The RacGAP β-Chimaerin is essential for cerebellar granule cell migration

During mammalian cerebellar development, postnatal granule cell progenitors proliferate in the outer part of the External Granule Layer (EGL). Postmitotic granule progenitors migrate tangentially in the inner EGL before switching to migrate radially inward, past the Purkinje cell layer, to achieve their final position in the mature Granule Cell Layer (GCL). Here, we show that the RacGAP β-chimaerin is expressed by a small population of late-born, premigratory granule cells. β-chimaerin deficiency causes a subset of granule cells to become arrested in the EGL, where they differentiate and form ectopic neuronal clusters. These clusters of granule cells are able to recruit aberrantly projecting mossy fibers. Collectively, these data suggest a role for β-chimaerin as an intracellular mediator of Cerebellar Granule Cell radial migration.


Introduction
Proper morphogenesis of the vertebrate Central Nervous System (CNS) relies on the tight spatiotemporal control of cell proliferation, differentiation, migration and guidance events. In the mammalian cerebellum, Granule Cells (GCs) undergo a prolonged and highly stereotyped migration that begins embryonically and completes late postnatally 1 . In the mouse, beginning at embryonic day 12 (E12), granule cell precursors (GCPs) are born from the rhombic lip and migrate tangentially to cover the cerebellar anlage 2 , forming a secondary germinal zone, the External Granule Layer (EGL). Postnatally, GPCs in the EGL exit the cell cycle and travel inwards, splitting the EGL into an upper, mitotically active (outer EGL, oEGL) and a lower, migratory layer (inner EGL, iEGL) (Fig. 1a). These postmitotic GCPs grow two horizontal processes and migrate tangentially in all directions, before growing a third perpendicular leading process. Using this leading process GCPs migrate radially inward along Bergmann Glial fibers, past the Purkinje Cell (PC) Layer, to occupy their final location in the mature Granule Cell Layer (GCL) 3,4 . Cerebellar GC migration has been shown to be influenced by a wide set of guidance cues, including the chemokine SDF-1 5 , Slit2/Robos 6 , Plexins/Semaphorins 7-9 , brain-derived neurotrophic factor (BDNF) 10 , Vascular Endothelial Growth Factor (VEGF) 11 , and others. However, the cytosolic machinery responsible for effecting and directing the cellular response downstream of these ligand-receptor pairs remains largely unexplored.
The Rho family of small G-Proteins, or GTPases, plays essential roles in vertebrate CNS development, influencing a wide range of developmental processes, including cell migration, cell polarity, axon pathfinding, and dendritic remodeling through their ability to modulate cytoskeletal structure 12,13 . GTPases exists in two states: an active GTP-bound state and inactive GDP-bound state 14 . Precise subcellular regulation of GTPase activity is essential in maintaining proper cellular function, and neurons achieve this using positive regulators, Rho Guanine Nucleotide Exchange Factors (or RhoGEFs) and negative regulators, Rho GTPase Activating Proteins (or RhoGAPs) 14,15 . Disruption of RhoGTPase activity or their regulators' function has been associated with a broad array of behavioral and developmental disorders 15,16 . The chimaerin family of RhoGAPs consists of two genes: α-chimaerin (CHN1) and β-chimaerin (CHN2). They posses specific GAP activity toward Rac family GTPases, which are key modulators of actin filaments 17 . In neural development, α-chimaerin has been shown to play roles in Ephrin-mediated circuit formation 18-21 , cortical migration 22 , optic tract axon guidance 23,24 , and hippocampal dendritic arbor pruning 25 . The in vivo role of β-chimaerin in neural development was unexplored until recently, where it was shown to effect hippocampal dentate gyrus axon pruning by regulating Rac1 activity downstream of Sema3F/Neuropilin-2 signaling 26 .
Of note, β-chimaerin has been shown to be strongly expressed in GCs in the adult 27 , but its function during cerebellar morphogenesis is unknown. Here, we show a functional requirement for β-chimaerin during cerebellar development. We find that β-chimaerin is necessary for a small subset of granule cells to complete their migratory route from the EGL to the GCL.

