Proc. Nati. Acad. Sci. USA Vol. 85, pp. 9346-9350, December 1988 Neurobiology Long-term potentiation differentially affects two components of synaptic responses in hippocampus (plasticity/N-methyl-D-aspartate/D-2-amino-5-phosphonovglerate/facilitation) DOMINIQUE MULLER*t AND GARY LYNCH Center for the Neurobiology of Learning and Memory, University of California, Irvine, CA 92717 Communicated by Leon N Cooper, September 6, 1988 (receivedfor review June 20, 1988)

ABSTRACT We have used low magnesium concentrations ing electrode was positioned in field CAlb between two and the specific antagonist D-2-amino-5-phosphonopentanoate stimulating electrodes placed in fields CAla and CAlc; this (D-AP5) to estimate the effects of long-term potentiation (LTP) allowed us to activate separate inputs to a common pool of on the N-methyl-D-aspartate (NMDA) and non-NMDA recep- target cells. Stimulation voltages were adjusted to produce tor-mediated components of postsynaptic responses. LTP in- field EPSPs of -1.5 mV and did not elicit population spikes duction resulted in a considerably larger potentiation of non- in any of the responses included for data analysis. NMDA as opposed to NMDA receptor-related currents. In- Paired-pulse facilitation was produced by applying two creasing the size of postsynaptic potentials with greater stimulation pulses separated by 30 or 50 ms to the same stimulation currents or with paired-pulse facilitation produced stimulating electrode and LTP was induced by patterned opposite effects; i.e., those aspects ofthe response dependent on burst stimulation-i.e., 10 bursts delivered at 5 Hz, each NMDA receptor's increased to a greater degree than did those burst being composed of four pulses at 100 Hz (see ref. 5). components mediated by non-NMDA receptors. These results Suppression of Inhibitory Potentials with "Priming" Stim- pose new constraints on hypotheses about the locus and nature ulation. The NMDA receptor ionophore is blocked in a ofLTP and strongly suggest that postsynaptic modifications ate voltage-dependent fashion by magnesium ions (14, 15). As a part of the effect. consequence, antagonists of the receptor [e.g., D-2-amino-5- phosphonopentanoate (D-AP5)] have little effect on field Long-term potentiation (LTP), a long-lasting increase in EPSPs elicited by single-stimulation pulses in slices main- synaptic efficacy observed in hippocampus (1) and elsewhere tained in normal medium. Two procedures were used in the in forebrain (2), has attracted interest as a possible substrate present experiments to reduce the blockade of the receptor of behavioral memory (3). While considerable progress has channel and allow response components mediated by NMDA been made in identifying the cellular events that trigger LTP receptors to appear. First, the concentration ofmagnesium in (4-7), the final and stable modifications that underlie the the medium was reduced from 1 mM to 10-20 uM. As increased synaptic potency remain to be resolved. In an previously shown (10), however, this manipulation by itself effort to restrict the list of proposed mechanisms, we have does not reliably result in the development of a large tested the possibility that LTP has selective effects on postsynaptic response component sensitive to D-AP5. Pre- different components of the postsynaptic response. Recent sumably, because of the small amount of magnesium still work has identified conditions under which a sizable portion present in the medium, the degree and duration of the of the field excitatory postsynaptic potential (EPSP) elicited depolarization produced by a single field EPSP are not by afferent stimulation is blocked by antagonists of the N- sufficient to counteract the voltage-dependent blockade of methyl-D-aspartate (NMDA) receptor (8-10). Here we report the receptor channel. However, when the feedforward in- that the NMDA and non-NMDA components of synaptic hibitory postsynaptic potentials (IPSPs) that normally ac- responses in hippocampus are differently affected by LTP, company and truncate synaptic responses in hippocampus the pattern of results being opposite that observed with (16) are removed, then a sizable portion of the field EPSP is paired-pulse facilitation, an effect attributed to increased blocked by NMDA receptor antagonists (8, 10). Accordingly, transmitter release (11-13). we used a technique referred to as priming (see ref. 5 and Fig. la for an illustration) to suppress feedforward IPSPs. Feed- forward IPSPs exhibit a short (-0.5 s) refractory period once MATERIAL AND METHODS having been activated (5, 6, 16), an effect that is readily Hippocampal slices (450 ,m thick) were prepared and main- apparent in experiments using two separate inputs (Schaffer- tained as described elsewhere in a surface recording chamber commissural fibers) to a common pool of intermeurons and and continuously perfused with a medium containing in mM: pyramidal cells. If one collection of afferents (the "priming NaCl, 124; KCl, 3; KH2PO4, 1.25; CaC12, 3; MgCI2, 1; input") is used to trigger the excitatory as well as inhibitory NaHCO3, 26; glucose, 10; L-ascorbate, 2 (pH 7.4). The slices potentials in the target region, then the responses to the were kept at 350C and oxygenated with 95% 02/5% Co2. second (or "test") input activated 200 ms later are largely After an hour of recovery, the perfusion medium was free of IPSPs. The intracellularly recorded afterhyperpolari- switched to a solution containing only 10 or 20 tLM magne- zation produced by feedforward IPSPs is suppressed (Fig. sium, which was perfused for another hour before the lb), while the repolarization phase of the field EPSP mea- experiments were started. sured with extracellular electrodes is prolonged (Fig. lc). As Responses were evoked by stimulation of the Schaffer- commissural projections in the stratum radiatum offield CA1 Abbreviations: LTP, long-term potentiation; NMDA, N-methyl- and recorded with glass micropipettes (1-5 Mfl). The record- D-aspartate; EPSP, excitatory postsynaptic potential; IPSP, inhibi- tory postsynaptic potential; D-AP5, D-2-amino-5-phosphonopen- tanoate. The publication costs of this article were defrayed in part by page charge *Present address: Department of Pharmacology, Geneva School of payment. This article must therefore be hereby marked "advertisement" Medicine, CMU, 1206 Geneva, Switzerland. in accordance with 18 U.S.C. §1734 solely to indicate this fact. tTo whom reprint requests should be addressed.

