333

Journal of Physiology (1990), 428, pp. 333-357 With 9 figures Printed in Great Britain

A KINETIC ANALYSIS OF THE MODULATION OF N-METHYL-DASPARTIC ACID RECEPTORS BY GLYCINE IN MOUSE CULTURED HIPPOCAMPAL NEURONES

BY MORRIS BENVENISTE, JOHN CLEMENTS*, LADISLAV VYKLICKY JRt AND MARK L. MAYER: From the Unit of Neurophysiology and Biophysics, Laboratory of Developmental Neurobiology, NICHD, Building 36, Room 2A21, National Institutes of Health, Bethesda, MD 20892, USA

(Received 17 October 1990) SUMMARY

1. Responses to N-methyl-D-aspartic acid (NMDA) were recorded from mouse embryonic hippocampal neurones in dissociated culture, using whole-cell patchclamp recording. A fast perfusion system, with an exchange time constant of less than 10 ms, was used to study modulation of NMDA receptor desensitization by glycine. 2. The onset of NMDA receptor desensitization was well fitted by a singleexponential function; with 30 nM-glycine the time constant was 250 ms, corresponding to a rate of 4 s-1. The rate of onset of desensitization became faster with increasing glycine concentration, with a slope of 0-87 x 107 M-1 s-1. Recovery from desensitization, studied with a twin-pulse technique, was also well fitted by a singleexponential function; with 30 nM-glycine the time constant of recovery was 1-95 s-'. The rate of recovery from desensitization became faster with increasing glycine concentration, with a slope of 0-76 x 107 M-1 s-1. These results are consistent with a model in which the effect of glycine occurs via an increase in the rate constant for recovery from desensitization, with little effect on the rate constant for onset of desensitization. Over the range 30-300 nM-glycine, the ratio of the rate constants calculated for recovery and onset of desensitization was a good predictor of the degree of desensitization recorded at equilibrium. 3. Concentration jump experiments with glycine were performed with 100 /LMNMDA present continuously, and for a single binding site model gave estimates of the association (1l1 x 107 m-l s-') and dissociation (3-1 s-1) rate constants for interaction of glycine with the NMDA receptor. In the presence of NMDA, concentration jumps from 3 ,tM-glycine to lower concentrations gave relaxations which became slower with decreasing glycine concentration over the range 1 uM-30 nM. A similar slowing of desensitization occurred when the glycine concentration was altered over the same range. * Present address: Vollum Institute for Advanced Biomedical Research, Oregon Health Sciences University, 3181 S.W. Sam Jackson Park Road, Portland, OR 97201, USA. t Present address: Insitute of Physiology, Czechoslovak Academy of Sciences, 142 20 Prague 4, Vfdenska 1083, Czechoslovakia. $ Author for correspondence.

MS 8021

3l. BENVVEXISTE AXD OTHERS 4. Glycine analogues of lower affinity produced desensitization with faster kinetics. D-Alanine, 150 nm, produced desensitization with a time constant of 175 ms, faster than recorded with an equipotent concentration of glycine (50 nm, time constant 259 ms). Responses of similar peak amplitude, recorded with 60 JM-Lalanine, and 500 /M-D,L-homoserine, did not produce strong desensitization, consistent with desensitization too rapid to resolve in our experiments. Concentration jump experiments with L-alanine and D,L-homoserine confirmed that the dissociation rate constants for these analogues (42 and 53 s-) were much faster than those obtained for glycine and D-alanine (341 and 4-9 s-'). 5. Numerical simulations were developed to test possible kinetic schemes for NMDA receptor activation and desensitizatiohn. Together with the experimental data they suggest that transitions to the open state do not occur unless glycine has first bound to a closed state of the NMDA receptor. Potentiation of NMDA responses by glycine results from the absolute requirement for glycine in promoting transitions to the open state. Glycine-sensitive desensitization is due to an agonist-triggered lowering of affinity for glycine, and subsequent dissociation of glycine from a closed state of the NMDA receptor channel complex, as a result of negative co-operativity between binding of glycine and NMDA. For kinetic schemes with either single or multiple glycine and NMDA binding sites, analysis of the response to fast applications of NMDA and glycine was used to obtain estimates for the glycine and NMDA association and dissociation rate constants. The best fit was obtained using a model in which the NMDA receptor has four independent amino acid binding sites: two for glycine and two for glutamate.

334

INTRODUCTION

Electrophysiological experiments on embryonic mouse neurones by Johnson & Ascher (1987) showed striking stimulation of NMDA receptor activity by glycine. Similar effects have been observed in experiments on NMDA receptors in Xenopus oocytes injected with messenger RNA (Kleckner & Dingledine, 1988; Kushner, Lerma, Zukin & Bennett, 1988). The augmentation by glycine of the binding of usedependent NMDA antagonists, which are thought to act within the ion channel portion of the NMDA receptor complex, provides further evidence that glycine promotes transitions of the NMDA receptor channel from closed to open states (Reynolds, Murphy & Miller, 1987; Snell, Morter & Johnson, 1988). However, the molecular mechanism by which glycine modulates NMDA receptor responses has not been established. Recently we described a reduction by glycine of NMDA receptor desensitization (Mayer, Vyklicky & Clements, 1989), and report here further experiments on this observation. Together with data presented in a companion paper (Vyklicky, Benveniste & Mayer, 1990) our results suggest that the molecular mechanism underlying the potentiating action of glycine on responses to NMDA does not involve block of desensitization per se, but rather that glycine is absolutely required for activation of the NMDA receptor channel complex. We propose that, over the glycine concentration range 10-1000 nm. responses to NMDA show glycine-sensitive desensitization because binding of NMDA to the agonist recognition site lowers the affinity of glycine for its binding site on the NMDA receptor. As a result, when

KINETIC EXPERIMENTS OX NMDA RECEPTORS

335

NMDA receptors are equilibrated first with glycine, the response to NMDA appears to desensitize as glycine dissociates from its binding site following the application of NMDA. Kinetic simulations based on rate constants measured in concentration jump experiments were used to test this hypothesis. METHODS

The recording and perfusion techniques were similar to those described in a companion paper (Vvklickv et al. 1990). Experiments were performed at room temperature (25-27 °C), 10-20 days after cultures were plated. Voltage clamp was achieved using whole-cell patch-clamp recording, and an Axon Instruments 'Axoclamp 2' discontinuous amplifier set at a gain of 2-4 nA mV-', with a switching frequency of approximately 10 kHz. NMDA was routinely applied in a low-calcium solution (0-2 mM) to reduce calcium-sensitive desensitization.

Fast perfusion techniques in whole-cell experiments The results presented are dependent on achieving fast perfusion of individual neurones. Flow pipes used in this series of experiments had large internal diameters (340 ,um); the solution velocity was 150 ,um ms-1 at the exit port. At any time solution was allowed to flow from only one barrel, and left the flow pipe as a diverging cone of solution. Thus the area perfused covered the complete width of the dendritic tree of individual neurones selected for recording. Experiments were performed on relatively young, low-density cultures (< 2 weeks after plating), and well isolated neurones were selected to minimize slow diffusional mixing that could occur in higher density, or older cultures, in which dendrites aggregate to form cables. Electron microscopy revealed that the dendritic processes of the young neurones used in our experiments do not burrow below the glial cell layer, as can happen when neurones and glial cells are plated together. Experimental results consistent with achieving fast perfusion include a time constant of exchange in sodium concentration jump experiments of less than ten milliseconds (Vyklicky et al. 1990), as well as very rapid, and substantial desensitization of responses to quisqualate (Mayer & Vyklicky. 1989), similar to that recorded using isolated cells (Kiskin, Krishtal & Tsyndrenko, 1986) or outside-out patches (Tang, Dichter & Morad, 1989; Trussell & Fischbach, 1989). The 'on' and 'off' responses to NMDA are very much faster than any of the glycine-sensitive relaxations described in the present series of experiments, suggesting that the speed of the application system does not limit the study of glycine-sensitive kinetics. For the study of kinetic processes of intermediate rates (time constants > 10 ms), the system used for the present experiments is probably adequate. For the study of faster processes the use of outside-out patches is preferable. A second concern in the present experiments is the accuracy of voltage clamp. Since agonist was applied to the whole of the dendritic tree this could result in amino acid responses arising from two compartments, one close to the electrode and well clamped, the second distant from the electrode and poorly clamped. We have several indications that poor voltage control is not a source of error. First, the reversal potential of desensitizing responses to NMDA and quisqualate, and of nondesensitizing responses to kainate, all occurred close to 0 mV, with no evidence of biphasic responses at the reversal potential, consistent with uniform voltage control in different parts of the dendritic tree. Second, responses to kainic acid, which do not show fast desensitization, and are thus simpler to interpret, show no evidence of activation of time- and voltage-dependent currents over the complete dose-response curve. This suggests that the more complex kinetics of desensitizing responses to NMDA are not due to activation of voltage-dependent conductance mechanisms in dendritic membranes as a result of poor voltage control. Results are presented as means+ standard deviation unless indicated differently.

