Novel role of Rac-Mid1 signaling in medial cerebellar development

During brain development, migration of immature neurons from germinal zones to their final destination is essential for the establishment of proper circuitry. Two types of neuronal migration – tangential and radial – are required, and radial migration may occur in both inward and outward directions during cerebellar development (Chédotal, 2010; Wang and Zoghbi, 2001). Cerebellar granule neuron (CGN) progenitors are derived at E13-16 from the upper rhombic lip (RL), which is the primary germinal zone of CGNs. Beginning at E13, these progenitor cells migrate tangentially to form the surface of the cerebellar primordium: the external granule layer (EGL), which is the secondary germinal zone of CGNs. After proliferation in the EGL, CGN progenitors begin to differentiate and migrate to deeper layers. Subsequently, CGNs develop inward dendritic processes for radial migration (Kawaji et al., 2004). Inward radial migration, which is completed by P20 in mice (Hatten et al., 1997), is a major characteristic of cerebellar development. By contrast, Purkinje cells (PCs) migrate outwardly from the ventricular zone.

Rac (Rac1-3) is a member of the Rho family of small GTPases and plays important roles in neuritogenesis, including axonogenesis and dendritogenesis, neuronal migration and modulation of gene transcription (Bosco et al., 2009; Funahashi et al., 2014; Govek et al., 2005). Previous studies had shown that, although conditional Rac1 knockout (KO) in telencephalon-derived neurons produced mild impairments in radial migration and axon guidance (Chen et al., 2007; Kassai et al., 2008), these phenotypes were apparently milder than those observed in studies using an electroporated dominant-negative Rac1 mutant (Kawauchi et al., 2003). Rac1-KO mice generated using a Nestin-Cre system exhibited impaired radial migration and axonogenesis in CGNs (Tahirovic et al., 2010). Furthermore, although a relatively large number of downstream molecules and putative disease-related pathways have been suggested to involve Rac signaling, the number of diseases in which involvement of Rac-associated molecules/pathways has been established is quite low (Govek et al., 2005; Stankiewicz and Linseman, 2014).

Midline 1 (MID1) is the gene responsible for Opitz G/BBB syndrome (OS), which is characterized by malformations that affect midline structures, such as laryngotracheoesophageal, cardiac and urogenital abnormalities. Mid1 is a microtubule-associated E3 ubiquitin ligase (Schweiger and Schneider, 2003) that regulates the degradation of the catalytic subunit of protein phosphatase 2A (Ppp2ca) via interaction with the α4 subunit of Pp2a (Trockenbacher et al., 2001). One-third of individuals with OS also exhibit developmental defects of the CNS, which mainly consist of hypoplasia/agenesis of the medial cerebellum (Fontanella et al., 2008; Pinson et al., 2004); however, clinical manifestations of OS in the CNS, as well as in the body, show variable expressivity. Indeed, although Mid1-KO mice have been reported to show hypoplasia of the anterior portion of the medial cerebellum, no other abnormalities were observed (Lancioni et al., 2010). Furthermore, no correlation between the position of the mutation in MID1 and clinical phenotypes of OS has been reported (Fontanella et al., 2008). Thus, the underlying pathogenesis of OS remains to be fully elucidated.

Here, we have investigated the association between downstream targets of Rac and developmental abnormalities using mice with double KO (DKO) of Rac1 and Rac3 in CGNs (Atoh1-Cre;Rac1^flox/flox; Rac3^−/−). Impaired final differentiation with dendritogenesis of CGNs resulted in agenesis of the medial internal granule layer (IGL) due to disruptions in the radial migration of CGNs. Furthermore, our findings suggest that Mid1 is a novel and transcriptionally regulated Rac target, and that Rac1-Mid1-mTORC1 signaling may be involved in producing agenesis of the medial IGL of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum.

To study the role of Rac GTPases in CGNs, we examined mRNA expression of Rac1 (widely expressed) and Rac3 (neuron specific) in the cerebellum using in situ hybridization. At P6, which is when CGNs begin to migrate radially to form the IGL (Hatten et al., 1997), Rac1 and Rac3 mRNAs were expressed almost exclusively in the EGL (Fig. 1A), where almost all cells are CGNs. At P21, patterns of Rac1 and Rac3 expression became PC dominant, particularly those of Rac3 (Fig. 1B). Slender cells in the molecular layer (ML), which are implicated in the radial migration of CGNs, were Rac1 positive (Fig. S1A). Although a previous study reported an absence of Rac3 expression in the cerebellum (Tahirovic et al., 2010), our results suggested that both Rac1 and Rac3 are expressed in the developing CGNs and that Rac1 expression is more dominant than that of Rac3 (Fig. S1B).

