MicrocircuitsCh28 Ito
28 Cerebellar Cortex Masao Ito
The microcircuit in the cerebellum is featured by the relative simplicity, precision, and geometric beauty of its arrangement. Its structure is identical all over the cerebellar cortex except for some regional differences (Fig. 28.1). The cerebellar cortex has three layers (molecular layer, Purkinje cell layer, and granular layer) and can be divided into more than a hundred subareas by horizonal grooves and longitudinal bands. Each subarea can be further subdivided into a number of microzones (there could be 10,000 microzones in the human cerebellum). A microzone, in combination with a small portion of the interior olive and in some regions also with that of parvocellular red nucleus, consists of a microcomplex, a functional unit of the cerebellum (Fig. 28.2). This chapter overviews the current knowledge on neuronal elements and their connections in the cerebellar microcircuit and its functional principles. Most of the relevant references can be found in Ito (2006).
Circuit Elements Mossy Fiber Mossy fiber afferents arise from numerous sources in peripheral nerves, the spinal cord, and the brainstem, and they convey major information to be processed in the cerebellar cortical circuit. Mossy fiber terminals in the granular layer of the cerebellar cortex form a characteristic rosette structure within a glomerulus. Within this structure, a mossy fiber terminal supplies excitatory synapses (mediated by both AMPA and NMDA receptors, but some can be cholinergic) to granule cell dendrites.
293
Handbook of Brain Microcircuits
294
LTD Cerebellar cortex Go
SC
PF
BC
Purkinje cell
Lg Gr
CF UB IO Pd R–O
5-HT fibres
N–O
N–C MF
pRN PCN
CN/VN
FIGURE 28–1. Microcircuit of the cerebellum. BC, basket cell; CF, climbing fiber; CN, cerebellar nucleus; GR, granule cell; GL, glomerulus; GO, Golgi cell; IO, inferior olive; LC, Lugaro cell; MF, mossy fiber; N-C, nucleocortical mossy fiber projection; N-O, nucleoolivary inhibitory projection; PCN, precerebellar neuron; PF, parallel fiber; pRN, parvicellular red nucleus; R-O, rubroolivary excitatory projection; SC, stellate cell; 5-HT, serotonergic; UB, unipolar brush cell; and VN, vestibular nucleus. (From Ito, 2008)
Cerebellar microcomplex Neuromodulation
Error learning LTD Error signals
Mode signals Mossy fiber
Climbing fiber
Beaded fiber Inferior olive
Main input
Output
FIGURE 28–2. A microcomplex of the cerebellum. LTD, long-term depression.
28: Cerebellar Cortex
295
Granule Cells Granule cells are individually the smallest (5–8 μm in diameter), yet the most numerous neurons (1010–1011 in humans) in the brain. A large divergence (from one mossy fiber to 400–600 granule cells) and a small convergence (from four to five mossy fibers to a granule cell) characterize the mossy fiber– granule cell pathway. The degree of mossy fiber–granule cell divergence is functionally regulated by Golgi cells, which supply inhibitory synapses to granule cells. Each granule cell issues an ascending axon, which branches in the molecular layer in T shape to form a parallel fiber (PF). Parallel fiber synapses on Purkinje cells are mediated by AMPA and mGluR1 receptors, and those on basket/stellate cells are mediated by AMPA, NMDA, and mGluR1a receptors. Purkinje Cells Purkinje cells extending magnificent dendritic trees lie in a single layer of the cerebellar cortex (about 1000 cells per mm2 in rat). Their dendrites receive excitatory input from numerous PFs (175,000 per Purkinje cell in rat) and inhibitory input from basket and stellate cells. They also receive climbing fibers and beaded fibers. Purkinje cells in turn supply GABA-mediated inhibitory synapses to their target neurons in cerebellar nuclei and certain brainstem neurons. Reciprocal inhibition occurs among Purkinje cells via their recurrent collaterals extending within 300 μm. Axon collaterals of Purkinje cells also inhibit basket cells. Climbing Fibers These are unique structures in the cerebellum with no homolog elsewhere in the central nervous system. Each Purkinje cell is innervated by one climbing fiber as a