Analysis and simulation Analysis of the time constant of relaxations produced by concentration jumps was performed using a Simplex algorithm to minimize the sum of square error between experimental responses and the response described by an equation of the form:

I(t)

=

I(oo) +Ai exp (-)

336

M. BEN VEXISTE AND OTHERS

in which I(t) is the current at time t, I(oo) is the current at steady state, T is the time constant of decay, and Ai the amplitude of the time-dependent component of the response at the start of the relaxation, such that Ai = I(0) -I(oo) where I(0) is the amplitude of the current at time zero. Numerical simulations and the fitting of experimental data to state diagrams describing receptor-channel gating were performed using FORTRAN programs run on a Microvax 3600 as described in the Appendix. For the most complex models described the analysis of responses to five concentrations of NMDA and glycine required around 10-30 h of central processing unit (CPU) time to converge, depending on initial starting values.

RESULTS

The goal of the experiments described here was to link the kinetics of desensitization at NMDA receptors with the kinetics of binding of glycine to NMDA receptors. A fast perfusion system was used to make sudden changes in the concentration of glycine, NMDA, or some glycine analogues, and the resulting relaxations of NMDA receptor current were used for kinetic analysis. Because our results (Vyklicky et al. 1990), and those of Kleckner & Dingledine (1988), suggest that glycine is absolutely required for activation of NMDA receptors by excitatory amino acids, we will use the terminology 'glycine agonists' when referring to glycine analogues which promote NMDA receptor activation in the presence of NMDA.

The effect of glycine on the kinetics of NMDA receptor desensitization When applications of NMDA were made with 5 mm-intracellular 1,2-bis(oaminophenoxy)-ethane-N,N,N',N'-tetraacetic acid (BAPTA), 0-2 mM-extracellular calcium, and 3 JLM or higher extracellular glycine, there was little desensitization to NMDA. In some neurones (e.g. control trace in Fig. 7, Vyklicky et al. 1990) desensitization was completely absent, while in others there was a gradual decrease in the NMDA-evoked current, which was too slow to quantify accurately. When the series resistance was initially high this slow desensitization was more pronounced, but decreased when suction was used to achieve better intracellular dialysis. We suggest that this effect of decreasing the series resistance is most likely due to a reduction of calcium-mediated desensitization, due to a more efficient introduction of BAPTA into the cytoplasm under optimal recording conditions (e.g. Fig. 2, Vyklicky et al. 1990). In our initial experiments we found a marked decrease in the time constant of recovery from desensitization on raising the concentration of glycine, with little effect on the time constant of onset of desensitization (Mayer et al. 1989). However, kinetic analysis of subsequent experiments also revealed a glycine concentrationdependent increase in the rate of onset of desensitization, such that desensitization became approximately 3 times faster on raising the concentration of glycine from 100 nm to 1 ,IM (Fig. 1). The onset of desensitization was well fitted by a singleexponential function of time constant Td, although a small improvement was usually obtained with a double exponential. A plot of the rate of onset of desensitization ('/Td) versus glycine concentration was linear, with a slope of 0-87 x 107 M-1 s-, and, assuming a background concentration of 20 nM-glycine (Vyklicky et al. 1990), gave a zero concentration intercept of 3-26 s-1, corresponding to a desensitization time

KINETIC EXPERIMENTS ON N\MDA RECEPTORS 30 nM-glycine

A

337

100 nM-glycine

0.5 2 nA

,

td319ms

rd 249 ms

0.3 jiM-glycine

1

iM-glycine 4 5 nA

rd81 ms J

t

d02 ms

750 ms B 14 12

10

r-

m

8

0o N

._a

C' 0

0

200

400

600

800

1000

Corrected [glycinel (nM)

Fig. 1. The rate of onset of NMDA receptor desensitization increases with glycine concentration. A, responses to 100 4uM-NMDA recorded in the presence of glycine at the concentrations indicated. Although increasing the concentration of glycine dramatically reduces desensitization, the onset of desensitization is faster at high concentrations of glycine. Curves drawn through the data points are single-exponential functions, the time constants of which are indicated adjacent to each record. B, desensitization rate (1/rd) versus glycine concentration, assuming a background glycine concentration of 20 nM; data points show means+S.E.M. of measurements made from twenty-six neurones. The linear regression line drawn through the data points has a slope of 0 87 x 107 M1s-1 (r = 0 99).

constant of 307 ms at low concentrations of glycine (Fig. 1). These values change by less than + 10 % when the background concentration of glycine is assumed to be either zero (highly unlikely) or 50 nm (also unlikely in view of the strong potentiation of responses to NMDA by 30 nM-glycine as shown in Fig. 6). Because of the small degree of desensitization with 0 6-1 ,tM-glycine, the study of the kinetics of

M. BENVENISTE AND OTHERS 338 desensitization at high concentrations of glycine required optimal recording conditions, and could only be performed in those cells in which slow desensitization described above was minimal. At concentrations of glycine above 3 ,M, it is likely that the rise time of the response to NMDA itself limits the resolution of fast A

100 pM-NMDA + 300 nM-Gly 0 0-19 0.56 0.96 s

0

I

I

I

I

I

100 pM-NMDA + 30 nM-Gly 0.3 1.3 2.3 3.3 4.8s

I

I

I

I

I

I1 nA 1s

C

B

4

c 0

o

krecovery

*

konset

c 0

.5

01 .6

0

c

0.4

-

a

0-4-

0

o0 0)

0

._

m

0.2

._i

-

az 0.0

0

1

2

3

4

0

5

Time (s)

100

200

300

400

Corrected [glycinel (nM)

Fig. 2. Glycine increases the rate of recovery from desensitization. A, twin-pulse applications of 100 ,uM-NMDA, recorded with 300 nm (left) or 30 nM-glycine (right). The interval between the 1st and 2nd pulses was varied to measure the time constant of recovery from desensitization. Sufficient time was allowed between applications for complete recovery from desensitization. B, peak amplitude of the response to the 2nd pulse of NMDA (Ite.) divided by the peak amplitude of the response to the 1st pulse of NMDA (Icontrol) versus interpulse interval. The recovery time constant (TrrC) was estimated by fitting single-exponential functions. C, recovery rate (l/r, ) versus glycine concentrations (0), plotted assuming a background level of 20 nM-glycine. The slope of 0-76x 107 M-1 s-1 was obtained by linear regression; data points show means+ s.E.M. of observations made on twelve neurones (r = 0 99). Control responses from the same data set were analysed to determine Td as shown in Fig. 1. The rate of onset of desensitization konset was calculated from the relationship Td l/(krecovery + konset) and was not greatly affected by changes in the concentration of glycine (@). =

desensitizing responses; in any event, we were unable to study the action of glycine on desensitization at concentrations greater than 1 gUM. Because recovery from desensitization is slow at low concentrations of glycine, we were able to use twin-pulse application of NMDA to study the kinetics of recovery with 30-300 nM-glycine. Examples of such experiments are shown in Fig. 2. The

