The Journal of Neuroscience, February 1993, 73(2): 674-684

Inactivation of NMDA Channels in Cultured Hippocampal Neurons by Intracellular

Pascal Legendre,‘va Christian Rosenmund,i~3 and Gary L. Westbrook’** ‘Vellum Institute and Departments of *Neurology and 3Physiology, Oregon Health Sciences University, Portland, Oregon 97201

Calcium-dependent inactivation of NMDA channels was ex- [Key words: NMDA, glutamate receptors, calcium, desen- amined on cultured rat hippocampal neurons using whole- sitization, channels, hippocampus] cell voltage-clamp and cell-attached single-channel record- ing. An ATP regeneration solution was included in the patch Calcium acting as a second messengerplays a key role in neu- pipette to retard current “rundown.” In normal [Ca*+L (l-2 ronal function. As a result, intracellular calcium concentrations mM) and 10 NM glycine, macroscopic currents evoked by 15 are tightly regulated by a seriesof cellular buffers and pumps set applications of NMDA (10 PM) inactivated slowly follow- (Carafoli, 1987; Blaustein, 1988). Thus, in most casesthe effects ing an initial peak. At -50 mV in cells buffered to [Ca2+li of intracellular calcium transients, either from transmembrane < 1 O-8 M with 10 mh! EGTA, the inactivation time constant influx or via releasefrom intracellular stores,are relatively lo- (T~“~~,) was -5 sec. Inactivation did not occur at membrane calized nearthe sourceof calcium entry or release.For example, potentials of +40 mV and was absent at [Ca2+10 10.2 mM, Ca2+transients in “domains” near presynaptic releasesites pro- suggesting that inactivation resulted from transmembrane vide the brief, high concentrations of calcium necessary for calcium influx. The percentage inactivation and 7inact were transmitter release(e.g., Smith and Augustine, 1988; Llinas et dependent on [Caz+lO. The rinact was also longer with BAPTA al., 1992). Focal increasesin calcium in dendritic spinesalso in the whole-cell pipette compared to EGTA, suggesting that occur following synaptic excitation (Miiller and Connor, 1991). ~~~~~~reflects primarily the rate of accumulation of intracellular This compartmentalization has important functional conse- calcium. quences.For example, the induction of long-term potentiation Inactivation was incomplete, reaching a steady state level in the CA1 region of the hippocampus is triggered by postsyn- of 40-50% of the peak current. At steady state, block of aptic calcium influx through NMDA, whereas depolarization open NMDA channels with MK-801 (( +)d-methyl-lo,1 1 -dih- and opening of voltage-dependent calcium channels is not as ydro-5H-dibenzo[a,d]cyclohepten-5, lo-imine) completely effective (Nicoll et al., 1988). blocked subsequent responses to NMDA, suggesting that One target of intracellular calcium is voltage-dependentand “inactivated” channels can reopen at steady state. Inacti- -gated ion channels (e.g., Scubon-Mulieri and Parsons, vation was fully reversible in the presence of ATP but was 1977; Chad et al., 1984; Inoue et al., 1986; Behrends et al., not blocked by inhibiting phosphatases or proteases. In cell- 1988; Marchenko, 1991). Calcium increasesthe activity ofCa2+- attached patches, transient increases in [Ca2+li following dependent K+ or Cl- channels(Marty, 1989), but decreasesthe cell depolarization also resulted in inactivation of NMDA activity of voltage-dependent calcium channels (Chad, 1989). channels without altering the single-channel conductance. Calcium can bind directly to ion channels or act via Ca2+- This suggests that Ca *+-dependent inactivation occurs in dependentenzymes (Chad and Eckert, 1986; Marty, 1989; Arm- intact cells and can be triggered by calcium entry through strong et al., 1991). Caz+-dependent inactivation of calcium nearby voltage-gated calcium channels, although calcium channels appearsto involve the action of a phosphatase,but entry through NMDA channels was more effective. We sug- other calcium-dependent processesmay also contribute (Chad gest that [Ca2+], transients induce NMDA channel inactiva- and Eckert, 1986; Belles et al., 1988; Korn and Horn, 1989). tion by binding to either the channel or a nearby regulatory For calcium-permeablechannels, inactivation could limit trans- protein to alter channel gating. This mechanism may play a membrane calcium flux during prolonged depolarization, and role in downregulation of postsynaptic calcium entry during thus could be an important protective factor to minimize Ca*+- sustained synaptic activity. induced neuronal death (e.g., Choi, 1988; Lynch and Seubert, 1989). The NMDA receptor is subject to modulation by a number Received June 15, 1992; revised Aug. 10, 1992; accepted Aug. 14, 1992. of extracellular agentsincluding MgZ+, glycine, Zn2+, and pro- This work was supported by U.S. Public Health Service Grants NS26494 and MH466 13 and the M&night Foundation (G.L.W.), and by NATO Foundation tons (for review, seeMayer and Westbrook, 1987; Collingridge and INSERM (P.L.). We thank Drs. Robin Lester and Craig Jahr for helpful and Lester, 1989), but intracellular regulation is much lessclear. discussions, and Jeff Volk for preparation of cell cultures. P.L. and C.R. contrib- Several experiments have suggestedthat calcium can to uted equally to these experiments. Correspondence should be addressed to Gary L. Westbrook, Vellum Institute, inactivation (or desensitization) of the NMDA channel, al- L474, Oregon Health Sciences University, 3 181 SW Sam Jackson Park Road, though both intracellular (Mayer and Westbrook, 1985; Mayer Portland, OR 97201. et al., 1987) and extracellular mechanismshave been proposed =Present address: Institut Pasteur, Departement des Biotechnologies, INSERM U 261, 75724 Paris Cedex 15, France. (Zorumski et al., 1989; Clark et al., 1990). Copyright 0 1993 Society for Neuroscience 0270-6474/93/l 30674-l 1$05.00/O We used whole-cell voltage clamp and cell-attached patch The Journal of Neuroscience, February 1993, 13(2) 675 recording to examine calcium-dependent inactivation of NMDA In some experiments, voltage jumps were used to evoke voltage- receptors in cultured hippocampal neurons. Calcium influx dependent calcium currents. In these experiments extracellular calcium was increased to 5 mM during the voltage jump. Jump protocols and through NMDA channels resulted in slow inactivation that membrane currents were recording on an IBM-PC using ~CLAMP soft- reached a maximum of 50% of the initial peak current. The rate ware (version 5.5). No leak subtraction was used. of inactivation appeared to reflect the rate of intracellular cal- Single-channelrecording. Cell-attached recording was made using cium accumulation rather than the kinetics of the Ca2+-depen- relatively large patch electrodes (l-2 Ma) containing Ca2+-free extra- dent process, and did not appear to involve Ca2+-dependent cellular solution buffered with 10 mM EGTA. L-Glutamate (1 NM) and glycine ( 10 PM) were included in the patch pipette. Single-channel cur- proteases or phosphatases. We suggest that calcium binds to rents were stored on a digital data recorder (VRlO, Instrutech), and then either the NMDA channel or a nearby regulatory protein to replayed, filtered at 2 kHz (eight-pole Bessel, Frequency