The trapping block of NMDA receptor channels in acutely isolated rat hippocampal neurones

Abstract

1. N-methyl-d-aspartate (NMDA) receptor responses were recorded from acutely isolated rat hippocampal neurones using the whole-cell patch-clamp technique. A rapid perfusion system was used to study the voltage-dependent block of NMDA channels by Mg²⁺, amantadine (AM) and N-2-(adamantyl)-hexamethylenimine (A-7).

2. Mg²⁺, AM and A-7-induced stationary blockade of NMDA channels increased with the blocker concentration but did not depend on the agonist (aspartate; Asp) concentration. Blockade by AM and A-7, but not Mg²⁺, was weakly use dependent.

3. ‘Hooked’ tail currents were observed after coapplication of Asp and Mg²⁺, AM or A-7. The hooked tail current kinetics, amplitude and carried charge indicated that Mg²⁺, AM and A-7 did not prevent closure and desensitization of NMDA channels nor agonist dissociation.

4. Tail currents following Asp application in the absence and continuous presence of Mg²⁺, AM or A-7 had similar kinetics.

5. Application of multiple stationary and kinetic criteria to the Mg²⁺, AM and A-7 blockade led us to conclude that their effects on NMDA channels can be described in terms of a ‘trapping’ model, which is fully symmetrical with respect to the blocking transition.

6. In general, the apparent blocking/recovery kinetics predicted by the fully symmetrical trapping model differ significantly from the microscopic kinetics and depend on the rate of binding and unbinding of the blocker, the NMDA channel open probability and the rate of solution exchange.

The N-methyl-d-aspartate (NMDA) subtype of glutamate receptor plays an important role both in physiological and pathophysiological processes occurring in the brain (Dingledine et al. 1999). Channel block is one of the characteristic features of its complex regulation. The first studies of NMDA receptors revealed a strong voltage-dependent block by physiological concentrations of Mg²⁺ ions (Mayer et al. 1984; Nowak et al. 1984). The effects of many drugs used in clinical practice for the treatment of a broad range of neurological disorders are based on NMDA channel blockade (Parsons et al. 1998). The most effective and tolerant drugs, such as memantine and amantadine, can be ‘trapped’ within the channel following dissociation of the agonist from the receptor (Blanpied et al. 1997; Chen & Lipton, 1997). When applied externally, these drugs can enter the open NMDA channel and bind to the ‘blocking site’ located deep within the pore. Occupancy of this site, however, does not prevent channel closure and the blocking molecule can remain within the pore for a relatively long time, being trapped by the closed activation gate. The subsequent opening of the gate allows the trapped blocker to leave the channel.

NMDA channels are blocked via the trapping mechanism by MK-801, ketamine and phencyclidine and its structural analogue N-ethyl-1,4,9,9α-tetrahydro-4αR-cis-4αH-fluoren-4α-amine (NEFA) (Huetter & Bean, 1988; MacDonald et al. 1991; Dilmore & Johnson, 1998). All these compounds have slow binding/unbinding kinetics and their trapping manifests itself in use dependence of both blockade and recovery from it (Neely & Lingle, 1986). However, trapping is not apparent for blockers which bind to and escape from the channel with kinetics which are faster than the channel opening and closure, respectively. Thus, trapping was a possible, but far from only, explanation for the effect of Mg²⁺ ions on NMDA channels (Ascher & Nowak, 1988; Jahr & Stevens, 1990).

Recently, a number of criteria have been developed in order to define the effect of fast blockers on the NMDA channel gating machinery (Sobolevsky et al. 1999a). In the present study we used these criteria to demonstrate that the physiological blocker of NMDA channels, Mg²⁺, the well-known antiparkinsonian and neuroprotective agent, amantadine (AM; Fig. 1), and a novel perspective antiparkinsonian drug, N-2-(adamantyl)-hexamethylenimine (A-7, Fig. 1; Valdman et al. 1999) can be attributed to the class of trapping blockers.

Figure 1.

The chemical structures of 1-adamantanamine (AM) and N-2-(adamantyl)-hexamethylenimine (A-7).

METHODS

All experiments were performed in compliance with standards for use of laboratory animals and were approved by the USSR Academy Commission on Laboratory Animals Usage Control. Two- to 4-week-old Wistar rats were killed by cervical dislocation. Hippocampal slices were prepared according to the procedure described by Vorobjev (1991). Pyramidal neurones were mechanically isolated from the CA1 region of the slice by vibrodissociation (Vorobjev, 1991). The experiments were begun no earlier than 3 h after incubation of the hippocampal slices in a medium containing (mM): NaCl, 124; KCl, 3; CaCl₂, 1.4; MgCl₂, 2; glucose, 10; NaHCO₃, 26. The solution was bubbled with carbogen and maintained at 32°C. During the whole period of isolation and current recording, nerve cells were washed with a Mg²⁺-free solution comprising (mM): NaCl, 140; KCl, 5; CaCl₂, 2; glucose, 15; Hepes, 10; pH 7.3. Glycine at the saturating concentration of 3 μm was always added to this solution. All the drugs were dissolved in water. Concentrated drug stock solutions were prepared and kept frozen until use. Fast replacement of solutions was achieved using a simple perfusion system (Vorobjev et al. 1996). The time constant of the solution exchange (τ_wash), measured by the method of sodium concentration jumps (Vyklickýet al. 1990; Chen & Lipton, 1997), was 20-30 ms. The currents were recorded at 18°C in the whole-cell configuration using micropipettes made from Pyrex tubes and filled with an ‘intracellular’ solution comprising (mM): CsF, 140; NaCl, 4; Hepes, 10; pH 7.2. The electrical resistance of the filled micropipettes was 3-7 MΩ. The analog current signals were digitized at 2 kHz and filtered at 1 kHz.

Statistical analysis was performed with the aid of the scientific and technical graphics computer program Microcal Origin 4.1 (Northampton, MA, USA). All the data are presented as means ±s.e.m. (n, number of cells), and comparisons were done by ANOVA, with P < 0.05 taken as significant.

