The Cerebellum: Brain for an Implicit Self (13 page)

The molecular layer of the cerebellum contains a group of glial progenitors termed “oligodendrocyte precursor cells” (
Palay and Chan-Palay, 1974
;
Levine and Card, 1987
). They express chondroitin-sulfate proteoglycan (NG2). Such NG2+ cells are found throughout the developing and mature CNS, and they are responsible for generating oligodendrocytes and myelin throughout life. NG2+ cells in the molecular layer of the cerebellum are small (soma diameter, 10–15 micrometers), and they extend numerous processes among the dendrites of Purkinje cells. Recently, Lin et al. (
2005
) found that climbing fiber stimulation induced AMPA receptor-mediated synaptic currents in NG2+ glial cells located in the molecular layer. A quantal analysis suggested that a climbing fiber formed up to 70 discrete junctions with one NG2+ cell and that each NG2+ cell could be innervated by more than one climbing fiber. Electron microscopic analysis of physiologically identified NG2+ cells revealed that anatomically defined climbing fibers formed direct synapses with the processes of NG2+ cells in the molecular layer. These recent findings require revision of our conventional concepts of neuronal circuits to include glial cells not only as supportive elements but also as important components for information processing.

In
Chapters 4
and
5
, more than ten unique cell types in the cerebellar cortex are defined. They interact in complex neuronal circuits in the cerebellar cortex. Among the inhibitory neurons in the molecular layer, basket/stellate cells add to operation of the mossy fiber-granule cell-Purkinje cell pathway. Among the inhibitory interneurons in the granular layer, Golgi cells are suggested to provide a clock circuit (
Chapter 9
), but functional roles of the other cell types still remain to be clarified. In addition to these neurons, two types of glial cells, Bergmann and NG2+, appear to be involved importantly in the neuronal circuit function of the cerebellar cortex.

In this chapter we decompose the neuronal circuits that connect the cerebellar cortex with other CNS structures. Some of the component cells are those of origin of afferents to the cerebellum that are located in the spinal cord and brainstem. These cells issue afferents known as mossy fibers, climbing fibers, and beaded fibers. Other component cells are the recipients of inhibitory signals from Purkinje cells, and they are located in cerebellar nuclei and selected brainstem nuclei. These precerebellar and postcerebellar cortex neurons, together with cerebellar cortical networks, constitute “global” neural systems that incorporate the cerebellum.

The vestibular nerve provides a unique case in which primary afferents project directly as mossy fibers to the vestibulocerebellum. These projections are prominent in the nodulus and uvula (lobules X and IX) and less represented in the flocculus (see
Chapter 10
, “
Ocular Reflexes
,”
Section 3
). The dorsal and ventral spinocerebellar tracts are examples of secondary afferents that are mossy fibers. These tracts convey proprioceptive and exteroceptive information from the body to the cerebellum. The dorsal spinocerebellar tract (DSCT) originates from neurons largely in Clarke’s column, which is located in the C8-L2/3 spinal segments. DSCT neurons project their axons through the ipsilateral region of the posteriolateral funiculus and then into the granular layers of cerebellar lobules I, II, III, and VIII. The ventral spinocerebellar tract (VSCT) originates from neurons located caudal to Clarke’s column with axons that ascend bilaterally through the ventral border of the lateral funiculi toward the cerebellum. They double cross the midline, first to the opposite side of the body and then back to their original side in the cerebellum.

