The Rack1^F/F lines were generated as previously described, in which exon 2 of Rack1 gene was flanked by loxP sites (Zhao et al., 2015). Homozygous Rack1^F/F mice were crossed with mice expressing a transgene encoding Cre recombinase driven by Pcp2 promoter (Barski et al., 2000). Conditional knockout mice were generated by the second generation, and Rack1^F/F littermates served as wild-type controls. All experiments with animals were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of Beijing Institute of Basic Medical Sciences. Mice were housed in specific pathogen-free conditions with 12/12-h light/dark cycles at Beijing Institute of Basic Medical Sciences.

It was performed as previously described (Wu et al., 2015; Yang et al., 2019). Briefly, frozen sections were washed 10 min with 0.5% phosphate-buffered saline with Tween 20 (PBS-T) for three times and then blocked with 3% bovine serum albumin for 1 hr. After that, sections were incubated overnight at 4°C with the primary antibodies as follows: Calbindin (C9848, Sigma, 1:400), NeuN (MAB377, Millipore, 1:400), brain lipid binding protein (BLBP) (ab32423, Abcam, 1:500), Rack1 (R1905, Sigma, 1:400). The sections were washed 10 min with 0.5% PBS-T for three times again and subsequently subjected to Alexa Fluor-conjugated secondary antibodies (Biotium, 1:500). Nuclear staining was visualized with a mounting medium with 4′,6-diamidino-2-phenylindole (ZSGB-BIO). All images were taken from a laser scanning confocal microscope (Olympus FV1200) and then were processed and analyzed by FV10-ASW or Image Pro Plus 6.0 software.

The sections (12 μm) of cerebellum mounted on gelatin-coated slides were washed 10 min with 0.5% PBS-T for three times and then immersed into 0.5% tar-violet solution for 20 min. The slices were then quickly rinsed in distilled water and differentiated in 95% ethanol for 2 min. Then, they were dehydrated in 75% ethanol twice, 3 min each. Finally, the slices were sealed with neutral resin.

The cerebellum were taken from mice at postnatal day 21 (P21) and then fixed in 2% formaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). After 12 h, the cerebellum were washed thoroughly and soaked in 0.1 M sodium dimethylarsenate buffer. The cerebellum was embedded in 4% agar and trimmed with a conventional microtome. After that, sections were fixed in 1% osmium tetroxide/1.5% potassium ferrocyanide solution for 1 h, washed three times in distilled water, incubated in 1% uranium peroxide acetate for 1 hr, washed twice in distilled water, and then dehydrated with gradient alcohol (50, 70, and 90%, 10 min each time; 100%, 10 min twice). Finally, the samples were incubated with propylene oxide for 1 h and then percolated overnight in a 1:1 mixture of propylene oxide and Epon (TAAB, United Kingdom). Next day, the samples were embedded in Epon and polymerized for 48 h at 60°C. Ultrathin sections (about 60–80 nm) were cut on Reichert Ultracut-S microtome sagittally and picked up on to a copper mesh stained with lead citrate. The formation of PF–PC synapses was observed by a transmission electron microscopy (Hitachi, H-7650) with an AMT 2k CCD camera.

Golgi staining was administrated with FD Rapid GolgiStain^TM Kit (PK401). Briefly, mice were deeply anesthetized before killing, and cerebellum was removed from the skull as quickly as possible, but handled carefully to avoid damaging or pressing of the tissue. Tissue was immersed in the impregnation solution made by mixing equal volumes of solutions A and B and was put aside at room temperature for at least 2 weeks in the dark, and then, tissue was transferred into solution C followed by storage at room temperature in the dark for 72 h. The 100-μm sections were cut on a vibrating slicer (Leica, VT1200 S). Each section was mounted on gelatin-coated microscope slides with solution C and dried naturally at room temperature. Sections were rinsed in double-distilled water twice, 4 min each, and then placed in a mixture consisting of one part solution D, one part of solution E, and two parts of double-distilled water for 10 min. Sections were dehydrated in 50, 75, 95, and 100% ethanol successively, 4 min each. Lastly, sections were cleared in xylene for three times, 4 min each, and finally sealed with neutral resin.

