We implanted a small piece of ethylene-vinyl acetate Small Molecule Compound Library copolymer (Elvax) containing 3-MP or vehicle on the surface of the cerebellar vermis (lobules 6–8) at P10 as described previously (Kakizawa et al., 2000, Kakizawa et al., 2003 and Kakizawa

et al., 2005). When CF innervation was tested within the lobules 6–8 at P23–P40, 80% of PCs (45/56) in vehicle-treated mice were innervated by single CFs (Figure 4D). By marked contrast, 48% of PCs (30/62) from 3-MP-treated mice were multiply-innervated by two or three CFs (Figure 4D) with significant difference in the frequency distribution (p = 0.002) (Figure 4D). Basic synaptic properties of CF-EPSCs were identical between the 3-MP- and vehicle-treated mice (Table S2). These results indicate that suppression of GAD activity in the cerebellum starting at P10 effectively perturbs CF synapse elimination presumably by weakening GABAergic transmission. Next, we studied whether there is a critical GSI-IX period for the GAD-sensitive CF synapse elimination. We implanted Elvax containing 3-MP or vehicle to the cerebellar lobules 6–8 at P17. As shown in Figure 4E, the frequency distribution histograms were similar between the vehicle- and 3-MP-treated mice (p = 0.725). No difference was observed in CF-EPSC kinetics or

paired-pulse ratio between the two experimental groups (Table S2). These results suggest that the critical period for the GAD-sensitive CF synapse elimination is around P10 to P16. 3-MP inhibits not only

GAD67 but also the other GAD isoform GAD65, and both were expressed in the mouse cerebellum during the second postnatal week (Figures S4A–S4L). To test the contribution of GAD65 to CF-synapse elimination, we analyzed GAD65 knockout (GAD65 KO) mice (Asada et al., 1996). GAD65 KO mice have been reported to maintain normal levels of GAD67 and GABA in the brain (Asada et al., 1996). We found both the amplitude and frequency of mIPSCs were normal in GAD65 KO PCs (Figures S4M–S4O). In mature stage, there was no significant difference between wild-type and GAD65 KO mice in terms of the number of CFs innervating each PC (p = 0.406) (Figure S4P), indicating that MYO10 CF synapse elimination is normal in GAD65 KO mice. These data indicate that the effects of 3-MP on mIPSCs and CF synapse elimination are attributable to the inhibition of GAD67. Next, we examined whether pharmacological enhancement of GABAergic inhibition by diazepam, a benzodiazepine-site agonist, facilitates CF synapse elimination in GAD67+/GFP mice. We assayed the effects of diazepam on mIPSCs in cerebellar slices from GAD67+/GFP mice at P10–P14 (Figures 5A–5C). Diazepam significantly enhanced the amplitude of mIPSCs which returned to the control level by subsequent application of flumazenil, a benzodiazepine-site antagonist (Ctrl: 66 ± 2.3 pA; DZ: 88 ± 3.3 pA; DZ+Flu: 62 ± 2.1 pA, n = 8; p < 0.001, Ctrl versus DZ; p = 0.