[1] 12w Comparing Long-Term Depression With Pharmacologically Induced Synaptic Attenuations in Young Rat Hippocampi

[1] 297w Field excitatory postsynaptic potentials (EPSPs) were recorded in the CA1 region of hippocampal slices from 12-18-day-old rats. The isolated N-methyl-Daspartate (NMDA) receptor mediated field EPSP as well as the composite field EPSP of both NMDA and a-amino-3-hydroxy-5-methylisoxazolepropionic acid (AMPA) receptor mediated components were obtained in low Mg 21 solutions with 10 µM or 1 µM of the AMPA receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), respectively. The isolated AMPA receptor mediated field EPSP was obtained either in normal Mg 21 solution or in a low Mg 21 solution in the presence of the NMDA receptor antagonist D-2-amino-5-phosphonopentanoic acid. The waveforms of the field EPSPs were studied and the effect of long-term depression (LTD) on these waveforms was compared with the effects of several pharmacological agents that attenuate the synaptic efficacy. It was shown that LTD occurred without changes in the waveforms of isolated AMPA and NMDA EPSPs. Reducing the number of release sites by lowering the stimulus strength or reducing the probability of transmitter release by an adenosine agonist N 6 -cyclohexyladenosine both mimicked the LTD-induced changes. Partial blockade of the AMPA receptors was also without effect on the waveforms of isolated AMPA EPSPs. In contrast, partial blockade of the NMDA receptors in several different ways resulted in waveform changes. A similar result could be inferred from experiments using composite field EPSPs. The synaptic attenuation caused by a partial blockade of NMDA receptors therefore appears to differ mechanistically from that involved in LTD, arguing against a postsynaptic locus of the modification involved in LTD. However, directly testing for alterations in transmitter release using the open channel blocker of NMDA receptors MK-801 failed in revealing such presynaptic changes during LTD. Our results therefore suggest that LTD might be due to a coordinated pre-and postsynaptic change instead of distinct pre-or postsynaptic modifications.

[1] 122w Long-term depression (LTD) is an activity-dependent decrease of synaptic efficacy that might be involved in mnemonic processes (Artola and Singer, 1993;Linden, 1994). There is a general consensus about its induction mechanism, especially for hippocampal homosynaptic LTD, which seems to involve activation of postsynaptic N-methyl-D-aspartate (NMDA) receptors leading to an increased Ca 21 level in the dendritic spine (Cummings et al., 1996;Dudek and Bear, 1992;Mulkey and Malenka, 1992). However, the subsequent biochemical steps, including the ultimate synaptic modification, are still elusive. Although a convincing case has been made for an initial activation of protein phosphatase (Mulkey et al., 1993(Mulkey et al., , 1994;;Xiao et al., 1995a) the available evidence diverges with respect to fundamental issues such as the implied location of the modification.

[2] 146w Our recent studies of LTD in the hippocampus, using a low Mg 21 solution to unblock the NMDA-mediated synaptic response, reveal that the a-amino-3-hydroxy-5methylisoxazolepropionic acid (AMPA) and NMDA mediated components of the synaptic potential undergo equal relative decreases (Xiao et al., 1995a,b). In addition, reducing transmitter release by pharmacological means led to changes resembling LTD in terms of equal changes of excitatory postsynaptic potential (EPSP) components. These results are in line with other work using quantal analysis, which indicated that LTD is due to a decrease in transmitter release (Stevens and Wang, 1994;Bolshakov and Siegelbaum, 1995). However, based on the finding of changes in the coefficient of variation (CV), other studies show that LTD mediated by AMPA and NMDA components represent independent mechanisms at the postsynaptic receptor level, suggesting that the observed equal relative change of the two components is just fortuitous (Selig et al., 1995).