Cerebellum was removed on ice and placed in a homogenate tube. Homogenate buffer (0.5 g tissue/5 ml homogenate, 0.32 M sucrose, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid–NaOH, pH 7.4) was added and placed under an electric homogenizer for 20 times. The whole protein fraction was centrifuged at 1,000 × g/4°C for 10 min. Then, 4 ml of 1.2 M sucrose solution was added into the ultracentrifuge tube in advance, and the previously obtained supernatant was poured in the top. The ultracentrifuge tube was centrifuged at 160,000 × g/4°C for 15 min. The synaptic layer (between 1.2 M sucrose solution and homogenate buffer) was carefully aspirated, and 4 ml of homogenate buffer was added to mix it well. Four milliliters of 0.8 M sucrose solution was added into the ultracentrifuge tube, and the solution obtained in the previous step was slowly added above the 0.8 M sucrose solution. Similarly, the ultracentrifuge tube was centrifuged at 160,000 × g/4°C for 15 min. The supernatant was discard, and the pellet was resuspend with 1.6 ml of the resuspension buffer (0.5% Triton X-100, 0.16 M sucrose, 6 mM Tris–HCl, pH 8.1). The synaptic component was centrifuged at 32,800 × g for 20 min. The supernatant was thrown away, and 0.4 ml resuspension buffer was added to the pellet to centrifuge for 1 h at 200,000 × g. The precipitate obtained after centrifugation was the postsynaptic density (PSD) component.

The experiments were performed as previously described (Wu et al., 2012). Briefly, PSD fraction isolated from cerebellum tissues was supplemented with 1 × protease and phosphatase inhibitor mixture. Protein concentration was measured using the BCA Protein Assay Kit. Samples (20–50 μg, including 5 μl of prestained protein standards) were loaded into the sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel, and electrophoresis was conducted at constant voltage (120 V) at 4°C and then transferred to polyvinylidene difluoride membranes. The membranes were then blocked with 5% skim milk in 0.1% Tris–buffered saline/Tween-20 for 1 h and incubated overnight at 4°C with indicated primary antibodies. In each experiment, horseradish peroxidase and enhanced chemiluminescence were used to image protein bands on film. The film signal was electronically scanned and statistically analyzed by Image Pro Plus software.

Serial cerebellar coronal cryostat sections were stained with cresyl violet. Golgi-stained brain slices of the whole PCs were taken with an Olympus microscope. One-micrometer-spaced Z-stack brightfield images for dendritic spines were taken with an Olympus BX60 microscope with Axiocam MRc Zeiss camera and Axiovision 4.8 Software (Zeiss, Germany). All images are processed and quantified using Adobe Photoshop CS6 version and ImageJ Software. Spine morphology was determinated based on previous studies (Lee et al., 2004). Spine density was evaluated as the relative spine number over 10-μm dendritic fragments with NeuronStudio software.

The data between the two independent groups were shown as mean ± SEM of at least three independent experiments. p-values were determined by Student’s t-test or non-parametric test, and p < 0.05 was considered statistically significant.

Expression of Rack1 protein was fairly rich in mice at birth but decreased gradually at approximately postnatal day P14 and remained constant thereafter (Yang et al., 2019). Immunofluorescent colocalization revealed that Rack1 protein was mainly expressed in PCs at P21 cerebellum. To investigate the potential function of Rack1 in PCs in vivo, the conditional knockout mice were generated with hybridization between Pcp2-Cre transgenic lines and Rack1 loxP mice. At first, the Pcp2-Cre recombinase was visualized by Ai9 reporter mice. It exhibited that abundant Pcp2-Cre recombinase was specifically expressed in PCs (Figures 1A,B). Selective deletion of Rack1 in PCs were confirmed by both Western blot from isolated synaptic fraction (Figure 1C) and immunofluorescence staining (Figure 1F). It showed robust Rack1 protein decline in Rack1 mutants (Figure 1D, 32.8 ± 1.96 in mutant vs. 100.0 ± 2.78 in control, p = 0.0009, n = 5). In general, Rack1 mutant mice appeared normal at P21, as shown by similar body weight and cerebellar surface fissure compared to wild-type littermates (Figure 1E).

Next, we further precisely examined the cerebellar size by quantitative histological analysis. Nissl staining of cerebellar sections indicated that the foliation of Pcp2-Cre;Rack1^F/F mutant mice was similar to that of the control littermates postnatally except for lobule VII until P60 (Figure 1G). Sagittal sections of cerebellar vermis indicates that the area of lobule VII but not other lobules was significantly smaller in Pcp2-Cre;Rack1^F/F mutant mice compared to control littermates at P14 (0.18 ± 0.02 mm² in mutants vs. 0.29 ± 0.02 mm² in wild-type controls, p = 0.007, n = 5) and P21 (0.21 ± 0.05 mm² in mutants vs. 0.31 ± 0.02 mm² in wild-type controls, p = 0.008, n = 5), but not P60 (0.33 ± 0.02 mm² in mutants vs. 0.35 ± 0.02 mm² in wild-type controls, p = 0.55, n = 5), suggesting that the ablation of Rack1 in PCs causes the delayed foliation and morphogenesis most specifically restricted to lobule VII, but not other lobules in vermis (Figure 1H). Interestingly, it should be noted that the alterations are only restricted to the vermis subregion but not other cerebellar hemispheres for some unknown reason. Since, the expansion of GNPs is crucial for cerebellar foliation formation, this phenotype is probably due to defects in delayed GNPs proliferation and migration at specific subregion in mutant mice.

