H. Daniel et al.

Cellular mechanisms of cerebellar LTD CEREBELLUM

Cellular mechanisms of cerebellar LTD

Herv Daniel, Carole Levenes and Francis Crpel

In the past decade there have been advances in understanding the cellular mechanisms of the long-
term depression (LTD) of synaptic transmission at parallel fiberPurkinje cell synapses in the
cerebellum. This review first summarizes current views on mechanisms involved in LTD induction,
from activation of voltage-gated Ca2+ channels, of ionotropic (AMPA) and metabotropic (mGluR1)
glutamate receptors, to stimulation of protein kinase C and nitric oxide formation. Second, we will
focus on recent findings that point towards the involvement of Ca2+ release from internal stores in
LTD induction, localize the sources and targets of nitric oxide and indicate a postsynaptic site
for LTD expression. Finally, a role for LTD in motor learning is now well supported by recent
experiments on transgenic mice.
Trends Neurosci. (1998) 21, 401407

L ONG-LASTING, ACTIVITY-DEPENDENT DEPRES-

SION of synaptic transmission, referred to as long-
term depression (LTD), has been identified at the syn-
apses between parallel fibers (PFs) and Purkinje cells
(PCs) in the cerebellum. Over the past decade, some of
the mechanisms controlling the induction of this form
of synaptic plasticity have been established. It is now
well accepted that the induction of LTD of PF-mediated
responses is triggered by the influx of Ca2+ into PCs
through voltage-gated Ca2+ channels (VGCCs) and by
the activation of two groups of glutamate (Glu) recep-
tors: ionotropic a-amino-3-hydroxy-5-methyl-4-isoxa-
zole propionic acid (AMPA) receptors and type-1 G-
protein-coupled metabotropic receptors (mGluR1). In
addition, there is also wide agreement that second
messenger cascades that occur after the rise of intra-
cellular Ca2+ concentrations involve protein kinase C
(PKC) activation and nitric oxide (NO) production.
This review will focus particularly on more recent
findings concerning the potential involvement of Ca2+
release from internal stores, the sources and targets of
NO, the modification of certain properties of AMPA
receptors during LTD expression, protein synthesis and
transcription factor expression and, finally, the role of
LTD in motor learning.
Initial experiments
Involvement of Ca2+ flow into PCs through VGCCs
Early in vivo experiments suggested that the flow of
Ca2+ through VGCCs was involved in LTD induction
because hyperpolarization evoked by the activation of
stellate cells prevented the occurrence of LTD (Ref. 1),
probably by blocking Ca2+-dependent plateau poten-
tials in PC dendrites following their activation by
climbing fibers (CFs)2. In subsequent experiments in
acute cerebellar slices maintained in vitro, it was also
shown that LTD of synaptic transmission at PFPC
synapses can be induced by conjunctive stimulation
of PFs and CFs impinging on the same PC (Ref. 3), and
that this long-term change in synaptic efficacy no
longer occurs when intracellular Ca2+ in PCs is buf-
fered by 1, 2-bis(2-aminophenoxy)ethane-N,N,N,N-
tetraacetic acid (EGTA). The crucial role of the rise in
Ca2+ resulting from the activation of VGCC in LTD
induction was demonstrated further in cerebellar slices
by showing that LTD of PF-mediated responses is in-
duced consistently by pairing these synaptic responses
with direct depolarization of PCs, giving rise to Ca2+
spike firing, which thus mimics activation of CFs (Refs
4,5) (Fig. 1A). On the other hand, as reported initially
by Crpel and Krupa in acute slices6 and confirmed
later on cultured PCs (Ref. 7), LTD of responses elicited
in these cells by iontophoretic application of Glu in their
dendritic fields is also induced when these responses
are combined with depolarization sufficient to produce
Ca2+ entry through VGCCs (Fig. 1B). Finally, in patch-
clamped PCs in acute slices, recordings of synaptic
currents combined with fluorometric measurements
of the intracellular Ca2+ concentration demonstrated
directly that PF stimulation paired with depolarization-
induced Ca2+ transients are sufficient to induce LTD
(Ref. 8) (Fig. 1C).
Glutamatergic receptors involved in LTD induction
In marked contrast to most other neurons in the
brain, adult PCs do not bear functional N-methyl-D-
aspartate (NMDA) receptors9. Thus, fast excitatory
synaptic transmission at PFPC synapses is mediated
entirely by non-NMDA ionotropic receptors10, most
probably of the AMPA type11. In addition, PFPC syn-
apses also bear mGluR1 (Ref. 12), known to be coupled
to phospholipase C, activation of which leads to pro-
duction of inositol 1,4,5-triphosphate (InsP3) and di-
acylglycerol (DAG) (cited in Ref. 13). It is now well
accepted that the induction of LTD of PF-mediated
responses in PCs in acute slices and of LTD of Glu-
induced currents in cultured PCs requires activation of
these two groups of receptors (Box 1).
Involvement of PKC activation and NO formation in LTD
Ca2+-dependent PKC is expressed abundantly in PCs
(Refs 14,15). As mGluR1 activation results in DAG pro-
duction (see above), it is tempting to postulate that
the cascade of events leading to LTD involves the acti-
vation of PKC by both Ca2+ entry through VGCCs and Herv Daniel,
DAG produced by mGluR activation following Glu Carole Levenes
release from PFs (Ref. 6). Indeed, it is now agreed that and Francis Crpel
cerebellar LTD induction requires PKC activation in are at the Institut
acute slices, as it does in cultures (Box 2). des Neurosciences,
Moreover, in neurons, Ca2+ can induce the formation CNRS-Universit
of NO from arginine by activating calmoduline (CAM)- Paris VI, 75005
dependent NO-synthase (NOS) (cited in Refs 16,17,18). Paris, France.