[3] 153w In the present study we attempted to discern between a presynaptic and a postsynaptic modification during LTD by examining the waveforms of isolated NMDA and AMPA EPSPs before and after the induction of LTD and, for a comparison, before and after reduction of synaptic transmission by pharmacological means. It was conjectured that changes in the EPSP waveforms would be indicative of a postsynaptic modification (cf. Ambros-Ingerson et al., 1991;Wilson, 1984). As will be described in the following, LTD did not lead to any such changes, while modifications at the NMDA receptor level changed the waveforms of isolated NMDA EPSPs. This result seems hard to reconcile with the idea of LTD as being a postsynaptic modification (Lisman, 1994;Mulkey et al., 1993). However, subsequent experiments to directly test alterations in presynaptic transmitter release, using the irreversible open channel blocker of the NMDAreceptor MK-801 as a probe, surprisingly failed in detecting such presynaptic changes during LTD.

[1] 68w We have previously shown that the AMPA and NMDA components of a composite field EPSP are subjected to equal relative changes during LTD (Xiao et al., 1995a). It was also reported that LTD of isolated NMDA EPSPs occurred as a uniform change of the potential. We have now extended this study of LTD in isolated NMDA EPSPs and in addition studied waveforms of isolated AMPA EPSPs during LTD.

[2] 161w Isolated NMDA EPSPs were obtained in the presence of CNQX (10 µM) and LTD was induced. On the average, the field EPSP was reduced to 66.0 6 1.8 % at 15-20 min after the end of the LFS (n 5 20, including four experiments from Xiao et al., 1995a;Fig. 1B). For an optimal analysis of the EPSP waveform, the potentials were corrected by subtracting the nonsynaptic potential, obtained after fully blocking synaptic transmission (see Materials and Methods). During LTD, the waveform of the NMDA EPSP remained unchanged as illustrated by the data from one experiment (Fig. 1C). For the set of 20 experiments, there were no significant changes in the rise time and the decay time constant, the values being 4.82 6 0.16 ms, 35.5 6 1.8 ms before and 5.00 6 0.22 ms, 36.8 6 2.5 ms during LTD, respectively (P.0.05, Fig. 1D1). Using percentage changes of the waveform parameters also revealed no significant effect of LTD (P.0.05, Fig. 1D2).

[3] 151w To obtain an isolated AMPA EPSP, without blocking LTD induction, experiments were carried out in ''normal'' Mg 21 solution with no specific blocker of glutamate receptors (see Materials and Methods). When inducing LTD, the AMPA EPSP was reduced to 78.8 6 2.2% of the baseline value at 15-20 min after the LFS (n512, Fig. 2A). There was no change in the waveform of the EPSP, as illustrated by the logarithmic EPSP plot from one experiment (Fig, 2B) and by the waveform parameters calculated for a set of 12 experiments. The rise time and decay time constant for the set of experiments were 2.18 6 0.05 ms, 7.31 6 0.13 ms before LTD and 2.09 6 0.06 ms, 7.23 6 0.16 ms during LTD, respectively (P..05 in both cases, Fig. 2C1). The percentage changes of the rise times and decay time constants before and during LTD were also insignificant (P.0.05, Fig. 2C2).

[4] 106w In our previous study, the adenosine agonist CHA was shown to decrease the AMPA and NMDA components of a composite EPSP to the same extent, similar to the change during LTD (Xiao et al., 1995a). Does this manipulation of the synaptic strength, which operates via a reduction of transmitter release, also share with LTD the property of preserving the waveforms of isolated AMPA or NMDA mediated EPSPs? To investigate this matter, we compared AMPA or NMDA EPSPs before and after partial blockade of transmitter release by CHA. The potential obtained after the full effect of the drug was used as nonsynaptic potential (see Materials and Methods).