The DSCT is usually considered to be purely sensory in transmission of proprioceptive information from the ipsilateral hindlimb. However, the presence of corticospinal input to DSCT neurons has long pointed to an integration of sensory input and motor command signals to these neurons (
Hongo et al., 1967
; Hongo and Okada, 1976). A recent study showed that DSCT neurons encode, in addition, a global representation of hindlimb mechanics during passive movement of a hindlimb. The discharge of these neurons was shown to be qualitatively but not quantitatively similar for active stepping versus passive, manually imposed steps (
Bosco et al., 2006
). The difference was suggested to indicate a contribution of DSCT to the spinal cord’s contribution to the control of locomotion (
Chapter 11
, “
Somatic and Autonomic Reflexes
”). In a very recent study by Hantman and Jessell (
2010
), proprioceptive afferents, corticospinal tract axons, and GABAergic and glycinergic synapses were labeled differentially by genetic/molecular markers to reveal their direct synaptic contact onto DSCT neurons. Moreover, patch clamp recording from DSCT neurons revealed that monosynaptic EPSPs were evoked by stimulation of both the dorsal roots and the dorsal column, the latter involving corticospinal axons. These observations indicated converging excitatory inputs to DSCT neurons from a proprioceptive pathway and the corticospinal tract. These neurons also have GABAergic and glycinergic synapses, possibly from the numerous GABAergic and glycinergic interneurons that surround Clarke’s column and receive dense innervation from corticospinal terminals. Consistent with these anatomical findings, dorsal column stimulation evoked long-lasting IPSPs in DSCT neurons. There was also indication of presynaptic inhibition, which occurred in corticospinal tract synapses on DSCT neurons. These synaptic connections point to the possibility that corticospinal descending signals for voluntary movements act on spinal segmental circuits not only for executing a voluntary movement, but also for adjusting signals ascending to supraspinal centers and the cerebellum via DSCT. This adjustment might be required for adapting posture and locomotion mechanisms to an intervening voluntary movement (see
Chapter 11
).

In contrast, the activity of VSCT neurons seems to be determined mainly by a central mechanism rather than sensory input. These neurons discharge rhythmically at their own consistent phase of the step cycle when the scratch reflex or locomotion is elicited in cats (
Arshavsky et al., 1978
,
1988
). During fictive scratching and locomotion in a paralyzed animal, the firing pattern of these neurons is similar to that during actual scratching. Therefore, the rhythmical burst firing of VSCT neurons must be determined centrally. Such a central mechanism may operate via segmental interneurons, which seemingly provide the VSCT with feedback information about their actions on motoneurons.

The “rostral spinocerebellar tract” is the forelimb equivalent of the hindlimb’s VSCT. It transmits, among probably other information, Ib afferent discharge to the cerebellum that arises from Golgi tendon organs of the cranial half of the body. The connectivity organizations of these two tracts are largely similar but with certain differences (
Oscarsson et al., 1965
). The “cuneocerebellar tract” in the cat consists of a proprioceptive component that arises from the external cuneate nucleus and is activated by group I muscle afferents. A second exteroceptive component arises from cells in the rostral part of the main cuneate nucleus and is activated by cutaneous afferents (
Cooke et al., 1971
). The cuneocerebellar tract terminates in the intermediate part of lobule V of the anterior lobe of the cerebellum and in the rostral four folia of the paramedian lobule.

The lateral reticular nucleus has three subnuclei: the parvocellular, magnocellular, and subtrigeminal subnuclei. The major part of the lateral reticular nucleus (mLRN) involves the parvocellular portion and the immediately adjacent magnocellular portion. These portions receive monosynaptic excitation from fibers ascending in the ventral part of the ipsilateral lateral funiculus of the spinal cord. In turn, mLRN projects almost exclusively to the classical spinocerebellum (
Clendenin et al., 1974
), involving bilaterally the intermediate part of the anterior lobe. Some fibers from the mLRN terminate also in the rostral part of lobule VI but hardly any in other parts of the cerebellar cortex. The rest of the magnocellular subnucleus receives inputs from higher brain structures and projects to the cerebellar hemispheres. The subtrigeminal nucleus sends its projections to the flocculonodular lobe.

Another prominent source of mossy fibers is the pontine nuclei in the brainstem. The vast majority of pontine nuclear neurons receive monosynaptic glutamatergic inputs from corticopontine fibers and, in turn, send their axons to the cerebellar cortex. Large and functionally diverse areas of the cerebral cortex find their “own” territories in the pontine nucleus. Similarly, each part of the pontine nucleus projects to specific areas of the cerebellar cortex. This topographic organization of the pontine nuclei may ensure that information from various functionally diverse parts of the cerebral cortex and subcortical nuclei is brought together and integrated in the cerebellar cortex (
Brodal and Bjaale, 1992
).