Devices), dig- regulate the kinetics of the NMDA channel. itized at 10 kHz, and analyzed on an IBM AT clone using PCLAMP software (version 5.5). Single-channel conductance was checked for each Materials and Methods patch using the observed reversal potential. Software developed by John Clements (AXOGRAPH, Axon Instruments) was used to plot cumulative Cell cultures.Cultures of hippocampal neurons were prepared as pre- point-per-point amplitude histograms and obtain estimates of channel viously described (Legendre and Westbrook, 1990). Hippocampi from opening probability (np)where n is the number of channels and p isthe postnatal day 1 Sprague-Dawley rat pups were incubated in a low- probability of opening for a single channel. calcium saline with papain (5-20 U/ml; Worthington Biochemicals) for Results are presented as mean values + SD. Statistical comparisons 1 hr. The papain was inactivated in bovine serum albumin and trypsin were made using the Student’s t test. inhibitor; the tissue was mechanically dissociated and plated onto a confluent layer of hippocampal astrocytes. The culture medium con- tained minimum essential medium, 0.6% glucose, 5% heat-inactivated Results horse serum (Hyclone), and a supplement including 200 pg/rnl trans- Whole-cell recordings were performed on hippocampal neurons ferrin, 200 PM putrescine, 60 nM sodium selenite, 40 in.4 progesterone, 40 rig/ml corticosterone, 20 r&ml triiodothyronine, and 10 Ilg/ml in- after 7-14 d in culture. When neurons were clamped near the sulin. Cultures were treated 1 d after plating with a mixture of S-fluoro- resting potential in normal extracellular calcium ([Ca2+], = l- 2-deoxyuridine and uridine (15 and 35 &ml, respectively) to suppress 2 mM>, NMDA-evoked inward currents in Mg2+-free solutions overgrowth of background cells; half-changes of medium were done showed Ca*+-dependent inactivation in all neurons tested. In twice weekly. order to limit rundown of the NMDA responsesduring long Whole-cellrecording and drugdelivery. Whole-cell voltage-clamp re- cordings were performed on neurons after l-2 weeks in culture. The recordings, an ATP “regeneration” solution (Forscher and Ox- extrac&ular solution contained (in mM) Na+, 162, K+, 2.4, Ca*+, 1.3; ford, 1985; MacDonald et al., 1989) was added to the intracel- Ma*+. 0: Cl-. 167: HEPES. 10: glucose. 10: DH adiusted to 7.3. Tetro- lular solution. This allowed stable recordings for periods up to do;oxin’(b.5 ;M), strychnine (~'LM), and pidrbtoxin”(50 PM) were added 30 min. In most experiments, 15 set drug applications were to block spontaneous electrical activity and glycine/GABA channels, delivered every 100 set to reducecumulative effectsof repeated respectively. Glycine (10 PM) was also added to facilitate activation of NMDA-evoked currents. In some experiments, [Ca2+10 was varied be- applications. tween nominally zero and 50 mM; sodium was adjusted to keep the osmolarity at 325 mOsm. Patch pipettes for whole-cell recording were Slow voltage-dependentinactivation by Ca2+ pulled from borosilicate glass (TWF 150, World Precision Instruments), coated with Sylgard, and fire polished. DC resistances were 2-10 Ma. Both glycine-dependent and glycine-independent desensitiza- Pipette solutions contained (in mM) CsCl, 95; EGTA, 10; HEPES, 10; tion of the NMDA receptor have been observed that are not and an ATP “reeeneration” solution (Forscher and Oxford, 1985; Mac- calcium dependent (Mayer et al., 1989; Sather et al., 1990). In Donald et al., 1389) including Na-A‘TP, 4; MgCl,, 4; disodium.phos- phocreatine, 20; and creatine phosphokinase, 50 U/ml. Patch solutions order to examine Caz+-dependentinactivation in isolation, neu- were prepared daily from frozen stocks and kept on ice until use. The rons were voltage clamped at -50 mV in the presenceof low pH was adjusted to 7.2 with CsOH; osmolarity was adjusted to 295 concentrations of NMDA (l-l 0 PM) and high concentrations of mOsm. In some experiments, intracellular calcium was increased using glycine (10 hi). Under theseconditions, there was little apparent Ca/EGTA mixtures; the pipette calcium concentration was calculated desensitization/inactivation of the NMDA-evoked inward cur- assuming an EGTA dissociation constant of 1O-7 M. The chamber was continuously perfused (l-2 ml/min) at room temperature (-20°C). rents in low [CaZ+10(Fig. 1A). However, when [Caz+lowas raised Whole-cell currents were recorded using an Axopatch 1B (Axon Instru- to 1.3 mM, a slow current relaxation appearedthat approached ments) with the current filtered at 10 kHz. Membrane current records a steady state level within 1O-l 5 sec. We have called this re- were monitored on a chart recorder and stored on either on computer laxation “inactivation” to avoid confusion with other forms of or videotape (VRlO, Instrutech Corp.). For display purposes, currents NMDA receptor desensitization. When the membranepotential were filtered at 30-100 Hz. NMDA (l-300 PM) was dissolved in the extracellular solution and was shifted to +40 mV, the Ca2+-dependentrelaxation was applied via an array of flow pipes (400 pm i.d.) positioned within lOO- markedly reduced (Fig. 1A). 200 pm of the cell. Each flow pipe was controlled by solenoid values; A small residual relaxation was often observed even at pos- solutions were changed by simultaneously closing one valve and opening itive membrane potentials. This relaxation was unaffected by another. The solution exchange time constant was ~20 msec as mea- sured by change in membrane current evoked by kainic acid in two changesin [Ca2+lo,was much slower than the Ca2+-dependent concentrations of sodium (Vyklickq et al., 1990). Ultrapure calcium and inactivation, and was not further analyzed. At a holding poten- sodium salts (Aldrich) were used in the flow pipes for drug delivery. In tial of -50 mV, there was also a small decreasein the peak exoeriments with Fm. CsF (50 mM) was substituted isotonically for CsCl current amplitude in 1.3 mM [Ca2+10(13 + 4%, n = 14) com- in-the patch solution. Stock solutions of calmidazolium (10 mM in pared to 0.2 mM [Ca2+10.This is consistent with a reduction in dimethyl sulfoxide, Calbiochem) and okadaic acid (100 fig/ml in N,N- dimethylformamide, Calbiochem) were diluted into the pipette solution the single-channelconductance of the NMDA channels as [CaZ+10 before each experiment. Fifteen minutes were allowed after the start of increases(Jahr and Stevens, 1987; Ascher and Nowak, 1988). whole-cell recording for the substance to diffuse into the cell; this is We examined several simple possibleexplanations for CaZ+- more than sufficient to allow exchange of these low-molecular-weight dependent inactivation including changesin ion gradients, ac- compounds (Oliva et al., 1988; Pusch and Neher, 1988). The access resistance was regularly monitored with small voltage steps. Excitatory tivation of secondaryCaz+-dependent conductances, or a Ca*+- amino acids were obtained from Cambridge Research Biochemicals or induced change in agonist affinity. However, the reversal po- Tocris Neuramin; other chemicals were obtained from Sigma. tential was unaffected as shown by a 130 msec ramp (1 mV/ 676 Legendre et al. l Ca2+-Dependent Inactivation of NMDA Channels

