Climbing fiber (CF) synapse formation onto cerebellar Purkinje cells (PCs) is critically dependent on the synaptogenesis from parallel fibers (PFs), the other input to PCs. Previous studies revealed that deletion of the glutamate receptor δ2 subunit (GluRδ2) gene results in persistent multiple CF innervation of PCs with impaired PF synaptogenesis, whereas mutation of the metabotropic glutamate receptor subtype 1 (mGluR1) gene causes multiple CF innervation with normal PF synaptogenesis. We demonstrate that atypical CF-mediated EPSCs (CF-EPSCs) with slow rise times and small amplitudes coexisted with typical CF-EPSCs with fast rise times and large amplitudes in PCs from GluRδ2 mutant cerebellar slices. CF-EPSCs in mGluR1 mutant and wild-type PCs had fast rise times. Atypical slow CF responses of GluRδ2 mutant PCs were associated with voltage-dependent Ca2+ signals that were confined to PC distal dendrites. In the wild-type and mGluR1 mutant PCs, CF-induced Ca2+ signals involved both proximal and distal dendrites. Morphologically, CFs of GluRδ2 mutant mice extended to the superficial regions of the molecular layer, whereas those of wild-type and mGluR1 mutant mice did not innervate the superficial one-fifth of the molecular layer. It is therefore likely that surplus CFs of GluRδ2 mutant mice form ectopic synapses onto distal dendrites, whereas those of wild-type and mGluR1 mutant mice innervate proximal dendrites. These findings suggest that GluRδ2 is required for consolidating PF synapses and restricting CF synapses to the proximal dendrites, whereas the mGluR1-signaling pathway does not affect PF synaptogenesis but is involved in eliminating surplus CF synapses at the proximal dendrites.
- climbing fiber
- parallel fiber
- Purkinje cell
- glutamate receptor
- postnatal development
- mutant mouse
In the adult cerebellum, each Purkinje cell (PC) is innervated by a single climbing fiber (CF) that originates in the inferior olive and forms strong excitatory synapses onto proximal PC dendrites (Palay and Chan-Palay, 1974; Ito, 1984;Strata and Rossi, 1998). This one-to-one relationship is preceded by a developmental stage of redundant innervation of PCs by multiple CFs (Crépel, 1982). Elimination of supernumerary CFs proceeds in parallel with synaptogenesis onto PCs from parallel fibers (PFs), bifurcated axons of granule cells and the other excitatory input to PCs (Ito, 1984). The importance of PF–PC synapses in CF synapse elimination has been substantiated by the fact that multiple CF innervation persists in animals with impaired PF–PC synaptogenesis, such as x-irradiated rats (Woodward et al., 1974; Crépel and Delhaye-Bouchaud, 1979; Bravin et al., 1995; Sugihara et al., 2000) and agranular cerebellar mutants (weaver, reeler, andstaggerer mice) (Crépel and Mariani, 1976; Mariani et al., 1977; Crépel et al., 1980; Mariani and Changeux, 1980).
Glutamate or its related amino acids is considered to be released from CF and PF terminals and to activate glutamate receptors (GluRs) on PCs. The GluR δ2 subunit (GluRδ2) is highly enriched in PCs (Araki et al., 1993; Lomeli et al., 1993) that are localized at the junctional site of PF–PC synapses but not at the site of CF–PC synapses (Landsend et al., 1997). In PCs of GluRδ2-deficient mice, the number of PF–PC synapses is reduced to nearly one-half that of wild-type mice, leaving approximately one-third of PC dendritic spines free of innervation (Kurihara et al., 1997). GluRδ2-deficient mice display persistent multiple CF innervation, impaired long-term depression (LTD), and motor discoordination (Hirano et al., 1995; Kashiwabuchi et al., 1995). The metabotropic GluR subtype 1 (mGluR1) is also abundant in PCs that are localized at the perijunctional site of both the PF–PC and CF–PC (Baude et al., 1993; Nusser et al., 1994). Inactivation of the mGluR1 gene results in persistent multiple CF innervation, impaired LTD, and motor discoordination, without apparent defects in PF–PC synaptogenesis (Aiba et al., 1994; Conquet et al., 1994; Kano et al., 1997; Lévénès et al., 1997; Ichise et al., 2000).
To differentiate the roles of these two GluR molecules in CF synapse formation and elimination, we comparatively examined the GluRδ2 and mGluR1 mutant mice by recording CF-mediated EPSCs (CF-EPSCs), imaging CF-induced Ca2+ signals in PCs, and labeling CFs anterogradely. We present evidence that surplus CFs form ectopic synapses onto distal dendrites of GluRδ2 mutant PCs, whereas CFs innervate proximal dendrites of mGluR1 mutant PCs. These findings suggest that GluRδ2 is required for consolidating PF synapses and restricting CF synapses to the proximal dendritic domain, whereas the mGluR1-signaling pathway is involved in eliminating excess CF synapses at the proximal dendrites.