consequence of the postnatal elimination of multiple innervation. Each climbing fiber forms numerous synaptic contacts with the dendrites of a single Purkinje cell (1,300 in proximal dendrites of rat Purkinje cells (Strata, 2002), but a much larger value, 26,000, is derived from the density ratio of climbing and parallel fiber synapses (Nieto-Bona et al., 1997)). This arrangement results in a particularly large excitatory postsynaptic potential (EPSP) superimposed with Ca2+ spikes. The major transmitter of climbing fibers is glutamate. Beaded Fibers The cerebellar cortex receives not only mossy and climbing fibers but also beaded fibers, which contain various amines (e.g., serotonin, dopamine, acetylcholine, norepinephrine, or histamine) or neuropeptides (angiotensin II,
296
Handbook of Brain Microcircuits
orexin, etc). This third type of afferent extends fine varicose fibers sparsely throughout the granular and molecular layers to form direct contact with Purkinje cells and other cerebellar neurons. On the basis of this morphology, the third type of afferent does not convey specific information to the cerebellar cortex. Rather, its role is modulatory and important in setting the activity level or switching the operational mode of cerebellar microcomplexes to match a behavioral demand. Unipolar Brush Cells Unipolar brush (UB) cells are located primarily in the granular layer of the vestibulocerebellum. On their dendritic brush, a UB cell receives a single mossy fiber terminal forming a giant glutamate-mediated synapse. NMDA, kainate, and AMPA receptors are expressed in the synaptic membrane and mGluR1 and mGluR2/3 receptors in the peri- and extrasynaptic parts of the spiny appendages of dendrites. Mossy fiber impulses induce in UB cells an AMPA-mediated fast excitation and a predominantly NMDA-mediated slow excitation. Unipolar brush cell axons branch within the granular layer and give rise to large terminals that synapse with both granule cell and UB cell dendrites within glomeruli. Unipolar brush cells may amplify mossy fiber inputs. Basket/Stellate Cells Basket cells are located near the Purkinje cell layer and receive numerous glutamate-mediated excitatory synapses from a bundle of PFs on their dendrites and, in turn, supply GABA-mediated inhibitory synapses to the bottleneck of a Purkinje cell soma, forming a unique complex structure called a “pinceau.” Stellate cells are located in the molecular layer and they also supply inhibitory synapses to Purkinje cell dendrites. Basket/stellate cells mediate the feedforward inhibition from PFs to Purkinje cells. A basket cell extends axons perpendicular to PFs and covers an area containing 10 × 7 rows of Purkinje cells, with a probable divergence number of 50. Twenty to 30 basket cell axons may converge onto one Purkinje cell. Basket cells receive also collaterals of climbing fibers and those of Purkinje cell axons. Golgi Cells Each Golgi cell receives ~4788 excitatory inputs to its dendrites in the molecular layer from PFs and also ~228 mossy fiber terminals on its descending dendrites. The major excitatory inputs from PFs to Golgi cells are mediated by both AMPA and NMDA receptors and mGluR2. Golgi cells also receive inhibitory synapses from Lugaro cells. A Golgi cell, in turn, extends a broadly
28: Cerebellar Cortex
297
branching axon to supply GABA-mediated inhibitory synapses up to ~5700 granule cells (cat). Lugaro Cells These cells are inhibitory neurons in the granular layer. Their fusiform cell soma is located in or slightly below the Purkinje cell layer. The cerebellar cortex contains approximately one Lugaro cell and one Golgi cell for every 15 Purkinje cells. Axons of more than 10 Lugaro cells converge onto only one Golgi cell, while the axon of one Lugaro cell diverges onto 150 Golgi cells. Lugaro cells express calretinin but not mGluR2 or somatostatin, whereas Golgi cells express mGluR2 and somatostatin but not calretinin. Lugaro cells in cerebellar slices are silent, but in the presence of serotonin, they discharge regularly at 5–15 Hz and induce inhibition in Golgi cells.