KINETIC EXPERIMENTS ON XN\MDA RECEPTORS

339

relatively long periods of stable recording required to characterize the recovery time constant made it difficult to complete this series of experiments, and it was not possible to obtain a sufficiently large number of data points to rigorously determine whether recovery from desensitization followed single or multiple exponential kinetics (sigmoidal kinetics also could not be rigorously excluded). With the limited data sets we were able to obtain, the recovery process was well fitted by a single exponential (Fig. 2). There was a dramatic decrease in the time constant of recovery from desensitization (Tre,) with increasing glycine concentration, and a plot of (1/Tree) versus glycine concentration was linear, with a slope of 0-76 x 107 M-1 s-1 (Fig. 2). At concentrations of glycine greater than 300 nm the speed of the apparatus used to move the flow pipes became limiting when a series of rapid sequential movements was required, and experiments with glycine concentrations greater than 300 nm were not performed. Further analysis of the data from Figs 1 and 2 is possible, if it is assumed that the conversion between activated and desensitized NMDA receptors is slow relative to the kinetics of binding of NMDA and glycine and subsequent ion channel activation. A highly simplified state diagram of NMDA receptor function, using the above assumptions is given by: Active

krecovery

=

Desensitized,

konset

in which konset and krecovery are the rate constants for onset of and recovery from desensitization. At any given concentration of glycine the value of kreeovery was calculated from i/Trec, measured as shown in Fig. 2B; konset was calculated from the relationship: 1 (1) Td = k krecovery ± konset Because the data in Fig. 2 gives estimates of both the amount of desensitization at equilibrium and the kinetics of desensitization, we compared desensitization predicted by the independently derived rate constants with the experimentally observed response at equilibrium (Fig. 3). This analysis was performed to see if the ratio krecovery/konset and the amount of desensitization at equilibrium (steadystate/peak current) were similar. The analysis makes no prediction about the mechanisms by which glycine alters the kinetics of desensitization, yet reveals a strong correlation between equilibrium desensitization (plotted as (steady-state peak)/1 - (steady-state peak) current) and the kinetics of desensitization (plotted as krecovery/konset). Our subsequent experiments were attempts to link the kinetics of desensitization with the kinetics of interaction of glycine with NMDA receptors.

Glycine concentration jumpps When neurones were continuously exposed to NMDA slow increases and decreases in current were recorded following the application and removal of glycine (Fig. 4). These changes in current were well fitted by single-exponential functions; slightly better fits were usually obtained with two exponentials, but the improvement in fit was marginal. These results are consistent with a bimolecular reaction between

M. BENVENISTE AND OTHERS

340

NMDA receptors and glycine, in which R-NMDA is a closed state of the receptor with NMDA bound, and R*-Gly-NMDA is a closed state of the receptor with both NMDA and glycine bound, and which is in rapid equilibrium with the open state.

Gly + R-NMDA ±=R*-Gly-NMDA k+ Consistent with this scheme the time constant of the current decrease recorded on terminating the application of glycine (T.ff) did not change appreciably with glycine concentration over the range 50 nM-100 /tM, and had a value of 331 +42 ms (mean +S.D., n = 204 observations, fifteen cells). The corresponding rate constant for dissociation of glycine from NMDA receptors (k_, calculated as l/rTff) was 3 07 +0 4 s-1. The time constant of the increase in current following application of glycine (Ton) became faster with increasing glycine concentration, and at 100LM1.2

0.8 'xL 0.8 0*6-

(pa,X0.4 Z. 0.2 co

0

0.4 0.6 0-8 1.0 Desensitization kinetics (krecovery/konset)

0.2

1.2

Fig. 3. Modulation of the kinetics of NMDA receptor desensitization predict measurements of desensitization obtained at equilibrium. The data illustrated in Fig. 2 gave values for krecovery and konset at three concentrations of glycine, as well as measurements of desensitization at equilibrium (steady-state/peak current). These parameters, plotted as (steady-state/peak current)/(1-(steady-state/peak current)) versus krecovery/konset' were highly correlated, as would be expected if the kinetics of desensitization determined the degree of desensitization at equilibrium. Data points show means + S.E.M. of observations made on twelve neurones.

glycine (the highest concentration examined) reached a value of 9-93 + 0-47 ms, suggesting activation of NMDA receptors by 100 /uM-glycine at rates greater than 100 s-5. The relationship between glycine concentration and Ton expected from the above relationship is given by

l/Ton = k+ [Gly] + k.

(2)

A plot of the l/Ton versus glycine concentration was linear over the range 50 nM-1 ,UMglycine, with k+ calculated from the slope as 1F1 x 107 M-1 s-1 (Fig. 4). However, the

KINETIC EXPERIMENTS OXN MDA RECEPTORS

341

zero concentration intercept (I 12 s-') was slightly slower than predicted for a simple reaction of the type described above (341 s-'), since similar values for the association and dissociation rates would be expected at low concentrations of glycine. One explanation for this, supported by analysis of data from simulations (see Appendix), is the need for a more complicated reaction scheme involving multiple binding sites for glycine. Although such multistate models predict current changes which should be fitted with the sum of several exponential terms, in practice, when the rate constants for several binding reactions are of similar value, the resulting relaxation can be nearly indistinguishable from a single exponential. In such cases the deviation from single-exponential kinetics is greatest at the beginning of the concentration jump response. With our experimental apparatus the solution change is not instantaneous, and time of the start of the concentration jump is not accurately determined. As a result it is difficult to distinguish sigmoidal from exponential kinetics. Simulated responses from appropriate multibinding site models, assuming non-instantaneous mixing following solution changes, were, after an initial delay, also well fitted with single-exponential functions and yielded similar plots to that shown in Fig. 4, with 1/Tro slower than l/Toff at low concentrations of ligand. Calculation of an apparent equilibrium dissociation constant for potentiation of NMDA receptor responses by glycine, from the ratio of the dissociation and association rate constants given above, gives a value of 280 nm. Although this calculation assumes only a single binding site for glycine, the result is in reasonable agreement with the EC50 value of 382 nm calculated from analysis of equilibrium dose-response curves (Fig. 5 in Vyklicky et al. 1990). The time constant of the decrease in current following removal of glycine, shown in Fig. 4, is similar to the time constant for desensitization of responses to NMDA shown in Figs 1 and 2. Together with the 3- to 4-fold difference in potency for the action of glycine and the glycine antagonist 7-chlorokynurenic acid on responses to NMDA measured at peak and steady state (Vyklicky et al. 1990), and the absolute requirement for glycine in allowing transitions to the open state of the receptor channel complex (Kleckner & Dingledine, 1988; Vyklicky et al. 1990), this raises the possibility that desensitization at NMDA receptors could be due to dissociation of glycine from the NMDA receptor, triggered by binding of excitatory amino acid. In this model, the affinity of glycine would by 3-4 times higher in the absence of agonist, than following binding of NMDA or L-glutamate. If this were true, then the time constant of desensitization (Td) at any given concentration of glycine should be similar to the equilibration time constant (Teq) recorded following a rapid change in glycine concentration: d

k+ [Gly] + k_

eq(

To measure Teq, the change in membrane current was recorded on switching from a nearly saturating concentration of glycine to a lower concentration of glycine at which the kinetics of desensitization were also measured. Experiments to test this hypothesis are shown in Fig. 5. NMDA receptors were first activated by rapid application of 100 ,tM-NMDA in the presence of 3 ,tm-glycine, and then the glycine concentration was stepped to a lower value, in the range 100-1000 nm. This produced a reduction of the NMDA-evoked current, thus simulating the process of

M. BEN VENISTE AND OTHERS

342 A

200 nM-glycine r0, 280 ms

|-0n 1 nA

T0ff321 1

ms

yM-glycine ro,,80ms 2 nA 500

ms

zff 282 ms B 14

o 1/T0n /roff

_*

12

10

7,

8

°64

2

J

0-

0

800 200 600 400 Corrected [glycinel (nM)

1000

Fig. 4. Kinetic experiments used to measure the rate of association and dissociation of glycine. A, responses recorded in 100 /LM-NMDA, during step applications indicated by bars of 200 nM (top) or 1 ,uM-glycine (bottom). Lines drawn through the data points are single-exponential functions, the time constants of which are indicated adjacent to the records. The rate of increase in NMDA receptor current is faster following a jump into 1 /LM-glycine, than following a jump into 200 nM-glycine. The rate of decrease of NMDA receptor current recorded on stepping into glycine-free solution is similar following activation by either 1 ,UM or 200 nM-glycine. B, association (1/TO,) and dissociation (1ITOff) rates versus glycine concentration for 50 nm to 1 /LM-glycine added experimentally. assuming a background glycine concentration of 20 nm. The dissociation rate is not sensitive to changes in glycine concentration; the association rate increases with glycine concentration, with a slope of 1.1 x 107 m-l s-'; data points show means+s.E.M. of observations made from fifteen neurones.