MATERIALS AND METHODS
Animals. The GluRδ2 mutant mice (Kashiwabuchi et al., 1995) and the mGluR1 mutant mice (Aiba et al., 1994) were produced as described. The GluRδ2 mutant mice and their littermates were of C57BL/6 × TT2 genetic background (Yagi et al., 1993). The mGluR1 mutant mice and their littermates were of the 129/Sv × C57BL/6 genetic background. Mice were kept in the same room at the animal facility with a 12 hr light/dark cycle. Because we have not found any electrophysiological or morphological difference so far between the wild-type (+/+) mice of C57BL/6 × TT2 genetic background and those of 129/Sv × C57BL/6 genetic background,nonmutant (+/+) littermates including both genotypes were used in electrophysiological analyses as wild-type controls.
Electrophysiology. Sagittal cerebellar slices of 200–250 μm thickness were prepared from wild-type, GluRδ2 mutant, or mGluR1 mutant mice as described previously (Edwards et al., 1989; Llano et al., 1991; Kano and Konnerth, 1992; Aiba et al., 1994). Whole-cell recordings were made from visually identified PCs using a 40× water-immersion objective attached to either an Olympus (BH-2 or BX50WI, Olympus Optical, Tokyo, Japan) or a Zeiss (Axioskop) upright microscope (Edwards et al., 1989; Llano et al., 1991) at 32°C. The resistance of patch pipettes was 3–6 MΩ when filled with an intracellular solution composed of (in mm): 60 CsCl, 30 Csd-gluconate, 20 TEA-Cl, 20 BAPTA, 4 MgCl2, 4 ATP, and 30 HEPES; pH 7.3, adjusted with CsOH. The composition of standard bathing solution was (in mm): 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgSO4, 1.25 NaH2PO4, 26 NaHCO3, and 20 glucose, bubbled continuously with a mixture of 95% O2 and 5% CO2. Bicuculline (10 μm) was always present in the saline to block spontaneous IPSCs (Konnerth et al., 1990; Kano et al., 1992). Ionic currents were recorded with either an Axopatch-1D (Axon Instruments, Foster City, CA) or an EPC-9 (HEKA, Lambrecht, Germany) patch-clamp amplifier. The pipette access resistance was compensated as described by Llano et al. (1991). Stimulation and on-line data acquisition were performed using the PULSE software on a Macintosh computer (HEKA). Signals were filtered at 3 kHz and digitized at 20 kHz. The decay phase of EPSCs was fitted with the PULSE-FIT software (HEKA). A glass pipette with a 5- to 10-μm-diameter tip that was filled with standard saline was used for stimulating CFs. To search for multiple CFs innervating the recorded PC, the stimulation pipette was moved systematically in the granule cell layer, and the stimulus intensity was gradually increased at each stimulation site (pulse width 0.1 msec; strength 0–100 V).
Ca2+ imaging.PCs were loaded for at least 20 min with a Ca2+ indicator (Oregon Green 488 BAPTA-1, Molecular Probes, Eugene, OR; 100 μm) through the patch pipette at room temperature. A high-speed confocal laser-scanning microscope (Noran Odyssey or Oz) attached to an upright microscope (Zeiss Axioskop or Olympus BX50WI) was used to acquire fluorescence images at 30 or 60 Hz in parallel to the whole-cell recordings (Eilers et al., 1995a; Takechi et al., 1998). Full-frame images were recorded on an optical disk (TQ2026F, Panasonic) and analyzed off-line with the Image-1 software (Universal Image). In some experiments, fluorescence images were acquired at 30 Hz by using a cooled CCD camera system (IMAGO, T.I.L.L. Photonics, Gräfelfing, Germany) attached to an upright microscope (Zeiss Axioskop). The Ca2+-dependent fluorescence signals from selected regions of interest were background corrected and expressed as increases in fluorescence divided by the prestimulus fluorescence values (ΔF/F 0) using Igor Pro software (Wavemetrics). In PCs with multiple CF innervation, it was often difficult to detect CF-induced Ca2+signals under normal resting potentials with an extracellular Ca2+ concentration of 2 mm. This was presumably because the CF-induced EPSPs were too small to evoke spikes or, even if they induced spikes, the associated Ca2+ entry was too small to be detected. In these cases, we depolarized the PCs by constant current injection so that CF-EPSPs evoked spikes and increased the extracellular Ca2+ concentration to 3 mm.