A B 2nA-, peak 0.2 mM Cd’ 1.3 mM Cak

+4OmV J--L -I---L

C a b t

_I 500 pA 15 set 200 ms

Figure 1. Characteristics of calcium-dependent inactivation of NMDA currents. A, Whole-cell currents were evoked by flow pipe application of NMDA (10 PM, 15 set) in either low external calcium (0.2 mM; left truces) or normal calcium (1.3 mM; right truces). The current showed no macroscopic desensitization in low calcium at a holding potential of - 50 mV, but in 1.3 mM calcium the current inactivated slowly following an initial peak, reaching a steady state level of 47% inactivation for this neuron. The small relaxations seen at +40 mV were calcium independent. B, Calcium-dependent inactivation was not due to a change in the ionic driving force. A 130 msec voltage ramp from -70 mV to +60 mV was applied 500 msec after the beginning of the response (thick he), and after inactivation reached steady state at 14.5 set (thin he). The reversal potential was unchanged. Currents are leak subtracted. C, Inactivation was incomplete but involved the entire population of NMDA channels. NMDA (10 PM) was applied to a neuron at -50 mV. After steady state of inactivation was reached, the current was irreversibly blocked by a 40 set application of MK-801 (10 PM) in the continuous presence of NMDA (left panel). A subsequent test pulses of NMDA (500 msec, 0.2 mM [Cal+],) 1 min after removal of NMDA + MK-801 showed no recovery (truce b, right panel), suggesting that all channels opened at some time during the steady state inactivation period. The test pulse prior to MK-801 is shown in trace A, right panel. msec) obtained at the peak and end of a 15 set NMDA appli- the absenceof [Ca2+lu,presumably reflecting an additional form cation (Fig. 1B). The reversal potential was 4.1 + 2.7 mV at of receptor desensitization (see,e.g., Mayer et al., 1989; Sather the peak and 4.3 f 2.9 mV (n = 7) after the relaxation reached et al., 1990). However, in 1.3 mM [Ca*+],,,the relaxation reached steady state. Of note, the slope conductance was reduced over 41.6 f 11.8% (n = 10) after an NMDA application of 3.5 sec. the entire voltage range, suggestingthat recovery from calcium- This wasslightly lessthan the steady stateinactivation; however, dependentinactivation was slowerthan the duration of the ramp longer drug applications at high agonist concentrations resulted (seealso Fig. 2). Substitution of the impermeant anions gluco- in irreversible channel rundown (seeMacDonald et al., 1989). nate or methylsulfonate for chloride in the intracellular solution Thus, inactivation is not due to a decreasein agonist affinity. also did not affect the degreeof Caz+-dependent inactivation. Ca*+-dependentinactivation was not complete even during The inactivation was 47.8 * 5.9% (n = 4) with gluconate-con- prolonged agonist applications, usually approaching a maxi- taining pipettes and 45.2 + 5.3% (n = 5) with methylsulfonate- mum inhibition of 45-50%. This suggestedthe possibility that containing pipettes, compared to 43.7 f 6.4% with chloride- 45-50% of the channels are completely inactivated by calcium containing pipettes (n = 20; seeFig. 3). Thus, activation of an and the remainder are calcium insensitive. As MK-801 blocks inward Ca*+-activated chloride current is not responsible for open NMDA channelsbut exits extremely slowly from the chan- the Ca*+-dependentrelaxation under our conditions. nel (Huettner and Bean, 1988), we usedthis property to test for To test whether Ca*+-dependentinactivation was a result of separatechannel populations. MK-80 1 rapidly blockedthe steady reduced ligand affinity, we measuredthe relaxation usinga su- state NMDA current (Fig. lC, left); however, no recovery was pramaximal concentration of NMDA (300 PM). At this concen- seento a test NMDA pulse 1 min following the end of the tration, a small relaxation (15.0 +- 3.3%, n = 10) was seenin NMDA + MK-801 application (trace b, Fig. 1C). For four The Journal of Neuroscience, February 1993, 73(2) 677 neurons, the block at 1 min following NMDA + MK-801 was 97 * 2%. In the absence of MK-801, recovery of the NMDA A current was complete within 1 min (not shown; see also Fig. 2). 1.3 mM Ca2’ This suggests that channels inactivated by calcium reopen dur- ing the 40 set exposure to NMDA + MK-801, and that most (if not all) of the current is carried by channels that undergo Caz+-dependent inactivation. The onset and recovery from inactivation were examined by a 15 set exposure to 1.3 mM [Ca2+lo during a prolonged appli- cation of NMDA in low [Ca2+lo. As shown in Figure 2A, the onset consisted of a rapid reduction in current resulting from the change in channel conductance followed by a slow relaxation as observed during single applications of NMDA in normal 250 pA [Ca2+10. Following return to low [Ca2+lo solutions, there was -l complete recovery of the current with a half-recovery time of 10 s 22.6 f 3.7 set (n = 13). The slow recovery did not appear to result from trapping of Ca2+ in the inactivated channel, because full recovery occurred even in the absence of NMDA (Fig. 2B). In addition, the rate of recovery was not increased during continuous exposure to NMDA. This contrasts with the behavior of slow open channel blockers such as phencyclidine, ketamine, and MK-80 1 (Huettner and Bean, 1988; Mayer et al., 1988; MacDonald et al., 199 1).

Dependence on [Ca2+l, Both the maximal inactivation and the rate of inactivation were Figure 2. Inactivation was reversible on removal of extracellular cal- dependent on the extracellular concentration of calcium. During cium. A, During a long application of NMDA (10 PM, 70 set), [Ca*+], applications of 10 PM NMDA, there was little or no appreciable was increased from 0.2 to 1.3 mM for 15 sec. This was accompanied inactivation during a 15 set application in the absence of added by an instantaneous reduction in current (due to the reduced single- [Ca2+10; however, the addition of 0.2, 0.6, and 1.3 mM [Ca2+10 channel conductance) followed by slow Ca2+-dependent inactivation. On return to the low-calcium solution, recovery was complete. The produced a dose-dependent increase in inactivation (Fig. 3A). recovery followed a nonexponential time course with a half-recovery Inactivation was 9.6 f 11% at 0.2 [CaZ+10(n = 14), 29.2 f 2.4 time of a20 sec. B, Channel activation did not acceleratethe recovery % at 0.6 mM [Ca2+10(n = 6) and 43.7 ? 6.4 % at 1.3 mM [Ca*+L rate (compare A with B), suggesting that closed channel block was not (n = 20). There was no further increase in inactivation when responsible for inactivation. NMDA (10~~) was applied in 0.2 Ca*+ before, and 50 set after, a 15 set application of NMDA in 1.3 mrvrCa2+. the [Ca2+10was increased to 10 mM (Fig. 3B). In some neurons, As in A, recovery was complete. there appeared to be a slight delay in the onset of inactivation at low concentrations of [Ca*+],, as is apparent in Figure 3A. However, the time constant of inactivation (T~“*~,)at 1.3 mM overcome by raising the [Ca2+10to 10 mM. For the neuron shown [Ca2+10was generally well fitted with a single exponential of 4.7 in Figure 4A, the inactivation was 5 1.6% with a time constant + 0.6 set (n = 11; Fig. 3C). The T~,,=~,decreased as [Ca2+],, of 3.8 set; similar results were obtained on six neurons. This increased, that is, the rate of inactivation was accelerated. Al- also suggests that H+ or Mg 2+ transients, resulting from ex- though the degree of inactivation was maximal at 1.3 mM [Ca2+10, change with Ca2+ bound to EGTA, do not trigger inactivation. ~~~~~~continued to decrease, reaching 2.5 f 0.7 set (n = 8) and As noted above, rinac, was dependent on [Ca2+10. This could 1.2 f 0.3 set (n = 7) at 10 and 50 mM [Ca2+lo, respectively (Fig. result from the kinetics of a Ca2+ binding site or could simply 30). Increasing the concentration of NMDA to 100 KM resulted reflect the rate of accumulation of calcium at an intracellular in a left shift of the calcium concentration-response curve. In- effector site near the channel. Our results suggest the latter, as activation was 28.5 f 2.6% (n = 4) at 0.2 mM [Ca2+10, 41.5 f reducing EGTA from 10 to 0.1 mM produced a marked decrease 2.8% (n = 5) at 0.6 mM [Ca2+],, and 46.6 f 3.0% (n = 5) at 1.3 in 7inac,(Fig. 4B). The T,,,,, was1.7fO.lsec(n=6)atO.lmM, mM [Ca2+lo. which was only slightly slower (p < 0.02) than 7inaain the absence of added buffer (1.2 f 0.2 set, n = 7). Dependence on [Caz+], These results strongly suggest that an intracellular calcium NMDA channels are calcium permeable; thus, Ca*+-dependent transient following the opening of NMDA channels is respon- inactivation could result from accumulation of intracellular cal- sible for inactivation. Thus, we tested whether inactivation could cium. Although the experiments above were performed using be occluded by tonic increases in [Ca2+]i. Ca2+-dependent in- intracellular buffering with 10 mM EGTA, the increased inac- activation was examined in cells dialyzed with either 1 WM Ca2+ tivation at high concentrations of agonist suggested that the (= 10 mM Ca2+ + 11 mM EGTA) or 1 mM Ca2+ (= 10 mM degree and rate of inactivation were dependent on the buffer EGTA + 11 mM Ca2+). As shown in Figure 4C, the percentage capacity in the cell. As BAPTA is a much more rapid and inactivation was significantly less using the high-calcium patch selective Caz+ buffer, we examined inactivation during dialysis solutions compared to control cells dialyzed with 10 mM EGTA with 15 mM BAPTA. As shown in Figure 4A, there was no with no added calcium. Inactivation was 26.4 + 5.8% (n = 6, apparent inactivation during a 15 set application of 10 PM NMDA p < 0.01) for 1 PM Ca2+ and 18.4 + 8.9% (n = 5,p < 0.01) for at 1.3 mM [Ca*+lO. However, the effect of BAPTA could be 1 mM Ca2+. This compares to 43.7 f 6.7% (n = 20) inactivation 676 Legendre et al. l Ca2+-Dependent Inactivation of NMDA Channels