Globular Neurons These are a newly identified group of large granular layer neurons having a globular soma located at variable depths in the granular layer (Laine and Axelrad, 2002). They extend three to four long radiating dendrites coursing through the three layers of the cortex. Their axons project into the molecular layer and expand a local plexus, with a pattern similar to that of Lugaro cell. The axons of several of these cells give off a collateral that courses for a long distance in the transverse direction, just above the Purkinje cell somata, parallel to PFs. Globular cells may be inhibitory neurons.
Circuits and Models Synaptic Plasticity Various synapses in cerebellar microcircuit express activity-dependent changes in synaptic efficacy, that is, synaptic plasticity, typically long-term potentiation (LTP) and long-term depression (LTD). In mossy fiber–granule cell synapses, LTP involving NMDA receptors appears to functionally regulate the degree of mossy fiber–granule cell divergence. In Purkinje cells, whereas repetitive stimulation of PFs alone induces LTP in the PF synapses, either presynaptic or postsynaptic, conjunctive activation of PFs and a CF induces PF-LTD underlain by complex signal transduction. Purkinje cells also exhibit a prolonged potentiation of GABAA receptor–mediated inhibitory postsynaptic potentials (IPSPs) (i.e., rebound potentiation [RP]) after the activation of climbing fibers, basket/stellate cell–induced IPSPs. Rebound potentiation is input-nonspecific in contrast to the input-specific PF-LTD.
298
Handbook of Brain Microcircuits
Inhibitory synapses supplied by Purkinje cell axons to cerebellar nuclear neurons exhibit input-nonspecific LTD due to a decreased postsynaptic GABA sensitivity that is caused by an increase in [Ca2+]. In basket/stellate cells, PF burst stimulation paired with CF activity induces LTP complementary to PF-LTD in Purkinje cells. Likewise, stimulation of a PF bundle unpaired with CF activity induces LTD complementary to PF-LTP in Purkinje cells. Large stable CF-evoked excitation in Purkinje cells exhibits a weak LTD when repetitively evoked. Neurocomputing The forward connections from cells of origin of mossy fibers to granule cells and to Purkinje cells form the three-layered neuronal networks, which have been modeled as simple perceptron, adaptive filter, or liquid-state machine (Yamazaki and Tanaka, 2007). Whereas stimuli are received by the cells of origin of mossy fibers on the first layer, granule cells receive mossy fiber afferent on the second layer, where the small convergence may suggest “sparse” coding; that is, each granule cell represents an integration of a small number of mossy fibers. However, a recent analysis of sensory signals suggests “similar” coding in that each granule cell retains characteristics of each mossy fiber (Bengtsson and Jörntell, 2009). On the third layer, Purkinje cells receive numerous synapses from PFs and so integrate information from numerous granule cells to generate outgoing signals. How a Purkinje cell processes incoming and outgoing signals, by linear algorithm or pause, has been a matter of recent discussions (Steuber et al., 2007; Alviña et al., 2008; Walter and Khodakah, 2009). Learning The finding that PF synapses of Purkinje cells undergo LTP or LTD depending on pairing or unpairing with CF provides the basis for learning and memory capability of the cerebellar cortical microcircuit. The more recently found reciprocal pattern of a combination of LTD in Purkinje cells and LTP in basket/stellate cells, or vice versa, may have a synergistic effect to augment the memory storage capacity of the cerebellar cortical microcircuit. The learning in the cerebellar cortex is considered to be primarily “error learning” because CF signals induce LTD in coactivated PF synapses in Purkinje cells like punishment and because climbing fibers convey information regarding errors in a motor performance. The predominance of functionally silent PF synapses in Purkinje cells may imply that 85% or more of PF synapses are long-term depressed during repeated learning, by which the neurocomputing circuits are shaped through experiences from the original preponderantly connecting PF synapses (Dean et al., 2010).