9346 Downloaded by guest on September 28, 2021 Neurobiology: Muller and Lynch Proc. Natl. Acad. Sci. USA 85 (1988) 9347

shown in Fig. ic, priming has a comparable effect on control a Priming paradigm responses as it does on field potentials that had been increased either by induction of LTP or with paired-pulse facilitation. The results show that after suppression ofthe fast IPSPs, the primed responses evoked in all three conditions are characterized by a similar time course, since they can be superimposed after normalization of their amplitude. 2 mU Effect of D-AP5 on Responses Evoked with Paired-Pulse Stimulation, Increased Stimulation Intensity, or Following 40 ms LTP Induction. Paired-pulse facilitation was produced by applying two pulses separated by 30 or 50 ms to a group of afferents terminating in a dendritic field that had been primed b Intracellular EPSPs by stimulation of a second input 200 ms earlier (Fig. 1). The area of the postsynaptic responses evoked by the first and second pulses to the test input was measured before and after application of D-AP5 (50-125 ,uM) to the medium. In some cases, the drug was washed out of the slices and the same experiment was repeated. Comparisons were thus made of the degree of paired-pulse facilitation found in the presence 5 mU and absence of the NMDA receptor antagonists. To assess the effect of D-AP5 on responses of different sizes, primed EPSPs were evoked by applying alternatively 20 ms stimulation pulses of different intensities to the same stimu- lation electrode. The area of both responses was then c Extracellular EPSPs measured before and after application of the drug. Two types ofexperiments were conducted to determine the Control Potentiation Facilitation effect of D-AP5 on potentiated responses. In nine cases, we tested for the effect of the drug on control responses, and then, following a washout period (40-60 min), LTP was induced and D-AP5 was reintroduced at the same concen- tration. In some cases, the control responses were evoked by using a paired-pulse paradigm, thus allowing for a direct 1 mu comparison of the effect of D-AP5 on facilitated and poten- 10 ms tiated responses (see Fig. 3). In five other experiments, we used two independent, equal-sized test inputs to the same Superimposed dendritic field in addition to the priming input (i.e., three stimulation electrodes activating converging afferents were used). LTP was then induced on one ofthe test inputs and the effect ofa single application ofD-AP5 (50 ,4M) was measured on potentiated and control responses. In some cases, control responses were also evoked using a paired-pulse paradigm, FIG. 1. Priming paradigm. (a) Illustration ofthe paradigm used to allowing for a direct comparison of the effect of the receptor suppress feedforward inhibitory responses. The records show the antagonist on control, facilitated, and potentiated responses. field potentials collected by one recording electrode and elicited on The two types of LTP experiments (i.e., single vs. sequential priming and test inputs by two different stimulating electrodes. The application of D-AP5) generated similar results. responses to a pair of priming pulses are shown at the beginning of the trace, while those elicited by paired-pulse (50-ms interpulse interval) stimulation ofthe test input 200 ms later are illustrated at the RESULTS end ofthe record. Recordings with (arrowhead) and without priming are superimposed. (b) Effect ofthe priming paradigm on intracellular As anticipated, the combination of priming and low magne- EPSPs. The afterhyperpolarization that reflects synaptically medi- sium medium