Ectopic Cerebellar Cell Migration Causes Maldevelopment of Purkinje Cells and Abnormal Motor Behaviour in Cxcr4 Null Mice

SDF-1/CXCR4 signalling plays an important role in neuronal cell migration and brain development. However, the impact of CXCR4 deficiency in the postnatal mouse brain is still poorly understood. Here, we demonstrate the importance of CXCR4 on cerebellar development and motor behaviour by conditional inactivation of Cxcr4 in the central nervous system. We found CXCR4 plays a key role in cerebellar development. Its loss leads to defects in Purkinje cell dentritogenesis and axonal projection in vivo but not in cell culture. Transcriptome analysis revealed the most significantly affected pathways in the Cxcr4 deficient developing cerebellum are involved in extra cellular matrix receptor interactions and focal adhesion. Consistent with functional impairment of the cerebellum, Cxcr4 knockout mice have poor coordination and balance performance in skilled motor tests. Together, these results suggest ectopic the migration of granule cells impairs development of Purkinje cells, causes gross cerebellar anatomical disruption and leads to behavioural motor defects in Cxcr4 null mice.


Introduction
CXC chemokine receptor 4 (CXCR4) is a seven-transmembrane G-protein-coupled receptor. It acts as a receptor for CXC chemokine stromal cell derived factor-1 (SDF-1, also called CXCL12). It is widely expressed in a variety of tissue types but is predominantly expressed by immune cells and in the brain. While the immune function of CXCR4 has been much studied, little is known about its role in the brain.
During embryonic mouse brain development, Cxcr4 is expressed in ventricular zones. These are sites of stem cell proliferation. In late embryonic stages, Cxcr4 is expressed in the hippocampus and cerebellum [1]. Embryonic data (E18.5 and P0) from Cxcr4 knockout (KO) mice show that the cerebellum develops abnormally with an irregular external granule cell layer (EGL) and ectopically located Purkinje cells [2,3]. These studies imply that defects in SDF-1/CXCR4 signaling result in premature migration from the EGL during embryonic cerebellar development. Indeed, SDF-1 has been shown to function as a chemoattractant and is secreted from the meninges. It attracts embryonic but not postnatal cerebellar EGL cells [4]. In SDF-1 KO mice at E15.5, premature granule cells have been detected migrating into the cerebellar anlage [5].
Cxcr4 is highly expressed from E18.5 to P4 in the cerebellum. Subsequently, expression becomes very low or non-detectable at P14 (according to the Allen Brain Atlas [6]). Currently, the effect of CXCR4 deficiency in postnatal cerebellar development is poorly understood. This is because Cxcr4 KO mice are embryonic lethal as a result of defects in cardiogenesis and hematopoiesis [3]. To date there has been no study into postnatal cerebellar development in CXCR4 KOs since the work of Zou et al. in 1998. Consequently, in order to study postnatal development and its impact on function we conditionally inactivated Cxcr4 in the central nervous system (CNS). We here report the functional characterization of conditional inactivation of Cxcr4 in postnatal cerebellar development.

Ethics Statement
All experiments were carried out in strict accordance with the recommendations in the Guide for Laboratory Animals Facilities and Care as promulgated by the Council of Agriculture. Executive Yuan, ROC. The protocol was approved by the Institional Animal Care and Use Committee of Chang Gung University (Permit Number: CGU11-007). In this protocol, all efforts were made to minimize suffering.