28: Cerebellar Cortex
299
Internal Model A microcomplex is inserted into various microcircuits in the spinal cord and brainstem and acts as an embedded adaptive component in the motor or autonomic control systems. Furthermore, a microcomplex is connected to the cerebral cortical area and acts as an internal model of the latter. For the motor cortex, a microcomplex may provide a forward model or an inverse model of the lower motor centers and motor apparatus, which enables the motor cortex to perform a precise movement without the external feedback about the movement (Kawato, 2009). A similar idea has been expanded to cognitive functions; a microcomplex would provide an internal model for mental activities going on in the cerebral association cortex (Ito, 2008). Oscillation A computer simulation suggests that reciprocal inhibition causes an oscillation at 100–250 Hz in the activity of basket/stellate cells. A computer simulation also predicts that feedback inhibition from Golgi to granule cells induces 10–50 Hz oscillations in spike discharges from the latter. Because of the large divergence from Lugaro cells to Golgi cells, an interesting possibility is that Lugaro cells play a role in synchronizing activity among Golgi cells situated along the parallel fiber beam as observed in anesthetized rats. Lugaro cells may switch the operation of Golgi cells from the individual rhythmic mode to the synchronous mode. Golgi Cell Clock The loop connection involving granule cells and Golgi cells (Fig. 28.1) has been interpreted to constitute a phase converter, whose function is to generate a set of multiphase versions of a mossy fiber input. Another unique model proposed by Yamazaki and Tanaka (2005) features a randomly connected granule cell–Golgi cell loop pathway, which operates as a clock to generate granule cell discharge in an ever-changing ensemble of patterns. The patterns do not repeat unless reset by a large mossy fiber input to the pathway. When this Golgi cell clock is incorporated in a neuronal circuit model for eyeblink reflex, it reproduces appropriately timed conditioned responses.
References Alviña K, Walter JT, Kahn A, Ellis-Davis G, Khodakhah K (2008) Questioning the role of rebound firing in the cerebellum. Nat Neurosci 11:1256–1258. Bengtsson F, Jörntell H (2009) Sensory transmission in cerebellar granule cells relies on similarly coded mossy fiber inputs. Proc Natl Acad Sci 106:2389–2394.
300
Handbook of Brain Microcircuits
Dean. P, Porrill J, Ekerot C-F, Jöntell H (2010) The cerebellar microcircuit as an adaptive filter: experimental and computational evidence. Nat Rev Neurosci 11:30–43. Ito M (2006) Cerebellar circuitry as a neuronal machine. Prog Neurobiol 78:272–303. Ito M (2008) Control of mental activities by internal models in the cerebellum. Nat Rev Neurosci 9:304–313. Kawato M (2009) Cerebellum: models. In: Squire LR, ed. Encyclopedia of Neuroscience, pp. 757–767. Oxford, England: Academic Press. Elsevier Limited. Lainé J, Axelrad H. (2002) Extending the cerebellar Lugaro cell class. Neuroscience 115: 363–374. Nieto-Bona MP, Garcia Sergura LM, Torres-Aleman I (1997) Transsynaptic modulation by insulin-like growth factor I of dendritic spines in Purkinje cells. Int J Devl Neurosci 15:749–754. Steuber V, Mittmann W, Hoebeek FE, Angus Silver R, De Zeeuw CI, Häusser M, De Schutter E (2007) Cerebellar LTD and pattern recognition by Purkinje cells. Neuron 54:121–136. Strata P (2002) Dendritic spines in Purkinje cells. The Cerebellum 1:230–232. Walter JT, Khodakhah K (2009) The advantages of linear information processing for cerebellar computation. Proc Natl Acad Sci 17:4471–4476. Yamazaki T, Tanaka S (2005) Neural modeling of an internal clock. Neural Computation 17:1032–1058. Yamazaki T, Tanaka S (2007) The cerebellum as a liquid state machine. Neural Networks 20:290–297.