We generated Rac1-KO mice using Atoh1-Cre mice and Rac1^flox/flox mice, hereafter referred to as Atoh1-Cre;Rac1^flox/flox mice. Atoh1 is a transcription factor that is essential for the genesis of CGNs (Ben-Arie et al., 1997). Atoh1-positive cells emerge as early as E9.5 in the RL (Akazawa et al., 1995), and Atoh1-positive CGN progenitors migrate from the RL to form the EGL between E12.5 and E16.5 (Wang et al., 2005). The pattern of Atoh1-Cre-driven recombination was assessed using Atoh1-Cre;tdTomato mice. Consistent with the findings of previous reports, tdTomato fluorescence was observed in the developing EGL at the surface of the primordial cerebellum of Atoh1-Cre;tdTomato mice at E16.5, in the EGL at P0, and in both the EGL and IGL at P9. However, no such findings were observed in control mice (CAG-Stop^flox-tdTomato) (Fig. 2A). Although Nestin-Cre;Rac1^flox/flox mice, in which the Nestin promoter functions in the cerebellum from E15.5 as well as in all neuronal progenitors in the cerebrum around E10.5 (Graus-Porta et al., 2001), were reported to exhibit poor balance (Tahirovic et al., 2010), we detected no obvious behavioral phenotype in Atoh1-Cre;Rac1^flox/flox mice. Therefore, we generated DKO mice of Rac1 and Rac3 (Atoh1-Cre;Rac1^flox/flox;Rac3^−/−) by crossing Atoh1-Cre;Rac1^flox/flox mice with Rac3-null mice, which have been reported to show normal microscopic development of the brain (Corbetta et al., 2005). Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice exhibited severe ataxic gait when they began walking, and the phenotype was retained throughout life (Movie 1).

Nissl-stained coronal sections of the cerebellum from 9-week-old Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice exhibited agenesis of the medial IGL (Fig. 2B). Rac1^flox/flox;Rac3^−/− (Rac3 KO) mice and Atoh1-Cre;Rac1^flox/+;Rac3^−/− mice had normal cerebella compared with Atoh1-Cre mice (Fig. 2B). Rac1^flox/flox;Rac3^−/− mice were subsequently used as controls, in addition to Atoh1-Cre mice and Rac1^flox/flox mice. Stepwise exacerbation of hypoplasia of the medial IGL was observed in the cerebellum of Atoh1-Cre;Rac1^flox/flox (Rac1 KO) mice and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− (Rac1/Rac3 DKO) mice following the additional deletion of Rac3 (Fig. 2B). Atoh1-Cre;Rac1^flox/flox;Rac3^−/+ mice exhibited mild ataxic gait when they began to walk, although the phenotype of these mice was much milder than that of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice throughout life. The thicknesses of the medial IGL in the control, Atoh1-Cre;Rac1^flox/flox and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebella were 143.8±3.1, 40.0±2.2 and 7.5±2.7 µm, respectively (Fig. 2B). In the medial Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum, the ML, PCL and IGL could not be fully distinguished (Fig. 2B). The affected regions (ML, PCL and IGL) of Atoh1-Cre;Rac1^flox/flox;Rac3^−/+ and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebella were occupied by multiple layers of lightly stained large cells and densely stained small cells (insets in Figs 2B and 3A,B), whereas Atoh1-Cre;Rac1^flox/flox cerebellum exhibited an aligned PCL and thin IGL (insets of Fig. 2B). The lightly stained large cells and some of the densely stained small cells were confirmed as calbindin-positive PCs and tdTomato-positive CGNs, respectively, using 12-week-old Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato mice (Fig. S2).

We then aimed to determine the proportion of the IGL affected by DKO of Rac1 and Rac3 using Nissl-stained serial sagittal (Fig. 3A) and coronal sections (Fig. 3B) of cerebellar tissue from 8-week-old mice. The anterior medial region (anatomically almost identical to the anterior cerebellar lobe) of the IGL (e.g. I-VII folia in the sagittal section) was hypoplastic/aplastic in the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum.

