The Cerebellum: Brain for an Implicit Self (40 page)

Schizophrenic patients were shown to be significantly more impaired than normal control subjects in adapting to prism distortion. They also had significantly greater difficulties in the reorientation that followed removal of the prisms (
Bigelow et al., 2006
). It seems likely that the cerebellar disorders of these patients affected both their voluntary movements (tested with prism-adaptation tests) and their mental processes, the latter being manifested as symptoms of schizophrenia. Hallucination and “passivity” are unique symptoms of schizophrenia, which can be explained as arising from the failure of a mental model and its copy in a cerebellar internal model to provide appropriate internal feedback to the prefrontal cortex (
Blakemore et al., 2000
). Hallucination includes the hearing of unreal spoken speech, which may result in the patient not recognizing that the unreal speech was produced internally. Passivity is a patient’s feeling that her/his will is being replaced by that of some other force or agent. This feeling might arise as a result of a lack of awareness of intended actions, which might be due to impairments of mental models and their copies in cerebellar internal models. This would result in a loss of prediction of the possible outcome of an intended action. Hence, an aberrant self-monitoring mechanism involving the cerebellum, as discussed in
Chapter 16
in connection with alien control, appears to apply also to mechanisms underlying hallucination and passivity in schizophrenia.

Neuroimaging data suggesting cognitive involvement of the cerebellum ever increase, but one should be aware of a caution such that the observed activities might be contaminated by eye movements that often accompany cognitive tasks (
Glickstein et al., 2009
). It is also important to realize that the internal model-based control of mental activities presented above is still a conceptual hypothesis based on psychological aspects of Craik’s mental model and Piaget’s schema. Neuroimaging has made plausible that (1) activation of certain regions of the temporoparietal cortex is the neural substrate of a mental model or schema, and (2) activation of the cerebellum is an indication of the involvement of cerebellar internal models. This hypothesis is also supported to some degree by other experimental and clinicopathological data. It must be admitted, however, that further testing of the hypothesis is challenged by difficulties arising from current technical limitations in human experimentation and the lack of relevant computational support.

Neuronal circuits in the cerebellum have now been decomposed to the extent that their various neuronal components are defined and synaptic plasticity (LTD/LTP) is as memory elements in neuronal networks; that is, at least for the early phase of memory processes (
Chapters 3
–
7
). Detailed signal flow charts underlying synaptic plasticity have been revealed, as have been some genetic and pharmacological means to manipulate this plasticity. Reconstruction of the cerebellar machine has also advanced markedly (
Chapter 9
, “
Network Models
”). Three-layered neuronal networks like a simple perceptron, which is based on the design of neuronal circuits, have been proposed as a model of the cerebellum that is capable of spatial pattern recognition. In addition, adaptive filter models are capable of temporal pattern recognition (
Chapter 9
). Real cerebellar networks integrate these two capabilities to major extent, and the networks have been simulated using liquid state machine models. Circuits have also been modeled for error learning during repeated attempts to execute a task. Experimentally-based knowledge about neuronal networks in the cerebellum has also become increasingly more detailed. For example, Lugaro cells and small inhibitory neurons have been added to the neuronal wiring diagram of the cerebellar cortical network, albeit their functional roles are still unclear. Also, many different neuropeptides and amines conveyed by beaded fibers to the cerebellum from the hypothalamus have been identified, and their functional roles are now being considered in terms of neuromodulation. Recently, it has become clear that the cerebellar and vestibular nuclei act as a memory site and as such, are a complement to the memory role of the cerebellar cortex. It will be intriguing to see how this ever-emerging new knowledge will modify and expand our present models of cerebellar function.

A microcomplex has been conceived as a modular unit of cerebellar neuronal circuits. It incorporates a microzone of the cerebellar cortex and a small group of cerebellar and/or vestibular nuclear neurons that are attached to a small group of inferior olive (IO) neurons and that of parvocellular red nuclear neurons (
Chapter 9
). Such a microcomplex is equipped with two types of memory (cortical and nuclear), which have complementary roles in learning. Microcomplex modules in the cerebellum seem capable of mimicking the input-output or output-input relationship of other neuronal systems in the CNS by use of an error learning mechanism. Modeling has suggested that discrepancies between the outputs of a microcomplex and the circuit it is copying can be obviated progressively throughout repeated trials. This is thought to be accomplished in main part by a change in the properties of the microcomplex brought on by conjunctive LTD in its Purkinje cells and associated LTP/LTD at mossy fiber-nuclear neuron synapses. In this way, a cerebellar internal model can be expected to simulate a controlled object or its inverse. This internal model concept is now being tested by unit recording from Purkinje cells (and hopefully also from nuclear neurons) during various forms of voluntary movement, and it is also being applied successfully in the field of robotics. Its application to complex motor actions and cognitive functions is still a matter of conceptual modeling, but it certainly points the way to future research.

Despite the above progress, a number of crucial questions, specific and general, and conceptual and technical, remain unanswered at all levels of analysis: that is, at the molecular, cellular, circuitry, and behavioral levels. The major purpose of this book is to collect so-far-uncovered facts and answered and unanswered questions in research on the cerebellum and then assort them according to systems control principles. Now, at the end of this exercise, it behooves me to ask seven questions, the answers to which are my opinion of what should guide the next phase of research on the cerebellum.

1.
Conjunctive LTD as synaptic plasticity.
The signal transduction properties of this synaptic plasticity have been analyzed extensively. Nevertheless, there could well be more as-yet-unknown molecules and processes to be discovered and their functional role determined. For example, we have shown recently that prostaglandin D
₂
and E
₂
play a crucial role at the stage of LTD induction when PKCα phosphorylates AMPA receptors. These receptors are then severed from the cytoskeleton and removed from the stable synaptic pool of AMPA receptors (
Figure 20
). However, just how prostaglandins D
₂
/E
₂
act on AMPA receptors is not yet known (
Le et al., 2010
).