The microcircuit in the cerebellum is featured by the relative simplicity, precision, and geometric beauty of its arrangement. Its structure is identical all over the cerebellar cortex except for some regional differences (Fig. 28.1). The cerebellar cortex has three layers (molecular layer, Purkinje cell layer, and granular layer) and can be divided into more than a hundred subareas by horizonal grooves and longitudinal bands. Each subarea can be further subdivided into a number of microzones (there could be 10,000 microzones in the human cerebellum). A microzone, in combination with a small portion of the interior olive and in some regions also with that of parvocellular red nucleus, consists of a microcomplex, a functional unit of the cerebellum (Fig. 28.2). This chapter overviews the current knowledge on neuronal elements and their connections in the cerebellar microcircuit and its functional principles. Most of the relevant references can be found in Ito (2006).
Circuit Elements Mossy Fiber Mossy fiber afferents arise from numerous sources in peripheral nerves, the spinal cord, and the brainstem, and they convey major information to be processed in the cerebellar cortical circuit. Mossy fiber terminals in the granular layer of the cerebellar cortex form a characteristic rosette structure within a glomerulus. Within this structure, a mossy fiber terminal supplies excitatory synapses (mediated by both AMPA and NMDA receptors, but some can be cholinergic) to granule cell dendrites.
293
Handbook of Brain Microcircuits
294
LTD Cerebellar cortex Go
SC
PF
BC
Purkinje cell
Lg Gr
CF UB IO Pd R–O
5-HT fibres
N–O
N–C MF
pRN PCN
CN/VN
FIGURE 28–1. Microcircuit of the cerebellum. BC, basket cell; CF, climbing fiber; CN, cerebellar nucleus; GR, granule cell; GL, glomerulus; GO, Golgi cell; IO, inferior olive; LC, Lugaro cell; MF, mossy fiber; N-C, nucleocortical mossy fiber projection; N-O, nucleoolivary inhibitory projection; PCN, precerebellar neuron; PF, parallel fiber; pRN, parvicellular red nucleus; R-O, rubroolivary excitatory projection; SC, stellate cell; 5-HT, serotonergic; UB, unipolar brush cell; and VN, vestibular nucleus. (From Ito, 2008)
Cerebellar microcomplex Neuromodulation
Error learning LTD Error signals
Mode signals Mossy fiber
Climbing fiber
Beaded fiber Inferior olive
Main input
Output
FIGURE 28–2. A microcomplex of the cerebellum. LTD, long-term depression.
28: Cerebellar Cortex
295
Granule Cells Granule cells are individually the smallest (5–8 μm in diameter), yet the most numerous neurons (1010–1011 in humans) in the brain. A large divergence (from one mossy fiber to 400–600 granule cells) and a small convergence (from four to five mossy fibers to a granule cell) characterize the mossy fiber– granule cell pathway. The degree of mossy fiber–granule cell divergence is functionally regulated by Golgi cells, which supply inhibitory synapses to granule cells. Each granule cell issues an ascending axon, which branches in the molecular layer in T shape to form a parallel fiber (PF). Parallel fiber synapses on Purkinje cells are mediated by AMPA and mGluR1 receptors, and those on basket/stellate cells are mediated by AMPA, NMDA, and mGluR1a receptors. Purkinje Cells Purkinje cells extending magnificent dendritic trees lie in a single layer of the cerebellar cortex (about 1000 cells per mm2 in rat). Their dendrites receive excitatory input from numerous PFs (175,000 per Purkinje cell in rat) and inhibitory input from basket and stellate cells. They also receive climbing fibers and beaded fibers. Purkinje cells in turn supply GABA-mediated inhibitory synapses to their target neurons in cerebellar nuclei and certain brainstem neurons. Reciprocal inhibition occurs among Purkinje cells via their recurrent collaterals extending within 300 μm. Axon collaterals of Purkinje cells also inhibit basket cells. Climbing Fibers These are unique structures in the cerebellum with no homolog elsewhere in the central nervous system. Each Purkinje cell is innervated by one climbing fiber as a consequence of the postnatal