resulted in field EPSPs that contained a ated IPSPs and that normally shortens the dendritic EPSP to significant component that was blocked by D-AP5 (Fig. 2a); single-stimulation pulses is eliminated in the primed condition. this effect was considerably smaller when using primed Primed (arrowhead) and nonprimed potentials are superimposed to responses in 1 mM magnesium. In 19 experiments conducted illustrate the difference in their decay phase. (c) Effect ofthe priming in the presence of 10-20 ,uM magnesium, we found that 50 or paradigm on extracellular field potentials. The records show super- 125 AtM D-AP5 reduced the area of the primed field potential imposed primed and nonprimed control, potentiated, and facilitated 28.0o ± 0.9% (mean ± SEM), whereas the reduction in responses. In the three conditions, priming (and thus suppression of by IPSPs) results in a significant prolongation of the repolarization area was only 7.8% ± 1.5% in six experiments carried out in phase, and all three primed responses (arrowheads) are then char- 1 mM magnesium. Fig. 2b shows that D-AP5 reduced in a acterized by a similar time course when superimposed at normalized similar way the size of the dendritic responses recorded scales. All responses were recorded in the presence of 1 mM intracellularly by the same paradigm. The effects of D-AP5 magnesium; records are mean of three or four individual traces. observed in a typical experiment on control, facilitated, and potentiated responses evoked on the same pathway are presence and absence of D-AP5. Subtraction of the two illustrated in Fig. 2c. potentials (Fig. 3b) yields the NMDA receptor-mediated To quantify the effects of LTP on the D-AP5 sensitive and component of the EPSP, whereas the response recorded in insensitive components of synaptic responses, we subtracted the presence of D-AP5 reflects current through non-NMDA the area ofthe averaged field EPSPs recorded in the presence receptors (Fig. 3c). Note that, as expected, the D-AP5 of the drug from that obtained in its absence. This was done sensitive component has a slower time course than the D-AP5 for control, potentiated, and facilitated responses in each insensitive response. The next sets of traces are, respec- slice, and the results are compared. Fig. 3 illustrates the tively, potentiated (middle column) and paired-pulse facili- method. Fig. 3a (Left) shows a control response in the tated (right column) responses from the same slice in the Downloaded by guest on September 28, 2021 9348 Neurobiology: Muller and Lynch Proc. Natl. Acad. Sci. USA 85 (1988)

a b

1 mU 5 mU

10 ms 20 Ms

C

o<

o-

30 Time, min FIG. 2. Effect of D-AP5 on primed synaptic responses. (a) Superimposed field potentials recorded in the absence and presence of 50 AM D-AP5. The drug reduces the size ofthe EPSP by affecting primarily the amplitude and the rate ofdecay ofthe response. (b) Primed intracellular EPSPs recorded in the absence and presence of 50 ,M D-AP5. The time course of responses is slower in intracellular recording but the effect ofthe drug is very similar. (c) Changes in area produced by 125 ,uM D-AP5 (a 10-min application beginning at the 10-min time point) on the primed responses evoked by stimulation of the same afferents in control conditions (o), using a paired-pulse paradigm at 30-ms intervals (facilitated response: *), and following LTP induction (e). Results are expressed as percent of the response area measured before D-AP5 application. The onset of the drug action was rapid and it affected a larger fraction of facilitated than potentiated responses. Each point represents the mean area of two field EPSPs.

