Abraham y Bear (1996) Met a Plasticity the Plasticity of Synaptic Plasticity. TRENDS in Neuroscience, 19, 126-130

8/6/2019 Abraham y Bear (1996) Met a Plasticity the Plasticity of Synaptic Plasticity. TRENDS in Neuroscience, 19, 126-130

1/5


2/5

W. Abraham and M. Bear - Plasticity of synaptic plasticity VIEWPOINT

(which produces no persistent Achanges in the evoked synaptic re-

a50

sponses) also inhibits LTP inducedg 200

150by strong tetanic stimulation. g ado ___---__----------Surprisingly, however, the same 1 505 Hz priming stimulation can & 200facilitate LTP when evoked by a p Iso -

Test

near-threshold tetanus containingw $OO_

fewer trains. The explanation for 500 10 20 30 40 so 60 70this apparent discrepancy lies in Time (mln)the fact that LTP induction in thelateral perforant path under normal B

conditions varies as an invertedZ 250

I I .U-shaped function of the number of 200

111111 : istimulus trains (Fig. 2). Prior prim- r: l -..

.0 150

ing stimulation shifts the whole 3 _@%\

l \W#

LTP induction function to the left,& 100

OF:_; _ _ ~)----WdV~~~~-___________

so weak stimulation is more likely,UJ 502

LO 30

and strong stimulation is less likely, So Tlma (min) 120 150

to Produce LTP. The complexity Ofthese data emphasizes the need to

Fig. 1. Effect of prior stimulation on long-term potentiotion (LTf) in oreo CA1 of the hippocompus. (A) The popu-

lation excitatory postsynaptic potential (EPSP) was recorded and two separate pathways (contro l and test) were stimu-test a wide range of tetanization /ated a/tern&e/y. At the times indicated by the small downward arrows, weak tetani (30 Hz, 0. I5 s) were delivered toprotocols to characterize fully the the test pathway. A/though this stimulation did not produce a lasting change in synaptic effectiveness, it did inhibiteffects of prior activity on LTP induc tion of LTP by a strong tetanus delivered 20 minut es l ater (indicated by the large downward arrow). LTP on theinduction. control ath was unaffected. (B) The inhibiti on of LTP caused by prior stimulation was transient, lasting no more than

M e t a p l as t i c i t y o f LT Dabout an hour. Figure adapted, with permission, from Ref. 2.

Christie and Abraham described a form of homo-synaptic LTD, termed associative LTD, that occurs inthe lateral perforant path when low-frequency stimu-lation of this pathway is delivered out-of-phase withbrief high-frequency trains delivered to the medialperforant path. In this study, associative LTD only

occurred, however, if prior 5 Hz priming stimulationwas given to the lateral perforant path; no associativeLTD was observed without priming stimulation.Consistent with the other metaplasticity resultsdescribed thus far, the priming effect was found to beinput specific and to involve activation of NMDAreceptors during the priming stimulation. The primingeffect in this case was unusual in its duration, as it wasshown to last for at least two hours. Prior synapticstimulation also can enhance LTD induced by low-frequency stimulation in the CA1 region. Thus, anumber of groups have now reported that a tetanus,which does not itself produce lasting changes in

synaptic efficacy, can facilitate the subsequent induc-tion of homosynaptic LTD (Refs 3,11,12).

To summarize the above data, prior activation ofNMDA receptors, regardless of whether LTP was pro-duced, leads to less LTP and more LTD being inducedby subsequent activity (Table 1). The pharmacology ofLTP facilitation remains to be clarified, although acti-vation of metabotropic glutamate receptors might beinvolved (see below).

M e c h a ni s m s o f m e t a p l a s t i c i t y

Most of the forms of synaptic plasticity we have dis-cussed depend on NMDA receptor activation and a

rise in intracellular Ca ([Ca+]J. Modulation ofNMDA-receptor activation, or the biochemical sequelaeto Ca entry, are likely targets for metaplasticity ex-pression. Thus, we can divide the possible sites ofmetaplasticity into two broad categories: (1) thoseprocesses that regulate the rise in postsynaptic [Ca],;and (2) the downstream processes that are activatedby the rise in [Ca+],.