We investigated the developmental processes of the EGL and IGL in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice from P4 to P17, when proliferation of CGNs in the EGL and the subsequent formation of the IGL are robust (Hatten et al., 1997). At P4, coronal sections of the cerebellum from Rac1^flox/flox;Rac3^−/− mice and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice showed comparable EGL thickness and lobular morphology (Fig. 4A). At P7, the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum was apparently smaller than the Rac1^flox/flox;Rac3^−/− cerebellum (Fig. 4A). Although the thickness of the EGL of the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum increased, the morphology of the EGL became abnormal, particularly in the medial cerebellum (Fig. 4A). At P10, the appearance of the IGL was abruptly absent in the medial Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum, and this phenotype was retained at P17 in combination with the disappearance of the EGL (Fig. 4A).

To examine the mechanism of agenesis in the medial IGL of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice between P7 and P10, we used active caspase 3 immunostaining to detect apoptotic cells. Many apoptotic cells were detected in the medial Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum but not in the Rac1^flox/flox;Rac3^−/− cerebellum at P8 (Fig. 4B), whereas no apoptotic cells were detected in the Atoh1-Rac1^flox/flox;Rac3^−/− cerebellum at E16.5, P1 or P3. The number of apoptotic cells in the Atoh1-Cre;Rac1^flox/flox;Rac3^−/+ cerebellum at P8 was ∼50% of that in the Atoh1-Rac1^flox/flox;Rac3^−/− cerebellum. Moreover, apoptotic cells were accumulated in the deep layer of the medial EGL of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice (Fig. 4B).

To verify the mechanism underlying decreased cerebellum size in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice relative to Rac1^flox/flox;Rac3^−/− mice, we examined sagittal sections of the cerebellum at E16.5. The distance from the RL to the most anterior point of the EGL, which is the longest path of tangential migration, was analyzed in four sections corresponding to those depicted in Fig. 3A; tangential migrations of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− CGNs were significantly shorter than those of Rac1^flox/flox;Rac3^−/− CGNs in all four sections (Fig. 5). However, no significant difference in the thickness of the primordium of the EGL was observed between Atoh1-Cre;Rac1^flox/flox;Rac3^−/− and Rac1^flox/flox;Rac3^−/− mice for any of the four sections (Fig. 5). To further examine the proliferative capability of CGNs, we applied EdU pulse labeling to cerebellum at P4, when proliferation of CGNs is high (Hatten et al., 1997). No significant differences in EdU labeling index (ratio of Edu-positive cells/DAPI-positive cells) in the EGL were observed between Atoh1-Cre;Rac1^flox/flox;Rac3^−/− and Rac1^flox/flox;Rac3^−/− mice at 4 h after EdU injection (Fig. 6A,B). At 30 h after injection, EdU-labeled CGNs were observed throughout the EGL (Fig. 6A).

Thus, these results suggested that the smaller size of the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum is due to impaired tangential migration, and that agenesis of the IGL in the medial Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum is caused by apoptosis in the EGL following normal proliferation of CGNs and development of the EGL.

To examine differentiation capabilities of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− CGNs, we performed immunostaining using: Pax6, a marker for developing CGNs as well as CGNs in the IGL (Yamasaki et al., 2001); p27^Kip1, a marker for CGNs located in the deep region of the EGL (Mulherkar et al., 2014; Wang and Zoghbi, 2001); and NeuN, a marker for postmitotic CGNs (Mulherkar et al., 2014) at P5. All CGNs in the EGL were Pax6 positive (Fig. 7A), whereas CGNs in the lower one-third to one half of the EGL were strongly p27^Kip1 positive (Fig. 7C). CGNs in the very lowest region (rows 1-3) of the EGL were strongly NeuN positive (Fig. 7D). No significant difference in cerebellar immunostaining for any marker was observed between Atoh1-Cre;Rac1^flox/flox;Rac3^−/− and Rac1^flox/flox;Rac3^−/− mice (Fig. 7A-D). The ratio of cell numbers (p27^Kip1/DAPI and NeuN/DAPI) in the EGL (statistically analyzed in the anterior and posterior lobe of cerebellum) did not significantly differ between Rac1^flox/flox;Rac3^−/− and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice (Fig. 7B,C). All lobules (I-X) were observed in the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum (Fig. 7B).

These results suggest that Atoh1-Cre;Rac1^flox/flox;Rac3^−/− CGNs undergo apoptosis following differentiation, which appears normal until expression of NeuN occurs.

To examine neuritogenesis in Rac-deleted CGNs, we used primary cultures of CGNs. Using βIII-tubulin immunostaining, we observed that Atoh1-Cre;Rac1^flox/flox CGNs exhibited moderately impaired neuritogenesis compared with Rac1^flox/flox and Rac1^flox/flox;Rac3^−/− CGNs, and that this impairment was exacerbated in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− CGNs (Fig. 8A).

