Calcium in long-term potentiation as a model for memory

Neuroscience Vol. 10, No. Printed in Great Britain

0306-4522/83$3.00+ 0.00

4, pp. 1071 to 1081, 1983

Pergamon

Press Ltd IBRO

COMMENTARY CALCIUM IN LONG-TERM POTENTIATION AS A MODEL FOR MEMORY J. C. ECCLES Abteihmg Neurobiologie Max-Planck-Institut

fiir Biophysikalische Chemie, Giittingen, W. Germany

CONTENTS Introduction Long-term potentiation (LTP) Calcium and long-term potentiation The postsynaptic densities @‘SD) The synaptic spine Possible presynaptic contributions to long-term potentiation Comprehensive hypothesis for long-term potentiation A model for cognitive memory built on long-term potentiation Cerebellar learning and conjunction depression Durations of long-term potentiation and cognitive memory Abstract-The granule, CA1 and CA3 cells of the hippocampus have been much investigated during the last decade because there is superimposed on the standard features of synaptic transmission a very prolonged potentiation lasting for weeks that is called long-term potentiation. Evidently long-term potentiation is a promising candidate in the construction of a model for memory. The thesis here developed is that the influx of calcium ions across the membrane of the granule and pyramidal cells plays the key role in the generation of long-term potentiation. This proposal makes it possible to account for the necessity of strong repetitive synaptic stimulation, preferably in bursts so as to optimize the conditions for the calcium influx. Studies on hippocampal slices with variations in the synaptic inputs to the granule cells give evidence of cooperativity, which is interpreted in relation to the threshold membrane depolarization for calcium influx. It is conjectured that the large increase of calcium in the granule and pyramidal cells results in its combination with the specific protein, calmodulin, to form a second messenger system, which produces metabolic changes leading to an increase in receptors of the postsynaptic membrane of the spine synapses, i.e. the postsynaptic densities, to the synaptic transmitter, glutamate. For example, Ca2+ could activate calcium-dependent kinases in the postsynaptic density resulting in the modification of protein components by phosphorylation. Other postsynaptic factors contributing to long-term potentiation are presumed to be protein synthesis with spine swelling and increased transport up the dendritic microtubules. There is discussion of the evidence for the alternative hypothesis that long-term potentiation is primarily presynaptic, being due to an increased output of transmitter. A unifying hypothesis is formulated, namely, that the primary event in long-term potentiation is in the increased sensitivity of the postsynaptic densities to the transmitter, and that, secondarily, this induces an increased output of transmitter from the presynaptic terminals by a trophic action across the synaptic cleft. It is shown how the proposed combination of calcium with calmodulin will account for the hypothesis of Marr that cognitive memory is due to conjunction potentiation. Furthermore, the Marr-Albus hypothesis for cerebellar learning is accounted for if the calcium-calmodulin messenger system causes the observed depression of the transmitter sensitivity of the spine synapses on Purkyne cells.

In the hippocampus there are three synaptic mechanisms that lend themselves to analytical investigations because of the simplicity of their monosynaptic connections (Fig. I). There are, firstly, the fibres from the entorhinal cortex (ento) via the perforating pathway (pp) that make synapses on the apical dendrites of granule cells of the dentate area. Secondly, the axons of granule cells make synapses as mossy fibres (mf) on the apical dendrites of CA3 pyramidal cells. Thirdly, the Schaffer axon collaterals (Sch) of the CA3 cells make synapses on the dendrites of CA1 pyramidal cells. Initially, the investigations were carried out in uivo usually on

rabbits, but in the last 10 years the hippocampal slice technique has become increasingly used because the lamella structure of the main hippocampal connectivities can be incorporated in a single slice (Fig. l), and the slice preparations exhibit responses similar to those found by the much more laborious in viva techniques. In developing an hypothesis of memory, the experimental observation on the perforating pathway from the entorhinal cortex to granule cells is of particular significance because of its inherent simplicity. It is necessary to give a brief review of the synaptic activation of granule cells before dealing

1071

J. (‘. Eccles

1. Drawing of the principal pathways in a Iamella of the hippocampus. The pathway of special interest is from

Fig.

the entorhinal cortex (ento) via perforating pathway (pp) to excite monosynaptically the dentate granule cell indicated by arrow.

with the special features on which a model for memory can be built. Stimulation of the entorhinal cortex (Fig. 1, ento) causes a monosynaptic activation of the dentate of a granule cells 7.37.46 (Fig. 2) with the production

typical excitatory postsynaptic potential. EPSP. The impulses in the perforating pathway (Fig. 1) (pp), cause the liberation of the synaptic transmitter glutamate,4Rh.5” from the synaptic vesicles (Fig. 3). The widely accepted liberation mechanism is that the

Fig. 2. Drawing of a dentate granule cell showing 4 dendrites in outline with synaptic spines drawn on both sides of one. The Ca2+ input across the dendrite surface membrane is indicated for the left dendrite are drawn in the right and also the intracellular flow of Ca 2t ions down to the soma. Microtubules dendrite by three interrupted lines. In the inset the microtubules are shown passing into the spine through the spine apparatus (cf. Fig. 3).

