Neuronal dystroglycan is not essential for cerebellar histogenesis
Dystrolygcan (DG) is broadly expressed in most cellular populations of the developing cerebellum—pre-migratory granule cells, Purkinje cells, and radial/Bergmann glia [21]. To evaluate how DG expression in each of the aforementioned cell types contributes to cerebellar histogenesis, we generated cell-specific conditional DG-null mice (Figure 1). We first examined the cerebellum of PCP2-Cre/DG-null mice, where DG is conditionally deleted only in PCs, then in malpha6-Cre/DG-null mice, where DG is conditionally deleted from GCs (Figure 2).
Dystroglycan is expressed in Purkinje cells (PCs) during early postnatal cerebellar development [21], and DG-positive puncta rimmed the PC somata and decorated PC dendrites across the molecular layer in the adult (P21) control cerebellum (Figure 2I). To assess whether loss of DG in PC is adequate to cause migration defects, we evaluated the histopathology of PCP2-Cre/DG-null mice. The PCP2 promoter drives Cre expression in PCs beginning on postnatal day 6 [23]. The brains of these mice were grossly normal and many histologic sections were free of histopathology (Figure 2G, H). After thorough examination of several adult mice (n = 9), small, infrequent GC heterotopia were identified in the molecular layer of each cerebellum (Figure 2C, F), just internal to an intact glia limitans (Figure 2N). No disruptions of the glia limitans or abnormalities of Bergmann glia orientation were observed (Figure 2M, N) despite a nearly complete loss of PC dystroglycan (Figure 2L).
Granule cells in the cerebellar cortex express DG during their migration and lose expression during maturation in the internal granule cell layer, suggesting that expression of DG in GC may play a role during their migration [21]. To determine whether the loss of DG in GC is sufficient to cause defects in GC migration, the histopathology of the cerebellum was examined in malpha6-Cre/DG-null mice, where DG is conditionally deleted from GCs. The malpha6 promoter drives Cre expression in GCs beginning at postnatal day 4 [24]. The brains of these animals (n = 7) were grossly normal (Figure 2D) and GC ectopia were not observed (Figure 2E).
Cerebellar growth in conditional dystroglycan-null mice
After establishing that neuronal DG is not necessary for cerebellar histogenesis, we next evaluated the role of glial DG in cerebellar development by using mice with deletion of DG from both neuronal and glial cells. The generation of the GFAP-Cre/DG-null and nestin-Cre/DG-null mice has been previously described [5, 7, 8]. The nestin promoter drives Cre expression in neuronal and glial precursors as early as embryonic day 10.5, while the ependyma and choroid plexus show no expression of the nestin-Cre transgene [27]. The GFAP promoter drives expression of Cre recombinase in radial glia, astrocytes, ependyma and some neuronal cell types beginning at embryonic day 13.5, but Purkinje cells and the choroid plexus do not express GFAP-Cre [28]. The cerebella of these mice were grossly normal; area measurements of mid-sagittal sections from three different post-natal developmental time points showed no significant differences among the two mutant genotypes and their littermate controls (Figure 3N). However, several ectopic foci of granule cells were noted in the cerebella of both GFAP-Cre and nestin-Cre conditional DG-null mice at post-natal day 8 (Additional file 1: Figure S1) and day 16 (Figure 3H, I). Under higher magnification, these ectopic cells appeared as clusters of non-migrating GCs that remained in the EGL along some fissures and at the surfaces of some lobules, with at least some granule neuron clusters migrating outward, through gaps in the basal lamina, into the leptomeninges (Figure 3, K-M). By P21 this pattern of dysgenesis appeared as fused interlobular fissures (Additional file 2: Figure S2). PCs in the meanwhile developed normally albeit for focal secondary irregularities in dendritic arborization—some dendrites reached across adjacent lobules at areas of discontinuous basal lamina and GC ectopia (Additional file 3: Figure S3).