Immunohistochemistry and in situ hybridization
Tissue was fixed in 4% paraformaldehyde. All sections for immunohistochemistry and in situ hybridzation were cut to a thickness of 40 mm on a sliding microtome. For antibody staining, sections were mounted on superfrost electrostatic slides and dried overnight. Subsequently, slides were incubated in the 0.01 mol/L citric buffer for 15 min at 90uC, 3% H 2 O 2 for 10 min, rinsed in PBS, and incubated overnight at room temperature. BrdU (Accurate, 1:250), NeuroD (Santa Cruz, 1:1000), Calbindin (Sigma, 1:1000), Cleaved Caspase-3 (Cell Signaling, 1:150) antibodies were used. Next day, following the ABC kit procedure (Vector Lab), slides were reacted with a Sigma DAB tablet. Sections were then cover-slipped with DPX. For immunofluorescence staining, sections were mounted on slides and dried overnight. On the following day, slides were incubated in the 0.01 mol/L citric buffer for 15 min at 90uC, rinsed in PBS, and incubated overnight in primary antibody solution. Pax6 (Millipore 1:1000), GFAP (Sigma, 1:2000), Calbindin (Sigma, 1:1000), CNPase (Abcam 1:1000), and NeuN (Millipore, 1:400) antibody concentrations were used. Subsequently, slides were rinsed in PBS and incubated with Alex Fluor 568 and Alex Fluor 488 secondary antibody (1:200) for one hour. Slides were then cover-slipped. For in situ hybridization, CXCR4 DIG-labeled probe was synthesized using a forward primer (ATGGAACCGATCAGTGTGAGTA) and reverse primer (ATGCTCTCGAAGTCACATCCTT). Slides were incubated in Proteinase K (Invitrogen, 20 ug/ml) for 10 min, acetylated by acetic anhydride, incubated with Cxcr4 probe in 50% formamide at 58uC overnight. Next day, washing with SSC and MABS, then slides were incubated with anti-DIG antibody (Roche, 1:5000) overnight. Signal was detected by NBT and BCIP for 12 hours.

Cell counting for Granule and Purkinje cells
Brain tissue for granule and Purkinje cells counting was cut to a thickness of 20 mm on a sliding microtome. One in eight representative sections was then stained for Pax6 or Calbindin using antibodies. Four sagittal sections close to the middle of the cerebellum were selected for cell counting. For Purkinje cells counting, the total number of cells present on the four sections was counted. For granule cells counting, we selected 24 areas (1006200 mm 2 ) inside the granule cell layer and counted the number of granule cell in each square using Image J.

Golgi staining
Brains were harvested and stained using an FD rapid Golgistain TM Kit in accordance with the manufacturer's instructions (FD Neuro Technologies). Brains were sectioned sagittally at a thickness of 120 mm using a microtome. The numbers of terminal dendrites of Purkinje cells were counted. A z-series of images were taken of purkinje cells and a focused image produced via photo-merge stacking in Adobe Photoshop CS5. mRNA quantitification P1 cerebella were collect from WT and KO mice. mRNA was extracted using a RNeasy Lipid Tissue Kit (Qiagen). cDNA was then made using SuperScript III reverse transcriptase (Invitrogen). Expression was measured by quantitative PCR using the following primers: Cxcr4 (F: GACTGGCATAGTCGGCAATG, R: AGAAGGGGAGTGTGATGACA AA) Neurog2 (F: AACTC-CACGTCCCCATACAG, R: TGCCAGTAGTCCACGTCT-GA) Ptf1a (F: CTGCGACAAGCCGCTAATG, R: GAAGGC-GTCGTTGATGGACT). b-actin (F: CATCACTATTGGCA-ACGAG, R: GGCATAGAGGTCTTTACGG). Experiments were performed in duplicate. Gene expression levels were calculated by the DDC t method and normalized against a b-actin control.

Cerebellum culture and staining
The cerebellum cell culture method was modified from a previous study [9]. Cerebella were dissected from WT or KO E17 embryos. They were then immersed in 12 ml of ice-cold HBSS in a 15 ml falcon tube. The medium was then discarded and the cerebella placed in 1 ml of HBSS supplemented trypsin (0.1%) and heated for 10 min at 37uC in the water bath. 1 ml of ice-cold Normal horse serum and 50 ml of 1% DNase was then added. The media was pipetted gently to dissociate cells. 1610 6 cells per well in 90 ml were plated onto glass cover slips (12 mm in diameter) coated with poly-D-lysine and placed in a humidified CO 2 incubator (5% CO 2 at 37uC). One milliliter of culture medium was added to each well after 3 hours. The medium was composed of DMEM-nutrient mixture of Ham's F-12 supplemented with insulin (7.5 mg/ml), transferrin (75 mg/ml), progesterone (15 nM), sodium selenite (22.5 nM), GlutaMAX (3 mM), triodothyronine (0.4 ng/ml), gentamicin (7.5 mg/ml), and putrescine (75 mM). For the CXCR4 antagonist experiment, AMD3100 (a gift from Dr. Kak-Shan Shia, NHRI, Taiwan) was added to the culture wells accordingly (0, 2.5 mg/ml, 10 mg/ml). The cerebellar cultures at 21 DIV were fixed with 4% paraformaldehyde in PBS for 15 minutes at room temperature. The cultures were incubated with calbindin antibody (Sigma, 1:1000) and beta-tubulin (abcam, 1:1000) in 0.5% Triton X-100 and 5% serum in PBS for 2 hours at room temperature. They were then incubated with fluoresceinlabeled secondary antibodies (1:200) for 1 hour at room temperature. The numbers of terminal dendrites of Purkinje cells were counted.