Anterograde labeling of CFs. Under deep anesthesia with chloral hydrate (350 mg/kg of body weight), a glass pipette (inner tip diameter 10–20 μm) filled with 2–3 μl of a 10% solution of biotinylated dextran amine (10,000 molecular weight; Molecular Probes) in 0.01 m PBS, pH 7.4, was inserted into the right inferior olive by dorsal approach (Rossi et al., 1995). Biotinylated dextran amine was injected iontophoretically by positive current (7 μA for 30 min with a protocol of 700 msec on and 1300 msec off). After 6–8 d of survival, mice were anesthetized with chloral hydrate (350 mg/kg) and perfused transcardially with 4% paraformaldehyde in 0.1 m phosphate buffer, pH 7.4. Coronal brainstem sections were examined to check the injection site of biotinylated dextran amine. Mice with the injection site restricted to the caudal part of the medial accessory olive were chosen for the preparation of microslicer cerebellar sections (50 μm in thickness). Parasagittal cerebellar sections were incubated overnight in avidin–peroxidase (Amersham Biosciences, Buckinghamshire, UK) diluted with PBS containing 1% Tween 20 and visualized with diaminobenzidine and cobalt. Low-power photographs were taken with a Normarski interference contrast microscope (Axiophoto, Zeiss), whereas high-power ones were taken with a bright-field microscope (AX-80,Olympus). To quantity the extent of arbors of labeled CFs, the relative vertical height of the most distal tip was measured for at least 30 CFs in each mouse and expressed as the percentage of the molecular layer thickness. In each genotype, the mean value and SEM were obtained from three mice, and the difference between the genotypes was tested by Student's t test.
A novel slow time course CF-EPSC in GluRδ2 mutant PCs
In parasagittal cerebellar slices, PCs were recorded in the whole-cell configuration, and CFs were stimulated with a glass pipette placed in the granule cell layer near the recorded PCs (Aiba et al., 1994; Kano et al., 1995, 1997). To search PCs with multiple CF innervation, the stimulation pipette was systematically moved by ∼20 μm steps, and the stimulus intensity was gradually increased at each stimulation site (pulse width 0.1 msec; strength 0–100 V) (Kano et al., 1995, 1997, 1998; Offermanns et al., 1997; Watase et al., 1998;Ichise et al., 2000). EPSCs were elicited in an all-or-none fashion in most of wild-type PCs (Fig.1 A), whereas EPSCs with more than one discrete step were recorded in mGluR1 mutant PCs (Fig.1 B) with much higher incidence than in wild-type PCs (Kano et al., 1997; Lévénès et al., 1997; Ichise et al., 2000). These EPSCs displayed clear paired-pulse depression (PPD) to a stimulus pair with varying interpulse intervals (Fig.2 A,B,wild-type) (Konnerth et al., 1990; Aiba et al., 1994;Kano et al., 1995, 1997; Hashimoto and Kano, 1998).
In the GluRδ2 mutant mice, stimulation in the granule cell layer elicited multiple EPSCs in 88 of 122 PCs (Fig. 1 C). Each of these multiple EPSCs displayed clear PPD to a similar or greater extent than that of wild-type CF-EPSCs (Fig.2 A,B, GluRδ2 mutant). These multiple EPSCs in GluRδ2 mutant PCs were judged to be elicited by stimulating CFs and not by activating mossy fiber–granule cell-PF circuits, because EPSCs elicited by PF stimulation display clear paired-pulse facilitation in GluRδ2 mutant PCs as in wild-type PCs (Kashiwabuchi et al., 1995; Kurihara et al., 1997). These results indicate that most of the GluRδ2 mutant PCs were innervated by multiple CFs.
In the multiple CF-EPSCs of the GluRδ2 PC shown in Figure1 C, the time course was significantly slower for the smaller two EPSCs than for the largest EPSC. The 10–90% rise time of the wild-type CF-EPSCs displayed a normal distribution with a peak around 0.5 msec (0.46 ± 0.11 msec; mean ± SD; n = 105) (Fig. 1 D). That of the mGluR1 mutant had a similar distribution (0.57 ± 0.58 msec; mean ± SD;n = 88) (Fig. 1 E). In contrast, the rise time of GluRδ2 mutant CF-EPSCs segregated into two distinct populations (Fig. 1 F). The mean value for the fast (<1.0 msec) population was 0.48 msec (±0.15 msec SD;n = 127), which was similar to wild-type, whereas that for the slow (>1.0 msec) population was 1.81 msec (±0.59 msec SD;n = 80). These results suggest that GluRδ2 mutants have two distinct populations of CF inputs in terms of the EPSC rise time. The CF-EPSCs with fast rise time were similar to those of wild-type and mGluR1 mutant mice, whereas the slower CF-EPSCs appear to be unique to GluRδ2 mutant PCs. We also measured the decay time constant of CF-EPSCs in the three strains of mice by fitting the CF-EPSC decay with single exponentials (Llano et al., 1991; Aiba et al., 1994; Kano et al., 1995, 1997). The decay time constants of CF-EPSCs in wild-type and GluRδ2 mutant mice were not significantly different, being 6.73 msec (±2.04 msec; n = 84) for wild-type and 6.48 msec (±2.92 msec; n = 174) for the GluRδ2 mutant. The decay time constants were 4.75 msec (±2.36 msec;n = 88) for the mGluR1 mutant, which is significantly shorter than wild-type mice as reported previously (Aiba et al., 1994;Kano et al., 1997).
