30 µM desipramine, respectively. Neither the fast time constants nor the proportion of decrease for which they were responsible changed significantly. Fast time constants ranged from 3.4 to 14.1 ms and the corresponding relative amplitude from 4.6 to 12.5%; neither showed any concentration-dependent tendency.

    Recovery from slow inactivation was investigated using a protocol similar to that used for recovery from fast inactivation, except that the duration of the first depolarization was 5 s (instead of 10 ms) (Fig. 5D, inset). The extent of recovery was measured after interpulse intervals ranging from 0.25 ms to 10 s. Results for control and for 3 and 30 µM concentrations of both drugs are shown in Fig. 5D. Biexponential equations were fit to recovery curves. Time constants (with relative contribution of the component to the amplitude) were 2.21 ms (45%) and 58.25 ms (55%) in control. The fast time constant probably reflects recovery from fast inactivation, because5sinthe absence of drugs is only enough for 50% slow inactivation. Both drugs markedly slowed recovery. At 3 µM, they changed both time constants by a factor of approximately 4 to 10; time constants for fluoxetine were 19.43 ms (68%) and 670.8 ms (32%) and for desipramine, 9.75 ms (72%) and 621.6 ms (28%). Either drug at 30 µM caused a roughly 15- to 30-fold reduction of time constants; for fluoxetine, they were 65.03 ms (40%) and 1059.1 ms (60%), and for desipramine, 36.07 ms (35%) and 1738.7 ms (65%). Particular time constants cannot be ascribed to specific transitions, because the process of recovery is too complex. At the beginning of recovery, the ion channel population is not homogenous but is distributed among bound and unbound fastinactivated and slow-inactivated subpopulations. The distribution is dependent on the concentration of the drug. (We also need to be aware of the existence of multiple slow inactivated states with different recovery time constants, as we have mentioned.) Furthermore, recovery of fast-inactivated and slow-inactivated bound channels can be started either by dissociation or by recovery from inactivation, and the fraction of the subpopulation that starts with the former or the latter remains unclear.

    Drug-induced changes in the time-dependent availability plots (Fig. 5, C and D) can be explained by two basic mechanisms. The conventional explanation assumes that altered rates reflect association and dissociation. However, it follows from the modulated receptor hypothesis that state-dependent affinity and alteration of relevant transition rates mutually presume one another; i.e., we can suppose that there is no preferential affinity without altered gating rates. If this is so, a presumed preferential affinity to slow inactivated state must be reflected by altered gating transition rates. What is the source of alterations in the time-dependent availability plots? Is it only increased affinity of the drugs to fast-inactivated channels or is slow inactivation also involved? If the slow-inactivated state was also preferred by the drugs, the rates of progression and regression of inhibition would be determined not only by the increased affinity of the drugs to slow inactivated channels but also by the altered gating rates. Resolving this problem is only possible by assessing the relative affinities of the drugs to the fast- and slow-inactivated states.

    Fig. 6. Test of association to fast-versus slow inactivated channels. A, voltage and perfusion protocol. Four consecutive trials were performed. In trials 1 (gray line) and 2 (black line), the effect of voltage protocols was tested in the absence of the drug. In trials 3 and 4, the same voltage protocols were repeated while drug was perfused during the train of three consecutive 100-ms depolarizations (for a total time of 310 ms; indicated by the shaded area). The inhibition caused by drug perfusion was significantly larger after the 20 s depolarization (trial 4 versus 2) than without depolarization (trial 3 versus 1). B, example for the effect of 30 µM fluoxetine. F-Ctr, trial 1; S-Ctr, trial 2; F-Flx, trial 3; S-Flx, trial 4. C, averaged amplitudes of test-pulse-evoked currents measured without drug application (Ctr) and after brief application of fluoxetine (Flx) or desipramine (DMI). Current amplitudes were tested after the protocol which favors either the fast-inactivated (F, gray columns) or the slow-inactivated (S, black columns) state. Current amplitudes were normalized to the current evoked in trial 1 by the first depolarization of the 3 x 100-ms train.

    Relative Affinity to Fast- and Slow-Inactivated States. The protocol we designed for studying the relative affinity to fast- and slow-inactivated states consisted of four trials. Trial 1 served as a control, trial 2 investigated the effect of slow inactivation, trial 3 examined the effect of drugs on predominantly fast-inactivated channels, and trial 4 tested the effects of drugs on predominantly slow-inactivated channels.

    For assessing the association to fast-inactivated state (trials 1 and 3), we designed a protocol that provides enough time for fluoxetine to bind to predominantly fast inactivated channels but does not allow significant slow inactivation. The fraction of ion channels in the fast-inactivated state was maximized by applying three consecutive 100-ms depolarizations (this duration-at least in the absence of drugs-is not enough for significant slow inactivation), with 5-ms gaps between them (enough to reach roughly 80% recovery from fast inactivation) (Fig. 6A). The protocol obviously cannot provide a homogenous population of fast-inactivated channels: a minor fraction of channels may undergo "intermediate" (Kambouris et al., 1998) or "prolonged" (Mickus et al., 1999) inactivation, depending on the type of channels. The former was described as a separate kinetic component, both its onset and recovery rates being between characteristic rates of fast and slow inactivation, and the latter as a fast onset-slow recovery inactivation process. Nevertheless, at any type of sodium channel, the 3 x 100-ms pulse depolarization protocol results in a distribution of ion channels, where the majority (> 85%) is in the fast-inactivated state. Drug application (trial 3) was provided throughout this short train. The extent of inhibition was tested 50 ms after the end of the third pulse. This interval, spent at -150 mV, was enough for almost full (96.4 ± 1.1%) recovery in the absence of drugs (trial 1). This was meant to be so, because trial 1 served as a control for trials 2, 3, and 4. To compare association to fast-inactivated state with association to slow-inactivated state, an identical protocol was applied, except that it was preceded by a long (20-s) depolarization (Fig. 6A, "S"). This protocol was performed both without (trial 2) and with drug application (trial 4). To summarize the protocol, the four trials were the following: 1) control (no drug application) with only the 3 x 100-ms pulse train of depolarization; 2) control with the 20-s depolarization right before the short train; (3) the 3 x 100-ms pulse train with concurrent drug application; and 4) the 20-s depolarization followed by the short train and concurrent drug application. Note that drug application in both trial 3 and 4 was only during the 310 ms of the short train. Peak amplitude values of the four trials are marked as FCtr, SCtr, Fdrug, and Sdrug, respectively. The depolarizing prepulse was chosen to be this long to make a significant fraction of ion channels reach slow inactivated state. The relatively long recovery interval (50 ms) was needed, because we wanted to study the ratio of drug-bound, slow-inactivated channels; during this interval, nearly 95% of fast-inactivated channels recovered, even in the presence of 30 µM drug (Fig. 5B), whereas from drug-bound, slow-inactivated channels, only 25% recovered (Fig. 5D). This interval was enough even for the majority of unbound slow-inactivated channels to recover (Fig. 6 and see below). Differences in the extent of recovery using this protocol, therefore, most sensitively reflected differences in the occupancy of the drug-bound, slow-inactivated state. (After the 20-s depolarization, the 3 x 100-ms depolarization did not cause further slow inactivation; in fact, a slight recovery from slow inactivation occurred, because during the 5-ms spent at -150 mV, more channels recovered from slow inactivation than the number of channels that slow inactivated during the 100-ms depolarization.)

    The effect of the protocol on sodium currents is illustrated in Fig. 6B; relative amplitudes of currents for control (no drug applied), fluoxetine, and desipramine are shown in Fig. 6C. All amplitude values are expressed as relative to the current evoked by the first 100-ms depolarization in trial 1.

    For fluoxetine and desipramine, drug-induced inhibition was 32.5 and 21.5%, respectively, when the 20-s prepulse was included (Sdrug/SCtr; ratio of black columns in Fig. 6C) but only 8.4 and 7.4% when the prepulse was excluded (Fdrug/FCtr; ratio of gray columns in Fig. 6C), suggesting that the slow-inactivated state was preferred by both drugs. The difference was significant at p < 0.001 in the case of 30 µM fluoxetine (n = 8) and at p < 0.01 in the case of 30 µM desipramine (n = 5) (paired t test). This indicates that the affinity of both drugs to the slow-inactivated state was higher than to the fast-inactivated state, and the difference was obvious even with this very brief (310 ms) pulse of antidepressants. (The drug application pulse could not be made longer, because we wanted to give the drug pulse to mostly depolarized channels, and longer depolarizations would have induced too much slow inactivation.) As expected, the inhibition caused by the 310-ms pulse of antidepressants caused a smaller inhibition than when drugs were continuously perfused (Fig. 5C), but a relatively small effect of the drugs without the long prepulse (trial 3) and a relatively small effect of the long prepulse without drugs (trial 2) allowed us to detect the superadditivity of the two effects when applied together (trial 4).