Metaplasticity by regulation of@ entryA rise in postsynaptic [Ca], following synaptic

stimulation depends on a number of factors thatmight be regulated. Because most CaL+ nflux is voltage-dependent, one powerful site of regulation is the effec-tiveness of converging inhibitory synapses. It is now

well established that auto-inhibition of GABA releaseduring high-frequency stimulation promotes LTP byallowing strong depolarization and thus larger NMDAresponses. Longer-lasting modifications of inhibitionmight contribute to metaplasticity. For example, LTP of

LTP in the Lateral Perforant Path

1 2 3 4

Number of TBS trains

+ Primed-O- Control

A\

--8 16

Fig. 2. Effect of priming stimulation on long-term potentiation (LTP) induction. The lateralperforant path inputs to the dentate gyrus in pentobarbit al-anoesthetized rats received prim-ing stimulation (80 pulses at 5 Hz) followed by LTPinduction IOmin later. An induction func-tion for control paths was established, which showed a threshold for LTP at three trains oftheta-burst stimu lation (TN) and a maximum /eve/of LTP after eight trains. LTPwas much lesswhen I6 trains were delivered. When prior priming stimu lation was given, the numbers of TBStrains required to produce threshold and peok levels of LTP were dramatically reduced. Thereduction of LTP caused by additional trains was also shifted to the left. Priming stimulation byitself did not affect synaptic efficacy. Data point s represent mean values f SET, s a percentageof baseline values. Figure adapted, with permission, from Ref. 10.

J-/NS Vol. 1). No 4, 19% 127


3/5

Effect of prior activity Pattern of prior activity Effect of prior activity Brain area Refson synaptic strength on synaptic plasticity

No change 5x IOOHz, I s+AP5 Increased LTP CAI 13IX ACPD 20 PM, IO min Increased LTP CAI 148x 5Hz, 2s Increased LTP DG IO

6X 30Hz, 150ms Decreased LTP CAI 2IX 5 Hz, 3 min Decreased LTP CAI 5IX I Hz, 8.3-l 6.5 min Decreased LTP CAI 6,78x 5Hz.2~ Decreased LTP DG 9.10

4x 30 Hz, 330ms Increased LTD CAI 38x 5Hz, 2s Increased LTD CAI II2X IOOHz, Is Increased LTD CAI I28x 5Hz, 2s Increased LTD DG 9

LTP 3x IOOHz, Is Decreased LTP CAI I56X 200 Hz, 75 ms Decreased LTP DG I5

180x 500Hz. IOms Decreased LTP MC I6

180x 500Hz, IOms Increased LTD MC I64x IO0 Hz, 500 ms Increased LTD CAI 3

IX IOOHz, I s Increased LTD CAI 6,17,18

2x IOOHz, Is Increased LTD CAI I2

IOX IO0 Hz, 40 ms Increased LTD CAI I9,20

80x IO0 Hz, 50 ms Increased LTD DG 21

Both the decreased LTP and the increased LTD appear to be due to prior NMDA-receptor activation. Increased LTP might be due to prior mtlu-

receptor activation. Abbreviations: APS, aminophosphonopentanoate; DG. dentate gyrus; CA I, CAI region of the hippocampus; MC, Mauchner cell.

inhibitory postsynaptic potentials (IPSPs) has recentlybeen demonstrated in the visual cortexz3, and this wouldbe expected to limit subsequent induction of LTP of

excitatory synaptic transmission severely. The reportedreduction of inhibition following LTP in the hippo-campus should have the opposite effect. Similarly, long-term regulation of K channels could indirectly affectNMDA-receptor function by modulating postsynapticexcitability (for review see Ref. 25). Other putative sitesof regulation include the NMDA receptors themselves,and the postsynaptic Ca2+ dynamics that result fromactivating them2h. Stimulation of the perforant path cancause a significant increase in the synthesis of calbindin-D28K, a high affinity Ca*+-binding protein, in the dentategyrus. Modelling studies indicate that even subtleactivity-dependent changes in Ca buffering, or struc-

tural changes in dendritic spines, can alter Ca*+ diffusionenough to be physiologically relevant2h,2R,2. A role forCa2+-buffering proteins in modulating synaptic plasticitywas confirmed by the recent demonstration that trans-fected cultured hippocampal pyramidal cells expressingcalbindin-D28K have markedly reduced post-tetanicpotentiation. Whether LTP and LTD are similarly af-fected remains to be tested. Finally, prior stimulationmight modify the storage or release of ]Ca2+11 n responseto afferent stimulation by modulating the Ca channelsor pumps in the endoplasmic reticulum.