β-chimaerin is specifically expressed in the Granule Cell Layer of the mouse cerebellum
β-chimaerin has been previously shown to be expressed in the adult cerebellum 27 . To explore the developmental expression profile of β-chimaerin in the cerebellum, we performed in situ hybridization in C57/BL6J mice to visualize β-chimaerin (Chn2) messenger RNA (mRNA) at several postnatal stages (Fig. 1b-h). We found Chn2 mRNA was strongly expressed in the GCL at all the postnatal ages tested. Interestingly, we observed Chn2 expression in small clusters of cells in the Molecular Layer (ML) of postnatal day 18 (P18) animals (Fig. 1f). This stage represents one of the last postnatal stages before the EGL dissolves. This ML expression did not persist into adulthood, disappearing by P35 (Fig. 1g).

β-chimaerin deficient mice display ectopic neuronal clusters on the cerebellar surface
As Chn2 transcript was found to be robustly expressed in the cerebellum at all postnatal stages examined, we asked whether β-chimaerin played a functional role during cerebellar development. We took advantage of a previously generated knock-in mouse that expresses betagalactosidase (βgal) from the endogenous Chn2 locus, rendering the Chn2 gene inactive 26 . We generated adult (P35) mice homozygous for a Chn2 null allele (Chn2 -/-) and compared their cerebellar structure to WT (Chn2 +/+ ) littermate controls (Fig. 2a-c). We observed no gross alterations to cerebellar lobule formation or cortical lamination in Chn2 -/mutants. However, we did see large ectopic clusters of cells aggregating in the ML of mutant animals (Fig. 2b, white arrows). These clusters strongly co-labeled with the pan-neuronal marker NeuN and an antibody raised against βgal, indicating that these clusters consist of ectopic cells that normally express Chn2 (Fig. 2 d,e). Sparse NeuN labeling was also seen in the ML of both WT and Chn2 -/-genotypes (Fig. 2d,e), and likely represents the stellate and basket cells known to occupy this region. We next asked if β-chimaerin function is required in a dose-dependent manner for normal cerebellar development. We quantified the number of neuronal ectopias in Chn2 +/+ , Chn2 +/-, Chn2 -/adult animals and found a highly significant increase in the number of ectopias in Chn2 -/mutants as compared to WT and Chn2 +/animals (p<0.01 for both comparisons) (Fig.   2c).
The adult cerebellum can be organizationally divided into four domains: Anterior, Central, Posterior, and Nodular. Each region, in turn, is physically divided into lobules, numbered I-X in mice 28 . Closely examining the P18 in situ hybridization data, we noticed that the majority of ML Chn2 transcript expression occurred in more posterior sections, particularly Lobules VII-IX and the fissures separating them (data not shown). Therefore, we asked if the NeuN-positive clusters we observed in Chn2 -/followed a similar pattern of distribution. Indeed, we found that NeuN-positive ectopias were more prevalent in the fissure separating lobules VII and VIII and on the posterior side of lobule IX (Fig. 2f for schematic and percent distribution).
These two locations collectively account for approximately 45% of all ectopic clusters scored (n=1068 ectopias across nine Chn2 -/animals). Collectively, these data suggest that Chn2 is expressed by a small subset of late radially migrating neurons prior to their arrival to the GCL, and that loss of β-chimaerin function causes these cells to fully arrest in the EGL.