    Study of the mechanism of action of antidepressants on sodium channels has thus far focused on tricyclic antidepressants. A detailed analysis of the mechanism has been performed using imipramine (Ogata and Narahashi, 1989; Yang and Kuo, 2002) and amitriptyline (Nau et al., 2000; Wang et al., 2004). All these studies suppose that fast-inactivated conformation is preferred by antidepressants. In this study, we provide evidence that the principal mechanism of usedependent sodium channel inhibition by the tricyclic desipramine was a preferred binding to (and stabilization of) the slow-inactivated state. In addition, we investigated the SSRI fluoxetine, which is believed to be more selective than tricyclic compounds, and proved that it inhibited sodium channels by a similar mechanism and with the same potency as desipramine.

    In the first four sections of Results (shown in Figs. 1, 2, 3, 4, and 5), we introduced the hypothesis of stabilization of the slow-inactivated state as a possible alternative explanation. Already in this section, certain hints suggested that this hypothesis is more plausible than the conventional explanation.

    In the last section (see Fig. 6), we directly compared affinities to fastversus slow inactivated states and found that both antidepressants preferred slow inactivated conformation.

    Let us first discuss the hints from the first four sections, which argue for stabilization of the slow-inactivated state as a likely mechanism:

    Slow inactivation was clearly detectable at a -60-mV holding potential, even in the absence of the drugs (Fig. 1). Slow inactivated channels are sure to be present during drug perfusion too; the only question is whether this conformation is made more or less energetically stable by drug binding. Because the delayed development of equilibrium was not reduced but was seen even at more negative potentials, it seemed likely that antidepressants promoted, rather than hindered transition into the slow-inactivated state.

    Slow inactivation upon prolonged depolarization was detected even in control (Fig. 5C).

    Recovery curves after prolonged depolarization in this article (Fig. 5D), and in other studies, are consistently biphasic even in the absence of drugs, indicating the presence of slow inactivation.

    The presence of the drug, again, is not likely to remove slow inactivation. Therefore, in the context of the modulated receptor hypothesis, it is sensible to ask how rates of slow inactivation and recovery from slow inactivation are affected by the presence of the drugs. We suggest that altered timedependent availability plots (Fig. 5, C and D) reflect not only binding rates but also are due in part to altered gating. This concept-however logical it may be-is a novel one. Similar studies on time-dependent availability were performed using the tricyclic antidepressants amitriptyline (Nau et al., 2000; Wang et al., 2004) and imipramine (Ogata and Narahashi, 1989; Yang and Kuo, 2002); similar results were obtained, but similar conclusions have not been drawn. Assuming that the fast-inactivated state was the only one preferred by antidepressants, the authors supposed that altered rates reflect association to and dissociation from ion channels. However, if-as we propose-the slow-inactivated state is equally or even more affected by drug binding, the effects on gating transition rates also have to be taken into account.

    Slow-inactivation plots (Fig. 5C) with drugs present, therefore, reflect rates of association to both fast- and slow-inactivated states, as well as the rate of transition from fast- to slow-inactivated state. These together form the overall rate of decrease in availability. It is therefore incorrect to discuss the overall effect as a rate of association.

    Plots of recovery from slow inactivation (Fig. 5D) with drugs present reflect a combination of dissociation and gating (recovery from inactivation) rates. If recovery from one of the inactivated states is slow enough to be comparable with dissociation rates (which is more likely in the case of the slow-inactivated state), then it will significantly affect the overall rate of recovery. To discuss the overall rate of recovery as "rate of dissociation," therefore, is incorrect.

    To summarize the latter two points, results of the doublepulse protocols can be explained by binding reactions only or by the complex interaction of binding and gating rates. Gating rates are especially likely to affect overall rates if the slow-inactivated state is preferred by the drug.

    To assess the relative affinity of the two antidepressants to fast-versus slow-inactivated states, we developed a special protocol. Although the test did not allow quantitative estimation of relative affinities or association rates to different conformational states, it did demonstrate in a qualitative manner that fluoxetine and desipramine have a preference toward the slow-inactivated state.

    Although stabilization of fast inactivated state used to be the established mechanism for the action of local anesthetics and sodium channel inhibitor anticonvulsants, scattered evidence has repeatedly been found for the similar importance of the slow-inactivated state (Khodorov et al., 1976; Quandt, 1988; Chen et al., 2000; Fozzard et al., 2005). This suggests that the hypothesis of preferential affinity to the fast-inactivated state as a general explanation of use-dependent inhibition may be worth re-examination in the case of other sodium channel inhibitors. The novelty in our approach is that the preference toward fast-versus slow-inactivated states can be determined by applying a relatively simple voltage- and drug-application protocol, so the question of relative affinities can be addressed without mutagenesis experiments.

    We have previously proposed stabilization of slow inactivated state as a mechanism of use-dependent inhibition in the case of the dopamine reuptake inhibitor GBR 12909 (Mike et al., 2004). The mechanism of inhibition by antidepressants differs from this previously proposed mechanism in two important aspects: a very slow rate of dissociation (with a time constant of several minutes) was found in the case of GBR 12909, and conduction of drug-bound channels was not impaired. Inhibition seemed to be principally caused by stabilization of nonconducting states, not by steric occlusion of the pore.

    In the case of antidepressants, the dissociation was much faster, and we had no reason to suppose that drug-bound channels could conduct. Open channel block has been previously demonstrated in the case of both imipramine and amitriptyline (Yang and Kuo, 2002; Wang et al., 2004) using inactivation deficient mutants. In these studies, a concentration-dependent acceleration of the current decay was observed. This evidence, by the way, is not fully convincing; because slow inactivation of these mutant channels was not disabled, the phenomenon could be explained in part by an acceleration of the rate of slow inactivation. The extent to which the acceleration of current decay is due to open channel block versus altered rate of slow inactivation is currently unknown.

    Our results suggest that either fluoxetine binds into a hydrophobic environment or it has to cross a hydrophobic barrier to reach its binding site. Based on our results, access of the charged form of antidepressants from the intracellular side can be excluded-unlike the case of the well described local anesthetic site. This suggests that either the accessibility or the location of the binding site is different (Barber et al., 1991).

    Thus far, the significance of sodium channel inhibition by antidepressants has been thought to be the cardiac side effects of tricyclic drugs (Marshall and Forker, 1982) and their analgesic potency (Namaka et al., 2004). The fact that these drugs obviously affect neuronal sodium channels at similar (supposedly therapeutic) concentrations raises the question of whether their sodium channel inhibition effect plays a role in their antidepressant action. The fact that the SSRI fluoxetine-which has a different monoamine transporter selectivity and less G protein-coupled receptor-mediated side effects than tricyclic compounds-was found to act on sodium channels with a similar mechanism and at a similar concentration range as desipramine also suggests that this effect might be of therapeutic significance. The inhibition was found to be highly sensitive to experimental parameters (holding potential, frequency, pulse duration, etc.), as well as to extracellular pH, which suggests that the inhibition in vivo is highly selective depending on local activity patterns. The identification of the mechanism of action has special importance because this makes it possible to identify the neurons and subneuronal structures likely to be most affected, the kinds of activity patterns that are more sensitive, and the ways in which neuronal information processing is altered during antidepressant treatment. In this respect, it is important to note that the susceptibility of neuronal sodium channels to slow inactivation is an important determinant of dendritic integration (Mickus et al., 1999) and is specifically targeted by G protein-coupled receptor-mediated phosphorylation (e.g., Carr et al., 2003).

    The message of this study is that for a more thorough understanding of the short-term effects of antidepressants on the brain, direct effects (such as the currently studied modulation of sodium channels) can be as important as indirect (monoamine-mediated) modulation.

    ABBREVIATIONS: SSRI, selective serotonin-reuptake inhibitor; RSI, resting state inhibition; SDI, state-dependent inhibition; GBR 12909, vanoxerine.

