TARPs γ-2 and γ-7 are essential for AMPA receptor expression in the cerebellum

Authors


  • The first two authors contributed equally to this work.

Dr Masahiko Watanabe, 2Department of Anatomy, as above.
E-mail: watamasa@med.hokudai.ac.jp

Abstract

The α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors require auxiliary subunits termed transmembrane AMPA receptor regulatory proteins (TARPs), which promote receptor trafficking to the cell surface and synapses and modulate channel pharmacology and gating. Of six TARPs, γ-2 and γ-7 are the two major TARPs expressed in the cerebellum. In the present study, we pursued their roles in synaptic expression of cerebellar AMPA receptors. In the cerebellar cortex, γ-2 and γ-7 were preferentially localized at various asymmetrical synapses. Using quantitative Western blot and immunofluorescence, we found severe reductions in GluA2 and GluA3 and mild reduction in GluA4 in γ-2-knockout (KO) cerebellum, whereas GluA1 and GluA4 were moderately reduced in γ-7-KO cerebellum. GluA2, GluA3 and GluA4 were further reduced in γ-2/γ-7 double-KO (DKO) cerebellum. The large losses of GluA2 and GluA3 in γ-2-KO mice and further reductions in DKO mice were confirmed at all asymmetrical synapses examined with postembedding immunogold. Most notably, the GluA2 level in the postsynaptic density fraction, GluA2 labeling density at parallel fiber–Purkinje cell synapses, and AMPA receptor-mediated currents at climbing fiber–Purkinje cell synapses were all reduced to approximately 10% of the wild-type levels in DKO mice. On the other hand, the reduction in GluA4 in γ-7-KO granular layer reflected its loss at mossy fiber–granule cell synapses, whereas that of GluA1 and GluA4 in γ-7-KO molecular layer was caused, at least partly, by their loss in Bergmann glia. Therefore, γ-2 and γ-7 cooperatively promote synaptic expression of cerebellar AMPA receptors, and the latter also promotes glial expression.

Introduction

Glutamate (Glu) mediates most of the fast excitatory synaptic transmission in the central nervous system, primarily through α-amino-3-hydroxyl-5-isoxazolepropionate (AMPA)-type glutamate receptors (GluRs; Hollmann & Heinemann, 1994). AMPA receptors comprise GluA1–GluA4 (GluRA–D or GluR1–4) subunits (Keinänen et al., 1990; Hollmann et al., 1991), and exist mainly as GluA1/GluA2 and GluA2/GluA3 heteromeric channels in brains (Wenthold et al., 1996). Inclusion of GluA2 edited at the ‘Q/R site’ from glutamine to arginine determines the Ca2+ permeability of AMPA receptors (Hollmann et al., 1991; Hume et al., 1991; Verdoorn et al., 1991; Mosbacher et al., 1994). Moreover, AMPA receptor trafficking and synaptic expression of AMPA receptors are controlled according to the ‘subunit-specific rule’. A long cytoplasmic tail of GluA1 or GluA4 binds to anchoring molecules SAP97 and protein 4.1, whereas a short tail of GluA2 or GluA3 interacts with GRIP1/2 and PICK1 (Jiang et al., 2006–2007). Phosphorylation and dephosphorylation of the C-termini alter the state of interaction with the anchoring molecules, which then regulates endocytosis and insertion of AMPA receptors at synapse in activity-dependent and subunit-dependent manners (Hirai, 2001; Shi et al., 2001; Malinow & Malenka, 2002; Song & Huganir, 2002; Lee et al., 2004).

Neuronal AMPA receptors also contain auxiliary subunits termed transmembrane AMPA receptor regulatory proteins (TARPs). The TARP family comprises six isoforms: four classical (γ-2, γ-3, γ-4 and γ-8) and two atypical (γ-5 and γ-7) TARPs (Kato et al., 2008; Soto et al., 2009). In the brain, their overall expressions are distinct but largely complementary both spatially and temporally: γ-2 in the cerebellum, γ-3 in the cerebral cortex, γ-4 in developing brain, γ-7 in the cerebellum and γ-8 in the hippocampus (Tomita et al., 2003; Fukaya et al., 2005; Kato et al., 2007). Ideas about the role of TARPs originally arose from the discovery of the virtual lack of AMPA receptor-mediated excitatory postsynaptic currents at mossy fiber–cerebellar granule cell synapses in the spontaneous mutant mouse stargazer or stg (Hashimoto et al., 1999), which carries an early transposon insertion in intron 2 of the γ-2 or Cacng2 gene (Letts et al., 1998). It is now evident that TARPs promote AMPA receptor expression at synaptic and extrasynaptic membranes (Chen et al., 2000; Tomita et al., 2004; Fukaya et al., 2006) and also modulate AMPA receptor gating both in vitro (Yamazaki et al., 2004; Priel et al., 2005; Tomita et al., 2005; Turetsky et al., 2005; Körber et al., 2007; Kott et al., 2007; Soto et al., 2007) and in vivo (Chen et al., 1999; Hashimoto et al., 1999, Rouach et al., 2005).

In the present study, we aimed at elucidating the roles of TARPs in the expression and function of cerebellar AMPA receptors. To this end, we generated mice deficient for γ-2 and γ-7 on the C57BL/6 genetic background, because these are two major TARPs expressed in cerebellar granule cells and Purkinje cells (Fukaya et al., 2005). We show that γ-2 and γ-7 are localized at various asymmetrical synapses, and their gene ablation severely reduces cerebellar contents, synaptic localization and synaptic responses of AMPA receptors.