Long-term potentiation, calcium and memory

Fig. 3. Drawing of spine synapse on a dentate granule cell showing dendritic microtubul~ going into a spine up to the spine apparatus. The postsynaptic density is indicated by the arrow. From the presynaptic membrane there are dense projections up to and between the synaptic vesicles (Grayz7). influx of Ca2 ’ ions into the presynaptic knobs results in the activation of the protein calmodulin,io by 4Ca2” ionsto’ to each calm~ulin molecule.g The activated calmodulin then causes the initial liberation of glutamate and further accounts for the role of Caz+ in causing 32a frequency potentiation during repetitive stimulation’ and the post-tetanie potentiation (PTP) after repetitive stimulation.45.47 All of these influences would be on the presynaptic liberation of glutamate from the synaptic vesicles and into the synaptic cleft (Fig. 3) and are over within a few minutes, even after a severe tetanization.47

LONG-TERM ~OT~NT[ATt~N (LTP)

Subsequent to the initial responses evoked by repetitive synaptic stimulation, there develops an extraordinarily long synaptic potentiation that begins after a latency of 10-20 s, reaches a maxims in minutes7,4s and persists for many days.4*6*i3*6sIt has been called long-term potentiation, LTP, to distinguish it quite sharply from PTF. Long-term potentiation is about 10,000 times longer.4 [Editor’s note: see also preceeding Commentary by Voronins9”].

In the usual investigation for LTP, a stimuIati~g electrode is appIied to the entorhinal cortex (ento) or perforating pathway (pp) and a recording microelectrode is inserted to the level of the dendrites of the dentate granule cells (Fig. 1). It records the excitatory postsynaptic potential (EPSP) generated by a population of granule cells as a negative wave (Fig. 4Aa). Synaptic potentiation is indicated by an increase in the EPSP evoked by a standard testing volley set up by the application of a single stimulus to ento or pp (Fig, 1). In order to demonstrate an LTP, there has to be strong repetitive presynaptic stimulation47 (ENTO or pp of Fig. 1) at a frequency above 51s” and with many stimuli. The optimal conditions have been empirically determined to be bursts of about S-10 stimuli at 400/s and repeated every 10s for perhaps 10 times, i.e. above 100 stimuli in just over a minute.*2 There has to be convergence of many perforating path impulses on a granule cell in order to induce an LTP, a finding named cooperativity.47 There are about 10,000 spines on a granule cell and activation of about 400 is required for generation of an impulse discharge,46 and possibily even more for producing an LTP. In LTP, the synaptic potency as

1073

J. (‘. Eccles test S+RA*

A

-g-J

-....,

. ..

;:

:

_..::*... . .

~

l&i!_& \

FC

EC

tt S&I

tt S&

Fig. 4. Effect of entorhinal conditioning stimulation S1 or S2 on testing responses evoked by a testing entorhinal cortex (EC) stimulation, S 1.Extracellular recording from dentate granule cells gives population excitatory postsynaptic potentials (EPSPs). In A, S, stimulation of EC excites the perforating path to the ipsilateral granule cells for both testing and conditioning stimulation. The conditioning was by 8 trains of 8-10 pulses at 400 Hz delivered every 10s. (a) Shows the population EPSP recorded by microelectrode R1 before conditioning and (b) at about 2min after the conditioning tetanus. A second conditioning tetanus was applied at 16 min after the first and (c) gives the population EPSP about 2 min later, revealing a slight further increase in LTP. In B, there were the same EC stimulation procedures by S1, but the recording by R2 was for the population EPSP of the contralateral side. Note absence of LTP in b and c. In C, there was recording by microelectrode R2 as in B, but there was now conditioning by a conjoint stimulation of both entorhinal cortices, S1 and Sz, as indicated. As a consequence, there was, b, c, a large LTP of the testing contralateral response S, alone, in great contrast to the absence of LTP in B. (Levy & Steward3’).

measured by the extracellular EPSP (Fig. 4) is rarely over double,7*37,45,47 but it persists for days. After 12 daily reinforcements by stimulation through a chronically implanted electrode the LTP declined with a time constant of 37 days.4 The pioneering experiments on LTP in the in uiuo preparation have been greatly extended by analytical experiments on hippocampal slice preparations (Fig. 1). In appropriate media, these slices give reliable performances for many hours and have the singular advantage of greatly simplifying the recording procedures and of allowing investigations on the effects of variations in the composition of the bathing medium. That LTP has independence

from the presynaptic

events referred to in the introduction is indicated by the action of 2-amino-4-phosphonobutyric acid (AFB)” which blocks the action of the transmitter (glutamate) on the postsynaptic receptors but has no presynaptic action. Strong repetitive presynaptic stimulation can be completely blocked by AFB through this postsynaptic blockade. After removal of AFB there was relief from the block but no LTP was observed,” which is contrary to what would be expected if LTP were presynaptically generated by an increased transmitter output. This finding clearly indicates that LTP results from the strong transsynaptic action, i.e. that it is initiated postsynaptitally. There will later be a consideration of the alterative hypothesis of presynaptic generation.