elimination of multiple innervation. Each climbing fiber forms numerous synaptic contacts with the dendrites of a single Purkinje cell (1,300 in proximal dendrites of rat Purkinje cells (Strata, 2002), but a much larger value, 26,000, is derived from the density ratio of climbing and parallel fiber synapses (Nieto-Bona et al., 1997)). This arrangement results in a particularly large excitatory postsynaptic potential (EPSP) superimposed with Ca2+ spikes. The major transmitter of climbing fibers is glutamate. Beaded Fibers The cerebellar cortex receives not only mossy and climbing fibers but also beaded fibers, which contain various amines (e.g., serotonin, dopamine, acetylcholine, norepinephrine, or histamine) or neuropeptides (angiotensin II,
296
Handbook of Brain Microcircuits
orexin, etc). This third type of afferent extends fine varicose fibers sparsely throughout the granular and molecular layers to form direct contact with Purkinje cells and other cerebellar neurons. On the basis of this morphology, the third type of afferent does not convey specific information to the cerebellar cortex. Rather, its role is modulatory and important in setting the activity level or switching the operational mode of cerebellar microcomplexes to match a behavioral demand. Unipolar Brush Cells Unipolar brush (UB) cells are located primarily in the granular layer of the vestibulocerebellum. On their dendritic brush, a UB cell receives a single mossy fiber terminal forming a giant glutamate-mediated synapse. NMDA, kainate, and AMPA receptors are expressed in the synaptic membrane and mGluR1 and mGluR2/3 receptors in the peri- and extrasynaptic parts of the spiny appendages of dendrites. Mossy fiber impulses induce in UB cells an AMPA-mediated fast excitation and a predominantly NMDA-mediated slow excitation. Unipolar brush cell axons branch within the granular layer and give rise to large terminals that synapse with both granule cell and UB cell dendrites within glomeruli. Unipolar brush cells may amplify mossy fiber inputs. Basket/Stellate Cells Basket cells are located near the Purkinje cell layer and receive numerous glutamate-mediated excitatory synapses from a bundle of PFs on their dendrites and, in turn, supply GABA-mediated inhibitory synapses to the bottleneck of a Purkinje cell soma, forming a unique complex structure called a “pinceau.” Stellate cells are located in the molecular layer and they also supply inhibitory synapses to Purkinje cell dendrites. Basket/stellate cells mediate the feedforward inhibition from PFs to Purkinje cells. A basket cell extends axons perpendicular to PFs and covers an area containing 10 × 7 rows of Purkinje cells, with a probable divergence number of 50. Twenty to 30 basket cell axons may converge onto one Purkinje cell. Basket cells receive also collaterals of climbing fibers and those of Purkinje cell axons. Golgi Cells Each Golgi cell receives ~4788 excitatory inputs to its dendrites in the molecular layer from PFs and also ~228 mossy fiber terminals on its descending dendrites. The major excitatory inputs from PFs to Golgi cells are mediated by both AMPA and NMDA receptors and mGluR2. Golgi cells also receive inhibitory synapses from Lugaro cells. A Golgi cell, in turn, extends a broadly
28: Cerebellar Cortex
297
branching axon to supply GABA-mediated inhibitory synapses up to ~5700 granule cells (cat). Lugaro Cells These cells are inhibitory neurons in the granular layer. Their fusiform cell soma is located in or slightly below the Purkinje cell layer. The cerebellar cortex contains approximately one Lugaro cell and one Golgi cell for every 15 Purkinje cells. Axons of more than 10 Lugaro cells converge onto only one Golgi cell, while the axon of one Lugaro cell diverges onto 150 Golgi cells. Lugaro cells express calretinin but not mGluR2 or somatostatin, whereas Golgi cells express mGluR2 and somatostatin but not calretinin. Lugaro cells in cerebellar slices are silent, but in the presence of serotonin, they discharge regularly at 5–15 Hz and induce inhibition in Golgi cells.