The most intriguing question about conjunctive LTD is the mechanism(s) by which its manifestation leads to more persistent memory-like processes. Synapses undergoing conjunctive LTD do not cause an immediate change in the shape of the dendritic spines on Purkinje cells, but the functional state of these synapses seems to be transformed sometimes into silence or even further, the synapses disappear (
Chapter 8
, “
Multiplicity and Persistency of Synaptic Plasticity
,”
Section 6
). An open question is just how LTD links to such synapse liability.

A relevant enduring issue is the length of time that LTD can be maintained (
Chapter 7
, “
Conjunctive Long-Term Depression (LTD)
”). Studies on the VOR and OKR have shown that short-term adaptation recovers in 24 hours. This duration seems to be unduly short for a memory formation in the usual sense. Nonetheless, there has been a tendency to generalize this duration to conjunctive LTD. It seems possible, however, that this 24-hour value for VOR/OKR adaptation is due, at least in part, to a relearning process, the goal of which is to reach the pre-adaptation level (e.g., like extinction of eye-blink conditioning). Climbing fiber discharge, which indeed occurs while animals are resting in the dark, could drive such a relearning mechanism (
Chapter 10
, “
Ocular Reflexes
,”
Section 3
). If such complications are avoided in some way, the VOR/OKR may continue to be a useful test paradigm for learning and memory mechanisms of the cerebellum.

2.
Conjunctive LTD versus motor learning.
A popular view is that the appearance of conjunctive LTD implies the formation of a memory trace, which is acquired during motor learning. This idea is usually based on the experimental finding of a coincident failure of conjunctive LTD and motor learning in response to genetic and/or pharmacological manipulations. LTD is usually observed in intracellular recording undertaken in tissue culture and slice preparations, or
in vivo
but under general anesthesia, and using synchronize electric shock stimuli. In contrast, motor learning is usually tested for by observing changes in an animal’s behavior under natural stimulating conditions. Welsh et al. (
2005
) challenged this experimental dichotomy in coincidence testing. In a study of anesthetized in vivo rat preparations, they confirmed that LTD detected in extracellular field recordings was blocked when a pharmacological agent (T-588, a blocker of calcium release from intracellular stores,
Kimura et al., 2005
) was continuously infused intravenously into rats. They also reported that awake rats administered perorally with T-588 exhibited normal eye blink conditioning even though the concentration of T-588 in the brain reached the level that was attained by intravenous infusion. This might suggest that LTD was not needed for motor learning but rather, its possible role was to protect the animal from excitotoxicity. In the above study,
however, there was a discrepancy between the experimental conditions for the manifestation of LTD and motor learning. The presence of the former was shown in anesthetized rats while they were receiving a continuous intravenous infusion of T-588, with the LTD evoked by electric pulse stimuli. In contrast the manifestation of motor learning was shown in awake rats, who received orally administered T-588 during their reactions to natural stimuli. In the latter case, there was no definitive proof that LTD was really blocked when the animals were exhibiting eye blink conditioning. Similar reservations apply to a recent report that three mutant mice with an internalization of AMPA receptors lacked conjunctive LTD, even though VOR adaptation and eye blink conditioning occurred in a seemingly normal fashion (
Schonewille et al., 2011
).

The above cited observations may compel us to reconsider the present LTD-based hypothesis of motor learning and attempt to formulate a new hypothesis. Before doing so, however, it deserves emphasis that none of the present studies have shown the lack or presence of LTD when the tested animals were undergoing motor learning. An unequivocal test would be to monitor LTD in behaving animals. To date, such an inference has been made (
Gilbert and Thach, 1977
;
Ojakangas and Ebner, 1992
;
Medina and Lisberger, 2008
) but not with the definitive observation of conjunctive LTD appearing during the learning process.

3.
Function of the IO.
Climbing fibers originating from the IO are a unique structure of the cerebellum and have been assigned the role of supervising teacher for the learning process in the cerebellar cortex (
Chapter 3
, “
The Cerebellum as a Neuronal Machine
”). The intricate network structure interconnected via electrical synapses in the IO generates highly regular rhythmic discharge under certain conditions, but stochastic, low-rate discharge under other conditions. This appears to imply a subtle mechanism for IO neurons to generate error signals consistent with the conditions to which the animals are subjected (
Chapter 6
, “
Pre- and Post-Cerebellar Cortex Neurons
,”
Section 3
). However, a contrasting interpretation has been put forward that the olivocerebellar system is a sort of clock for generating temporal patterns (Yarom and Cohen, 2002). These authors speculated that the olivocerebellar system generates a large repertoire of temporal patterns. When a specific temporal pattern is required, a group of IO neurons are assembled by pausing the activity in the cerebellar nuclear neurons and thereby removing the inhibition from appropriate gap junctions between IO neurons. The so-disinhibited IO ensemble begins oscillatory discharge, and thereby generates a temporal pattern that is uniquely suited to the task at hand. This hypothesis needs further testing, however.

4.
Nuclear memory.
Although a memory process is now known to be formed in vestibular and interpositus nuclear neurons, such a finding has not yet been reported for dentate neurons, which receive collaterals of pontocerebellar mossy fibers. Because these collaterals are so meager (
Chapter 6
), however, it is far from certain that an equally effective memory site could be located in the dentate nucleus. Dentate neurons also receive collaterals from mossy fibers originating from the nucleus reticularis tegmenti pontis. How these two pathways to dentate neurons contribute to nuclear memory is an interesting question for future studies.