Introduction

Top Summary Introduction Materials & Methods Results Discussion References

Epilepsy is a common neurological disorder. Different types of seizures may have different neurobiological bases and are controlled with different medications. DPH and CBZ have been the mainstay in the treatment of generalized tonic-clonic and partial seizure for a few decades (Rogawski and Porter, 1990). LTG is an anticonvulsant just in clinical use that shows similar effect to DPH and CBZ when used as monotherapy in newly diagnosed epilepsy (Steiner et al., 1994; Brodie et al., 1995). LTG, CBZ, and DPH also are similar in antiepileptic profiles in animal seizure models (Miller et al., 1986) and in use-dependent block of neuronal discharges at the cellular level (McLean and McDonald, 1983, 1986; Lees and Leach, 1993; Xie et al., 1995). The use-dependent block of discharges has been considered of great mechanistic importance because it readily explains why these anticonvulsants effectively inhibit only seizure discharges and spare most normal neuronal activities.

The molecular basis underlying the use-dependent block of discharges has been ascribed to voltage-dependent inhibition of Na⁺ currents. DPH, CBZ, and LTG all inhibit Na⁺ currents, and the inhibition is more pronounced at more depolarized membrane potentials (Matsuki et al., 1984; Willow et al., 1985; Lang et al., 1993; Kuo and Bean, 1994a; Xie et al., 1995; Kuo and Lu, 1997; Kuo et al., 1997). Detailed examination of the steady state effect and reaction kinetics of DPH, CBZ, and LTG on central neuronal Na⁺ currents further discloses very similar qualitative features in the molecular interactions between these anticonvulsants and the Na⁺ channel (Kuo and Bean, 1994a; Kuo and Lu, 1997; Kuo et al., 1997). For example, these drugs all bind to the channel via a simple bimolecular reaction (a one-to-one binding process), and they all have a much higher affinity toward the fast inactivated state than toward the resting (deactivated) state of Na⁺ channels.

Despite the striking similarities in the mode of action on neuronal Na⁺ channels by DPH, CBZ, and LTG, the chemical structures of these drugs apparently are not similar (Fig. 1A). DPH is a hydantoin containing the ureide structure (Fig. 1B), which is traditionally viewed as an important structural motif responsible for antiepileptic activities. CBZ, however, does not contain the ureide structure and is a tricyclic compound with a very short amide side chain. LTG is an even simpler compound composed of only two aromatic rings. Could the very similar mode of action still indicate that these drugs bind to the same binding site or "receptor" in the Na⁺ channel? And if so, can we tell more about the location and organization of the receptor? Because the highly selective binding of these anticonvulsants to the inactivated state rather than the resting state is a consequence of channel gating, characterization of the binding site for the anticonvulsants would not only be of pharmacological and pharmaceutical interest but also shed light on the gating conformational changes of the Na⁺ channel. In this study, I examine the inhibition of Na⁺ current by mixtures of different anticonvulsants and demonstrate that the Na⁺ channel cannot be doubly occupied by DPH, CBZ, or LTG. It also is found that DPH, CBZ, and LTG are effective Na⁺ channel inhibitors when applied externally, yet they are of no discernible effect when applied to the cytoplasmic side. Altogether these findings suggest that DPH, CBZ, and LTG bind to a common binding site located on the external side of neuronal Na⁺ channels.

View larger version (13K): [in this window] [in a new window]   Fig. 1.   Chemical structure of the drugs. A, Structures of phenytoin, carbamazepine, and lamotrigine. B, The ureide structure that could be found in many anticonvulsants, including phenytoin but not carbamazepine or lamotrigine. Black dot, substitutions in the ring, which could be ---C---N---, ---N---, ---C---C---, ---O---, or ---C---. R1 and R2 are side chain groups attached to the five- or six-member ring.

	Materials and Methods

Top Summary Introduction Materials & Methods Results Discussion References

Cell preparation. Coronal slices of the brain were prepared from 7-14-day-old Long-Evans rats. The CA1 region was dissected from the slices and cut into small chunks. After treatment for 5-10 min at 37° in dissociation medium (82 mM Na₂SO₄, 30 mM K₂SO₄, 3 mM MgCl₂, 5 mM HEPES, and 0.001% phenol red indicator, pH 7.4) containing 0.5 mg/ml trypsin (Type XI; Sigma Chemical, St. Louis, MO), tissue chunks were moved to dissociation medium containing no trypsin but 1 mg/ml bovine serum albumin (Sigma) and 1 mg/ml trypsin inhibitor (Type II-S, Sigma). Each time when cells were needed, two or three chunks were picked and triturated to release single neurons.