Copyright 1998, Elsevier Science Ltd. All rights reserved. 0166 - 2236/98/$19.00 PII: S0166-2236(98)01304-6 TINS Vol. 21, No. 9, 1998 401
CEREBELLUM H. Daniel et al. Cellular mechanisms of cerebellar LTD

in synaptic strength. Rather, its

role might be to allow a recruit-
ment of additional synapses within
the pool of those exhibiting LTD
following co-activation of PFs and
CFs, thus increasing the signal-to-
noise ratio of this process.
Recent findings
Role of Ca2+ release from internal stores
Although Ca2+ has been shown
to play a crucial role in cerebellar
LTD, the physiological significance
of Ca2+ release from internal stores
is still a matter of debate. PC den-
drites express two types of intra-
cellular Ca2+ stores: ryanodine- and
InsP3-sensitive stores2125, and it was
hypothesized that the cascade of
events leading to LTD might in-
volve Ca2+ release from these stores,
in addition to Ca2+ entry through
VGCCs (see above). In cultured PCs,
depletion of ryanodine- and InsP3-
sensitive internal Ca2+ stores by
thapsigargin, inhibition of Ca2+ re-
lease from ryanodine-sensitive stores
with ryanodine or ruthenium red,
and inhibition of Ca2+ release from
InsP3-sensitive stores with heparin,
all block LTD induced by conjunc-
tive stimulation with Glu and de-
polarization26,27. However, all these
compounds have several reported
non-specific effects28,29. In acute
cerebellar slices, there is also some
evidence that Ca2+ release from
internal stores plays a role in LTD
induction30,31, but the extent to
which this process is involved seems
to depend crucially upon experimen-
tal conditions. For example, preincu-
bation of slices with thapsigargin
blocks LTD of PF-EPSPs induced by
pairing Ca2+ spike firing with bath
Fig. 1. Role of voltage-gated Ca channels in LTD induction. Diagrams of experimental arrangements are shown on application of a mGluR agonist
2+

the left of each panel. (A) In cerebellar slices, LTD of parallel fiber (PF)-mediated EPSCs is induced by their pairing with (1S-3R ACPD) (without PF stimu-
Purkinje-cell (PC) depolarization sufficient to produce Ca2+ entry through voltage-gated Ca2+ channels. EPSC amplitudes lation), whereas treatment with
are plotted against time before and after the conventional pairing protocol (P). Insets display averaged EPSCs recorded thapsigargin does not prevent in-
during the control period (1) and 15 min after the end of the pairing protocol (2). Adapted from Ref. 5. (B) In cultured duction of LTD by a classic pairing
PCs, conjunctive stimulation with glutamate and depolarization conjunction (P) induces LTD of glutamate currents. protocol (that is, Ca2+ spike firing in
Insets display corresponding traces during the control period (1), 10 min (2) and 70 min (3) after the end of the pair- combination with PF stimulation)30.
ing protocol. Adapted from Ref. 7. (C) Time-course of changes of parallel fiber (PF)-mediated EPSC amplitudes and The situation is still more confus-
intracellular Ca2+ concentration before, during and after induction of LTD by pairing eight depolarizing pulses (10 ms
ing because in other experiments
duration, from 60 to 0 mV) with PF stimulations. Adapted from Ref. 8.
in acute slices, inhibition of Ca2+
release from InsP3-sensitive stores
NO is highly diffusible and its target is probably the with heparin, an antagonist of InsP3 receptors, blocks
soluble guanylate cyclase (cited in Ref. 19) in cells LTD induced by the classic protocol31. Even though in
where it is produced, as well as in surrounding cellular some protocols, Ca2+ release from internal stores might
elements (cited in Refs 17,18). Therefore, this trans- be necessary to induce LTD, it is still unclear whether
duction cascade can finally activate cGMP-dependent Ca2+ release from internal stores combined with PC
protein kinases in PCs, where this enzyme is particu- depolarization is sufficient to induce LTD in the
larly abundant20. Results obtained in acute slices by absence of PF stimulation. Indeed, activation of InsP3
several groups are entirely consistent with a role for NO receptors by photolytic release of InsP3 together with
in LTD induction (Box 2). Because of its great ability depolarization of PCs induces LTD in acute slices31,
to diffuse over large distances, the role of NO in LTD whereas this protocol is ineffective in cultured PCs,
cannot be to bring synapse specificity for this change unless it is combined with AMPA receptor activation27.