[5] 154w Figure 3A1 shows a series of NMDA EPSPs taken from one experiment, immediately before blockade and at increasing levels of blockade following application of a 200 µM droplet of CHA. When determined for a group of experiments, the rise time and decay time constant showed no significant changes for a range of EPSP sizes (P..05 for all data points, n511), indicating that there was no change in the waveform of the EPSP (Fig. 3A2). Notably, a reduction of the EPSP size to 60-70% of the baseline value (i.e., comparable to the reduction during LTD) leaved the waveform parameters virtually identical to the baseline values. Similar results were obtained for the CHA effect on isolated AMPA EPSPs (using 50 µM AP5 to eliminate the NMDA component). As illustrated in Figure 3B2, the waveform parameters of the isolated AMPA EPSP were not significantly changed for a range of EPSP sizes (P..05 for all data points, n511).

[6] 95w Similar to the above, reducing transmitter output by decreasing the number of activated synapses was also associated with a constancy of the EPSP waveform. In the experiment illustrated in Figure 4, the NMDA EPSP recorded at the initial stimulus strength was compared to EPSPs obtained at successively decreasing strengths. When comparing the changes of the rise times and decay time constants for a group of experiments (n511, Fig. 4B), there were no significant changes of those parameters for the whole range of tested EPSPs (P.0.05 in all cases; see Fig. 4 legend for experimental details).

[7] 49w As shown above, a presynaptic manipulation, such as partially blocking transmitter release by CHA, resembled LTD in that the isolated NMDA and AMPA EPSP waveforms were unchanged. As a next step we tested the effects on EPSP waveforms of drugs that attenuate synaptic transmission by interacting with postsynaptic receptors.

[8] 136w The data in Figure 5A,B, obtained immediately before and during application of the NMDA antagonist AP5 in one experiment, reveal a change of the EPSP waveform. It should be noted that the final concentration of AP5 (50 µM) lead to a total blockade of the NMDA EPSP, this recording being used as nonsynaptic potential (see Materials and Methods). The percentage changes of the rise times and decay time constants of the EPSPs before and after AP5 application for a group of experiments were plotted against the EPSP size (n511, Fig. 5C). The data show a substantial increase of the decay time constant, and thus a waveform change, being significant already at a reduction to 90% of the pre-AP5 level (P,.01 in all cases). The rise times were also changed but to a lesser extent (Fig. 5C).

[9] 65w In contrast to the above results on isolated NMDA EPSPs, partial blockade of isolated AMPA EPSPs by CNQX did not alter the waveform of the latter EPSPs. This is illustrated in Figure 6 (n58; P..05 for all data points). The final concentration of CNQX (10 µM) was sufficient to completely block the AMPA EPSP, this recording being used as nonsynaptic potential (see Materials and Methods).

[10] 197w The observed waveform change for isolated NMDA EPSPs in the presence of AP5 might be specific to the use of this drug, acting as an competitive antagonist at NMDA receptors, or it could reflect a general characteristic of synaptic attenuation by a postsynaptic mechanism. To dissociate between these possibilities we employed 5,7-dichlorokynurenic acid (DCKA; 10 µM), an antagonist acting at the glycine site of the NMDA receptor, and MK-801 (5-10 µM), an irreversible open channel blocker, to attenuate NMDA mediated synaptic transmission. These manipulations resulted in changes of the waveforms of the isolated NMDA EPSPs comparable to those seen with AP5 (Fig. 7, left and middle diagrams). A similar result was obtained by increasing the Mg 21 concentration of the perfusion solution to 0.2 mM, leading to an increase of the voltage dependent blockade of the NMDA receptor channels (Fig. 7, right). In all three cases the decay time constants increased significantly in terms of percentage change (P,.05, n58 for all groups) when the EPSPs were reduced to 2/3 of the baseline size, which is comparable to the degree of reduction during LTD. The nonsynaptic potential (see Materials and Methods) was obtained in the presence of CHA.