Whole-cell recording. The dissociated neurons were put in a recording chamber containing Tyrode's solution (150 mM NaCl, 4 mM KCl, 2 mM MgCl₂, 2 mM CaCl₂, and 10 mM HEPES, pH 7.4). Whole-cell voltage clamp recordings were obtained using pipettes pulled from borosilicate micropipettes (o.d., 1.55-1.60 mm; Hilgenberg, Malsfeld, Germany), fire polished, and coated with Sylgard (Dow-Corning, Midland, MI). Except for the internal anticonvulsant experiments (see Fig. 8), the pipettes were filled with the standard internal solution containing 75 mM CsCl, 75 mM CsF, 2.5 mM MgCl₂, 5 mM HEPES, and 5 mM EGTA, pH adjusted to 7.4 by CsOH. For the experiments studying the effect of internal anticonvulsants, 300 µM LTG, 300 µM CBZ, or 100 µM DPH was added to the standard internal solution. Seal was formed, and the whole-cell configuration was obtained in Tyrode's solution. The cell then was lifted from the bottom of the chamber and moved in front of an array of flow pipes (microcapillary; Hilgenberg; content, 1 µl, length, 64 mm) emitting external recording solutions, which were Tyrode's solution with or without different concentrations of drugs. DPH, CBZ, and LTG were dissolved in dimethylsulfoxide to make a 100 mM stock solution, which then was diluted into Tyrode's solution to attain the final concentrations desired. The final concentration of dimethylsulfoxide (0.3%) was not found to have significant effect on Na⁺ currents. Currents were recorded at room temperature (~25°) with an Axoclamp 200A amplifier, filtered at 5 kHz with four-pole Bessel filter, digitized at 20-50-µsec intervals, and stored using a Digidata-1200 analog/digital interface along with the pCLAMP software (Axon Instruments, Foster City, CA). Residual series resistance was generally smaller than 1 M after partial compensation (typically >80%). DPH and CBZ were from Sigma, and LTG was a kind gift from Wellcome Foundation (Kent, England). All statistical values are given as mean ± standard deviation.

	Results

Top Summary Introduction Materials & Methods Results Discussion References

The shift of inactivation curve by 100 µM CBZ plus 100 µM LTG argues against simultaneous occupancy of the channel by CBZ and LTG. It has been shown that steady state inactivation curve of Na⁺ channels is shifted by DPH, CBZ, and LTG (Matsuki et al., 1984; Willow et al., 1985; Lang et al., 1993; Kuo and Bean, 1994a; Xie et al., 1995; Kuo and Lu, 1997; Kuo et al., 1997). The shift is well explained by a scheme that the anticonvulsants bind to the inactivated channels with a much higher affinity than to the resting channels (Fig. 2A). According to this scheme, in the control condition the fraction of channels in state R at different membrane potentials (V) can be approximated by a Boltzmann distribution: 1/(1 + exp[(V  V_h)/k] (the "inactivation curve"). In the presence of an anticonvulsant, the shape of the inactivation curve (the slope factor k) remains the same, but the midpoint of the curve (V_h) would be shifted by V, which is given by the following equation (Bean et al., 1983; Kuo and Bean, 1994a):

(1)

where D is the concentration of the anticonvulsant, and K_R and K_I are the dissociation constants for the resting and inactivated states, respectively. Because K_R is generally in the millimolar range for these anticonvulsants in rat hippocampal neurons (Kuo and Bean, 1994a; Kuo and Lu, 1997; Kuo et al., 1997), item D/K_R is small with ~100 µM anticonvulsant. The interaction between anticonvulsants and the resting channels thus is disregarded for a more straightforward (but still valid) conceptualization of the V in the presence of multiple drugs. Deletion of D/K_R becomes

(2)

Based on rationales of eqs. 1 and 2, one may derive the V in the presence of two different anticonvulsants. If the two drugs have the same binding site in the channel, then the binding of one drug to the inactivated channel would preclude the binding of the other drug (the one-site model; Fig. 2B), and V is given by

(3)

where D1 and D2 are the concentrations of the two drugs, and K_I,1 and K_I,2 are the K_I values for drug 1 and drug 2, respectively. However, if the inactivated channel can be doubly occupied by the drugs (i.e., there are different binding sites for different drugs), then there would be occupancy of a new state (ID₁D₂), the inactivated Na⁺ channel simultaneously occupied by two different drugs (the two-site model, Fig. 2C), and the shift of inactivation curve is given by