Materials and methods

Generation of TARP γ-2- and γ-7-knockout (KO) mice

All animal experiments were carried out in accordance with the guidelines laid down by the animal welfare committees and the ethics committees of Niigata University. Mice deficient in γ-2 or γ-7 were produced by homologous recombination using the C57BL/6N ES cell line RENKA (Mishina & Sakimura, 2007). We isolated γ-2 (Cacng2) and γ-7 (Cacng7) genes by screening the genomic DNA library derived from the C57BL/6 mouse, and each gene fragment was yielded by PCR and sequenced. A γ-2 targeting vector contained exons 3 and 4 of Cacng2 gene with the 6.8-kb upstream and 4.8-kb downstream homologous genomic DNA fragments and the diphtheria toxin gene for negative selection (Fig. 1A). A DNA fragment, which carried a 34-bp loxP sequence and pgk-1 promoter-driven neomycin phosphotransferase gene (pgk-neo) flanked by two Flp recognition target (frt) sites, was inserted into the site 158 bp upstream of exon 3. The other loxP site was introduced into the site 159 bp downstream of exon 3 in order to eliminate the exon 3 containing the two putative transmembrane domains after Cre-mediated recombination. The γ-7 targeting construct contained exons 1–3 of the Cacng7 gene with the 6.6 kb upstream, 4.5 kb downstream homologous genomic DNA (Fig. 1C). The loxP sequence and pgk-neo flanked by two frt sites was inserted into the site 340 bp upstream of exon 2. The other loxP site was introduced into the site 54 bp downstream of exon 3 in order to eliminate exons 2 and 3 after Cre-mediated recombination. Homologous recombinants were identified by Southern blot analysis under the following conditions: Kpn I-digested DNA hybridized with γ-2-5′ probe, 8.7 kb for wild-type (WT) and 8.2 kb for targeted allele; EcoR V-digested DNA hybridized with γ-2-3′ probe, 12.2 kb for WT and 10.2 kb for targeted allele; EcoR I-digested DNA hybridized with neo probe, 7.2 kb for γ-2-targeted allele; Spe I-digested DNA hybridized with γ-7-5′ probe, 20.3 kb for WT and 16 kb for targeted allele; EcoR I-digested DNA hybridized with γ-7-3′ probe, 8.2 kb for WT and 9.3 kb for targeted allele; Hinc II-digested DNA hybridized with neo probe, 11 kb for γ-7-targeted allele. ES cell clones with correct recombination were used to yield chimeric mice as described previously (Fukaya et al., 2006). Chimeric mice were mated with C57BL/6 mice, and offspring were further crossed with TLCN-Cre mice (Nakamura et al., 2001; Fuse et al., 2004) to yield heterozygous KO mice (Fig. 1B and D). Homozygous γ-2- and γ-7-KO mice were obtained by crossing heterozygous pairs. The first offspring was genotyped by Southern blotting under the following conditions: EcoR I-digested DNA hybridized with γ-2-inner probe, 7.2 kb for WT and 6.7 kb for KO allele; EcoR I-digested DNA hybridized with γ-7-3′ probe, 8.2 kb for WT and 7.2 kb for KO allele. Genotypes for all subsequent breeding were determined by PCR analysis of digested mice tail samples. PCR genotyping of mouse tail DNA was performed with the following primers: γ-2-forward, 5′- GGTGCTAGAGTCCTGATCCTA -3′; γ-2-reverse, 5′- AGTGGGTTGCATGGAGTCTC -3′, γ-7-forward, 5′-ACAGGAATCCTTATTCCCAG -3′; γ-7-reverse, 5′-CTGAGCTCATGACTTCATCC -3′.

Figure 1.

 Production of γ-2-KO, γ-7-KO and γ-2/γ-7-DKO mice by the Cre/loxP recombination system. (A and C) Schematic representation of cDNA, WT allele, targeting vector, targeted allele and KO allele after Cre-mediated recombination. Gray boxes represent the transmembrane segments TM1 to TM4, while open boxes represent exon sequences, neo cassette (Neo) and diphtheria toxin (DT) cassette. loxP sites indicated by black triangles. Gray bars indicate probe regions used for Southern blot analysis. EI, EcoR I; EV, EcoR V; K, Kpn I; S, Spe I; H, Hinc II. (B and D) Southern blot analysis for genomic DNAs from chimeric, WT, heterozygous (Het) and homozygous (KO) mice. (E) Western blot analysis for γ-2 and γ-7 in the PSD fraction of the adult cerebellum prepared from WT, γ-7-KO, γ-2-KO and DKO mice. Proteins loaded in each lane are 2 μg. (F) Typical footprint patterns obtained from each genotype at 8 weeks of age. Ink was applied to the hind paws of the mice. Arrow indicates the direction of walking. DKO mice showed the most severe ataxic gait.

To evaluate the ataxic gait, footprints of the mice were recorded. Ink was applied to the hind paws of the mice, which were allowed to walk on white paper along a narrow path.

Antibodies

In Western blot analysis, we used the following primary antibodies (host species): TARP γ-2 (rabbit; see below), TARP γ-7 (rabbit; see below), GluA1 (rabbit; Watanabe et al., 1998), GluA2 (mouse; MAB397, Millipore), GluA3 (mouse; MAB5416, Millipore), GluA4 (guinea pig; Nagy et al., 2004), synaptophysin (mouse; MAB5258, Chemicon), PSD-95 (rabbit; Fukaya & Watanabe, 2000) and actin (mouse; MAB1501R, Chemicon). For immunohistochemistry we used GluA4 (guinea pig; Nagy et al., 2004) and glutamate–aspartate transporter (GLAST) antibodies (rabbit and guinea pig; Shibata et al., 1997), and also produced γ-2, γ-7, GluA1, GluA2 and GluA3 antibodies as described below.

Affinity-purified antibodies to γ-2 and γ-7 were raised in the rabbit and guinea pig using synthetic peptide CIQKDSKDSLHANTANR (302-318 amino acid residues, Genbank accession number AF077739) and CPAIKYPDHLHISTSP (260–274, AF361349), respectively, which were conjugated to keyhole limpet hemocyanin. We also immunized rabbits, guinea pigs and goat to produce polyclonal antibodies to the C-termini of AMPA receptor GluA1–A3 subunits. Due to partial homology in the C-terminal sequences between GluA1 and GluA4 and between GluA2 and GluA3 (Fig. S1A), we selected the following sequences: amino acid residues 880–907 and 841–907 of GluA1 (GenBank, X57497) were used for antigen, affinity purification or for dot blot assay, respectively, and 853–883 of GluA3 (AB022342) were used for antigen, affinity purification and dot blot assay, while residues 847–863 and 847–877 of GluA2 (X57498) were for antigen and affinity purification or for dot blot assay, respectively (Fig. S1A). Procedures for bacterial protein expression, immunization and purification of antibodies have been described previously (Fukaya et al., 2006). The specificity of the AMPA receptor subunit antibodies as well as no crossreactivity with other subunits was tested by immunoblot with brain extracts (Fig. S1B) and dot blot assay for C-terminal fragments (Fig. S1C), respectively. As a result, subunit-specific antibodies were obtained for GluA1 and GluA2 in the rabbit and guinea pig, and for GluA3 in the rabbit, guinea pig and goat.

Subcellular fractionation and Western blot analysis

Preparation of fractionated protein samples and Western blotting was performed as previously described (Abe et al., 2004; Fukaya et al., 2006). Briefly, adult (8–16 weeks of age) animals were decapitated by cervical dislocation, and their cerebella were homogenized in homogenate buffer (0.32 m sucrose, 5 mm EDTA, 1 μm pepstatin, 2 μm leupeptin and 0.5 mm phenylmethylsulfonly fluoride) and centrifuged at 1,000 g for 10 min to obtain the supernatant post-nuclear fraction (homogenate, Fig. 4A). The supernatant was further centrifuged at 10,000 g for 10 min and the pellet was separated on a sucrose density gradient (0.32, 0.8 and 1.2 m sucrose), and the synaptosome fraction was obtained between 0.8 and 1.2 m sucrose. For the postsynaptic density (PSD) fraction, the synaptosome sample was further solubilized with 0.5% Triton X-100 and the pellet, after centrifugation at 200 000 g for 1 h, was suspended with 40 mm Tris–HCl pH 8.0 and 1% sodium dodecyl sulfate (SDS). Protein samples from homogenate (20 μg), synaptosome (3 μg) and PSD (2 μg) fractions were loaded onto each lane and subjected to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) for Western blotting (Fig. 4A). Signal intensities of immunoreacted bands were determined by densitometric measurement using ImageJ software (available from the US National Institutes of Health) and normalized with actin signal intensities. Statistical significance was assessed by two tailed, one-sample t-test using PRISM (GraphPad Software, San Diego, CA, USA). All results are expressed as mean ± SEM.

