Long-term Depression in Cerebellum Research Paper

1. Neuronal Elements Invol edvin LTD

Mossy fibers arising from various precerebellar nuclei in the brain stem and spinal cord enter the granular layer of the cerebellar cortex and form synaptic contact with granule cells there. Axons of granule cells ascend through the Purkinje cell layer to the molecular layer and bifurcate to parallel fibers, which extend 2–3 mm to each side of the bifurcation point through the extensive dendritic arbor of Purkinje cells. Estimates of the number of synapses that each Purkinje cell receives from parallel fibers vary from 60,000 to 175,000 (Ecles et al 1967, Palay and Chan-Palay 1974, Ito 2001). While parallel fibers make up the majority of granule cell inputs to Purkinje cells, the ascending axons of granule cells also contribute about 20 percent of the total granule cell inputs, preferentially on the smallest diameter distal regions of the Purkinje cell dendrites.

Another input to Purkinje cells, a climbing fiber, originates from the inferior olive in the medulla oblongata. In contrast to the convergence of numerous granule cell axons to a Purkinje cell, only one climbing fiber comes into contact with each Purkinje cell and forms numerous discrete synaptic junctions on stubby dendritic spines of the Purkinje cell. The number of junctions formed between a climbing fiber and a Purkinje cell has previously been estimated to be approximately 300 in the frog cerebellum. However, in rats, a much larger number (26,000) is suggested because a 100 µm length of PC dendrite contacts 11.45 granule cell axons and 1.7 climbing fibers on the average (Ito 2001).

2. Electrical Signals Invol edvin LTD

Impulses of granule cell axons evoke fast excitatory postsynaptic potentials (EPSPs) in Purkinje cells as mediated by AMPA receptor. These EPSPs, when sufficiently large, evoke Na+ spikes in the somatic region, which passively spread into the dendrites. During extracellular recording, these spikes are observed as ‘simple spikes,’ which are spontaneously discharged in i o at a rate of 50–100 Hz. Impulses in climbing fibers evoke responses composed of large EPSPs and Ca + spikes evoked by the EPSPs in Purkinje cell dendrites. During extracellular recording, climbing fiber responses take the form of short bursts of spikes, called ‘complex spikes,’ which are discharged in i o at irregular, low rates around 1 Hz.

During in i o extracellular recording, LTD is represented by the persistent reduction in the probability (firing index) of a Purkinje cell to generate simple spikes in response to the stimulation of granule cell axons. Under slice or tissue culture conditions, LTD is observed as a persistent reduction in the granule cell axon-evoked AMPA-mediated EPSPs or excitatory postsynaptic currents (EPSCs) (see Eccles et al. 1967, Ito 2000).

3. Conditions for Induction of LTD

LTD is optimally induced by conjunctive stimulation of climbing fibers and granule cell axons at 1–4 Hz by 100–300 pulses. LTD can also be induced by a more natural stimulus paradigm such as a series of eight pulses (80 msec, 100 Hz) to granule cell axons and immediately thereafter three pulses to climbing fibers (Ito 2000). LTD is input-specific; it occurs only in those granule cell–Purkinje cell synapses conjunctively stimulated by climbing fibers. However, spread of LTD to synapses in the close neighborhood has recently been reported (Reynolds and Hartell 2000, Wang et al 2000).

Various reduced forms of LTD have been generated by replacing either granule cell axon stimulation or climbing fiber stimulation, or both, with chemical stimuli or electrical currents directly applied to Purkinje cells. In cerebellar slices, climbing fiber stimulation can be replaced by the application of dep olarizing pulses, which bring about the entry of Ca + ions into Purkinje cells through voltage-gated channels. The simplest procedure for inducing LTD is to apply quisqualate or to increase the extracellular K+ concentration. In cultured Purkinje cells devoid of both granule cells and climbing fibers, a reduced form of LTD is induced by a combination of glutamate (or quisqualate) pulses and membrane depolarization. The LTD is detected as a reduced sensitivity of PCs to glutamate or AMPA. This form of LTD occurs in very simple preparations devoid of dendritic spines, indicating that LTD induction does not require such morphological specialization.