(4)

The shift of inactivation curve predicted by eq. 3 is obviously different from that predicted by eq. 4. One may use saturating concentrations of the drugs to make the difference large and easily discernible. Fig. 3, A-C, shows the shift of inactivation curve (V) in high concentrations of CBZ and LTG. The V documented in these neurons in the presence of CBZ or LTG alone is consistent with previous observations (Kuo and Lu, 1997; Kuo et al., 1997), where the shift of inactivation curve over a wide range of drug concentration yielded a K_I value of ~25 µM for CBZ and a K_I value of ~9 µM for LTG. Based on these K_I values and a slope factor k = 6, the V in the presence of 100 µM CBZ plus 100 µM LTG would be 16.7 mV by eq. 3 and 24.6 mV by eq. 4. The experimental data (~16.6 mV, Fig. 3C) is much closer to the former than to the latter. Also, the inactivation curve in 100 µM CBZ plus 100 µM LTG typically lies between that in 200 µM CBZ and that in 200 µM LTG, rather than being far more negative to that in 200 µM LTG (Fig. 3, B and C). These findings argue against significant existence of the doubly occupied state ID₁D₂. In other words, it seems that the Na⁺ channel cannot be occupied simultaneously by CBZ and LTG.

View larger version (6K): [in this window] [in a new window]   Fig. 2.   Simplified gating schemes illustrating possible actions of the anticonvulsants phenytoin, carbamazepine, and lamotrigine. A, R and I, Resting and inactivated states of the Na+ channel, respectively. D, Anticonvulsant drugs. RD and ID, anticonvulsant drug-bound resting and inactivated states, respectively. The anticonvulsant drugs have much higher affinity to the I state than to the R state. Thus, the +D arrow (drug binding rates) between I and ID is much larger than the D arrow (drug unbinding rates), but the relative size of the arrows between R and RD is just the opposite. Membrane depolarization (+V) tends to move the channel from the resting to the inactivated state, whereas membrane hyperpolarization (V) has the opposite effect. B, During depolarization, most channels are moved to the inactivated state. If anticonvulsants drug 1 and drug 2 (D1 and D2) share the same receptor in the inactivated channel, then the channel can be bound by either D1 or D2 but not both, even when saturating concentrations of D1 and D2 are present. C, If different anticonvulsants have separate receptors, then the inactivated Na+ channels may be doubly occupied by both D1 and D2, especially when saturating concentrations of D1 and D2 are present.

View larger version (17K): [in this window] [in a new window]   Fig. 3.   Shift of the inactivation curve by CBZ and LTG. A, and B, The cell was held at 130 mV and stepped every 15 sec to the inactivating pulse (130 to 60 mV) for 9 sec. The channels that remain available after each inactivating pulse were assessed by the peak currents during a following short test pulse to 0 mV for 5 msec. The fraction available is defined as the normalized peak current (relative to the current evoked with an inactivating pulse at 130 mV) and is plotted against the voltage of the inactivating pulse (V). Both parts A and B contain the same two sets of control data, which were obtained before and after the five sets of data in the presence of drugs to demonstrate the lack of significant voltage drift during this long experiment. The lines are fits with a Boltzmann function 1/(1 + exp[(V  Vh)/k)], with Vh values (in mV) of 83.1, 83.1, 91.3, 95.1, 96.4, 98.1, and 97.7, and k values of 6.2, 5.9, 6.3, 6.2, 5.8, 6.1, and 6.1 for control (before drugs), control (after drugs), 100 µM CBZ, 100 µM LTG, 200 µM CBZ, 200 µM LTG, and 100 µM CBZ plus 100 µM LTG, respectively. C, Shift of the inactivation curve by CBZ and/or LTG. The shift (V) is determined in each cell by the difference between Vh in control and in the presence of anticonvulsant drugs. The values of V (in mV) are 7.8 ± 1.1, 11.3 ± 1.4, 12.5 ± 1.8, 17.2 ± 1.8, and 16.6 ± 2.1 (all from five sets) for 100 µM CBZ, 100 µM LTG, 200 µM CBZ, 200 µM LTG, and 100 µM CBZ plus 100 µM LTG, respectively.