desensitization. The equilibration time constant recorded in such experiments was faster at 1 /LM-glycine (740 + 2-7 ms) than at 100 nM-glycine (295 + 29 ms), and plots of the equilibration rate (l/Teq) showed a strong correlation with the desensitization rate (I/Td) measured at corresponding concentrations of glycine in independent

*~ ~ ~ |nA

KINETIC EXPERIMENTS ON NMDA RECEPTORS A __

100 nM-glycine

.q 247 ms

;

1 gM-glycine

r,^,..o ; 109 ms B 14

500 ms

12-

.$10X lo

-

-

._ 0-

0

2

4

6

8

10

12

14

Desensitization rate, 1/;d (s-1)

Fig. 5. Concentration jump experiments with glycine mimic NMDA receptor desensitization. A, responses to 100 ,SM-NMDA recorded following a step decrease in glycine concentration. Initially neurones were incubated with 3, M-glycine, prior to a step application of 100,UM-NMDA indicated by arrows above the traces. Lines above the traces indicate a reduction of glycine concentration in the presence of NMDA, from 3/SM to the concentrations indicated. Lines drawn through the data points are singleexponential functions, the time constants of which are indicated adjacent to the experimental records. The NMDA-activated current reaches equilibrium faster following a step to 1 /uM-glycine than following a step to 100 nM-glycine. B, the glycine equilibration rate (1/lr,) plotted against 1/'r, recorded at the equal concentrations of glycine (10, 30, 100 and 300 nM) in independent experiments during application of 100 uM-NMDA, as shown in Fig. 2. Observations are means+ S.E.M. recorded from nine neurones.

343

344

M. BE.NVEX'VISTE AND OTHERS

experiments (Fig. 5). Further support for our hypothesis comes from comparison of the results shown in Figs 1 and 4. The dissociation rate constant for glycine (341 s-1) is similar to the rate of onset of desensitization with no added glycine (3 3 s-1), estimated from the intercept of 11Trd versus glycine concentration. Mlodulation of responses to NMDA by lower affinity glycine agonists The above model for the action of glycine on NMDA receptors predicts that the rate constant of desensitization recorded with different glycine agonists should be highly correlated with the rate constant for dissociation of these agonists from the glycine binding site on NMDA receptors. Recent studies on structure-activity relationships for glycine analogues have identified several compounds with lower affinity than glycine (Bristow, Bowery & Woodruff, 1986; Snell et al. 1988; Kessler, Terramani, Lynch & Baudry, 1989). The maximum rate at which drugs can bind to receptors is determined by the diffusion limit; for low molecular weight ligands in aqueous solution this is ~ 101 m-1 s-' (Smoluchowski, 1916). In practice association rates are slower for neurotransmitter and drug receptors, and, typically, have values between 108 and 106 M-1 s-' (e.g. Burgen, 1966), hence differences in affinity will often reflect changes in dissociation rate constant, low-affinity ligands dissociating more rapidly than high-affinity ligands. Because the gating of NMDA receptors is intrinsically slow under the present experimental conditions (the activation time constant of NMDA receptor currents recorded following concentration jumps with high concentrations of NMDA has a minimum value of approximately 7 ms (M. L. Mayer & L. Vyklicky Jr, unpublished observations), low-affinity glycine agonists should fail to produce strongly desensitizing responses because there will be simultaneous rapid dissociation of the glycine agonists during activation by rapid applications of NMDA. This model also predicts that if higher affinity glycine agonists are developed, the time constant of desensitization should become slower in the presence of such ligands. Four potential lower affinity glycine agonists that have been identified as active in ligand-binding experiments were tested for potentiation of responses to NMDA (Snell et al. 1988; Kessler et al. 1989). These were initially tested at concentrations expected to give comparable receptor occupancy based on equilibrium constants determined in binding experiments. Subsequently doses were chosen to give peak amplitude responses to 100 1aM-NMDA similar to those recorded in the same neurones when 50 nM-glycine was added to the extracellular solution. Potentiation of responses to NMDA was observed with 150 nM-D-alanine, 20/tM-L-alanine, and 150 /M-D,L-homoserine, but not 150/tM-L-valine. At the above concentrations of glycine analogue, NMDA responses measured at peak were 4-7 times larger than those recorded in the absence of any added glycine. Currents measured at steady state for L-alanine, and for D,L-homoserine, were on average 4 7 and 3-2 times larger than those recorded with 50 nM-glycine, while those recorded with 150 nM-D-alanine were comparable in amplitude to responses recorded with 50 nM-glycine (Fig. 6). As a result, calculation of the ratio steady-state/peak current showed strong desensitization for glycine and D-alanine, but not for L-alanine and D,L-homoserine. Although glycine and D-alanine appear to have a similar action as modulators of NMDA receptors, and at low concentrations produce strongly desensitizing responses

KINETIC EXPERIAIENTS ON.NAIMDA RECEPTORS

345

to NMDA, the potency of D-alanine appeared to be slightly lower than that of glycine, consistent with results from binding studies (Snell et al. 1988). The model proposed above thus suggests that, compared to responses recorded with glycine, NMDA receptor desensitization should be faster in the presence of D-alanine. A A

Background 150 nM-D-Ala

B

20 LuM-L-Ala

1 50 =_

E a)

O

v

ur

>

-~

,2

1 00

0 50

000

Bckgnd

50 nM-Gly

1

D-Ala

L-Ala

Gly

DL-Hser

C

150 pM-DL-Hser

1.00 c

~~~~~~~~~~~~~~~~~~~~~~~~~~ =

a.

0.75

-o +m 0.50nA

>.)> m

1 s

0.25 0.00

I

Bckgnd

D-Ala

L-Ala

Gly

DL-Hser

Fig. 6. NMDA responses recorded with low-affinity glycine analogues. A, top left traces are superimposed responses to applications of 100 ,yM-N'MDA recorded in control solution, with no added glycine (background), or with 150 nm-D-alanine; adjacent traces show responses recorded in the presence of 20 ,uM-alanine, 50 nm-glycine. or 150 /M-D.Lhomoserine (DL-Hser). Compared to NMDA responses recorded in control solution, all of the analogues used produce marked potentiation. but there is strong desensitization only with glycine or D-alanine. The residual desensitization recorded with L-alanine or D,Lhomoserine was variable in degree and time course. B, bar graphs of the amplitude of the NMDA-evoked peak and steady-state current, normalized with respect to the peak current recorded with 150 nM-D-alanine; values are given as means+ S.D. Each analogue was tested on nine neurones. C, NMDA receptor desensitization (steady-state/peak current) for the same series of glycine analogues, plotted as means+S.D. Responses to 20 ,uM-L-alanine and 150 ,LM-D,L-homoserine do not show strong desensitization.

comparison of the rate of desensitization was made at concentrations of D-alanine and glycine producing similar amplitude initial current responses to NMDA. The presence of a background concentration of glycine complicates the interpretation of such experiments, since at low concentrations of D-alanine there will be substantial occupancy of the glycine binding site by glycine. Despite this, we found that desensitization recorded with 150 nM-D-alanine, which produced peak responses on average 0-85 times those recorded with 50 nM-glycine, was faster (time constant

346

M. BEN,. VENAISTE AND OTHERS

175 +3 ms) than desensitization recorded in the same cells with 50 nM-glycine (time constant 259+20 ms); this difference was highly significant P = 000 1, one-way analysis of variance). Representative traces are shown in Fig. 7. Concentration jump application of D-alanine. L-alanine and D,L-homoserine To confirm that the absence of desensitization recorded with L-alanine, and D,Lhomoserine. and the faster desensitization recorded with D-alanine was indeed due to A

100iuM-NMDA+

B

50 nM-glycine or 150 nM-D-alanine

8.0,

~~~~~~~~~~~~~~60 c4.0 0 D-Ala rd 157 ms Gly zd272 ms 4 200 ms

|0

--;:

20 c

075nA

Bckgnd

Gly

D-Ala

Fig. 7. NMDA receptor desensitization is faster in the presence of D-alanine. A, responses of the same hippocampal neurone to 100 /,M-NMDA recorded in the presence of either 50 nM-glcine or 150 nM-D-alanine. Lines drawn through the data points are singleexponential functions. the time constants of which are indicated adjacent to the experimental records. B. desensitization rate constants (l/Td) for responses to 100MUMNMDA recorded in control solution, with no added glycine (bekgnd). or with 50 nMglycine (Gly) or 150 nM-D-alanine (D-Ala); error bars show I S.D. The rate of desensitization recorded with control solution (3 91 + 0-58 s-') is not significantly different, P> 05 from the rate of desensitization recorded mith 50 nM-glycine (388±030 s-') desensitization with 150 nm-D-alanine was significantly faster (5 80+0-70 s-1), P < 0 001, one way analysis of variance.

the faster rate constant for dissociation of these ligands from the NMDA receptor we performed a series of concentration jump measurements similar to those shown for glycine in Fig. 4, but with the D and L isomers of alanine and with D,L-homoserine. Representative traces illustrating measurement of dissociation rate constants for these glycine agonists are shown in Fig. 8. For D-alanine responses were evoked with concentrations of 200-800 nm, and their time constant of growth and decay was slightly faster than that recorded with glycine at concentrations producing similar potentiation of responses to NMDA. For L-alanine much higher concentrations (10-60,UM) were required to evoke similar amplitude responses to NMDA, and the time constants of association and dissociation were much faster than recorded with glycine and D-alanine. For D,L-homoserine (100-900 /aM) the relaxations were even faster. For all of these glycine agonists the time constant of dissociation did not vary with concentration of ligand, and was of a characteristic value for each ligand (Fig. 8). The dissociation rate constants derived from these measurements (calculated as 1/Toff) became faster as the potency of the

KINETIC EXPERIMENTS ON NMDA RECEPTORS A

347

B

D-Alanine

0.2nA

L-Alanine 50-

10.4nA

Toff 199 ms

D,L-Homoserine

co

25-

0

'roff 25 ms..-

0.4 nA 0

D-Ala

200

roff 16ms'

L-Ala

Gly

D,L-Hser

50ms

Fig. 8. The dissociation rate constant for low-affinity glycine agonists is faster than that recorded for glycine. A, responses recorded in the presence of 100 ,uM-NMDA, following termination of application of either 600 nM-D-alanine, 60 /M-L-alanine or 500 /LM-D,Lhomoserine. The decline in NMDA-evoked current is due to dissociation of the glycine analogues from NMDA receptors, and is fitted with single-exponential functions the time constants (Toff) of which are indicated adjacent to the experimental traces. The time bar for the trace recorded with D-alanine is 4 times slower than that for L-alanine and D,L-homoserine. B. bar graphs of the dissociation rate constants (1/T.f,) for D-alanine, L-alanine, glycine and D,L-homoserine; values are plotted as means+ S.D.

glycine analogue decreased, and followed the sequence glycine (3 07 +04 s-') < Dalanine (4 94+0 74 s-') < L-alanine (41-6+5±74 s-') < D,L-homoserine (53-1 + 14-2 s-1). DISCUSSION

Glycine-sensitive desensitization at NVMDA receptors Our preliminary experiments emphasized the strong reduction of NMDA receptor desensitization by glycine which occurs in solution with a low extracellular concentration of calcium (Mayer et al. 1989). Several lines of evidence from the experiments described here, and in a companion paper (Vyklicky et al. 1990), suggest that such desensitization is actually due to dissociation of glycine from the NMDA receptor-channel complex, triggered by binding of excitatory amino acid. This hypothesis can be illustrated in a state diagram, referred to as model 1, in which -K are the microscopic equilibrium dissociation constants for glycine and NMDA. Model 1 Fast GIN Rl

RGl y K2

KDA

R

_

K,

RNMDA

Open

348

M. BEXVIENISTE AND OTHERS

Model 1 requires the following assumptions: (1) NMDA receptor channel complexes open only when both glycine and NMDA are bound to the receptor; (2) at the concentrations of NMDA used in the present experiments, the binding and dissociation of NMDA is fast relative to the rates for binding and dissociation of glycine; (3) transitions to and from the open state are fast relative to the rate constants for binding and dissociation of glycine; (4) the dissociation constant (K2) for the interaction of glycine with unliganded receptors is of higher affinity than the dissociation constant (K3) for the interaction of glycine with receptors occupied by NMDA; (5) the law of microscopic reversibility requires that K1 must equal [K2 x K4]/K3. As a result the equilibrium dissociation constants for NMDA (K1 and K4) also have different affinities, with K4 of lower affinity than K1. Model 1 thus predicts a negative allosteric interaction between binding of glycine and binding of NMDA. The small amplitude of responses to NMDA in solutions with only a background concentration of glycine, and the complete block of responses to NMDA by glycine antagonists provides direct support for assumption (1) (Kleckner & Dingledine, 1988; Huettner, 1989; Vyklicky et al. 1990). Rapid dissociation of NMDA (assumption (2)) was tested by concentration jump application of NMDA in Mg2+free solutions containing 30-1000 nM-glycine (Fig. 9). After a switch to NMDA-free solution the current decays with a time constant of approximately 25 ms. This value is independent of glycine concentration, and 13-fold faster than the decay following a switch to glycine-free solution in the presence of NMDA. Rapid binding of 100 /tmNMDA is supported by the observation that glycine-sensitive desensitization is essentially absent below 10 ,tM-NMDA (e.g. Fig. 9). This is predicted by the model shown above, because the net rate of binding of NMDA varies with NMDA concentration, and at 10 /um is slow relative to the kinetics of dissociation of glycine (Fig. 4). There is no direct evidence available concerning assumption (3). Measurement of ,? and a from single-channel experiments has not yet been reported for NMDA receptor channels, but many laboratories are in agreement that burst lengths of 5-10 ms account for the major gating behaviour observed with a variety of agonists (Jahr & Stevens, 1987; Ascher, Bregestovski & Nowak, 1988; Howe, Colquhoun & Cull-Candy, 1988; Cull-Candy & Usowicz, 1989). However, the glycine concentration at which these measurements were obtained is not known. The lower potency of glycine for potentiation of equilibrium versus initial responses to NMDA, and the opposite potency sequence for the glycine antagonist 7-chlorokynurenic acid supports assumption (4) (Vyklicky et al. 1990). The model is similar to the cyclic model of desensitization for nicotinic acetylcholine receptors first proposed by Katz & Thesleff (1957), and subsequently adopted by many other groups. The major difference is that desensitized states in the classical model are replaced by states with no glycine bound. The kinetics of onset of, and recovery from, desensitization in the classical model, are replaced by the kinetics of dissociation of glycine, and rebinding of glycine to the NMDA receptor. Unfortunately, the model gives no insight into the molecular mechanism by which glycine and NMDA promote transitions to the open state. Model 1 predicts dose-response curves for the action of NMDA and glycine with Hill coefficients of 1. This is not supported by experimental observations. The Hill

KINETIC EXPERIMENTS ON NMDA RECEPTORS

349

coefficient for the action of several excitatory amino acids as NMDA receptor agonists on hippocampal neurones is > 1. For NMDA responses recorded in the presence of 3 /tM-glycine we obtained an EC50 of 34 /um, Hill coefficient = 1 41 (M. L. Mayer & L. Vyklicky Jr, unpublished observations; see also Verdoorn & Dingledine, 1988). For glycine responses at equilibrium, in the presence of 100 um-NMDA, we obtained an EC0 of 382 nM, Hill coefficient = 1P44 (Vyklicky et al. 1990). The simplest model which satisfies the above five conditions, and yields Hill coefficients consistent with experimental data, contains two binding sites for NMDA and two binding sites for glycine, and is illustrated below (Model 2). Model 2 4. 4,off

RiGlY L_R2>G^>< 2 x k2f off

k2.on RGIY

k2,orf

'4I

RS'INMDA

2 x k,2.on

R R ±4~ If 2 X kiLon

2 X k-4,off

IV_

RO(Nl1NA

k off 2 x3.

k303 on

Open

IV _R2NMDA

k3. kl.off R,M) RNMDA

Fast

2 x k-3on

2 x k-l'~ofr

tI _'R2NMDA R2|D kL.on

Hatch marks labelled I, II. III and IN' signify binding reactions of equivalent microsco)ic Kd.