A B 60-4

Figure 3. Maximal inactivation and the inactivationrate were dependent on the extracellular calcium concentra- tion. A, The currentsevoked by NMDA (10 PM) in 0.2, 0.6, and 1.3 mM [Ca’+], werenormalized to the initial peakand superimposed.The inactivation pro- gressivelyincreased from a 27%reduc- tion in 0.6 mM to 4 1%in 1.3mM [Ca2+10 for this neuron.B, The percentagein- activationis shownas a functionof ex- tracellularcalcium. The nominallycal- WV o cium-freesolution had an estimated calciumconcentration of 20 fih4.Inac- tivation wasmeasured at the endof a C D 15set applicationof NMDA (10 PM). Eachpoint is the averageof 6-20 neu- rons. Intracellularbuffer was 10 mM EGTA with noadded calcium. The data werefitted with the logisticequation Z = I,,, l l/(1 + (IC,,/[Ca2+],~,where Z = percentageinactivation and I,,,,, = maximalinactivation (45.2%). The half- maximalinactivation (IC,,) was4 11 pM with a slopefactor (n) of 1.97.C, In- activationwas reasonably well fitted by a singleexponential with a time con- stantof =4 set in 1.3 mM [Caz+lo(10 mM EGTA buffering).D, Increasesin [Ca2+10above 1.3 mM acceleratedthe inactivationtime constantwithout al- 250 pA 1 10 100 tering the maximalinactivation. Each -I point is the averageof 6-l 1 neurons. 3s VW,

in control conditions (seeFig. 3B). This is the expected result test pulseof NMDA was immediate, but then decreasedduring if high intracellular calcium tonically inactivates NMDA chan- the pulse, presumably reflecting clearanceof calcium from the nels. The 7inac,was also decreased(Fig. 4C) similar to cells with effector site. Prepulsesof NMDA (50 PM, 5 mM Ca2+,150 msec) low concentrations of EGTA (Fig. 4C), indicating a rapid ac- in Ca*+-containing medium produced a similar degree of in- cumulation of free calcium. It was somewhat surprising that hibition (34.0 + 8.9%, n = 6; Fig. 5B). This could not be at- inactivation wasnot completely occluded even with 1 mM Ca2+ tributed to Ca*+-independent desensitization, as there was little in the pipette solution; that is, in the virtual absenceof a calcium macroscopic desensitization following application of 50 PM gradient, [Ca2+10= 1.3 mM. Thus, the free calcium concentration NMDA with [Caz+]O= 0. in the submembranecompartment is likely to be lessthan in Although NMDA channels are calcium permeable, the cal- the pipette solution, presumably reflecting continuing clearance cium flux representsonly a fraction of the total current (Mayer of calcium via transmembranepumps and exchangers. and Westbrook, 1987) compared to the high selectivity of the Under physiological conditions, severalsources may contrib- calcium channel (Lee and Tsien, 1984). However, chargetrans- ute to increasesin cytoplasmic calcium including voltage- and fer was 114 + 35 pC compared to 249 f 58 pC for the prepulse ligand-gated channelsas well as releaseof calcium from intra- of NMDA. If calcium entry through Ca2+channels was equally cellular stores (e.g., Miller, 1991). We examined whether cal- effective in producing inactivation of the NMDA channel, this cium flux through voltage-gated calcium channelscould also would require that 45% of the NMDA current was carried by induce inactivation of NMDA channels.Increases in cytoplas- calcium ions. This is much in excessof the estimatesfrom the mic calcium wereelicited either by voltage jumps or by prepulses GHK current equation (Mayer and Westbrook, 1987) and sug- of NMDA in Caz+-containing medium. The prepulsewas fol- geststhat the Ca2+ influx through NMDA channels is more lowed by a test pulseof NMDA in Ca2+-freemedium. As shown effective. in Figure 54, a 240 msecvoltage stepto + 10mV (5 mM [Ca2+]J The recovery from inhibition following a voltage stepto + 10 produced a sustainedinward calcium current. Following 250 mV is shown in Figure 5C. Recovery had a time constant of msecin calcium-free medium, the NMDA current activated by 12.7 f 6.3 set (n = 3) with a final recovery of 88 + 11% (n = the test pulse was inhibited by 32.0 f 9.8% (n = 14). The 4). This was more rapid than recovery from Ca2+-dependent reduction in the size of the NMDA current could be attributed inactivation produced by long application of NMDA (seeFig. to transmembranecalcium influx, as stepsto + 70 mV near the 2). These observations suggestthat calcium entry, not receptor Ca*+ equilibrium potential had no effect. The inhibition of the activation, is responsiblefor inactivation. The Journal of Neuroscience, February 1993, 13(2) 679

A B [BAPTAJ= 15 mM 7 j 10 mM 01% 5: Figure 4. Intracellularcalcium buffer- ing alsoinfluenced inactivation. A, In- activationwas significantly reduced us- ing BAPTA compared to EGTA in the whole-cell pipette. For this neuron, there was little inactivation in 1.3 mM [Ca*+10 in the presence of 15 mM BAPTA. However, even buffering with 15 mM BAPTA could be overcome with in- creases in [CaZ+],, to 10 mM, resulting in maximal (= 50%) inactivation. B, The T...,...“_. decreased when EGTA was re- 0 0.1 10 duced to 0.1 mM or omitted from the EGTA (mlK) whole-cell pipette. [Ca*+], was 1.3 mM. This suggests that rinac,reflects primarily C the rate of [Ca2+], accumulation. Each 10 mM EGTA 1 mM [Cak 1, point is the average of 6-20 cells. Data for 10 mM EGTA are the same as shown in Figure 30 for 1.3 mM [Ca2+10. C, Ca2+-dependent inactivation was also f reduced by increasing the buffered con- centration of calcium in the whole-cell dialysate to 1 PM (11 EGTA + 10 mM Ca*+) or 1 mM (10 EGTA + 11 mM Ca2+j. The control patch solution con- tained 10 mM EGTA with no added calcium. [Ca*+], was 1.3 mM, and 5 min of whole-cell dialysis preceded appli- cation of NMDA. With high intracel- lular calcium, the percentage inactiva- tion was markedly reduced and r,.,,, decreased. Examples from three neu- rons are shown.