Long-term

potentiation,

CALCIUM AND LONG-TERM POTENTIATION If the calcium of the bathing medium was reduced for 2.5 to 1 mM, synaptic transmission and PTP were well maintained, but LTP failed.r6 A special role for Ca’+ in LTP was also suggested by experiments in which Ca’+-free bathing medium was substituted after LTP was fully developed.17 All transmission was temporarily blocked during this episode, but on restoration of the normal Ca*+-containing medium, the recovered synaptic transmission exhibited the LTP exactly as before the episode.i7 This indicates that LTP is an enduring process fully set up in a few minutes by tetanization in the presence of Ca’+ and thereafter surviving at full strength for many minutes in the absence of CaZ + It is the thesis of this article that LTP is primarily postsynaptic and that the influx of Ca’+ ions into the granule cell soma and dendrites (Fig. 2) sets in train a series of processes that result in the enduring LTP. The LTP of CA1 cells seems more localized. The hypothesis is that the large and prolonged membrane depolarization46 set up by the intense and prolonged synaptic stimulation causes the opening of voltage-dependent Ca2+ gates over the whole soma and dendritic membrane. There is a rapid influx of Ca2+ into the granule cell (cf. Fig. 2) down an electrochemical gradient that is as high as - 200 mV inside to outside, the inside Ca2+ concentration being as low as lo-* Mg. The Ca2+ gates only open with a large depolarization, perhaps as high as 20mV.62 They are to be sharply distinguished from voltage-dependent Na+ gates.2g In some hippocampal neurones, but apparently not granule cells, there may be an initial self-regenerative Ca2+ depolarization, a calcium spike.52,62 However, the influx of Ca2+ would be fully operative through voltage-dependent channels of the whole dendritic membrane because synaptically-induced depolarizations spread very effectively by electrotonus, as is evidenced by the intracellular recording of EPSPS.~~ The empirical finding that LTP is most effectively induced by brief stimulus bursts repeated every 10 s is attributable to the necessity for renewal of the extracellular Ca2’. For example, after brief seizures of the piriform cortex, the extracellular Ca2+ was reduced to half in a few seconds and recovery was far advanced 10 s after the seizure.25 We now confront the problem of how the increased intracellular CaZf can increase the effectiveness of the glutamate liberated into the synaptic cleft in close proximity to the postsynaptic density (Fig. 3, arrow) that incorporates the glutamate receptors. There is now good evidence that immediately Ca2+ ions enter a nerve cell at a sufficient concentration there is combination with calmodulin (CaM), a protein that has been very intensely studied.g~g”~24~26~48 Calmodulin is a chain of 148 aminoacids (MW 16,700, fully sequenced) with 4 specific Ca2 + sites that become fully occupied when the Ca2+ reaches a sufficient concentration. This

calcium

1075

and memory

complexed CaM (Ca2+ - CaM) becomes a powerful activator of enzyme systems and proteins. For example, it activates adenylate cyclase to convert ATP to cyclic-AMP, and phosphodiesterase to degrade cyclic-AMP to the inactive AMP.g*48 Complexed CaM also has a wide range of actions on protein kinases,2894* so that it can be regarded as a second messenger system. g*51It is of special interest that trifluoperazine inhibits the action of CaM on enzyme systems by preventing its combination with Ca2+ 9a.36aand that could cause the observed failure of LTP.24.40 THE POSTSYNAPTIC

DENSITIES @‘SD)

The PSDs of the spine synapses of the cerebral cortex (including the hippocampus) are very thick structures (Fig. 3, arrow) that have been studied in detail by a variety of techniques.43 Several constituent proteins have been recognized, actin, tubulin, calmodulin and an unidentified protein of about 51,OOOMW (PSD51)26*33,43*44 which form a cytoskeletal framework,43,44 apparently for mobilizing glutamate receptors. The simplest explanation of LTP is that Ca2+-activated protein kinases (presumably by means of Ca2+ - CaM) cause the uncovering of glutamate receptor sites, which has been demonstrated by the increased uptake of glutamate. 4a,4b*40It was there postulated to be an effect restricted to activated dendritic spines with no cooperativity, which is the situation described for the LTP of CA1 pyramidal cells (Fig. l).2,40 However, the threshold and cooperativity exhibited for the LTP of granule cells47 necessitate an explanation in terms of EPSPs generated by many activated spine synapses summating in producing dendritic depolarization that surpasses the threshold for opening Ca2+ gates. Then, there could be the sequence: influx of Ca’+; production of the second messenger system, Ca2+ - CaM; action of Ca*+ - CaM to produce the LTP, which will now be discussed. The fastest action would be by the Ca2’ - CaMactivated kinases on the PSD,24 phosphorylating its proteins5i and activating or exposing the glutamate receptors.4a*4b.40 The high concentration of CaM in the PSDs, about 5 to lO%,26 would facilitate the Ca2’ action on the PSDs. The next fastest action could be on the protein transport up the dendrites by microtubules and into the spines (Fig. 3), this transport presumably being accelerated by Ca2+ - CaM just as with axonal microtubules. Another relatively short latency effect would be by the Ca2+ - CaM-activating protein synthesis in the polyribosomes that are located in the dendrites, in proximity to the spine origin55 with transport to the PSD of the adjacent spine. The most delayed action would be by the Ca2 + - CaM second messenger system acting on the protein manufacture by the nucleus and perikaryal ribosomes with subsequent transport up the dendritic microtubules (Figs 2,3). It would seem that

this latter process is too slow to contribute to the O~IW of an LTP, which reaches 66% of its maximum in 76s after the initiating tetanus4’ Thus. it is envisaged that LTP is probably generated by a temporal sequence of actions, as described above, but much more investjgat~on is required. THE SYNAPTICSPINE The LTP of granule cells is associated with an hypertrophy of the spine heads by 3&40%.21~22 This hypertrophy is prevented by a preceding injection of anisomycin, which induces a brief, less than 2 h suppression of the translation phase of protein synthesis2’ A limited hypertrophy, about 20%, appears as the anisomycin suppression passes off. A related observation is that LTP in hippocampal slices is accompanied by protein synthesis.” The hypertrophy of the spine head in LTP is accompanied by an increase in the diameter of the spine stalks by up to 60%, which was significant at 4mm after the conditioning tetanus, but which tended to decline at 90min.20 RaIlso first suggested that decrease in the resistance of the spine stalk could provido a basic mechanism for learning, which is related to the observed increase in caliber of spine stalks. However, calculations by Jack, Noble & Tsien3’ suggested that this mechanism for control of synaptic potency is inefficient. A doubling of the electrical length constant of the spine stalk would produce only a 10% change in the membrane voltage that a spine synapse induces in the parent dendrite. Koch & Poggio34 now give a more effective action in their calculation. However, there is much uncertainty in the parameters, and they entertain the possibility “that the spines would not have any specific electrical properties that could play a role in learning.” The repetitive stimulation inducing the LTP also causes within minutes swelling of the nucleus and increase in perikaryal ribosomes (E. Fifkova, personal communication), which indicates that the whole neurone is implicated in the activity of Ca*+ - CaM in generating the LTP. An outstan~ng problem, therefore, is to account for the restriction of the LTP to the activated synapses. It is here suggested that this restrictive action is due to the spine apparatus that is located in the neck of dendritic spines of the cerebral cortex and that lies athwart the microtubules entering the spine27 (Fig. 3). It could act as a gating device blocking microtubule transport except for a seclective channelling into those spines recently activated. In this connection, it is of interest that the spine apparatus sequesters &“++8’2” possibily as Ca’* - CaM, and so the spine apparatus may receive the signal for opening to microtubule transport from the high Ca2’ influx into its activated spine. The spine apparatus of an inactivated spine would receive no such signal. In contrast to granule cells and Cal pyramidal