402 TINS Vol. 21, No. 9, 1998

 H. Daniel et al. Cellular mechanisms of cerebellar LTD CEREBELLUM

Box 1. Glutamatergic receptors involved in LTD induction

There is wide agreement that induction of cerebellar LTD requires
the activation of two groups of glutamatergic receptors: ionotropic
AMPA receptors and G-protein-coupled metabotropic receptors
(mGluRs).
Consistent with previous in vivo observationsa, the involvement
of AMPA receptors has been demonstrated in whole-cell clamped
Purkinje cells (PCs) in acute slices, as application of CNQX, a selec-
tive antagonist of this group of receptors, prevents induction of LTD
of parallel fiber (PF)-mediated responses by a classic pairing proto-
col (PC depolarization and PF stimulation), but not its expressionb
(Fig. A). Likewise, in whole-cell clamped cultured PCs, induction of
LTD using conjunctive stimulation with quisqualate (an agonist of
both mGluRs and AMPA receptors) and depolarization is blocked by
bath application of CNQX (Ref. c) (Fig. B).
In cultures, application of glutamate, quisqualate or a mixture
of mGluR and AMPA receptor agonists can replace the activation of
PFs for the induction of LTD during a pairing protocol with Ca2+
spike firing, whereas the application of AMPA receptor agonists
alone during the pairing protocol is ineffectivec. Moreover, the
application of mGluR1-inactivating antibodies completely blocks
LTD induced by conjunctive stimulation using Glu and depolariz-
ation in cultured PCs (Ref. d) (Fig. C). Finally, in acute slices, LTD
induction by a classic pairing protocol is impaired significantly in
knockout mice lacking functional mGluR1 (Refs e,f) (Fig. D).
References
a Kano, M. and Kato, M. (1987) Nature 325, 276279
b Hemart, N. et al. (1995) Eur. J. Neurosci. 7, 4553
c Linden, D. et al. (1991) Neuron 7, 8189
d Shigemoto, R. et al. (1994) Neuron 12, 12451255
e Conquet, F. et al. (1994) Nature 372, 237243
f Aiba, A. et al. (1994) Cell 79, 377388

Fig. Involvement of AMPA receptors and

mGluR1 in LTD induction. (A) In cer-
ebellar slices, pairing PF-mediated EPSPs
with depolarization of PCs giving rise to
Ca2+ spikes induces LTD in standard bath
conditions (A1), but not in the presence
of 4 M CNQX (horizontal bar), as dem-
onstrated after wash-out of the drug (A2).
The EPSP amplitudes are plotted against
time before and after the conventional
pairing protocol (P, indicated by thick
bar). Insets display averaged EPSPs at the
indicated times, except inset 2, which illus-
trates the pairing protocol with depolariz-
ing current pulses giving rise to Ca2+ spike
firing, applied in conjunction with PF stimu-
lation. Adapted from Ref. b. (B) In cul-
tured PCs, conjunctive stimulation with
quisqualate and depolarization (P) in the
presence of 20 M CNQX fails to induce
LTD of quisqualate currents, as seen after
washout of the compound. A second
series of conjunctive stimulations (applied
at t = 30 min) in the absence of CNQX
induces LTD. Adapted from Ref. c. (C) In
cultures, PC depolarization in conjunction
with glutamate application induces per-
sistent depression of glutamate currents in
the absence (C1), but not in the presence
(C2) of antibodies inactivating mGluR1.
Inset displays representative single sweeps
of glutamate-induced currents recorded
12 min before (1) and 20 min after (2)
the depolarization (04 min) to 0 mV.
Adapted from Ref. d. (D) In acute slices,
induction of LTD of PF-mediated EPSPs is
impaired significantly in knockout mice
lacking functional mGluR1. Amplitudes
of PF-mediated EPSPs of wild-type (filled
circles) and mGluR1 mutant (empty
squares) PCs are plotted against time
before and after a conventional pairing
protocol (P). Each data point in the graph
represents mean ( SEM) of separate
experiments. Insets on the left and on the
right represent superimposed averaged
EPSPs recorded in wild-type and mutant
mouse PCs, respectively, during the
control period (1) and 15 min after the
end of the pairing protocol (2). Adapted
from Ref. e.