The long-term regulation of NMDA receptors orchannels deserves further comment. Already there is

considerable evidence that NMDA responses can bepersistently upregulated by tetanizations that also pro-duce LTP of non-NMDA responses3m33. Conversely, adepression of NMDA receptor-mediated responses hasalso been observed following low-frequency stimu-lation that produces LTD (Refs 34-36). These changesin NMDA-receptor function are likely to mean thatthere will be alterations in the ability to induce further

128 1I.U Vol. 11, NC,. 4. 1996

synaptic plasticity later on, although this has not beeninvestigated explicitly. Lasting modifications inNMDA-receptor function are not dependent on a con-

current change in non-NMDA receptors, however,since NMDA-receptor potentiation alone has beenobserved following a short period of anoxia, andNMDA-receptor depression alone has been observedafter weak tetanic stimulationix. Clearly, NMDA-receptor function should be a primary target for studiesof metaplasticity mechanisms.Metoplosticit y by regulation of biochemical processes

There are numerous possibilities for the regulationof downstream components following an elevation in[Ca+],. Levels of [Ca2+li are believed to be translatedinto synaptic modifications by the actions of a net-work of protein kinases and phosphatases (for review

see Ref. 39), and the activity or amount of theseenzymes might be regulated by prior synaptic or cel-lular activation. For example, activation of Ca*+-calmodulin kinase II (CaMK II) appears to be necessaryfor the establishment of LTP in the CA1 region (forreview see Ref. 40). Activity-dependent regulation offunctional CaMK II occurs at three levels: (1) geneexpression-; (2) availability of calmodulin,; and(3) post-translational modification of the enzyme,H.Metaplasticity could be explained by any or all of thesemodifications. Of particular relevance is the observationby Mayford et aLJ9 that a point mutation of CaMK IIthat mimics the effect of autophosphorylation and

makes the kinase less Ca*+-dependent both raises thethreshold for LTP induction and makes LTD more likelyafter low-frequency stimulation (see also Ref. 50).

There are good reasons to believe that metabotropicglutamate (mGlu) receptors, as well as NMDA receptors,are involved in metaplasticity phenomena, since a num-ber of long-lasting effects of mGlu-receptor activationhave recently been reported. A brief application of the


4/5

W. Abraham and M. Bear - Plasticity of synaptic plasticity VIEWPOINT

selective agonist aminocyclopentane-(ls,3R)-dicar-boxylate (ACPD) can directly cause a lasting (~30 min)enhancement of pharmacologically isolated NMDAreceptor-mediated excitatory postsynaptic currents(EPSCs) and induce a long-lasting depression of GABA-mediated IPSPs that should indirectly enhance NMDA-receptor responses These are promising candidatemechanisms for metaplasticity since they might be rela-

tively silent when conventional recordings of field EPSPsare used, yet should profoundly influence the inductionof NMDA receptor-dependent synaptic plasticity. Inagreement with this prediction, there is now evidencethat ACPD application can cause a persistent enhance-ment of LTP induction mechanisms3f14, although thebiochemical mechanisms by which this occurs remainto be elucidated. It is possible that other signallingpathways, for example cholinergic muscarinic recep-tor activation, might work synergistically with mGlu-receptor activation to regulate synaptic plasticitylO.

implic ations of me tapla sticity

The data reviewed above demonstrate that thethresholds for synaptic plasticity are not static proper-ties of synaptic connections but, instead, vary dynami-cally according to the recent history of synaptic ac-tivity (that is, synapses are metaplastic). Clearly, oneimplication of such metaplasticity is that the degree ordirection of synaptic plasticity induced by a particularpattern of conditioning stimulation cannot be pre-dicted unless the previous stimulation history of thetissue is known. There are other implications, how-ever, of which two will be considered.The effect of Tf on subsequent induction ofsynaptic plasticity