[11] 127w In addition, we carried out experiments with composite field EPSPs in which AP5 or CNQX was applied at increasing concentrations. By subtracting the composite EPSP (baseline potential or partially blocked one) with the remaining potential during full blockade by AP5 (50 µM) or CNQX (10 µM), the isolated NMDA EPSP or isolated AMPA EPSP could be calculated, respectively. Again, AP5 altered the waveform of the (calculated) NMDA EPSP as shown from one experiment (Fig. 8A); evidenced by a significant change of the decay time constant for a set of experiments (P,0.01, n58; Fig. 8A3, right). On the contrary, CNQX had no effect on the waveform of the calculated AMPA EPSP (Fig. 8B, P.0.05, n513), in agreement with the results on CNQX application using pharmacologically isolated AMPA EPSPs.

[12] 100w As shown above, the waveforms of isolated NMDA EPSPs were altered by the postsynaptic manipulations used whereas the presynaptic ones, and LTD, did not induce such changes. To test more directly the inferred presynaptic locus of LTD, the irreversible open channel blocker of the NMDA receptor MK-801 was used. In the presence of this drug, the size of the NMDA mediated EPSP declines at a rate which depends on the amount of transmitter released per synapse and time unit; the decline rate can thus be used to monitor changes in release probability (cf. Hessler et al., 1993;Manabe and Nicoll, 1994).

[13] 180w LTD was induced for isolated NMDA EPSPs in one pathway by 2-3 periods of LFS (average reduction to 47.0 6 3.3%, n58) and the stimulus strength of the control pathway was then lowered to produce an EPSP of the same size as the depressed one. Following this procedure, the stimulation was turned off, 50 µM MK-801 was applied and stimulation was resumed 10 min later (Fig. 9; cf. Manabe and Nicoll, 1994). After stimulation was resumed, the size of the EPSP decreased for both pathways in a parallel manner, show- ing no clear difference in the rate of decay (Fig. 9A,B1). To quantify the relation between the decaying EPSPs the relation between them (cf. Fig. 9B1) was fitted by a second degree polynomial and the coefficient for the second degree term (C 2 ) was used as a curvature measure (Manabe and Nicoll, 1994; for details see Fig. 9 legend). For a set of eight experiments, C 2 was not significantly different from zero (0.099 6 0.077, P.0.05), suggesting that LTD is not due to a change in release probability.

[14] 149w In view of this result, which appears to conflict with our previous waveform data, additional experiments were undertaken to test the validity of the MK-801 method. In these experiments the stimulus frequency of the test pathway was reduced to half (from 0.1 to 0.05 Hz). Since the rate of MK-801 blockade depends on the transmitter output per time unit, a reduction of the stimulus frequency is equivalent to a reduction of the release probability by the same relative amount. In the experiment in Figure 9C, the isolated NMDA EPSPs progressively decreased in the presence of MK-801, the decay rate appearing to be somewhat smaller for the pathway stimulated at a low frequency. For a group of eight experiments, the C2 value was significantly greater than zero (0.386 6 0.046, P,.001) and was also significantly different from the C2 value obtained above for a test pathway subjected to LTD (P,.05).

[15] 105w Theoretically, a 1:2 ratio in transmitter output between the ''test'' and ''control'' pathways, either due to different frequencies or due to differences in the release probabilities between the two pathways, will result in a relative size of the control EPSP that equals the square of the relative size of the test EPSP. This means a C 2 value of one in this case. The actual C 2 value of around 0.4 implies that the method is less sensitive than expected. Nevertheless, if LTD were presynaptic, it should have been possible to detect this; note that the EPSP was reduced to less than half during LTD.

[1] 128w The results show that LTD occurred without changes in the waveform of isolated NMDA and AMPA EPSPs. Reducing the number of release sites by lowering the stimulus strength or reducing the probability of transmitter release by an adenosine agonist both mimicked the LTD-induced changes. Partial AMPA receptor blockade was also without effect on the waveform of isolated AMPA EPSPs. In contrast, partially blocking NMDA receptors in a number of ways resulted in waveform changes. These manipulations thus appear to differ mechanistically from LTD, suggesting that LTD is due to a presynaptic rather than a postsynaptic modification. Nonetheless, testing for alterations in transmitter release by the open channel blocker MK-801 failed to reveal any such changes. These seemingly contradictory results call for a reevaluation of possible mechanisms for LTD.