or rate of Ca2+-dependent inactivation (n = 4). The percentage Divalent selectivity and role of Ca’+-dependentenzymes inactivation was 44.4 -t 5.8% at the beginning of recording In addition to Ca2+, several other divalent cations including compared to 43 + 5.7% following dialysis with okadaic acid. BaZ+,ST*+, and MrP+ can permeatethe NMDA channel (Mayer The calmodulin inhibitor calmidazolium (10 PM) also had no and Westbrook, 1987; Ascher and Nowak, 1988). However, effect on inactivation in four neurons. most Caz+/calmodulin-dependentenzymes show marked selec- tivity for Ca2+compared to other divalent cations (Chao et al., NMDA channel activity in cell-attached patches 1984). For example, the Ca*+-dependentphosphatase that de- Cell-attached recording provides one method to analyze ion phosphorylates voltage-dependent calcium currents cannot be channel regulation without disrupting the intracellular milieu. activated by barium (Chad and Eckert, 1986). However, in our NMDA-activated channelsin cell-attached patcheshave kinet- experiments, Ba*+ and Sr2+were partially effective in mimicking icssimilar to outside-out patches(Gibb and Colquhoun, in press). Ca2+-dependentinactivation of NMDA channels. The results In cell-attached patches from cultured hippocampal neurons, for one neuron are shown in Figure 6; the peak current varied low concentrations of L-glutamate (1 PM) activated inward sin- according to the predicted slopeconductances for thesedivalents gle-channel currents with amplitudes of 3-3.5 pA in normal (Mayer and Westbrook, 1987). In the presenceof 1.3 mM [Ba*+] extracellular [Ca”]. The channelswere absentwhen L-glutamate or [S++], inactivation was 28.4 + 1.4% (n = 7) and 27.2 f 2.2% was omitted from the pipette solution or when 50 PM 2-amino- (n = 8), respectively. This was significantly less than that due 5-phosphonovalerate was added (n = 5), consistent with selec- to equimolar Ca2+on the sameneurons (41.4 -t 4.5%, II = 8). tive activation of NMDA channels.Depolarization of the patch For the neuron shown in Figure 6, rinac,for Ca2+ was 3.9 set resulted in outward singlepotassium channel currents, but these compared to 2.8 and 1.3 set for Ba2+and SrZ+. were rare at a command potential of 0 mV (the cell resting We directly tested whether a proteaseor phosphatasecould potential). Although the cell membrane potential was not con- be responsible for Caz+-dependent inactivation. The protease trolled, measurementof the reversal potential of the NMDA inhibitor leupeptin (100 PM, n = 3) had no effect on inactivation channelsgave a predicted resting potential of - 60 to - 70 mV. when included in the patch pipette. Likewise, inactivation was The goal of theseexperiments was to assessthe effect of in- reversible, inconsistent with a proteolytic mechanism. Several creasesin cytoplasmic calcium on NMDA channelsin the patch. phosphataseinhibitors were also added to the whole-cell pipette To eliminate calcium flux through channels in the patch, the and allowed to diffuse into the cell over 10-l 5 min. Neither the cell-attached pipette contained 10 mM EGTA. In the absence nonselective phosphataseinhibitor FP (50 mM) nor the type of [CaZ+lO,the single-channelcurrent was 4.5-5 pA. When the I/2A inhibitor okadaic acid (5 PM) had any effect on the degree cell was briefly depolarized with 5 set pulsesof either 150 mM 660 Legendre et al. * Ca*+-Dependent Inactivation of NMDA Channels

Figure 5. The onset of inactivation re- quired less than 500 msec. A, Voltage jump from a holding potential of -50 mV (240 msec; V, = + 10 mV, [Cal+], = 5 mM) resulted in an inward calcium current. Following a delay of 250 msec in Ca*+-free medium, -the NMDA cur- rent (10 PM, [Ca2+10 = 0) was instan- taneously reduced by 40% in this neu- ron. However, a voltage jump to the calcium equilibrium potential (+ 70 mV) evoked an outward current, but had no effect on the NMDA current 100 ms 2s amplitude. The NMDA current follow- ing the jump to +70 mV is superim- C posed on the current in the absence of a voltage prepulse. The whole-cell pi- 10 UM NMDA. 0 Ca 2’ pette contained 1 mM EGTA with no added calcium and 100 FM leupeptin. B, A brief pulse of NMDA (50 PM) in Ca2+-containing medium (5 mM) also caused a rapid inhibition of the test NMDA pulse. Conditions were the same as for A. C, Recovery from inactivation was monitored by 300 msec NMDA pulses in Ca2+-free medium. Test puls- es immediately preceding (a), and fol- lowing (b), a 250 msec voltage jump to + 10 mV (see A) are shown in the left 600 i - , panel. Recovery was nearly complete I I I I I (88%) at 60 set following the voltage 0 20 40 60 pulse (right panel). Time (s)

KC1 + 5 mM Ca*+, or NMDA (100 PM) + 5 mM CaZ+, there ficient to induce maximal Ca2+-dependentinactivation, and that was an abrupt outward current and disappearanceof NMDA depolarization alone cannot account for the decreasein channel channels as the driving force collapsed. Following removal of activity. the KC1 solution, the membrane potential was quickly restored The amplitude histogramsimmediately before and after KC1 as indicated by the return of the single-channelcurrent to its are shown in Figure 7B, the probability dropped from 0.64 to basallevel (Fig. 7A). However, the activity of NMDA channels 0.26. The time courseof the drop in channel activity after KC1 in the patch was markedly decreased.Similar decreasesin ac- application wasestimated by calculating the probability in 1 set tivity with applications of KC1 or NMDA were not seenwhen epochs.For the patch shown in Figure 7C, the activity returned the pipette solutionscontained 1.3 mM [Ca2+].This suggeststhat to baselinelevels within 15 sec.For three patches,the inhibition the continuous activity of NMDA channels in the patch is suf- was 57 * 8%, which was similar to the maximal inactivation in whole-cell recording.

Discussion Ca2+ Ba2+ Sr2’ Our results demonstrate that intracellular calcium leadsto in- activation of the NMDA channel. The rate and degreeof in- activation were related to the influx and clearance of intracel- lular calcium, and inactivation could be triggered by opening of either NMDA or voltage-dependent calcium channels. In- activation was unaltered by phosphataseinhibitors, suggesting that dephosphorylation of the channel doesnot underlie Ca*+- dependent inactivation. 250 pA Ca2+-dependentinactivation resultsfrom increasesin 5.5 intracellular calcium Figure 6. Barium and strontium are less effective than calcium. NMDA Several distinct mechanismsappear to underlie desensitization (10 PM) was applied in the presence of [Ca*+],, Ba2+, or Sr2+, all at 1.3 mM. Responses are from one neuron at a holding potential of -50 mV. of NMDA channels.At submaximal concentrations of glycine, The percentage inactivation was 41.8%, 29%, and 23.8%, respectively. glutamate binding results in an allosteric reduction in glycine The T,,,~ was decreased for Ba*+ and Sr2+ compared to Ca2+ (see text). affinity with subsequentunbinding of glycine and channel clo- The Journal of Neuroscience, February 1993, 73(2) 661