Microarray analysis
Gene expression analysis was performed with multiple biological replicates for WT (n = 8) and Cxcr4 KO (n = 7). RNA was prepared from cerebella from one-day-old WT and Cxcr4 KO mice. RNA was reverse transcribed into cRNA and biotin-UTP labeled using the Illumina TotalPrep RNA amplification kit (Ambion) and hybridized to the Illumina mouse Rsfseq-8v2 Expression BeadChips. Image data were extracted using the Illumina GenomeStudio software. Raw expression values were log2 transformed and median normalized. The expression array data is available on Gene Expression Omnibus (GSE48788).

Pathway enrichment analysis
For pathway enrichment analysis, the normalized data of entire 17925 transcripts were uploaded to the GSEA software (v 2.0, http://www.broad.mit.edu/gsea/). Signal-to-noise ratio was used to generate the ranked list for all genes. The association between a given gene set and a treatment group was measured by the nonparametric running sum statistic termed the enrichment score (ES).The gene sets used are from Molecular Signatures Database (MsigDB), catalog C2 functional sets, subcatalog canonical pathways, which include 241 gene sets from pathway databases (version 2.5, updated by April, 2008). These gene sets are canonical representations of 241 biological process compiled by KEGG. The cut-off p-values were assigned based upon 1000 random permutation tests. The normalized enrichment score (NES) was calculated based on the size of the gene set and its enrichment score. The nominal p-value was calculated after permutation testing of the microarray samples. The false discovery rate (FDR) was calculated to correct for multiple hypothesis testing. As recommended by the software developer, a gene set is considered significantly enriched when its p-value is less than 0.05 or FDR score is less than 0.25.

Heatmap and hierarchical clustering
All genes in the leading edge and the trailing edge of the eight significantly altered pathways listed in Table S1 were selected for hierarchical clustering analysis. A total of 69 genes were included in the analysis. Two-dimensional hierarchical clustering was performed by Ward's linkage method using Pearson's correlation as a measure of similarity. Intensity on the heatmap represents standardized expression level of individual gene in each sample. Hierarchical clustering analysis and intensity standardization were performed using Partek Genomics Suite (v 6.4).

Behavioral tests
Activity box. Mice were housed separately in transparent plastic cages for two hours. Two laser beams were fired across opposite ends of the cage. If a mouse positioned itself between the beam emitter and sensor, a ''beam break'' would be automatically counted by a computer. If a mouse broke both beams in succession, thereby transversing the length of the cage, a ''crossover'' was counted. The counts for beam breaks and crossovers were measured over 2465 minute intervals.
Inverted mesh grid grip test. Mice were placed on the centre of a 43 cm square wire mesh grid consisting of 12 mm squares of 1 mm diameter wire. The grid was surrounded by a 4 cm thick wooden frame which prevented mice climbing to the converse side. Once a mouse was positioned, the grid was turned upside down and elevated in order to force the mouse to grip the wire to avoid falling. The time taken for a mouse to fall was recorded. If an animal did not fall, the experiment was concluded after one minute.
Grip strength test. We followed the manufacturer's instruction for the Grip Strength Meter (UGO Basile). Mice were placed over a base plate, in front of a grasping bar. The bar was fitted to a force transducer connected to the peak amplifier. Mice were pulled by the tail, and the mice grasp the bar. Within 20 seconds, maximal grip force was measured.
Accelerating rotarod. Mice were held by the tail and placed on the rotarod, facing away from the direction of rotation. The rotarod moved at an initial speed of 4 rpm. After 10 seconds, the rod speed was accelerated at a rate of 20 rpm per minute. Once acceleration had been triggered, the time taken for mice to fall was noted.
Gait analysis. When mice spontaneously walked at a velocity of 20 cm/sec their motion was recorded by BCamCapture Version (Clever Sys Inc). TreadScan (Clever Sys Inc) was then used to analysis the standing and swing phases during gait cycles. The stand time is the time elapsed while the foot is in contact with the tread, in its stance phase. The swing time is the time elapsed while the foot is in the air, in its swing phase.
Ladder Rung Walking Test. Mice were placed on a Ladder Rung walking test apparatus. Mice were trained to walk through an irregular rung arrangement. Mice were given three trials in total, and the number of missing steps counted. The ladder rung task apparatus was modified from the published version for mice [10].