The shift of inactivation curve by 100 µM DPH plus 100 µM CBZ plus 100 µM LTG suggests no double occupancy of the channel by any two drugs. Fig. 4A summarizes the V in high concentrations of DPH and LTG. The V in 200 µM DPH is not available because of the low solubility of DPH in Tyrode's solution (~100 µM, pH 7.4, 25°). The V in the presence 100 µM DPH also is consistent with previous observations in the same system (Kuo et al., 1997), where a K_I value for DPH of ~9 µM is documented. The experimental data of V are 16.1 mV in 100 µM DPH plus 100 µM CBZ and 17.8 mV in 100 µM DPH plus 100 µM LTG. These values again are much closer to the predictions by eq. 3 (16.7 and 18.9 mV, respectively) than to the predictions by eq. 4 (24.6 and 29.9 mV, respectively). Moreover, the V in the presence of 100 µM DPH plus 100 µM CBZ plus 100 µM LTG is 18.5 mV (Fig. 4, B and C). This is slightly smaller than the V value in 300 µM LTG (19.1 mV) and is close to the predicted value according to the equation (19.8 mV, calculated using aforementioned K_I values of DPH, CBZ, and LTG, and a slope factor k = 6):

(5)

where D1, D2, and D3 are the concentrations of the three drugs, and K_I,1, K_I,2, and K_I,3 are the K_I values for drugs 1, 2, and 3, respectively. Eq. 5 is an extension of eq. 3. On the other hand, if the three drugs can all bind to the channel simultaneously, then an extension of eq. 4 based on incorporation of the triply occupied state ID₁D₂D₃ and three doubly occupied states would predict a much larger V value (39.6 mV) by 100 µM DPH plus 100 µM CBZ plus 100 µM LTG. These findings suggest that one inactivated channel can be occupied by only one drug molecule, even though high concentrations of DPH, CBZ, and LTG are present.

View larger version (13K): [in this window] [in a new window]   Fig. 4.   Experiments similar to those in Fig. 3 to demonstrate shift of the inactivation curve (V) by different combinations of anticonvulsants DPH, CBZ, and LTG. A, The values of V (in mV) are 13.4 ± 1.6, 16.1 ± 2.0, and 17.8 ± 2.3 (all from five sets) for 100 µM DPH, 100 µM CBZ plus 100 µM DPH, and 100 µM LTG plus 100 µM DPH, respectively. B, and C, The inactivation curves in 300 µM LTG and in 100 µM CBZ plus 100 µM LTG plus 100 µM DPH. The lines are fits with a Boltzmann function 1/(1 + exp[(V  Vh)/k)], with Vh values (in mV) of 77.3, 76.6, 95.4, and 94.8 and k values of 5.9, 6.0, 6.3, and 6.1 for control (before drugs), control (after drugs), 300 µM LTG, and 100 µM CBZ plus 100 µM LTG plus 100 µM DPH, respectively. The values of V (in mV) are 19.2 ± 2.5 and 18.5 ± 1.9 (both from eight data sets) for 300 µM LTG and 100 µM CBZ plus 100 µM LTG plus 100 µM DPH, respectively.

The recovery kinetics of the inhibited Na⁺ current are similar in different drug concentrations. In addition to the effect on the steady state inactivation curve, one may examine the existence of doubly occupied inactivated channels from a kinetic point of view. Fig. 5, A-D, shows the time course of recovery of Na⁺ current from inhibition by either CBZ or LTG. Here, 100 and 200 µM drug concentrations (quite higher than the apparent K_I value for each drug) are used to ensure that a major portion of the inactivated channels is bound by the anticonvulsant at the end of the long inactivating prepulse. In the control (drug-free) condition, a major part (~70%) of the Na⁺ current recovers within 10 msec at 120 mV after the long inactivating pulse (Fig. 5, A-D). When the anticonvulsant is present, there is a small but very fast component of recovery in the first 10 msec. This component presumably represents the recovery from residual drug-free inactivated Na⁺ channels in the presence of CBZ or LTG and therefore is smaller in higher drug concentrations. On the other hand, the majority of the recovery after 10 msec should be from the drug-bound channels. The recovery courses after 10 msec are quite similar in different drug concentrations, which could be demonstrated by the time constants of the forced monoexponential fits to this part of the recovery courses. This is consistent with the notion that these anticonvulsants interact with Na⁺ channels via a simple bimolecular reaction, where the drug unbinding rate should be unrelated to ambient drug concentrations.