【参考文献】
  Barber MJ, Starmer CF, and Grant AO (1991) Blockade of cardiac sodium channels by amitriptyline and diphenylhydantoin. Evidence for two use-dependent binding sites. Circ Res 69: 677-696.[Abstract/Free Full Text]

Bolo NR, Hode Y, Nedelec JF, Laine E, Wagner G, and Macher JP (2000) Brain pharmacokinetics and tissue distribution in vivo of fluvoxamine and fluoxetine by fluorine magnetic resonance spectroscopy. Neuropsychopharmacology 23: 428-438.[CrossRef][Medline]

Bou-Abboud E and Nattel S (1998) Molecular mechanisms of the reversal of imipramine-induced sodium channel blockade by alkalinization in human cardiac myocytes. Cardiovasc Res 38: 395-404.[Abstract/Free Full Text]

Carr DB, Day M, Cantrell AR, Held J, Scheuer T, Catterall WA, and Surmeier DJ (2003) Transmitter modulation of slow, activity-dependent alterations in sodium channel availability endows neurons with a novel form of cellular plasticity. Neuron 39: 793-806.[CrossRef][Medline]

Castren E (2005) Is mood chemistry? Nat Rev Neurosci 6: 241-246.[Medline]

Chen Z, Ong BH, Kambouris NG, Marban E, Tomaselli GF, and Balser JR (2000) Lidocaine induces a slow inactivated state in rat skeletal muscle sodium channels. J Physiol 524 Pt 1:37-49.[Abstract/Free Full Text]

Choi JS, Choi BH, Ahn HS, Kim MJ, Han TH, Rhie DJ, Yoon SH, Jo YH, Kim MS, and Hahn SJ (2004) Fluoxetine inhibits A-type potassium currents in primary cultured rat hippocampal neurons. Brain Res 1018: 201-207.[CrossRef][Medline]

Deak F, Lasztoczi B, Pacher P, Petheo GL, Valeria K, and Spat A (2000) Inhibition of voltage-gated calcium channels by fluoxetine in rat hippocampal pyramidal cells. Neuropharmacology 39: 1029-1036.[CrossRef][Medline]

Deffois A, Fage D, and Carter C (1996) Inhibition of synaptosomal veratridineinduced sodium influx by antidepressants and neuroleptics used in chronic pain. Neurosci Lett 220: 117-120.[CrossRef][Medline]

Eisensamer B, Rammes G, Gimpl G, Shapa M, Ferrari U, Hapfelmeier G, Bondy B, Parsons C, Gilling K, Zieglgansberger W, et al. (2003) Antidepressants are functional antagonists at the serotonin type 3 (5-HT3) receptor. Mol Psychiatry 8: 994-1007.[CrossRef][Medline]

Fozzard HA, Lee PJ, and Lipkind GM (2005) Mechanism of local anesthetic drug action on voltage-gated sodium channels. Curr Pharm Des 11: 2671-2686.[CrossRef][Medline]

Gumilar F, Arias HR, Spitzmaul G, and Bouzat C (2003) Molecular mechanisms of inhibition of nicotinic acetylcholine receptors by tricyclic antidepressants. Neuropharmacology 45: 964-976.[CrossRef][Medline]

Hamill OP, Marty A, Neher E, Sakmann B, and Sigworth FJ (1981) Improved patch-clamp techniques for high-resolution current recording from cells and cellfree membrane patches. Pflueg Arch Eur J Physiol 391: 85-100.[CrossRef][Medline]

Hennings EC, Kiss JP, De Oliveira K, Toth PT, and Vizi ES (1999) Nicotinic acetylcholine receptor antagonistic activity of monoamine uptake blockers in rat hippocampal slices. J Neurochem 73: 1043-1050.[CrossRef][Medline]

Hille B (1977) Local anesthetics: hydrophilic and hydrophobic pathways for the drug-receptor reaction. J Gen Physiol 69: 497-515.[Abstract/Free Full Text]

Kambouris NG, Hastings LA, Stepanovic S, Marban E, Tomaselli GF, and Balser JR (1998) Mechanistic link between lidocaine block and inactivation probed by outer pore mutations in the rat μ1 skeletal muscle sodium channel. J Physiol 512: 693-705.[Abstract/Free Full Text]

Khodorov BI, Shiskova L, Peganov E, and Revenko S (1976) Inhibition of sodium currents in frog ranvier node treated with local anesthetics. Role of slow sodium inactivation. Biochim Biophys Acta 433: 409-435.

Kuo CC and Bean BP (1994) Slow binding of phenytoin to inactivated sodium channels in rat hippocampal neurons. Mol Pharmacol 46: 716-725.[Abstract]

Kuo CC, Huang RC, and Lou BS (2000) Inhibition of Na+ current by diphenhydramine and other diphenyl compounds: molecular determinants of selective binding to the inactivated channels. Mol Pharmacol 57: 135-143.[Abstract/Free Full Text]

Marshall JB and Forker AD (1982) Cardiovascular effects of tricyclic antidepressant drugs: therapeutic usage, overdose, and management of complications. Am Heart J 103: 401-414.[CrossRef][Medline]

McNeal ET, Lewandowski GA, Daly JW, and Creveling CR (1985) [3H]Batrachotoxinin A 20 alpha-benzoate binding to voltage-sensitive sodium channels: a rapid and quantitative assay for local anesthetic activity in a variety of drugs. J Med Chem 28: 381-388.[CrossRef][Medline]

Mickus T, Jung H, and Spruston N (1999) Properties of slow, cumulative sodium channel inactivation in rat hippocampal CA1 pyramidal neurons. Biophys J 76: 846-860.[Medline]

Mike A, Karoly R, Vizi ES, and Kiss JP (2003) Inhibitory effect of the DA uptake blocker GBR 12,909 on sodium channels of hippocampal neurons. Neuroreport 14: 1945-1949.[CrossRef][Medline]

Mike A, Karoly R, Vizi ES, and Kiss JP (2004) A novel modulatory mechanism of sodium currents: frequency-dependence without state-dependent binding. Neuroscience 125: 1019-1028.[CrossRef][Medline]

Mike A, Pereira EF, and Albuquerque EX (2000) Ca2+-sensitive inhibition by Pb2+ of alpha7-containing nicotinic acetylcholine receptors in hippocampal neurons. Brain Res 873: 112-123.[CrossRef][Medline]

Muscettola G, Goodwin FK, Potter WZ, Claeys MM, and Markey SP (1978) Imipramine and desipramine in plasma and spinal fluid: relationship to clinical response and serotonin metabolism. Arch Gen Psychiatry 35: 621-625.[Abstract/Free Full Text]

Namaka M, Gramlich CR, Ruhlen D, Melanson M, Sutton I, and Major J (2004) A treatment algorithm for neuropathic pain. Clin Ther 26: 951-979.[CrossRef][Medline]

Nau C, Seaver M, Wang SY, and Wang GK (2000) Block of human heart hH1 sodium channels by amitriptyline. J Pharmacol Exp Ther 292: 1015-1023.[Abstract/Free Full Text]

Nicholson GM, Blanche T, Mansfield K, and Tran Y (2002) Differential blockade of neuronal voltage-gated Na+ and K+ channels by antidepressant drugs. Eur J Pharmacol 452: 35-48.[CrossRef][Medline]

Ogata N and Narahashi T (1989) Block of sodium channels by psychotropic drugs in single guinea-pig cardiac myocytes. Br J Pharmacol 97: 905-913.[Medline]

Pacher P, Magyar J, Szigligeti P, Banyasz T, Pankucsi C, Korom Z, Ungvari Z, Kecskemeti V, and Nanasi PP (2000) Electrophysiological effects of fluoxetine in mammalian cardiac tissues. Naunyn-Schmiedeberg's Arch Pharmacol 361: 67-73.[CrossRef][Medline]

Pancrazio JJ, Kamatchi GL, Roscoe AK, and Lynch C, 3rd (1998) Inhibition of neuronal Na+ channels by antidepressant drugs. J Pharmacol Exp Ther 284: 208-214.[Abstract/Free Full Text]

Quandt FN (1988) Modification of slow inactivation of single sodium channels by phenytoin in neuroblastoma cells. Mol Pharmacol 34: 557-565.[Abstract]

Sernagor E, Kuhn D, Vyklicky L Jr, and Mayer ML (1989) Open channel block of NMDA receptor responses evoked by tricyclic antidepressants. Neuron 2: 1221-1227.[CrossRef][Medline]

Shytle RD, Silver AA, Lukas RJ, Newman MB, Sheehan DV, and Sanberg PR (2002) Nicotinic acetylcholine receptors as targets for antidepressants. Mol Psychiatry 7: 525-535.[CrossRef][Medline]

Starmer CF, Grant AO, and Strauss HC (1984) Mechanisms of use-dependent block of sodium channels in excitable membranes by local anesthetics. Biophys J 46: 15-27.[Medline]

Torres GE, Gainetdinov RR, and Caron MG (2003) Plasma membrane monoamine transporters: structure, regulation and function. Nat Rev Neurosci 4: 13-25.[CrossRef][Medline]

Wang GK, Russell C, and Wang SY (2004) State-dependent block of voltage-gated Na+ channels by amitriptyline via the local anesthetic receptor and its implication for neuropathic pain. Pain 110: 166-174.[CrossRef][Medline]

Woodhull AM (1973) Ionic blockage of sodium channels in nerve. The J Gen Physiol 61: 687-708.[CrossRef]