The voltage dependence of the degree of stationary current inhibition (1 –I_BS/I_CS), where I_CS and I_BS are the stationary control and blocked currents, respectively (see Fig. 2A), was fitted according to the Woodhull model (Woodhull, 1973) using the following equation:

(1)

Figure 2. Voltage dependence of the A-7-induced stationary current inhibition.

A, experimental traces. Asp (100 μm) was applied alone (left) or was coapplied with 50 μm A-7 (right) for 3 s at different membrane holding potentials (V_h): -100, -80, -60, -40, -20, 20 and 40 mV. The degree of stationary current inhibition (1 –I_BS/I_CS) diminished with membrane depolarization. Inset, control and blocked stationary current-voltage curves. B, mean 1 –I_BS/I_CS plotted against V_h. Fitting with eqn (1) (continuous line) gave the following parameter values: δ= 0.76 ± 0.02; K_0.5(0) = 191 ± 10 μm (n = 5).

where [B] is the concentration of the blocker, V_h is the membrane holding potential, K_0.5(0) is the equilibrium dissociation constant at a V_h of 0 and δ is the fraction of the electric field sensed by the blocker at the binding site. z, F, R and T are the charge of the blocker, Faraday's constant, the gas constant and absolute temperature, respectively.

The dependencies of the degree of stationary current inhibition (1 –I_BS/I_CS) and the amplitude of the hooked tail current ((I_P–I_BS)/I_CS), where I_P is the maximum value of the hooked tail current (see Fig. 4A), on the blocker concentration were fitted by the following logistic equation:

(2)

where F([B]) is 1 –I_BS/I_CS or (I_P–I_BS)/I_CS, A is a constant, [B]₀ is the blocker concentration resulting in 50 % effect and n_H is the Hill coefficient describing the steepness of the fit.

Figure 4. Concentration dependence of Mg²⁺-induced blockade.

A, superposition of currents in response to Asp (100 μm) alone and its coapplication with Mg²⁺ at different concentrations (1.2, 3.7, 11, 33 and 100 μm). Top panel, experimental currents. Bottom panel, currents predicted by Model 1. B, the dependencies of the stationary current inhibition (1 –I_BS/I_CS) and the amplitude of the hooked tail current ((I_P–I_BS)/I_CS) on Mg²⁺ concentration. ○, experimental values of 1 –I_BS/I_CS; •, experimental values of (I_P–I_BS)/I_CS. The continuous lines represent the fit of the experimental data by eqn (2) with the following parameter values: A = 1, [B]₀= 7.8 ± 0.6 μm and n_H= 0.81 ± 0.05 (n = 6) for 1 –I_BS/I_CS; A = 0.26 ± 0.05, [B]₀= 11.2 ± 6.3 μm and n_H= 1.02 ± 0.39 (n = 6) for (I_P–I_BS)/I_CS. The dashed lines represent the fit of the modelling data by eqn (2) with the following parameter values: A = 1, [B]₀= 7.9 ± 0.1 μm and n_H= 0.98 ± 0.01 for 1 –I_BS/I_CS; A = 0.20 ± 0.01, [B]₀= 5.6 ± 0.1 μm and n_H= 1.77 ± 0.02 for (I_P–I_BS)/I_CS. C, dependence of the normalized plateau/peak ratio ((I_BS/I_B0)/(I_CS/I_C0)) on the stationary current inhibition (1 –I_BS/I_CS). •, experimental values of (I_BS/I_B0)/(I_CS/I_C0). The continuous line is the prediction of Model 1. Note that (I_BS/I_B0)/(I_CS/I_C0) is lower than unity. The values of the rate constants for Model 1 are the same as those given in the legend to Fig. 8A.

According to the simple bimolecular reaction model, the apparent binding (k_on,app) and unbinding (k_off,app) rate constants of the blocker were defined using the following equations:

(3)

(4)

where τ_on and τ_off are the time constants for channel blockade and recovery, respectively.

The kinetic model used to simulate the blocking action was based on the conventional rate theory and utilized independent forward and reverse rate constants for simultaneous solution of first-order differential equations which represented the transitions between all possible states of the channel. The processes of NMDA channel activation, opening and desensitization were described in terms of the kinetic model proposed by Lester & Jahr (1992). The choice of values of the kinetic constants was made as described previously (Sobolevsky & Koshelev, 1998). Thus, the kinetic constants for agonist binding (l₁) and unbinding (l₂) were taken as 2 μm⁻¹ s⁻¹ and 25 s⁻¹, respectively (Benveniste & Mayer, 1991). The kinetic constant of the channel closure (α) was adopted as 200 s⁻¹ based on the measurements of the mean open time in single-channel studies. The kinetic constant of the channel opening (β) was accepted as 8.45 s⁻¹ (Sobolevsky, 2000). This β value stipulates the value of the NMDA channel open probability (P_o=β/(α+β) = 0.04). The time constant of the solution exchange was accepted as 25 ms. The rate constants for the entrance and recovery from desensitization, γ and ε, respectively, were defined by analysing the control aspartate (Asp)-induced currents (see Appendix A in Sobolevsky & Koshelev, 1998). As the degree and the time constant of the desensitization decay varied little with the time of the current recording, γ and ε were considered as constants for each cell. The initially unknown values of the rate constants of the blocker binding and unbinding, k_on and k_off, respectively, were estimated.

Differential equations were solved numerically using the algorithm analogous to that described previously (Benveniste et al. 1990).

Amantadine was synthesized by MERZ (Frankfurt-am-Main, Germany). N-2-(adamantyl)-hexamethylenimine was synthesized at the Institute of Pharmacology (Moscow, Russia). Glycine was obtained from Serva (Heidelberg, Germany). Aspartate was purchased from Sigma (St Louis, MO, USA).

RESULTS

Voltage-dependent block of NMDA channels by Mg²⁺, AM and A-7

Ionic currents through NMDA channels were elicited by fast application of 100 μm Asp in the continuous presence of 3 μm glycine. At the membrane holding potential (V_h= -100 mV), Asp induced an inward current which, after an initial fast rise to a peak value (I_C0) decreased gradually (decay time constant (τ_D) = 517 ± 43 ms, n = 22) to a certain plateau level (I_CS). Such current decay during continued action of the agonist is considered to be due to desensitization of the NMDA receptor-channel complex. The fraction of desensitized channels (d = 1 – I_CS/I_C0) varied over a wide range (0.07-0.81) among the cells studied.