View larger version (26K): [in this window] [in a new window]   Fig. 5.   Recovery from inhibition by CBZ or LTG. A, In control or in the continuous presence of 100 or 200 µM CBZ, the cell was held at 120 mV and then prepulsed to 40 mV for 9 sec to reach a maximal (steady state) block of Na+ current by CBZ. The cell then was stepped back to a recovery gap potential at 120 mV for variable length before being stepped again to a short test pulse at 0 mV for 5 msec to assess the available current. The pulse protocol was repeated every 15 sec. After long recovery period (~5 sec), the peak currents at the test pulse reached a "steady state" level, which was slightly reduced in CBZ than in control because of the very mild inhibition of resting channels by high concentrations of CBZ. The fraction recovered is defined as the normalized peak current at the test pulse (relative to the peak current at the test pulse after a 5-sec recovery gap period) and is plotted against the length of the recovery gap period. Lines, monoexponential fits to the recovery course after 10 msec in drugs and are of the form: fraction recovered = 0.97  0.64exp[(t  10)/180)] for 100 µM CBZ (t denotes length of the recovery gap potential in msec), and fraction recovered = 0.96  0.75exp[((t  10)/197)] for 200 µM CBZ. B, The first 1 sec of the recovery courses in part A is replotted with a different time scale for a better illustration of the data. The lines are the same as those in part A. C, Similar experiments to those in part A were repeated in the same cell to demonstrate the recovery kinetics in 100 and 200 µM LTG. Lines, monoexponential fits for the recovery course after 10 msec and are of the form: fraction recovered = 0.97  0.73exp[((t  10)/325)] for 100 µM LTG, and fraction recovered = 0.97  0.82exp[((t  10)/317)] for 200 µM LTG. D, The first 1 sec of the recovery courses in part C are replotted with a different time scale for a better illustration of the data. Lines, same as those in part C. E, The recovery courses in part A are not purely from the CBZ-bound channels but are "contaminated" by the small amount of drug-free channels. The "contamination" is corrected (for details, please refer to the text) to yield the recovery courses of the CBZ-bound channels in this plot. Dashed lines, connecting the data points and have no mathematical meanings. Note that the recovery courses are very similar to each other whether it is derived from the 100 µM or 200 µM CBZ data in part A. F, Similar treatment of the data in part C yields the recovery courses of the LTG-bound channels. Dashed lines, connecting the data points and have no mathematical meanings. Note again that the courses derived from the 100 or 200 µM LTG data in part C are quite similar to each other.

Recovery of the inhibited Na⁺ currents in 100 µM CBZ plus 100 µM LTG is faster than that in 100 µM LTG alone. Fig. 6, A-D, examines the recovery kinetics in 100 µM CBZ plus 100 µM LTG. If the channel can be doubly occupied by CBZ and LTG, then the recovery from inhibition by 100 µM CBZ plus 100 µM LTG is expected to be slower than those in 100 µM CBZ or in 100 µM LTG. This is because in the presence of saturating concentrations of both drugs, most channels would be in the double-occupancy state (state ID₁D₂ in Fig. 2C) if there are separate drug binding sites. Recovery from state ID₁D₂ conceivably would be slower than that from ID₁ or ID₂ because it is one step farther from the R state. Fig. 6, A and B, however, shows that the recovery course in 100 µM CBZ plus 100 µM LTG is not slower but is even faster than the recovery in 100 µM LTG. Although drug-free inactivated channels must be less prevalent in 100 µM CBZ plus 100 µM LTG than in 100 µM LTG at the end of the long inactivating prepulse, the absolute value of fraction recovered is higher in 100 µM CBZ plus 100 µM LTG than in 100 µM LTG at almost every time point. In all four cells examined, the half-recovery time is always shorter, or at least not longer, in 100 µM CBZ plus 100 µM LTG than in 100 µM LTG alone (Fig. 6C). The faster recovery in the presence of 100 µM CBZ plus 100 µM LTG strongly suggests that the binding of one drug (e.g., the faster unbinding drug CBZ) decreases the binding of the other (e.g., the slower unbinding drug LTG). This is consistent with the foregoing argument that CBZ binding and LTG binding to the channel are mutually exclusive.