[1] 247w The above results may seem to contradict each other, a fact that mainly stems from the presumption that LTD be either pre-or post-synaptic. If, on the other hand, LTD is due to both a presynaptic and a postsynaptic change, especially if these changes are coordinated, the above experimental approaches will not be able to detect presynaptic or postsynaptic changes. Imagine, for instance, that LTD lead to a decrease in the area of synaptic contact without changing the density of membrane-associated factors related to synaptic transmission. The transmitter release would then be smaller for each synapse but constant when calculated per postsynaptic ion channel. No change in the blocking rate would be seen during MK-801 application. Similarly, one would expect the waveform of the NMDA EPSP to be preserved due to the constant conditions for each channel. This reasoning assumes that different regions within the synaptic contact area operate independently. Such independent behavior of subsynaptic regions would be facilitated if synapses change by addition/removal of spinules (Edwards, 1995). Synaptic changes of this kind can also account for the lack of changes in paired pulse facilitation during LTP/LTD (McNaughton, 1982;Mulkey and Malenka, 1992;Muller and Lynch, 1989;Xiao et al., 1995a;but see Schultz et al., 1994). It can finally be noted that a decoupling, or elimination, of a fraction of the population of synapses during LTD is also fully consistent with the present data. It is hard to see, however, how synapses decoupled during LTD could be reengaged by subsequent LTP.

[2] 125w In the above we have made no difference between LTD of AMPA and NMDA receptor mediated EPSPs. This was based on our previous results on a close linkage between the relative changes of the two EPSPs during LTD (Xiao et al., 1995a,b), implying a common LTD mechanism. However, the published data on LTD diverge with respect to the issue of a functional AMPA-NMDA linkage (cf. Selig et al., 1995;Xiao et al., 1995a), similar to the situation with LTP (cf. Aniksztejn and Ben-Ari, 1995;Clark and Collingridge, 1995). If the premise of a common LTD mechanism is dropped, the idea that LTD is associated with a coordinated presynaptic-postsynaptic change remains as an explanation of our NMDA EPSP data but leaves open the possibility of additional AMPA-specific changes.

[1] 215w Experiments were performed on 12-18-day-old Sprague-Dawley rats (n569). Hippocampal slices were prepared according to standard procedures (cf. Xiao et al., 1995a). The 400 µm thick slices were kept in a constant flow recording chamber (Brain Slice Chamber, Medical Systems Corp., Greenvale, NY) where they were maintained at 29-30°C in the interface between an oxygenated (95% O 2 , 5% CO 2 ) salt solution and humidified gas (95% O 2 , 5% CO 2 ). The solution contained (in mM) NaCl 119, KCl 2.5, CaCl 2 2, MgCl 2 0.1, NaHCO 3 26, NaH 2 PO 4 1, glucose 10, and 10 µM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) for experiments on isolated NMDA receptor-mediated field EPSPs (''isolated NMDA EPSPs'') or 1 µM CNQX for experiments on composite AMPA/NMDA receptor-mediated field EPSPs (''composite EPSPs''). For the experiments on isolated AMPA receptor-mediated field EPSPs (''isolated AMPA EPSPs'') the solution contained (in mM) NaCl 119, KCl 2.5, CaCl 2 4, MgCl 2 4, NaHCO 3 26, NaH 2 PO 4 1, glucose 10, and 100 µM picrotoxin. Experiments on isolated AMPA EPSPs were also carried out in the solution for composite EPSPs but using 50 µM D-2-amino-5-phosphonopentanoic acid (AP5) to block the NMDA response. A surgical cut was made between CA3 and CA1 regions to prevent the propagation of epileptiform activity.