Mg²⁺, A-7 and AM, when coapplied with Asp, inhibited NMDA currents in a strongly voltage-dependent manner. The current responses to Asp and to Asp coapplied with A-7 (50 μm) at V_h values varying from -100 to 40 mV (in 20 mV increments) are shown in Fig. 2A (the control currents before and after A-7 treatment practically coincided). Current-voltage curves for the control and inhibited stationary current are shown in the inset. The degree of stationary current inhibition (1 –I_BS/I_CS) diminished with membrane depolarization (Fig. 2B). The voltage dependence of the A-7-induced stationary current inhibition was fitted by eqn (1) with δ= 0.76 ± 0.02 and K_0.5(0) = 191 ± 10 μm (n = 5). The values of δ and K_0.5(0) estimated for AM and Mg²⁺ block are given in Table 1 and are consistent with previously reported values (δ= 0.72 and K_0.5(0) = 261 μm for AM: Blanpied et al. 1997; δ= 0.8-1.0 and K_0.5(0) = 1.8-8.8 mM for Mg²⁺: Ascher & Nowak, 1988; Jahr & Stevens, 1990).

Table 1.

Kinetic, concentration- and voltage-dependence parameters

	Voltage dependence		Concentration dependence		Kinetics
Compound	K0.5(0)	δ	IC50	nH	Kon	Koff
	(μm)		(μm)		(μm−1s−1)	(s−1)
Mg2+	6817 ± 1677	0.78 ± 0.03	7.8 ± 0.6	0.81 ± 0.05	> 100	> 1000
AM	694 ± 17	0.90 ± 0.01	14.5 ± 4.4	1.02 ± 0.13	12.0 ± 4.9	175 ± 18
A-7	191 ± 10	0.76 ± 0.02	11.8 ± 0.6	1.09 ± 0.06	13.5 ± 2.5	159 ± 21

Values are means ±s.e.m., n = 4–11.

The voltage-dependent inhibition of NMDA responses by Mg²⁺, A-7 and AM indicates that the mechanism of their action consists in channel block. The experiments described below were carried out at a V_h of -100 mV.

Weak use dependence for AM and A-7 block and use independence for Mg²⁺ block

An illustrative example of currents obtained in response to 2 s consecutive AM (250 μm) plus Asp (100 μm) coapplications is shown in Fig. 3A. Independent of the time interval between the coapplications (1-10 s), the initial current in response to the first coapplication (I_B0), which was much smaller than the initial control current (I_C0), was always greater than the initial current in response to the second coapplication (I₁). The initial current in response to the third coapplication did not differ from I₁ (not shown). The level of stationary current (I_BS) for all the responses was the same indicating that, during the 2 s coapplication, AM blockade of the NMDA channels reached saturation. Recovery from AM block was achieved by repetitive applications of Asp. The initial current obtained in response to the first Asp application (I₂), independent of the time interval between the beginning of the Asp application and the termination of the last Asp plus AM coapplication (10-40 s), was smaller than the initial control current (I_C0) but much greater than the stationary blocked current (I_BS). The initial current in response to the second Asp application 1-10 s later did not differ from I_C0.

Figure 3.

Use dependence of the AM-induced blockade and use independence of the Mg²⁺-induced blockade

A, responses to consecutive 2 s application of Asp (100 μm), two Asp and AM (250 μm) coapplications and two Asp applications. Note that I₁ and I₂ are smaller than I_B0 and I_C0, respectively. B, responses to consecutive 2 s application of Asp (100 μm), two Asp plus Mg²⁺ (100 μm) coapplications and one Asp application. Note that I₁ and I₂ are the same as I_B0 and I_C0, respectively. The insets show the control tail currents (c) superimposed on the tail currents obtained after coapplication of the agonist and blocker (b).

The weak use dependence for both the AM block and the recovery from it shown in Fig. 3A confirms the conclusion of Blanpied et al. (1997) that after agonist removal from the external solution some portion of the NMDA channels can close with the blocker inside. An analogous use dependence was observed for A-7 (not shown).

In contrast to AM and A-7, blockade by Mg²⁺ and recovery from it were not use dependent (Fig. 3B). The initial blocked currents (I_B0 and I₁) were identical and the initial current in response to Asp application after its coapplication with Mg²⁺ (I₂) was indistinguishable from the control (I_C0).

Therefore, in contrast to MK-801, ketamine and phencyclidine, the weak use dependence or its absence does not allow us to conclude that AM, A-7 and Mg²⁺ can be fully trapped within the NMDA channel. This is due to their fast binding/unbinding kinetics. Other criteria are needed in order to check whether the effects of Mg²⁺, AM and A-7 can be described by a kinetic model of trapping block: where C, D, and O designate the channel in the closeds, desensitized and open states, respectively. The subscripts A, AA and B indicate the binding of one agonist, two agonists and one blocker molecule to the channel, respectively. [A] and [B] are the agonist and the blocker concentrations, respectively. The conducting state is indicated with an asterisk. Model 1 implies that the blocker does not prevent channel closure and desensitization nor dissociation of the agonist from the blocked channel.

Model 1.

Hooked tail currents

The first criterion for determining whether the actions of Mg²⁺, A-7 and AM can be described by Model 1 is the kinetics of the hooked tail currents compared with those of the control currents. Hooked tail currents, normally observed only for fast blockers after agonist and blocker coapplication, are illustrated for AM and Mg²⁺ in Fig. 3. The superposition shown in the insets clearly demonstrates that hooked and control tail currents do not intersect. This finding is predicted by Model 1 (see, e.g. Fig. 4A, bottom panel). In the case of blockers which prevent channel closure, desensitization or agonist dissociation, hooked and control tail currents do intersect (Sobolevsky et al. 1999a). Therefore, the kinetics of the hooked tail currents observed for AM, A-7 and Mg²⁺ are in good correspondence with the predictions of Model 1.

