The Cerebellum: Brain for an Implicit Self (10 page)

Early in the 2000’s, Strick’s group undertook precise anatomical remapping of the cerebrocerebellar communication loops (
Kelly and Strick, 2003
;
Dum and Strick, 2003
). One of the two loops so defined is attached to the cortical motor areas, whereas the other is attached to areas 46 and 9 of the cerebral prefrontal cortex (Figures 42 and 53). Ramnani (
2006
) reviewed anatomical data about these loops and pointed out that, whereas in macaque brains, fibers from the cortical motor system occupied the largest proportion of the cerebral peduncle, a comparatively small proportion was occupied by fibers from the prefrontal cortex. Importantly, he also pointed out that in the human brain, the largest contribution came not from the cortical motor areas but from the prefrontal cortex, suggesting that in humans the cerebellum has a more important role than in macaques in processing information from the prefrontal cortex. This is probably information at a more abstract level than that processed in the motor cortex. To explain such a non-motor role of the cerebellum, Ramnani (
2006
) adopted (as did I) the formalistic analogy between motor and cognitive systems as internal model-based control systems. Thus, we may conceive that a provisional thought system includes the prefrontal cortex as a controller, which is assisted by an internal model formed in the cerebellar hemisphere (
Ito, 2008
). Supportive evidence for such a thought system is available in the wealth of ever-increasing brain imaging data, and in the brain lesion and disease data that had accumulated over many years (see
Chapter 17
, “Cognitive Functions”).

The history of research on the cerebellum is certainly replete with excitement and thrilling experimental possibilities. Over the past five decades, in particular, basic concepts of synaptic plasticity, error learning, adaptive control, and model-based control have been formulated and substantiated experimentally. This has changed the once widely held belief that the function of the cerebellum was strictly for a relatively simple form of motor control to the current idea that it is an elaborate neuronal machine equipped with learning capabilities and devoted to far-more-advanced forms of systems control for posture and movement and probably also for participation in the control of complex motor actions and cognitive functions.

We are now ready to begin decomposing neuronal circuits in the cerebellar cortex into component neurons and examining them one by one. This exercise provides the basis for considering principles operating in these circuits and also for testing the validity of thus-far-derived hypotheses, which are discussed in later chapters. The focus in this chapter is on the neuronal elements that comprise the input and output interactions with the cerebellar cortex via mossy fibers, granule cells, and Purkinje cells. Unipolar brush cells and beaded fibers are also considered.

Mossy fibers are the most numerous afferent fibers that reach the cerebellar cortex through the white matter and terminate in the cerebellum’s granular layer, forming a mosslike structure (
Figure 14
). Some mossy fibers originate as sensory peripheral nerves, but most mossy fibers arise from neurons located within the spinal cord and brainstem. They also arise from unipolar brush cells in the granular layer (see below) and from cerebellar nuclei (
Chapter 6
, “
Pre- and Post-Cerebellar Cortex Neurons
”). Mossy fibers terminate within a glomerulus, which forms a characteristic rosette structure. Within a glomerulus, granule cell dendrites receive α-amino-3-hydroxy-5-methyl-4-isoxazolone propionic acid (AMPA)-mediated excitatory synapses from a mossy fiber terminal. Granule cells receive also inhibitory synapses supplied by a Golgi cell axon terminal. Descending dendrites of mostly deep Golgi cells also receive excitatory synapses directly from a mossy fiber terminal.

Most mossy fibers release glutamate as a transmitter, but some in the vestibulocerebellum release acetylcholine (
Barmack et al., 1992a
,
b
;
Jaarsma et al., 1997
).
Both AMPA and N-methyl-D-aspartate (NMDA) receptors mediate mossy fiber-granule cell synapses in glomeruli (
Traynelis et al., 1993
). Within a glomerulus, NMDA receptor subunits (NR1, NR2A, and NR2C) are co-located between the centrally positioned rosette structure and the peripherally positioned, tiny Golgi cell axon terminals at the postsynaptic junction with granule cell dendrites (
Yamada et al., 2001
).

A marked “spillover” phenomenon has been reported to occur in both glutamate released from a mossy fiber terminal and gamma-amino-butyric acid (GABA) released from Golgi cells. Single AMPA-receptor-mediated excitatory postsynaptic currents (EPSCs) or potentials (EPSPs) at the mossy fiber-granule cell connection are mediated by both the direct release of glutamate and the rapid diffusion of glutamate from neighboring synapses. Spillover currents contribute about one-half of the synaptic charge and improve transmission efficacy by increasing both the amplitude and duration of EPSPs (
DiGregorio et al., 2002
). Fluctuation analysis indicates that these indirect release sites are at least fourfold more numerous than those directly connected to the postsynaptic cell. As a result, spillover is predicted to improve the reliability and reduce the variability of transmission at this glomerular synapse. The unique firing behavior of granule cells may also be relevant; a single impulse in a mossy fiber tends to induce bursting spikes in a granule cell (
Chadderton et al., 2004
).

