Post-training infusions of muscimol into the cerebellar cortex prevent the
consolidation of conditioned NMR. Comparable infusions into the deep nuclei have no such effect. These findings dissociate the function of the cerebellar cortex and the deep nuclei in learning for the first time. It had previously proved impossible to separate the contributions of elements of the olivo-cortico-nuclear loop to motor
learning. Targeting consolidation in the cerebellum was an untested approach to alleviating the experimental problems posed by the loop. This study has therefore been at least partially successful in its aims because consolidation has been affected by interference at one point in the loop but not at another. Despite this overall success the results are open to many interpretations. Careful consideration of the results and further work is required before concluding that the cerebellar cortex is the site of motor learning and that the deep nuclei do not store information related to conditioned eyeblinks. We can appreciate the breadth of possible interpretations if we consider muscimol’s potential mode of action on consolidation.
How does muscimol in the cerebellar cortex prevent consolidation?
Muscimol is clearly a G A B Aa receptor agonist of some potency. The dampening
effect of muscimol infusions on previously acquired CRs in phase 5 can last for between four and eight hours. It is not clear whether the receptor agonism is itself maintained over this period or whether muscimol binding has short-term
physiological consequences for the cells to which it binds. By testing with 2-
deoxyglucose uptake, the lasting effects of muscimol have previously been attributed to a metabolic interference (Martin, 1991), although it is not clear why agonism of an ionotropic membrane receptor such as the G A B Aa receptor should lead to general
alterations in metabolism inside the cell. Possibly, second messenger systems are affected by muscimol binding.
Muscimol competes with G ABA by binding to the G ABA binding site itself. In so doing it gates an anion channel that predominantly allows chloride ions [Cf] to flow into the cell to which it is bound. The main consequence of such an influx might be expected to be a hyper-polarisation of the cell membrane. As a result of this change epsps should become less likely to contribute to an action potential. Muscimol could therefore potentially affect consolidation in two ways. It may either impair the
internal cell mechanisms of memory storage through metabolic interference or it may prevent perseveration of activity within the system by inhibiting crucial circuit
components. Interpretation of the results presented in this thesis may differ depending on how muscimol action is considered to affect consolidation processes. Which cerebellar component the results implicate as the site of memory storage depends very
much on whether we consider consolidation to rely upon continued activity in the system or not.
Muscimol is likely to act at synapses between inhibitory interneurons and the other cell types
We could assume that muscimol acts on consolidation by affecting elements in the cerebellar cortex because there is no clear effect on consolidation from deep nuclear muscimol infusions. Muscimol must bind in the cortex to post-synaptic receptors at synapses between inhibitory intemeurons and other cortical elements. Intemeurons include the large Golgi cell, the soma of which can be found in the granule cell layer, and the basket and stellate cells, which respectively occupy inner and outer fields of the molecular layer (see tig 1.06). These three cell types are known to be inhibitory (Andersen et al., 1964) (Eccles et al., 1966a), and thought to be GABAergic
(Kawamura and Provini, 1970) (Curtis and Felix, 1971), (Dupont et al., 1979), although stellate cells have been suggested to be taurinergic (McBride and
Frederickson, 1980). The stellate and basket cells receive input from parallel fibres and contact Purkinje cells with relatively short axons. The stellate cell, lying within the outer molecular layer, contacts the outer part o f the Purkinje cell dendritic arborisation, and the basket cell contacts both the inner dendrites and the soma itself of the Purkinje cell. Muscimol is therefore likely to bind at a large number of widely distributed sites across the Purkinje cell.
The Golgi cell projects to a specialised unit found within the granular layer. This is the cerebellar glomerulus and consists of one of twenty to thirty swellings spaced along the mossy fibre length, called rosettes, surrounding granule cell dendrites and Golgi cell axons (Ramon y Cajal, 1995) (Eccles et al., 1967). Like the molecular layer intemeurons, the Golgi cell receives input from parallel fibres, and thereby seems to provide feedback to the granule cells (Eccles et al., 1966b). It has been suggested that this cell buffers mossy fibre input to granule cells (Marr, 1969) (Tyrrell and
Willshaw, 1992). Muscimol is likely to bind at these sites and thereby inhibit granule cells. Both granule cell and Purkinje cell inhibition could affect intra-cortical
Is continued neural activity required for consolidation?