View larger version (22K): [in this window] [in a new window]   Fig. 6.   Recovery from inhibition by 100 µM CBZ plus 100 µM LTG. A, Recovery from inhibition by 100 µM CBZ plus 100 µM LTG were documented in the same cell as that in Fig. 5. The experimental protocol is also the same as that in Fig. 5A. The recovery courses in 100 µM CBZ and 100 µM LTG are taken from Fig. 5, A and C, and are replotted here for comparison. B, The first 1 sec of the recovery courses in part A is replotted with a different time scale. Note the faster recovery course in 100 µM CBZ plus 100 µM LTG than in 100 µM LTG. C, Because the recovery courses in 100 µM CBZ plus 100 µM LTG cannot be reasonably approximated by monoexponential functions, the half recovery time is used as a simple parameter to compare the recovery kinetics in different conditions. The half recovery time is defined as the length of the recovery gap potential to recover half the current (fraction recovered equals 0.5). The same symbols represent data from the same cell. In every cell tested, the half recovery time is shorter in 100 µM CBZ plus 100 µM LTG than in 100 LTG alone. D, Based on rationales of a bimolecular reaction and the data in Fig. 5, E and F, the "theoretical" recovery courses in 100 µM CBZ plus 100 µM LTG are calculated according to the one-site model and the two-site model (for details, refer to the text). The lines are connecting the data points and have no mathematical meanings. The experimental findings (the same data as those in part A) are closer to the course predicted by the one-site model than to the course predicted by the two-site model.

(6)

(7)

Recovery of the inhibited Na⁺ currents in 100 µM CBZ plus 100 µM DPH also is no slower than that in 100 µM DPH alone. Fig. 7, A-D, demonstrates the recovery course in 100 µM CBZ plus 100 µM DPH. The findings are very similar to those in Fig. 6, A-D. Fig. 7, A and B, shows the recovery courses in 100 µM CBZ, 100 µM DPH, and 100 µM CBZ plus 100 µM DPH. The recovery in 100 µM CBZ plus 100 µM DPH is not slower than the recovery in 100 µM DPH (although the drug-free inactivated channels must be less prevalent in 100 µM CBZ plus 100 µM DPH than in 100 µM DPH at the end of the long prepulse). The half recovery time is also shorter, or at least not longer, in 100 µM CBZ plus 100 µM DPH than in 100 µM DPH (Fig. 7C). Based on the same rationales underlying Figs. 5, E and F, and 6D, and a K_I value of ~25 µM for CBZ as well as a K_I value of ~9 µM for DPH, the recovery courses in 100 µM CBZ plus 100 µM DPH predicted by the one- and two-site models are derived from the data in Fig. 7A and are plotted in Fig. 7D. The observed recovery course in 100 µM CBZ plus 100 µM DPH again is well predicted by the one-site model. Altogether, the kinetic data in Figs. 6 and 7 strongly argue against the existence of channels doubly occupied by saturating concentrations of CBZ, LTG, or DPH and are thus suggestive of a common binding site for these anticonvulsants in inactivated Na⁺ channels.

View larger version (22K): [in this window] [in a new window]   Fig. 7.   Recovery from inhibition by 100 µM CBZ plus 100 µM DPH. A, Recovery from inhibition by 100 µM CBZ, 100 µM DPH, and 100 µM CBZ plus 100 µM DPH were examined by the same protocol as that in Fig. 5A. B, The first 1 sec of the courses in part A are replotted with a different time scale. Note that the recovery in 100 µM CBZ plus 100 µM DPH is not slower than the recovery in 100 µM DPH alone. C, In every cell tested, the half recovery time (defined in the legend of Fig. 6C) is not longer in 100 µM CBZ plus 100 µM DPH than in 100 µM DPH alone. The same symbols represent data from the same cell. D, The recovery courses in 100 µM CBZ plus 100 µM DPH predicted by the one-site model and the two-site model are derived based on the same rationales underlying those in Figs. 5, E and F, and 6D. The lines are connecting the data points and have no mathematical meanings. The experimental findings (the same data as those in part A) again are closer to the course predicted by the one-site model than to the course predicted by the two-site model.