We then used cerebellar microexplant cultures to further investigate the involvement of Rac in neuritogenesis and migration. Previous research has indicated that, after their final mitosis, CGNs remain in the EGL for 20-48 h. During this period, CGNs change their morphology from round to unipolar to bipolar to tripolar. In the bipolar phase, CGNs reach the deep layer of the EGL. CGNs then inwardly extend a single vertical process and initiate radial migration to form the IGL (Komuro et al., 2001). The horizontal and vertical processes of CGNs are reported to differentiate into the parallel fiber axons and dendrites, respectively (Kawaji et al., 2004). Parallel neurites projecting from the microexplant and neurites perpendicular to the parallel neurites have been reported to mimic the horizontal axons and vertical dendrites of CGNs in vivo, respectively. Biphasic migration, which includes early migration along the parallel neurites and later migration along the perpendicular neurites, observed in the microexplant culture have been used as models of tangential and radial migration in vivo, respectively (Kawaji et al., 2004). The parallel neurite length stained by βIII-tubulin projected from the microexplant (Fig. 8B) and tangential migration, defined as the migration distance along the parallel neurite from the microexplant (Fig. 8B and Fig. S3A), were significantly shorter in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− microexplants than in Rac1^flox/flox;Rac3^−/− microexplants. Furthermore, although Rac1^flox/flox;Rac3^−/− microexplants exhibited regularly aligned parallel neurite bundles in combination with Map2-positive perpendicular neurites, Atoh1-Cre;Rac1^flox/flox;Rac3^−/− microexplants exhibited impaired alignment of βIII-tubulin-positive neurites with few Map2-positive neurites (Fig. 8C), suggesting severely impaired dendritic development and radial migration in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− CGNs.

Next, we examined the vertical neurites projecting from CGNs for radial migration in vivo using Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato mice, in which Rac1/Rac3-DKO CGNs were labeled by tdTomato fluorescence.

Because tdTomato fluorescence in neurites of CGNs was not strong enough for detailed evaluation, and because Map2 immunostaining exhibits a strong non-specific background, tdTomato fluorescence from vertical neurites was enhanced by immunostaining using an RFP polyclonal antibody. Although vertical neurites became visible, horizontal neurites were impossible to evaluate due to the strong tdTomato fluorescence from crowded somata of CGNs in the EGL. Vertical inward neurites projecting from radially migrating CGNs were observed in the Atoh1-Cre;tdTomato cerebellum at P8 (Fig. 8D). In sharp contrast, they were short and poorly visualized in the anterior medial cerebellum of Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato mice (Fig. 8D). However, vertical neurites were visible in the posterior medial Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato cerebellum (Fig. S3B).

To verify the underlying mechanism of impaired development of the anterior medial cerebellum in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice, we applied DNA microarray analyses to the medial cerebellum at P6, which is just prior to the robust radial migration of CGNs (Hatten et al., 1997) and the emergence of the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− phenotypes. Among the genes upregulated or downregulated by more than threefold in the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum relative to the Rac1^flox/flox;Rac3^−/− cerebellum, we focused on those encoding cytoskeleton-associated proteins. We revealed that Mid1, which is responsible for abnormalities in midline structures observed in OS, was decreased in the Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebellum compared with control (0.297-fold when normalized using the 75th percentile, 0.289-fold when normalized to Actg1 and 0.297-fold when normalized to Tuba1a). The effect of Rac KO on Mid1 expression detected by DNA microarray analyses was confirmed in the dissected cerebellar cortex, which mainly consists of the EGL, using quantitative PCR (qPCR) (Fig. 9A). Furthermore, we confirmed specific localization of Mid1 mRNA in the EGL as well as decreased Mid1 mRNA in both the EGL of the anterior and posterior lobe of the cerebellum in Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice using in situ hybridization (Fig. S4).

Because the thickness of the medial IGL in the Atoh1-Cre;Rac1^flox/flox cerebellum was 25-30% of that in the control (Rac3-KO) cerebellum (Fig. 2B), we expected that the transcriptional regulation of Mid1 by Rac could be determined by the knockdown (KD) experiment of Rac1 using a validated Rac1-siRNA (Ueyama et al., 2006). A mild decrease in Mid1 mRNA in response to Rac1 KD was observed in rat-derived RN46A neuronal cells using qPCR (Fig. 9B). Transcriptional regulation of Mid1 by Rac was further confirmed by the following results in RN46A cells: decreased Mid1 mRNA levels following treatment with a novel Rac-specific inhibitor (EHop-016; Fig. 9C); and increased Mid1 mRNA levels following expression of a Rac-specific guanine nucleotide-exchange factor (GEF) (Tiam1; Fig. 9D). Increases in Mid1 protein levels following expression of Tiam1 were confirmed in human-derived HEK293 cells (Fig. 9E).