To further evaluate the effect of Rack1 knockout in PCs on locomotion, 8-week-old Pcp2-Cre;Rack1^F/F and control mice were selected for balance-related behavioral testing. Pcp2-Cre;Rack1^F/F mice did not show obvious ataxia in standard cages. However, they performed poorly, with a remarkably longer time when walking on a narrow elevated beam (Figure 2A, 3.6 ± 0.22 s in mutant vs. 8.0 ± 0.86 s in control, p < 0.0001, n = 10), which indicated that the balance ability of mutant mice was significantly decreased. In addition, the time of Pcp2-Cre;Rack1^F/F mice staying on the accelerating rotarod was significantly shorter than that of the control mice (Figure 2B, 211.4 ± 20.11 s in mutant vs. 298.1 ± 1.81 s in control, p < 0.0001, n = 11), indicating the deficits in fine motor coordination skills in Pcp2-Cre;Rack1^F/F mutants. Moreover, Rack1 mutant mice also showed impaired motor learning, in which mutant mice exhibited declined improvement after multiple sessions on the accelerating rotarod compared with controls (Figure 2B).

In addition to motor function, the cerebellum has been implicated in various cognitive and social behaviors. The dysfunction of PCs have been observed in autism and schizophrenia (Andreasen and Pierson, 2008; Yeganeh-Doost et al., 2011; Peter et al., 2016). Thus, a series of behavioral tests were performed to assess whether Rack1 ablation in PCs would affect different domains of mouse behavioral repertoire. In the open field tests, Rack1 mutant mice exhibited dramatic hyperactivity (Figure 2C). In a limited time (5 min), the mutant mice travel more distances compared to controls (Figure 2D, 53.6 ± 8.6 cm in mutant vs. 35.2 ± 9.6 cm in control, p = 0.0062, n = 8), especially at the center zone of the place (Figure 2E, 8.3 ± 4.2 cm in mutant vs. 3.6 ± 1.9 cm in control, ^∗∗p = 0.0096, n = 8; Figure 2F, 65.6 ± 18.6 s in mutant vs. 45.2 ± 19.4 s in control, ^∗∗p = 0.0624, n = 8). Together, these results indicate that Pcp2-Cre;Rack1^F/F mice exhibit hyperactivity and a significant deficit in motor coordination.

Moreover, immunofluorescence staining was employed to identify PCs and Bergmann glial cells using Calbindin and BLBP antibodies, respectively, to analyze the fine lamination and morphological differences in mutant cerebellum. As was shown in Figure 3A, in the mutant cerebellum, the Calbindin⁺ PCs and BLBP⁺ Bergmann glial cells were both well and neatly organized, with a single layer of polarization distributed in the PC layer, suggesting the normal cerebellar cortex stratification in Rack1 mutant mice. Nevertheless, due to the excessive number of Calbindin⁺ PCs in the brain slices, it was almost impossible to accurately count the number of dendritic branches and dendritic spines of PCs. Therefore, the Golgi staining method was adopted to sparsely illustrate the morphology of PCs. In general, it showed that there is no obvious distinction of dendritic branches and spine density in the PCs between control and Rack1 mutant mice (Figure 3B). Statistical results also confirmed that dendritic spine density was comparable to that of control littermates (Figure 3C, 13.3 ± 0.39/10 μm in mutant vs. 15.0 ± 1.15/10 μm in control, p = 0.1878, n = 5). Thus, specific deletion of Rack1 in PCs does not impair its morphogenesis and cytoarchitecture.

Owing to the fact that PC dendrites could convert excitatory PF input from granule cells into signals, they play an important role in synaptic plasticity and motor learning (Rowan et al., 2018). Next, we asked whether the impaired motor coordination in Rack1 mutant mice was caused by the synaptogenesis deficiency or synaptic dysfunction. Therefore, we assessed the effect of Rack1 ablation in PCs on the synapse formation between PCs dendrites and PFs from granule cells by electronic microscope analysis. As shown in Figure 3D, the black high-density postsynaptic materials marked by the red asterisks was the excitatory synapse formed by PCs and PFs. Quantitative analysis shows that the density of presynaptic PF boutons in Rack1 mutant mice was substantially lower than that of control (Figure 3E, 5.4 ± 0.56/μm² in mutant, n = 17, vs. 7.8 ± 0.69/μm² in control, n = 22, p < 0.001). We also found there were more free or mismatched spines in Rack1 mutant cerebellum, suggesting the defective synaptogenesis in mutant mice. Together, these ultrastructural results suggest that Rack1 is able to promote synaptogenesis between PC dendrites and PFs in the cerebellar cortex.