presence and absence of the drug. The results of the sub- balance of NMDA and non-NMDA components of field traction are shown underneath, with the size of the compo- EPSPs. When two stimulation pulses are given in rapid nent observed in control responses superimposed (dotted succession (<200 ms apart), the response to the second pulse trace). As is evident, LTP had a much greater effect on the is considerably larger than that to the first. This effect has non-NMDA aspect of the postsynaptic response than it did been analyzed in detail at peripheral synapses and shown to on the NMDA-mediated component, a result that contrasts reflect an increase in transmitter release (11, 12, 13). Paired- with that observed following facilitation (see below). pulse facilitation in hippocampus is characterized by a time Fig. 4 summarizes the data for all 14 LTP experiments. For course and a sensitivity to extracellular calcium levels that each slice, we measured the drug-sensitive and drug- are very similar to those described at neuromuscular junc- insensitive portions offield EPSP for control and potentiated tions, suggesting strongly that it reflects the operation of the responses. The figure expresses the increase in response size same presynaptic variables. induced by high-frequency stimulation on, respectively, the The area of the second response in the paired-pulse NMDA and non-NMDA receptor-mediated components of experiments was found to be 65% and 57% larger than that the synaptic responses. For all 14 experiments, LTP had a recorded after the first pulse at, respectively, 30- and 50-ms much greater effect on the D-AP5 insensitive aspects of the interpulse intervals. Figs. 2c and 3 illustrate the results of a response than it did on the NMDA component (63% + 3% vs. typical experiment in which the effect ofD-AP5 was tested on 19% ± 6%, respectively; P < 0.0001). control, facilitated, and potentiated responses evoked by The more than 3-fold difference in the effects of LTP on the stimulation of the same afferents. As shown in Fig. 3 (Right), NMDA vs. non-NMDA components of the postsynaptic subtracting the responses collected in the presence of the responses was not reproduced by simply increasing the size drug from those obtained in its absence (Fig. 3a) again yields of the field EPSP with greater stimulation voltages. In five NMDA (Fig. 3b) and non-NMDA (Fig. 3c) receptor-mediated experiments, we tested the effects of D-AP5 on responses aspects of the field EPSP. For comparison, the components that were 161% ± 5% of control values (i.e., slightly greater observed in the first responses evoked in the paired-pulse than potentiated responses). In each of these cases the paradigm (Left) are shown superimposed (dotted traces). In NMDA component of the field potential increased at least as marked contrast to the effect observed with LTP, the D-AP5 much as and on the average more than the non-NMDA sensitive component increased to a greater degree than did component (58% ± 4% for the non-NMDA and 75% ± 6% for the D-AP5 insensitive response. In 18 of 19 experiments in the NMDA component; P < 0.05; see Fig. 4). which we used a paired-pulse paradigm at 30- or 50-ms A similar pattern of results was also obtained when intervals, we found that the increase in the D-AP5 sensitive analyzing the effect ofpaired-pulse facilitation on the relative component was larger than that of the D-AP5 insensitive Downloaded by guest on September 28, 2021 Neurobiology: Muller and Lynch Proc. Natl. Acad. Sci. USA 85 (1988) 9349 CONTROL POTENTIATION FACILITATION

a) EPSP before and after D-AP5

b) D-AP5 sensitive ...... component .1-.

c) D-AP5 insensitive component

FIG. 3. Typical experiment illustrating the effect of LTP and paired-pulse facilitation on the D-AP5 sensitive and D-AP5 insensitive components of synaptic responses. (a) Representative field potentials recorded in each condition before and after application of 125 ,AtM D-AP5. All responses were evoked by stimulation of the same afferents. (b) Subtraction of the responses recorded after D-AP5 application from those obtained before yields the D-AP5 sensitive component ofthe synaptic response. The component observed in control conditions is superimposed as a dotted trace on the results ofthe subtraction obtained following LTP induction and facilitation. (c) EPSPs recorded after D-AP5 application. Again the control response has been superimposed as a dotted trace on the potentiated and facilitated potentials. Each record represents the mean of four individual EPSPs; scales are 10 ms and 1 mV in a and c and 0.5 mV in b.