Granule cells are individually the smallest (soma diameter, 5–8 micrometers (μm)) and the most numerous neurons in the CNS (
Braitenberg and Atwood, 1958
;
Zagon et al., 1977
). A large divergence and a small convergence characterize the mossy fiber-granule cell pathway. Each granule cell receives mossy fiber terminals via only four to five excitatory synapses (
Eccles et al., 1967
;
Chadderton et al., 2004
). The functional significance of this small convergence number will be considered later in
Chapter 8
, “
Multiplicity and Persistency of Synaptic Plasticity
.” In contrast, one mossy fiber supplies excitatory synapses to 400–600 granule cells in a folium and probably more when the branches of a mossy fiber reach two or more folia. The efficacy of synaptic transmission from a mossy fiber to granule cells may vary probabilistically from glomerulus to glomerulus. Such relative efficacy may also be affected by the following three factors: activity-dependent induction of long-term potentiation (LTP) (
Chapter 8
), enhancement of intrinsic excitability (
Armano et al., 2000
), and Golgi cell inhibition (
Chadderton et al., 2004
; see also below).

Parallel fiber axons of granule cells run along the folia of the cerebellar surface after ascending vertically from the granular layer to the molecular layer and then
bifurcating into two parallel fiber collaterals. The formation of parallel fibers is controlled genetically. This is known because in Pax6 mutant rats, granule cells in the external germinal layer fail to form parallel fiber axons (
Yamasaki et al., 2001
). In normal animals, the length of a parallel fiber from terminal to terminal across its T-junctions has been reestimated to be as long as 4–6 mm (
Mugnaini, 1983
;
Harvey and Napper, 1988
;
Pichitpornchal et al., 1994
). Optical recording in mice shows that the local stimulation of a parallel fiber bundle excites Purkinje cells along the bundle over a distance of more than 3 mm (
Coutinho et al., 2004
). This extent of excitation was also observed by optical recording in neonatal rats on postnatal day 5, although it reduced to 1.5–2 mm at postnatal days 6–7 (
Arata and Ito, 2004
).

Ascending segments of granule cell axons form synapses with spines, which are located exclusively on the smallest diameter distal regions of Purkinje cell dendrites (
Gundappa-Sulur et al., 1999
;
Lu et al., 2009
). This contrasts to parallel fibers, which form synapses on the intermediate or large diameter regions of spiny branchlets, as well as the smallest diameter distal regions. The ascending segments form about 20% of the granule cell-Purkinje cell synapses. A differential stimulation of parallel fibers and ascending segments of granule cell axons in cerebellar slices revealed substantial differences in the properties of EPSCs generated in Purkinje cell dendrites (
Sims and Hartell, 2005
). Ascending segment synapses release a transmitter with a higher mean release probability and larger mean quantal amplitude than parallel fiber synapses, and they do not exhibit LTD (
Chapter 7
, “
Conjunctive Long-Term Depression (LTD)
”). These different properties of parallel fiber versus ascending segment synapses suggest that they have different roles in Purkinje cell function.

Unipolar brush cells of unique morphology are located primarily in the granular layer of the vestibulocerebellum. This portion of the cerebellum, which corresponds roughly to the flocculonodular lobe, receives primary vestibular afferents in the form of mossy fibers. These cells receive excitatory synapses on their dendritic “brush” from a single mossy fiber terminal (Color Plate VI). This connection has the form of a giant glutamate-mediated synapse (
Diño et al., 1999
). The unipolar brush cell’s axon forms branches within the granular layer, which give rise to large terminals that synapse with both granule cell and unipolar brush cell dendrites. This arrangement is within glomeruli that resemble those formed by extrinsic mossy fibers. Hence, unipolar brush cells are an intracortical source of mossy fibers.