Yang YC and Kuo CC (2002) Inhibition of Na+ current by imipramine and related compounds: different binding kinetics as an inactivation stabilizer and as an open channel blocker. Mol Pharmacol 62: 1228-1237.[Abstract/Free Full Text]

作者单位：Department of Pharmacology, Institute of Experimental Medicine, Hungarian Academy of Sciences, Budapest, Hungary

【关键词】  nigriventer Toxin

    A toxin was purified to homogeneity from the venom of the South American armed spider Phoneutria nigriventer and found to have a molecular mass of 8600 Da and a C-terminally amidated glycine residue. It appears to be identical to Toxin 1 (Tx1) isolated previously from this venom. Tx1 reversibly inhibited sodium currents in Chinese hamster ovary cells expressing recombinant sodium (Nav1.2) channels without affecting their fast biophysical properties. The kinetics of inhibition of peak sodium current varied with membrane potential, with on-rates increasing and off-rates decreasing with more depolarized holding potentials in the –100 to –50 mV range. Thus, the apparent affinity of Tx1 for the channel increases as the membrane is depolarized. A mono[125I]iodo-Tx1 derivative displayed high-affinity binding to a single class of sites (KD = 80 pM, Bmax = 0.43 pmol/mg protein) in rat brain membranes. Solubilized binding sites were immunoprecipitated by antibodies directed against a conserved motif in sodium channel  subunits. 125I-Tx1 binding was competitively displaced by µ conotoxin GIIIB (IC50 = 0.5 µM) but not by 1 µM tetrodotoxin. However, the inhibition of 125I-Tx1 binding by µ conotoxin GIIIB was abrogated in the presence of tetrodotoxin (1 µM). Patch-clamp and binding data indicate that P. nigriventer Tx1 is a novel, state-dependent sodium-channel blocker that binds to a site in proximity to pharmacological site 1, overlapping µ conotoxin but not tetrodotoxin binding sites.Voltage-gated sodium channels underlie the rapid depolarizing phase of the action potential and play a crucial role in the propagation of electrical signals in neurons and in cardiac and skeletal muscle. Sodium channels consist of a poreforming  subunit associated with either one or two 1–4 auxiliary subunits. Mammals express at least 10 sodium-channel genes, and the coexpression of different combinations can confer distinct electrical properties on neurons and myocytes (Catterall et al., 2003). Sodium channels constitute a major target for therapeutic drugs used in the treatment of epilepsy, pain, and cardiac arrhythmias. Many naturally occurring neurotoxins also bind to sodium channels and modify their properties (Cestèle and Catterall, 2000; Barbier et al., 2003; Li and Tomaselli, 2004; Terlau and Olivera, 2004). Their specificity and high affinity make them ideal probes for defining the different pharmacological sites located on the channel, dissecting the allosteric interactions between these sites and mapping structure-function relationships. Five distinct pharmacological sites have been mapped within the sodium channel sequence (Cestèle and Catterall, 2000). Sites 1 to 5 are typically defined with tetrodotoxin, veratridine,  (Old World) scorpion toxins,  (New World) scorpion toxins, and brevetoxin, respectively.

    Spider venoms are a rich source of agents that target voltage-gated ion channels, and we thus examined the pharmacological action of peptides from the venom of the South American spider Phoneutria nigriventer. P. nigriventer, known as the "armed" spider, causes most of the severe human envenomations by spider bite in Southeast Brazil (Gomez et al., 2002). Experiments with whole venom and partially characterized fractions have suggested that toxicity principally involves voltage-gated sodium currents (Fontana and Vital-Brazil, 1985; Love and Cruz-Höfling, 1986; Araùjo et al., 1993).

    We purified P. nigriventer Tx1 and evaluated its mode of action by patch-clamp and binding studies. Our results indicate that Tx1 is a novel inhibitor of neuronal sodium channels that binds in proximity to site 1 and displays increasing affinity as the membrane potential is depolarized.

    Purification and Biochemical Characterization of Tx1. Wild P. nigriventer were collected in the area of Santa Barbara, near Belo Horizonte, MG, Brazil, and kept at the Fundação Ezequiel Dias, Belo Horizonte. Venom was obtained by electrical stimulation of anesthetized spiders. Tx1 was first purified from fractions lethal in mice using the method of Diniz et al. (1990), producing a fraction (designated sample R21) that was lyophilized. Because sample R21 proved to contain two peptides (see Results), it was submitted to an additional HPLC step using a reverse-phase Vydac C18 (218TP54; Grace Vydac, Hesperia, CA) analytic column (0.46 x 25 cm) eluted with a linear gradient of acetonitrile (0–60% for 60 min) in 0.1% trifluoroacetic acid. Purity was assessed by MALDI-TOF mass spectrometry and sequencing. This fraction was designated Tx1 sample R41.

    N-terminal sequencing was performed by Edman degradation using a 476A automatic pulsed liquid sequencer (Applied Biosystems, Foster City, CA). Amino acid analyses were carried out on a 6300 Beckman Analyzer (Beckman Coulter, Fullerton, CA). For peptide mass fingerprinting, 8.6 nmol concentration of native Tx1 were reduced with dithiothreitol (60-fold excess over cysteine, 20 h, 40°C, under N2) in 0.25 M Tris/6 M guanidine buffer and alkylated with iodoacetamide (1.5-fold excess over cysteine, 1 h at 25°C). The reaction was stopped with citric acid, and then alkylated Tx1 was desalted by HPLC. Complete alkylation was assessed by MALDI-TOF MS. Alkylated Tx1 (1.46 nmol) was digested with Lys-C protease [final buffer: 25 mM Tris-HCl, 1 mM EDTA, and 10% enzyme/protein (w/w), 22 h, 37°C]. The solution was acidified and submitted to C18 HPLC. Each peak was collected for mass measurement and/or sequencing.

    Cell Culture. The stable CHO line CNaIIA (West et al., 1992) kindly provided by Professor W. A. Catterall (Seattle, WA), expressing rat brain Nav1.2, was cultured in RPMI 1640 medium supplemented with 5% fetal calf serum, 2 mM glutamine, 10 mM glucose, penicillin (100 U/ml), streptomycin (0.1 g/l), and geneticin (0.2 g/l) at 37°C in a 5% CO2 atmosphere.

    Electrophysiology. Whole-cell currents were recorded from isolated cells using wide-tipped patch pipettes (1.0–1.8 M). pClamp6 and -8 (Molecular Devices, Sunnyvale, CA) and Sigma Plot (SPSS Inc., Chicago, IL) softwares were used for experimental protocols and analysis. Whole-cell currents (inward current traces downward) were digitized (40–100 kHz) after low-pass filtering (5 kHz). The voltage stimulus used to elicit sodium currents is indicated in the figure legends. A voltage protocol used for leak subtraction was applied after each stimulus [P/n protocol from the holding potential (EH) with the same polarity as stimulus, n = 3 and 5 in Figs. 1 and 2, respectively). The peak sodium current value (peak INa) was measured from the zero current level in leak-subtracted traces and plotted against the episode starting time. The pipette solution contained 135 mM CsCl, 1 mM MgCl2, 10 mM HEPES, and 10 mM BAPTA, with pH adjusted to 7.2 with CsOH. The extracellular solution contained 150 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, 15 mM glucose, and 10 mM HEPES, with pH adjusted to 7.4 with NaOH. Tx1 or tetrodotoxin was added manually near the cell as a single dose left to diffuse for the time indicated by the bars. In experiments shown in Fig. 2, before and after toxin applications, a continuous extracellular solution flow was applied to the cell using a local superfusion system (70–100 µl/min).

    Fig. 1. Effect of P. nigriventer toxin 1 on sodium currents recorded from a CHO cell expressing recombinant Nav1.2 channels. A, superimposed current traces recorded before and after the addition of Tx1 (10 µl, 1 µM) near the cell. A 5-ms step depolarization from –90 to 10 mV was applied every 6 s. The illustrated traces are separated by 60 s, and each trace is the average of leak subtracted currents recorded in 10 consecutive tests. B, plot of the peak sodium current (peak INa) against test time in the absence and in the presence of Tx1 ().

    Fig. 2. Effect of the holding potential on the peak current reduction by P. nigriventer toxin 1. Sodium currents from CHO cells expressing Nav1.2 were elicited by test pulses (2 ms to 0 mV, 0.1 Hz) in the absence and in the presence of Tx1 (10 µl, 1 µM; ) or tetrodotoxin (TTX, 10 µl, 10 µM; ), and peak INa was plotted against time. Two intertest holding potentials were used. At EH = –50 mV, test pulses were preceded by a deinactivating prepulse (200 ms to –100 mV). A, EH was shifted from –100 to –50 mV for 420 s in the presence of Tx1 and then back to –100 mV 180 s before washout. The small (0.2 nA) reduction observed between times 2 and 3 min was due to the arrest of perfusion 0.6 min before the addition of Tx1. B, EH was shifted from –100 to –50 mV in the absence of toxin. Then Tx1 was applied for 100 s and washed out at EH =–50 mV.