Figure 4.

 Changes in cerebellar contents of four GluA subunits in WT, γ-2-KO, γ-7-KO and γ-2/γ-7-DKO mice. (A) Western blotting of cerebellar homogenate, synaptosome fraction and PSD fraction. Synaptosome samples were extracted in 0.5% Triton X-100 to yield PSD components. The quality of the synaptosome and PSD fractions is shown by enrichment and equal intensities for synaptophysin or PSD-95, respectively. Proteins loaded in each lane are 20 μg for homogenate, 3 μg for synaptosome and 2 μg for PSD fraction. (B and C) Bar graphs showing the mean intensities of immunoblot bands in γ-2-KO, γ-7-KO and DKO cerebella relative to those in WT mice. Bars on individual columns represent SEM; n = 3-5 experiments. *< 0.05 compared to WT (one-sample t-test).

Immunohistochemistry

Under deep pentobarbital anesthesia (100 mg/kg of body weight, i.p.), mice were perfused transcardially with 4% paraformaldehyde in 0.1 m sodium phosphate buffer (PB; pH 7.2) for light microscopic immunohistochemistry or with 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 m PB for postembedding immunogold electron microscopy. Brains to be compared simultaneously were embedded in single paraffin blocks, and paraffin sections (4 μm in thickness) were made using a sliding microtome (SM1000R; Leica, Nussloch, Germany). Microslicer sections were also used for immunofluorescence (50 μm; VT1000S, Leica) and for postembedding immunogold (400 μm). All immunohistochemical incubations were done at room temperature.

For light microscopic immunohistochemistry, paraffin sections were first subjected to pepsin pretreatment for antigen exposure, i.e., incubation in 1 mg/ml of pepsin (DAKO, Carpinteria, CA, USA) in 0.2 N HCl for 10 min at 37°C. Then sections were incubated successively with 10% normal donkey serum for 20 min, primary antibodies (1 μg/ml) overnight, biotinylated secondary antibodies for 2 h and avidin–biotin–peroxidase complex for 1 h, using a Histofine SAB-PO(R) kit (Nichirei Corp., Tokyo, Japan). Immunoreaction was visualized using the tyramide signal amplification kit (Perkin-Elmer, Boston, MA, USA). To detect nonsynaptic AMPA receptors, double immunofluorescence without pepsin pretreatment was done for GLAST and GluA1 or GluA4 using microslicer sections. Images of whole brain sections were taken with a dissecting microscope, while those of cerebellar cortex were with a confocal laser scanning microscope (FV1000; Olympus).

For postembedding immunogold, cerebellar slices were cryoprotected with 30% sucrose in 0.1 m PB, and frozen rapidly with liquid propane in a Leica EM CPC unit. Frozen sections were immersed in 0.5% uranyl acetate in methanol at −90°C in a Leica AFS freeze-substitution unit, infiltrated at −45°C with Lowicryl HM-20 resin (Lowi, Waldkraiburg, Germany) and polymerized with UV light. After etching with saturated sodium ethanolate solution for 3 s, ultra-thin sections on nickel grids were treated successively with 1% human serum albumin (Wako, Osaka, Japan) with 0.1% Tween 20 in Tris-buffered saline (HTBST; pH 7.5) for 1 h, primary antibodies to GluA subunits (15 μg/ml for each) in HTBST overnight, and colloidal gold (10 nm)-conjugated antirabbit IgG (1:100; British Bio Cell International, Cardiff, UK) in HTBST for 2 h. Finally, grids were stained with uranyl acetate for 15 min. Electron micrographs were taken with an H7100 electron microscope (Hitachi, Tokyo, Japan). For quantitative analysis, the number of metal particles and the length of synaptic membrane were measured on electron micrographs, using IPLab software (Scanalytics, Fairfax, VA, USA).

Fluorescent in situ hybridization (FISH)

Procedures for FISH have been reported previously (Yamasaki et al., 2010). Briefly, fresh frozen sections were hybridized with mixtures of digoxigenin (DIG)- or fluorescein-labeled cRNA probes for mouse γ-7 (nucleotide residues 181–828, AF361349.1) and 67-kDa glutamic acid decarboxylase (GAD67; 1036–2015, NCBI Reference Sequence NM_008077) or GLAST (1571–2473, AF330257.1). Supporting Fig. S2A–C shows overall patterns of FISH labeling, which were consistent with those of in situ hybridization using radiolabeled probes (Shibata et al., 1996; Fukaya et al., 2005; Uchigashima et al., 2007). DIG and fluorescein were detected using the two-step method: the first detection with peroxidase-conjugated antifluorescein antibody (Roche Diagnostics, 1:500) for 1 h and the FITC-TSA plus amplification kit (PerkinElmer), and the second detection with peroxidase-conjugated anti-DIG antibody (Roche Diagnostics, 1:500) for 1 h and the Cy3-TSA plus amplification kit (PerkinElmer). Residual activities of peroxidase introduced in the first detection were inactivated by incubation of sections with 0.6% H2O2 for 30 min. TOTO3 (Invitrogen) was used for fluorescent nuclear counterstaining.

Electrophysiological analysis

Animals were anesthetized with carbon dioxide and parasagittal cerebellar slices (250 μm thickness) were prepared from mice aged postnatal day (P)24 to P95 as described previously (Edwards et al., 1989; Hashimoto & Kano, 2003). Whole-cell recordings were made from visually identified Purkinje cell somata using an upright microscope (BX50WI; Olympus, Tokyo, Japan). Ionic currents were recorded with an Axopatch 1D (Molecular Devises, Sunnyvale, CA, USA) patch-clamp amplifier. Resistances of patch pipettes were 2–3 MΩ when filled with an intracellular solution composed of (in mm): CsCl, 60; Cs D-gluconate, 10; TEA-Cl, 20; BAPTA, 20; MgCl2, 4; ATP, 4; and HEPES, 30 (pH 7.3, adjusted with CsOH). The pipette access resistance was compensated by 70–80%. The holding potential was corrected for liquid-junction potential. The composition of the standard bathing solution was (in mM): NaCl, 125; KCl, 2.5; CaCl2, 2; MgSO4, 1; NaH2PO4, 1.25; NaHCO3 26; and glucose, 20; bubbled with 95% O2 and 5% CO2. Bicuculline (10 μm) or picrotoxin (100 μm) was always added to block inhibitory synaptic transmission. The signals of membrane currents were filtered at 3 kHz and digitized at 20 kHz for recording evoked climbing fiber-mediated excitatory postsynaptic currents (CF-EPSCs) or at 10 kHz for recording postsynaptic AMPA receptor-mediated currents. On-line data acquisition and off-line data analysis were performed using PULSE software (HEKA, Lambrecht/Pfalz, Germany). Climbing fibers were stimulated via the stimulation pipette placed in the granule cell layer. Stimuli (duration, 0.1 ms; amplitude, 0–90 V) were applied at 0.2 Hz. In the experiment for the I–V relationships of the postsynaptic AMPA receptor-mediated currents, spermine (100 μm) was added to the intracellular solution and cyclothiazide (100 μm) and tetrodotoxin (0.5 μm) were added to the external solution. All experiments were carried out at 31°C.

