The synaptic organization of the elaborated neuronal networks of the cerebellar cortex, composed of PCs, molecular layer interneurons (basket and stellate cells), Golgi cells, PFs from granule cells, mossy fibers and climbing fibers (CFs), have been elucidated in terms of excitation/inhibition and divergence/convergence, and the well-organized circuitry diagram has suggested that the cerebellum is a “neuronal machine”1, though there was previously no idea about purpose of this “machine”. Later Marr proposed that the PFs input to PCs constitute a triple layer associative learning network2. He also suggested that each CF conveys a cerebral instruction for elemental movement2. He assumed that simultaneous activation of PFs and CF would enhance PF-PC synapse activity, and cause long-term potentiation (LTP) of the PF-PC synapse. On the other hand, Albus assumed that synchronous activation of PFs and CF resulted in LTD at the PF-PC synapses3. Both the above studies interpret the cerebellum as a unique memory device, the incorporation of which into the cerebellar cortical network leads to the formation of the Marr–Albus model learning machine model.
Following these theoretical predictions, two lines of evidence suggest the presence of synaptic plasticity in the cerebellum. The first line of evidence was suggested by the anatomical organization of the flocculus; here MF pathways of vestibular organ origin and CF pathways of retinal origin converge on the PCs4. This unique convergence pattern suggests that a synaptic plasticity occurring in the flocculus causes the remarkable adaptability of the vestibulo-ocular reflex. Second, the recording of the PCs response in the flocculus and the lesioning of the flocculus also supported the above hypothesis5,6,7. Furthermore, the PC discharge pattern during adaptation of a monkey’s hand movement8 supported the synaptic plasticity hypothesis, especially Albus’s LTD-hypothesis3.
To determine the nature of the synaptic plasticity directly, repeated conjunctive stimulation (Cjs) of a bundle of PFs and the CF that specifically innervates the PC in vivo was shown to induce LTD for the transmission efficacy of the PF–PC synapses9,10,11. In the subsequent in vitro explorations using a cerebellar slice12 and cultured PCs, conjunction of co-cultured granule cell stimulation and olive cell stimulation13 or conjunction of iontophoretically applied glutamate and somatic depolarization14,15 caused LTD. The signal transduction mechanism underlying the LTD-induction was also intensively investigated using in vitro preparations16,17.
Adaptations of the VOR and the OKR were often used for quantitative evaluation of gene-manipulation effects on cerebellar motor learning, because the vestibule-cerebellar cortex was proven to be the essential origin in the adaptive learning of the VOR18,19,20 and the OKR19,21 The correlation between failure of LTD-induction and impairment of behavioral motor learning has been taken as evidence that LTD plays an essential role in motor learning mechanisms22. These views are collectively referred to as the LTD hypothesis of motor learning, or Marr-Albus-Ito hypothesis23,24,25,26.
Adaptive learning of eye movement was measured using similar protocols, while various experimental conditions were used to induce LTD in slice preparation27,28,29,30,31. Recently, Schonewille et al.26 reported that some gene-manipulated mice demonstrated normal motor learning, but the cerebellar slices did not show LTD, and thereby concluded that LTD was not essential for motor learning. However, the induction of LTD was only attempted using one type of protocol at room temperature. Hence, we used several types of LTD-inducing protocols under recording conditions at around 30 °C, and we confirmed that the LTD was reliably induced in the gene-manipulated mice by using these protocols at near physiological temperatures32.
However, there remain some questions regarding the basic properties of conjunctive stimulation. The first is the relationship between the complex spike’s shape and the amplitude of LTD. Second, in conjunction with PF-stimulation and somatic depolarization, whether the number of stimuli used were necessary or not was elusive. In the present study, these questions were investigated using wild type (WT) mice.