and DCNCs must affect synaptic efficacy. Thus, the character of simultaneous modification of
synapses at different layers of cerebellar network depends on the presence of a signal from IO.
Proposed long-term bidirectional changes in the efficacy of diverse cerebellar synapses are summa-
rized in Table 3. In turn, the activity of neurons in IO is influenced by cerebellar cells Fig. 1. Thus,
DCNCis inhibit olivary neurones, while DCNCes interpositus, for example excite olivary cells via
mesodiencefalic nuclei red nucleus, for example. Cells of red nucleus activate also DCNCes, form-
ing a reverberation loop and supporting modifica- tion of synapses on DCNCes.
A signal from IO is considered as a training signal. It could teach PCs to recognize specific
pattern of input signals Marr, 1969 or train PCs to reduce error Albus, 1971. With respect to
conventional opinion, the rise of output cerebellar signal requires CF activity, since this rise is the
consequence of associative LTDe in the PF – PC pathway and disinhibition of DCNCes. However,
such an effect is inconsistent with experimentally found data that activity of olivary cells is inhib-
ited during the reaching and all phases of the behavior Horn et al., 1998. Unlike, according to
our model, induction of LTPe in PF – PC synapses that occurs in the absence of a CF activity, this
could cause an increase of inhibition of DCNCes, and advance LTP of excitatory inputs from MFs
to DCNCes. Thus, a potentiation of output cere- bellar signal increase in the rate of DCNCes
discharges could be induced only during the si- lence of olivary neurones. This result is in a good
accordance with the aforementioned experimental data Horn et al., 1998.
Neural networks with numerous interconnected inhibitory cells have been implicated in the gener-
ation of high-frequency oscillations Whittington et al. 1995. The olivary-cerebellar network in-
cludes different reciprocally connected inhibitory neurones Fig. 1. Frequency of oscillations pro-
duced by this network tends to shift to lower frequencies after an inhibition of GABAa recep-
tors on IO cells decrease of inhibition of IO cells Lang et al., 1996. This result leads us to the
assumption that external stimulation andor DC- NCis discharges that cause inhibition of IO cells
could result in a rise in the frequency of oscillations.
5. The comparison with known experimental data
The necessity of two distinctive modifiable synapses was recently proposed in some models of
cerebellar learning Fiala et al., 1996; Kenyon, 1997; Raymond and Lisberger, 1997; Thompson
et al. 1997. The modification of different cerebel- lar synapses was revealed in only a few experi-
ments. LTPe was found in the MF – GC pathway Racine et al., 1986; D’Angelo et al., 1999 and
the MF – DCNC pathway Rossi et al., 1996. LTDi was demonstrated in PC – DCNC synapses
Table 3 The influence of olivary signal on simultaneous modification of cerebellar synapses
Synaptic inputs Postsynaptic cyclic nucleotide
The character of modification CF signal presence
CF signal absence Ca
2+
B Ca
2+
o Ca
2+
\ Ca
2+
o Ca
2+
B Ca
2+
o Ca
2+
\ Ca
2+
o cAMP
LTPe MF Golgi cell
LTDe cAMP
LTPe PF Golgi cell
LTDe cGMP
MF Granule cell LTPe
LTDe cGMP
LTDi LTPi
Golgi Granule cell PF PC
LTDe cGMP
LTPe LTPi
PC DCNCe LTDi
cGMP LTPe
LTDe MF DCNCe
cGMP
Morishita and Sastry, 1993, and Golgi cell – GC synapses Robello et al., 1996; Amico et al., 1998.
The character of synaptic modification is in the accordance with the suggested mechanism of cere-
bellar plasticity. Thus, LTDi in PC – DCNC synapses was found in the absence of excitation
Morishita and Sastry, 1993. Owing to the ab- sence of glutamate, mGlu receptors could not be
activated, and PKC was inactive. Neither depolar- ization of DCNC nor intracellular Ca
2 +
rise was observed in this study. In the absence of Ca
2 +
ions, the activation of protein phosphatase 1 and CaMKII must be excluded Fig. 2. NO synthase
is absent in PC axon terminals Ross et al., 1990, therefore it cannot participate in cGMP produc-
tion. However, cGMP levels could increase due to the action of GABA at GABAb receptors, and
LTDi could be induced as a result of phosphory- lation of GABAa receptors by PKG.
In terms of our model, experimentally observed LTPe in MF – GC synapses Racine et al., 1986
and LTDi in Golgi cell – GC synapses Amico et al., 1998 indicates that inhibitory action of Golgi
cell at GC is usually strong. This inhibition pro- motes activation of PKG, and phosphorylation of
AMPA and GABAa receptors on GC. Subse- quent long-term strengthening of an input signal
by GC can be restricted by disinhibitory action of the Lugaro cell that inhibits the Golgi cell. It
follows from our model that LTPe in MF – GC synapses can be also induced in the absence of
inhibition from Golgi cell. In this case, LTPe can be the result of activation of mGlu receptors and
phosphorylation of AMPA receptors by PKC, while PKG is inactive. Such an effect has been
obtained in GCs after MF stimulation that caused excitatory postsynaptic potential EPSP but not
discharges of GCs D’Angelo et al., 1999. Possi- bly, this stimulation of MFs has also been too
weak to fire Golgi cells directly.
6. Conclusion