Results

Generation of mutant mice lacking γ-2 and γ-7

To investigate the roles of TARP γ-2 and γ-7 in synaptic expression and function of cerebellar AMPA receptors, we generated mice deficient in γ-2 or γ-7 on the C57BL/6N background (Fig. 1A–E). A previous study reported that, when backcrossed to the C57BL/6J background, mice carrying the stg mutation died before weaning (Letts et al., 2003). However, our γ-2-KO mice were viable after weaning and exhibited essentially the same phenotype as the original stg mouse, including ataxic gait and head-lifting behavior. In addition, γ-2-KO mice were small in size with 73% of the body weight of their WT littermates at 8–10 weeks of age, similarly to original stg mice. On the other hand, γ-7-KO mice were viable, fertile and indistinguishable from their WT littermates. Then we crossed the two mouse lines to obtain γ-2/γ-7 double-KO (DKO) mice, which had approximately 70% of the body weight of their WT littermates. DKO mice showed much more severe ataxia than γ-2-KO mice did, as they could not walk straight and displayed frequent tumbling and rolling as appreciated from footprint patterns (Fig. 1F).

Synaptic localization of γ-2 and γ-7 in the cerebellar cortex

The distribution of γ-2 and γ-7 at the protein level was examined in the cerebellar cortex by producing specific antibodies. The specificity was verified by the lack of immunoreacted bands in the corresponding KO cerebella (Fig. 1E). We further noted that cerebellar content of γ-7 was reduced in γ-2-KO cerebellum, while that of γ-2 was not altered in γ-7-KO cerebellum (Fig. 1E). By immunohistochemistry, γ-2 and γ-7 were distributed at the highest levels in the cerebellum (Fig. 2A and E), the specificity of which was verified by blank immunostaining in the corresponding KO brains (Fig. 2B and F). Within the cerebellum, γ-2 was detected as clustered staining in the granular layer (i.e., synaptic glomeruli) and as punctate staining in the molecular layer (Fig. 2C and D). γ-7 showed similar distributions in the cerebellar cortex (Fig. 2G and H). However, γ-7 was more intense in the molecular layer than in the granular layer, and puncta sometimes showed vertical lining (arrows in Fig. 2H), suggesting its distribution along Bergmann glial fibers. The glial expression was ascertained with double-labeling FISH, in which γ-7 mRNA was detected not only in GAD67 mRNA-expressing Purkinje cells and molecular layer interneurons but also in GLAST mRNA-expressing Bergmann glia (supporting Fig. S2D and E).

Figure 2.

 Light microscopic immunohistochemitry for (A–D) TARP γ-2 and (E–H) γ-7 in the adult mouse brain. (A, B, E and F) Immunoflurescence in parasagittal brain sections of (A and E) WT, (B) γ-2-KO and (F) γ-7-KO brains. (C, D, G and H) Closer images of (C and G) the cerebellum and (D and H) cerebellar cortex. GL, granular layer; ML, molecular layer; PC, Purkinje cell layer. Arrows in H indicate γ-7 labeling along putative Bergmann glial fibers. Scale bars, 1 mm (A–C and E–G), 20 μm (D and H).

Postembedding immunogold microscopy demonstrated that γ-2 and γ-7 were selectively detected on the postsynaptic membrane of various asymmetrical synapses, including the parallel fiber–Purkinje cell synapse (Fig. 3A and G; supporting Fig. S1A and B), climbing fiber–Purkinje cell synapse (Fig. 3E and K), parallel fiber–molecular layer interneuron synapse (Fig. 3D and J) and mossy fiber–granule cell synapses (Fig. 3B and H). However, we rarely found immunogold labeling at symmetrical synapses between terminals of molecular layer interneurons (basket and stellate cells) and Purkinje cell dendrites. The specificity of immunogold labeling for γ-2 and γ-7 was confirmed by almost blank labeling at the parallel fiber–Purkinje cell synapse of γ-2-KO and γ-7-KO mice, respectively (Fig. 3C and I). By counting the number of immunogold particles at given types of cerebellar synapses, γ-2 was distributed with two- or three-fold higher densities at the parallel fiber–molecular layer interneuron synapse compared to other asymmetrical synapses (Fig. 3F). On the other hand, γ-7 was distributed at similar densities at various asymmetrical synapses (Fig. 3L). Although no significant immunogold labeling was noted for extrasynaptic cell membrane, intracellular organelles and glial elements, this may be due not only to their low expression, if any, at nonsynaptic sites, but also to the low detection sensitivity of postembedding immunogold. Therefore, it is safe to conclude that γ-2 and γ-7 highly accumulate on the postsynaptic membrane of various asymmetrical synapses in the cerebellar cortex.

Figure 3.

 Postembedding immunogold electron microscopy for (A–F) TARP γ-2 and (G–L) γ-7 at various cerebellar synapses. (A and G) Asymmetrical synapses between parallel fiber terminals (PF) and Purkinje cell (PC) spines (Sp). (B and H) Asymmetrical synapses between mossy fiber terminals (MF) and granule cell dendrites (Gr). (C and I) Control immunogold labeling at PF-PC synapses of (C) γ-2-KO and (I) γ-7-KO mice. (D and J) Asymmetrical synapses between PF and molecular interneuron dendrites (In). (E and K) Asymmetrical synapses between climbing fiber terminals (CF) and Purkinje cell spines. (F and L) The density of immunogold labeling per μm of the postsynaptic density (PSD) at PF-PC, CF-PC, PF-In, MF-Gr and In-PC (symmetrical) synapses of WT mice, and PF-PC synapses of γ-2-KO and γ-7-KO mice. Synapses from three WT and three γ-2-KO mice or from three WT and three γ-7-KO mice were analyzed and pooled. The numbers of analyzed synapses (total length of PSD) were 144 (47.1 μm) for PF-PC synapses, 40 (10.3 μm) for MF-Gr synapses, 32 (6.8 μm) for PF-In synapses, 32 (8.3 μm) for In-PC synapses and 18 (5.9 μm) for CF-PC synapses in WT mice for γ-2 immunogold, 82 (27.1 μm) for PF-PC synapses in γ-2-KO mice for γ-2 immunogold, and 125 (41.1 μm) for PF-PC synapses, 60 (15.5 μm) for MF-Gr synapses, 50 (10.5 μm) for PF-In synapse, 34 (8.8 μm) for In-PC synapses and 18 (5.6 μm) for CF-PC synapses in WT mice for γ-7 immunogold, and 76 (25.1 μm) for PF-PC synapses in γ-7-KO mice for γ-7 immunogold. Scale bars, 100 nm.