4. Duration of LTD

The observation time for LTD is usually limited to one half to one hour, and occasionally for two to three hours, but in certain special cases, it has been extended to one to two days (Ito 2000). LTD induced in the amplitude of mEPSCs in cultured PCs by conjunctive application of 50 mM K+ and 100 µM glutamate was observed to last for 36 hours and return to the original level after 48 hours. When the rabbit cerebellum was sliced 24 hours after the rabbit had been trained for eyeblink conditioning (Eyelid Classical Conditioning), sequentially applied granule cell axon and climbing fiber stimuli failed to induce LTD, which occurred in slices dissected from control rabbits. This suggests that LTD underlying the eyeblink conditioning persists for at least 24 hours and precludes eliciting another LTD.

5. Signal Transduction for LTD

Diverse chemical reactions have been found to underlie LTD induction (Levenes et al. 1998, Ito 2000). They can be roughly classified into the following seven major pathways:

(a) The glutamate released from granule cell axons reacts with AMPA receptors in Purkinje cells, which open associated cation channels for Na+ and K+ ions to generate AMPA-EP SPs. These EPSPs act to open voltage-dependent Ca + channels.

(b) Granule cell-derived glutamate also reacts with mGluR1s, which in turn activate phospholipase C through G protein (Gαq ). Phospholipase C produces diacylglycerol (DAG) which activates protein kinase C (PKC), as well as inositol trisphosphate (IP₃), which acts upon IP receptors to release Ca²⁺ ions from intracellular stores.

(c) mGluR1s activated by granule cell-derived glutamate also stimulate phospholipase A2 via G proteins, eventually producing unsaturated free fatty acids including arachidonic acid and oleic acid.

(d) Granule cell-released NO diffuses into Purkinje cells and activates guanylyl cyclase which produces cyclic GMP (cGMP). cGMP in turn activates protein kinase G (PKG), which, following phosphorylation, converts a protein, G-substrate, into a potent inhibitor of protein phosphatase.

(e) A climbing fiber-derived amino acid transmitter (most likely glutamate, yet unidentified) reacts with receptors and opens associated channels for Na⁺ and K⁺ ions to evoke a large EPSP, which in turn causes Ca²⁺ influx through voltage-dependent Ca²⁺ channels.

(f) A climbing fiber-derived corticotropin-releasing factor (CRF) reacts with type-1 CRF receptors and associated G proteins, and then yields an as yet unidentified second messenger, which interacts with PKC.

(g) A climbing fiber-derived type-1 insulin-like growth factor (IGF-1) may facilitate internalization of receptors by endocytosis.

There is cross-talk between these pathways. Protein tyrosine kinases (PTKs), mitogen-associated protein kinases (MAPKs), glial fibrillary acid protein (GFAP), an as yet unidentified rapidly turned-over protein(s) and immediate early gene, Jun-B, are also implicated in LTD induction. A recent hypothesis is that these complex chemical reactions eventually lead to removal of AMPA receptors by internalization (Wang and Linden 2000, Matsuda et al. 2000).

6. Error Representation by Climbing Fiber Signals

What information do climbing fibers convey to induce LTD? In simple situations such as reflexes, climbing fibers convey sensory signals such as pain or loud sounds. These sensory signals suggest a harmful consequence of an inadequately executed movement. In various types of ocular movements, climbing fiber signals are known to represent retinal slips, which reflect a deviation of a realized eye movement from a desired one, that is, an error in controlling eye movement. These observations support the view that, during the repeated exercise of a movement, errordriven LTD reshapes the neuronal circuit of the cerebellum in the direction which minimizes errors consequent to the movement (Ito 1984, 2001).

During locomotion, climbing fiber signals are evoked by a sudden descent of a stepped-on rung at the beginning, but not at the cessation, of the fall of a paw. These signals appear to reflect the discrepancy between command signals descending from the brain stem center to the spinal cord and the signals that the command signals evoke in the spinal cord neuronal circuit. During hand arm movements, climbing fiber signals arise in three phases of movement: at the beginning, before the end, and after. Climbing fiber signals therefore encode not only the consequence errors detected through sensory systems, but also other types of errors arising from the internal mechanisms of the motor systems.

7. In olvement of LTD in Motor Learning

Recent data that the inhibition of LTD following pharmacological or genetic manipulation impairs various types of motor learning supports the view that error-driven LTD is the basis of motor learning (Ito 2001).