Globular Neurons These are a newly identified group of large granular layer neurons having a globular soma located at variable depths in the granular layer (Laine and Axelrad, 2002). They extend three to four long radiating dendrites coursing through the three layers of the cortex. Their axons project into the molecular layer and expand a local plexus, with a pattern similar to that of Lugaro cell. The axons of several of these cells give off a collateral that courses for a long distance in the transverse direction, just above the Purkinje cell somata, parallel to PFs. Globular cells may be inhibitory neurons.
Circuits and Models Synaptic Plasticity Various synapses in cerebellar microcircuit express activity-dependent changes in synaptic efficacy, that is, synaptic plasticity, typically long-term potentiation (LTP) and long-term depression (LTD). In mossy fiber–granule cell synapses, LTP involving NMDA receptors appears to functionally regulate the degree of mossy fiber–granule cell divergence. In Purkinje cells, whereas repetitive stimulation of PFs alone induces LTP in the PF synapses, either presynaptic or postsynaptic, conjunctive activation of PFs and a CF induces PF-LTD underlain by complex signal transduction. Purkinje cells also exhibit a prolonged potentiation of GABAA receptor–mediated inhibitory postsynaptic potentials (IPSPs) (i.e., rebound potentiation [RP]) after the activation of climbing fibers, basket/stellate cell–induced IPSPs. Rebound potentiation is input-nonspecific in contrast to the input-specific PF-LTD.
298
Handbook of Brain Microcircuits
Inhibitory synapses supplied by Purkinje cell axons to cerebellar nuclear neurons exhibit input-nonspecific LTD due to a decreased postsynaptic GABA sensitivity that is caused by an increase in [Ca2+]. In basket/stellate cells, PF burst stimulation paired with CF activity induces LTP complementary to PF-LTD in Purkinje cells. Likewise, stimulation of a PF bundle unpaired with CF activity induces LTD complementary to PF-LTP in Purkinje cells. Large stable CF-evoked excitation in Purkinje cells exhibits a weak LTD when repetitively evoked. Neurocomputing The forward connections from cells of origin of mossy fibers to granule cells and to Purkinje cells form the three-layered neuronal networks, which have been modeled as simple perceptron, adaptive filter, or liquid-state machine (Yamazaki and Tanaka, 2007). Whereas stimuli are received by the cells of origin of mossy fibers on the first layer, granule cells receive mossy fiber afferent on the second layer, where the small convergence may suggest “sparse” coding; that is, each granule cell represents an integration of a small number of mossy fibers. However, a recent analysis of sensory signals suggests “similar” coding in that each granule cell retains characteristics of each mossy fiber (Bengtsson and Jörntell, 2009). On the third layer, Purkinje cells receive numerous synapses from PFs and so integrate information from numerous granule cells to generate outgoing signals. How a Purkinje cell processes incoming and outgoing signals, by linear algorithm or pause, has been a matter of recent discussions (Steuber et al., 2007; Alviña et al., 2008; Walter and Khodakah, 2009). Learning The finding that PF synapses of Purkinje cells undergo LTP or LTD depending on pairing or unpairing with CF provides the basis for learning and memory capability of the cerebellar cortical microcircuit. The more recently found reciprocal pattern of a combination of LTD in Purkinje cells and LTP in basket/stellate cells, or vice versa, may have a synergistic effect to augment the memory storage capacity of the cerebellar cortical microcircuit. The learning in the cerebellar cortex is considered to be primarily “error learning” because CF signals induce LTD in coactivated PF synapses in Purkinje cells like punishment and because climbing fibers convey information regarding errors in a motor performance. The predominance of functionally silent PF synapses in Purkinje cells may imply that 85% or more of PF synapses are long-term depressed during repeated learning, by which the neurocomputing circuits are shaped through experiences from the original preponderantly connecting PF synapses (Dean et al., 2010).