    125I-Tx1 Binding Experiments. Tx1 was radioiodinated as described for  P. nigriventer toxin IIA (Gouvêa dos Santos et al., 2002), with minor modifications. Tx1 (0.2 nmol) was reacted for 2 min with 0.2 nmol of carrier-free sodium 125I (GE Healthcare, Little Chalfont, Buckinghamshire, UK) in the presence of lactoperoxidase and H2O2 in 50 mM phosphate buffer at pH 7.2. Monoiodotoxins were separated by reverse-phase liquid chromatography on a C18 column (Vydac; 0.46 x 25 cm, 5-µm particles). The column was eluted with a linear gradient of 19 to 27% acetonitrile in 0.1% trifluoroacetic acid for 65 min at a flow rate of 1 ml/min. Native toxin was eluted after 27 min (22% acetonitrile), and two radioactive peaks corresponding to the two possible monoiodo-Tx1 derivatives were detected at 28 and 31 min. The second peak was used in binding assays. The binding of 125I-Tx1 to rat brain synaptosomal membranes (prepared as described in Martin-Moutot et al., 1995) was measured at 30°C in 200 µl of binding buffer: 140 mM NaCl, 10 mM Tris base, and 0.1% bovine serum albumin, adjusted to pH 7.4. Binding was initiated by the addition of 5 µg of membrane protein, stopped by rapid filtration on glass fiber filters (GF/C; Whatman, Clifton, NJ) pretreated with 0.3% polyethyleneimine, and washed three times with 2 ml of ice-cold binding buffer. Bound ligand was measured by -counting. Tetrodotoxin was purchased from Latoxan (Valence, France); veratridine, brevetoxin 2, bupivacaine, lidocaine, tetracaine, and pompilidotoxin were from Sigma (St. Louis, MO); and µ conotoxin GIIIB was from the Peptide Institute (Osaka, Japan). Deltamethrin was from Calbiochem (San Diego, CA), and Leiurus quinquestriatus quinquestriatus toxin V and Centruroides suffusus suffusus toxin VI were kindly supplied by Dr. M. F. Martin-Eauclaire (Marseille, France).

    Immunoprecipitation. Synaptosomal membranes (2 mg) were incubated with 0.2 nM 125I-Tx1 overnight at 4°C in 6 ml of binding buffer, washed by centrifugation, and solubilized in 1.5% Triton X-100 in 300 µl of solubilization buffer (10 mM HEPES and 0.1 M KCl, pH 7.4, containing Sigma protease inhibitor cocktail). After centrifugation at 100,000g, aliquots of supernatant containing 8 fmol of 125I-Tx1 were incubated with antibodies in a final assay volume of 50 µl. After 4 h at 4°C, immune complexes were recovered by mixing for 30 min with Protein A-Sepharose CL4B. After centrifugation, the pellet was washed with 0.5% Triton X-100 and 0.1% bovine serum albumin in solubilization buffer, and the immunoprecipitated radioactivity was counted.

    Purification of Tx1. Tx1 was purified (sample R21) from P. nigriventer venom using a method published previously (Diniz et al., 1990). Two molecular masses were detected at 8600 and 4920 Da in this fraction, and two sequences were assigned by N-terminal Edman sequencing. Whereas the major component (AELTSXFPVG) seemed to be the N-terminal sequence of Tx1 (Diniz et al., 1990), the minor one, accounting for approximately 5% (XGXAQAYKS), showed a clear homology to the N-terminal segment of the Tx3-3 neurotoxin (GCANAYKS; Cordeiro et al., 1993), if we assume that Xs are half cystines. A supplementary chromatographic step was thus introduced to isolate the two peptides. The contaminating Tx3-3-like peptide (4920 Da) had no effect on sodium currents (data not shown). However, these results suggest that previous reports concerning the pharmacology of Tx1 may involve the effects of a contaminating peptide (see Discussion).

    The homogeneity of the purified major component (sample R41) was assessed by MALDI-TOF MS (8600.4 ± 1.6 Da) and N-terminal sequencing (AELTSXFPVG, without any detectable contamination i.e., less than 0.5%). We found that 8600 Da was significantly higher than the 8557 Da calculated from the published Tx1 sequence (Diniz et al., 1990). However a discrepancy has been reported between the C-terminal sequence of Tx1 determined by peptide sequencing (Diniz et al., 1990) and that predicted by the cDNA (Diniz et al., 1993), -RREC versus -RRNCGG, respectively. To confirm that sample R41 was identical with Tx1, the peptide MS fingerprint was established after Lys-C protease digestion (Table 1). Alkylated peptides yielded masses in good agreement with those predicted, except for the C-terminal peptide. This peptide was identified as KPCRRNCG, which is the 71 to 78 sequence predicted by the cDNA (Diniz et al., 1993) but with a single terminal glycine residue and C-terminal amidation. The corresponding mass predicted for Tx1 (8598.83 Da) is in good agreement with the observed value.

    TABLE 1 Peptide fingerprint of P. nigriventer toxin 1

    Tx1 was purified to homogeneity, reduced, alkylated, and digested with Lys-C protease. Peptides were separated by C18 HPLC, and peaks were collected for mass spectrometry. The masses measured for three proteolytic peptides (4633.78, 1374.56, and 1491.63 Da) were within 100 ppm of the masses predicted for Tx1 proteolytic peptides 1 to 39, 42 to 53, and 56 to 67, as indicated above the published P. nigriventer toxin 1 sequence (Diniz et al., 1993; accession number P17727). The last mass measured (1046.31 Da) corresponds (within 192 ppm) to a C-terminal peptide with N at position 76 and a single, amidated terminal glycine.

    Initial electrophysiological experiments were performed using Tx1 sample R21 with the contaminating peptide (4920 Da). The 8600-Da Tx1 component was subsequently purified to homogeneity (sample R41). Tx1 samples R21 and R41 were found to have identical effects on sodium currents; thus, no distinction was made between these two preparations in our patch-clamp studies.

    P. nigriventer Tx1 Inhibits Voltage-Gated Sodium Currents. Voltage-gated sodium currents were recorded in CHO cells expressing Nav1.2 channels using the whole-cell configuration. Repetitive step depolarizations evoked transient inward currents that were reduced by the addition of Tx1 (10 µl, 1 µM) near the cell (Fig. 1A), yielding approximately 85% inhibition of sodium current within 10 min (Fig. 1, A and B).

    Tx1 did not modify the fast biophysical properties of Nav1.2 channels. First, the kinetics of inward current rise and fall during a 2-ms step depolarization were not modified after application of Tx1, although inhibition of peak INa developed from 0 to 85%, suggesting that Tx1 did not alter activation and fast inactivation. Second, conductance-voltage and inactivation characteristics were established using standard voltage protocols in the 1- to 100-ms range (Sarkar et al., 1995) from a holding potential of –100 mV. Each test peak INa value was normalized to a reference peak INa measured during a pulse (2 ms to 0 mV) applied a few milliseconds or tens of milliseconds before each conditioning pulse/test pulse pair. Experiments done in the absence of toxin or in the presence of 100 nM Tx1 in conditions leaving 20 to 40% residual current showed that conductance voltage and fast inactivation characteristics were unaffected in the presence of Tx1 (data not shown). Finally, in addition, complete inhibition was observed at high Tx1 concentrations in favorable conditions (see below). These results suggest that in the presence of a nonsaturating concentration of Tx1, individual channels either function with unmodified fast biophysical properties or are fully blocked.

    The experiments illustrated in Fig. 2 were performed to assess the effects of the holding potential on inhibition of sodium currents by Tx1. To measure a pool of functional channels, at holding potentials positive to –100 mV, a deinactivating prepulse of 200 ms to –100 mV was used to reverse the fast inactivation of sodium channels before the test. Under these conditions, in the absence of Tx1, a slow decrease of peak INa was observed corresponding to a slow voltage-dependent inactivation of Nav1.2 channels (Fig. 2B before Tx1 application). Switching the holding potential from –100 to –50 mV in the absence of toxin resulted in a 15% decrease in peak INa within 5 min. In Fig. 2A, currents were evoked initially from a holding potential of –100 mV. In these conditions, the addition of Tx1 had no effect within 5 min. However, shifting the holding potential to –50 mV in the presence of Tx1 rapidly produced inhibition, reaching >85% reduction within 7 min. Inhibition was reversed by returning to a holding potential of –100 mV in the presence of Tx1, and toxin washout at –100 mV had no effect on the kinetics of peak INa recovery (slow off = 201 s before, off = 238 s after washout). Tetrodotoxin produced full, reversible inhibition when applied at –100 mV. In Fig. 2B, Tx1 (10 µl, 1 µM) was applied at a holding potential of –50 mV, inducing a 92% reduction of peak INa within 110 s. After washout of Tx1, inhibition was only slowly reversible at –50 mV (off > 35 min), but recovery accelerated significantly when the holding potential was hyperpolarized to –100 mV (off = 172 s). These results, added to those of experiments at EH intermediate between –100 and –50 mV (data not shown), are consistent with the view that depolarizing the holding potential both increases the on-rate and decreases the off-rate for sodium-channel inhibition (i.e., the Nav1.2 sodium channel displays relatively higher apparent affinity for Tx1 at more depolarized membrane potentials).