(a) Adaptation of the vestibuloocular reflex (233) has been shown to be blocked by superfusing the cerebellum in rabbits and monkeys with a NO scavenger, hemoglobin, or by injection of an inhibitor of NO synthase (NOS) into the goldfish cerebellum.

Transgenic mice that selectively express the pseudosubstrate PKC inhibitor in Purkinje cells lacked the vestibuloocular reflex adaptation consistent with the loss of LTD induction in cerebellar slices obtained from these mice. Neural NOS (nNOS)-deficient mice lacked adaptation of the optokinetic eye movement response, which normally occurs under continued rotation of the visual field around a stationary animal. Subdural applications of NO scavenger or NOS inhibitor to the paraflocculus-flocculus depressed the adaptation of smooth pursuit in monkeys.

(b) Rabbits administered a NO synthase inhibitor exhibited learning deficits in the conditioned eyeblink response. GFAP-deficient mutant mice exhibited a loss of LTD and also impaired eyeblink conditioning. Injection of an IGF-1 antisense oligonucleotide in the inferior olive, which reduces IGF-1 levels in the cerebellum, blocked conditioned eyeblink learning in rats.

(c) Postural compensation after unilabyrinthectomy was delayed in rats after intracerebroventricular application of PKC inhibitors. When a decerebrate cat walking on a treadmill experienced a sudden increase in the speed of only the running belt under the left forelimb, regular stable locomotion was restored in 50–100 steps. Injection of a NOS inhibitor into the cerebellum blocked this adaptation. mGluR1-deficient mice walking on a treadmill exhibited abnormally dispersed locomotor cycles of two limbs and did not adapt to an increase in the belt velocity. This sign of impaired interlimb coordination in locomotion diminished when the mGluR1-deficiency was rescued in the cerebellum.

(d) Motor coordination is evaluated by measuring the duration that an animal can remain on a horizontally placed rotating rod. In five types of genemanipulated mice showing significant reduction in staying time, three types (mGluR1-, GluRδ2-, and Gαq-deficient) lacked LTD. However, the other two (PKCγ and mGluR4-deficient) retained LTD, and their motor discoordination is explained by the persistent multiple innervations of Purkinje cells by climbing fibers, abnormal transmission in granule cell–Purkinje cell synapses. GFAP-deficient mice and PKC-inhibitor-transfected mice were reported to exhibit motor coordination indistinguishable from wildtype mice in spite of the lack of LTD. These observations are inconsistent with the hypothesis of motor learning. The reasons for this discrepancy need to be investigated.

8. Comments

How is LTD eventually converted to permanent memory? One possibility is that LTD as a functional depression is consolidated as a structural change in synaptic contacts and spines. How can the error-driven LTD-based learning mechanism be expanded to apply beyond motor learning to implicit learning in mental functions such as cognition, language, and thought? This expansion has been achieved at experimental and conceptual levels, but in the future it should be made at computational levels as well.

Bibliography:

1. Eccles J C, Ito M, Szentagothai J 1967 The Cerebellum as a Neuronal Machine. Springer, Berlin
2. Ito M 1984 The Cerebellum and Neural Control. Raven, New York
3. Ito M 2000 Cerebellar long-term depression: Characterization, signal transduction and functional roles. Physiology Review 81: 1143–95
4. Levenes C, Daniel H, Crepel F 1988 Long-term depression of synaptic transmission in the cerebellum: Cellular and molecular mechanisms revisited. Progress in Neurobiology 55: 79–91
5. Matsuda S, Launey T, Mikawa S, Hirai H 2000 Disruption of AMPA receptor GluR2 clusters following long-term depression induction in cerebellar Purkinje neurons. EMBO Journal 19: 2765–74
6. Wang Y T, Linden D 2000 Expression of cerebellar long-term depression requires postsynaptic clathrin-mediated endocytosis. Neuron 25: 635–47
7. Palay S L, Chan-Palay V 1974 The Cerebellar Cortex. Springer, New York
8. Reynolds T, Hartell N A 2000 An evaluation of the synapse specificity of long-term depression induced in rat cerebellar slices. Journal of Physiology (London) 527: 563–77
9. Wang S S H, Khiroug L, Augustine G J 2000 Quantification of spread of cerebellar long-term depression with chemical twophoton uncaging of glutamate. Proceedings of the National Academy of Science, USA 97: 8635–40