The blocking action of Mg²⁺, A-7 and AM on NMDA channels depends on their concentration. A representative superposition of control and blocked currents at different Mg²⁺ concentrations is shown in Fig. 4A (top panel). The value of the stationary current inhibition (1 –I_BS/I_CS) increased with Mg²⁺ concentration (Fig. 4B) and was well fitted by eqn (2) with the following parameter values: A = 1, IC₅₀=[B]₀= 7.8 ± 0.6 μm and n_H= 0.81 ± 0.05 (n = 6). The parameters of the stationary current inhibition for AM and A-7 are given in Table 1. The values of the half-blocking concentrations for AM and Mg²⁺ estimated in our experiments are consistent with previously reported values (for AM: IC₅₀= 71 μm at -70 mV, Parsons et al. 1995; IC₅₀= 39 μm at -67 mV, Blanpied et al. 1997; for Mg²⁺: IC₅₀= 64 μm at -50 mV, Kupper et al. 1996; IC₅₀= 78 μm at -80 mV, Zhang et al. 1996; IC₅₀= 19 μm at -70 mV, Williams et al. 1998).

The value of the hooked tail current amplitude ((I_P–I_BS)/I_CS) also increased with Mg²⁺ concentration (Fig. 4B) and was well fitted by eqn (2) with the following values: A = 0.26 ± 0.05, [B]₀= 11.2 ± 6.3 μm and n_H= 1.02 ± 0.39 (n = 6). The hooked tail currents for AM and A-7 were distinguishable only at a high degree of stationary current inhibition (1 –I_BS/I_CS > 0.5) and their amplitudes were, on average, lower than that for Mg²⁺. For this reason, hooked tail currents were not observed in Fig. 2A at V_h values more positive than -60 mV. The amplitude of the hooked tail current for all the blockers under study was always lower than the value of the stationary current inhibition. This observation is in good agreement with predictions of the fully symmetrical trapping model. The currents simulated using Model 1 for Mg²⁺ are shown in Fig. 4A (bottom panel). Clearly, the amplitude of the hooked tail current was lower than the value of the stationary current inhibition. The same conclusion follows from simulation using Model 1 of AM and A-7 blockade (not shown). The dependencies of the stationary current inhibition and the hooked tail current amplitude on the blocker concentration predicted by Model 1 for Mg²⁺ proved to be qualitatively similar to those determined experimentally. Fitting by eqn (2) (dashed lines in Fig. 4B) gave the following values: A = 1, IC₅₀=[B]₀= 7.9 ± 0.1 μm and n_H= 0.98 ± 0.01 for 1 –I_BS/I_CS; and A = 0.20± 0.01, [B]₀= 5.6 ± 0.1 μm and n_H= 1.77 ± 0.02 for (I_P–I_BS)/I_CS.

Model 1 also predicts a rise in the amplitude of the hooked tail current with an increase in the rate of solution exchange (see also Sobolevsky et al. 1999a). Thus, a decrease in τ_wash from 25 to 1 ms leads to a 3-fold increase in (I_P–I_BS)/I_CS. However, even at an infinitely fast solution exchange (τ_wash= 0), the amplitude of the hooked tail current predicted by Model 1 is lower than the stationary current inhibition and approaches it at an infinitely high value of the blocker unbinding rate constant (k_off). In other words, for Model 1, I_P≤I_CS.

If I_P > I_CS, the blocking action could not be described by Model 1 (Sobolevsky et al. 1999a). Thus, for 9-aminoacridine which prohibited both agonist and coagonist dissociation (Benveniste & Mayer, 1995), for tetrabutylammonium which prohibited desensitization (Koshelev & Khodorov, 1992, 1995; Sobolevsky et al. 1999a; Sobolevsky, 2000) and for tetrapentylammonium which prohibited NMDA channel closure (Sobolevsky et al. 1999a), the amplitude of the hooked tail current exceeded the stationary current inhibition. Therefore, the low amplitudes of hooked tail currents are consistent with the description by Model 1 of Mg²⁺, A-7 and AM blockade.

A study of the electric charge carried during the hooked tail current provides important information about the interaction of the blocker with the NMDA channel gating machinery (Sobolevsky, 1999a, 2000). If the electric charge, measured by integrating the current curve (Fig. 5A), carried during the hooked tail current (Q_hook) is greater than that carried during the control tail current (Q_control), the blocker prevents the dissociation of the agonist from the channel. Vice versa, if Q_hook < Q_control, the blocker does not prevent agonist dissociation. In the former case, the ratio Q =Q_hook/Q_control is greater than unity but in the latter, Q < 1. Figure 5B shows the Q values for Mg²⁺, A-7 and AM plotted against stationary current inhibition. For all the blockers under study, Q was always lower than unity and decreased with 1 –I_BS/I_CS. For AM and A-7, which possess slower binding/unbinding kinetics, Q was lower than that for Mg²⁺. This finding is in good agreement with the predictions of Model 1. The dependence of Q on stationary current inhibition predicted by Model 1 is shown in Fig. 5B for Mg²⁺ (continuous line) and A-7 (dashed line). The dependence of Q on stationary current inhibition predicted by Model 1 for AM was practically the same as that for A-7 (not shown). Therefore, the Q dependence for all the three blockers under study is consistent with the predictions of trapping Model 1 and these blockers most probably do not prevent agonist dissociation from the blocked channel.

Figure 5. The electric charge carried during the hooked tail current.

A, experimental currents in response to Asp (100 μm) alone and its coapplication with A-7 (100 μm). The filled areas represent the electric charge carried during the control (Q_control) and hooked (Q_hook) tail currents. B, dependence of the normalized value of the electric charge carried during the hooked tail current (Q =Q_hook/Q_control) on the stationary current inhibition (1 –I_BS/I_CS). The continuous and dashed lines are the predictions of Model 1 for Mg²⁺ and A-7, respectively. Note that Q is lower than unity. The values of the rate constants for Model 1 are the same as those given in the legend to Fig. 8A.