TINS Vol. 21, No. 9, 1998 403

CEREBELLUM H. Daniel et al. Cellular mechanisms of cerebellar LTD

Box 2. Involvement of protein kinase C (PKC) activation and nitric oxide (NO)
formation in LTD

It is now well accepted that both PKC activation and NO formation are required
cellular events leading to cerebellar LTD. In acute slices and in cultures, direct
activation of PKC by phorbol esters induces LTD of the responsiveness of
Purkinje cells (PCs) to exogenously applied glutamate (Glu) or quisqualate,
whereas inactive analogs are without any effecta,b. Moreover in acute cerebellar
slices, the potent protein kinase inhibitor, polymixin B, or the selective PKC-
inhibitor, peptide 1936, almost totally prevents LTD induction by conventional
pairing protocols [PC depolarization and parallel fiber (PF) stimulation]c,d. This
selective inhibitor also blocks LTD induction by conjunctive stimulation using
Glu and depolarization in cultured PCs (Ref. b) (Fig. A).
In acute slices, application of N-monomethylarginine (L-NMMA), a potent
NO synthase (NOS) inhibitor, almost totally prevents induction of LTD of PF-
EPSPs by a conventional pairing protocolc,e (Fig. B), and also prevents long-last-
ing desensitization of AMPA receptors of PCs resulting from successive exposures
of slices to quisqualatef. The use of NO-sensitive electrodes inserted into the mol-
ecular layer of cerebellar slices also demonstrates that protocols used to induce
LTD lead to the release of NO (Ref. g), most probably from PFs (Ref. h). Moreover,
a LTD-like phenomenon, consisting of a long-lasting depression of PF-mediated
responses, is induced by bath application of NO donors or 8-bromo-cGMP, a
membrane permeable cGMP analoge,i. Such a long-lasting depression of PF-EPSPs
is also induced by direct dialysis of cGMP into the recorded PCs through the
recording patch-pipettee. Finally, these NO donor- or cGMP-induced LTD-like
phenomena partially occlude subsequent induction of LTD by classic pairing
protocolse. Thus, all the results obtained in PC in situ are consistent with a role
for NO in LTD induction and establish that its target is probably the soluble
guanylate cyclase of PC.

References
a Crpel, F. and Krupa, M. (1988) Brain Res. 458, 397401
b Linden, D.J. and Connor, J.A. (1991) Science 254, 16561659
c Crpel, F. and Jaillard, D. (1990) NeuroReport 1, 133136
d Hemart, N. et al. (1995) Eur. J. Neurosci. 7, 4553
e Daniel, H. et al. (1993) Eur. J. Neurosci. 5, 10791082
f Ito, M. and Karachot, L. (1990) NeuroReport 1, 129132
g Shibuki, K. and Okada, D. (1991) Nature 349, 326328
h Shibuki, K. and Kimura, S. (1997) J. Physiol. 498, 443452
i Blond, O. et al. (1997) Neuroscience 77, 945954

Fig. Involvement of PKC activation and NO formation in LTD induction. (A) In cultured PCs, conjunctive stimulation with glutamate and depolarization (P)
does not induce LTD of glutamate currents with the following perfusates in the pipette: vehicle plus BAPTA (s, 20 M) or vehicle plus PKC (1936), a selective
PKC inhibitory peptide ( l, 10 M). In contrast, LTD occurs with only vehicle () or vehicle plus [glu27] PKC (1936), a noninhibitory control peptide ( q, 10 M).
Insets: EPSCs at the indicated times. Scale bars, 100 pA, 2 s. Adapted from Ref. b. (B) In acute slices, a conventional pairing protocol (P) which induces LTD of
PF-mediated EPSPs in standard bath conditions (B1), fails to induce LTD in the presence of the NOS inhibitor L-NMMA (30 M) (B2). Insets represent averaged
EPSPs at the indicated times. Adapted from Ref. e.

Altogether, these results provide compelling evidence
that, at least in certain experimental conditions, re-
lease from InsP3-sensitive or ryanodine-sensitive store,
or both, is required for LTD induction. However, these
studies provide only limited evidence for an impor-
tant role for Ca2+ release from these intracellular stores
in LTD induction in more physiological conditions
because, for example, synaptically driven Ca2+ mobil-
ization from intracellular stores by PF stimulation has
not been detected32.
Sources and targets of NO
All experiments in acute slices support the view that
NO is a crucial signal in cellular events leading to LTD
(see above). In the cerebellum, neuronal NOS (nNOS)
has not been identified in PCs, even following reverse
transcriptionpolymerase chain reaction (RTPCR)
analysis of mRNAs harvested directly from these neur-
ons during patch-clamp experiments33. In contrast,
nNOS is expressed at high levels in neighbouring el-
ements such as PFs and basket cells34,35. Given such
localizations of nNOS, it was hypothesized that, follow-
ing a large entry of Ca2+ into PCs during a conventional
pairing protocol, a resulting K+ efflux through Ca2+-
dependent K+ conductances depolarizes neighbouring
presynaptic cellular elements to such an extent that they
can produce a sufficient amount of NO to activate sol-
uble guanylate cyclase in nearby PCs (see above) by a
paracrine effect. This hypothesis is supported by ex-
periments in acute slices showing that raising the extra-
cellular K+ concentration is sufficient to induce a robust
LTD in patch-clamped PCs, and that this LTD is nearly
blocked completely by NOS inhibitors33 (Fig. 2).
Moreover, the fact that LTD induced by a classic pair-
ing protocol is inhibited totally by intracellular injec-
tion into PCs of the selective and potent inhibitor of
soluble guanylate cyclase, 1H-(1,2,4)oxadiazolo(4,3-
a)quinoxalin-1-one (ODQ), confirms that the soluble
guanylate cyclase of PCs is the target of NO (Ref. 36).
According to Ito and Karachot37, production of cGMP
would in turn activate a cGMP-dependent protein