The requirement for the Rack1 in synaptogenesis in the cerebellar cortex led to the postulation that synaptic transmission at PF–PC synapses might be affected in Rack1 mutants. Our electrophysiological analysis in acute cerebellar slices revealed that the amplitude of evoked EPSC at PC–PF synapses was normal between Pcp2-Cre;Rack1^F/F and Rack1^F/F mice (Figures 4A,B, 280.2 ± 44.45 pA, n = 10 in mutant vs. 279.2 ± 28.63 pA, n = 14 in control, p = 0.9855). However, the ratio of paired-pulse facilitation measured at an interval of 80 ms was reduced nearly 20% (Figures 4C,D, 1.7 ± 0.10 in mutant vs. 2.2 ± 0.14, p = 0.0078, n = 12), suggesting that the presynaptic glutamate release was impaired. Moreover, there was no difference in the amplitude of the presynaptic volley in mice between Pcp2-Cre;Rack1^F/F mutant and Rack1^F/F control mice, illustrating that the impairment of evoked EPSC in Rack1 mutant mice is not due to the difference in axon excitability (data not shown).

Then, we asked whether the neurotransmission defects in Rack1 mutant mice might be due to the secondary effect of the changed synapse numbers. As expected, the frequency of mEPSCs in PCs was reduced in acute cerebellar slices from Pcp2-Cre;Rack1^F/F mice compared to Rack1^F/F mice (Figure 4F, 0.14 ± 0.02 pA, n = 17 in mutant vs. 0.30 ± 0.03 pA, n = 20 in control, p = 0.0002), consistent with the previous conclusion that synapse number is reduced in Rack1 mutants. The amplitude of mEPSCs was hardly reduced in Rack1 knockout mice (Figure 4E, 11.99 ± 0.54 pA, n = 17 in mutant vs. 12.02 ± 0.55 pA, n = 20 in control, p = 0.9724), suggesting that the reactivity of the postsynaptic membrane was normal. Collectively, our electrophysiological and electronic microscope results suggest that the deficiency of synaptic transmission between PF-PCs in Rack1 mutants might be due to the impaired synaptogenesis.

Long-term depression has been proposed as a potential factor contributing to motor learning in the cerebellar cortex (Kakegawa et al., 2018; Zamora Chimal and De Schutter, 2018). Given that Pcp2-Cre;Rack1^F/F mutant mice showed impaired motor coordination and hyperactive locomotion, we next examined the LTD at PF–PC synapses with voltage-clamp mode both in control and Rack1 mutants (Figure 4G). Our results showed that PCs in Rack1^F/F mice performed robust PF-LTD induction (Figure 4H, last 5 min, 47.4 ± 5.73% of baseline; n = 9) in response to repetitive PF stimulation. However, the induction of LTD in PCs was significantly impaired in Pcp2-Cre;Rack1^F/F mutant cerebellum (Figure 4H, last 5 min, 82.6 ± 7.14% of baseline; n = 8, p = 0.0015).

Motor PF-LTD deficits may result from altered ion channel or glutamatergic-transmission-associated protein. Thus, the expression of ionotropic glutamate receptors were examined. We found that the level of AMPA-associated proteins such as GluA1/2/4 and transmembrane AMPA-receptor-regulated protein γ2/8 were not altered significantly in relation to those of PSD95 and GAD65 (Figures 5A,B, p > 0.05). Glutamatergic-transmission-associated protein such as membrane-associated guanylate kinase, Dynamin1, and Synapsin were also indistinguishable between Rack1^F/F control and Pcp2-Cre;Rack1^F/F mutant cerebellum (Figures 5A,B, p > 0.05), suggesting that deletion of Rack1 in PCs has no significant effects on the expression of synaptic proteins.

Our previous work has shown that Rack1 promotes the development of granule cells via regulating the stability of HDAC1/2 in GCPs (Yang et al., 2019). Here, we also found that the expression of HDAC1/2 proteins were significantly decreased in Pcp2-Cre;Rack1^F/F mutant cerebellum compared to Rack1^F/F control (Figures 5A,B, 93.5 ± 1.37% in mutant vs. 100.0 ± 1.58% in control, n = 3, p = 0.0092). Interestingly, previous work has shown that HDAC1/2 together with Chd4, RbAp48, Mbd3, and Mta1/2 constitute the remodeling of nucleosome and deacetylation complex, which programs the differentiation of presynaptic sites and triggers synaptic connectivity in the PF–PCs (Yamada et al., 2014). Thus, the decreased expression of HDAC1/2 in PCs might be responsible for the malformation of PF–PCs synaptogenesis.