The plateau/peak ratio

The plateau/peak ratio criterion allows one to distinguish fast blockers which prevent channel desensitization from those which do not (Sobolevsky et al. 1999a). If the normalized plateau/peak ratio ((I_BS/I_B0)/(I_CS/I_C0)) is lower than unity, the blocker does not prevent NMDA channel desensitization and vice versa, if (I_BS/I_B0)/(I_CS/I_C0) > 1, the blocker prevents desensitization. (I_BS/I_B0)/(I_CS/I_C0) for Mg²⁺ was lower than unity at all blocker concentrations. Its dependence on stationary current inhibition (Fig. 4C) is consistent with the prediction of Model 1 (continuous line). These data demonstrate that Mg²⁺ probably does not prevent NMDA channel desensitization.

The plateau/peak ratio criterion is the only one used in the present study which is applicable to Mg²⁺ but not to AM and A-7. The reason for this lies in the slower binding/unbinding kinetics for AM and A-7 than for Mg²⁺. These kinetics to a large extent stipulate the current decline during Asp coapplication with AM or A-7 (see below).

Tail currents in the continuous presence of the blocker

Further, we compared the kinetics of the control tail current (c) and the tail current after termination of agonist application in the continuous presence of the blocker (b; Fig. 6A). The superposition of ‘b’ and ‘c’ currents for Mg²⁺, A-7 and AM (Fig. 6B) shows that the kinetics of these currents were very similar. This became especially evident when the b and c curves were normalized (Fig. 6B, insets). In agreement with the experimental data, Model 1 predicts that the normalized b and c curves coincide. Tail currents simulated using Model 1 for A-7 are shown in Fig. 6C. The currents predicted by Model 1 for Mg²⁺ and AM were very similar (not shown). The absence of blocker-induced delay in the tail current kinetics indicates that Mg²⁺, A-7 and AM do not prevent agonist dissociation from the blocked channel (Sobolevsky et al. 1999a).

Figure 6. Tail currents after termination of Asp application in the continuous presence of the blocker.

A, experimental protocol with Mg²⁺ as an example. Asp (100 μm) was applied for 2 s in the control external solution (left) or in the continuous presence of 10 μm Mg²⁺ (right). B, control tail currents (c) superimposed with tail currents obtained in the continuous presence of A-7 (15 μm), AM (10 μm) or Mg²⁺ (10 μm) (b). Each inset shows the superposition of the normalized curves c and b. Note that the normalized curves coincide for all the blockers under study. C, tail currents predicted by Model 1 for A-7 (15 μm). The values of the rate constants are the same as those given in the legend to Fig. 8A.

Agonist independence of the stationary current inhibition

A representative superposition of currents in response to Asp application and its coapplication with Mg²⁺ (20 μm) at different Asp concentrations is shown in Fig. 7A. As seen, the degree of Mg²⁺-induced stationary current inhibition did not depend on the Asp concentration. The mean values of 1 –I_BS/I_CS for Mg²⁺ (20 μm), AM (15 μm) and A-7 (15 μm) in relation to the Asp concentration are shown in Fig. 7B. These values were not significantly different at different agonist concentrations (P > 0.5, n = 5 for Mg²⁺; P > 0.5, n = 5 for AM; P > 0.1, n = 8 for A-7). If the blocker prevents channel closure, desensitization or agonist dissociation, the agonist dependence of the stationary current inhibition will increase or decrease, but will never be constant (Sobolevsky et al. 1999a). In contrast, Model 1 predicts that 1 –I_BS/I_CS does not depend on the agonist concentration. Thus, the method described previously (Sobolevsky & Koshelev, 1998; Sobolevsky, 1999b) allows derivation of the following equation for the fully symmetrical trapping Model 1

Figure 7. Agonist independence of the stationary current inhibition.

A, superposition of currents in response to 3 s Asp application and Asp plus 20 μm Mg²⁺ coapplication at different Asp concentrations (6.25, 12.5, 25, 50 and 100 μm). B, mean values of the degree of stationary current inhibition (1 –I_BS/I_CS) for Mg²⁺ (20 μm), A-7 (15 μm) and AM (15 μm) did not depend on the Asp concentration.

(5)

Therefore, the independence of Mg²⁺-, AM- and A-7-induced blockade of agonist concentration is consistent with the description of their effects on the NMDA channel by Model 1.

Kinetics in the continuous presence of the agonist

The experimental protocol is shown in Fig. 8A with A-7 (10 μm) as an example. The blocker was applied when the Asp-induced current had already reached its stationary level (I_CS). The recovery kinetics after termination of the blocker application in the continuous presence of the agonist for Mg²⁺, A-7 and AM did not show any ‘overshoot’. Overshoot occurs when the recovery current exceeds the stationary control level (I_CS); it has been observed for tetrabutylammonium, tetrapentylammonium, tacrine, 9-aminoacridine and IEM-1754 (Koshelev & Khodorov, 1992, 1995; Costa & Albuquerque, 1994; Antonov et al. 1995; Sobolevsky, 1999a,b, 2000; Sobolevsky et al. 1999a). The appearance of an overshoot in the recovery kinetics indicates that the blocker prevents NMDA channel desensitization (Sobolevsky et al. 1999a). Therefore, Mg²⁺, A-7 and AM most probably do not prevent desensitization, which is consistent with the description of their action by Model 1.

Figure 8. Blocking/recovery kinetics in the continuous presence of the agonist.

A, experimental protocol. A-7 (10 μm) was applied for 2 s in the continuous presence of Asp (100 μm) when the inward current reached its stationary level (I_CS). The continuous line shows the prediction of Model 1 with k_on= 11 μm⁻¹ s⁻¹, k_off= 130 s⁻¹, [B]= 10 μm, γ= 0.59 s⁻¹ and ε= 0.96 s⁻¹. The inset shows the current in response to 2 s Mg²⁺ (100 μm) application in the continuous presence of Asp (100 μm). The continuous line shows the prediction of Model 1 with k_on= 1282 μm⁻¹ s⁻¹, k_off= 10000 s⁻¹, [B]= 100 μm, γ= 1.19 s⁻¹ and ε= 0.81 s⁻¹. B, inverse time constants for A-7-induced blockade (1/τ_on) and recovery from it (1/τ_off) plotted against A-7 concentration. The linear fits to eqns (3) and (4) (continuous lines) allow estimation of the apparent binding (k_on,app= 0.42 ± 0.04 μm⁻¹ s⁻¹) and unbinding (k_off,app= 4.7 ± 0.1 s⁻¹; n = 11) rate constants for A-7.