404 TINS Vol. 21, No. 9, 1998

 H. Daniel et al. Cellular mechanisms of cerebellar LTD CEREBELLUM

kinase (PKG), thereby allowing

phosphorylation of its specific en-
dogenous substrate G-substrate,
which is a potent inhibitor of phos- GLU
phatases38. In keeping with this hy- NO-S
pothesis, LTD is induced in acute
slices by bath application of the
protein phosphatase inhibitor, Na+ K+
calyculin A (Ref. 39). Finally, the K +
Depol. mGluR1
conjunctive PF stimulation and de- NO
polarization required for LTD in- Ca2+ Ca2+
AMPA-R Ca 2+
duction can be replaced by photo-
lytic releases of NO and Ca2+ inside G PLC
PCs, suggesting that NO release is
Ca2+
sufficient to replace PF stimulation40. DAG
In addition, this study with caged IP3 LTD
compounds showed that only a PLA2
AA
<10 ms gap between NO release PKC
and Ca2+ elevation is allowed in the
induction of LTD, whereas G-cyclase
cGMP
200300 ms is allowed between IP3 St.
photolytic release of cGMP and Phosphatases
Ca2+ increase40. However, these G-Substrate
results are somewhat puzzling, as
Fig. 2. Schematic diagram of signal transduction processes proposed to participate in LTD induction. The thickness
they contradict previous experi-
of arrows reflects the postulated importance of the various cascades in LTD induction. Abbreviations: AA, arachidonic
ments in acute slices and in cul- acid; AMPA-R, AMPA receptor; cGMP, cyclic guanosine monophosphate; DAG, 1,2-diacylglycerol; G, G protein; G-
tures showing that the activation cyclase, guanylate cyclase; GLU, glutamate; IP3 St., inositol-1,4,5-triphosphate-sensitive internal Ca2+ stores; mGluR1,
of AMPA receptors and of mGluRs type-1 metabotropic glutamate receptor; NO, nitric oxide; NO-S, nitric oxide synthase; PKC, protein kinase C; PLC, phos-
is necessary for LTD induction (see pholipase C; PLA2, phospholipase A2.
previous section). They are also at
variance with a recent voltage-clamp study in acute LTD of Glu-induced responses. Accordingly, further
slices showing that LTD of PF-EPSCs can be achieved experiments using the coefficient of variation (CV)
by pairing these responses with the bath application applied to PF-EPSCs recorded in voltage-clamped PCs
of NO donors, that is, without the depolarization of in acute slices, also suggested that LTD induced by a
postsynaptic cells5. Even though an adequate voltage classic pairing protocol is expressed entirely at the
control of adult PC dendrites is not easy to achieve, postsynaptic level5. A similar result was obtained with
this later study suggests that NO plays the role of CFs, the LTD-like phenomenon induced by NO released into
rather than of PFs, in such pairing protocols. Altogether, the extracellular medium by NO donors5. Moreover,
these discrepancies point out the difficulties of studying in acute slices, a true modification of Glu receptors
physiological processes with pharmacological tools, as during LTD expression has also been shown. Thus, the
they can exaggerate the importance of certain pathways nootropic compound aniracetam, which reduces
in LTD induction, which might be much less important, desensitization of AMPA receptors or decreases the
or not even used, in physiological conditions. closing rate constant for ion channel gating, or both44,45,
Finally, if results obtained in acute slices are all con- has a larger potentiating effect on PF-mediated re-
sistent with a role for NO in the induction of LTD, it sponses during LTD expression than in control condi-
is fair to say that in cultured PCs, Linden and Connor tions46. These results support the view that the change
have demonstrated that the induction of LTD of Glu- in synaptic efficacy involves a genuine change in the
induced currents is unaffected by treatment that functional characteristics of postsynaptic AMPA recep-
stimulates or inhibits NO signaling41. This discrepancy tors (Fig. 2). Moreover, it has also been shown that
with results obtained in acute slices might be because protocols known to induce LTD, such as perfusion of
in cultures putative NO donors, such as PFs and basket slices with 8-bromo-cGMP, dibutyryl-cGMP or the
cells, are very scarce. Alternatively, it is also conceiv- phosphatase inhibitor calyculin A, lead to a persistent
able that mechanisms of LTD induction are different (over 30 minutes) phosphorylation of AMPA receptor
in neurons grown in dissociated cultures versus those subunits in PC dendrites47. However, the latter result is
in slices from young adults. rather surprising because the topology of AMPA recep-
Site of expression of LTD and modification of properties of tors established recently48 shows that the polypeptidic
AMPA receptors of PCs sequences used in this study to prepare polyclonal
In early in vivo experiments, co-activation of PC by antibodies against the phosphorylated form of AMPA
CF stimulation and iontophoretic application of Glu receptor subunits have an extracellular location.
in their dendritic fields induced a long-lasting decrease Therefore, there is still a need to establish more firmly
of their responsiveness to this agonist42. This result led the requirement of the phosphorylation of AMPA
to the proposal that induction of LTD might ultimately receptors during LTD expression.
lead to a long-term desensitization of the postsynaptic Protein synthesis, transcription factors and LTD
ionotropic Glu receptors of PCs. Consistent with this Currently, there is only a small amount of evidence
hypothesis, it was shown both in acute slices6 and in concerning the possible mechanisms of the mainten-
dissociated cultures43, that pairing iontophoretic ance phase of LTD (which could be involved in long-
application of Glu and depolarization of PCs induces term memory). However, recently, it has been shown