It may be the case that the first stages of motor memory encoding occur just as suggested by the Marr/Albus model, but changes at the parallel fibre-Purkinje cell synapse may then require rehearsal in order to be made permanent. For instance, repeated parallel fibre input may serve to prevent dephosphorylation o f receptors, and thereby perpetuate temporary changes induced during learning until more permanent alterations can be made (see LTD section in the final discussion). Reduction of parallel fibre input to Purkinje cells by muscimol-mediated inhibition of granule cells would prevent such changes from being rehearsed. Purkinje cell inhibition by
muscimol may also prevent putative rehearsal processes due to reduced synaptic transmission. Alternatively, rehearsal of changes that may have occurred at mossy fibre-granule cell synapses could equally well be prevented by muscimol, because inhibition of granule cells would prevent synaptic transmission here.
Another possibility is that already encoded memories require subsequent modulatory input if they are to be stored long-term. This modulation could take the form of monoaminergic input, as has previously been included in models o f cerebellar learning (Gilbert, 1974), or GABAergic input itself. Monoaminergic reinforcing signals to either Purkinje cells or granule cells during the post-training period may be affected by membrane hyperpolarisation. Monoamine and G ABA interactions have already been suggested to be important in memory formation (Mitoma and Konishi,
1999). GABAergic modulation could also be directly affected by introducing
muscimol into the system because muscimol would compete successfully with G ABA for crucial binding sites.
That GABA can act to modulate memory storage is implied not only by the fact that agonism o f GABA receptors by muscimol can impair memory formation, as
mentioned in the introduction to this chapter, but also that antagonism o f GABA receptors with picrotoxin can enhance the formation of various memories (McGaugh et al., 1990) (Castellano and McGaugh, 1989) (Brioni and McGaugh, 1988). Further consideration of the relative merits of these different candidate storage mechanisms will be included later in the thesis, but for the purposes of this discussion it is clear that muscimol could be affecting consolidation processes intrinsic to the cerebellar
cortex in several ways. The simple interpretation of the results described in this
chapter is that the storage of essential components of the classically conditioned NMR occurs somewhere within cerebellar cortical circuitry. This interpretation is perhaps overly simple, however, because muscimol action in the cortex could also be held responsible for failure o f memory storage in the deep nuclei.
Could cerebellar cortical muscimol prevent consolidation through secondary effects on the deep nuclei?
Inactivations of different components of the olivo-cortico-nuclear loop during
classical eyeblink conditioning have been shown to prevent acquisition o f CRs (Clark et al., 1992) (Krupa and Thompson, 1997) (Hardiman et al., 1996) (Welsh and
Harvey, 1998) (Attwell et al., 2001). The results of these studies taken collectively contrast with the consolidation experiment described in this chapter in that acquisition is prevented by cortical and nuclear inactivation whereas consolidation is affected by cortical inactivation only. It must be assumed, therefore, that consolidation does not rely upon the same processes as acquisition. Straightforward perseveration o f activity within the olivo-cortico-nuclear loop cannot be considered a mechanism of memory storage because disruption of activity in the deep nuclei, a component of the loop, is not sufficient to prevent consolidation.
Perseverative processes cannot, however, be completely ruled out as a means of consolidating motor learning. There are obvious ways in which cortical inactivations could influence putative nuclear storage mechanisms. Information encoded in the cortex could be shifted by consolidation to the deep nuclei in the manner that has been proposed by some authors for the consolidation of episodic memory (McClelland et al., 1995) (Buzsaki, 1998). Alternatively, memory could be encoded in the deep nuclei and consolidated by a modulatory signal from the cortex. Either one of these two processes would be predicted to be interfered with by cortical muscimol infusions.
It should be remembered, though, that Purkinje cells are themselves GABAergic (Obata et al., 1970) (Ito et al., 1970). Neurons in the vestibular nuclei, an efferent target of Purkinje cells, are hyperpolarized by GABA application (Obata et al., 1967).
Glutamic acid decarboxylase (GAD), the enzyme that converts glutamate to GABA, is found to be greatly reduced in the interpositus nucleus by cerebellar cortical lesions that destroy Purkinje cells (Fonnum et a l, 1970) and GAD activity is very high in the axon terminals of Purkinje cells. Muscimol would therefore be expected to act in the deep nuclei at Purkinje cell synapses, among others, and greatly inhibit deep nuclear neurons by hyperpolarising their cell membranes. In theory, this action could mimic that of GABA in the normal functioning system. It would be surprising, however, if the specificity of input required for shifting encoded information from the cortex to the nuclei could be maintained through a broad application of a receptor agonist. Any putative deep nuclear consolidation processes cannot be dependent upon a uniquely patterned cortico-nuclear signal, because deep nuclear muscimol would surely have strongly affected such patterning.