The ectopic clusters contain mature granule cells, but not other types of cerebellar neurons
While the prior data suggest that the neuronal ectopias observed in Chn2 -/mutants contain Chn2 expressing cells, we sought to more thoroughly examine the composition of these ectopias. To test for the presence of mature GCs, we made use of the marker Gamma-Amino Butyric Acid Receptor subunit α6 (GABARα6) (Fig. 3a,b). Most cells in the neuronal ectopias in Chn2 -/animals colabeled with GABARα6, confirming the presence of mature, fully differentiated GCs (71 ± 5%, mean ± SD). Furthermore, we did not detect the immature GC marker Pax6 in the ectopias ( Supplementary Fig. S1a,b). To explore the possibility of other cell types contributing to the composition of these ectopic clusters, we immunolabeled with antibodies raised against the Purkinje Cell marker Calbindin, but did not find any Calbindin + cells within the clusters (Fig. 3c,d). Interestingly, Purkinje cell dendrites failed to invade the space occupied by the neuronal clusters (Fig. 3d). We also immunolabeled for the general GABAergic interneuron marker Parvalbumin (Fig. 3e,f) and found no co-labeling in the neuronal ectopias. Finally, we immunolabeled with the GABAergic marker Glutamic Acid Decarboxylase 67 (GAD67) (Fig. 3g,h). No GAD67 + cell bodies were detected in the ectopic clusters. We did detect evenly spread ML labeling of GAD67-positive processes, even in areas containing neuronal ectopias, suggesting these ectopias could potentially receive GABAergic input from stellate or basket cells (Fig. 3g,h). Collectively, these data suggest that the neuronal ectopias found in Chn2 -/animals are composed primarily of GCs, but not other cerebellar neuronal types.
To further study the granule cell migration defect in Chn2 -/animals, we performed pulse labeling of migrating cells with Bromo-deoxy-Uridine (BrdU), a thymidine analog that incorporates specifically into cells in the S-phase of mitosis. WT and Chn2 -/animals were injected with BrdU at P10 and cerebella were collected at adult stages. Whereas BrdU labeling in WT animals under these conditions is restricted to the GCL and a few cells scattered in the ML, Chn2 -/animals display considerable accumulation of BrdU + cells in the ectopias (Fig. 4a, b).
This arrest of GCs born after P10 in the ML of Chn2 -/mutants suggest that GC migration within the EGL or from the EGL requires Chn2.
During radial migration, GCPs in the iEGL migrate along Bergmann glial fibers to navigate toward the GCL 4 . Failure of GCPs to properly associate with glial tracts, or errors in glial scaffold architecture itself could inhibit GC radial migration, and could explain the ectopic phenotype observed in Chn2 -/mutants. Therefore, we examined the structure of the glial scaffolds surrounding ectopic clusters using an antibody raised against Glial Fibrillary Acidic Protein (GFAP) (Fig. 4c, d). We observed no gross alterations to Bergmann Glial structure, arguing against the possibility of an architectural cause underlying the phenotype. However, upon co-labeling with βgal, which strongly marks most cells in neuronal clusters ( Chn2 -/and Chn2 +/vs. Chn2 -/-). This observation reinforces the idea that GCPs lacking βchimaerin function stall during radial migration.

Granule cell ectopias recruit presynaptic partners
In the mature cerebellar circuit, granule cells in the GCL receive glutamaergic input from mossy fibers originating from the spinal cord, pontine nucleus, and other CNS regions. GCs in turn provide glutamatergic input via parallel fibers onto local purkinje cell dendrites 1 . Since the neuronal ectopias contain differentiated, GABARα6-positive GCs (Fig. 3b), we asked if they could form local circuits. We assayed for the expression of the synaptic marker vesicular glutamate transporter 2 (Vglut2), which labels a subset of cerebellar glutamatergic synapses formed by climbing fibers and mossy fibers, and found robust colabeling with βgal-positive cells within neuronal ectopias (Fig. 5a,b). Furthermore, Vglut2 staining in the ectopias displayed a pattern highly reminiscent of the rosette structures formed by mossy fiber terminals with full penetrance and expressivity (100% of the ectopias displayed this pattern).
To test whether the Vglut2-positive staining on the ectopic neuronal clusters indeed represented mossy fiber synaptic terminals, we performed stereotactic injections of an Adenosine precursors become postmitotic and extend two horizontal processes, moving inward to generate the inner EGL (iEGL) as a distinct population from the more superficial precursors that remain mitotically active in the outer EGL. In the inner EGL these postmitotic precursors will migrate tangentially, eventually arresting and growing a third perpendicular process. They then begin migrating radially inward, past the PC layer, to form the mature GCL. Given the complex migratory path GCPs take in their development, we asked if an earlier, subtler defect in EGL structure may precede the development of neuronal ectopias.
We examined P10 Chn2 -/and control animals for the overall distribution of GCPs. We first immunolabeled with antibodies against the transcription factor Pax6, which is active in GCPs in the EGL and maturing GCs in the GCL. We noticed no major difference in Pax6 distribution between Chn2 -/mutants and controls in the lobules that frequently develop ectopias ( Fig. 6 a, b). We also examined the expression profile of the cell adhesion molecule L1-NCAM (L1), which labels migrating granule cells in the inner EGL 8 . We found no major difference in its distribution between Chn2 -/mutants and controls (Fig. 6 c, d). These results suggest that there is no altered distribution of GCPs preceding the development of neuronal ectopias. As stated earlier, one possible explanation of ectopia formation is alterations to Bergmann glial tracts. We analyzed the structure of the Bergmann glial scaffold using an antibody against GFAP and found no structural differences in the lobules that more frequently develop neuronal ectopias (Fig. 6 ef). Collectively, these results suggest that there are no major early postnatal lamination or architectural defects that could predispose certain GCs to arrest.
The data presented in Fig. 4 suggests that in Chn2 -/mutants GCs are arrested during radial migration. However, defects in tangential migration and/or proliferation could be indirectly contributing to this phenotype. Initial tangential migration of GC progenitors from the rhombic lip appears to be normal in Chn2 -/animals, as the length of the EGL in WT and Chn2 -/animals is not significantly different at P0 (Fig. 7a-c; two-tail t-test, p=0.082, n=5) 29 . A second phase of tangential migration occurs after precursors become postmitotic and move inward to generate the iEGL. It would be predicted that, if GCs are arresting during tangential migration or during the tangential to radial migration switch, the iEGL would become thicker in the folia that develop ectopias in Chn2 -/mice compared to control animals. To label tangentially migrating GCs in the inner EGL we performed immunostaining with anti-Sema6a antibody at P10 (Fig.   7d,e) 8,9 . We then measured the thickness of the iEGL relative to the EGL overall in caudal folia ( Figure 7f). The iEGL/EGL ratio was comparable in WT and Chn2 -/mice (Figure 7f; two-tail ttest, p=0.528, n=10), suggesting that tangential migration is not notably disrupted in Chn2 -/mice.
During mammalian cerebellar development, granule cell precursors normally continue to proliferate postnatally in the oEGL 30 Fig. 8g,h).
No proliferating cells were found in the neuronal ectopias, suggesting these ectopic neuronal clusters consist entirely of post-mitotic cells (Fig. 8g,h). Overall these data suggest that the formation of ectopias and the arrest of GCs in the molecular layer of Chn2 -/animals are primarily due to a defect in radial migration.