The GluRδ2 mutant PCs could be classified into four categories in terms of the pattern of CF innervation. These are as follows: (1) PC innervated by a single CFs that generates typical fast CF-EPSCs (Fig.3 A); (2) PC multiple-innervated by more than one CF that generates fast CF-EPSCs (Fig. 3 B); (3) PC multiple-innervated by one CF that generates fast CF-EPSCs and one or more CFs that generate slow CF-EPSCs (Fig. 3 C); and (4) PC multiple-innervated by more than one CF that generates fast CF-EPSCs and one or more CFs that generate slow CF-EPSCs (Fig. 3 D). Frequency distribution of all PCs in terms of number of CF-EPSC steps indicates that 72.1% of PCs in GluRδ2 mutant mice were innervated by multiple CFs (Fig.3 E). The histogram constructed only from PCs without slow CF-EPSCs indicates that 36.2% of such PCs were innervated by two CFs and 5.2% by three CFs (Fig. 3 F). On the other hand, the histogram constructed only from PCs with slow CF-EPSCs shows that most of the PCs were multiple-innervated (Fig. 3 G). In particular, 68.7% are innervated by three or more CFs (Fig.3 G). These results indicate that the existence of the slow CF-EPSCs significantly contributes to the high percentage of multiple-innervated PCs in the GluRδ2 mutant mice.
Slow CF-EPSCs of GluRδ2 mutant PCs appear during the second postnatal week
In rodents, PF synaptogenesis onto PC dendrites is known to continue postnatally up to approximately postnatal day (P) 20 (Altman and Bayer, 1997). The structure of the PF–PC synapse is immature around P7, when no significant differences in the morphology and electrophysiology are found between the wild-type and GluRδ2 mutant mice (Kurihara et al., 1997). Impaired PF synapse formation is obvious in the GluRδ2 mutant mouse at P14 (Kurihara et al., 1997). In contrast, PF–PC synapse formation appears normal in mGluR1 mutant mice (Kano et al., 1997). We followed the developmental course of CF-EPSCs in the three strains of mice to examine whether the impaired PF synapse formation correlated with the appearance of atypical slow CF-EPSCs in GluRδ2 mutant mice.
The 10–90% rise time of the wild-type CF-EPSC was longer during P3–P9 than at later developmental stages (Fig.4 A) (1.0 ± 0.4 msec; mean ± SD; n = 43). The rise times of the mGluR1 mutant and GluRδ2 mutant CF-EPSCs from P3 to P9 had a similar distribution (1.0 ± 0.5 msec, n = 71 for the mGluR1 mutant; 0.9 ± 0.4 msec, n = 64 for the GluRδ2 mutant) (Fig. 4 B,C) to that of wild-type. During P10–P14, the rise times of the wild-type and mGluR1 mutant CF-EPSCs decreased to adult levels (Fig.4 A,B). In contrast, CF-EPSCs with significantly slower rise times appeared during P10–P14 in the GluRδ2 mutant (Fig. 4 C). These data indicate that atypical slow CF-EPSCs become evident during the second postnatal week when a defect in the PF to PC synapse formation occurs in the GluRδ2 mutant mouse.
Electrophysiological properties of slow CF-EPSC in GluRδ2 mutant PCs
To examine whether the CF-EPSC amplitudes of mGluR1 mutant and GluRδ2 mutant mice differ from those of wild-type mice, the peak amplitudes of CF-EPSCs were measured at a fixed holding potential (−10 mV). In cases of multiple-innervated PCs (Fig.1 B,C), the amplitude of a given CF-EPSC was measured by digitally subtracting other CF-EPSCs that were induced by lower stimulus strength. The peak CF-EPSC amplitude was plotted against the 10–90% rise time. CF-EPSCs in wild-type (Fig.5 A) and mGluR1 mutant (Fig.5 B) mice were in a single population with only a few CF-EPSCs with slow rise time and small amplitude. In contrast, two distinct populations of CF-EPSCs, those with fast rise time and relatively large amplitude and those with slow rise time and small amplitude, were present in GluRδ2 mutant mice (Fig.5 C).
We examined other electrophysiological and pharmacological properties of atypical slow CF-EPSCs of GluRδ2 mutant PCs. They displayed a linear current–voltage (I–V) relationship (data not shown). In Mg2+-free saline containing glycine (10 μm), atypical slow CF-EPSCs were not affected by an NMDA receptor antagonist,dl-2-amino-5-phosphonopentanoate (50 μm) but were totally suppressed by a non-NMDA antagonist, 6-cyano-7-nitroquinoxaline-2,3-dione (10 μm) (data not shown). These results indicate that atypical CF-EPSCs of GluRδ2 mutant PCs are mediated exclusively by non-NMDA receptors, which is similar to typical CF-EPSCs of wild-type and those of GluRδ2 PCs (Kashiwabuchi et al., 1995).