Although a slight trend towards improved learning is seen in the DN group, it would also be surprising if memory consolidation in the deep nuclei is aided by an inhibitory input, whether this be mediated by GABA in the normally functioning cerebellum or mimicked by muscimol in the case o f this experiment. The literature strongly suggests that the GABAergic system acts in a modulatory capacity throughout the brain, but only to reduce memory storage ((Nabeshima et al., 1988) (Brioni et al., 1989) and see (Izquierdo and Medina, 1991) for review).
The impairment in consolidation of classical NMR seen in the CTX group seems likely to have resulted from a relatively localised action of muscimol on consolidatory mechanisms within the cortex. The dissociation of effect seen in these experiments is not consistent with any interpretation that posits a role for post-training electrical signalling from the cortex to the nuclei in consolidation. Any such signalling would surely be prevented by deep nuclear as well as cortical infusions o f muscimol. The results of these experiments therefore strongly support the view that there is an intra- cortical consolidation process. Whether muscimol directly affects intracellular storage processes or instead manipulates intercellular perseverative processes within the cerebellar cortex is not at all clear.
Do post-training muscimol infusions into the cerebellar cortex completely prevent consolidation?
The storage of some major component of memory relating to NMR conditioning seems to occur within the cerebellar cortex. However, the possibility that this is accompanied by storage of other components elsewhere cannot be ruled out. The deficit in consolidation caused by post-training muscimol infusions in the cortex is profound but perhaps not complete. Although the statistical methods employed in this chapter do not reveal any consolidation in the CTX group under the influence of muscimol it appears that there is a trend towards faster acquisition in this group during phase 2 than is seen in the CON group during phase 1. There is no significant difference between the percentage of trials on which a CR is produced in phase 2 of the CTX group and phase 1 of both the CON and DN group.
Unfortunately, the raw data from this experiment was not normally distributed due to the large number of 0 scores on the first days of training and asymptotic scores for most of the other days. Only during a few intervening sessions did subjects produce intermediate scores. This raw data cannot be converted to normality in any
conventional way so non-parametric statistical tests have been used. The power of these tests is not as great as parametric tests and no sense of the progression in learning can be gleaned from them as it might have been with the repeated measures design available within parametric tests. Consequently, relatively crude statistical analyses o f CR percentages overall, within each phase, have been compared between groups. This approach is correct, being highly unlikely to find significant differences where in reality they do not exist, and is therefore greatly preferable to the use of parametric tests on data that is not normally distributed. However, the lack of power in non-parametric tests means that they will possibly fail to detect significant
differences that in reality exist between the groups. The trend towards fast learning relative to controls seen in the CTX group during phase 2 may therefore reflect the fact that these subjects were not fully impaired on consolidation during phase 1.
There are a number of reasons why the effect on consolidation could have been incomplete. The muscimol may not be fully effective in obstructing whatever mechanism mediates consolidation in the cortex. Perhaps because consolidation can
occur unhindered by anything other than competing presentations of the stimuli during the training period itself. The subjects may have become habituated to the conditioning apparatus during phase 1 and thereby have an advantage in phase 2 over less habituated controls in phase 1. Subjects may also be more alert during phase 2 due to sensitisation in phase 1. Finally, it should be considered that some aspect o f the conditioning may rely upon memory storage that occurs elsewhere, and that this will be unhindered by muscimol in the cerebellar cortex.
Another non-significant trend that is consistent with this latter interpretation is seen in the acquisition rate of the DN subjects. These subjects appear to leam more quickly than the CON group. It could be, therefore, that consolidation in these subjects is enhanced by muscimol action in the deep nuclei, which may be consistent with there being some memory storage at this site. Alternatively, muscimol in the nuclei may disinhibit the inferior olive, increase activity in the Purkinje cells via climbing fibres and thereby in some way facilitate consolidation in the cortex. The most likely explanation has nothing to do with consolidation. Cannulation in the cortex is more likely to cause damage to eyeblink circuitry. The DN group may therefore have learned at a rate similar to naïve uncannulated animals, whereas the CON group may have been slightly retarded by damage.
Summary
Overall, the success of the work presented in this chapter has been to identify a process or processes of memory consolidation that is sensitive to muscimol invasion in the cerebellar cortex but not in the deep nuclei. This is the first study to succeed in separating the function of these two structures in memory formation. Furthermore, infusions of muscimol effective in preventing consolidation are consistently localised to an area of lobule HVI that has previously been implicated in NMR conditioning. The study described in the next chapter further investigates the sensitivity o f cortically dependent consolidation processes and in so doing characterises a time-course for the consolidation of NMR conditioning.