Most mammals have a basic pattern of 10 major lobules (I-X) in the cerebellar vermis. We demonstrated that Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice exhibited agenesis of the IGL, predominantly in the anterior medial cerebellum. Similarities in abnormal alterations observed in the anterior medial cerebellum of Mid1-KO mice (I-III) (Lancioni et al., 2010) and our Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice (I-VII) prompted us to further examine the function and signaling of Mid1 in CGNs. Normal development of the posterior medial cerebellum in both Mid1-KO mice and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice may be explained by the lateral-to-medial movement of CGNs that specifically occurs in the posterior lobe during cerebellar development (Chédotal, 2010; Sgaier et al., 2005).

A recent study demonstrated that inhibition/depletion of Mid1 results in the downregulation of mTORC1 signaling (Liu et al., 2011). This report prompted us to examine a new signaling pathway: Rac1-Mid1-mTORC1. In rat-derived PC12 neuronal cells with Rac1 KD, the protein levels of Ppp2ca, a well-established target of Mid1, were increased relative to control (Fig. 10A). In addition, levels of S6 ribosomal protein and 4E-BP1 phosphorylation, which are well-known downstream targets of mTORC1, were decreased (Fig. 10A). The decreased levels of Mid1 were confirmed using qPCR (Fig. 10A). Decreased levels of Mid1 and mTORC1 signaling (represented by the decreased phosphorylation of 4E-BP1) following Rac deletion were observed in both the medial and lateral cerebellum of Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice, using the dissected cerebellar cortex, which mainly consists of the EGL (Fig. 10B). Taken together, these results indicate that Rac is involved in the regulation of the Mid1-mTORC1 signaling axis.

Finally, we examined the effects of Mid1 KD on neuritogenesis, mTORC1 signaling and migration using mouse primary CGNs and microexplants. Two effective siRNAs for Mid1, siMid1-832 and siMid1-1107 (Fig. 10C) revealed a mild reduction of neuritogenesis (Fig. 10C). The decreased levels of Mid1 were confirmed using qPCR (Fig. 10C). A similar Mid1-KD effect on neuritogenesis was also observed in the experiment using rat primary CGNs and rat-specific siRNA for Mid1 (Fig. 10D). The increased levels of Ppp2ca and the reduced mTORC1 signaling (represented by decreases in the phosphorylation of S6 ribosomal protein) were also confirmed in Mid1-KD (siMid1-832) primary CGNs (Fig. 10E). Furthermore, decreased cell migration was observed in Mid1-KD CGNs of microexplant cultures (Fig. S5).

In the present study, we revealed that both Rac1 and Rac3 mRNAs are expressed at the highest levels in CGNs during cerebellar development. Notably, although Rac1 mRNA was still expressed in CGNs at P21, Rac3 mRNA expression disappeared following cerebellar development. Furthermore, the phenotypes of Rac1-KO mice were exacerbated and became tangible by the additional KO of Rac3. These results suggest that both Rac1 and Rac3 function in CGNs during cerebellar development, and that Rac3 is particularly important for CGNs in this period.

Although we demonstrated that both axonogenesis and dendritogenesis, and both tangential and radial migration were impaired in Rac1/Rac3-DKO mice, dendritogenesis and radial migration were more severely impaired both in vitro (cerebellar microexplant) and in vivo (cerebellum). Our findings are in accordance with those of previous reports that have revealed: (1) decreased levels of Map2 and impaired dendrite branching in primary CGNs from Vav3, a GEF for Rac, KO mice (Quevedo et al., 2010); (2) impaired dendritogenesis but not axonogenesis in Rac1-KD primary hippocampal neurons from Rac3-KO mice (Gualdoni et al., 2007); (3) Rac activation by laminin (Gu et al., 2001); and (4) radial migration observed only on laminin-coated dishes in cerebellar microexplant cultures (Kawaji et al., 2004; Nagata and Nakatsuji, 1990).