Behavioral statistics
The activity box data was analyzed via a paired T test, with KO results analyzed against their WT counterparts for a given time bin. The inverted mesh test was analyzed via a survival analyzed test. T tests were used to analysis grip strength, rotarod results, gait analysis and ladder rung walking. Values represent means 6 SEM. All stats were performed on Graphpad Prism version 5.

CXCR4 inactivation causes granule cell ectopia and agenesis of the cerebellar folia
Mice lacking CXCR4 in the CNS were generated by crossing mice harboring loxP sites flanking exon 2 of the Cxcr4 gene (Cxcr4 flox/flox ) [7] with Sox1-Cre mice [8] (Fig. 1A). Sox1 expression defines the neuronal precursors of the embryonic central nervous system [11,12]. Sox1-Cre mice express Cre throughout the neural tube at E9.5 [8]. To check the specificity of this Sox1-Cre expression, we crossed Sox1-Cre mice with Rosa26-EGFP reporter mice. We showed that the Cre recombinase is expressed in the neural tube at E11 (Fig. 1B). In our study, we used Cxcr4 flox/flox ; Sox1-Cre mice as KO mutants and the littermates without Cre were used as wild-type (WT) controls. KO mice were viable and fertile. Prior to weaning, approximately half of KO pups (Cxcr4 flox/flox ; Sox1-Cre) were observed to feature a limping gait phenotype with dragging of their hind limbs (Fig. 1C). From our observations, this limb dragging phenotype became less severe after weaning age. Mice were sacrificed at P9, and KO brains were found to be smaller in size and significantly reduced in weight compared with WT littermates (p = 0.0064, n = 4) (Fig. 1D). On macroscopic examination, KO mice were also found to feature a smaller cerebellum which lacked the well-defined cerebellar folia observable in the WT mice (Fig. 1E,F).
We further characterized the abnormalities of the cerebellar folia by injecting BrdU (50 mg/kg, i.p.) into P1 animals. We sacrificed them two hours later to label the proliferating cells. The BrdU imaging showed a subset of proliferating granule cells of the EGL invading the cerebellar anlage (Fig. 1G,H). Cxcr4 expression in the cerebellum is greatest between P3 to P5 [13]. We used in situ hybridization to show Cxcr4 expression was very low in our P4 KO mice whilst abundant in WT, particularly in the Purkinje cells layer and EGL (Fig. 1I,J). The expression pattern of NeuroD, a transcription factor required for differentiation of granule cells [14], showed that WT littermates featured a well-organized granule cell layer (GCL). In contrast, differentiating granule cells were ectopically located in the KO mice. Furthermore, no folia were formed in the KO cerebellum (Fig. 1K,L). These images suggest that knocking out Cxcr4 not only affects the EGL, but also the GCL and Purkinje cell layers during cerebellar development. To investigate whether there was increased apoptosis in the irregularly distorted cerebellum, we stained for cleaved caspase-3. We found no evidence of an increase in the number of cleaved caspase-3 labeled cells in the P4 KO (Fig. 1M,N).