cells with an exclustvcly homosynaptic L.Tp. rhV (_‘A? pyramidal cells display a dell-d~vclopcd hcrcrosynaptic L.TP.“‘“,“” It is of great rclcvancc to the present hypothesis that the synapses of the mossy fibres on lhe CA3 pyramidal cells do not make simple contacts with dcndritic spines as in Figs 2 and 3. hur there are instead very complicated protrusions from the dcndrites into which the prcsynaptic terminals projcct.2Ja.2ga This structure is not conducive to the selective control of input that is required for a homosynaptic LTP, and that is postulated to be due to gating by the spine apparatus: hence. the property of heterosynaptic LTP of CA3 pyramidal cells would be expected on the gating hypothesis. In summary, there is much evidence that LTP is induced postsynaptically by the following scqucnce of events: (1) Ca’ ’ influx into the dendrites and soma, which are strongly depolarized by the powerful bursting synaptic stimulation; (2) the Ca2’ activation of CaM, 4Ca2 ’ ions to each calmodulin molecule; (3) Ca” -- CaM acting as a second messenger system producing a variety of actions on protein and enzyme systems that result in an increase in glutamale receptor sites of the PSDs and a swelling of the spines. PO.SSIBLE PRESYNAPTIC

DISTRIBUTIONS

TO

LONC-

TERM POTENTIATION

In several recent publications,“*’ rz3 a presynaptic locus for LTP was supported by evidence that a conditioning repetitive stimulation produced for one hour an increase in the resting presynaptic release of a glutamate analogue53 or of labelled glutamate.” Such observations suggest that LTP may be induced in part by an increased presynaptic output of glutamate by the testing stimulus. The other evidence adduced in support of a presynaptic location for LTP could equally well be attributed to a postsynaptic location as in the present hypothesis, for example, the increase in uptake and retention of Ca2 ‘,” that paralleled the LTP of CA1 pyramidal cells.” More si~ni~~ant evidence against the postsynaptic location of LTP is the finding ‘* that there was no increased responsiveness to iontophoretically-applied glutamate after a brief increase (5 IO mm) in the Ca2 ’ of the bathing medium from 2 to 4mM. There was. however, a prolonged increase, for at least 3 h. in the response to presynaptic inputs. This resembled a LTP in both EPSP and population spike,4H but it could be prcsynaptic or postsynaptic or both. Against this evidence for a presynaptic generation of LTP, there is the finding4b that stimulation of the hippocampal slice so as to induce a LTP was accompanied by a long-lasting (30min) highly significant increase by about 20% in the glutamate accumulation by the slice, which is restricted to the stimulated area. This indicates that there is an increase in the number of glutamate receptors during LTP. Furthermore, in isolated hippOGimPaJ mem-

1077

Long-term potentiation, calcium and memory branes the reversible glutamate uptake was increased by 80% in the presence of 250pM calcium.40 Presumably, this reversible uptake was on the glutamate receptor sites of the PSDs. COMPREHENSIVE

HYPOTHESIS

FOR LONG-TERM

POTENTIATION

As discussed by Bliss & Dolphin,’ there is good evidence that LTP is in part postsynaptic. The question now arises: is it possible to develop a comprehensive hypothesis that recognizes the primacy of the postsynaptic origin of LTP, but also accounts for such presynaptic effects as an increased transmitter output? It is important to comprehend the spine synapse (Fig. 3) as a functional element, which we may term a synaptic unit. It is not simply a site for chemical transmission in the conventional sense, the chemical transmitter ejected from the presynaptic terminal into the synaptic cleft, diffusion of the transmitter to the postsynaptic receptor sites and its molecular linkage to these sites, the consequent opening of ionic gates across the postsynaptic membrane. There are more subtle trophic actions that are as yet rather ill defined.59 For example specific polypeptides, retrophins, are secreted postsynaptically and taken up presynaptically to be retrogradely transported up to the presynaptic neurone for necessary trophic actions. Failure of this transport, as after axonal section, results in chromatolysis which may lead to cell death. Nerve growth factor36 is an example of a retrophin for sympathetic ganglion cells, and other examples are being discovered. Close interaction between the presynaptic and postsynaptic components of a synapse is indicated by their topographic relationship. In an asymmetrical synapse (cf. Fig. 3) there is a close matching of the active presynaptic dense projections with the PSD. There is now abundant evidence that synapses may develop a central perforation that may lead to a horseshoe shape and an eventual fragmentation.8b.49.60*61 Through all these structural vicissitudes there continues to be close matching between the presynaptic and postsynaptic components, as is indicated diagrammatically in Fig. 1 of ref. 8b. This matching is particularly well shown by the beautiful technique of Vrensen and associates 60*61which enables synaptic discs to be viewed full-m-face and in section. Evidently, there is an extraordinary degree of information flow across the synaptic cleft. Hence, it is conceivable that when PSDs are transformed to give the increase in glutamate receptors during LTP,4b.40 there can be across the synaptic cleft a complementary influence on the presynaptic terminal with increased output of transmitter. Thus, though initially LTP may be an exclusive postsynaptic process, as proposed above, there could be secondary presynaptic changes such as those reported.3’5.11,53.58