It is helpful to consider precedents from experiments on other ligand gated channels before concluding that the above proposal for multiple NMDA and glycine binding sites is unrealistically complicated. Expression of functional GABA and glycine receptor channels in cell lines or Xenopus oocytes occurs following injection of messenger RNA coding for a single protein subunit (Blair, Levitan. Marshall. Dionne & Barnard, 1988; Sontheimer, Becker. Prichett, Schofield, Grenningloh, Kettenmann. Betz & Seeburg, 1989). It is probable that every subunit contains an agonist binding site, and that functional channels are formed from pentamers (Langosch, Thomas & Betz, 1988), or tetramers (Mamalaki, Stephenson & Barnard, 1987). Thus GABA and glycine receptor channels could contain up to five binding sites for agonists. Despite this, Hill coefficients for activation of these channels are usually close to 2 (Sontheimer et al. 1989), suggesting that not all agonist binding sites need to be occupied to open the channel, or strong co-operativity of binding masks the requirement for occupation of five ligand binding sites. Kinetic models for glycine-sensitive potentiation and desensitization Numerical techniques were used to fit model 2 to experimental results (Fig. 9A and B). We assumed that equilibrium with the open state was sufficiently rapid to preclude the need for modelling this state directly (e.g. Cachelin & Colquhoun, 1989). This has the effect that the probability of channel opening is directly proportional to number of receptors with both NMDA and glycine bound. The data sets used for

350

M. BENIVENISTE AND OTHERS

analysis were recorded from three neurones using fast application of solutions with 10, 30 and 100 ,tM-NMDA, with 50, 200 or 1000 nm-glycine added experimentally, assuming a background concentration of 20 nM-glycine. The control solution was Mg21 free, and thus produced slightly larger peak responses than recorded previously (e.g. Vyklicky et al. 1990). Model 1 was fitted to the same data for comparison and had a sum of squares error 19-7 times greater than obtained with model 2. Model 1 consistently failed to give enough desensitization at low concentrationls of glycine, although fits were good at higher concentrations. The average values for association rate constants. in units of M1-1 s-' and dissociation rate constants, in units of s-' obtained using model 2 were: k o 1[ 1x 10'; k off, 46; k2o 1.1 x 10'; k2-ff, 0-51; k30on 5a9x 106; k3off, 2-0; k14,01 1 3 x 106; k4,off, 39. Although model 2 accurately simulated glycine-sensitive activation of NMDA receptors small discrepancies in the best fits are noticeable at the peaks of the responses (Fig. 9A and B). It is likely that problems with determining the exact timing of the start of the agonist application, and small differences in the effective rate of solution exchange for barrel to barrel ill the flow pipe array contribute to this. Also, a small amount of calcium-sensitive desensitization could distort the experimental traces. By simulating the response to a series of NMDA concentration jumps recorded at different concentrations of glycine, we can determine the effect of glycine on the rate of onset of desensitization (rd) and the degree of desensitization at equilibrium predicted by model 2. Simulations were performed using the mean values for microscopic rate constants given above. A plot of steady-state/peak current versus glycine concentration for simulated responses to NMDA (Fig. 90). had an EC50 of 488 nm, in reasonable agreement with the value of 264 nm measured experimentally (Vyklickv' et al. 1990). A plot of (l/Td) versus glycine concentration for simulated responses to NMDA was linear (Fig. 9D). with a slope of 0 57 x 107 M-1 s- consistent with the value of 0-87 x 10 M-1 -1 measured experimentally (Fig. 1). Model 2 predicts that the decay of NMDA receptor current, following termination of a concentration jump application of glycine, should occur with a time constant approximately equal to 2 x k3,off; the value predicted, 4 s-1, is close to the experimentally determined value of 1/Toff for glycine (3-1 s-1). Recent experiments on the stoichiometry of ligand binding sites on NMDA receptors in membrane preparations suggest a ratio of two NMDA and two glycine binding sites per channel (Thedinga, Benedict & Fagg, 1989). Model 2 shows negative allosteric interactions similar to model 1: binding of glycine lowers the affinity for NMDA and vice versa (K1 and K2 are higher affinity than K3 and K4). However, because the pair of NMDA binding reactions to receptors with both glycine sites occupied have equal dissociation constant (IV: top row), the model predicts a Hill coefficient for NMDA > t in the presence of a high concentration of glycine. A similar argument applies for glycine binding reactions when both NAIDA recognition sites are occupied (K3; right side of model). Analysis of simulated responses generated by model 2, using the logistic equation, gave the following values for equilibrium responses: NMDA EC50 = 704um, n = 1-24; glycine EC50 = 723 nm, n = 1-21. An alternative version of model 2 is one in which the binding of NMDA to receptors with one glycine binding site occupied, and also the binding of glycine to receptors with

KINETIC EXPERIMENYTS O-.NNMDA RECEPTORS A

B

100 ,iM-NMDA + 50, 200. 1000 nM-Gly

351

1000 nM-Gly + 10, 30, 100 gM-NMDA

--I

1- *---"~"'

-

2 nA

,;

Im

500 ms

C

D 7

10-

1-0 o M 000.8 0044 a

'Z .he

4

046

C

Z

0

.N

cn

5 X Q C

-

Xm 6 -

0

_S

2-

cn

0-0

aD.01

cx)

°

250 500 750 1000 0 [Glycine] (pM) [Glycine] (nM) Fig. 9. Simulated transients predicted by a model 2. A, responses to rapid application of 100 ,1M-NMDA, recorded with 50, 200 and 1000 nM-glycine added experimentally. Superimposed smooth lines are the responses predicted by model 2. B, experimental and simulated responses from the same cell to rapid application of 10, 30 and 100 ,tM-NMDA, in the presence of 1000 nM-glycine. The rapid current decay following removal of NMDA reflects the kinetics of dissociation of NMDA and the solution exchange time constant, and is not due to the onset of channel block by Mg2+ (the control solution had no added Mg2+). Fits in A and B were obtained assuming a background contamination of 20 nmglycine. C and D, analysis of simulated responses predicted by model 2 to rapid application of 100 ,uM-NMDA (10 ms exchange time constant). C, steady-state/peak current versus glycine concentration obtained was fitted with the logistic equation 1 steady-state peak I + (EC50/[Gly])n the plot is similar to that obtained by analysis of experimental data. D, rate of onset of desensitization, fitted with single-exponential functions. The increase in Td with increasing glycine concentration is similar to that recorded experimentally as described in Fig. 1. 0.1

1

10

352

A3. BEXNVENISTE AND OTHERS

one NMDA binding site occupied, was positively co-operative. There is some evidence from ligand-binding experiments that glycine increases binding of Lglutamate to NMDA receptors (Fadda, Danysz, Wroblewski & Costa, 1988; Monaghan, Olverman, Nguyen, Watkins & Cotman, 1988), though this point is controversial (Bonhaus, Yeh, Skaryak & McNamara, 1989). However, the alternative version of model 2 was slightly less well fitted to experimental data (sum of squares error 1 33 times higher than model 2). All other models containing two NMDA binding sites and two glycine binding sites, with two different affinities for each ligand were not consistent with our experimental data. Also models with two binding sites for NMDA, and one binding site for glycine, were unsatisfactory. Model 2 predicts glycine-sensitive desensitization in response to high concentrations of NMDA for the following reasons. In our experiments, prior to the application of NMDA, receptors were first equilibrated with glycine (promoting transitions from bottom left to top left). Then, with glycine still present, NMDA was applied rapidly, promoting transitions to states on the right, including a closed state which has bound both NMDA and glycine and which is assumed to be in rapid equilibrium with the open state. Because binding of NMDA lowers the affinity for glycine, glycine then dissociates from the NMDA receptor (transitions from top right to bottom right) reducing the probability of transitions to the open state. At high concentrations of glycine (1-3/,M), the rate of binding of glycine to the NMDA receptor (transitions from bottom right to top right) is fast relative to the rate at which glycine dissociates, and thus there is little desensitization. At intermediate concentrations of glycine (100-300 nM), the rate of desensitization is determined by both the dissociation rate constant for glycine, and the forward rate at which glycine binds to the receptor. At concentrations of glycine < 100 nm the desensitization rate is determined largely by the rate constant for dissociation of glycine.