    125I-Tx1 Binding Properties. Results of an equilibrium binding experiment demonstrating saturable binding of Tx1 to synaptosomal brain membranes are shown in Fig. 3A. Increasing concentrations of 125I-Tx1 were added to membranes in the absence (total binding) or presence of a large excess of unlabeled Tx1 (nonspecific binding). The difference between these curves yields the saturable binding component. A Scatchard plot of specific binding (Fig. 3A, inset) shows a single class of sites with KD= 80 pM and Bmax = 0.43 pmol/mg cell protein. The kinetics of 125I-Tx1 association and dissociation from membrane binding sites are shown in Fig. 3, B and C. The slopes of the linear semilogarithmic plots (Fig. 3C, inset) gives the apparent association rate constant (kapp) =–43 x 10–4/s and the dissociation rate constant (k–1) = 11 x 10–4/s. The equation k+1 = (kapp – k–1)/[125I-Tx1] yields the association rate constant k+1 = 32 x 106 M–1/s. The equilibrium dissociation constant KD calculated from the kinetic data (KD = k–1/k+1) was 33 pM.

    Fig. 3. Binding of 125I-tagged P. nigriventer toxin 1 to rat brain membranes. A, rat brain membranes were incubated with the indicated concentrations of 125I-Tx1 for 30 min at 30°C in the absence (, total binding) or presence (, nonspecific binding) of 50 nM unlabeled Tx1 and then filtered and washed. Bound toxin was measured by  counting. Inset, Scatchard plot of the specific binding component (total minus nonspecific). B, rat brain membranes were incubated with 0.1 nM 125I-Tx1 in the presence and absence of unlabeled Tx1. Incubations were stopped by filtration at the indicated time points, and the kinetics of specific binding were plotted. Inset, linear semilogarithmic plot. C, rat brain membranes were incubated with 0.1 nM 125I-Tx1 for 30 min at 30°C. At equilibrium, 50 nM unlabeled Tx1 was added, and the residual specific binding component was measured by filtration at the indicated times to yield dissociation kinetics. Inset, linear semilogarithmic plot.

    Because a previous report suggested that Tx1 acts on N-type (Cav2.2) calcium channels (Gouvêa dos Santos et al., 1999), immunoprecipitation experiments were performed with solubilized 125I-Tx1-labeled binding sites and antibodies directed against conserved sequences in the Nav1 or Cav2 families of proteins (pan anti-Nav, antibody directed against a conserved sequence in domain III-IV of the  subunit and an antibody against a conserved sequence in the N-terminal domain of Cav2). The results (data not shown) indicated that anti-Nav antibodies recognized the 125I-Tx1/channel complex and that immunoprecipitation was blocked by the cognate peptide. In contrast, anti-Cav2 antibodies that recognize N-type channels only precipitated background amounts of 125I-Tx1/channel complexes. These results are consistent with our patch-clamp data and indicate that Tx1 binding sites are associated with sodium channels.

    Competition experiments were performed to determine whether Tx1 interacts with any of the defined pharmacological sites on sodium channels. At the indicated concentrations, none of the drugs or toxins active at sodium channel sites 2, 3, 4, or 5 (1 µM veratridine, 0.2 µM Leiurus quinquestriatus quinquestriatus toxin V and Centruroides suffusus suffusus toxin VI, and 0.1 µM brevetoxin, respectively) nor local anesthetics (bupivacaine, lidocaine, and tetracaine, 10 µM), the wasp venom peptide pompilidotoxin (0.1 µM), nor the pyrethroid deltamethrin (10 µM) significantly modified 125I-Tx1 binding. In contrast, although 1 µM tetrodotoxin had no effect, µ conotoxin GIIIB (3 µM), a peptide active at site 1 (Barbier et al., 2003; Li and Tomaselli, 2004; Terlau and Olivera, 2004) displaced 75% of specific 125I-Tx1 binding. Assays performed to determine the concentration dependence for displacement (Fig. 4A) gave an IC50 value of 300 pM for native Tx1, yielding a calculated Ki value of 75 pM, and showed that radioiodination does not affect the affinity of Tx1. µ Conotoxin GIIIB displaced 50% of 125I-Tx1 binding at 0.5 µM. µ Conotoxin GIIIB and tetrodotoxin are known to display binding interactions; thus, there is an apparent discrepancy in the fact that 125I-Tx1 binding was displaced by µ conotoxin GIIIB but not by tetrodotoxin. Experiments designed to address this issue (Fig. 4A, inset), indicated that 1 µM tetrodotoxin completely reverses inhibition of 125I-Tx1 binding by 0.3 µM µ conotoxin GIIIB. Thus, 125I-Tx1 and tetrodotoxin bind to two distinct sites, but µ conotoxin GIIIB overlaps both.

    Fig. 4. Interactions between 125I-tagged P. nigriventer toxin 1 and µ conotoxin binding sites in rat brain membranes. A, 125I-Tx1 (0.1 nM) was incubated with rat brain membranes for 30 min at 30°C in the presence of the indicated concentrations of unlabeled Tx1 () or µ conotoxin GIIIB (). Specific binding was measured by filtration. Inset, specific binding of 125I-Tx1 as indicated above (C = control) was measured in the presence of 1 µM tetrodotoxin (TTX) or 0.3 µM µ conotoxin GIIIB alone (µ) or in the presence of 1 µM tetrodotoxin (µ + TTX). B, saturation curves were established as in Fig. 3A in the presence (open symbols) or absence (filled symbols) of 50 nM unlabeled Tx1, and in the presence () or absence () of 3 µM µ-conotoxin GIIIB. Inset, Scatchard plot of specific binding in the presence () or absence () of µ conotoxin GIIIB.

    To assess the nature of the interaction between the two ligands, 125I-Tx1 saturation curves were determined in the presence or absence of µ conotoxin GIIIB (3 µM; Fig. 4B). Scatchard plots (Fig. 4B, inset) show a single class of sites in both conditions with KD = 30 pM, Bmax = 520 fmol/mg protein in the absence of µ conotoxin GIIIB, and KD = 140 pM and Bmax = 520 fmol/mg in the presence of µ conotoxin GIIIB. Inhibition of binding in the presence of µ conotoxin GIIIB is essentially due to an apparent decrease in affinity, with no significant modification of binding site capacity. These data are consistent with competitive interaction between Tx1 and µ conotoxin GIIIB.

    Tx1 was initially purified and sequenced in 1990 (Diniz et al., 1990). Intracerebroventricular injection of this toxin in mice was reported to produce behavioral excitation and spastic paralysis. Our analysis of fractions prepared using the original protocol indicated a major component of 8600 Da and a minor contaminant of 4920 Da. Peptide fingerprinting, mass spectrometry, and amino acid sequencing have unequivocally identified the 8600-Da peptide as Tx1. Tx1 differed from the sequence predicted by the cDNA (Diniz et al., 1993) only at the C terminus, and our data are consistent with post-translational maturation of Tx1, which involves carboxypeptidase cleavage of a glycine residue and -amidation.

    Tx1 was purified to homogeneity, and our data indicate that this toxin is an inhibitor of neuronal voltage-gated sodium channels. Previous reports suggested that Tx1 targets calcium channels, mainly based on sequence homologies to agatoxins and on its ability to inhibit the binding of 125I- conotoxin GVIA, a ligand specific for Cav2.2 (N-type) channels (Gouvêa dos Santos et al., 1999). Our results suggest that the contaminating 4920-Da peptide in the initial Tx1 sample R21 could be toxin Tx3-3 based on its partial N-terminal sequence. The Tx3 fraction includes several toxins that are antagonists of high-voltage-activated calcium channels with a preference for Cav2 channels (Cassola et al., 1998; Leão et al., 2000; Gouvêa dos Santos et al., 2002; Vieira et al., 2005). Thus, the contaminating peptide may account for Tx1 initially being designated inappropriately as a calcium channel blocker.