28: Cerebellar Cortex
299
Internal Model A microcomplex is inserted into various microcircuits in the spinal cord and brainstem and acts as an embedded adaptive component in the motor or autonomic control systems. Furthermore, a microcomplex is connected to the cerebral cortical area and acts as an internal model of the latter. For the motor cortex, a microcomplex may provide a forward model or an inverse model of the lower motor centers and motor apparatus, which enables the motor cortex to perform a precise movement without the external feedback about the movement (Kawato, 2009). A similar idea has been expanded to cognitive functions; a microcomplex would provide an internal model for mental activities going on in the cerebral association cortex (Ito, 2008). Oscillation A computer simulation suggests that reciprocal inhibition causes an oscillation at 100–250 Hz in the activity of basket/stellate cells. A computer simulation also predicts that feedback inhibition from Golgi to granule cells induces 10–50 Hz oscillations in spike discharges from the latter. Because of the large divergence from Lugaro cells to Golgi cells, an interesting possibility is that Lugaro cells play a role in synchronizing activity among Golgi cells situated along the parallel fiber beam as observed in anesthetized rats. Lugaro cells may switch the operation of Golgi cells from the individual rhythmic mode to the synchronous mode. Golgi Cell Clock The loop connection involving granule cells and Golgi cells (Fig. 28.1) has been interpreted to constitute a phase converter, whose function is to generate a set of multiphase versions of a mossy fiber input. Another unique model proposed by Yamazaki and Tanaka (2005) features a randomly connected granule cell–Golgi cell loop pathway, which operates as a clock to generate granule cell discharge in an ever-changing ensemble of patterns. The patterns do not repeat unless reset by a large mossy fiber input to the pathway. When this Golgi cell clock is incorporated in a neuronal circuit model for eyeblink reflex, it reproduces appropriately timed conditioned responses.
References Alviña K, Walter JT, Kahn A, Ellis-Davis G, Khodakhah K (2008) Questioning the role of rebound firing in the cerebellum. Nat Neurosci 11:1256–1258. Bengtsson F, Jörntell H (2009) Sensory transmission in cerebellar granule cells relies on similarly coded mossy fiber inputs. Proc Natl Acad Sci 106:2389–2394.
300
Handbook of Brain Microcircuits
Dean. P, Porrill J, Ekerot C-F, Jöntell H (2010) The cerebellar microcircuit as an adaptive filter: experimental and computational evidence. Nat Rev Neurosci 11:30–43. Ito M (2006) Cerebellar circuitry as a neuronal machine. Prog Neurobiol 78:272–303. Ito M (2008) Control of mental activities by internal models in the cerebellum. Nat Rev Neurosci 9:304–313. Kawato M (2009) Cerebellum: models. In: Squire LR, ed. Encyclopedia of Neuroscience, pp. 757–767. Oxford, England: Academic Press. Elsevier Limited. Lainé J, Axelrad H. (2002) Extending the cerebellar Lugaro cell class. Neuroscience 115: 363–374. Nieto-Bona MP, Garcia Sergura LM, Torres-Aleman I (1997) Transsynaptic modulation by insulin-like growth factor I of dendritic spines in Purkinje cells. Int J Devl Neurosci 15:749–754. Steuber V, Mittmann W, Hoebeek FE, Angus Silver R, De Zeeuw CI, Häusser M, De Schutter E (2007) Cerebellar LTD and pattern recognition by Purkinje cells. Neuron 54:121–136. Strata P (2002) Dendritic spines in Purkinje cells. The Cerebellum 1:230–232. Walter JT, Khodakhah K (2009) The advantages of linear information processing for cerebellar computation. Proc Natl Acad Sci 17:4471–4476. Yamazaki T, Tanaka S (2005) Neural modeling of an internal clock. Neural Computation 17:1032–1058. Yamazaki T, Tanaka S (2007) The cerebellum as a liquid state machine. Neural Networks 20:290–297.