Dystroglycan expression in the developing cerebellum
To evaluate DG expression in the developing cerebellum, mid sagittal sections from cerebella of nestin-Cre/DG-null and littermate controls were double stained for α-DG and laminin, a basement membrane protein and ligand of αDG. DG was found to be concentrated at the choroid plexus and the glia limitans from at least embryonic day 14 (E14.5), and it colocalized with laminin at the glia limitans (Figure 4, A-C). Loss of DG in the nestin-Cre/DG-null cerebellar anlage was confirmed at E14.5 (Figure 4D). Laminin staining remained indistinguishable from control tissue, indicating an intact basement membrane at this developmental time point. Soon after birth (P0), laminin expression at the basal lamina was absent in small foci (Figure 4E), but the extent of abnormal laminin immunostaining spread as the cerebellum grew to P3 (Figure 4F). These breaks at the basal lamina were found at the tips of cerebellar lobules and along fissures between lobules of the nestin-Cre/DG-null mice. In the developmentally mature control cerebellum (P21), DG was expressed in Purkinje cells (PCs) in addition to being concentrated at glial endfeet (Figure 4G). Both neuronal and glial DG were lost in the nestin-Cre/DG-null mice (Figure 4H), whereas the GFAP-Cre/DG-null mice showed loss of DG from glial cell endfeet, while maintaining expression in PCs (Figure 4I).
Disruptions at the glia limiting membrane correlate with radial glia/Bergmann glia irregularities
Since DG is normally expressed and concentrated at glial endfeet abutting the glia limitans, we next assessed the morphology and orientation of radial glia (Bergmann glia) in the absence of DG. Brain lipid binding protein (BLBP) and nestin were used as markers for radial glia at P0 and P3 due to their expression early in cerebellar development, whereas glial fibrillary acidic protein (GFAP) was employed as a glial marker from P8 onward. Perlecan and laminin are protein constituents of the extracellular matrix that are ligands for α-DG. Radial glia (and later Bergmann glia) are oriented in such a way that their endfeet abut and form the glia limiting membrane underneath the basement membrane (Figure 5A, C). In the nestin-Cre/DG-null cerebellum, small breaks at the glia limitans that were first observed at P0 then became much more visible at P3, and were accompanied by projections of glial processes into these gaps (Figure 5B, D). Breaches became exacerbated as the basement membrane, visualized by both DG ligands (laminin and perlecan) and a non-DG ligand (type IV collagen; Additional file 4: Figure S4), became severely disrupted by P8, while Bergmann glial endfeet projected into the gaps at the basement membrane or retracted underneath the EGL (Figure 5F). Electron microscopy studies confirmed the presence of disrupted basement membrane, abnormally organized glial endfeet, and ectopic cells within the leptomeninges of nestin-Cre/DG-null cerebella (Figure 6). Additionally, β1 integrin (β1itg), a laminin receptor, was expressed in Bergmann glia and concentrated at glial endfeet (Figure 5G); a compensatory change in β1itg expression was not observed at the glia limitans in the nestin-Cre/DG-null mice, although focal areas of gliosis did show high levels of expression (Figure 5H). Furthermore, we observed reactive gliosis at sites with disrupted basement membrane in developmentally more mature mice (Figure 7). Taken together, our data indicate that the expression of DG in glial cells is important for their morphological development as well as the integrity of the surface basement membrane during cerebellar growth.
We further evaluated the expression of proteins known to associate with DG and the dystrophin-glycoprotein complex (DGC) in glial cells, namely aquaporin 4 (AQP4) and potassium inward rectifying channel 4.1 (Kir4.1). The membrane pore protein AQP4 comprising the major water channels in brain is clustered together with the DGC via interaction between AQP4 and α-syntrophin [29], which in turn binds to dystrophin and β-DG. In the cerebellum, AQP4 concentrated at the glia limitans in control (Additional file 5: Figure S5A, B) but was reduced at P8 and lost by P16 in the nestin-Cre/DG-null mice (Additional file 5: Figure S5D, E). Whether loss of AQP4 at glial endfeet in these mice causes other pathologic phenotypes remains to be investigated, but AQP4-null mice show no brain morphological defects, suggesting that the presence of ectopic cells in the nestin-Cre/DG-null mice is independent from AQP4. Kir4.1 is enriched at Müller glial endfeet abutting the inner limiting membrane in the retina; defects in Kir4.1 clustering and retinal physiology have been demonstrated in the nestin-Cre/DG-null mice [6]. In the cerebellum, Kir4.1 is expressed in the molecular layer, but is not concentrated at the glia limitans. Expression of Kir4.1 was indistinguishable between nestin-Cre/DG-null and littermate control (Additional file 5: Figure S5C, F), indicating that Kir4.1 localization in the cerebellum, unlike in the retina, is not dependent on DG.