component (average increase at, respectively, 30 or 50 ms evidence that this paradigm differentially affected control, interpulse intervals: 96% ± 13% and 86% ± 9% for the facilitated, or potentiated responses. NMDA and 55% ± 3% and 47% ± 3% for the non-NMDA We assume that the size of the D-AP5 sensitive aspect of receptor-mediated aspects of the EPSP; see Fig. 4). the synaptic responses is controlled by two major factors: (i) the amount of transmitter released and hence the number of NMDA receptors activated and (ii) the depolarization of the DISCUSSION dendritic spines, which results from the effects ofthe released Much evidence indicates that the postsynaptic NMDA re- transmitter on the non-NMDA receptors. Depolarization can ceptors play a major role in the induction of LTP in area CA1 be expected to affect currents through NMDA receptors of hippocampal slices (6, 7). In the present experiments, since a substantial amount of magnesium was left in the however, we analyzed the contribution of these receptors to medium (14). In the case ofthe paired-pulse experiments, one the responses recorded before and after LTP induction to would expect these two factors to be operative, if, as shown obtain information about the processes that are responsible in peripheral systems, the facilitation effect results from an for the maintenance ofthe potentiation effect. Measurements increase in transmitter release. The fact that we observed in of NMDA receptor-mediated components of synaptic re- 18 of 19 experiments a larger increase in D-AP5 sensitive than sponses were obtained after suppression of IPSPs using a in D-AP5 insensitive aspects of field potentials supports the priming technique. As illustrated in Fig. 1, we did not obtain interpretation that both increased release and a greater depolarization of the spines were involved. It cannot be more may also 125 - excluded, however, that other subtle effects * NMDA have (e.g., a thle non-NMDA participated delayed repolarization of spines or a residual activation of NMDA receptor channels). It 100 - should be noted, however, that the repolarization of synaptic 0 responses as well as of the D-AP5 sensitive component were 75. 0 both completed by 50 ms (see Fig. 2b), making it unlikely that delayed influence of the first response would have contrib- 50 uted significantly to the larger D-AP5 sensitive component observed on the second response. 25 The major result ofthe present experiments is that LTP has markedly different effects on components ofthe postsynaptic responses mediated by different transmitter receptors. This finding places new constraints on hypotheses concerning the LTP Facilitation Facilitation Increased locus and nature of the LTP effect and in particular suggests (30 ms) (50 ms) stimulation that it is not due to a simple increase in release (7) or to the addition of new contacts with properties like those found FIG. 4. Changes in NMDA and non-NMDA receptor-mediated before have been components of synaptic responses observed in the different condi- potentiation. Facilitated release would tions tested. Results are expressed as the increase in size of the expected to affect those aspects of the field EPSP mediated D-AP5 sensitive (solid bar) and insensitive (hatched bar) components by the NMDA receptors at least to the same degree that it did of field potentials measured by the subtraction procedure in each non-NMDA receptors, something that was observed with condition. Results are mean ± SEM of 5-14 experiments. paired-pulse facilitation but clearly did not occur following Downloaded by guest on September 28, 2021 9350 Neurobiology: Muller and Lynch Proc. Natl. Acad. Sci. USA 85 (1988) LTP induction. A possible explanation for this result would geometry accord with the structural modifications that cor- be that NMDA receptors are only present in limited quanti- relate with LTP. Information ofthis type will clearly be ofuse ties and thus are unaffected by greater release. Two very in further localizing the changes responsible for LTP. different observations argue against this. First, binding stud- ies indicate that NMDA receptors are present in at least as We thank M. Baudry, J. Turnbull, and J. Larson for helpful great numbers as the non-NMDA amino acid receptors, discussion, and J. Porter and M. Lay for secretarial help. This work including the "quisqualate type" thought to mediate trans- was supported by Air Force Office of Scientific Research Grant mission (17). Second, paired-pulse facilitation, which very 86-0099 to G.L. D.M. holds a fellowship from the Swiss National probably involves increased release, had greater effects on Science Foundation, Grant 83.392.0.86. the NMDA component of the response than it did on the 1. Bliss, T. V. P. & Lomo, T. (1973) J. Physiol. (London) 232, non-NMDA components. 