Biochemical reduction in GluA subunits in γ-2- and γ-7-KO cerebella

We next analyzed changes in cerebellar contents of the four AMPA receptor subunits GluA1–GluA4 by preparing the homogenate, synaptosome fraction and PSD fraction from WT, γ-2-KO, γ-7-KO and DKO mice (Fig. 4A, top). The quality of fractionated protein samples was tested by immunoblot for synaptophysin and PSD-95, while the amount of loaded samples was normalized with actin signal intensities (Fig. 4A, bottom). Quantitative Western blot analysis with cerebellar homogenates showed that, in γ-7-KO cerebellum, protein levels were reduced significantly by approximately 40% for GluA1 (56.3 ± 3.3% of the WT level; = 0.0002, one-sample t test, two-tailed) and GluA4 (55.0 ± 8.7, = 0.01), while no significant changes were found for GluA2 (87.7 ± 17.1%, = 0.51) or GluA3 (74.3 ± 7.0%, = 0.07; Fig. 3A and B). In contrast, in γ-2-KO cerebellum, protein levels were reduced significantly by 60% for GluA2 (35.8 ± 5.1%, = 0.0002) and GluA3 (35.9 ± 7.0%, = 0.01) and by 40% for GluA4 (57.6 ± 6.1%, = 0.002), while no reduction was found for GluA1 (75.6 ± 16.7%, = 0.24; Fig. 4A and B). These reductions became more remarkable in the synaptosome fraction (Fig. 4A, middle panel) and PSD fraction (Fig. 4A, right panel; Fig. 4C).

All four subunits displayed further reductions in DKO cerebellum (Fig. 4C). In the PSD fraction, protein levels relative to those in WT mice were 38.3 ± 7.2% for GluA1, 9.5 ± 4.6% for GluA2, 15.2 ± 3.3% for GluA3 and 37.8 ± 5.4% for GluA4, showing significant differences (= 0.0011, <0.0001, 0.0001 and 0.0014, respectively).

Immunohistochemical reduction in GluA subunits in γ-2- and γ-7-KO cerebella

Next, immunohistochemical changes in GluA1–GluA4 were examined using subunit-specific antibodies (supporting Fig. S1) and pepsin pretreatment, an antigen-exposing method particularly effective in detection of postsynaptic molecules (Fukaya & Watanabe, 2000). In WT mice, the molecular layer was stained intensely for all four subunits, while the granular layer was stained weakly for GluA2 and GluA4 (Figs 5 and 6). These patterns of immunohistochemical distribution appeared to reflect cell type-specific subunit expression shown by previous in situ hybridization and single-cell PCR: GluA1–GluA3 mRNAs in Purkinje cells, GluA1 and GluA4 mRNAs in Bergmann glia, and GluA2 and GluA4 mRNAs in granule cells (Keinänen et al., 1990; Pellegrini-Giampietro et al., 1991; Lambolez et al., 1992).

Figure 5.

 Light microscopic immunohistochemistry for four GluA subunits in parasagittal brain sections of WT, γ-2-KO and γ-2/γ-7-DKO mice. Overall immunostaining patterns in the brain or the cerebellar molecular layer (ML) are shown for (A and E) GluA1, (B and F) GluA2, (C and G) GluA3 and (D and H) GluA4. Note severe reduction in GluA2 and GluA3 in the molecular layer of γ-2-KO mice and further reduction in DKO mice, when compared to WT mice. Also note mild reductions in GluA1 and GluA4 in the molecular layer of γ-2-KO mice. Scale bars, 1 mm (A), 0.1 mm (E).

Figure 6.

 Light microscopic immunohistochemistry for four GluA subunits in parasagittal brain sections of WT and γ-7-KO mice. Note (A–C)moderate reduction in GluA1 in the molecular layer (ML), (D–F) no apparent reduction in GluA2, (G–I) slight reduction in GluA3 in the molecular layer, and (J–L) marked reduction in GluA4 in the granular and molecular layers. Scale bars, 1 mm (A), 0.1 mm (B).

Brains from WT, γ-2-KO and DKO mice were embedded in single paraffin blocks, mounted on single glass slides and processed simultaneously for immunoreaction (Fig. 5). Compared to the intensity in WT mice, striking reductions were noted in the cerebellar cortex for GluA2 and GluA3, with intensities in the order WT > γ-2-KO > DKO (Fig. 5B, C, F and G). In particular, GluA2 became almost blank in the granular layer of γ-2-KO mice and in the molecular layer of DKO mice (Fig. 5B and F; supporting Fig. S4A). On the other hand, GluA4 was reduced mildly in the molecular layer and severely in the granular layer of γ-2-KO and DKO mice (Fig. 5D and H; supporting Fig. S4B). GluA1 was reduced mildly in the molecular layer of γ-2-KO and DKO mice (Fig. 5A and E).

Likewise, WT and γ-7-KO brains embedded in single paraffin blocks were examined (Fig. 6). In contrast to the staining in γ-2-KO cerebellum, moderate reduction was noted for GluA1 and GluA4 in the molecular layer of γ-7-KO mice (Fig. 6A–C and J–L). GluA4 was also reduced in the granular layer of γ-7-KO mice (Fig. 6J–L; supporting Fig. S4D). On the other hand, the reduction in immunohistochemical intensity was relatively mild for GluA3 (Fig. 6G–I), while no difference was noticed for GluA2 in the molecular and granular layers (Fig. 6D–F; supporting Fig. S4C). Therefore, consistent with the changes observed with Western blot, the loss of γ-2 and γ-7 did affect immunohistochemical intensities of cerebellar AMPA receptors in subunit-dependent manners: preferential reduction in GluA2 and GluA3 in γ-2-KO cerebellum, preferential reduction in GluA1 and GluA4 in γ-7-KO cerebellum, and severe reduction in GluA2 and GluA3 and moderate reduction in GluA1 and GluA4 in DKO cerebellum.

Reduced synaptic localization of GluA subunits in γ-2 and γ-7-KO cerebella

The extent of reduction for synaptic AMPA receptors was assessed by postembedding immunogold electron microscopy. By this method, most immunogold particles fell on the postsynaptic membrane of asymmetrical synapses, whereas labeling of extrasynaptic membrane, intracellular organelles or glial elements was very rare and nearly at the background level (supporting Fig. S3C–E), as is the case for γ-2 and γ-7. From our preliminary experiments, we focused on major subunits expressed at given types of synapses, i.e. GluA1–GluA3 at the parallel fiber–Purkinje cell and climbing fiber–Purkinje cell synapses, GluA1–GluA4 at the parallel fiber–interneuron synapse and GluA2 and GluA4 at the mossy fiber–granule cell synapse (Fig. 7).