A MODEL

FOR COGNITIVE LONG-TERM

MEMORY

BUILT

ON

POTENTIATION

The proposed postsynaptic origin of LTP is of great significance in respect of the hypothesis developed by Marr 4**42that long-term memory is encoded in synaptic potentiations that are set up by a conjunction process, and that have an indefinitely long duration. An essential feature is the interaction between a strong conditioning synaptic input to a neurone and a weak synaptic input to that same neurone at about the same time. For example, in the hypothesis of cerebral learning,42 it was proposed that, when there was conjunction of the strong excitation of the apical dendrites of a pyramidal cell by the climbing fibres, the cartridge synapse of Szentagothai 57 (Fig. 5) with the weak excitation of the terminal branches of that same dendrite by horizontal fibres in lamina I, there would be induced a prolonged potentiation of the horizontal fibre synapses. i8 The deficiency of the hypothesis was that there was no specification of the mode of interaction either in its nature or its timing. Figure 5 represents the essential features of interaction together with the specificity of the synaptic potentiation. On the basis of investigations on hippocampal LTP by Levy & Steward,37r38 it is possible to construct a model of the Marr conjunction hypothesis. It is based on the strong projection from the ipsilateral enthorhinal cortex (EC) to the middle and distal thirds of the granule cell dendrites (Fig. 1) and on the weak contralateral projection to the proximal third of these same dendrites. In Fig. 4(A) a conditioning tetanus of the ipsilateral EC set up an LTP, whereas in Fig. 4(B) after a similar stimulation of the contralateral EC, there was no LTP because the synaptic excitation was too weak. However, when the conditioning stimulation was applied to both the contra and ipsi-ECs, there was a large LTP of the response to the contra-EC alone (Fig. 4Cb, c). On the calcium hypothesis of LTP, the contralateral EC synapses can take advantage of the increased intracellular Ca*+ produced by the strong ipsilateral EC stimulation. This provides an excellent model for the Marr conjunction hypothesis with a safeguard of the necessary selectivity, because only when the contra-EC synapses are activated at about the same time as the conditioning ipsiEC synapses do they participate in the LTP. Inactive synapses are unaffected. Levy & Stewardja have further demonstrated the requisite temporal relationship of the conjunction. The LTP occurs only if the weak contralateral stimulus burst occurred at a time before the ipsilateral of not more than 20ms. Simultaneity also is effective, but not the reverse sequence. The hypothesis of the postsynaptic origin of LTP based on increased intracellular CaZ+ can thus be converted into a more developed variant of the Marr hypothesis. This is justified because Lee35 has found that slice preparations of several areas of the cerebral

1078

J.C. Eccles

[ "A

~

~

"'C ~

i T-

'

,

i i |

Ir ~..

II

i

'!

i

i

l

,"

'

"

"

: •

:

l ,Jt,

: I

I

i

i

J

:

i i :

~i;!~i~

i

:~!~%.<:

l

tv ~i, '#: -::. :":': . . . . . . .

:SSt~

i

l...,_~_..:2_I2q224,.,

|

i

iii::i{::~: Py

..:::: -........

::~':.:!~:N~I~!.:~

i l

:

Fig. 5. Simplified diagram of connectivities in the neocortex that is constructed in order to show pathways and synapses in the proposed theory of cerebral learning (cf. Eccles18). The diagram shows three modules, A, B, C. In laminae 1 and 2 there are horizontal fibres arising as bifurcating axons of commissural (COM) and association (ASS) fibres and also of Martinotti axons (MA) from module C. The horizontal fibres make synapses with the apical dendrites of the stellate pyramidal cell in module C and of pyramidal cells in modules A and B. Deeper, there is shown a spiny stellate ceil (Sst) with axon, AX, making cartridge synapses with the shafts of apical dendrites of pyramidal cells (Py). Due to conjunction hypertrophy the association fibre from module C has enlarged synapses on the apical dendrites of the pyramidal cell in module A (Eccles 1s). cortex exhibit a LTP matching that for the hippocampus. Thus, cerebral learning can be attributed to the calcium-induced changes in the postsynaptic densities on the spine synapses of the pyramidal cells, particularly for synapses made by horizontal fibres in lamina I, as outlined in a recent commentary. 18 There is urgent need for the experimental testing of this comprehensive hypothesis. F r o m the unitary effect illustrated in Fig. 5, there has been developed an hypothesis of the manner in which cognitive memories can be stored and retrieved in the cerebral cortex. 1 S CEREBELLAR LEARNING AND CONJUNCTION DEPRESSION

Marr 41 first developed the conjunction hypothesis for cerebellar learning, proposing that climbing fibre impulses conditioned the Purkyn6 cells so that

parallel fibre synapses activated at about the same time underwent a prolonged potentiation, exactly as with the cerebral cortex. A little later, Albus I made the inverse proposal, namely, that conjunction interaction induced depression of the parallel fibre synapses on the Purkyn6 cell dendrites. In a recent rigorous testing Ito, Sakurai & Tongroach 31 have demonstrated the conjunction depression proposed by Albus. Nevertheless, it appears that the climbing fibre activation of Purkyn6 cells causes a large Ca 2 ÷ influx into Purkyn6 cells. 19'39 A modification of the calcium LTP hypothesis can be proposed, namely, that in the Purkyn6 cell the calcium-calmodulin second messenger system has an inverse effect, causing the postsynaptic densities to be less receptive to the transmitter, glutamate, as is actually observed. 31 Two comments may be made. Firstly, it has been recognized that the spines on Purkyn6 cells differ from those in the cerebral cortex and hippo-