Passive membrane properties of GluRδ2 mutant PCs
It has been shown morphologically that CF synapses are located on the proximal dendrites of PCs, whereas PF synapses oppose spines at the distal dendrites (Palay and Chan-Palay, 1974; Ito, 1984; Strata and Rossi, 1998). The 10–90% rise times of PF-EPSCs are significantly slower than those of CF-EPSCs in both the rat (Llano et al., 1991) and mouse (Aiba et al., 1994; Kano et al., 1995, 1997), suggesting that the EPSC rise time in PCs reflects the electrotonic length from the soma to the site of the synapse. It is likely, therefore, that the atypical slow CF-EPSCs in GluRδ2 mutant PCs arise from synapses located electrotonically distant from the soma, presumably on the distal dendrites. Because the EPSC rise time depends on the passive membrane properties, we compared several parameters of wild-type and GluRδ2 mutant PCs in response to hyperpolarizing voltage steps from −70 to −80 mV. As reported previously in rat PCs (Llano et al., 1991), the decay of the current was biphasic and could be described by the sum of two exponentials in both wild-type and GluRδ2 mutant PCs (data not shown). The time constants for the fast (τ1) and slow (τ2) components were 0.35 ± 0.04 and 13.4 ± 1.9 msec (mean ± SD; n = 9) for the wild-type and 0.33 ± 0.11 and 11.6 ± 2.0 msec (n = 11) for the GluRδ2 mutant. Both τ1 and τ2 values were not significantly different between the wild-type and the GluRδ2 mutant (p> 0.05; Student's t test). We then calculated several parameters representing passive properties of PCs based on the model equivalent circuit of PCs described by Llano et al. (1991) (Table1). As shown in Table 1, the lumped dendritic capacitance (C2) was smaller (p < 0.01; Student's t test) and lumped resistances were larger (p < 0.05 for R2; p < 0.01 for R3; Student's t test) in the GluRδ2 mutant than in the wild-type mice. These results suggest that the average total membrane area of the dendritic tree of the GluRδ2 mutant PCs is smaller than that of wild-type PCs. This is consistent with morphological data that the molecular layer of the GluRδ2 mutant is thinner than that of the wild-type cerebellum (Kurihara et al., 1997). These results indicate that the slow rise times of a subpopulation of GluRδ2 mutant CF-EPSCs cannot be explained by the altered passive properties of the PC dendrites.
Imaging CF-induced Ca2+ transients suggests that slow CF-EPSCs of GluRδ2 mutant PCs arise from synapses located at distal dendrites
We measured Ca2+ signals associated with CF responses by using a high-speed confocal laser-scanning microscope or a cooled CCD camera system. Whole-cell recording from PCs was conducted simultaneously with the Ca2+imaging. The Ca2+ signals were measured in small regions at three different levels of PC dendritic arborization: the primary smooth branch, the tertiary smooth branch, and terminal branchlets. In the following figures, the primary and tertiary smooth branches are collectively termed “proximal dendrites,” and the terminal branchlets are termed “distal dendrites.” In wild-type animals, it is shown morphologically that CFs form extensive synaptic contacts onto proximal dendrites, whereas PFs form en passant synapses onto spines of distal dendrites (Palay and Chan-Palay, 1974; Ito, 1984; Strata and Rossi, 1998).
In the mono-innervated wild-type PC (Fig.6 A), stimulation of the CF induced a typical complex spike (Fig. 6 B) that accompanied a clear Ca2+ signal over the entire dendritic tree (Fig. 6 C,D). The Ca2+ rose in both proximal (R1,R2) and distal (R3, R4) dendrites (Fig. 6 C,D). In the double-innervated mGluR1 mutant PC (Fig. 7 A), stimulation of each CF induced a large EPSP with spike(s) (Fig.7 B,E) and induced a clear Ca2+ transient (Fig.7 C,F). Stimulation of one CF (CF1) caused elevation of Ca2+that was confined to the left dendritic tree (Fig.7 C,D), whereas stimulation of the other CF (CF2) caused Ca2+ transients in the right dendritic tree (Fig.7 F,G). It should be noted that Ca2+ was elevated in both proximal and distal portions of each dendritic tree. In 14 multiple-innervated mGluR1 mutant PCs, CF-induced Ca2+ signals always involved proximal dendrites. In 16 mono-innervated PCs from mGluR1 mutant mice, the pattern of CF-induced Ca2+ signals was identical to that of the mono-innervated wild-type PCs (data not shown). These results suggest that mGluR1 mutant PCs are innervated by CFs at proximal dendrites as wild-type PCs whether they are mono- or multiple-innervated. The Ca2+ signals in the distal dendrite presumably reflected the spread of depolarization and activation of voltage-gated Ca2+ channels.