Therefore, we used multiple criteria to demonstrate that the effects of AM, A-7 and Mg²⁺ can be described by Model 1. To complete the description, we estimated the unknown values of the rate constants of blocker binding and unbinding, k_on and k_off, respectively.

The fitting of blocking and recovery kinetics for AM and A-7 by single exponential functions showed that the blocking rate increased with the blocker concentration, while the rate of recovery did not depend on the blocker concentration. This is illustrated in Fig. 8B where the inverse time constants for the A-7-induced blockade (1/τ_on), and recovery from it (1/τ_off), are plotted against the A-7 concentration. 1/τ_on significantly correlated with the A-7 concentration (P < 10⁻⁵), while 1/τ_off did not (P > 0.6). In this situation, it was tempting to use a simple bimolecular reaction model to define the apparent binding (k_on,app) and unbinding (k_off,app) rate constants for A-7. The linear regression for 1/τ_on in accordance with eqn (3) allowed us to calculate values of k_on,app and k_off,app of 0.42 ± 0.04 μm⁻¹ s⁻¹ and 4.4 ± 0.6 s⁻¹, respectively (n = 11). According to eqn (4), the value of k_off,app was 4.7 ± 0.1 s⁻¹ (n = 11). Calculation of the k_off,app/k_on,app ratio yielded an apparent K_d,app value of 11.0 μm, which is comparable to the IC₅₀ value (11.8 μm) obtained from the concentration dependence of the A-7 stationary block (Table 1).

Analogous calculations for AM revealed values for k_on,app and k_off,app of 0.35 ± 0.04 μm⁻¹ s⁻¹ and 5.3 ± 0.2 s⁻¹, respectively (n = 6). The k_off,app/k_on,app ratio was 15.2 μm and, as in the case of A-7, was also comparable to the IC₅₀ (14.5 μm) for the concentration dependence of AM stationary block (Table 1).

From eqn (5):

(6)

A good correlation between K_d,app and IC₅₀ for both A-7 and AM tempts one to identify k_on,app and k_on, k_off,app and k_off as well as K_d,app and K_d. We checked this possibility by simulating currents for an experiment in which the blocker was applied in the continuous presence of the agonist. The simulations were carried out under the conditions of eqn (6) correctness by varying the value of k_off. The result of such a simulation for A-7 superimposed with the experimental curve is shown in Fig. 8A. The good correspondence of the experimental and modelling data was disturbed only for the recovery of the current after termination of the Asp application (the experimental recovery contained a slow component which is not practically resolved in the modelling recovery, see also Fig 4A (bottom panel) and Fig 6C). This discrepancy is not surprising because the activation model used in the present study (Lester & Jahr, 1992) is simple and cannot reproduce many of the NMDA receptor properties described in single-channel studies (see, e.g. Gibb & Colquhoun, 1992). Thus, the existence of a slow component in the control current relaxation can be explained, for example, by more complex NMDA receptor desensitization (Sather et al. 1992) or by the infringement of the principle of independence in the binding of two agonist molecules to the receptor (Benveniste & Mayer, 1991). Simulations analogous to that shown in Fig. 8A allowed us to calculate the following parameter values for A-7: k_on= 13.5 ± 2.5 μm⁻¹ s⁻¹ and k_off= 159 ± 21 s⁻¹ (n = 11). The values of the binding and unbinding rate constants for AM proved to be very similar (Table 1). Our value of the binding rate constant for AM was close to that estimated by Blanpied et al. (1997) (k_on= 14.9 μm⁻¹ s⁻¹), while our value of the unbinding rate constant was 14 times lower than their estimate (k_off= 2520 s⁻¹).

Therefore, despite the equality, K_d,app=K_d, appearing to be valid for A-7 and AM, the values of the apparent rate constants k_on,app and k_off,app differed significantly from the microscopic kinetic constants k_on and k_off, respectively.

Compared with A-7 and AM block, the kinetics of Mg²⁺ block were much faster (Fig. 8A, inset) and were determined mainly by the rate of the solution exchange. Thus, the time constants of the Mg²⁺-induced blockade and recovery from it were about 20-30 and 30-50 ms, respectively, and did not depend on Mg²⁺ concentration. If the rate of the current recovery reflected the microscopic dissociation of Mg²⁺ from the channel, the value of k_off would be 20-30 s⁻¹. However, simulations of experimental kinetics using Model 1 showed that the k_off value should be much higher than 1000 s⁻¹. The result of such simulation with k_on of 1282 μm⁻¹ s⁻¹ and k_off of 10000 s⁻¹ superimposed with the experimental curve is shown in Fig. 8A (inset). The assumption that Mg²⁺ has a high unbinding rate constant was confirmed by the recent study of Antonov & Johnson (1999) where a k_off value of 16493 s⁻¹ was reported at a membrane potential of -100 mV.

Therefore, as in the case of A-7 and AM, an essential difference was observed between the apparent and microscopic kinetics of Mg²⁺ block. There are two reasons why the simple bimolecular reaction model is not valid for the calculation of microscopic rate constants using Model 1, namely: (1) the non-instantaneous solution exchange, and (2) P_o lower than 1.

For fast blockers with k_off > 1000 s⁻¹ (e.g. Mg²⁺), the apparent blocking/recovery kinetics strongly depend on the time constant of the solution exchange, τ_wash (Fig. 9A), but not on P_o. In contrast, the apparent blocking/recovery kinetics for slow blockers with k_off < 1 s⁻¹ (e.g. ketamine, phencyclidine and MK-801) strongly depend on P_o (Fig. 9B), but not on τ_wash. The apparent blocking/recovery kinetics for blockers with intermediate k_off values of ∼100 s⁻¹ (e.g. AM and A-7) depend on both P_o and τ_wash. This is clearly illustrated in Fig. 9C-E where the ratio of microscopic and apparent dissociation rate constants (k_off/k_off,app) for blockers with fast, intermediate and slow kinetics is plotted against P_o at two different values of τ_wash. The behaviour of the k_on/k_on,app ratio was very similar (not shown).