A 150 mM KC1

Figure 7. Ca2+-dependent inactiva- tion is present in intact cells. A, In cell- attached recording, L-glutamate (1 PM) activated inward single-channel cur- rents of 4.5 pA (165 mM Na+, 0 Ca*+); the patch pipette contained 10 mM EGTA with no added Ca*+ to eliminate Ca2+ influx through channels in the patch. The patch was clamped at 0 mV. The cell membrane potential was de- termined by measuring the channel re- versa1 potential, which was usually near 15 pA -60 mV for stable natches. Cell de- 10 pA polarization (5 set, 150 mM KCl, 5 mM I Ca2+) caused an abrupt outward cur- 5s/300ms rent. Following removal of the depo- larizing stimulus, the membrane poten- tial quickly returned, as indicated by B C the return of the single-channel current amplitude to its basal level. However, 500- the activity of NMDA channelsin the patch was transiently decreased. One- 400- second epochs immediately before and 3 : after KC1 are shown on an expanded 0.6 5300- time scale. B, Point-by-point amplitude 0 histogram ofdata segment immediately z . LO.5 before (thin line) and after KC1 (thick %200- line) show the drop in the current peaks # . at 4.4 and 8.8pA. The openprobability loo- dropped from 0.64 to 0.26. C, Time course of recovery of NMDA channel activity is plotted in 1 set epochs. The 01 half-recovery time was 8 set for this -20 -15 -10 0 5 0 5 10 IS 20 25 30 35 40 patch.Solid bar indicatesperiod of KC1 Current (i*) Time (s) application.Same patch as A and B. sure (i.e., glycine-dependent “desensitization”; Mayer et al., periments using the calcium indicator dye arsenazo III also 1989). However, in outside-out membrane patches,prominent suggestedthat increasesin intracellular calcium could decrease glycine-independent desensitization leadsto 2 90% loss of the NMDA responses(Mayer et al., 1987). However, recently it has evoked current (Sather et al., 1990; Lester and Jahr, 1992). been suggestedthat the calcium acts at an extracellular site to Profound glycine-independent desensitization also appearsin enhancereceptor desensitization (Clark et al., 1990). Increases whole-cell recording of small or acutely dissociated neurons in extracellular calcium do reduce the conductance of NMDA (Sather et al., 1990; Zilberter et al., 1991). Neither of these channels(Jahr and Stevens, 1987; Mayer and Westbrook, 1987; phenomenacontributed to the Ca2+-dependentinactivation seen Ascher and Nowak, 1988), but this processis instantaneousand in our experiments becauseglycine was usedat saturating con- thus clearly distinguishablefrom the slow inactivation observed centrations, and macroscopic glycine-independent desensitiza- in our experiments. Our evidence strongly favors an intracellular tion was small, representingat most 20% of the peak current at site of action for inactivation. Both the degreeand rate of in- high agonist concentrations. In addition, no macroscopic de- activation were dependent on the extracellular [Ca2+], but this sensitization was observed with the low agonist concentrations was highly dependent on the concentration of EGTA in the (e.g., 10 PM NMDA) used. whole-cell pipette. Lowering the EGTA concentration would be The first indication that calcium might inactivate NMDA expected to increase the rate of accumulation of intracellular channels came from voltage-clamp studies of cultured spinal free calcium. Thus, the decreasein T,,,,~as the EGTA concen- cord neurons (Mayer and Westbrook, 1985). Subsequent ex- tration was decreased(or [Ca2+10increased) suggeststhat ac- 662 Legendre et al. . Ca2+-Dependent Inactivation of NMDA Channels cumulation of [Caz+]i determines the inactivation rate. Similar zymes have also been suggestedas contributing to calcium results were seen with BAPTA, excluding the possibility that channelinactivation in GH3 cellsand guineapig myocytes (Belles intracellular protons or Mg*+ (Johnson and Ascher, 1990) is et al., 1988; Kalman et al., 1988). Although Ca*+-dependent responsible for inactivation. However, even strong buffering inactivation of the NMDA currents is similar in many of the with high concentrations of BAPTA could be overcome when criteria proposedfor inactivation of molluscancalcium channels the calcium influx was sufficiently large. This may explain the (Chad, 1989) the mechanismsdiffer in two important respects. failure of other investigators to see a dependence on intracellular First, Caz+-dependentinactivation of NMDA currents doesnot buffering (e.g., Clark et al., 1990). appear to involve phosphatases,and was fully reversible, sug- Calcium-dependent inactivation had several characteristic gestingthat proteasesare not involved. Second, equimolar sub- features. The inactivation was slow and was enhanced at neg- stitution of Caz+ with BaZ+or Sr*+ is partially effective in pro- ative holding potentials that could be explained by calcium ducing inactivation. Ba*+ and Sr*+ have larger ionic radii and influx through open NMDA channels (Mayer et al., 1987). Al- thus are less effective agoniststhan Ca in stimulating calmo- though we did not determine the calcium sensitivity of the in- dulin-dependent processes(Chao et al., 1984). Ba2+ and Srz+ tracellular site, even quite high pipette calcium concentrations did evoke relaxations that were more rapid than Ca2+,perhaps did not completely occlude inactivation. This may suggest that suggestingthat buffering and clearancemechanisms are also less the site of calcium action has a relatively low affinity. This would efficient in handling divalents other than calcium (see,e.g., Ah- predict a rapid off-rate; thus, the slow recovery from inactivation med and Connor, 1979). Thus, a mechanism other than de- suggests that [Caz+]i clearance is the rate-limiting step in recov- phosphorylation appearsto underlie Ca2+-dependentinactiva- ery. Inactivation was maximal at physiological levels of extra- tion of the NMDA receptor. However, this does not exclude cellular calcium and also was present in intact cells as measured the possibility that the NMDA channel can be phosphorylated, by cell-attached recording. Thus, inactivation may be an im- as the NR-1 subunit does have a consensusCa2+/calmodulin- portant process during normal synaptic transmission. Inacti- kinase II/protein kinase C phosphorylation site (Moriyoshi et vation was also incomplete, reaching a steady state near 50% al., 199l), and protein kinase C-dependent potentiation of of the peak current; however, this could not be explained by NMDA responseshas been reported (Chen and Huang, 1991). complete inactivation of a subpopulation of channels, although a small population of Caz+-insensitive channels cannot be ex- Mechanism and site of action of intracellular cqlcium cluded. Other than dephosphorylation, calcium could conceivably act by direct binding to the channel, by binding to regulatory pro- Comparison to Ca2+-dependent inactivation of other ion teins, or by screeningnegative chargeson the membrane. Al- channels though our experiments do not directly addressthese possibil- Desensitization or inactivation of several ligand-gatedion chan- ities, they provide constraints on the possiblemechanisms and nels is calcium dependent. In the case of calcium-permeable site of action. For example, neither ligand binding nor channel channelsthis could serve as a regulatory mechanism.Although openingwas required to elicit Caz+-dependentinactivation, sug- the molecular action of calcium is not known in all cases,several gestingthat the putative binding site is accessiblefrom the cy- mechanismsappear to be involved. Manthey (1966) first re- toplasm, but unlikely to be deep within the channel pore. Like- ported increasesin desensitization of the muscle ACh receptor wise, inactivation was not overcome by increasing the agonist (AChR) by increasesin [CaZ+],, which were subsequently at- concentration; thus, a competitive reduction in the affinity of tributed to increasesin intracellular calcium (Scubon-Mulieri ligand binding is excluded. Recovery alsocontinued in the pres- and Parsons, 1977). However, this has not been observed in all ence of agonistwhen extracellular calcium was removed, unlike preparations (Anwyl and Narahashi, 1980) and potentiation of the use dependencethat occurs with closed channel block (e.g., neuronal AChRs via an extracellular action of Caz+ has also Huettner and Bean, 1988). Intracellular Mgz+ can bind within been reported (Vemino et al., 1992). Increasesin intracellular the NMDA channelsand reduceits conductance, but Ca*+even calcium can also suppressGABA, receptors in some neurons at 1 mM did not produce this effect (Johnsonand Ascher, 1990). (Inoue et al., 1986; Behrends et al., 1988; but see Shirasaki et In our experiments,the I-Vrelationship of the inactivated whole- al., 1992). Phosphorylation enhancesmuscle AChR desensiti- cell current was unchangedfrom control, which excludesa volt- zation, and it has been proposedthat calcium entry through the age-dependentblock by an intracellular divalent cation as the channels could activate protein kinase C leading to’channel causeof inactivation. The reversal potential of the inactivated phosphorylation (reviewed in Huganir and*Greengard, 1990). current was also unchanged; thus, membrane charge screening On the other hand, rundown of GABA, currents