In GluRδ2 mutant PCs, we detected different pattern of CF-induced Ca2+ signals from those seen in wild-type or mGluR1 mutant PCs. The GluRδ2 mutant PC in Figure8 was innervated by at least three distinct CFs. Stimulation of one CF (CF1) generated typical complex spike (Fig. 8 B), whereas stimulation of other two CFs (CF2 and CF3) elicited slow responses (Fig. 8 E,H). Stimulation of CF1 induced Ca2+ transients that spread over the entire dendritic tree involving both proximal and distal dendrites (Fig. 8 C,D,R1–R6). In contrast, stimulation of CF2 induced Ca2+ transients that were confined to the distal portions (R4, R5) of the left dendritic branch (Fig. 8 F,G). Stimulation of CF3 induced small Ca2+transients that were confined to the most distal dendrites (R6) near the pial surface (Fig.8 I,J). Another GluRδ2 mutant PC shown in Figure 9 was innervated by at least three distinct CFs that generated a fast response (Fig. 9 B, CF1) and slow responses (Fig.9 E,H, CF2 andCF3). Stimulation of CF1 induced Ca2+ transients that spread over the middle dendritic tree and spared the small trees in the left and the right branches (Fig. 9 C,D). The Ca2+ rose in both proximal (R1,R2) and distal (R4, R5) portions of the middle dendritic tree (Fig. 9 C,D). Stimulation of CF2 induced Ca2+ transients that were confined to the distal portions (R5) of the right dendritic tree and spared its proximal (R1, R2) portions (Fig. 9 F,G). Stimulation of CF3 induced Ca2+ transients that were confined to the left dendritic branch (Fig.9 I,J). The Ca2+ rose slightly in the tertiary dendrite (R3) and clearly in distal dendrites (R6) but showed no change in the primary dendrite (R1). In 24 GluRδ2 mutant PCs, we found that slow CF responses were always associated with Ca2+transients that were confined to the distal dendrites. In a few slow CF responses, small Ca2+ signals could be detected also in the tertiary dendrites (Fig. 9, CF3); however, the Ca2+ transients never involved the primary dendrites. This pattern of CF-induced Ca2+ transients was observed in no PCs from the wild-type and in only one PC from the mGluR1 mutant.
To quantify the differences in CF innervation patterns, the distance from the soma to the dendrites with detectable Ca2+ signals was measured along the main dendrites (Fig. 10). As exemplified in Figure 10 A, the actual dendritic branches were traced, and the shortest distance was measured from the origin of the primary dendrite to the most proximal point of the dendritic tree where the Ca2+ signal was detected. In CF responses of wild-type and mGluR1 mutant mice, the distance was almost 0 μm (Fig. 10 B), indicating that the Ca2+ rose in the primary dendrite. In GluRδ2 mutant mice, the length was almost 0 μm for the fast response, whereas it was ∼50 μm on an average for the slow response (Fig. 10 B). These results suggest that atypical slow CF responses of GluRδ2 PCs arise from synapses located on the distal dendrites.
Morphology of CFs from GluRδ2- and mGluR1-deficient mice
To investigate the CF innervation patterns morphologically, anterogradely labeled CFs were visualized in parasagittal cerebellar sections (Fig. 11). Genotypic differences were noted in the extent of CF arbors within the molecular layer. Labeled CFs of GluRδ2 mutant mice extended up to more superficial regions of the molecular layer than those of wild-type and mGluR1 mutant mice (Fig. 11 A–C). Because the thickness of the molecular layer is slightly reduced in GluRδ2 mutant mice (Kurihara et al., 1997), we evaluated the extent of CF arbors by measuring the height of the most distal tip of each labeled CF relative to the thickness of the molecular layer. We selected cerebellar sections from each mouse that contained staining of entire CF arbors from the Purkinje cell layer to the top of the molecular layer. Then, we examined at least 30 labeled CFs in the straight portion of the lobule VI. The most distal tip of CFs in GluRδ2 mutant mice reached 95.0 ± 0.1% (mean ± SEM; n = 3) of the molecular layer thickness. This score was significantly higher than that of the wild-type (83.1 ± 0.5%;n = 3; p < 0.0001) or the mGluR1 mutant (77.8 ± 0.7%; n = 3; p < 0.0001) mouse. Our previous study indicates that PC dendrites extend to the surface of the molecular layer in GluRδ2 mutant mice as well as wild-type mice (Kurihara et al., 1997). Taken together, these results suggest that CFs of GluRδ2 mutant mice innervate more distal portions of PC dendrites than those of wild-type and mGluR1 mutant mice.