CXCR4 effects Purkinje cell dendritogenesis in vivo, but not in vitro
We explored the development of Purkinje cells in the Cxcr4 KO mice. The Purkinje cell marker, Calbindin, showed that Purkinje cells were ectopically located in P9 KO mice. The Purkinje cell layer was much less organized compared with WT mice (Fig. 2A,B). The greater magnification of the Calbindin images (Fig. 2C,D) revealed that whilst dendritic arborization was less expansive and axons were disorganized in KO mice these axons do project to deep cerebellar nuclei. We also observed that dendritic elaboration is reduced in Purkinje cells which have inappropriately migrated inwardly, away from the molecular layer. This suggests the ectopic migration affects dendritic development.
Thus, we sought to establish whether disrupting SDF-1/ CXCR4 signaling impaired dendritogenesis in Purkinje cells directly or if the Purkinje cell defects were a consequence of ectopic migration. To answer this question, we isolated WT and KO cerebella from E17 embryos and co-cultured granule cells and Purkinje cells together. The cultured cells were in the same medium for three weeks to allow the Purkinje cells to receive factors from the granule cells and fully develop. Three weeks later, we analyzed the dendritic arboration of the Purkinje cells from KO and WT by using a Purkinje cell marker, Calbindin, and a neuronal marker, beta-tubulin class III. The results showed that there is no difference in dendritic complexity upon quantifying the number of terminal branches (p = 0.19, n = 12) (Fig. 2E,F). We also did not observe any obvious difference in axon development between KO and WT Purkinje cells. To confirm this result, we repeated this experiment by adding AMD3100 (0, 2.5 mg/ml, 10 mg/ml), a CXCR4 antagonist, into the cultured cells. There was no difference in dendritic complexity between the three groups (p = 0.85, n = 12). Both in vitro results indicated that the Purkinje cell defect detected in Cxcr4 KO mice were not a direct effect of CXCR4 loss. Consequently, our findings strongly suggest that the abnormal migration and alien extracellular environment underlie the mal-development of the KO Purkinje cells in vivo.

Radial glia cells, granule cell migration, and Purkinje cell projections
We used Calbindin staining to analyse the projections of Purkinje cells. Surprisingly, we found that Purkinje cell axons end correctly in the deep cerebellar nuclei in KO mice (Fig. 2G,H). We also stained for the oligodendrocyte marker, CNPase. We found that whilst axons in both WT and KO mice successfully projected to deep cerebellar nuclei, the projections in KO mice were more sporadic and untidy (Fig. 2I,J). To determine whether some of these stained myelinated fibers were Purkinje cells axons, we  a double labelling experiment for Calbindin and CNPase (Fig. S1A,B). We found a substantial amount of Calbindin staining colocalized with CNPase in both WT and KO cerebellum (Fig. S1C,D). These results suggest that the ectopically located Purkinje cells in KO mice still send myelinated axons to their targets in the deep cerebellar nuclei. Next, we studied granule cell migration using GFAP-labeled radial glia cells. In WT mice, Pax6 positive granule cells migrate along the radial glia fibers which extend to the apical surface. In KO mice, the radial glia scaffolds were not well aligned resulting in disrupted granule cell migration (Fig. 2K,L). Meanwhile, it has been demonstrated that the decrease in the number of granule cells may affect the location of Purkinje cells and the development of their dendritic trees [15]. To better understand whether there are difference in the number of granule cells and Purkinje cells in this CXCR4 KO model, we used thinner section (20 mm) for cell counting. The result show there is a significant lower density of granule cells in the KO cerebella (p = 0.001, n = 6) (Fig. 3A,B,E). However there is no significant difference in the number of Purkinje cells between KO and WT (p = 0.35, n = 6) (Fig. 3C,D,F).

Granule cells, Purkinje cells, and axon projection in adult KO mice
To establish if this defect was corrected during later development, and to explore the cell fate of granule and Purkinje cells, we examined the cerebellum histology in adult mice (three months old). We observed that ectopic granule cell clusters remain and the cerebellar lobules are still absent in adult KO mice (Fig. 4A,B). The Calbindin staining images revealed that the abnormally aligned Purkinje cells in KO mice remain alive but featureless dendritic elaboration than WT (Fig. 4C,D). To better understanding the development of adult Purkinje cells, we use Golgi staining to investigate the dendrites (Fig. S2A,B). KO Purkinje cells featured less complex dendritic arborization compared with WT (p,0.001, n = 12). Next, we investigated the axon projection by using CNPase staining. Our findings are similar to what we observed in the developmental stage: axons project to the deep cerebellar nuclei but fibers are disorganized in KO mice (Fig. 4E,F).