in hippocampal by [Ca2+]i at concentrations that induce inactivation must be neuronsis acceleratedby intracellular calcium, suggestinga role negligible. The calcium-induced reduction in channel activity for dephosphorylation by a Ca2+/calmodulin-dependentphos- in cell-attached patchessuggests an alteration in channel gating phatase(Chen et al., 1990). with no apparent effect on single-channelconductance. Overall, Perhaps the best-studied examplesof calcium-dependentin- our results are consistent with an allosteric binding site on the activation are L-type calcium channels in molluscan neurons channel, but do not exclude an action on a Ca*+-dependent (Chad et al., 1984). Channel activity in thesecells is dependent regulatory protein. on CAMP-dependent phosphorylation. Inactivation by intra- Although the site of calcium action does not appear to be cellular calcium appears to result from Ca2+-dependent de- within the channel pore, calcium influx through NMDA chan- phosphorylation, as it is mimicked by perfusion with the Caz+- nels was more effective in the induction of inactivation than dependent phosphatasecalcineurin (Armstrong, 1989; Chad, influx through voltage-dependent calcium channels. This sug- 1989). An additional irreversible component is blocked by leu- geststhat the site of calcium action is relatively near the intra- peptin, suggestingan additional component due to a Ca*+-de- cellular domain of the receptors. Although the relative distri- pendent protease(Chad and Eckert, 1986). Ca*+-dependenten- bution of voltage-dependent calcium and NMDA channels is The Journal of Neuroscience, February 1993, 13(2) 663 unknown, several observations suggest compartmentalization Carafoli E (1987) Intracellular calcium homeostasis. Annu Rev Bio- of intracellular calcium transients following activation of glu- them 56:395-433. Carafoli E (199 1) The calcium pumping ATPase of the plasma mem- tamate channels (Malenka et al., 1988; Mtiller and Connor, brane. Annu Rev Phvsiol 53:53 l-547. 1991). This may involve either channel location or variations Chad J (1989) Inactivation of calcium channels. Comp Biochem Phys- in calcium clearance in different compartments (Guthrie et al., iol93A:95-105. 1991). Although an action of calcium near the channel might Chad J, Eckert R, Ewald D (1984) Kinetics of calcium-dependent be expected to have a rapid onset, inactivation was slow during inactivation of calcium current in voltage-clamped neurones of Anlv- sia californica. J Physiol (Lond) 347:279-300. NMDA applications due to the time required for calcium ac- Chad JE. Eckert R (1986) An enzvmatic mechanism for calcium____ cur---- cumulation. Maximal inactivation could occur in ~500 msec rent inactivation in dialysed Heiix neurones. J Physiol (Lond) 378: following rapid increases in [Caz+]i by voltage steps. An inter- 31-51. esting possibility is that calcium accumulation to pro- Chao S-H, Suzuki Y, Zysk JR, Cheung WY (1984) Activation of calmodulin by various metal cations as a function of ionic radius. gressive changes in channel gating, as has been proposed for Mol Pharmacol 26:75-82. calcium-mediated inactivation of voltage-dependent calcium Chen L, Huang Y-YM (1991) Sustained potentiation of NMDA re- channels (Yue et al., 1990). This could also be significant during ceptor-mediated glutamate responses through activation of protein high-frequency stimulation of excitatory synapses in the hip- kinase C by a p opioid. Neuron 7:3 19-326. pocampus. Chen QX, Stelzer A, Kay AR, Wong RKS (1990) GABA, receptor function is regulated by phosphotylation in acutely dissociated guinea- pig hippocampal neurones. J Physiol (Lond) 420:207-22 1. Role of high-energy phosphates Choi DW (1988) Glutamate neurotoxicity and diseases of the nervous The activity of NMDA channels during whole-cell recording system. Neuron 1:623-634. has been reported to require high-energy phosphatespossibly Clark GD, Clifford DB, Zorumski CF (1990) The effect of -o------aeonist for direct receptor phosphorylation (MacDonald et al., 1989). concentration, membrane voltage and calcium on N-methyl-n-as- Dartate recentor desensitization. Neuroscience 39:787-797. We have also noted loss of NMDA current in the absenceof Cdllingridge GL, Lester RAJ (1989) Excitatory amino acid receptors ATP (“rundown”), and thus an ATP regeneration solution was in the vertebrate central nervous system. Pharmacol Rev 40: 143-2 10. included in our experiments.Under theseconditions, the NMDA Forscher P, Oxford GS (1985) Modulation of calcium channels by current was stable for recording periods up to 30-60 min. How- norepinephrine in internally dialyzed avian sensory neurons. J Gen Physiol 85~743-763. ever, in the absenceof ATP, not only did the peak current Gibb AJ, Colquhoun D (in press) Activation of NMDA receptors by decrease, but there was a nearly complete loss of the Caz+- L-glutamatein cellsdissociated from adultrat hippocampus.J Physiol dependentrelaxation (Rosenmundet al., 1991). This could mean (Lond), in press. that a channelthat has “run down” is no longer subjectto Ca*+- Guthrie PB, Segal M, Kater SB (1991) Independent regulation of cal- dependent inactivation, or that rundown itself is a Ca2+-depen- cium revealed by imaging dendritic spines. Nature 354:76-80. Huettner JE, Bean BP (1988) Block of N-methyl-n-aspartate-activated dent phenomenon as has been reported for GABA, channels current by the anticonvulsant MK-801: selective binding to open (Chen et al., 1990) and in our preliminary experiments with currents. Proc Nat1 Acad Sci USA 85: 1307-l 3 11. NMDA channels (Rosenmund et al., 1991). The second pos- Huganir RL, Greengard P (1990) Regulation of neurotransmitter re- sibility might indicate that ATP is required simply to enable ceptor desensitization by protein phosphorylation. Neuron 5:555- the clearance of calcium from the cytoplasm by the Na/Ca ex- 567. Inoue M, Oomura Y, Yakushiji T, Akaike N (1986) Intracellular cal- changer and calcium pump (Blaustein, 1988; Carafoli, 199 1). cium ions decrease the affinity of the GABA receptor. Nature 324: Thus, intracellular Ca2+may play a central role in the regulation 156-158. of NMDA channelsand provide negative feedback control over Jahr CE, Stevens CF (1987) Glutamate activates multiple single chan- calcium influx into dendritic spines. nel conductances in hippocampal neurons. Nature 325:522-525. Johnson JW, Ascher P (1990) Voltage-dependent block by intracellular Mg2+ of N-methyl-D-aspartate-activated channels. Biophys J 57: 1085- References 1090. Ahmed Z, Connor JA (1979) Measurement of calcium influx under Kalman D, G’Lague PH, Erxleben C, Armstrong DL (1988) Calcium- voltage clamp in molluscan neurones using the metallochromic dye dependent inactivation of the dihydropyridine-sensitive calcium arsenazo III. J Phvsiol (Land) 286:61-82. channels in GH, cells. J Gen Physiol 92:53 l-548. Anwyl R, Narahashi T (1980) Desensitization of the acetylcholine Kom SJ, Horn R (1989) Influence of sodium-calcium exchange on receptor of denervated rat soleus muscle and the effect of calcium. Br calcium current rundown and the duration of calcium-dependent J Pharmacol 69:91-98. chloride currents in pituitary cells, studied with whole cell and per- Armstrong DL (1989) Calcium channel regulation by calcineurin, a forated patch recording. J Gen Physiol 94:789-8 12. Ca2+-activated phosphatase in mammalian brain. Trends Neurosci Lee KS, Tsien RW (1984) High selectivity of calcium channels in 12:117-122. single dialysed heart cells of the guinea-pig. J Physiol (Lond) 354: Armstrong DL, Rossier MF, Shcherbatko AD, White RE (199 1) En- 253-272. zymatic gating of voltage-activated calcium channels. Ann NY Acad Legendre P, Westbrook GL (1990) The inhibition of single N-methyl- Sci 635~26-34. D-aspartate-activated channels by zinc ions on cultured rat neurones. Ascher P, Nowak L (1988) The role of divalent cations in the N- J Physiol (Lond) 429:429-449. methyl-D-aspartate responses of mouse central neurones in culture. J Lester RAJ, Jahr CE (1992) NMDA channel behavior depends on Physiol (Lond) 399~247-266. agonistaffinity. J Neurosci12:635-643. Behrends JC, Maruyama T, Tokutomi N, Akaike N (1988) Ca*+- mediated suppression of the GABA-response through modulation of Llinas R, Sugimori M, Silver RB (1992) Microdomains of high calcium chloride channel gating in frog sensory neurones. Neurosci Lett 86: concentrations in a presynaptic terminal. Science 256:677-679. 311-316. Lynch G, Seubert P (1989) Links between long-term potentiation and Belles B, Hescheler J, Trautwein W, Blomgren K, Karlsson JO (1988) neuronatholoav. Ann NY Acad Sci 568: 17 l-l 80. A possible physiological role of the Ca-dependent protease calpain MacDonald JF,Mody I, Salter MW (1989) Regulation of N-methyl- and its inhibitor calpastatin on the Ca current in guinea pig myocytes. D-aspartate receptors revealed by intracellular dialysis of murine neu- Pfluegers Arch 412:554-556. rones in culture. J Physiol (Lond) 414: 17-34. Blaustein MP (1988) Calcium transport and buffering in neurons. MacDonald JF, Bartlett MC, Mody I, Pahapill P, Reynolds JN, Salter Trends Neurosci 11:438-443. MW, Schneiderman JH, Pennefather PS (199 1) Actions of ketam- 664 Legendre et al. * Ca2+-Dependent Inactivation of NMDA Channels