Long-term potentiation, calcium and memory

campus in having postsynaptic densities of only about half the thickness*” and in an absence of the spine apparatus. Baudry & Lynch reported that in contrast to hippocampus and cerebral cortex Ca2+ caused almost no increase in the glutamate binding sites in the cerebellum.4” One wonders if the long-term depression (LTD) of these synapses is non-specific and heterosynaptic. Secondly, the inverse effect, LTD instead of LTP, may be correlated with the inhibitory action of PurkynE: cells on their target neurones in the cerebellar nuclei, LTD being converted to LTP in the output of these nuclei with the further projection from the cerebellum. DURATIONS

OF

LONG-TERM COGNITIVE

POTENTIATION

AND

MEMORY

As we enquire into the applicability of the postsynaptic calcium model for providing an explanation of cerebral memory, we have to consider the duration of the LTP. At this level of enquiry we have to consider the question of the long duration of the LTP, particularly after repeated reinforcement. Can the LTP of the hippocampal granule cells be a model for long-term cognitive memory that may last for a life-time? The elemental hypothesis of memory is that a spine synaptic unit such as that of Fig. 3 can be permanently changed to be a more powerful mechanism. This could be effected by inducing more enduring LTPs, by such procedures as repeated reinforcements over much longer periods than the 12 daily hitherto applied,4 which gave an LTP lasting for many weeks. It is relevant to this enquiry that cognitive

1079

memories do not become permanently established until they have been replayed in the cerebral circuitry for about 3 years. 54 We consolidate our memories by repeated recall-voluntary or involuntary. One can predict that with repeated reinforcement over many months the duration of hippocampal LTPs may match that of cognitive memories. It is most important that Lee’s35 pioneer investigations on LTP in slices of cerebral cortex should be made the basis of many research programs. The principal objection that can be raised against an hypothesis that attributes long-term memory to long-lasting changes in the effectiveness of spine synapses is that these synapses appear to be labile.8b,4g,60*6’ Cotman and associates49 give evidence that the PSDs of spine synapses on dentate granule cell undergo cyclic changes following ablation of the ipsilateral entorhinal cortex. There appears to be a cycle of growth, central perforation, complex perforations, fragmentation and regrowth of the PSD from fragments. Even if there is this lability in the life cycle of any one PSD,8b it is possible still to envisage that LTP is a property of an assemblage of spine synapses which collectively would preserve the potentiated response. The LTP with a time constant of decay of 17 weeks4 is a functional test that necessitates some very long overall survival of whatever structural changes subserve LTP. The conclusion is that LTP provides to an amazing degree the basis of a model for cognitive or cerebral memory. However, we are confronted by great unknowns in the attempt’s to build a comprehensive hypothesis of the manner in which memories are coded in the brain and retrieved therefrom.

REFERENCES

Albus J. S. (1971) A theorv of cerebellar function. Math. Biosci. 10. 25-61 Andersen P:, Sundberg S-H., Sveen O., Swann J. N. and Wigstrom H. (1980) Possible mechanisms for long-lasting potentiation of synaptic transmission in hippocampal slices from guinea-pigs. J. Physiol.. Lond. 302,463482. Baimbridge K. G. and Miller J. J. (1981) Calcium uptake and retention during long-term potentiation of neuronal activity in the rat hippocampal slice preparation. Brain Res. 221, 299-305. Barnes C. A. and McNaughton B. L. (1980) Spatial memory and hippocampal synaptic plasticity in senescent and middle-aged (ed. Stein D.) pp. 253-272. Elsevier-North Holland, Amsterdam. _ . __ rats.. In _ Psychology . - _---..of Aging 4a. Bauory M. and Lynch G. (1980) Regulation of hippocampal glutamate receptors: evidence for the involvement of calcium activated protease. Proc. natn Acad. Sci. U.S.A. 77,2298-2302. 4b. Baudry M., Oliver M., Greager R., Wieraszko A. and Lynch G. (1980) Increase in glutamate receptors following repetitive electrical stimulation in hippocampal slices. Life Sci. 27, 325-330. 5. Bliss T. V. P. and Dolphin A. C. (1982) What is the mechanism of long-term potentiation in the hippocampus? Trends Neurosci. 5.289-290. 6. Bliss T. V. P. and Gardner-Medwin

A. R. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the unanaesthetized rabbit following stimulation of the perforant path. J. Physiol., Lond. 232, 357-374. 7. Bliss T. V. P. and Lsmo T. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J. Physiol., Lond. 232, 331-356. 8. Burgoyne R. D., Gray E. G. and Barron J. (1983) Cytochemical localization of calcium in the dendritic spine apparatus of the cerebral cortex and at synaptic sites in the cerebellar cortex. J. Anat. In press. 8a. Carlin R. K., Grab D. J., Cohen R. S. and Siekevitz P. (1980) Isolation and characterization of postsynaptic densities from various brain regions: enrichment of different types of postsynaptic densities. J. Cell Biol. 86, 83 l-843. 8b. Carlin R. K. and Siekevitz P. (1983) Plasticity in the central nervous system: do synapses divide? Proc. natn Acad. Sci. U.S.A. 80, 3517-3521. 9. Cheung W. Y. (1980) Calmodulin plays a pivotal role in cellular regulation. Science, N. Y. 207, 19-27. 9a. Cheung W. Y. (1982) Calmodulin. Scient. Am. 246, 6, 48-56. 10. DeLorenzo R. J. (1981) The calmodulin hypothesis of neurotransmission. Cell Calcium 2, 365-385.