Cerebellar Structure in mice expressing hyperactive β-chimaerin
Genetic ablation of Rac1 and Rac3 results in severe disruption of cerebellar granule cell migration 29,31 . Could increasing β-chimaerin RacGAP activity cause similar phenotypes? To test whether enhanced β-chimaerin activity could also affect cerebellar development, we made use of a knock-in mouse that harbors a hyperactive Chn2 allele. This allele consists of a single amino acid substitution introduced into the endogenous gene locus 26 . The I130A substitution yields a protein with a more "open" conformation, which renders it more sensitive to induction 32 . We collected adult (P35) mice that were homozygous for the hyperactive allele (Chn2 I130A/I130A ) and stained for the mature granule cell marker GABARα6 (Fig. 9a, b) and glutamatergic synapse marker Vglut2 (Fig. 9c,d) to label fully differentiated GCs and glutamatergic synapses, respectively. In contrast to Chn2 -/mutants, Chn2 I130A/I130A animals did not develop ectopic clusters of cells. Further, GC lamination appeared no different from controls. We next looked for other errors in cerebellar structure or lamination by immunostaining for the markers GFAP ( Fig.   9 e, f), Parvalbumin (Fig. 9g, h), and GAD67 (Fig. 9i, j). We found no difference in the Bergmann Glial scaffold or GABAergic cell populations, respectively. Collectively, these data suggest that hyperactivity of β-chimaerin does not negatively affect cerebellar morphogenesis.