ine, phencyclidine and MK-80 1 on NMDA receptor currents in cul- Nicoll RA, Kauer JA, Malenka RC (1988) The current excitement in tured mouse hippocampal neurones. J Physiol (Lond) 432:483-508. long-term potentiation. Neuron 1:97-103. Malenka RC, Kauer JA, Zucker RS, Nicoll RA (1988) Postsynaptic Oliva C, Cohen IS, Mathias RT (1988) Calculation of time constants calcium is sufficient for potentiation of hippocampal synaptic trans- for intracellular diffusion in whole cell patch clamp configuration. mission. Science 242:8 l-84. Biophys J 54~791-799. Manthey AA (1966) The effect of calcium on the desensitization of Pusch M, Neher E (1988) Rates of diffusional exchange between small membrane receptors at the neuromuscular junction. J Gen Physiol cells and a measuring patch pipette. Pfluegers Arch 4 11:204-2 11. 49~963-976. Rosenmund C, Legendre P, Westbrook GL (1991) Rundown of NMDA Marchenko SM (199 1) Mechanism of modulation of GABA-activated currents: dependence on ATP and calcium. Sot Neurosci Abstr 17: current by internal calcium in rat central neurons. Brain Res 546: 957. - 355-357. Sather W, Johnston JW, Henderson G, Ascher P (1990) Glycine- Marty A (1989) The physiological role ofcalcium-dependent channels. insensitive desensitization of NMDA responses in cultured mouse Trends Neurosci 12:420-424. embryonic neurons. Neuron 4~725-73 1. Mayer ML, Westbrook GL (1985) The action of N-methyl-D-aspartic Scubon-Mulieri B, Parsons RL (1977) Desensitization and recovery acid on mouse spinal neurones in culture. J Physiol (Lond) 361:65- at the frog neuromuscular junction. J Gen Physiol 69:431447. 90. Shirasaki T, Aibara K, Akaike N (1992) Direct modulation of GABA, Mayer ML, Westbrook GL (1987) The physiology of excitatory amino receptor by intracellular ATP in dissociated nucleus tractus solitarii acids in the vertebrate central nervous system. Prog Neurobiol 28: neurones of rat. J Physiol (Lond) 449:55 l-572. 197-276. Smith SJ, Augustine GJ (1988) Calcium ions, active zones and syn- Mayer ML, MacDermott AB, Westbrook GL, Smith SJ, Barker JL aptic transmitter release. Trends Neurosci 11:458-464. (1987) Agonist- and voltage-gated calcium entry in cultured mouse Vemino S, Amador M, Luetje CW, Patrick J, Dani JW (1992) Calcium spinal cord neurones under voltage clamp measured using arsenazo modulation and high calcium permeability of neuronal nicotinic ace- III. J Neurosci 7:3230-3244. tylcholine receptors. Neuron 8: 127-l 34. Mayer ML, Westbrook GL, Vyklicky L Jr (1988) Sites of antagonist Vyklicky L Jr, Benveniste M, Mayer ML (1990) Modulation of N- action on N-methyl-D-aspartate receptors studied using fluctuation methyl-D-aspartic acid receptor desensitization by glycine in mouse analysis and a rapid perfusion techniaue. J Neurovhvsiol60:645-663. cultured hippocampal neurones. J Physiol (Lond) 428:3 13-33 1. Mayer-ML, Vyklicky L Jr, Clements J- (1989) Regulation of NMDA Yue DT, Backx PH, Imredy JP (1990) Calcium-sensitive inactivation receptor desensitization in mouse hippocampal neurons by glycine. in the gating of single calcium channels. Science 250: 1735-1738. Nature 338:425-427. Zilberter Y, Uteshev V, Sokolova S, Khodorov B (1991) Desensiti- Miller RJ (1991) The control of neuronal Ca2+ homeostasis. Prog zation of N-methyl-o-aspartate receptors in neurons dissociated from Neurobiol 37:255-285. adult rat hippocampus. Mol Pharmacol 40:337-34 1. Moriyoshi K, Masu M, Ishii T, Shigemoto R, Mizuno N, Nakanishi S Zorumski CF, Yang J, Fischbach CD (1989) Calcium-dependent, slow (199 1) Molecular cloning and characterization of the rat NMDA desensitization distinguishes different types of glutamate receptors. receptor. Nature 354:31-37. Cell Mol Neurobiol 9:95-104. Miiller W, Connor JA (199 1) Dendritic spines as individual neuronal compartments for synaptic Ca*+ responses. Nature 354:73-76.