TINS Vol. 21, No. 9, 1998 405

CEREBELLUM H. Daniel et al. Cellular mechanisms of cerebellar LTD
clearly in cultured PCs that the establishment of a late
phase of cerebellar LTD (beginning 3045 min after
the induction protocol) requires postsynaptic protein
synthesis, as applications of transcription or translation
inhibitors, immediately, but not 30 minutes after the
induction protocol, attenuate this late phase49.
Moreover, experimental protocols known to induce
LTD also trigger the expression of the immediateearly
genes, c-fos and jun-B, suggesting that the expression
of these genes might help to establish cerebellar long-
term plasticity50,51.
Concluding remarks
Before providing conclusions, it is worth mentioning
other players recently discovered in the cellular mecha-
nisms leading to LTD induction. In cultured PCs, it
has been shown that the increase in the cytosolic Ca2+
concentration required for LTD induction probably
involves activation of a Na+Ca2+ exchanger52. In the
same preparation, it has also been established that the
full activation of PKC required for LTD induction in-
volves activation of the Ca2+-dependent enzyme phos-
pholipase A2 (PLA2), in addition to the intracellular
cascade leading to DAG production53. Finally, in acute
slices, still more recent experiments suggest that protein-
tyrosine kinases (PTKs) are also necessary for LTD in-
duction at PFPC synapses54, probably because they
facilitate receptor-mediated G protein activation55.
On the other hand, the recent development of
genetically engineered mice provides a powerful new
tool to dissect the cellular and molecular mechanisms
involved in the induction of LTD. Thus, by using
mGluR1a knockout mice, the role of mGluR1 in LTD
at PFPC synapses has been established successfully
(see above). More recently, other knockout mice have
been developed with sometimes puzzling results con-
cerning LTD induction. For example, mice with a null
mutation in protein kinase C-g exhibit apparently
normal cerebellar LTD (Ref. 56). However, the authors
have shown that this LTD is still abolished by specific
PKC inhibitors, suggesting that compensatory processes
involving other subtypes of PKC might be activated in
these mutants, to sustain apparently normal LTD. In-
deed, selective expression of a PKC inhibitor in PCs in
transgenic mice leads to a complete blockade of LTD
induction, supporting the hypothesis that activation
of PKC is necessary for LTD induction57. Moreover, the
application of the genetic approach to LTD has
allowed the identification of additional players in this
form of synaptic plasticity. Thus, the involvement in
LTD induction of the newly discovered d2 subunits of
the glutamate receptor family, which are heavily
abundant in PC synapses58, has been shown recently,
as knockout mice devoid of the d2 subunits of Glu
receptors exhibit deficient cerebellar LTD (Ref. 59).
Finally, the fact that mice devoid of glial fibrillary acidic
protein (GFAP), a protein expressed specifically in
astrocytes, show impaired LTD (Ref. 60) suggests that
glial cells might also be involved directly or indirectly
in this form of synaptic plasticity.
In general, our understanding of the cellular and
molecular mechanisms of cerebellar LTD has made
rapid progress during the past decade. Concerning its
possible role in motor learning, recent in vivo experi-
ments point towards a role for LTD in the adaptation
of the vestibuloocular reflex (VOR)61, a conclusion that
has been strengthened by the observations that im-
406 TINS Vol. 21, No. 9, 1998
pairment of LTD by selective expression of PKC in-
hibitors in PCs in transgenic mice is also accompanied