CGN progenitors migrate tangentially from the RL to form the EGL during the initial stage of cerebellar development (until E16.5) (Sgaier et al., 2005). Rac1-dependency in the tangential migration of olfactory and cortical interneurons has been reported (Chen et al., 2007). In the present study, we demonstrated impaired tangential migration during the formation of the most anterior region of the cerebellar primordium, which requires the longest path of migration, in Rac1/Rac3-DKO mice at E16.5. Because lobules I-X and their EGL were formed at P5, the impaired tangential migration may be related only to the smaller size of the Rac1/Rac3-DKO cerebellum. Impaired differentiation and apoptosis have been reported in Rac1-KO telencephalon-derived neurons (Chen et al., 2009). In addition, selective inhibition of Rac has been reported to induce apoptosis in CGNs (Stankiewicz et al., 2015). Rac1/Rac3-DKO CGNs exhibited: (1) a normal pattern of differentiation until the expression of NeuN, a marker for postmitotic CGNs, was seen in the deepest layer of the EGL at P5; (2) few Map2-positive neurites in cerebellar microexplants and disrupted development of vertical dendrites at P8; and (3) robust apoptosis at the deep layer of the EGL, a premigratory zone for preparing the subsequent radial migration (Yamasaki et al., 2001), at P8. Such results suggest that agenesis of the IGL is due to disruption of the radial migration in CGNs, which is caused by impaired final differentiation of CGNs with dendritogenesis. However, the precise cause of apoptosis in the premigratory zone of the EGL remains unknown.

In the present study, we revealed that Mid1 is a novel and transcriptionally regulated downstream target of Rac. The dynamic phosphorylation/dephosphorylation cycles regulated by Pp2a, a well-established downstream target of Mid1, play important roles in neuronal functions, including migration, neurite initiation and protein synthesis (Ayala et al., 2007; Basu, 2011; Laplante and Sabatini, 2012). Previous studies have reported that 4E-BP1, which is a phosphorylation target of mTORC1 for the promotion of protein synthesis (Nanahoshi et al., 1998), and Map2, which is a well-known dendrite marker involved in neurite outgrowth (Sanchez et al., 2000), are Pp2a targets. Indeed, we observed impairments in mTORC1 signaling and Map2 expression (dendritogenesis) via Rac KD and KO. The protein mTOR assembles into two complexes, mTORC1 and mTORC2, which can be distinguished by their associated proteins, raptor (Rptor) and rictor, respectively (Laplante and Sabatini, 2012). A recent study showed that neuron-specific inactivation of mTORC2 following Rictor KO resulted in a small brain, including a small cerebellum, although the development of the medial IGL was normal (Thomanetz et al., 2013). Meanwhile, another study revealed that neuron-specific inactivation of mTORC1 following Rptor KO resulted in microcephaly and lethality within a few hours of birth, prior to the completion of cerebellar development (Cloetta et al., 2013). These reports exclude the involvement of mTORC2, but not of mTORC1, signaling in CGNs of the medial cerebellum. Rptor-KO mice also exhibit reduced Map2 staining and reduced dendritic complexity in the cerebral cortex (Cloetta et al., 2013). Additionally, Mid1 has been reported to regulate the interaction between mTOR and raptor (Liu et al., 2011), and to regulate microtubule-associated mRNA transport and protein translation (Aranda-Orgillés et al., 2008). Activation of mTORC1 by brain-derived neurotrophic factor (BDNF), which induces Rac activation (Yoshizawa et al., 2005), has been reported to be involved in local protein synthesis in neuronal dendrites (Takei et al., 2004). Impaired protein synthesis in dendrites through impaired Rac-Mid1-mTORC1 signaling may be one of the causes of impaired dendritogenesis and apoptosis of Rac1/Rac3-DKO CGNs in the deep layer, a premigratory zone, of the EGL. However, rescue experiments using GFP-Mid1 into Rac1/Rac3-DKO primary CGNs showed no recovery of impaired neuritogenesis. Although relatively weak involvement of Rac1-mediated transcriptional regulation of Mid1 cannot be completely excluded, these results suggest the following possibilities: (1) in spite of the involvement of Mid1 in neuritogenesis, migration and mTORC1 signaling, Rac is more crucially and broadly involved in these functions than Mid1; and (2) Rac, Mid1 and mTORC1 synergistically act to carry out these functions. Indeed, the extent of cerebellar agenesis in Rac1/Rac3-DKO mice (I-VII) was greater than that of agenesis/hypoplasia reported in the Mid1-KO cerebellum (I-III) (Lancioni et al., 2010). Moreover, we identified the possibility that Rac1/Mid1/mTOR functions as a complex (Fig. S6), in accordance with the findings of previous reports: interactions of Mid1 with mTORC1 (Liu et al., 2011) and Rac1 with mTOR (Saci et al., 2011).