Gene expression profiling of Cxcr4 null cerebellum
To further understand the molecular changes associated with the cerebellar phenotypes, we conducted a microarray study to compare the transcriptome profile between WT and Cxcr4 KO cerebella from P1 mice (n = 8 for WT and 7 for KO). Cxcr4 KO animals showed a significant decrease in Neurog2 (foldchange = 21.65, p,0.001) and Ptf1a (fold change = 21.86, p,0.001) gene expression levels. Recent evidence suggests that Neurog2, a direct downstream target of Ptf1a, regulates Purkinje cell dendritogenesis [16,17]. Cerebellar GABAergic neurons are generated from Ptf1a expressing neuroepithelial cells. Ptf1a is essential for cerebellar development in mice and humans  [18,19,20]. Therefore, to confirm our array result, we used this P1 cerebella tissue to measure the relative mRNA levels via quantitative PCR. We showed a large reduction of Cxcr4 expression in KO mice (p,0.001, n = 11). There were also significant reductions in both Ptf1a and Neurog2 expression (p,0.001, n = 11). This result implies alterations to Ptf1a and Neurog2 pathways may affect the dendritogenesis of Purkinje cells in Cxcr4 deficient mice.
In addition, the Cxcr4 KO also displayed a significant decrease in Cxcl12, suggesting a positive regulatory loop between CXCR4 and its physiological ligand. We also performed a gene set enrichment analysis to identify biological pathways that were significantly altered (nominal p,0.05) (Table S1). Several biological pathways related to chemokine receptor interaction, cell mobility, and axonal guidance were found to be significantly down-regulated in Cxcr4 KO animals (Fig. 5A). The two most significantly affected pathways were the ''ECM receptor interaction'' and ''focal adhesion'' (Fig. 5B). A closer examination of differentially expressed genes in these pathways revealed that the Cxcr4 KO is associated with an extensive decrease in genes involved in cell-cell adhesion (Fig. 5C). Together, these results indicated that a widespread inhibition of genes involved in chemotaxis, cell migration and axon guidance may underlie the phenotype observed in Cxcr4 KO animals.

Knocking out CXCR4 compromises the motor and coordination systems
Finally, to assess the effect on walking and motor behavior, we conducted a series of behavioral experiments. First, we assessed locomotor activity (Fig. 6A). In a two hour long test, KO mice showed higher activity than WT (p,0.001, n = 9). Next, to test muscle strength and grip, we performed an inverted mesh grid grip test. During the 60 second test, more than half of mutant mice fell (p = 0.0053, n = 9), while all the WT mice still remained on the inverted grid (Fig. 6B left). All of the mutant mice that fell had failed to use their hind limb to grasp. We investigated whether the tendency to fall was related to grip strength, lack of coordination, or both. We tested forelimb grip strength using a grip strength meter. KO mice had a lower forelimb grip strength (p,0.0033, n = 6) (Fig. 6B right), suggesting that they fell could be due to inability to grip. Cxcr4 KO mice feature ectopic collateral branching from the corticospinal tract and this may explain this grip strength result [21]. To assess balance and coordination, we performed a rotarod test. The mutant mice fell sooner (p = 0.003, n = 9) and at a lower speed (p = 0.0023, n = 9), suggesting they also have poor balance and coordination (Fig. 6C). We then assessed normal walking and skilled walking gait of the animals. For normal walking, we utilized a gait analysis treadmill to examine the stance time and swing time. Surprisingly, there was no significant difference for stance time (n = 6) (Fig. 6D left). Skilled walking was tested using a ladder rung walking task in which the spacing of the rungs was uneven. In this task, the mutant mice made more missteps (p,0.001, n = 6) (Fig. 6D right). Overall, our results clearly demonstrate that motor and coordination systems, have been compromised in the Cxcr4 KO mice.