    Whole-cell patch-clamp recording from CHO cells expressing recombinant Nav1.2 channels indicate that Tx1 inhibits neuronal voltage-gated sodium channels in a reversible manner. Its effect resembles the tetrodotoxin effect, in that it can induce full inhibition and does not affect the fast biophysical properties of the residual current in nonsaturating conditions. These results suggest that in conditions of incomplete inhibition by Tx1, individual channels are either functional with unmodified properties or fully blocked. The simplest explanation would be that channels with bound Tx1 become nonfunctional. These properties contrasts with those reported for two P. nigriventer venom peptides shown to target sodium channel site 3 and slow down sodium channel inactivation in frog muscle (PnTx2-6; Matavel et al., 2002) and insect axon [Tx4(6-1); De Lima et al., 2002], respectively, and suggest that Tx1 may strongly contribute to lethality in human envenomation.

    The kinetics of peak sodium current inhibition by Tx1 were voltage-dependent, with on-rates increasing and off-rates decreasing with more depolarized holding potentials in the –100 to –50 mV range, corresponding to an apparent increase in the affinity of Tx1 for sodium channels at more depolarized membrane potentials. Whether the voltage-dependent inhibition is due to voltage effect on Tx1 or the channel may be of question. A number of points do not support the hypothesis that the voltage-dependence is due to an effect of the electric field on the toxin molecule: 1) a 78-amino acid peptide with a molecular mass of 8600 Da is not likely to enter the channel pore; 2) the net charge estimated for Tx1 at pH 7.4 from the published amino acid sequence (Diniz et al., 1990; Cordeiro et al., 1993) and our data concerning the C terminus (N76 and C-terminal amidation) is positive (+4.6); a positively charged open-channel blocker would undergo a relief of block with depolarization contrary to what is observed; and 3) the idea that a negatively charged arm of the toxin molecule would behave as an open channel blocker is hardly compatible with the very slow on and off kinetics observed. Therefore, it is likely that the voltage-dependence of inhibition by Tx1 is due to voltage-dependent conformational changes of the sodium channel, making Tx1 a channel state-dependent inhibitor. Tx1 had no effect on sodium currents when applied at a hyperpolarized holding potential (–100 mV) in experiments with brief (2 ms), infrequent (0.1 Hz) depolarizing test pulses. This suggests that the deactivated, deinactivated sodium channel is not a target for Tx1 and that Tx1 rather binds one or several of the channel states reached through membrane depolarization.

    Tetrodotoxin and saxitoxin are known to induce use-dependent block of neuronal sodium channels due to channel state-dependent binding (Lönnendonker, 1989a,b; Patton and Goldin, 1991). Using a combination of nonstationary fluctuation analysis and use-dependent block analysis, Lönnendonker (1989a) showed that Ranvier node sodium-channel affinity for tetrodotoxin and saxitoxin is independent of the holding potential when channels are stimulated at 1 Hz, whereas unstimulated channels may have a lower affinity at more negative holding potentials. In the case of Tx1, the very slow kinetics of on and off effects on neuronal sodium currents allows direct demonstration of a change in apparent affinity at various holding potentials. Slow kinetics in the case of Tx1 may be an advantage for further mechanistic studies.

    Slow voltage-dependence has not been studied for Huwentoxin-IV (Peng et al., 2002), the only other spider toxin known to block tetrodotoxin-sensitive sodium channels. However, properties similar to those of Tx1 may be expected for the uncharacterized toxins 1A and 1B from Phoneutria keyserlingi, which display 95 and 94% sequence identity, respectively, to Tx1 (Table 2). The next nearest sequences found in databanks (34–38% identity to Tx1) correspond to spider toxins known to inhibit voltage-gated calcium channels (variants of -phonetoxins from P. nigriventer and -agatoxin III from Agelenopsis aperta).

    TABLE 2 Sequence similarities between Phoneutria toxins 1 and µ conotoxins

    The sequence of Tx1 was initially aligned to that of Conus stercusmuscarum, µ conotoxin SmIIIA, using the LALIGN programme (Huang and Miller, 1991) and then with µ conotoxins SIIIA from C. striatus, KIIIA from C. kinoshitai, PIIIA from C. purpurascens, GIIIA and GIIIB from C. geographus using Table 1 in Bulaj et al. (2005), and P. keyserlingi toxins 1A and 1B (PkTx1A,B, accession numbers P84062 and P84063). Numbers on the left indicate the position of the illustrated motif in each sequence. Bold letters indicate the identity between toxins 1 and µ conotoxins, and bold and underlined letters indicate conserved basic residues with demonstrated functional importance in GIIIA, numbered below according to their position in this toxin.

    A 125I-Tx1 derivative displayed high-affinity binding to a single class of sites in rat brain membranes. In keeping with the electrophysiological data, two lines of evidence indicate that these binding sites are associated with voltage-gated sodium channels. First, detergent-solubilized 125I-Tx1/binding site complexes were immunoprecipitated by antibodies directed against a conserved motif from sodium channel  subunits. Second, 125I-Tx1 binding was inhibited by µ conotoxin GIIIB, a sodium channel antagonist from the venom of the predatory marine gastropod Conus geographus.

    µ Conotoxin GIIIB and the homologous peptide µ conotoxin GIIIA block sodium channels by interacting with the extracellular pore region and compete with tetrodotoxin for binding to site 1. Tetrodotoxin and µ conotoxin binding sites seem to overlap but are not identical (Cestèle and Catterall, 2000; Barbier et al., 2003; Li and Tomaselli, 2004; Terlau and Olivera, 2004). µ Conotoxins are potent blockers of skeletal muscle (Nav1.4) sodium channels but display relatively weak affinity for other sodium-channel subtypes (Cruz et al., 1985). For example, µ conotoxin GIIIB inhibited [3H]saxitoxin binding to sodium channel site 1 in rat muscle membranes by 80% with an apparent KD = 0.14 µM but only displaced approximately 20% of binding to brain membranes (Moczydlowski et al., 1986). Our present data indicate that at the highest concentration tested, µ conotoxin GIIIB inhibited 125I-Tx1 binding to brain membranes by approximately 70% and that the interaction between the two ligands is competitive. These findings and the fact that Tx1 inhibited sodium currents but did not modify their biophysical properties suggest that Tx1 acts in proximity to sodium channel site 1. However, 125I-Tx1 binding was not inhibited by tetrodotoxin, nor were any interactions between [3H]saxitoxin and Tx1 detected (data not shown). Nevertheless, tetrodotoxin did block the ability of µ conotoxin GIIIB to displace 125I-Tx1 binding. These findings can be explained by assuming that Tx1 and tetrodotoxin bind to distinct sites and that µ conotoxins overlap both.

    Sequence comparisons between Tx1 and the µ conotoxin family provided some support for this hypothesis (Table 2). A 19-residue overlap between amino acids 22 to 40 of Tx1, toxins 1A and 1B from Phoneutria keyserlingi, and six µ conotoxins indicates overall similarity in the spacing of four conserved cysteines and a tryptophan present in the three toxins 1 and three µ conotoxins. Extensive structure-function studies with µ conotoxin GIIIA analogs have indicated an important role for basic residues at positions 13, 16, and 19 that are conserved in µ conotoxins (Huang and Miller, 1991; Sato et al., 1991; Li and Tomaselli, 2004; Bulaj et al., 2005). Arginine 13 is particularly critical for interaction with site 1, and the requirement for a guanidinium group in the channel blocking activity of tetrodotoxin, saxitoxin, and the µ conotoxins has led to their classification as the "guanidinium" group of toxins. The toxin 1 variants also carry basic residues that align with µ conotoxin residues 16 and 19, but position 13 is occupied by a glycine or an alanine. We may speculate that similarities in cysteine scaffolding and the position of two positive charges may underlie competitive interactions between Tx1 and the µ conotoxins, whereas the absence of a basic residue equivalent to arginine 13 might preclude competition between Tx1 and tetrodotoxin. Further studies are required to clarify the molecular mechanisms by which Tx1 produces state-dependent inhibition in proximity to sodium channel site 1.

    Acknowledgements

    We thank Professor F. Couraud for his contribution.

     ABBREVIATIONS: Tx1, Phoneutria nigriventer toxin 1; BAPTA, 1,2-bis(2-aminophenoxy)-ethane-N,N,N',N'-tetraacetic acid; CHO, Chinese hamster ovary; HPLC, high performance liquid chromatography; MALDI-TOF, matrix-assisted laser desorption ionization time-of-flight; MS, mass spectrometry; Nav1.2, brain II isoform of the Na+ channel.

    1 Current affiliation: Centre National de la Recherche Scientifique Unité Propre de Recherche 2589, Parc Scientifique de Luminy, Marseille, France.