The most frequent CNS phenotype of OS involves agenesis/hypoplasia of the cerebellar vermis, followed by agenesis/hypoplasia of the corpus callosum (Fontanella et al., 2008; Pinson et al., 2004), suggesting that the activity of Mid1 in neurons may be modulated by other factors. Indeed, researchers have revealed agenesis of the vermis (Lancioni et al., 2010) and mild abnormalities in the corpus callosum (Lu et al., 2013) in Mid1-KO mice. In the latter study, elongated axons were observed in Mid1-KD/Mid1-KO cortical neurons (Lu et al., 2013). The cause of the discrepancy between the present study, which revealed impaired neuritogenesis following Mid1 KD, and the latter study remains unknown, although it may be related to differences in the specificity of cortical neurons and CGNs. Alternatively, the activation states of Rac and redundancy of Mid2 (Short et al., 2002) may vary according to the neuronal type/brain region, as well as developmental stage, inducing variable expressivity of OS phenotypes. In any case, phenotypes of Rac1/Rac3-DKO and OS are restricted to the midline. However, we observed that Mid1 and Rac are expressed not only in the medial structures but also the lateral structures/regions (Fig. 10B). Moreover, Mid1 expression and mTORC1 signaling in the Rac1/Rac3-DKO cerebellum was reduced in both the medial and lateral EGL. Such evidence strongly suggests that a specific molecule(s)/signaling(s) exists to regulate the function of Mid1 only around the medial structures of the body. In the present study, we were unable to identify additional factor(s) responsible for the restriction of lesions in OS. However, our findings indicate that Rac1 regulates levels of Mid1 expression and functions synergistically in a complex with Mid1 and mTORC1. A previous study using Nestin-Cre-driven Rac1-KO mice focused on impaired axon formation and neuronal migration in CGNs; however, the authors reported more modest medial cerebellar agenesis (II-V) than that observed in our Rac1/Rac3-DKO mice (I-VII), without referring to any mechanism (Tahirovic et al., 2010). The novel Rac-Mid1-mTORC1 signaling pathway proposed for midline structures in the present study may be further supported by the findings that telencephalon-specific Rac1-KO mice exhibit agenesis of the corpus callosum (Chen et al., 2007; Kassai et al., 2008).

Atoh1-Cre transgenic (TG) mice (Matei et al., 2005) and Rac1^flox/flox mice (Kassai et al., 2008) were backcrossed to generate Atoh1-Cre;Rac1^flox/flox mice. Atoh1-Cre;Rac1^flox/flox mice and Rac3^−/− mice (Corbetta et al., 2005) were backcrossed for more than five generations to obtain experimental animals (Rac^flox/flox;Rac3^−/−, Atoh1-Cre;Rac^flox/+;Rac3^−/−, Atoh1-Cre;Rac1^flox/flox;Rac3^−/+ and Atoh1-Cre;Rac1^flox/flox;Rac3^−/−). CAG-STOP^flox-dtTomato (Ai9) TG reporter mice were purchased from the Jackson Laboratory and backcrossed with Atoh1-Cre mice and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice. See the supplementary Materials and Methods for further details of genotyping.

RN46A (Sakai et al., 2003), PC12 (Sakai et al., 2003) and HEK293 (ATCC) cells were used in the present study (for details on cell maintenance, see the supplementary Materials and Methods). EHop-016, which inhibits Rac GEF interaction and has 100-fold higher specificity than the established Rac inhibitor NSC23766 (Montalvo-Ortiz et al., 2012), was obtained from Millipore. The details of plasmids and RNA interference are described in the supplementary Materials and Methods.

In situ hybridization was performed as previously described (Ueyama et al., 2014) using P6 and P21 wild-type mice and a pair of Rac1^flox/flox;Rac3^−/− and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− mice at P5. See the supplementary Materials and Methods for details of DNA probes.

Total RNA was extracted using an RNeasy Mini kit (Qiagen). Reverse transcription was performed on 1-5 µg of total RNA using SuperScript III reverse transcriptase (Invitrogen) and random primers. qPCR was performed using gene-specific primers (Mid1, Rac1, Rac3, Actb and Gapdh) as previously described (Ueyama et al., 2016). Details of primers used for qPCR and RT-PCR can be found in the supplementary Materials and Methods.

Immunoblotting was performed as previously described (Ueyama et al., 2016). For details of antibodies used in immunoblotting and histochemistry, see the supplementary Materials and Methods.