In the examples cited above, we focused on the

induction of metaplasticity in cases where it could beeasily distinguished from the induction of concurrentmodifications of synaptic efficacy. However, it isreasonable to assume that the two activity-dependentprocesses would often be induced simultaneously, asboth can be triggered by NMDA receptor activation. Iftrue, what might be some of the consequences? Oneobvious prediction is that LTP-inducing tetanic stimu-lation should inhibit further LTP induction. However,in situations where tetanizations are delivered repeat-edly until no further LTP is induced, this is typicallyreferred to as saturation of LTP. Does this phenom-enon represent a saturation of LTP expression mecha-

nisms (as conventionally believed), or is metaplasticityresponsible (for example, by downregulating NMDAreceptors, thereby preventing further LTP induction)?Frey et LZ~.~ecently addressed this issue by saturatingLTP and then looking to see if further LTP beyond thesaturated level could be induced hours later. Ad-ditional LTP was found to occur eventually, possiblybecause the metaplasticity processes inhibiting LTPinduction decayed over time.

Another prediction is that LTP induction shouldfacilitate subsequent LTD induction. Indeed, in hip-pocampal CA1 pyramidal cells~y~z~17-zo, s well as inthe goldfish Mauthner ce1116, ynaptic depression is more

reliably produced following prior induction of LTP.These results might be explained by LTP raising thesynaptic weights off the floor of their dynamic range,but this seems unlikely given the evidence in both sys-tems that LTD can occur in the absence of prior LTP.We believe that the data are better accounted for byassuming that a strong tetanus, in addition to producing

LTP, also induces metaplasticity which then facilitatesthe subsequent induction of LTD. Thus, the phenom-enon of depotentiation, which results when low-frequency stimulation is delivered to a potentiatedpopulation of synapses, might be considered better asprimed LTD.

If metaplasticity and synaptic plasticity can occurconcurrently, what implications does this have for the

interpretation of events occurring during LTP and LTDinduction? Recall that in the dentate gyrus, LTP in-duction by weak stimulation was facilitated by priorpriming stimulation. Perhaps a similar facilitationregularly occurs during any of the long or repeatedstimulus protocols that are often used to induce synapticplasticity, such that the stimuli occurring early in astimulation sequence prime the induction of plasticityby subsequent stimuli within the same sequence. In thecase of LTD, for example, it seems entirely plausible that,during the 900 pulses (at 1-3 Hz) commonly used toinduce LTD homosynaptically in the hippocampus,metaplasticity is set up early in the stimulus train and

this facilitates LTD induction occurring later in the sametrain. This possibility is ripe for experimental examin-ation since already it has been shown that significantmetaplasticity can develop over the course of minutes.Biochemical correlates of synopti c plasticit y

One area of intense interest in the field of synapticplasticity is the identification of biochemical and mol-ecular correlates of LTP or LTD induction. Correlationshave been made with changes in gene expression, pro-tein synthesis, extracellular protein release, proteinphosphorylation, kinase activity, receptor binding andsynaptic structure. While it is understood that thesestudies do not causally link such changes to LTP or LTD,

the connection is usually hypothesized, especially ifthe correlated cellular change is blocked by an agentthat also blocks the synaptic plasticity, for example, aNMDA-receptor antagonist. However, we now pointout that an equally likely hypothesis is that the changesunder investigation pertain to metaplasticity, ratherthan directly to synaptic plasticity per se.

Concluding remarks

The data reviewed above indicate that prior synapticactivation can leave an enduring trace that affects thesubsequent induction of synaptic plasticity. Many dif-ferent mechanisms probably contribute to metaplas-

ticity, as is also the case for synaptic plasticity. Indeed,there are likely to be many induction mechanisms thatoverlap between the two phenomena, for exampleNMDA-receptor activation and rises in postsynaptic[Ca],. Any biochemical processes set in motion byneural activity could, therefore, play a role in synapticplasticity, metaplasticity, both of them, or neither,and we are challenged to distinguish between thesepossibilities. On the other hand, recognition of thepresence of metaplasticity might yield new interpre-tations of old data and provide new insights into one ofthe key questions in neurobiology: how is informationstored in the nervous system?