Proliferation and migration of postnatal granule cells
To assess GC proliferation immediately after birth (P0) and again at the peak of cerebellar growth (P8), 5-bromodeoxyuridine (BrdU) was acutely given to mice 30 minutes prior to sacrifice, and cerebellar tissues were evaluated in cryosections using an antibody against BrdU. BrdU strongly labeled the EGL at P0 and the outer proliferative EGL (EGLo) at P8 in the control cerebellum (Figure 8A, F). The nestin-Cre/DG-null mice at P0 showed a similar expression profile of BrdU positive cells and were not significantly different from the control tissues (Figure 8B, I). At P8, ectopic GCs labeled by neuronal nuclei protein (NeuN) and residing in areas of highly disruptive basal lamina (Figure 8E, H) have stopped proliferating. Quantitative analysis showed a significantly reduced number of BrdU-positive cells in regions with a disrupted basal lamina (Figure 8J) compared to cells beneath an intact basement membrane (Figure 8G) and compared to control cerebellum (Figure 8F).
To investigate the fate of GC produced during the latter half of cerebellar histogenesis, pulse-chase migration studies were performed by BrdU administration at P8 and evaluation of cerebellar tissues at P16. In the control mice, the vast majority of cells labeled at P8 were found within the IGL at P16. In contrast, many of these later-born GCs remained at the cerebellar surface within ectopia of the nestin-Cre/DG-null cerebellum (Figure 9).
We additionally evaluated expression levels of sonic hedgehog (Shh)—a critical morphogen involved in the patterning of many brain structures during development—at P8 and P16 to confirm that GC proliferation and migration defects were independent from Shh pathways. Shh expression in the cerebellum was found to be comparable between nestin-Cre/DG-null and control littermates (Additional file 4: Figure S4).
In comparing the degree of glia limitans pathology and ectopic GC over entire mid-sagittal sections, sparing of lobules IV-V and VI was noted (Additional file 6: Table S1). A quantitative comparison of lobules IV-V and IX was carried out in nestin-Cre/DG-null mice (Figure 9G), and complete mapping of mid-sagittal sections was performed in both nestin-Cre/DG-null and GFAP-Cre/GFAP-null mice (Figure 10A). The spatially asymmetric pathology had the greatest degree of sparing limited to the tips of lobules IV-V; the tip of lobule VI was spared to a lesser degree. Fissures between these lobules were not spared in comparison to any other fissures. This analysis also showed gradually more severe pathology over an increasingly larger surface area of the developing cerebellar cortex in both nestin-Cre/ and GFAP-Cre/DG-null mice, with the former showing a somewhat more severe phenotype at each postnatal age evaluated. Evaluation of parasagittal sections showed that glia limitans and GC pathology extended laterally to involve the cerebellar hemispheres in both models. The spatial distribution of pathology away from the midline in one nestin-Cre/DG-null cerebellum was mapped (Figure 10B).
Characterization of ectopic granule cells at sites of disrupted basal lamina
To assess GC maturation, the expression of several markers was evaluated in the course of post-natal cerebellar development. The inner, post-mitotic, pre-migratory zone of the EGL (EGLi) was labeled with transient axonal glycoprotein 1 (tag1) and microtubule-associated protein 2 (MAP2) at P8 (Figure 11A, C). Tag1 is down-regulated as GCs begin to migrate inward across the molecular layer (ML) to settle in the internal granule cell layer (IGL); tag1-null mice displayed a mild GC migratory defect [30]. MAP2 is a cytoskeletal protein that plays essential roles in a number of cellular developmental processes, including neuronal morphogenesis. In the nestin-Cre/DG-null cerebellum, tag1-positive cells were visible within the EGLo as ectopia bridging between adjacent lobules (Figure 11B); MAP2 showed similar expression to tag1 in small foci of ectopic GCs (Figure 11D). Meanwhile, the paired box 6 (pax6) transcription factor, a DNA-binding protein crucial for CNS development, was shown in cells of the EGL underneath an intact basement membrane (Figure 11E). GCs from Pax6-null mice, due to defect in cytoskeletal organization and polarization, do not form leading edges and migrate in random directions instead of radially inward [31]. The expression level of pax6 in ectopic cells, at sites of disrupted basement membrane, was similar to that of surrounding cells and in control (Figure 11H). These data indicate that the cellular machinery responsible for GC maturation is largely unperturbed in the nestin-Cre/DG-null cerebellum, insofar as we have demonstrated, and that the cause of ectopia is seemingly due to aberrant glial organization and reactive gliosis rather than GC-intrinsic abnormality. In support of this interpretation, GC ectopia at later developmental stages (P16 and P21) expressed markers of mature neurons normally found in the granule cell layer (GCL), for instance NeuN and GABAA receptor α6 (GABARA6) (Figure 11I, J).