Figure 9. The dependence of the apparent blocking/recovery kinetics predicted by Model 1 on the microscopic binding/unbinding kinetics, channel open probability and the rate of solution exchange.

A, currents predicted by Model 1 for a fast blocker with k_off= 10000 s⁻¹ at different values of the time constant of solution exchange (τ_wash; 1, 10 and 100 ms). B, currents predicted by Model 1 for a slow blocker with k_off= 0.1 s⁻¹ at different values of P_o (0.02, 0.09 and 0.5). C-E, ratio of microscopic and apparent dissociation rate constants (k_off/k_off,app) as a function of P_o for blockers with fast (k_off= 10000 s⁻¹; C), intermediate (k_off= 130 s⁻¹; D) and slow (k_off= 0.1 s⁻¹; E) kinetics at two different values of τ_wash (1 and 100 ms).

Figure 9 shows that the apparent and microscopic kinetics predicted by Model 1 coincide only if (1) the solution exchange is much faster than the rates of the blocker binding and unbinding, and (2) P_o is equal to 1. For NMDA channels, these conditions are never accomplished simultaneously because P_o is always lower than 1 (Huetter & Bean, 1988; Jahr, 1992; Benveniste & Mayer, 1995; Rosenmund et al. 1995; Dzubay & Jahr, 1996; Sobolevsky, 2000). Therefore, even for slow blockers, such as ketamine, phencyclidine and MK-801, for which the first condition is accomplished, the apparent kinetics are much slower than the microscopic kinetics due to the low open probability of the NMDA channel.

DISCUSSION

Trapping block of NMDA channels by Mg²⁺, AM and A-7

The possibility that blocking particles can be trapped within the NMDA channel was reported previously for MK-801, ketamine, phencyclidine, NEFA, memantine, tetraethylammonium, tetrapropylammonium and tetrabutylammonium (Huetter & Bean, 1988; MacDonald et al. 1991; Blanpied et al. 1997; Chen & Lipton, 1997; Dilmore & Johnson, 1998; Sobolevsky et al. 1999a). In contrast to trapping blockers, so-called ‘sequential’ blockers prevent the closure of the NMDA channel activation gate. 9-Aminoacridine, tacrine, long-chain adamantane derivatives and tetrapentylammonium are believed to belong to this second class of blockers (Costa & Albuquerque, 1994; Vorobjev & Sharonova, 1994; Benveniste & Mayer, 1995; Koshelev & Khodorov, 1995; Antonov & Johnson, 1996; Sobolevsky et al. 1999a; Sobolevsky, 2000).

Previously, trapping was considered as a possible explanation for the action of Mg²⁺ on NMDA channels in single-channel studies (Ascher & Nowak, 1988; Jahr & Stevens, 1990). Indeed, the results of these studies are in good agreement with the predictions of trapping Model 1. Thus, Model 1 predicts the absence of the Mg²⁺-induced increase in burst duration observed by Ascher & Nowak (1988).

The possibility that AM might be trapped within the NMDA channel was reported in the study of Blanpied et al. (1997). However, they did not resolve whether the trapping was partial or complete and whether AM, when bound to the blocking site, altered NMDA channel gating or the interaction of the agonist with the receptor. These issues remained unresolved because Blanpied et al. (1997) used an experimental protocol which was analogous to that shown in Fig. 3. Due to the fast binding/unbinding kinetics of AM, this protocol always demonstrates a macroscopically incomplete trapping and does not allow precise estimation of the fraction of channels in which the blocker is trapped.

Although the action of all the blockers used in the present study was consistent with the fully symmetrical trapping model of open-channel block, these blockers may in addition bind to the NMDA channel not only in the open but also in the closed state. Thus, along with trapping block, AM induced NMDA channel blockade without any agonist assistance (Blanpied et al. 1997; Sobolevsky et al. 1998). However, this type of AM-induced blockade was slower and 100 times less effective than the open-channel block (Blanpied et al. 1997). The same conclusion can be drawn for A-7 (authors’ unpublished observation). Therefore, the overwhelming majority of channels blocked by AM and A-7 in our experiments was blocked via a trapping mechanism. In contrast, Mg²⁺, which is thought to be an open NMDA channel blocker (Ascher & Nowak, 1988), has fast binding/unbinding kinetics and cannot be tested in experiments which are usually used for studying the effect of the blocker in the absence of agonist (Blanpied et al. 1997; Sobolevsky et al. 1998). Another approach is required to elucidate whether Mg²⁺ is able to block closed NMDA channels.

Mg²⁺, AM and A-7 do not prevent NMDA channel desensitization

We concluded that Mg²⁺, AM and A-7 do not prevent NMDA channel desensitization. Other trapping blockers may interact with NMDA channel desensitization in different ways. This can be illustrated by the example of tetraalkylammonium compounds which, similar to Mg²⁺, AM and A-7, can be attributed to the class of fast NMDA channel blockers. Thus, small tetraalkylammonium compounds, tetraethylammonium and tetrapropylammonium, do not prevent NMDA channel desensitization, while the larger trapping blocker, tetrabutylammonium, does (Sobolevsky et al. 1999a).

It is unclear how trapping blockers with slow binding/unbinding kinetics (MK-801, ketamine, phencyclidine, NEFA and memantine) affect NMDA channel desensitization. Almost all the kinetic criteria used in the present study to classify the effect of the blocker on NMDA channel desensitization are applicable only to fast blockers (Sobolevsky et al. 1999a). The only criterion which can be applied to slow blockers is the agonist dependence of the stationary current inhibition. However, additional criteria are required to reliably discriminate the effects of slow blockers on NMDA channel desensitization. The major difficulty in finding kinetic criteria for such blockers lies in the fact that their binding/unbinding kinetics are much slower than the rates of entrance to and recovery from desensitization, thus being rate-limiting for microscopic kinetics.