331-356. Electron microscopic studies have shown that LTP is 2. Racine, R. J., Milgram, N. W. & Hafner, S. (1983) Brain Res. accompanied by an apparent increase in two types of con- 260, 217-231. tacts, sessile and shaft synapses (18-20). A simple increase 3. Morris, R. G. M., Anderson, E., Lynch, G. & Baudry, M. in synapse number would be expected to affect equally both (1986) Nature (London) 319, 774-776. NMDA and non-NMDA components of the field EPSP and 4. Collingridge, G. L., Kehl, S. J. & McLennan, H. (1983) J. indeed we found that the percentage reduction caused by Physiol. (London) 334, 33-46. D-AP5 increased slightly when postsynaptic responses were 5. Larson, J. & Lynch, G. (1986) Science 232, 985-988. increased by delivering stronger afferent stimulation pulses 6. Larson, J. & Lynch, G. (1988) Brain Res. 441, 111-118. (i.e., when more synapses were activated; see Fig. 4). 7. Collingridge, G. L. & Bliss, T. V. P. (1987) Trends Neurosci. Therefore, if increased numbers of synapses were to account 10, 288-293. 8. Wigstrom, H., Gustafsson, B. & Huang, Y.-Y. (1985) Acta for LTP, one would have to assume that the added contacts Physiol. Scand. 124, 475-479. contained an abnormally low density of NMDA receptors. 9. Collingridge, G. L., Herron, C. E. & Lester, R. A. J. (1988) J. There is also ultrastructural evidence that modifications in Physiol. (London) 399, 283-300. the shape ofdendritic spines occur when LTP is induced (18- 10. Muller, D. & Lynch, G. (1988) Synapse, in press. 21). In fact, the added contacts described above could be due 11. Katz, B. & Miledi, R. (1968) J. Physiol. (London) 195, 481-492. to some type ofspine transformation effect. Two possibilities 12. Wernig, A. (1972) J. Physiol. (London) 226, 751-760. relevant to LTP might be considered here. First, the ana- 13. Zucker, R. S. (1973) J. Physiol. (London) 229, 787-810. tomical changes could alter the surface chemistry of the 14. Nowak, L., Bregestovski, P., Asher, P., Herbet, A. & Pro- the calcium dependency oftransmission in chiantz, A. (1984) Nature (London) 307, 462-466. spine. Analyses of 15. Mayer, M. L., Westbrook, G. L. & Guthrie, P. B. (1984) hippocampus and of paired-pulse facilitation suggest that a Nature (London) 309, 261-264. simple increase in receptor number cannot account for LTP 16. Alger, B. E. & Nicoll, R. A. (1982) J. Physiol. (London) 328, (22), but it remains possible that changes in the properties of 105-123. the non-NMDA receptors (e.g., increased coupling with ionic 17. Cotman, C. W., Monaghan, D. T., Ottersen, 0. P. & Storm- channels, changes in the biophysical properties of the chan- Mathisen, J. (1987) Trends Neurosci. 10, 274-280. nel) participate in the LTP effect. Second, morphological 18. Lee, K., Oliver, M., Schottler, F. & Lynch, G. (1981) Electrical adjustments should change the biophysics of the spine, and Activity in Isolated Mammalian CNS Preparations (Academic, Rall and others (22) have proposed effects of this kind as the New York), pp. 189-212. substrate of In light of our results, 19. Lee, K., Schottler, F., Oliver, M. & Lynch, G. (1980) J. synaptic potentiation. Neurophysiol. 44, 247-258. these biophysical changes would have to favor rapidly 20. Chang, F. L. & Greenough, W. T. (1984) Brain Res. 309, 35- developing synaptic potentials (i.e., the non-NMDA compo- 46. nent) and have little effect on more slowly evolving currents 21. Fifkova, E. & Anderson, C. L. (1981) Exp. Neurol. 74, 621- (i.e., the NMDA component). Computer simulations would 628. be useful in asking if such effects are plausible and if so 22. Rall, W. (1967) in Studies in Neurophysiology (Cambridge whether the necessary changes in spine and spine neck Univ. Press, Cambridge, U.K.), pp. 203-209. Downloaded by guest on September 28, 2021