Distinct CF innervation patterns in GluRδ2 mutant and mGluR1 mutant PCs
GluRδ2 mutant PCs had atypical CF-EPSCs with slow 10–90% rise times that coexisted with typical fast rise time CF-EPSCs, suggesting that slow CF-EPSCs arise from synapses located electrotonically distant from the soma. Imaging of CF-induced Ca2+signals through voltage-gated Ca2+channels revealed that GluRδ2 mutant PCs had atypical small CF responses with Ca2+ signals that were confined to PC distal dendrites. In contrast, CF-induced Ca2+ signals involved both proximal and distal dendrites in wild-type PCs. Morphologically, CFs of GluRδ2 mutant mice extended to the superficial regions of the molecular layer, whereas CFs of wild-type mice extended to the four-fifths of the molecular layer and spared the external one-fifth. On the other hand, the CF innervation pattern of mGluR1 mutant mice resembled that of the wild-type mice in the following respects: (1) CF-EPSCs had fast rise times (≤1 msec), (2) CF-induced Ca2+signals involved proximal dendrites, and (3) anterogradely labeled CFs extended to the four-fifths of the molecular layer and spared the external one-fifth.
Previous electron microscopic studies in mouse and rat indicate that CFs form extensive synaptic contacts on protuberance of the proximal dendrites of PCs, whereas PFs form en passant synapses on spines of the distal dendrites of PCs (Larramendi and Victor, 1967;Palay and Chan-Palay, 1974; Ito, 1984). Staining of CFs in rats also revealed that CFs innervate PC proximal dendrites (Strata and Rossi, 1998; Sugihara et al., 1999; Kreitzer et al., 2000). The range of CF-induced Ca2+ signals, however, does not precisely reflect the location of CF synapses on PC dendritic trees. Previous studies consistently indicate that CF-induced Ca2+ transients spread over the entire dendritic tree of PCs involving both proximal and distal dendrites (Ross and Werman, 1987; Kano et al., 1992; Konnerth et al., 1992;Miyakawa et al., 1992; Eilers et al., 1995b). The Ca2+ transients in distal dendrites are thought to reflect the spread of depolarization from proximal dendrites. CF-EPSPs are generated at proximal dendrites of PCs. However, because they are quite large in size, they can readily activate voltage-gated Ca2+ channels not only at proximal dendrites but also at distal dendrites. In contrast, the Ca2+ transients induced by PF stimulation are confined to the distal dendrites, and the range of the Ca2+ signals reflects the site of PF synapses (Eilers et al., 1995a; Finch and Augustine, 1998; Takechi et al., 1998; Wang et al., 2000). This is presumably because PF-EPSPs are relatively small in size and can only activate voltage-gated Ca2+ channels locally at the site of PF-EPSP generation. The atypical CF responses in GluRδ2 mutant mice were also relatively small in size and presumed to induce Ca2+ signals that reflect the site of their generation. In these respects, the atypical CF responses resemble PF-mediated responses and presumably reflect synaptic activity at PC distal dendrites.
Taken together, the results of the present study suggest that surplus CFs persist in GluRδ2 mutant mice by forming ectopic synapses onto distal dendrites of PCs, whereas CFs of mGluR1 mutant mice form synapses onto PC proximal dendrites similar to wild-type mice.
Roles of GluRδ2 in PF and CF synapse formation
Our previous study in the GluRδ2 mutant indicates that the abnormality of PF synaptogenesis is not found at P7 but is evident at P14 and thereafter (Kurihara et al., 1997). Thus, the appearance of atypical slow CF-EPSCs coincides temporally with the manifestation of impaired PF synaptogenesis. Previous studies indicate that the developmental change in CF from multiple- to mono-innervation of cerebellar PCs depends on the presence of intact granule cells and the formation of PF to PC synapses (Woodward et al., 1974; Crépel and Mariani, 1976; Mariani et al., 1977; Crépel and Delhaye-Bouchaud, 1979; Crépel et al., 1980; Mariani and Changeux, 1980; Mariani et al., 1990; Bravin et al., 1995; Sugihara et al., 2000). Thus, impairment of PF to PC synapse formation in GluRδ2 mutant mice (Kashiwabuchi et al., 1995; Kurihara et al., 1997) can cause persistent multiple CF innervation. In wild-type mice, massive formation of PF to PC synapses occurs onto dendritic spines of PC distal dendrites during the second and third postnatal weeks. CF synapses may not be maintained at distal dendrites and become confined to the thick proximal dendrites. Then, a competition may follow between the remaining CFs for postsynaptic sites on the proximal dendrites, and finally, a single CF may survive to innervate the PC. In GluRδ2 mutant mice, CF synapses may remain at distal dendrites because of the impairment of PF to PC synapse formation during the second and third postnatal weeks. However, the competition may occur between CFs for postsynaptic sites on the proximal dendrites, and in the majority of PCs, one CF may win to form extensive synapses on the proximal dendrites. As a consequence, many PCs lacking GluRδ2 have one typical CF that generated fast CF-EPSCs and additional atypical CFs that generated slow CF-EPSCs.