In the vermis and flocculonodular lobe, the major excitatory inputs to the target neurons of Purkinje cells are provided by collaterals of mossy fiber afferents (
Figure 13
). However, the presence of collaterals of pontocerebellar mossy fibers to cerebellar nuclei has been controversial. Shinoda et al. (
1992
) used intra-axonal horseradish peroxidase staining in cats to trace single axons of pontine nuclear neurons that received cerebral inputs. Forty such axons were shown to terminate as
typical mossy fiber rosettes in the granular layer of the cerebellar cortex. Of these, 22 axons projected collaterals to the lateral cerebellar nucleus. Virtually all of the axon branches observed in the lateral cerebellar nucleus were collaterals of mossy fibers from pontine nuclear neurons to the cerebellar cortex. One to three collaterals were projected from their parent axons to the lateral cerebellar nucleus. They were very thin (mean diameter, 0.6 micrometers) compared to the large parent axons (2.1 micrometers). Some collaterals were very simple with only one terminal branch with or without short branchlets. In other cases, single collaterals ramified before forming a terminal arborization. Hence, in contrast to the large divergence and convergence of the corticopontine-cerebellar projections, pontine nuclear neurons appear to form rather specific connections with lateral cerebellar nucleus neurons.

Another source of mossy fiber afferents in the brainstem is the nucleus reticularis tegmenti pontis (NRTP). It contains neurons that relay visual signals to the flocculus (
Kano et al., 1991
) and other cells that are sensitive to head rotation (
Taillanter and Lannou, 1988
) (
Chapter 10
). The NRTP in monkeys also receives major inputs from the primary motor cortex and less input from the premotor region (area 6), the somatosensory cortex (areas 3, 1, and 2), and area 5 (
Figure 2
). These cortical afferents terminate throughout the NRTP except for its dorsomedial part (
Brodal, 1980
).

The pontine nuclei and the NRTP project to the lateral cerebellar nuclei. This was shown by use of biotinylated dextran labeling in rats. The pontine nuclei also project to the posterior interpositus nucleus. In contrast, NRTP projects to the lateral cerebellar nucleus, the nucleus interpositus, and the caudal part of the nucleus medialis. Cerebellar nuclear neurons are innervated mainly by projections from contralateral pontine nuclei. They are also innervated bilaterally, albeit to a lesser degree by NRTP. These anatomical differences imply different functional roles for these two sources of mossy fibers (
Parenti et al., 2002
).

Climbing fibers originate solely from the inferior olive (IO) in the opposite side of the medulla. In rodents, this complex consists of three primary subdivisions: the principal olive (PO), medial accessory olive (MAO), and the dorsal accessory olive (DAO) (Color Plate III). In cats, the total number of IO neurons is ten times greater than that of Purkinje cells, such that one climbing fiber appears to project axonal branches to innervate ten Purkinje cells. This observation was verified in a study using the retrograde fluorescent double-labeling method in rats. The lateral parts of the IO project more rostrally within a longitudinal zone, and the medial
parts project more caudally in the same zone. Double-labeled olivary neurons with axons branching rostrally and caudally within a single zone were found to lie in an intermediate position between the two groups of single-labeled neurons (
Wharton and Payne, 1985
).

The IO contains two morphologically distinct types of neurons. One group is “curly” cells, with relatively small cell somata and curly dendrites as described by Scheibel and Scheibel (1975). The other type is “straight” cells with relatively large cell somata and long, relatively straight dendrites (
Devor and Yarom, 2002
) (Color Plate XI A, B). A characteristic feature of IO neurons is that they communicate with each other via gap junction-mediated electrical synapses (
Llinás et al., 1974
;
Sotelo et al., 1974
) as demonstrated in Color Plate XI. Electrical synapses provide reciprocal pathways for ionic current and small organic molecules, and they are maintained by the protein connexin36 (Cx36) (
Condorelli et al., 2001
).

Neurobiotin spreads through gap junctions. When injected into an IO neuron, it was shown to produce indirect labeling of 1–11 nearby cells (average, 3.8 cells). All indirectly labeled cells were found within 70 micrometers of the injected cell, that is, within or immediately adjacent to the dendritic field of the stained cell (Color Plate XI). Dye passage through gap junctions from a straight type IO neuron stained only straight type neurons. Likewise, dye passage from a curly type IO neuron stained only curly type neurons. Each IO neuron was calculated to be directly coupled with about 50 other IO neurons. The coupling coefficient was defined as the ratio between voltage responses of the post- and prejunctional cell. As measured by simultaneous double patch recordings from IO neurons in rat brain slice preparations, the coupling coefficient varied between 0.002 and 0.17; that is, most of the pairs were weakly coupled (
Devor and Yarom, 2002
).