Discussion
The effect of CXCR4 deficiency on fetal cerebellar development has been known for more than a decade. However, study of postnatal cerebellar development has previously not been investigated. Thanks to Cre-Lox system, we were able to knock out the CXCR4 gene conditionally and bypass embryonic lethality. To our knowledge, by using a conditional KO model to bypass embryonic lethality, ours is the first study to explore the importance of Cxcr4 in postnatal cerebellar development.
Our postnatal results are consistent with previous Cxcr4 KO studies which demonstrated the irregular EGL and numerous granule cell clumps within the cerebellar anlage in the fetal cerebellum [2,3]. This suggests CXCR4 is required for proliferating granule cells to detach in the EGL. In addition, we observed ectopic and abnormal development of Purkinje cells in the KO mice. Ectopic Purkinje cells had far less dendritic arborization and featured disorganized and sporadic axon projections. To establish if SDF-1/CXCR4 signaling directly influences dentritogenesis, we co-cultured Purkinje cells and granule cells from both WT and KO E17 embryos. There was no notable morphological difference between WT and KO Purkinje cells for either the CXCR4 antagonist treated or non-treated groups. CXCR4 is not required for the Purkinje cell dendritic development in vitro. This suggests the in vivo abnormalities caused by knocking out Cxcr4 may be due to abnormal migration and a foreign extracellular environment rather than a direct effect of knocking out the gene. Additionally, we detected ectopic location and lower density of granule cells in the KO cerebella. This may contribute to abnormal development of the dendrites of Purkinje cells. Furthermore, although our cell count per section showed no difference compared to WT cerebella, since the volume of the KO cerebellum is reduced there should be a decrease in the total number of PCs in the KO cerebellum.
We went on to assess the granule and Purkinje cells in adult Cxcr4 KO mice. A previous paper demonstrated Cxcr4 deficient mice feature abnormal development of the dentate gyrus neonatally with distinct inner and outer blades failing to form. This hippocampal defect is corrected from P14 into adulthood [22,23]. However, the postnatal developmental process does not correct the irregular cerebellum in adult Cxcr4 KO mice. The abnormal cerebellum, reduced dendrite density and chaotic axon projections of Purkinje cells are still present in Cxcr4 null mice in adulthood. However, these axon projections do successfully project to the deep cerebellar nuclei.
Thus, to look for the factors that may contribute to the cerebellar phenotype caused by Cxcr4 deficiency, we measure the gene expression pattern change from the P1 cerebellum. The gene expression profile showed the two most significantly affected pathways were the ''ECM receptor interaction'' and ''focal adhesion''. For example, genes encoding extracellular matrix components such as collagens (Col1a1, Col1a2, Col2a1, Col3a1, Col4a1, Col4a5, Col4a6, Col5a1, Col5a2, Col6a1, Col6a2 and Col11a2) and laminins (Lama1, Lama2, and Lama3) were significantly reduced in KO animals. Similarly, genes encoding for signaling molecules that promote cell migration, such as Met and Mylk, were also down-modulated in KO animals. Conversely, Slit3 and Sema5a, were significantly increased in knockout mice. SLIT3 is a negative regulator for CXCR4 and CXCL12 and inhibits neurite outgrowth in cultured embryonic and fetal stem cells [24,25]. SEMA5A inhibits glioma cell migration through RAC1 inactivation [26]. These results further support the hypothesis that a cell migration defect results in extra cellular environment unsuitable for the maturation of developing Purkinje cells. Recently, it has been suggested that Neurog2 is a key regulator of Purkinje cell development and maturation [17]. Ptf1a is a direct regulator of Neurog2 and is also required for the specific generation of Purkinje cells [16,27]. In our expression array, we detected lower level of expression of these two genes in the KO. We confirmed this finding with quantitative PCR. This suggested that a lower level of Neurog2 and Ptf1a may partially contribute to the defect of Purkinje cell dentritogenesis.
Finally, we assessed the motor and coordination systems of the KO mice. Historical evidence shows that the cerebellum is heavily involved in regulating these systems and that the behavioral defects observed in the KO are consistent with the cerebellar and Purkinje cell abnormalities we observed histologically [28,29,30]. Not surprisingly, the KO mice performed poorly in coordination, muscle strength, balance, and skilled walking tests.
In summary, we used Cxcr4 conditional KO mice to investigate the role of CXCR4 in postnatal cerebellar development, transcriptome profile, and motor behavior. We have provided a systematic study of the effects of knocking out Cxcr4 within the cerebellum from early postnatal to adult life. We found that CXCR4 is essential for cell migration and that the ectopic positioning of granule cells causes the Purkinje cell ectopia and aberrant dendritogenesis. Our data suggests that Purkinje cells develop abnormally in vivo in Cxcr4 null mice due to an altered cerebellar environment resulting from changes to ECM receptor interactions and focal adhesion. In a series of behavioral tests, the KO mice exhibit poor performance in balance, coordination and skilled walking behavior. This is consistent with functional disruption of the cerebellum. In this study, we provide data on postnatal phenotypes to demonstrate that failure of CXCR4 mediated signaling during early development has profound implications for the development of the cerebellum throughout life. Figure S1 (A,B) The CNPase and Calbindin staining shows that Purkinje cells send axons to deep cerebellar