Glycine-resistant desensitization Sather, Johnson, Henderson & Ascher (1990) have found that application of NMDA to neurones cultured from mouse cortex evoked rapidly desensitizing responses, especially in outside-out patches, as well as calcium-dependent slow desensitization similar to that noted in our experiments. This fast desensitization is not blocked at high concentrations of glycine. In some of our experiments we noted that following prolonged periods of whole-cell recording ( > 20 min), combined with repeated applications of NMDA, responses to NMDA showed substantial glycineresistant fast desensitization. The onset of this behaviour was strongly accelerated by the use of CsF-based intracellular solutions, similar to those used by Sather et al. (1990). The mechanism underlying the greater desensitization recorded with intracellular solutions containing fluoride versus methanesulphonate is unknown, but second messenger-mediated modification of intracellular amino acid residues of NMDA receptor channel protein molecules may be involved. Because glycine still potentiates initial responses to NMDA after intracellular dialysis with CsF, it is tempting to speculate how the majority of our results (where glycine does block desensitization) might be related to glycine-resistant desensitization described by Sather et al. (1990). A modification of the state diagrams presented above is capable of accounting for both sets of observations, if an additional desensitized state of the

KINETIC EXPERIMENTS ONXMDA RECEPTORS

353

NMDA receptor is included (for simplicity a model with only one binding site each for glycine and NMDA is illustrated): Desensitized t |Cytoplasmic regulation

RGlY

K

Fast

Open

K3

K2 R

RGDA

KK1

RNMDA

In the majority of our experiments we suggest there is little entry into the desensitized state. We suggest that entry into the desensitized state can become significant following some modification of the NMDA receptor, which is promoted by high extracellular concentrations of calcium, by intracellular dialysis with fluoride during whole-cell recording, or by outside-out patch recording. This could occur either if the rate of recovery from the desensitized state becomes slower, or because the rate of entry into this state becomes faster. In this scheme glycine is still required for, and strongly potentiates responses to NMDA, but does not block desensitization. Because it seems probable that during the initial period of whole-cell recording there will be less disturbance of the cytoplasmic components of transmembrane proteins, than occurs following the formation of outside-out patches, our results with glycine-sensitive desensitization are likely to be of physiological significance for the behaviour of NMDA receptors in intact cells. However, because our experiments were conducted with only 0'2 mM-extracellular Ca2+, a complete description of NMDA receptor desensitization will need to include a study of the effect of divalent cations on the glycine-sensitive component of desensitization. APPENDIX

Experimental measurement revealed that the change in agonist concentration at the surface of a neurone following a switch from one flow pipe to another has an approximately exponential time course (Vyklicky et al. 1990). For the purposes of the model developed below, if at time, to, a switch is made from one flow pipe which contains control solution with no agonist, to another which contains agonist at concentration, C0, then the cell is assumed to experience a change in agonist concentration, C(t) = Co(1-exp((t0-t)/T)), (1) where C(t) is the agonist concentration at the surface of the cell (assumed constant over the entire surface) at time t and t > to and where T is the solution exchange time constant. If at a later time t1 a switch is made from the test solution back to the control solution, the agonist concentration is then assumed to be

C(t) = C0(exp ((t1- t)/r) - exp ((to- t)/r)),

(2)

where t > t1. 12

PHY 428

354

M. BENVENYISTE AND OTHERS

The receptor channel states are numbered in an arbitrary manner, and the rate constant for a transition between the ith and the jth state is labelled kij. Given the distribution of states at time t it is possible to estimate the new distribution at time t+At and the evolution of the number of channels in each of the n states 81(t), S2(t)I... S(t) is calculated by applying this estimation iteratively. Consider a pair of states numbered i and j. We make the approximation that the number of transitions from i to j, is given by

Si (t) rij(t) At, where the transition rate,

(3)

ri,(t), is ri (t)

=

kij(4)

for agonist independent transitions, and

ri,(t) = C(t) k

(5)

for agonist dependent transitions. This estimate of the number of transitions will only be accurate when (rij(t) At) is much less than one (i.e. when the reaction rate is slow relative to the time increment). Similarly, the number of transitions from j to i will be (6) Sj(t) rji(t) At. Let the change that occurs during the time increment from t to t + At in the number of channels that are in state i due to transitions between i and j be ASij(t). From approximations (3) and (6) we have

ASi,(t) = (Sj(t) rji(t) -Si(t) rij(t)) At. (7) When the value of (ri,(t) At) or of (rji(t) At) is close to or greater than one, the above approximations no longer hold. Under these circumstances we assume that during At the reaction between the two states will come to approximate equilibrium, and then use the equilibrium condition to determine the number of transitions that will occur between the states. At equilibrium, the rate of transitions from state i to j is equal to the rate of transitions from state j to i:

Si(t + At) rij(t + At) = Sj(t + At) rji(t + At), (8) but the total number of channels remains constant,

Si(t) +Sj(t) = Si(t + At) + Si(t + At). Combining eqns (8) and (9) so as to eliminate multiplying top and bottom by At gives

S,(t + At),

(9) then rearranging and

ASii(t) = (Sj (t) rji (t) - Si(t) rij (t)) At/ (rij (t) + rji (t)) At. (10) Note that the numerator of eqn (10) is identical to the expression for ASij(t) deduced from the assumption of slow reaction rates (see eqn (7)). This suggests a simple approach to the problem of incorporating both slow and fast transition rates, and transition rates which vary from slow to fast with increasing agonist concentration, into the numerical model. At each iteration, the denominator of eqn (10), (rij(t) + rj(t)) At,

KINETIC EXPERIMENTS ON NMDA RECEPTORS

355

is calculated. and if the result is less than a pre-defined cut-off value, P'ut, it is set to one (the slow reaction approximation of eqn (7)), otherwise the calculated value is used (the fast reaction approximation of eqn (10)). The selection of a suitable cutoff value is considered below. So far the view of the model has been restricted to the pair of states numbered i and j. and these have been viewed in isolation from the rest of the multistate model. In order to calculate the new distribution of states at time t + At for the whole model, the above calculations are applied to each pair of states between which transitions occur. The number of channels in the i state at time t ±At is given by n

Si(t+At)

=

Si(t)+

E

ASZk(t),

(11)

k=1

where ASik(t) = 0 when k = i, or when there are no transitions between the i and k states; and ASik(t) is given by eqn (7) when (rij(t)+rji(t))At < Pcut, or by eqn (10) when (ri(t) + rji(t)) At >- Peut Equation (11) is used to calculate the number of channels in each of the n states at time t + At. It forms the basis of the iterative scheme. The value of Pcut is determined from stability considerations. First all agonistdependent transition rates are set to the maximum values they will take during the run. Next, Si(t) is set to one and the number of channels in all other states are set to zero. The value of Si(t+At) is calculated from eqn (11) with Peut set to one. Each of the n states is examined in the same way. If the minimum Si(t + At) calculated in the above manner, Smini is greater than zero, then the model is stable, and P?ut is left set to one, otherwise the following formula is used to compute a more conservative value of Pcut

Pcut = 1/(1 Smin)

(12)

The accuracy of the simulation was tested by comparing its predictions for a threestate model with the analytical solution for that model (Colquhoun & Hawkes, 1977). Several combinations of parameters were chosen which caused the numerical simulation to use both the slow and fast reaction approximations. It was found to be accurate ( < 2 % error at times > 20At) for all parameter combinations. In general, At was chosen to be 1/10th of the sample interval being modelled. Fitting model parameters to experimental data A search procedure was developed which identified the combination of model parameters that results in the optimum fit between a given family of current transients recorded in response to the application of a series of different agonist concentrations, and the corresponding family of simulated transients. The procedure is based on the Simplex multidimensional optimization algorithm (Kowalik & Osborne, 1968), and it seeks to minimize the sum of the squared errors between the recorded transients and the corresponding model transients. Most of the model parameters are optionally free variables in the search procedure. These include the number of channels, the association and dissociation rate constants, and microscopic equilibrium dissociation constants. It is redundant to allow both the single-channel current and the number of channels to be free variables, so single-channel current is 12-2

356

M. BEN VEN'VISTE AND OTHERS

held constant during the search. In addition, several rate constants may be held equal to one another while still being allowed to vary in the search procedure, and constraints imposed by the law of microscopic reversibility are automatically accounted for. The program permits a search based on the assumption of several identical and independent drug binding sites. When the search procedure was applied to a family of simulated transients, it converged accurately (< 5 % error) on the parameter values used in the original simulation. We thank C. Winters and S. Fitzgerald for preparing and maintaining the cultures used in our experiments, Dr B. Smith, Dr S. Hsiao, W. Holsinger, J. Ries and N. Simmons for building the rapid perfusion apparatus, and K. WVeeks for maintaining the Microvax 3600. Electron microscopic analysis was performed by Dr L. Williamson. J. C. and L. V. were Fogarty visiting fellows. M. B. is an NRC research fellow. REFERENCES