by a lack of adaptation of VOR in these animals57. Such
a correlation between the impairment of LTD and a
deficit in motor learning has also been found consist-
ently in other types of transgenic mice, such as mice
deficient in d2 Glu receptors62, mGluR1 (Ref. 63) or
GFAP (Ref. 60). Thus, it is hoped that a consensus will
also occur soon on the role of cerebellar LTD in motor
learning. In this respect, an important future develop-
ment in the field would be the study of the develop-
mental aspects of LTD in relation to the acquisition of
motor skills.
Selected references
1 Ekerot, C.F. and Kano, M. (1985) Brain Res. 342, 357360
2 Ekerot, C.F. and Oscarsson, O. (1981) J. Physiol. 318,
207221
3 Sakurai, M. (1987) J. Physiol. 394, 463480
4 Crpel, F. and Jaillard, D. (1991) J. Physiol. 432, 123141
5 Blond, O. et al. (1997) Neuroscience 77, 945954
6 Crpel, F. and Krupa, M. (1988) Brain Res. 458, 397401
7 Linden, D.J. et al. (1991) Neuron 7, 8189
8 Konnerth, A., Dreesen, J. and Augustine, G.J. (1992) Proc. Natl.
Acad. Sci. U. S. A. 89, 70517055
9 Crpel, F., Dhanjal, S.S. and Sears, T. A. (1982) J. Physiol. 329,
297317
10 Konnerth, A., Llano, I. and Armstrong, C.M. (1990) Proc. Natl.
Acad. Sci. U. S. A. 87, 26622665
11 Renard, A., Crpel, F. and Audinat, E. (1995) Neuro-
pharmacology 34, 335346
12 Baude, A. et al. (1993) Neuron 11, 771787
13 Pin, J-P. and Duvoisin, R. (1995) Neuropharmacology 34,
126
14 Hidaka, H. et al. (1988) J. Biol. Chem. 263, 45234526
15 Nishizuka, Y. (1986) Science 233, 305311
16 Garthwaite, J., Charles, S.L. and Chess-Williams, R. (1988)
Nature 336, 385388
17 Garthwaite, J., Southam, E. and Anderton, M. (1989)
J. Neurochem. 53, 19521954
18 Ross, C.A., Bredt, D. and Snyder, S.H. (1990) Trends Neurosci.
13, 216222
19 Tremblay, J., Gerzer, R. and Hamet, P. (1988) in Advances in
Second Messengers and Phosphoprotein Research (Greengard, P. and
Robinson, G.A., eds), pp. 319368, Raven Press
20 Lohmann, S.M. et al. (1981) Proc. Natl. Acad. Sci. U. S. A. 78,
653657
21 Ross, C. et al. (1992) Proc. Natl. Acad. Sci. USA 89, 42654269
22 Ellisman, M.H. et al. (1990) Neuron 5, 135146
23 Walton, P. et al. (1991) J. Cell Biol. 113, 11451157
24 Nakanishi, S., Kuwajima, G. and Mikoshiba, K. (1992)
Neurosci. Res. 15, 130142
25 Kuwajima, G. et al. (1992) Neuron 9, 11331142
26 Kohda, K., Inoue, T. and Mikoshiba, K. (1995) J. Neurophysiol.
74, 21842188
27 Kasono, K. and Hirano, T. (1995) NeuroReport 6, 569572
28 Michelangeli, F. et al. (1995) Neuropharmacology 34, 11111122
29 Ehrlich, B.E. et al. (1994) Trends Pharmacol. Sci. 15, 145149
30 Hemart, N. et al. (1995) Eur. J. Neurosci. 7, 4553
31 Khodakhah, K. and Armstrong, C.M. (1997) Proc. Natl. Acad.
Sci. U. S. A. 94, 1400914014
32 Batchelor, A.M. et al. (1996) NeuroReport 7, 19491952
33 Crpel, F. et al. (1994) Neuropharmacology 33, 13991405
34 Bredt, D.S., Hwang, P.M. and Snyder, S.H. (1990) Nature 347,
768770
35 Southam, E., Morris, R. and Garthwaite, J. (1992) Neurosci.
Lett. 137, 241244
36 Boxal, A.R. and Garthwaite, J. (1996) Eur. J. Neurosci.
22092212
37 Ito, M. and Karachot, L. (1992) Neurosci. Res. 14, 2738
38 Nairn, A.C., Hemmings, H.C. and Greengard, P. (1985) Annu.
Rev. Biochem. 54, 931976
39 Ajima, A. and Ito, M. (1995) NeuroReport 6, 297300
40 Lev Ram, V. et al. (1997) Neuron 18, 10251038
41 Linden, D.J. and Connor, J.A. (1992) Eur. J. Neurosci. 4,
1015
42 Ito, M., Sakurai, M. and Tongroach, P. (1982) J. Physiol. 324,
113134
43 Linden, D.J. and Connor, J.A. (1991) Science 254, 16561659
44 Ito, I. et al. (1990) J. Physiol. 424, 533543
45 Vyklicky, L., Patneau, D.K. and Mayer, M.L. (1991) Neuron 7,
971984