Figure 7.

 Quantitative analyses on four GluA subunits by post-embedding immunogold at various cerebellar synapses in WT, γ-2-KO, γ-7-KO and γ-2/γ-7-DKO mice. (A–L) Electron micrographs for (A–D) GluA1, (E–H) GluA2 and (I–L) GluA3 at parallel fiber–Purkinje cell (PF-PC) synapses of (A, E and I) WT, (B, F and J) γ-7-KO, (C, G and K) γ-2-KO and (D, H and L) γ-2/γ-7-DKO mice. Arrows indicate gold particles in the postsynaptic density (PSD). (M–X) Changes in immunogold labeling for (M, P and S) GluA1, (N, Q, T and W) GluA2, (O, R and U) GluA3 and (V and X) GluA4 at (M–O) the parallel fiber–Purkinje cell synapse, (P–R) climbing fiber–Purkinje cell synapse, (S–V) parallel fiber–interneuron synapse and (W and X) mossy fiber–granule cell synapse. Scores on each bar show the percentage relative to labeling at the corresponding synapses in WT mice. The numbers of measured synapses were (M–O) 80-142 for PF-PC (from three mice for each genotype), (P–R) 10-30 for climbing fiber–Purkinje cell (CF-PC), (S–V) 10-49 for parallel fiber–interneuron (PF-In), and (W and X) 20-36 for mossy fiber–granule cell (MF-Gr) synapses. Bars on columns represent SEM. Scale bar, 100 nm.

At the parallel fiber–Purkinje cell synapse (Fig. 7A–L), synaptic labeling in γ-2-KO mice showed severe reductions for GluA2 and GluA3 (30.5 and 28.7%, respectively, of WT levels) and mild reductions for GluA1 (62.1%) in γ-2-KO mice (Fig. 7M–O). On the other hand, mild reduction was only noted for GluA3 in γ-7-KO mice (60.5%). All three subunits were further reduced in DKO mice: GluA1 (46.5%), GluA2 (11.6%) and GluA3 (12.6%). This tendency was largely similar at the climbing fiber–Purkinje cell synapse (Fig. 7P–R). A notable difference at this synapse was severe loss of GluA1 at the climbing fiber–Purkinje cell synapse in DKO mice (12.7%), as was the case for GluA2 (0.0%) and GluA3 (31.3%). At the parallel fiber–interneuron synapse (Fig. 7S–V), GluA2–GluA4 were substantially reduced in γ-2-KO mice (45.4, 23.1 and 41.3%, respectively), whereas in γ-7-KO mice GluA3 was the only subunit displaying a significant reduction (32.3%). In DKO mice, all four subunits showed moderate to severe reductions (60.0% for GluA1, 31.6% for GluA2, 9.2% for GluA3 and 22.1% for GluA4). At the mossy fiber–granule cell synapse (Fig. 7W and X), GluA2 and GluA4 were severely reduced in γ-2-KO mice (4.9 and 28.9%, respectively), whereas GluA4 (52.6%), but not GluA2, showed moderate reduction in γ-7-KO mice and was further lowered to 11.3% in DKO mice. These results indicate that γ-2 and γ-7 synergistically promote expression of AMPA receptors, particularly GluA2–GluA4, at almost all cerebellar synapses, although the extent of reductions in γ-2-KO, γ-7-KO and DKO mice varied depending on the type of synapse. Considering that major synapses in the molecular layer, i.e., parallel fiber synapses on Purkinje cells and interneurons, had almost normal levels of GluA1 and GluA4 in γ-7-KO mice, reduced immunohistochemical intensities for GluA1 and GluA4 in γ-7-KO molecular layer (Fig. 6) should reflect their loss from the other cellular elements.

Reduction in glial AMPA receptors in γ-7-KO cerebellum

In the molecular layer, GluA1 and GluA4 are known to be expressed in Bergmann glia (Keinänen et al., 1990; Pellegrini-Giampietro et al., 1991). We then tested whether the reductions in GluA1 and GluA4 in the molecular layer were due to their reduced expression in Bergmann glia. To this end, we employed double immunofluorescence for the glutamate transporter GLAST, an astrocyte-specific molecule particularly enriched in Bergmann glia (Shibata et al., 1997; Yamada et al., 2000), and we omitted pepsin pretreatment to preferentially detect nonsynaptic AMPA receptors (Fukaya et al., 2006). Immunofluorescent signals for GluA1 or GluA4 overlapped well with GLAST in the molecular layer of WT and γ-7-KO mice, and the intensities were substantially reduced in the latter mice as compared to the former (Fig. 8). These results suggest that the ablation of γ-7 reduces expression of AMPA receptors in Bergmann glia.

Figure 8.

 Immunofluorescence without pepsin treatment for GluA1 and GluA4 in the cerebellar molecular layer of WT and γ-7-KO mice. Note marked reduction in GluA1 and GluA4 in the molecular layer (ML) of (B and D) γ-7-KO mice compared to (A and C) WT mice. (E–H) Double immunofluorescence for (E and F) GluA1 or (G and H) GluA4 with astrocyte marker GLAST in the molecular layer of (E and G) WT and (F and H) γ-7-KO mice. Asterisks and arrows indicate Purkinje cell and Bergmann glia soma, respectively. GL, granular layer. Scale bars, 100 μm (A), 50 μm (E).

Reduced AMPA receptor-mediated currents at γ-2- and γ-2/γ-7-DKO Purkinje cell synapses

Finally, we examined functional reductions in AMPA receptors by electrophysiology. Whole-cell patch-clamp recording was conducted from Purkinje cells in acute slices prepared from WT and respective KO mice. First, we examined the climbing fiber-mediated excitatory postsynaptic current (EPSC) that is solely mediated by AMPA receptors (Konnerth et al., 1990; Kano et al., 1995). In γ-7-KO mice, climbing fiber EPSCs were normal (Fig. 9A and B). On the other hand, the peak amplitude of climbing fiber EPSCs decreased progressively, in the order WT = γ-7-KO > γ-2-KO > DKO (Fig. 9A and B).

Figure 9.