[2] 182w Stimulation was provided by electrolytically sharpened tungsten wires to which 0.1 ms negative constant current pulses (12-45 µA) were applied. Two stimulating electrodes were placed in the apical dendritic layer of CA1 pyramidal cells, on either side of the recording electrode, to provide stimulation of two separate sets of afferents that activate independent synaptic populations (test pathway and control pathway). The stimulating electrodes were activated alternately at a test stimulus frequency of 0.1 Hz with stimuli separated by 5 s except for one type of experiment in which a test stimulus frequency of 0.05 Hz was used. Extracellular responses were recorded in the apical dendritic layer of CA1 pyramidal cells, using a glass micropipette (resistance 1.5-4.5 MV) filled with 3 M NaCl. The stimulus strength was adjusted to yield a field EPSP of about one third of the size at which distortions of the field EPSP curve were seen. LTD was induced by applying 2 Hz stimulation for 10 min to one of the pathways at test stimulus strength (low frequency stimulation, LFS). The other pathway received no stimulation during this period.

[3] 106w In order to obtain EPSPs at various levels during the drug application, e.g., for comparison with EPSPs during LTD, drug was applied at increasing concentrations in the perfusion solution or as a droplet spreading slowly to the recording region (usually the case for N 6 -cyclohexyl-adenosine, CHA). For some drugs, the blockade occurred slowly enough to allow recording of a series of partially blocked EPSPs with drug application just once. The records were usually corrected by subtracting the nonsynaptic potential (representing stimulus artifact and presynaptic volley), obtained by blocking synaptic transmission by CHA (Lupica et al., 1992, Xiao et al., 1995a) or both CNQX and AP5.

[4] 53w Signals were amplified and filtered at 3 kHz on an Axoclamp-2A (Axon Instruments, Burlingame, CA), and transferred via a CAMAC interface to a Nord 10/S microcomputer, providing an on-line display and analysis of the field EPSPs. Data were then transferred to a PC-clone personal computer for a more detailed off-line analysis and plotting.

[5] 133w The size of the EPSP was quantified by an area measurement, equal to the absolute value of the integral of the recorded potential from 3-4 ms (roughly representing the onset of the EPSP) to 100 ms after the stimulus. The waveform of the EPSP was quantified by the rise time and the decay time constant (for simplicity the latter is referred to as ''decay time'' in the figures). Rise time was determined as the time required for the EPSP to advance from 20 to 80% of its peak amplitude (see Fig. 1A). Decay time constant was obtained via a linear regression fit to the logarithm of the decay of the EPSP curve, using the decaying portion of the curve falling within 20-80% of the peak amplitude. Statistical comparisons were made using Student's t-test.

[6] 40w D(-)-2-amino-5-phosphonopentanoic acid (AP5), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), and 5,7dichlorokynurenic acid (DCKA) were obtained from Tocris Cookson, Ltd., UK; N6-cyclohexyl-adenosine (CHA) and picrotoxin (PTX) were from Sigma Chemical Company (St. Louis, MO); and (1)-MK-801 hydrogen maleate (MK-801) from Research Biochemicals International (Natick, MA).