46 Hemart, N. et al. (1994) Neurosci. Res. 19, 213221
47 Nakazawa, K., Mikawa, S., Hashikawa, T. and Ito, M. (1995)
Neuron 15, 697709
48 Hollmann, M., Maron, C. and Heineman, S. (1994) Neuron 13,
13311343
49 Linden, D.J. (1996) Neuron 17, 483490
50 Nakazawa, K. et al. (1993) NeuroReport 4, 12751278
51 Yamamori, T., Mikawa, S. and Kado, R. (1995) NeuroReport 6,
793796
52 Linden, D.J., Smeyne, M. and Connor, J.A. (1993) Neuron 11,
120
53 Linden, D. (1995) Neuron 15, 13931401
H. Daniel et al. Cellular mechanisms of cerebellar LTD CEREBELLUM
54 Boxall, A., Lancaster, B. and Garthwaite, J. (1996) Neuron 16,
805813
55 Umemori, H. et al. (1997) Science 276, 18781881
56 Chen, C. et al. (1995) Cell 83, 12331242
57 De Zeeuw, C.I. et al. (1998) Neuron 20, 495508
58 Lomeli, H. et al. (1993) FEBS Lett 315, 318322
59 Kashiwabuchi, N. et al. (1995) Cell 81, 245252
60 Shibuki, K. et al. (1996) Neuron 16, 587599
61 Nagao, S. and Ito, M. (1991) NeuroReport 2, 193196
62 Funabiki, K., Mishina, M. and Hirano, T. (1995) NeuroReport 7,
189192
63 Aiba, A. et al. (1994) Cell 79, 377388

Plasticity of the olivocerebellar pathway

Piergiorgio Strata and Ferdinando Rossi

The adult olivocerebellar axons and their terminal arbours, the climbing fibres, are capable of
remarkable structural plasticity,regulated through their interaction with Purkinje cells.When these
cells are deleted,terminal climbing fibre branches retract.In contrast,there is a vigorous outgrowth
of entire terminal arbours when extra postsynaptic neurones are available.The new connections
lead to a functional, highly specific pattern of innervation at the single Purkinje cell level and are
topographically organized according to the principles of the original projection map.A reversible
climbing fibre retraction occurs following depression of electrical activity of the cerebellar cortex.
These remarkable plastic properties, together with the fact that these neurones express several
growth-associated genes constitutively, suggest that the climbing fibre synapses might be adjusted
dynamically to participate in physiological plasticity.
Trends Neurosci. (1998) 21, 407413

A MAJOR ISSUE in cerebellar function is its involve-

ment in motor learning, according to what is now
called the MarrAlbusIto theory1. This theory assumes
that the olivocerebellar axon, which terminates in the
cerebellar cortex as climbing fibres, is of basic impor-
tance in inducing a long-term depression (LTD) at the
parallel fibrePurkinje cell synapses2,3. This occurs when
coincident signals along the two inputs arrive at the
Purkinje cell. The climbing fibre action is mediated by
an influx of Ca2+ into the dendrites, an event necessary
to induce phosphorylation and, therefore, a reduced sen-
sitivity of the a-amino-3-hydroxy-5-methyl-4-isoxazole
propionic acid (AMPA) receptors4. In this form of plas-
ticity, the role attributed to the climbing-fibre synapse
is the generation of precisely timed Ca2+ transients in
Purkinje dendrites, and the possibility that the olivo-
cerebellar input itself might undergo modifications of
synaptic efficacy or structural remodelling has never
been considered. Here, we provide evidence that such
a synapse presents a high degree of remodelling in re-
sponse to experimental manipulation. In addition to its
role in postlesional plasticity, such an ability for struc-
tural modification might be important for physiological
processes.
Morphology of the climbing fibrePurkinje cell
synapse
In the rat, the inferior olive contains about 48 000
cells. Each inferior olive neurone innervates an average
of seven Purkinje cells, forming an extensive terminal
arbour, the climbing fibre, which abuts on to the prox-
imal dendrites of a single target cell5. Each arbour has
an average of 111 branches, organized in four orders, for
a total length of 1535 mm, bearing 288 varicosities6. Each
varicosity makes 16 synaptic contacts with stubby
spines on the primary, secondary and tertiary dendrites7.
As shown in Fig. 1 and Table 1, the majority of branches
are in the two most distal orders, where 88% of the syn-
apses are located6. On the other hand, 200 000 parallel
fibrePurkinje cell synapses are present in the distal den-
dritic compartment, made by the spiny branchlets9.
Target deletion leads to climbing fibre atrophy in
its distal compartment
During development, the survival and maturation of
neurones both depend on target-derived support10,11.
In the adult, axontarget interactions are important to
maintain the neuronal phenotype and regulate plastic
processes12,13. In fact, in the mature brain, target depri-
vation severely affects axonal arbours, which become
atrophic, but remain for a long time in their terminal
fields1416. Conversely, an increase in the extent of
available target space could elicit growth and synapto-
genesis by afferent axons17,18. Because of its extensive
terminal territory covering a large surface of a single Piergiorgio Strata
target neurone and the unique electrophysiological and Ferdinando
properties, the climbing fibre provides an ideal model Rossi are at the
to study the remodelling of central axons following Department of
different manipulations. Therefore, in adult rats, we have Neuroscience,
studied changes of the climbing fibre when its target University of
has been deleted and when a new target space is made Turin, C.so
available. Raffaello 30,
Following Purkinje-cell degeneration induced by I-10125 Torino,
intracerebellar injections of kainic acid or propidium Italy.

Copyright 1998, Elsevier Science Ltd. All rights reserved. 0166 - 2236/98/$19.00 PII: S0166-2236(98)01305-8 TINS Vol. 21, No. 9, 1998 407