Unipolar brush cells receive inputs from glutamate-mediated primary vestibular fibers and choline-acetyltransferase-positive mossy fibers. Some of the latter
originate from the medial and descending vestibular nuclei (
Diño et al., 2001
). An excitatory effect of muscarine, but not nicotine, was detected in ~15% of granule cells tested in the vestibulocerebellum (
Takayasu et al., 2003
). Evidence suggests that this effect is caused by the inhibition of an intrinsic outward K
⁺
current via the activation of muscarinic M3 receptors. Two subtypes of unipolar brush cells have been distinguished: one expresses calretinin, and the other expresses metabotropic glutamate receptor type 1a (mGluR1a) (
Nunzi et al., 2002
). Both subtypes express glutamate receptor subunit 2 (GluR2) (
Sekerková et al., 2004
). Tbr2/Eomes, a T-domain transcription factor (Tbr2), has been considered to be a specific marker of both subtypes of unipolar brush cells in the adult and developing cerebellum (
England et al., 2006
) (Color Plate VII). Unipolar brush cells express NMDA, kainite, and AMPA receptors in the synaptic membrane. They also express metabotropic glutamate receptors (mGluR1 and mGluR2/3) on the perisynaptic and extrasynaptic parts of the spiny appendages of dendrites (
Jaarsma et al., 1995
,
1998
;
Billups et al., 2002
). Mossy fiber impulses induce an AMPA-mediated fast EPSC and a predominantly NMDA-mediated slow EPSC in unipolar brush cells (
Rossi et al., 1995
). It has been suggested that unipolar brush cells may amplify mossy fiber inputs in the vestibulocerebellum (
Kalinichenko and Okhotin, 2005
;
Barmack and Yakhnitsa, 2008
).

Purkinje cells are the largest neurons in the cerebellum, extending magnificent planar dendrites to receive numerous synaptic inputs (Color Plate VII). Purkinje cells mediate the sole outputs of the cerebellar cortex, which are exclusively inhibitory in action upon their target neurons. Parallel fibers form excitatory synapses on dendritic spines of Purkinje cells. The synaptic membrane, lined with postsynaptic density (PSD), is located on the side (but not top) of a spine head and is therefore located at an optimal distance from the endoplasmic reticulum that protrudes to the spine head (
Launey et al., 2004
) (Color Plate VIII). A large divergence and an enormous convergence characterize the parallel fiber-Purkinje cell connection. While a single parallel fiber extends for ~3 mm (i.e., ~1.5 mm on each side of the T-junction), it passes through the dendrites of ~450 Purkinje cells and thereby forms synaptic contacts with the dendritic spines of at least 300 Purkinje cells (
Eccles et al., 1967
). On the other hand, the number of parallel fibers making synaptic contacts with the dendritic arborization of a Purkinje cell can be as large as 180,000 in the human (
Fox and Bernard, 1957
) or either ~60,000–80,000 (
Palay and Chan-Palay, 1974
) or ~175,000 (
Napper and Harvey, 1988a
,
b
) in the rat. Note that parallel fibers form synaptic contacts with only ~54% of the Purkinje cells through whose dendritic arborization they pass. Simultaneous whole-cell recording from synaptically connected granule and Purkinje cells in cerebellar slices revealed that an impulse from a single granule cell evoked a fast EPSC of 2–60 pA in a Purkinje cell (
Barbour, 1993
). This suggests that ~50 simultaneously active granule cells are sufficient to excite a single Purkinje cell.

Parallel fiber impulses release glutamate as a transmitter, which evokes two pharmacologically distinct types of synaptic potential in Purkinje cells. One is mediated by AMPA receptors, and the other by mGluR1. AMPA-mediated EPSPs are fast and evoked individually by each granule cell’s impulses (
Figure 15C
), whereas mGluR1-EPSPs are slow and observed only after a brief tetanus of parallel fibers (8 pulses at 50 Hz) in the presence of an AMPA receptor antagonist (
Batchelor and Garthwaite, 1993
) (
Figure 15D
). Slow EPSPs are accompanied by an increase in intradendritic sodium concentration, but the mechanism underlying this excitation remains unclear. The above two types of EPSP have different frequency characteristics. For example, following a single-shock stimulation of parallel fibers, fast EPSPs predominate, whereas in a burst stimulation of parallel fibers, slow EPSPs are facilitated. When a parallel fiber bundle is repetitively stimulated with 10 pulses at 100 Hz, the mGluR1- and AMPA receptor-mediated activations of Purkinje cells are equally potent (
Coutinho et al., 2004
). Metabotropic GABA
_B
receptors are expressed in the extra-postsynaptic sites of parallel fiber-Purkinje cell synapses. The activation of GABA
_B
receptors leads to the augmentation of mGluR1-mediated parallel fiber-Purkinje cell transmission (
Hirono et al., 2001
). This is an interesting case of interaction between two types of metabotropic receptor.