DPH, CBZ, and LTG have no significant effect on the Na⁺ current if applied internally. In previous experiments, DPH, CBZ, and LTG are applied externally (applied to the extracellular side). However, the uncharged form of these anticonvulsants cross the membrane easily. Because there is no rapid "wash" of the intracellular space, the intracellular drug concentration may be similar to the external drug concentration in those experiments. Thus one cannot tell whether the anticonvulsants inhibit Na⁺ channels from outside or inside based on the previous experiments. Taking advantage of the rapid and continuous solution change of the extracellular space in the experimental system, I examined the effect of internal LTG by adding 300 µM LTG to the pipette solution. Now, the LTG crossing the membrane and reaching the outside is quickly washed away. Thus, there will be no significant build-up of LTG concentrations in the external solution. In ~10 min after breakthrough into the cell, when the intracellular space should be completely dialyzed by the LTG-containing internal solution, Na⁺ current is still elicitable by a test pulse from a holding potential of 65 mV, and the current amplitude is ~10% of that elicited from a holding potential of 120 mV (Fig. 8A). This is similar to the case with drug-free internal solution and is very different from the effect of 300 µM external LTG (see the inactivation curves in Fig. 4). Moreover, with 300 µM internal LTG and a holding potential of 65 mV, the elicited Na⁺ current still is significantly inhibited by 30 µM external LTG, further supporting that the "control" current in Fig. 8A is not a residual current already under significant inhibition by 300 µM LTG. To elucidate this point more quantitatively, the inactivation curves in the absence and presence of 30 µM external LTG are obtained with 300 µM LTG in the intracellular space. The shift of inactivation curve (V) by 30 µM external LTG in the presence of 300 µM internal LTG is very similar to the V by 30 µM external LTG with drug-free internal solution (Fig. 8B). In contrast, if the Na⁺ current is already under significant inhibition by 300 µM LTG, there should be only negligible effect by the addition of 30 µM LTG (V < 0.6 mV by eq. 2). Significant effects of 30 µM external DPH and 30 µM external CBZ also are observed with 100 µM internal DPH and 300 µM internal CBZ, respectively (Fig. 8, C and D). Thus, internal DPH, CBZ, and LTG all seem to show no significant effect on the Na⁺ current, which implies an external rather than an internal location of the binding site for these anticonvulsants in the Na⁺ channel.

View larger version (17K): [in this window] [in a new window]   Fig. 8.   Internal LTG, DPH, or CBZ has no effect on Na+ currents. A, Sample Na+ currents in a cell in which the pipette contained standard internal solution plus 300 µM LTG. After the whole-cell configuration was obtained, the Na+ current was elicited by a short step depolarization to 0 mV for 15 msec from holding potentials 120 mV or 65 mV (the "control" sweeps in both panels). For the next 10 min, the current had shown no significant change. The cell then was moved into external solution containing 30 µM LTG, which significantly inhibits the Na+ current elicited from a holding potential of 65 mV. B, The shift of inactivation curve (V) by 30 µM external LTG in five cells containing 300 µM internal LTG is 6.9 ± 0.8 mV (, determined by the same procedures as those in Fig. 3). As a comparison, the V by 30 µM external LTG in five cells containing no internal anticonvulsants is 6.7 ± 1.2 mV (). C, Similar experiments as those in part B. The V by 30 µM external DPH in five cells containing 100 µM internal DPH is 7.4 ± 0.8 mV (), and the V by 30 µM external DPH in five cells containing no internal anticonvulsants is 7.5 ± 1.0 mV (). D, Similar experiments as those in part B. The V by 30 µM external CBZ in five cells containing 300 µM internal CBZ is 5.5 ± 0.9 mV (), and the V by 30 µM external CBZ in five cells containing no internal anticonvulsants is 5.3 ± 1.2 mV ().

	Discussion

Top Summary Introduction Materials & Methods Results Discussion References

DPH, CBZ, and LTG have qualitatively very similar inhibitory effects on central neuronal Na⁺ currents. Based on the steady state and kinetic data in this study, I propose that such similar effects arise, at least in part, from a common anticonvulsant binding site located on the external side of neuronal Na⁺ channels.

No double occupancy of the Na⁺ channel by anticonvulsants DPH, CBZ, or LTG. DPH, CBZ, and LTG all preferentially bind to the fast inactivated state rather than to the resting state of Na⁺ channels (Matsuki et al., 1984; Willow et al., 1985; Lang et al., 1993; Kuo and Bean, 1994a; Xie et al., 1995; Kuo and Lu, 1997; Kuo et al., 1997), and the interaction between these anticonvulsants and the inactivated Na⁺ channel is a bimolecular reaction (one-to-one binding process; Kuo and Bean, 1994a; Kuo and Lu, 1997; Kuo et al., 1997). In this study, it is shown that the rule of "single occupancy" is still observed when two or more anticonvulsants are simultaneously present in the system. The shift of inactivation curve is quantitatively incompatible with a model that the channel can be simultaneously occupied by two different drug molecules yet is well in line with a scheme that the binding of one drug precludes the binding of the others. The faster recovery from inhibition by a mixture of a fast-unbinding drug (e.g., CBZ) and a slow-unbinding drug (e.g., LTG) than from the inhibition by one single slow-unbinding drug further strengthens the mutual exclusiveness among DPH, CBZ, and LTG binding to the Na⁺ channel.