[1] 243w Several previous studies report that LTP is essentially devoid of changes in the waveform of the AMPA EPSP. In cases where waveform changes were observed they could often be accounted for by secondary effects or by effects unrelated to LTP (Asztely and Gustafsson, 1994;Hanse et al., 1991;Hess and Gustafsson, 1990; but see Ambros-Ingerson et al., 1991). We have now extended this analysis of EPSP waveform to LTD, showing that the isolated AMPA EPSP is changed in a uniform (i.e., waveform-preserving) manner. A constant waveform of the AMPA EPSP during LTP/LTD is hard to reconcile with certain types of postsynaptic modification. For instance, changes in spine neck resistance will (theoretically) influence fast and slow synaptic currents differently (Lynch et al., 1990;Wilson, 1984) and will thus alter the waveforms of EPSPs, both composite and isolated ones. Certain changes in AMPA channel kinetics affect the waveform of isolated AMPA EPSPs (Ambros-Ingerson et al., 1991, Asztely et al., 1992). However, a postsynaptic scenario is by no means ruled out as evidenced by our finding that partial blockade of AMPA channels by CNQX, in similarity with LTD, preserved the waveform of the isolated AMPA EPSP. Notably, this waveform was also unchanged when synaptic efficacy was attenuated by lowering transmitter release through application of the adenosine agonist CHA. Thus our results on AMPA EPSPs do not discriminate between a presynaptic and a postsynaptic modification but can be used to exclude certain postsynaptic changes (and possibly also some presynaptic ones).

[2] 296w LTD was also found to be associated with a uniform change of the isolated NMDA EPSP (for similar results on LTP, see Xiao et al., 1995b) and, as in the case of the isolated AMPA EPSP, this change could be mimicked by reducing transmitter release through application of CHA. In contrast, a number of manipulations at the postsynaptic receptor level all lead to changes in the waveform of the NMDA EPSP. The drugs used included AP5, an antagonist at the glutamate binding site; DCKA, an antagonist at the glycine site; and MK-801 Fig. 7. Several ways of partially blocking NMDA receptors all result in changes of the waveforms of isolated NMDA EPSPs. Left to right: DCKA(10 µM), MK-801 (5-10 µM) and increasing Mg 21 from 0.1 mM to 0.2 mM. The decreased responses were taken at a time when the size of the EPSP was reduced to 2/3 of the baseline value, which degree of reduction is comparable to LTD. All records (averages of ten sweeps) have been corrected by subtracting the nonsynaptic potential (using CHA). The average rise times and decay time constants before and after applying drug or changing Mg 21 for a set of experiments are shown (using EPSPs reduced to 2/3 of the baseline ones as illustrated by the sample records; n58 for all three groups). and magnesium, both of which bind to locations within the channel pore. We did not test drugs acting at the polyamine site, since these drugs have been shown in previous studies to affect NMDA channel kinetics in a manner expected to influence the waveform of the EPSP (cf. Lerma, 1992). So far, we have also not studied the effect of redox agents but some data published by others show signs of waveform changes (cf. Tauck, 1992).

[3] 71w In the experiments using MK-801, the drug was applied at a relatively low concentration to avoid waveform changes directly related to the activity-dependent block, which tends to fasten the decay of the synaptic response (Hessler et al., 1993;Manabe and Nicoll, 1994). There are still other complexities with this drug. For instance, synapses with a high basal release probability would be expected to be blocked before synapses with a lower release probability.

[4] 57w It should be noted that the observed waveform changes are unlikely to be secondary to the change of EPSP size or to other unspecific effects, due to the simple fact that several of the manipulations used (LTD, CHA, and stimulus strength reduction) lead to a decrease of the size of the NMDA EPSP while preserving the waveform.

[5] 124w The results were also supported by data from experiments with composite EPSPs. Using receptor antagonists to partially block the AMPA or NMDA responses of a composite EPSP demonstrated changes of the waveform of the NMDA response (calculated by subtraction) during application of AP5 but no change of the waveform of the AMPA response (calculated by subtraction) during application of CNQX. In addition, previous studies showed that the entire time course of composite field EPSPs changed in a uniform manner, during both LTD and application of CHA (Xiao et al., 1995a). This result is most easily accounted for by a constancy of the waveforms of both EPSP components (during LTD and CHA application), in combination with attenuation of the two components by a common factor.