 AMPA receptor-mediated synaptic and membrane currents of Purkinje cells in WT, γ-2-KO, γ-7-KO and γ-2/γ-7-DKO cerebella. (A) Representative traces of climbing fiber-mediated EPSCs (CF-EPSCs). Two traces are superimposed at each threshold stimulus intensity. Climbing fibers were stimulated in the granule cell layer at 0.2 Hz. Holding potential was -20 mV. (B) Summary bar graph showing peak amplitudes of CF-EPSCs recorded at the holding potential of -20 mV. The numbers of analyzed CF-EPSCs are indicated in parentheses on each column. (C) I-V relationships of the AMPA (5 μm)-induced currents recorded from Purkinje cells in WT (n = 6), γ-2-KO (n = 7) and DKO (n = 4) mice. Currents were measured during voltage ramp from +40 mV to −60 mV (duration, 1.7 s). For leak subtraction, evoked currents measured in the control solution were subtracted from those in the presence of AMPA. Records were taken in the presence of cyclothiazide (100 μm) and tetrodotoxin (0.5 μm). Intracellular solution was supplemented with spermine (100 μm). Error bars represent SEM at −60, −50, −40, −30, −20, −10, 0, 10, 20, 30 and 40 mV.

Next, we measured membrane currents in Purkinje cells induced by bath-applied AMPA (Fig. 9C). The current recorded in the standard external solution during voltage ramp (holding potential of +40 to −60 mV, 1.7 s) was subtracted from the current recorded in the presence of 5 μm AMPA. The AMPA receptor-mediated currents also decreased in the order WT > γ-2-KO > DKO (Fig. 9C). Because parallel fiber synapses (105–106 per Purkinje cell) far outnumber climbing fiber synapses (presumably by a factor of several hundred; Napper & Harvey, 1988; Kurihara et al., 1997), the reduced AMPA-induced currents in Purkinje cells are considered to virtually reflect functional loss of AMPA receptors at parallel fiber–Purkinje cell synapses. These electrophysiological data are consistent with the anatomical data and suggest that γ-2 and γ-7 cooperatively promote synaptic expression of AMPA receptors at climbing fiber and parallel fiber synapses in Purkinje cells, while the ablation of γ-7 by itself causes no apparent changes.

Discussion

Of the six TARP members (Chen et al., 2000; Tomita et al., 2003; Kato et al., 2008) we focused on γ-2 and γ-7, the highest expression levels of which are in two major cerebellar neuron types, i.e., granule cells and Purkinje cells (Fukaya et al., 2005). In the present study, we produced specific antibodies against γ-2 and γ-7 to determine their synaptic localization in the cerebellum, and also produced mutant mice lacking these TARPs to pursue their role in synaptic expression of cerebellar AMPA receptors.

Selective expression of γ-2 and γ-7 at postsynaptic regions of asymmetrical synapses

We investigated the distribution of γ-2 and γ-7 by an immunohistochemical staining method whose specificity was confirmed by blank immunolabeling in the brain of the respective KO mice. By light microscopic immunohistochemistry, the granular and molecular layers of the cerebellum were labeled most intensely for both γ-2 and γ-7 in the brain. Clustered labeling in the granular layer probably reflects their synaptic distribution in granule cells, while punctate labeling in the molecular layer probably represents synaptic distribution in Purkinje cells and molecular layer interneurons, and putative glial expression. Of these elements, postembedding immunogold microscopy revealed robust labeling of γ-2 and γ-7 at the mossy fiber–granule cell synapse, parallel fiber–Purkinje cell synapse, climbing fiber–Purkinje cell synapse and parallel fiber–interneuron synapse. All these synapses are classified as asymmetrical (or type I) synapses, a neuroanatomical feature of excitatory synapses (Llinas et al., 2004). However, they were absent at the interneuron–Purkinje cell synapse, a GABAergic symmetrical (or type II) synapse. Moreover, immunogold labeling of γ-2 or γ-7 was preferentially localized to the postsynaptic membrane at all these asymmetrical synapses. This distribution pattern is identical to that of γ-8, which is highly concentrated at various asymmetrical synapses in the hippocampus (Fukaya et al., 2006; Inamura et al., 2006). As γ-2 and γ-7 mRNAs are expressed in deep cerebellar nucleus neurons and Golgi cells as well (Fukaya et al., 2005; Kato et al., 2007), they may be also expressed at asymmetrical synapses of these neurons. Taken together, γ-2 and γ-7 are the major TARPs at various excitatory synapses in the cerebellum.

Synergistic function of γ-2 and γ-7 for cerebellar AMPA receptors

Using quantitative Western blot analysis and immunohistochemical techniques, we found that protein contents and immunohistochemical signal intensities of AMPA receptor subunits were decreased in γ-2-KO and γ-7-KO cerebella, and further reduced in DKO cerebellum. Importantly, the extent of reduction was apparently larger in the PSD fraction than in the homogenate. For example, in DKO cerebellum, GluA2 levels were reduced to 30% of the WT level in the homogenate, whereas it was reduced to approximately 10% in the PSD fraction. This suggests that the ablation of γ-2 and γ-7 severely affected expression of synaptic AMPA receptors. Indeed, in DKO mice the density of GluA2 immunogold labeling was reduced to 11.6% of the WT level at the parallel fiber–Purkinje cell synapse, the most prevalent synapse in the molecular layer. Furthermore, AMPA receptor-mediated EPSCs also reduced to 9.5% at the climbing fiber–Purkinje synapse. Previous experiments using heterologous cells (Chen et al., 2000; Tomita et al., 2004; Vandenberghe et al., 2005; Kato et al., 2007) and brain extracts (Fukata et al., 2005; Nakagawa et al., 2005; Inamura et al., 2006) demonstrate that γ-2 and γ-7 tightly interact with AMPA receptors and regulate their proper folding, trafficking and stability. Presumably through these molecular functions, γ-2 and γ-7 synergistically promote cerebellar contents, synaptic localization and postsynaptic currents of AMPA receptors to ten-fold that of cerebellum lacking both γ-2 and γ-7.

Such a synergistic function by the TARP family is also indicated from a study using stg/γ-3-DKO mice (Menuz et al., 2008). γ-3 is highly expressed in cerebellar Golgi cells (Fukaya et al., 2005), and its sole gene ablation did not affect cerebellar contents of AMPA receptors or AMPA receptor-mediated responses in Golgi cells. In stg/γ-3-DKO mice, however, all four AMPA receptor subunits, particularly GluA2 and GluA3, were severely reduced in the cerebellum, and AMPA receptor-mediated responses were reduced to nearly 10% in Golgi cells (Menuz et al., 2008). Multiple TARP members, being expressed with differential combination and stoichiometry in given neuronal populations, may regulate AMPA receptor expression in a cooperative manner.

Promotion by γ-2 and γ-7 of expression of synaptic AMPA receptors

In quantitative Western blot analysis, we found severe reductions in GluA2 and GluA3 and mild reductions in GluA4 in γ-2-KO cerebellum. GluA2–GluA4 were further reduced in γ-2/γ-7-DKO cerebellum. Light-microscopic immunohistochemistry gave a closely similar result, which was also consistent with their severe reductions at almost all cerebellar synapses examined by postembedding immunogold. In γ-7-KO mice, reductions in AMPA receptor subunits were more modest, i.e., mild reductions in GluA3 at the parallel fiber–Purkinje cell and parallel fiber–interneuron synapses and moderate reduction in GluA4 at the mossy fiber–granule cell synapse. As to GluA1, we found mild reductions at the parallel fiber–Purkinje cell and climbing fiber–Purkinje cell synapses in γ-2-KO mice, and found no reduction at any synapses examined in γ-7-KO mice. Nevertheless, following the ablation of both TARPs, GluA1 was reduced severely at climbing fiber–Purkinje cell synapse and moderately so at the parallel fiber–Purkinje cell and parallel fiber–interneuron synapses. These results suggest that γ-2 or γ-7 per se preferentially promotes synaptic expression of GluA2–GluA4, and that they come to promote GluA1 expression too, when expressed together.