【参考文献】
  Araùjo DA, Cordeiro MN, Diniz CR, and Beir?o PSL (1993) Effect of a toxic fraction, PhTx2, from the spider Phoneutria nigriventer on the sodium current. Naunyn-Schmiedeberg's Arch Pharmacol 347: 205–208.[Medline]

Barbier J, Benoit E, and Molgó J (2003) Conotoxins targeting voltage-sensitive Na+ channels, in Bacterial, Plant and Animal Toxins (Ascenti P, Polticelli F, and Visca P eds) pp 127–145, Research Signpost, India.

Bulaj G, West PJ, Garrett JE, Marsh M, Zhang MM, Norton RS, Smith BJ, Yoshikami D, and Olivera BM (2005) Novel conotoxins from Conus striatus and Conus kinoshitai selectively block TTX-resistant sodium channels. Biochemistry 44: 7259–7265.[CrossRef][Medline]

Cassola AC, Jaffe H, Fales HM, Castro-Afeche S, Magnoli F, and Cipolla-Neto J (1998) omega-Phonetoxin-IIA: a calcium channel blocker from the spider Phoneutria nigriventer. Pflueg Arch Eur J Physiol 436: 545–552.[CrossRef][Medline]

Catterall WA, Goldin AL, and Waxman SG (2003) International Union of Pharmacology. XXXIX. Compendium of voltage-gated ion channels: sodium channels. Pharmacol Rev 55: 575–578.[Abstract/Free Full Text]

Cestèle S and Catterall WA (2000) Molecular mechanisms of neurotoxin action on voltage-gated sodium channels. Biochimie (Paris) 82: 883–892.

Cordeiro MN, de Figueiredo SG, Valentim AC, Diniz CR, von Eickstedt VRD, Gilroy J, and Richardson M (1993) Purification and amino acid sequences of six Tx3 type neurotoxins from the venom of the Brazilian `armed' spider Phoneutria nigriventer (Keys). Toxicon 31: 35–42.[Medline]

Cruz LJ, Gray WR, Olivera BM, Zeikus RD, Kerr L, Yoshikami D, and Moczydlowski E (1985) Conus geographus toxins that discriminate between neuronal and muscle sodium channels. J Biol Chem 260: 9280–9288.[Abstract/Free Full Text]

De Lima ME, Stankiewicz M, Hamon A, de Figueiredo SG, Cordeiro MN, Diniz CR, Martin-Eauclaire M, and Pelhate M (2002) The toxin Tx4 (6–1) from the spider Phoneutria nigriventer slows down Na+ current inactivation in insect CNS via binding to receptor site 3. J Insect Physiol 48: 53–61.[CrossRef][Medline]

Diniz CR, Cordeiro MN, Rezende L Jr, Kelly P, Fischer S, Reimann F, Oliveira EB, and Richardson M (1990) The purification and amino acid sequence of the lethal neurotoxin Tx1 from the venom of the Brazilian `armed' spider Phoneutria nigriventer. FEBS Lett 263: 251–253.[CrossRef][Medline]

Diniz MRV, Paine MJI, Diniz CR, Theakston RDG, and Crampton JM (1993) Sequence of the cDNA coding for the lethal neurotoxin Tx1 from the Brazilian "armed" spider Phoneutria nigriventer predicts the synthesis and processing of a preprotoxin. J Biol Chem 268: 15340–15342.[Abstract/Free Full Text]

Fontana MD and Vital-Brazil O (1985) Mode of action of Phoneutria nigriventer spider venom at the isolated phrenic nerve-diaphragm of the rat. Braz J Med Biol Res 18: 557–565.[Medline]

Gomez MV, Kalapothakis E, Guatimosim C, and Prado MA (2002) Phoneutria nigriventer venom: a cocktail of toxins that affect ion channels. Cell Mol Neurobiol 22: 579–588.[CrossRef][Medline]

Gouvêa dos Santos R, Diniz CR, Cordeiro MN, and De Lima ME (1999) Binding sites and actions of Tx1, a neurotoxin from the venom of the spider Phoneutria nigriventer, in guinea pig ileum. Braz J Med Biol Res 32: 1565–1569.[Medline]

Gouvêa dos Santos R, Van Renterghem C, Martin-Moutot N, Mansuelle P, Cordeiro MN, Diniz CR, Mori Y, De Lima ME, and Seagar M (2002) Phoneutria nigriventer -phonetoxin IIA blocks the Cav2 family of calcium channels and interacts with -conotoxin-binding sites. J Biol Chem 277: 13856–13862.[Abstract/Free Full Text]

Huang X and Miller W (1991) A time-efficient, linear-space local similarity algorithm. Adv Appl Math 12: 337–357.[CrossRef]

Le?o RM, Cruz JS, Diniz CR, Cordeiro MN, and Beir?o PS (2000) Inhibition of neuronal high-voltage activated calcium channels by the -Phoneutria nigriventer Tx3–3 peptide toxin. Neuropharmacology 39: 1756–1767.[CrossRef][Medline]

Li RA and Tomaselli GF (2004) Using the deadly mu-conotoxins as probes of voltagegated sodium channels. Toxicon 44: 117–122.[Medline]

L?nnendonker U (1989a) Binding of tetrodotoxin and saxitoxin to Na+ channels at different holding potentials: fluctuation measurements in frog myelinated nerve. Biochim Biophys Acta 985: 161–167.[Medline]

L?nnendonker U (1989b) Use-dependent block of sodium channels in frog myelinated nerve by tetrodotoxin and saxitoxin at negative holding potentials. Biochim Biophys Acta 985: 153–160.[Medline]

Love S and Cruz-H?fling MA (1986) Acute swelling of nodes of Ranvier caused by venoms which slow inactivation of sodium channels. Acta Neuropathol (Berl) 70: 1–9.[CrossRef][Medline]

Martin-Moutot N, Lévêque C, Sato K, Kato R, Takahashi M, and Seagar M (1995) Properties of conotoxin MVIIC receptors associated with 1A calcium channel subunits in rat brain. FEBS Lett 366: 21–25.[CrossRef][Medline]

Matavel A, Cruz JS, Penaforte CL, Araùjo DAM, Kalapothakis E, Prado VF, Diniz CR, Cordeiro MN, and Beir?o PSL (2002) Electrophysiological characterization and molecular identification of the Phoneutria nigriventer peptide toxin PnTx2–6. FEBS Lett 523: 219–223.[CrossRef][Medline]

Moczydlowski E, Olivera BM, Gray WR, and Strichartz GR (1986) Discrimination of muscle and neuronal Na-channel subtypes by binding competition between [3H]saxitoxin and μ-conotoxins. Proc Natl Acad Sci USA 83: 5321–5325.[Abstract/Free Full Text]

Patton DE and Goldin AL (1991) A voltage-dependent gating transition induces use-dependent block by tetrodotoxin of rat IIA sodium channels expressed in Xenopus oocytes. Neuron 7: 637–647.[CrossRef][Medline]

Peng K, Shu Q, Zhonghua L, and Liang S (2002) Function and solution structure of Huwentoxin-IV, a potent neuronal tetrodotoxin (TTX)-sensitive sodium channel antagonist from Chinese bird spider Selenocosmia huwena. J Biol Chem 277: 47564–47571.[Abstract/Free Full Text]

Sarkar SN, Adhikari A, and Sikdar SK (1995) Kinetic characterization of rat brain type IIA sodium channel alpha-subunit stably expressed in a somatic cell line. J Physiol (Lond) 488: 633–645.[Abstract/Free Full Text]

Sato K, Ishida Y, Wakamatsu K, Kato R, Honda H, Ohizumi Y, Nakamura H, Ohya M, Lancelin J-M, Kohda D, et al. (1991) Active site of μ-conotoxin GIIIA, a peptide blocker of muscle sodium channels. J Biol Chem 266: 16989–16991.[Abstract/Free Full Text]

Terlau H and Olivera BM (2004) Conus venoms: a rich source of novel ion channel-targeted peptides. Physiol Rev 84: 41–68.[Abstract/Free Full Text]

Vieira LB, Kushmerick C, Hildebrand ME, Garcia E, Stea A, Cordeiro MN, Richardson M, Gomez MV, and Snutch TP (2005) Inhibition of high voltage-activated calcium channels by spider toxin PnTx3-6. J Pharmacol Exp Ther 314: 1370–1377.[Abstract/Free Full Text]

West JW, Scheuer T, Maechler L, and Catterall WA (1992) Efficient expression of rat brain type IIA Na+ channel subunits in a somatic cell line. Neuron 8: 59–70.[CrossRef][Medline]

作者单位：Institut National de la Santé et de la Recherche Médicale U641 (N.M.-M., G.A., M.S., C.V.R.), Centre National de la Recherche Scientifique Formation de Recherche en Evolution 2738 (P.M.), and Université de la Méditerranée, Institut Fédératif de Recherches 11 (N.M.-M., P.M., G.A., M.S., C.V.R.), Mars