Sections were prepared as previously described (Ishii et al., 2017). Nissl staining was performed using Cresyl Violet solution (Muto Pure Chemicals). P8 cerebellar sections were used for detecting apoptosis via active caspase 3 immunostaining, followed by diaminobenzidine staining using a Vectastain ABC kit (Vector Laboratories) and Hematoxylin counterstaining. P5 sections obtained from Rac1^flox/flox;Rac3^−/−;tdTomato and Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato cerebellum were immunostained for Pax6, p27^Kip1 and NeuN to evaluate the differentiation of CGNs. P8 sections obtained from Atoh1-Cre;tdTomato and Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato cerebellum were used for immunofluorescence imaging of vertical neurites projecting from radially migrating GCNs. Coronal cerebellar sections obtained from 12-week-old Rac1^flox/flox;Rac3^−/−;tdTomato and Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato mice were immunostained for calbindin. Samples were imaged using a light microscope with a DP26 camera (Olympus) or a LSM700 confocal laser microscope (Carl Zeiss).

P4 mice (Atoh1-Cre;Rac1^flox/flox;Rac3^−/−;tdTomato and Rac1^flox/flox;Rac3^−/−;tdTomato) were intraperitoneally injected with 15 mg/kg 5-ethynyl-2′-deoxyuridine (EdU) dissolved in PBS. For the analysis of proliferation of CGNs in the EGL, cerebella were fixed 4 h after injection. For the analysis of radial migration of CGNs, cerebella were fixed 30 h after injection. In sagittal sections (8 µm), EdU detection was performed using a Click-iT EdU Alexa Fluor 488 Imaging kit (Invitrogen) and a confocal laser microscope. The EdU labeling index was defined as the ratio of EdU-Alexa488-labeled cells/DAPI-labeled cells.

Primary CGN cultures and cerebellar microexplant cultures (Nagata and Nakatsuji, 1990) were generated according to a previously described protocol.

For primary CGN cultures, the P8 cerebellum was dissected and dissociated with 0.25% trypsin. Washed and centrifuged cells mixed in DMEM supplemented with B27 (Invitrogen) were plated on poly-D-lysine (PDL)-coated Lab-TekII chamber slides (Nalge Nunc International). In the case of siRNA experiments, 1.0×10⁶ cells were electroporated, and plated on six-well plate dishes (Falcon). CGNs were incubated for the indicated period after addition of Ara-C (final 10 µM) at 24 h.

For cerebellar microexplant cultures, the cerebellum was dissected at P4-P6, and sagittal slices (150 μm) were subsequently prepared with a McIlwain tissue chopper (Mickle Laboratory Engineering). White matter and deep cerebellar nuclei were removed from the tissue slices. Rectangular pieces were then dissected from the remaining tissue, which mainly consisted of the EGL. The prepared rectangular microexplants were placed into DMEM supplemented with B27 on PDL-coated glass-bottomed dishes (MatTek) coated with laminin (Wako) and incubated for 3 days. The rectangular microexplants (from whole, medial and lateral cerebella) were also used for qPCR and immunoblotting analyses.

Total RNA from the medial region of the Rac1^flox/flox;Rac3^−/− and Atoh1-Cre;Rac1^flox/flox;Rac3^−/− cerebella at P6 was extracted using TRIzol (Invitrogen). The quality and quantity of RNA were determined using the Agilent 2100 BioAnalyzer. Gene expression profiles were examined using the SurePrint G3 Mouse Gene Expression 8×60K Microarray kit (Agilent Technologies). The background-subtracted signal intensity of Mid1 was normalized to that of the 75th percentile, Actg1 (γ-actin) or Tuba1a (α-tubulin). The ratios of normalized Mid1 signal (Atoh1-Cre;Rac1^flox/flox;Rac3^−/−/control) are presented.

Immunoprecipitation experiments to detect Rac1/MID1/mTOR complex were performed using HEK293 cells. For details, see the supplementary Materials and Methods.

All data are presented as the mean±s.e.m. Details of measurements are included in the supplementary Materials and Methods. Two groups were compared using unpaired Student's t-tests or Kolmogorov–Smirnov test. For comparisons of more than two groups, one-way or two-way analyses of variance (ANOVA) were performed, followed by Bonferroni's post-hoc test of pairwise group differences. Statistical analyses were performed using Prism 6.0 software (GraphPad).

We thank Prof. Bernd Fritzsch (University of Iowa, USA) for providing Atoh1-Cre TG mice. We also thank Ms Aya Shimizu, MSc and Mr Isao Sakamoto, MPharm for their technical assistance.