ASCHER, P., BREGESTOVSKI, P. & NOWAK, L. (1988). N-Methyl-D-aspartate-activated channels of mouse central neurones in magnesium-free solutions. Journal of Physiology 399, 207-226. BLAIR, L. A. C., LEVITAN, E., MARSHALL, J., DIONNE, V. E. & BARNARD, E. A. (1988). Single subunits of the GABAA receptor form ion channels with properties of the native receptor. Science 242, 577-579. BONHAUS, D. WV., YEH, G. C., SKARYAK. L. & McNAMARA, J. 0. (1989). Glycine regulation of the N-methyl-D-aspartate receptor-gated ion channel in hippocampal membranes. Mlolecular Pharmacology 36, 273-279. ;RISTOW, D. R., BOWERY, N. & WOODRUFF, G. N. (1986). Light microscopic autoradiographic localization of [3H]glycine and [3H]strychnine binding sites in rat brain. European Journal of Pharmacology 126, 303-307. BURGEN, A. S. V. (1966). The drug-receptor complex. Journal of Pharmacy and Pharmacology 18, 137-149. CACHELIN, A. B. & COLQUHOUN, D. (1989). Desensitization of the acetylcholine receptor of frog end-plates measured in a Vaseline-gap voltage clamp. Journal of Physiology 415, 159-188. COLQUHOUN, D. & HAWKES, A. G. (1977). Relaxation and fluctuation of membrane currents that flow through drug-operated channels. Proceedings of the Royal Society B 199, 231-262. CULL-CANDY, S. G. & Usowicz, M. M. (1989). On the multiple-conductance single channels activated by excitatory amino acids in large cerebellar neurones of the rat. Journal of Physiology 415, 555-582. FADDA, E., DANYSZ, W., WVROBLESKI, J. T. & COSTA. E. (1988). Glycine and D-serine increase the affinity of the N-methyl-D-aspartate sensitive glutamate binding sites in rat brain synaptic membranes. Neuropharmacology 27, 1183-1185. HOWE, J. R., COLQUHOUN, D. & CULL-CANDY, S. G. (1988). On the kinetics of large-conductance glutamate-receptor ion channels in rat cerebellar granule neurons. Proceedings of the Royal Society B 233, 407-422. HUETTNER, J. E. (1989). Indole-2-carboxylic acid: a competitive antagonist of potentiation by glycine at the NMDA receptor. Science 243, 1611-1613. JAHR, C. E. & STEVENS, C. F. (1987). Glutamate activates multiple single channel conductances in hippocampal neurones. Nature 325, 522-525. JOHNSON, J. W. & ASCHER, P. (1987). Glycine potentiates the NMDA response of mouse central neurones. Nature 325. 529-531. KATZ, B. & THESLEFF, S. (1957). A study of the 'desensitization' produced by acetylcholine at the motor end-plate. Journal of Physiology 138, 63-80. KESSLER, M., TERRAMANI, T., LYNCH, G. & BAUDRY, M. (1989). A glycine site associated with N-methyl-D-aspartic acid receptors: characterization and identification of a new class of antagonists. Journal of Neurochemistry 52, 1319-1328. KISKIN, N. I., KRISHTAL 0. A. & TSYNDRENKO, A. Y. (1986). Excitatory amino acid receptors in hippocampal neurones: kainate fails to desensitize them. Neuroscience Letters 63, 225-230.

KINETIC EXPERIMENTS OXN MDA RECEPTORS

357

KLECKNER, N. WV. & DINGLEDINE, R. (1988). Requirement for glycine in activation of N-methylD-aspartic acid receptors expressed in Xenopus oocytes. Science 241, 835-837. KOWALIK, J. & OSBORNE, M. R. (1968). Methods for Unconstrained Optimization Problems. Elsevier, New York. KUSHNER, L., LERMA, J., ZUKIN, R. S. & BENNETT, M. V. L. (1988). Coexpression of N-methyl-Daspartate and phencyclidine receptors in Xenopus oocytes injected with rat brain mRNA. Proceedings of the National Academy of Sciences of the USA 85, 3250-3254. LANGOSCH, D., THOMAS, L. & BETZ, H. (1988). Conserved quaternary structure of ligand-gated ion channels: the postsynaptic glycine receptor is a pentamer. Proceedings of the NVational Academy of Sciences of the tSA 85, 7394-7398. NIAMALAKI, C., STEPHENSON, F. A. & BARNARD, E. A. (1987). The GABAA/benzodiazepine receptor is a heterotetramer of a and /? subunits. European M.1olecular Biology Journal 6, 561-565. MAYER, M. L. & VYKLICKY, L. JR (1989). Concanavalin A selectively reduces desensitization of mammalian neuronal quisqualate receptors. Proceedings of the National Academy of Sciences of the USA 86, 1411-1415. MIAYER, M. L., VYKLICKY, L. JR & CLEMENTS, J. D. (1989). Regulation of NMDA receptor desensitization in mouse hippocampal neurones by glycine. NVature 338, 425-427. MONAGHAN, D. T., OLVERMAN, H. J., NGUYEN, L., WATKINS, J. C. & COTMAN, C. W. (1988). Two classes of N-methyl-D-aspartate recognition sites: differential distribution and differential regulation by glycine. Proceedings of the National Academy of Sciences of the USA 85, 9836-9840. REYNOLDS, I. J., MURPHY, S. N. & MILLER, R. J. (1987). 3H-Labelled MK-801 binding to the excitatory amino acid receptor complex from rat brain is enhanced by glycine. Proceedings of the National Academy of Sciences of the UJSA 84, 7744-7748. SATHER, W., JOHNSON, J. W., HENDERSON, G. & ASCHER, P. (1990). Glycine-inseInsitive desensitization of NMDA responses in cultured mouse embryonic neurons. Neuron (in the Press). SMOLUCHOWSKI, M. V. (1916). Drei Vortraige iuber Diffusion, brownsche Molekularbewegung und Koagulation von Kolloidteilchen. Physikalische Zeitschrift 17, 557-571. SNELL, L. D., MORTER, R. S. & JOHNSON, K. M. (1988). Structural requirements for activation of the glycine receptor that modulates the N-methyl-D-aspartate operated ion channel. European Journal of Pharmacology 156, 105-110. SONTHEIMER, H., BECKER, C. M., PRITCHETT, D. B., SCHOFIELD, P. R., GRENNINGLOH, G., KETTENMANN, H., BETZ, H. & SEEBURG, P. H. (1989). Functional chloride channels by mammalian cell expression of rat glycine receptor. Neuron 2, 1491-1497. TANG, C. M., DICHTER, M. & MORAD, M. (1989). Quisqualate activates a rapidly inactivating high conductance ionic channel in hippocampal neurons. Science 243, 1474-1477. THEDINGA, K. H., BENEDICT, M. S. & FAGG, G. E. (1989). The N-methyl-D-Aspartic acid (NMDA) receptor complex: a stoichiometric analysis of radioligand binding domains. Neuroscience Letters 104, 217-222. TRUSSELL, L. 0. & FISCHBACH, G. D. (1989). Glutamate receptor desensitization and its role in synaptic transmission. Neuron 3, 209-218. VERDOORN, T. A. & DINGLEDINE, R. (1988). Excitatory amino acid receptors expressed in Xenopus oocytes: agonist pharmacology. Molecular Pharmacology 34, 298-307. VYKLICKY, L. JR, BENVENISTE, M. & MAYER, M. L. (1990). Modulation of N-methyl-D-aspartic acid receptor desensitization by glycine in mouse cultured hippocampal neurones. Journal of Physiology 428, 313-331.

A kinetic analysis of the modulation of N-methyl-D-aspartic acid receptors by glycine in mouse cultured hippocampal neurones.

1. Responses to N-methyl-D-aspartic acid (NMDA) were recorded from mouse embryonic hippocampal neurones in dissociated culture, using whole-cell patch...
3MB Sizes 0 Downloads 0 Views