[6] 113w The similar effect on the waveform of the NMDA EPSP by substantially different NMDA receptor blocking mechanisms suggests that a common external factor might be involved. A possible one is the Ca 21 concentration in the postsynaptic spine, which will be inversely related to the degree of NMDA receptor blockade during activation of the receptor. A tentative explanation is then that some aspect of channel kinetics, such as the rate of desensitization, is influenced by the intracellular Ca 21 (Rosenmund et al., 1995), and thus will be affected by the various kinds of NMDA receptor blockade. This scenario predicts a slower decay for smaller EPSPs, which is the same as the observed change.

[7] 209w Another possible scenario takes into account the possible coexistence of different heteromeric NMDA receptors (channels) in the hippocampus, consisting of NR1 subunits in combination with a differential incorporation of subunits NR2A-2B (Ishii et al., 1993;Monyer et al., 1992Monyer et al., , 1994) ) and perhaps also NR2C (Pollard et al., 1993). These molecularly distinct channels differ with respect to some key functional properties. For instance, channels containing NR2A subunits are associated with faster decaying currents than those containing NR2B-2D subunits (Monyer et al., 1992(Monyer et al., , 1994)). The NR2A-containing channels are also more sensitive to antagonists (AP5 and 7-chlorokynurenic acid) and Mg 21 block (Ishii et al., 1993;Monyer et al., 1992Monyer et al., , 1994; see also Kutsuwada et al., 1992 for data on MK-801) as well as redox agents (Ko ¨hr et al., 1994). As a consequence, a partial blockade of the NMDA response is expected to preferentially eliminate the fast-decaying currents, leading to an apparent increase in the decay time constant of the population current. Accordingly, the absence of waveform changes during LTD implies that NMDA receptors with different subunit compositions contribute equally to LTD. This is an extension of the previous result that AMPA and NMDA receptors contribute equally to LTD (Xiao et al., 1995a).

[8] 294w Our main result relevant for a pre-vs. post-synaptic location of LTD is the finding of waveform changes when interfering with NMDA receptors in a number of ways, while such changes were missing during LTD. This result argues against LTD as being a postsynaptic receptor modification. As pointed out above, a preservation of waveform is also difficult to reconcile with some other postsynaptic changes. The alternative interpretation, that is that LTD is due to a presynaptic modification, is consistent with our findings that the waveform of the NMDA EPSP was constant during a reduction of B2: Similar to B1, but showing the average of binned data from a set of experiments (n58). Computational details: Let x and y denote test and control measurements after stimulation was resumed. These values were first normalized with respect to their average during the first 30 s and their mean (x1y)/2 was then used to assign bins. Binned data were fitted by a curve y 5 C 0 1 C 1 x 1 C 2 x 2 . The curve shown in B2 is based on the average coefficients C 0 , C 1 , and C 2 . The value of C 2 , which represents curvature, is indicated. C: Similar to A, but instead of LTD, the stimulus frequency of one pathway was kept at 0.05 Hz (instead of 0.1 Hz), leading to a reduction of the amount of transmitter released per time unit. For the test used this is equivalent to a reduction of the release probability to half. D1: Similar to B1, but showing the somewhat curved relation between the EPSPs of the low and normal frequency pathways. D2: Similar to B2 but for a set of experiments of the same type as in C (n58).

[9] 22w the number of release sites (lowering the stimulus strength) as well as during a decrease of the release probability (application of CHA).

[10] 139w Using the irreversible open channel blocker MK-801 to directly probe for presynaptic changes, we observed a closely parallel attenuation of EPSPs in pathways subjected to LTD and in naive ones. This result was evidenced by a nearly linear relation between the magnitudes of the EPSPs of the two pathways, in contrast to the expected curved relation if LTD were associated with a decrease in transmitter release. To assess the sensitivity of this method in detecting differences in transmitter release, the rate of blockade (per time unit) was compared for pathways stimulated at different frequencies. For unknown reasons, the sensitivity was found to be lower than expected theoretically. Until a better understanding of this matter is available some caution is necessary in applying the present results. Anyhow, using them as they stand, clearly argues against a presynaptic location for LTD.