S e l e c t e d r e f e r e n c e s

1 Malenka, R.C. (1991) Neuron 6, 53-602 Huang, Y-Y. et al. (1992) Sci ence 255, 730-7333 Wexler, E.M. and Stanton, P.K. (1993) NewoReport 4, 591-5944 Coan, E.J., Irving, A.J. and Collingridge, G.L. (1989) Neurosci.

L&t. 105, 205-2105 ODell, TJ. and Kandel, E.R. (1994) LearningMem. , 129-1396 Fuji i, S . et al. (1991) Bruin Res. 555, 112-122

TINS Vol. 19, No. 4. 1996 129


5/5

AcknowledgementsThis work was

partly supported bya grant rom theHealth Research

Council of NewZealand to

W.C. Abraham andby a Fogarty Senior

International

Fellowship toM. Bear. We thank

A. Cohen,A. Heynen nndR. Sayer for helpful

comments on earlierversions of the

manuscript.

7 Larkman, A. et al. (1992) Nature 360, 70-738 Izumi, Y., Clifford, D.B. and Zorumski, C.F. (1992) Sciencr

257, 1273-12769 Christie, B.R. and Abraham, W.C. (1992) Neuron 9, 79-84

10 Christie, B.R., Stellwagen, D. and Abraham, W.C. (1995)Hippocampus 5, 52-59

11 Christie , B.R., Abraham, W.C. and Bear, M.F. (1993) Sot.Neurosci. Abstr. 19, 1324

12 Wagner, J.J. and Alger, B.E. (1995) 1. Nrurosci. 15, 1577-158613 Bortolotto, Z.A. et al. (1994) Nature 368, 740-743

14 Cohen, A.S., Kerr, D.S. and Abraham, W.C. (1995) Sot.Neurosci. Abstr. 21, 60215 Frey, U. et al. (1995) Neuroscience 67, 799-80716 Yang, X-D. and Faber, D.S. (1991) Proc. Nat1 Acad. Sci. USA 88,

4299430317 Bashir, Z.I. and Collingridge, G.L. (1994) Exp. Brain Rrs. 100,

437-44318 Barrionuevo, G., Schottler, F. and Lynch, G. (1980) Life Sci.

27, 2385-239119 Larson, J., Xiao, P. and Lynch, G. (1993) Brain Xes. 600, 97-10220 Staubli, U. and Lynch, G. (1990) Brclin Res. 513, 113-11821 Christie , B.R. and Abraham, W.C. (1992) Synapse 10, l-622 Davies, C.H. et al. (1991) Nature 349, 609-61123 Komatsu, Y. (1994) I. Neurosci. 14, 6488-649924 Stelzer, A. et al. (1994) Proc. Natl Acad. Sci. USA 91, 3058-306225 Ben-Ari, Y., Aniksztejn, L. and Bregestovski, P. (1992) Trends

Neurosci. 15, 333-339

26 Gold, J.I. and Bear, M.F. (1994) Proc. Nat1 Acad. Sci. USA 91,3941-394527 Lowenstein, D.H. et al. (1991) hleuron 6, 627-63428 Holmes, W.R. and Levy, W.B. (1990) I. Neurophysiol. 63,

1148-116829 Zador, A., Koch, C. and Brown, T.H. (1980) Proc. Nat/ Acad. Sci.

USA 87, 6718-6722

30 Chard, P.S. et al. (1995) Proc. NatIAcad. ci. USA 92, 5144514831 Bashir, Z.I. et al. (1991) Nature 349, 156-15832 Asztely, F., Wigstrdm, H. and Gustafsson, B. (1992) Eur. I.

Neurusci. 4, 681-69033 Clark, K.A. and Collingridge, G.L. (1995) I. Physiol. 482,

39-5234 Xie, X., Berger, T.W. and Barrioneuvo, G. (1992)

1. Neurophysiol. 67, 1009-101335 Gean, P-W. and Lin, J-H. (1993) hreurosci. Lett. 158, 170-17236 Xiao, M .Y., Wigstrdm, H. and Gustafs son, B. (1994) Eur. 1.