I OS0

J. c‘. Ecclev

IOa. Dodge F. A. and Rahamimon‘ R. (1967) Co-operative action of calcium ions in transnutter rcleasc ,(I the neuromuscular junction. J. Physiol., Lond. 193, 419432. I I. Dolphm A. C., Errington M. L. and Bliss T. V. P. (1982) Long-term potentlation of the pcrt’orant path 1,~ I ICOI\ associated with increased glutamate release. Nuture, Lond. 297,496498. I?. Douglas R. M. (1977) Long lasting synaptic potentiation in the rat dentate gyrus following brlcf high frrqucncy stimulation. Brrrin Rex. 126, 361-365. 13. Douglas R. M. and Goddard G. V. (1975) Long-term potentiation of the perforant path-granule cell synapse m the rat hippocampus. Brain Rex. 86, 205- 2 I 5. 14. Duffy C. J.. Teyler T. J. and Shashona V. E. (1981) Long-term potentiation in the hippocampal slice: cvidencc 101 stimulated secretion of newly synthesized proteins. Science, N. Y. 212, I l48- I I5 I. 15. Dunwiddie T. and Lynch G. (1978) Long-term potentiation and depression of synaptic responses In the rat hippocampus: localization and frequency dependency. J. Physiol.. Land. 276, 353-367. 16. Dunwiddic T. V. and Lynch C. (1979) The relationship between extracellular calcium concentrations and the induction of hippocampal long-term potentiation. Brain Res. 169, 103-l IO. 17. Dunwiddie T. V., Madison D. and Lynch G. (1978) Synaptic transmission is required for initiation of long-term potentiation. Brain Res. 150,413-417. IX. Eccles J. C. (1981) The modular operation of the cerebral neocortex considered as the material basis of mental events. Neurosc~iencr 6, 1839-l 856. 19. Ekcrot C. F. and Oscarsson 0. (1981) Prolonged depolarization elicited in Purkinje cell dendrites by climbing fibre impulses in cat. J. Physiol., Lond. 318, 207-221, 20. Fifkovi E. and Anderson C. L. (1981) Stimulation-induced changes in dimensions of stalks of dendritic spmes in the dentate molecular layer. E.xpl Neural. 74, 621-627. 21. Fifkovli E., Anderson C. L.. Young J. J. and van Harreveld A. (1982) Effect of anisomycin on stimulation-induced changes in dendritic spines of dentategranule cells. .I. Neurocytol. 11, 183S210. 22. Fifkovi E. and van Harreveld A. (1977) Long-lasting morphological changes in dendritic spines of dentate granular cells following stimulation of the entorhinal area. /. Neurocytol. 6, 21 I-230. 23. Fifkova E., Markham J. A. and Delay R. J. (1983) Calcium in the spine apparatus of dendritic spines in the dentate molecular layer. Bruin Rex. 266, 163-168. 24. Finn R., Browning M. and Lynch G. (1980) Trifluoperazine inhibits hippocampal long-term potentiation and the phosphorylation of a 40,000 dalton protein. Neurosci. Lefts 19, 103-108. 24a. Frotscher M., Misgeld U. and Nitsch C. (1981) Ultrastructure of mossy fibre endings in vitro hippocampal slices. Eupl. Bruin Res. 41, 247-255. 25. Galvan M., Grafe P. and ten Bruggencate G. (1982) Convulsive actions of 4-aminopyridine on neurons and cxtraccllular K’ and Cal’ activities in guinea-pig olfactory cortex slices. In Physiology und Pharmacology of Epileptogenic Phenomena (eds Klee M. R., Lux H. D. and Speckman E. J.) pp. 353-360. Raven Press, New York. 26. Grab D. J., Carlin R. K. and Siekevitz P. (1980) The presence and functions of calmodulin in the postsynaptic density. Ann. N. Y. Acud. Sci. 356, 55-12. 27. Grav E. G. (1982) Rehabilitating the dendritic soine. Trends Neurosci. 5. 5-6. 28. Gre&ard fi. (19?9) Cyclic nuciotides, phospcorylated proteins and the nervous system. Fedn Proc. Fedn Am. Sacs erp. Biol. 38, 2208.-22 I I. 29. Hagiwara S. and Byerly L. (1981) Calcium channel. A. Rev. Neurosci. 4, 69-125. 29a. Hamlyn L. H. (1962) The fine structure of the mossy fibre endings in the hippocampus of the rabbit. .I. .-1~~(1r. 96, 112-120. 30. Iqbal Z. and Ochs S. (1980) Calmodulin in mammalian nerve. J. Neurobiol. 11, 3 I l-3 18. 3 I. Ito M., Sakurai M. and Tongroach P. (1982) Climbing fibre-induced depression of both mossy fibre responsiveness and glutamate sensitivity of cerebellar Purkinje ceil. J. Physiol., Land. 324, 113-134. 32. Jack J. J. B.. Noble D. and Tsien R. W. (1975) Electric Current Flow in Exciruble Cr/l.c, pp. 218~~223. Clarendon Press, Oxford. 32a. Katz B. and Miledi R. (1968) The role of calcium in neuromuscular facilitation. J. Physiol., Lond. 195,481-492. 33. Kelly P. T. and Cotman C. W. (1978) Synaptic proteins. Characterization of tubulin and actin and identification of a distinct postsynaptic density protein. J. Cell Biol. 79, 173-183. 34. Koch C. and Poggio T. (1983) Electrical properties of dendritic spines. Trends Neurosci. 6, 80-83. 35. Lee K. S. (1983) Sustained modification of neuronal activity in the hippocampus and cerebral cortex. In Moleculur. Cellulur and Behaaioral Neurobiology of the Hippocumpus (ed. Seifert W.) pp. 265-272. Academic Press. New York. 36. Levi-Montalcini R. (1980) New features of the nerve growth factor-target cells interaction. In Nerve Cells, N[,urofrunsmifters and Behauiour (ed. Levi-Montalcini R.) pp. I t-39. Pontificia Academia Scientiarum. Citti del Vaticano. 36a. Levin R. M. and Weiss B. (1977) Binding of trifluoperazine to calcium-dependent activation of cyclic nucleotide phosphodiesterase. Molec. Pharmac. 13, 69&697. 37. Levy W. B. and Steward 0. (1979) Synapses as associative elements in the hippocampal formation. Brain Res. 175, 233 245. 38. Levy W. B. and Steward 0. (1983) Temporal continuity requirements for long-term associative potentiationidepression in the hippocampus. Neuroscience 8, 791-797. dendritic spikes in Avian Purkinje cells. Proc. natn Acud. Sci. 39. Llin2is R. and Hess R. (1976) Tetrodotoxin-resistant U.S.A. 73, 2520-2523. and biochemical effects of high frequency stimulation in 40. Lynch G., Halpain S. and Baudry M. (1983) Structural hippocampus. In Molecular, Cellular and Behavioral Neurobiology of the Hippocampus (ed. Seifert W.) pp. 253 -262. Academic Press, New York. 41. Marr D. (1969) A theory of the cerebellar cortex. J. Physiol., Lond. 202, 437470. 42. Marr D. (1970) A theory for cerebral neocortex. Proc. R. Sot. B. 176, 161-234. 43. Matus A. (1981) The postsynaptic density. Trends Neurosci. 4, 51-53. 44. Matus A., Ackermann M., Pehling G., Byers H. R. and Fujiwara K. (1982) High actin concentrations m brain dendritic spines and postsynaptic densities, Proc. nafn Acad. Sci. L:.S.A. 79, 759@7594.