Discussion
Here we show that the RacGAP β Chimaerin is essential for cerebellar GC development.
Many ligand-receptor pairs have been shown to regulate GC proliferation and migration, but less is known about the cytoplasmic effectors that link these extracellular signals with the cytoskeleton 5-9 . Guided by the previously reported robust expression of Chn2 in the adult GCL 27 , we examined whether this cytoplasmic protein could be playing a functional role during cerebellar development. We found that the genetic ablation of Chn2 results in the formation of ectopic clusters of neurons in the outer ML. These ectopias are primarily formed by GCs. Since we initially established that Chn2 was mainly expressed in the GCL of early postnatal and adult cerebella (Fig. 1), which represents the mature post-migratory GC population, how could the RacGAP known to negatively regulate Rac1-GTP levels in neurons 26,39 , is probably moving the scale in the opposite direction. Thus, balanced Rac activity might be essential for proper GC migration. In this regard, expression of a hyperactive version of β Chimaerin (I130A) from the endogenous Chn2 locus was not enough to disrupt GC migration (Fig. 9). This could be in part due to the regionally and temporally restricted expression of Chn2 in premigratory GCs.
This novel role of Chn2 during cerebellar development is the newest addition to a growing list of functional requirements for these RacGAPs during neural development: chimaerins have been shown to regulate axon guidance, pruning in the hippocampus, and cortical Previous studies have demonstrated that Chimaerin function can be modulated by Class 3 semaphorins (Semas) during axon guidance and pruning 23,26 . In particular, β-chimaerin's Rac activity was shown to be regulated by Sema3F/Neuropilin 2 function during hippocampal pruning 26 . Even though cerebellar circuitry proceeds normally in Sema3F -/animals 40 , other Semaphorins and Sema receptors have well-established roles during cerebellar GC migration and proliferation 7-9,41,42 . It is plausible that some of these other members of the Sema family could modulate β-chimaerin function in GCs, since it is still mostly unclear how plexins regulate the actin cytoskeleton 8,9 .
As mentioned above, only a subset of granule cells are susceptible to an arrest in migration in Chn2 -/cerebella, while the GC population at large is phenotypically normal. Are these ectopic cells able to recruit the right presynaptic partners in a sea of normally positioned GCs? The surprising answer to this question appears to be yes. Anterograde labeling of the pons using viral approaches revealed that the ectopic clusters found in Chn2 -/cerebella were innervated by pontine axon fibers, one of the normal presynaptic partners for cerebellar GCs (Fig.   5). These ectopic presynaptic terminals are Vglut2 + and display the rosette morphology characteristic of normal pontine mossy fibers. Whether these synaptic terminals are active and mature remains to be explored. While the data presented here provides developmental insight into cerebellar circuit assembly at the anatomical level, it is unlikely that the small number of ectopias present in Chn2 -/animals will result in obvert behavioral or physiological changes. Far more severe histological defects are observed in other mutant animals without any measurable behavioral or motor changes 7,8 . Exploration of subtle changes in behavior and physiology in Chn2 -/animals will be the subject of further studies.

Animals and Genotyping
The day of birth in this study is designated as postnatal (P) day 0. The generation of Chn2 -/and Chn2 I130A/+ mice has been described elsewhere 26

Immunohistochemistry
Mice were perfused and fixed with 4% paraformaldehyde for 2 hours at 4°C, rinsed and sectioned on a vibratome (150 µm). Immunohistochemistry of floating parasagittal cerebellar sections was carried out essentially as described 43  . Confocal fluorescence images were taken using a Leica SPE II microscope. Area and length were measured using ImageJ. For cell counts, the ImageJ cell counter plugin was used 44 .

In situ Hybridization
In situ hybridization was performed on floating cerebellar vibratome sections (150 μ m thickness) using digoxigenin-labeled cRNA probes, essentially as described for whole-mount RNA in situ hybridization 45 . Generation of the Chn2 cRNA probes has been described in 26

Injections of AAV
Synapsin-EGFP AAV8 was obtained from the University of North Carolina viral core.
The concentrated viral solution (0.2 μ l), was delivered into the pons by stereotactic injection (0.25 μ l per min), using the following coordinates: anterior-posterior, -5.1 mm; lateral, ±0.6 mm; and vertical, -4.1 mm. For all injections, Bregma was the reference point.

BrdU labeling
BrdU labeling agent was purchased from Life Technologies (#000103) and was delivered via intraperitoneal injection at 1ml BrdU solution/100g animal weight, following manufacturer instructions. Brains were perfused and collected 2hrs post injection for proliferation assessment, or as adults for pulse-chase experiments. Perfused brains were fixed for 2 hours and sectioned on a vibratome (150 μ m thickness). Sections underwent antigen retrieval: incubated in 1M HCl in 1xPBS for 30 mins at room temperature, washed 3x10 min in 1xPBS, incubated in 10mM sodium citrate for 30min at 80C, and washed 3x10min in 1xPBS. Following antigen retrieval, immunohistochemistry was performed as described above using a mouse monoclonal antibody anti-BrdU (Invitrogen, clone BU-1, MA3-071 at 1:250).   Neuronal ectopias do not co-label with any of these three markers and therefore do not contain PCs, stellate, or basket cells that normally occupy the ML. Interestingly, Purkinje cell dendrites appear to avoid invading the clusters. Scale bar, 50μm for all.