Neurosci. 6, 1055-105737 Hammond, C. et al. (1994) Trends Neurosci. 17, 497-50838 Selig, D.K. et al. (1995) Neuron 15, 417-42639 Bear, M.F. and Malenka, R.C. (1994) Curr. @in. Neurobiol. 4,

389-39940 Lisman, J. (1994) Trends Neurosci. 17, 406-41241 Neve, R.L. and Bear, M.F. (1989) Proc. Nat1 Acad. Sci. USA 86,

4781-478442 Hendry, S .C. and Kennedy, M.B. (1986) Proc. Nat1 Acad. Sci.

USA 83, 1536-154043 Mackler, S.A., Brooks, B.P. and Eberwine, J.H. (1992) hreuron

9, 539-54844 Thomas, K.L. et al. (1994) Neuron 13, 737-74645 Skene, J.H.P. (1990) Neurosci. Res. 13 (Suppl.) , S112-S12546 Klee, C.B. (1991) Neurochem. Res. 16, 1059-106547 Miller, S.G. and Kennedy, M.B. (1986) Cell 44, 861-87048 Hanson, P.I. et al. (1994) Neuron 12, 943-956

49 Mayford, M. et al. (1995) Cell 81, 891-90450 Bear, M.F. (1995) Neuron 15, l-451 OConnor, J.J., Rowan, M.J. and Anwyl, R. (1994) Nature 367,

557-55952 Liu, Y.B., Disterhoft, J.F. and Slater, N.T. (1993)

1. Neurophysiol. 69. 1000-100453 Dudek, SM. and Bear, M.F. (1993) 1. Neurosci. 13, 2910-2918

Peter Kiinig is atThe Neuroscirnces

Institute, 10640

John lay Hopkins

Drive, San Diego,CA 92121, USA,

and Andreas K.

Engel and WolfSinger are at the

Mux-Planck-Institutfiir Hirnforschung,

Deutschordenstr. 46,60.528 Frank/&,

Germany.

130

Integrator or coincidence detector?The role of the cortical neuron revisitedPeter Kiinig, Andreas K. Engel and Wolf Singer

Neurons can operate in two distinct ways, depending on the duration of the interval over which

they effectively summate incoming synaptic potentials. If this interval is of the order of the mean

interspike interval or longer, neurons act effectively as temporal integrators and transmit temporal

patterns with only low reliability. If, by contrast, the integration interval is short compared to the

interspike interval, neurons act essentially as coincidence detectors, relay preferentially synchronized

input, and the temporal structure of their output is a direct function of the input pattern. Recently,

interest in this distinction has been revived because experimental and theoretical results suggest

that synchronous firing of neurons might play an important role for information processing in the

cortex. Here, we argue that coincidence detection, rather than temporal integration, might be a

prevalent operation mode of cortical neurons. We base our arguments on established biophysical

properties of cortical neurons and on particular features of cortical dynamics.

Trends Neurosci. (1996) 19, 130-137

A THOUGH UR KNOWLEDGEbout the morpho-logical and physiological features of cortical cellshas increased substantially over the past 20 years, thebasic operational mode of cortical neurons has

remained controversial. The traditional view, whichstill predominates in cortical physiology and mostneural network models, considers the cortical neuronas an integrate-and-fire device. This view was advo-cated first by Sherrington and later supported by evi-dence obtained from the spinal cord. An alternativeconcept, proposed about a decade ago3f4, suggests that

neurons in the cortex operate primarily as detectorsfor the temporal coincidence of synaptic inputs. Thisproposal is motivated by the assumption that corre-lated activity of neurons is of crucial importance for

cortical processing and that synchrony might, inparticular, contribute to solving the so-called bindingproblem, that is, the problem of integrating distrib-uted information into coherent representational pat-tern?. Such a temporal code can only be employedby the nervous system if neurons are sensitive to coin-cidence. Otherwise, it would be impossible to convey

TINS Vol. 1 9, No. 4, 1996 Copyright 0 1996, Clsevier Science Ltd. All rights reserved. 0166 2236/96/S15.00 PII:S016622369620008h

Abraham y Bear (1996) Met a Plasticity the Plasticity of Synaptic Plasticity. TRENDS in Neuroscience, 19, 126-130

Documents