A previous study indicates that the δ receptor immunoreactivity is prominent at CF synapses at P10 and P14 and then becomes weaker at later developmental stages (Zhao et al., 1998). On the other hand, δ receptors are abundant on postsynaptic membranes at PF synapses from P10 to adult (Takayama et al., 1995, 1996; Landsend et al., 1997; Zhao et al., 1998). Thus, the developmental stage at which the impaired PF synapse formation becomes obvious in GluRδ2 mutant mice corresponds well with the period during which the δ receptors are transiently expressed at CF synapses. These results suggest that GluRδ2 may also play active roles in CF synapse formation during P10–P14 in addition to stabilizing PF synapses.
It is reported that the spontaneously occurring ataxic mouse mutationlurcher is a gain-of-function mutation in the GluRδ2 gene (Zuo et al., 1997). Mutated GluRδ2 channels behave as depolarizing leak channels independent of glutamate. Consequently, PCs display a very high membrane conductance and are constitutively depolarized and, perhaps, permit continuous Ca2+ inflow. This appears to disturb various signal transduction cascades within PCs, including the pathway from mGluR1 to the γ isoform of protein kinase C (PKCγ) that is shown to be required for CF synapse elimination (Kano et al., 1995, 1997, 1998; Offermanns et al., 1997;Hashimoto et al., 2000). This may be a cause of persistent multiple CF innervation in lurcher mutant mice (Rabacchi et al., 1992). We studied the kinetics of CF-EPSCs in lurcher mice, but atypical slow CF-EPSCs abundant in GluRδ2 mutant mice were not found (K. Hashimoto and M. Kano, unpublished observation).
Roles of mGluR1 in CF synapse elimination
We have reported that multiple CF innervation persists into adulthood in four strains of mutant mice lacking mGluR1 (Kano et al., 1997), G-protein αq subunit (Gαq) (Offermanns et al., 1997;Hashimoto et al., 2000), phospholipase C β4 (PLC β4) (Kano et al., 1998; Hashimoto et al., 2000), and PKCγ (Kano et al., 1995) caused by the defects in CF synapse elimination during the third postnatal week. In contrast to GluRδ2 mutant mice, these mutant mice have normal PF–PC synaptogenesis and normal rise time of the CF-EPSCs. We also found that blockade of NMDA receptors during P15 and P16, but not before or after this period, resulted in a higher incidence of multiple CF innervation and caused a mild but persistent loss of motor coordination (Kakizawa et al., 2000). Neither basic synaptic functions nor cerebellar morphology was affected by this manipulation. During P15 and P16, NMDA receptors were not functional at either PF to PC or CF to PC synapses but were rich in mossy fiber to granule cell synapses (Kakizawa et al., 2000). We also demonstrated that the mGluR1 to PKCγ cascade must work within PCs to complete CF synapse elimination (Ichise et al., 2000). We introduced the mGluR1 transgene into mGluR1-null mutant mice using a PC-specific promoter, L7. The mGluR1-rescue mice showed normal regression of multiple CF innervation and apparently normal motor coordination (Ichise et al., 2000). These results suggest that the signal transduction cascade from mGluR1 to PKCγ is activated at PF–PC synapses (Kano et al., 1995, 1997) in a manner dependent on NMDA receptor-mediated activity at mossy fiber to granule cell synapses (Kakizawa et al., 2000).
In GluRδ2 mutant mice a substantial number of PCs had multiple CF-EPSCs with fast rise-times in addition to atypical slow CF-EPSCs (Fig. 3). It is possible that the impaired PF–PC synapse formation in GluRδ2 mutant leads to a reduction of PF-induced drive of mGluR1 and causes persistent multiple CF innervations that are similar to those of the mGluR1, Gαq, PLC β4, and PKCγ mutant PCs.
In conclusion, we propose that GluRδ2 is required for consolidating PF synapses onto PC distal dendrites and restricting CF synapses to the proximal dendritic domain, whereas the mGluR1 to PKCγ signaling pathway is involved in eliminating excess CF synapses at the proximal dendrites. By the actions of these two glutamate receptors, synaptic connections of PCs mature normally during postnatal cerebellar development.
This work has been partly supported by grants from the Japanese Ministry of Education, Science, Sports, Culture and Technology (M.K.), the Kato Memorial Bioscience Foundation (K.H.), and the Human Frontier Science Program (M.K., A.K.). This work has also been supported by Special Coordination Funds for promoting Science and Technology from the Japanese Government (K.H.) and Research for the Future Program of Japan Society for the Promotion of Science (A.A). We thank Dr. N. Kawai for continuous encouragement throughout the course of this study, K. Matsumoto and Y. Hirano for excellent technical assistance, and T. Hensch for critically reading this manuscript.
Correspondence should be addressed to Masanobu Kano, Department of Physiology, Kanazawa University School of Medicine, Takara-machi, Kanazawa 920-8640, Japan. E-mail:.