A common receptor for DPH, CBZ, and LTG versus an allosteric interaction among the drugs. The mutual exclusiveness among the binding of different drugs to a channel may result from either an allosteric mechanism or a direct competition mechanism. The direct competition mechanism assumes that different drugs bind to the same receptor site. The allosteric mechanism assumes that different drugs bind to different receptors in the channel, but these different receptors do not coexist in one channel conformation. The scheme in Fig. 2B depicts only one inactivated state, which is connotative of a direct competition mechanism. For the allosteric model, the I state in Fig. 2B should be subdivided into at least three different inactivated states, I', I'', and I''', which have the receptor for DPH, CBZ, and LTG, respectively. Binding of DPH to I', for example, would stabilize the channel to state I' and thus prevent the binding of CBZ and LTG.

Nature of the anticonvulsant receptor in Na⁺ channels. In view of the apparent dissimilarity in chemical structure of DPH, CBZ, and LTG, it is interesting that these anticonvulsants would bind to a common receptor in Na⁺ channels. DPH has two phenyl rings connected to a carbon atom in a ureide core. CBZ has two phenyl rings attached to another "bridge ring" to form a tricyclic structure. LTG is an even simpler compound and has just two phenyl rings connected to each other (Fig. 1A). Thus the only common structural motif shared by these three drugs is two phenyl groups separated by one to two C---C or C---N single bonds (1.5-3 Å). Such a motif presumably contains the major ligands interacting with the anticonvulsant receptor in Na⁺ channels. It has been shown that the neutral, rather than charged, form of DPH is the active form inhibiting Na⁺ currents (Morello et al., 1984). Thus, the binding between DPH and its receptor probably is nonionic. Other than ionic bond, drug/receptor interaction most likely involves hydrophobic bond (which represents a freeing of water molecules and a gain in entropy) or (induced) dipole-induced dipole bond [London forces or Debye forces; for review, see Zimmerman and Feldman (1989)]. The dissociation constants between these anticonvulsants and the inactivated channel are ~9-25 µM, which may be translated into a binding energy of 10.6-11.6 RT, or ~6.5 kcal/mol. Both the binding energy value and the notion that phenyl groups may be the major binding ligands are consistent with the proposal that the aforementioned nonionic bonds play a major role in the drug/receptor interaction under consideration here.

Implications on the gating conformational changes of Na⁺ channels. On depolarization, the Na⁺ channel is activated quickly and then inactivated quickly because of binding of an inactivating particle (the "ball and chain model"; Armstrong and Benzanilla, 1977; Armstrong, 1981) or a hinged peptide flap [the "hinged-lid model" (West, 1992)] to the internal pore mouth to block the pore. The open state is omitted in the schemes in Fig. 2 because the binding rates of DPH, CBZ, and LTG are so slow that no significant drug binding may happen to the very transient open state (Kuo and Bean, 1994a; Kuo and Lu, 1997; Kuo et al., 1997). Because the inactivated state may be considered as a state that is activated (open) but blocked at the internal pore mouth, the high affinity between anticonvulsants and the inactivated channel may arise from either channel activation or binding of the inactivating peptide (the hinged flap). In enzymatically treated channels that lack fast inactivation, DPH still shows significant inhibitory effect (Quandt, 1988). Quantitative comparison between the voltage dependence of the recovery from fast inactivation and the voltage dependence of the recovery from DPH inhibition also argues that DPH binding to the Na⁺ channel does not require the presence of fast inactivation (Kuo and Bean, 1994b). Thus, the binding site for the anticonvulsants most likely is formed or is shaped into the "right" conformation during the activation processes of the channel [the "modulated receptor hypothesis" (Hille, 1977, 1993)]. Along with the findings that the anticonvulsant receptor is located on the extracellular side of the channel (Fig. 8), it seems that Na⁺ channel activation involves multiple conformational changes not only in the pore (to become conducting of Na⁺ ions) and on the cytoplasmic side (to bind the inactivating peptide or hinged flap) but also on the external side of the channel (to make the anticonvulsant receptor by realignment of side chains of some aromatic amino acids). The external conformational changes associated with channel gating also is supported by the previous finding that external application of some macromolecules impermeable to the membrane, such as antibodies against part of the primary sequence of Na⁺ channel, shifts the steady state inactivation curve to more negative potentials (Meiri et al., 1987).