AMPA receptors containing an edited GluA2 exhibit either linear or outwardly rectifying current–voltage (I-V) relationships and have low permeability to Ca2+, whereas those lacking GluA2 show strong inward rectification and high Ca2+ permeability (Hollmann et al., 1991; Hume et al., 1991; Verdoorn et al., 1991; Mosbacher et al., 1994; Tsuzuki et al., 2000). In Purkinje cells, AMPA receptors exhibit a linear I-V relationship and thereby little Ca2+ permeability (Tempia et al., 1996; Momiyama et al., 2003), indicating that GluA2-containing receptors are the major form in this neuron. Consistent with this notion, high levels of GluA2 mRNA are expressed together with GluA1 and GluA3 mRNAs in Purkinje cells (Keinänen et al., 1990; Pellegrini-Giampietro et al., 1991; Lambolez et al., 1992). In the present study, we showed that these three subunits were localized on Purkinje cell spines forming synaptic contact with parallel fiber terminals and climbing fiber terminals. Taken together, AMPA receptors expressed in Purkinje cells are considered to be GluA1/GluA2 or GluA2/GluA3 heteromeric channels. In contrast, AMPA receptors lacking GluA2, such as GluA1/GluA3 heteromeric channels and GluA1 or GluA3 homomeric channels, are little expressed, if at all, in Purkinje cells. Notably, AMPA receptors remaining in γ-2-KO, γ-7-KO and DKO Purkinje cells all preserved the linear I-V relationship, even although GluA2 expression was significantly reduced in Purkinje cells of these KO mice.

From these findings, it can be assumed that in Purkinje cells the ablation of γ-2 causes severe reduction in GluA2/GluA3 channels, which results in severe reduction in AMPA receptor-mediated currents. The remaining GluA1/GluA2 channels probably mediate residual currents in γ-2-KO Purkinje cells. This large current deficit in γ-2-KO Purkinje cells suggests that GluA2/GluA3 channels are the predominant channel in Purkinje cells. This possibility appears to be supported by consistently much lower density of immunogold labeling for GluA1 than for GluA2 and GluA3 at the climbing fiber–Purkinje cell synapse (M. Fukaya, M. Yamasaki and M. Watanabe, unpublished observation). The large deficit may also reflect tonic enhancement of AMPA receptor channel function by γ-2 (Yamazaki et al., 2004; Kato et al., 2007, 2008). In contrast, similar levels of GluA1–GluA3 localization and AMPA receptor-mediated currents at γ-7-KO climbing fiber–Purkinje cell synapses suggest normal synaptic expression of GluA2/GluA3 and GluA1/GluA2 channels. By the ablation of both TARPs, however, GluA2/GluA3 channels are depleted almost completely and GluA1/GluA2 channels are also reduced substantially, leading to more severe deficits at all the biochemical, electrophysiological and behavioral levels. In future studies, it would be intriguing to pursue whether such a subunit-dependent regulation by multiple TARPs plays a role in activity-dependent insertion, internalization and recycling of GluA1/GluA2 and GluA2/GluA3 channels. These are considered to be key mechanisms underlying the changes in synaptic strength observed during several forms of long-term potentiation and long-term depression (Shi et al., 2001; Malinow & Malenka, 2002; Song & Huganir, 2002; Lee et al., 2004).

Promotion by γ-7 of expression of glial AMPA receptors

The synergistic promotion of synaptic GluA2–GluA4 expression by γ-2 and γ-7 was demonstrated reproducibly by Western blot, light microscopic immunohistochemistry and postembedding immunogold electron microscopy. By contrast, the lack of apparent reductions in synaptic localization of GluA1 and GluA4 in γ-7-KO mice (except for GluA4 at the mossy fiber–granule cell synapse) was inconsistent with their substantial reductions in cerebellar contents and immunohistochemical signals in the molecular layer. This discrepancy was explained by substaintal loss of GluA1 and GluA4 in Bergmann glia. Considering the vertical lines of immunostaining in the molecular layer (Fig. 2H), γ-7 may be expressed in Bergmann glia and promote AMPA receptor trafficking and expression in these glia. Secondary reduction of γ-7 in γ-2-KO cerebellum (Fig. 1E) might also account for the mild reduction in GluA1 and GluA4 signals in the molecular layer of γ-2-KO mice (Fig. 5). We can not exclude the possibility that GluA1 and GluA4 are also reduced at extrasynaptic or intracellular sites of Purkinje cells and interneurons in γ-2-KO and γ-7-KO mice.

Bergmann glia are specialized astrocytes thoroughly enwrapping the soma, dendrites and synapses of Purkinje cells (Yamada & Watanabe, 2002). Ca2+-permeable AMPA receptors are highly expressed in these glia (Burnashev et al., 1992; Müller et al., 1992), and the Ca2+ permeability has been shown to regulate the enwrapping of Purkinje cell synapses, efficient glutamate removal and rearrangement of neural circuits (Iino et al., 2001). Therefore, the promoting role of glial AMPA receptor expression by γ-7 probably plays an important role in synaptic development and function of Purkinje cells. Considering that Bergmann glia also express TARPs γ-4 and γ-5 (Fukaya et al., 2005), regulation of glial AMPA receptors by γ-4, γ-5 and γ-7 needs to be addressed in a future study.

Acknowledgments

We thank E. Kushiya for technical assistance. This investigation was supported in part by Grants-in-Aid for Scientific Research 17023021 (M.K.), 21220006 (M.K.), 21300118 (K.S.) and 17023001 (M.W.), Special Coordination Funds for Promoting Science and Technology, Grant-in-Aid for Young Scientists (B), 18700311 (M.Y.) and the Strategic Research Program for Brain Sciences (Development of Biomarker Candidates for Social Behavior) from the Ministry of Education, Culture, Sports, Science and Technology, Japan.

Abbreviations
AMPA

α-amino-3-hydroxyl-5-isoxazolepropionate

CF-EPSC

climbing fiber-mediated EPSC

DKO

double-KO

EPSC

excitatory postsynaptic current

FISH

fluorescent in situ hybridization

GLAST

glutamate–aspartate transporter

Glu

glutamate

GluR

Glu receptor

I-V

current–voltage

KO

knockout

PSD

postsynaptic density

TARP

transmembrane AMPA receptor regulatory protein

WT

wild-type

Ancillary