Binding of trapping blockers within the NMDA channel pore

High δ values for Mg²⁺, AM and A-7 (Table 1) indicate that these blockers bind to the same or at least overlapping blocking sites in the depth of the channel pore. Other NMDA channel trapping blockers most probably bind to the same blocking site(s). This conclusion is supported by the competition between block by Mg²⁺ and MK-801 (Huetter & Bean, 1988), Mg²⁺ and phencyclidine (Lerma et al. 1991), and Mg²⁺ and memantine (Chen et al. 1992; Sobolevsky et al. 1998).

Recently, it has been hypothesized that the NMDA receptor channel is endowed with a desensitization gate along with an activation gate (Sobolevsky et al. 1999a). Within the frame of this hypothesis, Mg²⁺, AM and A-7 bind in the depth of the channel pore and do not prevent the closure of the activation and desensitization gates above them.

The blocking site of Mg²⁺ is located at the narrow constriction of the NMDA channel pore and is formed by adjacent NR2-subunit asparagines, the N-site and N + 1 site (Wollmuth et al. 1998). The δ values for Mg²⁺ (0.78) and A-7 (0.76) proved to be slightly lower than that for AM (0.90). This can be explained under the stipulation that the charged amino group of amantadine penetrates through the narrow constriction (with the bulky adamantane head remaining shallower than the narrow constriction). In contrast, the hydration shell of Mg²⁺ and the hexamethylenimine group of A-7 (Fig. 1) do not allow the charged centres of these blockers to occupy a deeper position within the channel pore.

Apparent and microscopic blocking kinetics

The simple bimolecular reaction model was used in many previous studies (e.g. Parsons et al. 1995; Chen & Lipton, 1997; Sobolevsky et al. 1999b) to estimate apparent binding (k_on,app) and unbinding (k_off,app) rate constants for trapping blockers. However, despite the similarity between K_d,app and IC₅₀ values, simulation of macroscopic blocking/recovery kinetics using trapping Model 1 showed that microscopic binding (k_on) and unbinding (k_off) rate constants essentially differ from their apparent analogues, k_on,app and k_off,app, respectively. The discrepancy between microscopic and macroscopic blocker binding/unbinding kinetics has been mentioned previously for slow trapping blockers (Huettner & Bean, 1988; MacDonald et al. 1991; Dilmore & Johnson, 1998).

Our results (Fig. 9) suggest that k_on and k_off microscopic rate constants can be determined from the apparent blocking/recovery kinetics when (1) the P_o of the NMDA channel is known and (2) the solution exchange is much faster than the apparent blocking/recovery kinetics.

The binding/unbinding kinetics of Mg²⁺ were too fast to be measured precisely in the present study. A faster perfusion system is needed to resolve this question using whole-cell recordings. Alternatively, the binding/unbinding kinetics of Mg²⁺ can be measured in single-channel studies. In previous single-channel studies the blocking effect of Mg²⁺ was described using very simplistic kinetic models (Ascher & Nowak, 1988; Jahr & Stevens, 1990). Analysis of single-channel records using a more complicated but realistic model, such as Model 1, may yield values of Mg²⁺ binding and unbinding rate constants which are essentially different from the previous estimates.

Functional implications

A-7, a novel potential antiparkinsonian drug which is now at the last stage of preclinical trials, is equally or more effective than AM and L-dopa in animal models of Parkinson's disease. Thus, similar to AM, A-7 significantly decreased the rigidity and oligokinesis caused by 1-methyl-4-phenyl-1,2,3,6-tetrapyridine (MPTP) in mice (Valdman et al. 2000). A-7 was more effective than AM and L-dopa in preventing catalepsy caused by neuroleptics and, in contrast to AM, decreased the duration of tremor induced by cholinergic substances (Valdman et al. 1999). A better pharmacological spectrum indicates that A-7 promises to be a more effective antiparkinsonian drug than those existing in clinical practice and, possibly, it will manifest a neuroprotective activity.

A-7 and AM blocked NMDA channels via a trapping mechanism with similar kinetics and concentration dependencies (Table 1). The difference in the pharmacological properties of these drugs can be attributed partly to the lower voltage dependence of A-7 but, more probably, is stipulated by different effects of AM and A-7 on other channels or receptors, especially dopaminergic receptors (Valdman et al. 1999).

The kinetics of an NMDA channel blocking drug are considered to be the main factor determining its clinical safety (Rogawski, 1993). Thus, clinical safety increases with the unblocking rate constant for (from the slowest to the fastest blocker) MK-801, phencyclidine, ketamine, memantine and amantadine (MacDonald et al. 1991; Chen et al. 1992; Parsons et al. 1995; Blanpied et al. 1997). The similar kinetics of A-7 and AM blockade (Table 1) raises hopes that clinically A-7 will be as safe as AM.

Why is Mg²⁺, an effective blocker of NMDA channels under physiological conditions, less effective than A-7 and AM in protecting against acute neurotoxic effects of glutamate, despite the obvious similarity in their mechanism of action (trapping block)? A possible answer to this question lies in the stronger voltage dependence observed for Mg²⁺ due to its charge (z = 2; Parsons et al. 1995). Under physiological conditions, high (millimolar) concentrations of glutamate are present in the synaptic cleft for a few milliseconds (Clements et al. 1992) inducing a pronounced depolarization of the cell membrane. In contrast, during acute excitotoxic insults, such as ischaemia, relatively low micromolar concentrations of glutamate can be present in the interstitial medium for much longer periods of time (Andine et al. 1991), thus inducing moderate but prolonged depolarization. During such moderate depolarization, the effectiveness of AM and, to a grater extent, A-7 blockade (the δ value for A-7 is smaller than that for AM) remains sufficiently high for occupation of the majority of NMDA channels by these compounds. At the same time, the strongly voltage-dependent Mg²⁺ block is partially relieved, thus allowing the flow of Ca²⁺ through NMDA channels. As a result, an increase in the intracellular concentration of Ca²⁺ ions triggers intracellular processes which ultimately lead to cell death.