Long-term potentiation, calcium and memory

1081

45. McNaughton B. L. (1983) Activity-dependent modulation of hippocampal synaptic efficiency: some implications for memory processes. In Molecular, Cellular and Behavioral Neurobiology qf the Hippocampus (ed. Seifert W.) pp. 233-249. Academic Press, New York. 46. McNaughton B. L., Barnes C. A. and Andersen P. (1981) Synaptic efficacy and EPSP summation in granule cells of rat fascia dentata studied in vitro. J. Neurophysiol. 46,952-966: 47. McNaughton B. L., Douglas R. M. and Goddard G. V. (1978) Synaptic enhancement in fascia dentata: cooperativity among coactive afferents. Brain Res. 157,277-293. 48. Means A. R., Tash J. S. and Chafouleas J. G. (1982) Physiological implications of the presence, distribution and regulation of calmodulin in eukaryotic cells. Physiol. Rev. 62, l-39. 48a. Misgeld U., Sarvey J. M. and Klee M. R. (1979) Heterosynaptic postactivation potentiation in hippocampal CA3 neurons: long-term changes of the postsynaptic potentials. Expl Brain Res. 37,217-229. 48b. Nadler J. V., Vaca K. W., White W. F., Lynch G. S. and Cotman C. W. (1976) Aspartate and glutamate as possible transmitters of excitatory hippocampal afferents. Nature, Lond. 260, 538-540. 49. Nieto-Sampedro M., Hoff S. W. and Cotman C. W. (1982) Perforated postsynaptic densities: probable intermediates in synapse turnover. Proc. natn Acad. Sci. U.S.A. 79,5718-5722. 50. Rall W. (1970) Cable properties of dendrites and effects of synaptic location. In Excitatory Synaptic Mechanisms (eds Andersen P. and Jensen J. K. S.) pp. 175-187. Universitetsforlaget, Oslo. 51. Schulman H. and Greengard P. (1978) Stimulation of brain membrane protein phosphorylation by calcium and an endogenous heat-stable protein. Nature, Lond., 271,478%479. 52. Schwartzkroin P. A. and Slawsky M. Probable calcmm spikes in hippocampal neurons. Brain Res. 135, 157-161. 53. Skrede K. K. and Mahhe-Sorenssen D. (1981) Increased resting and evoked release of transmitter following repetitive electrical tetanization in hippocampus: a biochemical correlate to long-lasting synaptic potentiation. Brain Res. 208, 436-441. 54. Squire L. R. (1982) The neuropsychology of human memory. A. Rev. Neurosci. 5,241-273. 55. Steward 0. and Levy W. B. (1982) Preferential localization of polyribosomes under the base of dendritic spines in granule cells of the dentate gyrus. J. Neurosci. 2,284-291.

56. Storm-Mathisen J. (1977) Localization of transmitter candidates in the brain: the hippocampal formation as a model. Proy. Neurobiol. 8, 119-181. 57. Szentagothai J. (1978) The neuron network of the cerebral cortex: a functional interpretation.

Proc. R. Sot. B 201, 219-248. 58. Turner R. W., Baimbridge K. G. and Miller J. J. (1982) Calcium-induced long-term potentiation in the hippocampus. Neuroscience 7, 141 I-1416. 59. Varon S. S. and Bunge R. P. (1978) Trophic mechanisms in the peripheral nervous system. A. Rev. Neurosci. 1,327-361, 59a. Voronin L. L. (1983) Long-term potentiation in the hippocampus. Neuroscience 10, 1051-1069. 60. Vrensen G. and Nunes Cardozo J. (1981) Changes in size and shape of synaptic connections after visual training: an ultrastructural approach of synaptic plasticity. Brain Res. 218,79-97. 61. Vrensen G., Nunes Cardozo J., Miiller L. and Van der Want J. (1980) The presynaptic grid: a new approach. Brain Res.

l&$2340. 62. Wong R. K. S. and Prince D. A. (1978) Participation of calcium spikes during intrinsic burst firing in hippocampal neurons. Brain Res. 159,385-390. 63. Yamamoto C. and Chujo T. (1978) Long-term potentiation in thin hippocampal sections studied by intracellular and extracellular recordings. Expl Neural. S&242-250. (Accepted 15 June 1983)