Abl Family Tyrosine Kinases Are Essential for Basement Membrane Integrity and Cortical Lamination in the Cerebellum

Generation of CNS-specific Abl knock-out mice on the Arg^−/− background

To investigate the role of Abl in CNS development, a floxed allele of Abl (Abl^F) was generated using Cre/loxP technology. LoxP sites upstream of exon 5 and downstream from exon 6 were introduced into ES cells by homologous recombination (Fig. 1A). Upon Cre-mediated recombination, excision of the fifth exon and sixth exon would delete part of the tyrosine kinase domain and fully inactivate the gene as determined by Western blot (supplemental Fig. S1, available at www.jneurosci.org as supplemental material, Fig. 1E), creating an Abl-null allele (Abl^Δ). Splicing of the primary transcript to join the flanking exons, if it occurred, would produce a frameshift mutation. Abl^F/F mice appeared healthy and indistinguishable from their wild-type littermates. Abl^F mice were first bred to males expressing an EIIa-Cre transgene (Agah et al., 1997), which deletes loxP-targeted DNAs in the germline, generating the Abl^Δ allele. Mice with combinations of different Abl alleles (wild-type, Abl^F and Abl^Δ) were genotyped by Southern blot (Fig. 1B) and PCR analyses (Fig. 1C) of tail DNA. Intercrosses to generate homozygotes of the Abl^Δ allele resulted in partial neonatal lethality with various developmental defects, such as runting, splenic, and thymic atrophy (supplemental Fig. S1, available at www.jneurosci.org as supplemental material), consistent with the previous descriptions of Abl-null mice (Schwartzberg et al., 1991; Tybulewicz et al., 1991).

CNS-specific inactivation of Abl was accomplished by crossing Abl^F mice to a transgenic line expressing Cre under the control of the rat nestin promoter and enhancer (Nestin-Cre) (Tronche et al., 1999). Nestin-Cre was shown to induce potent and widespread recombination in precursors of neurons and glia starting around E10 (Graus-Porta et al., 2001). Genomic DNA from multiple tissues dissected from Abl^F/F;Nestin-Cre or control newborns were analyzed by Southern blot analysis, confirming efficient recombination within the floxed Abl allele specifically in the brain (Fig. 1D). Furthermore, Western blot analysis of extracts from Abl^F/F;Nestin-Cre newborns showed nearly complete loss of Abl protein in the brain, but not in the adjacent meninges (Fig. 1E). Abl^F/F;Nestin-Cre mice were grossly normal and histological analysis of the brain revealed no obvious morphological defects. This was not a surprise, since no brain malformation was detected in Abl knock-out mice (Schwartzberg et al., 1991; Tybulewicz et al., 1991). We then crossed Abl^F/F;Nestin-Cre mice onto an Arg^−/− background (Koleske et al., 1998) and generated Abl^F/F;Arg^−/−;Nestin-Cre mice, which we refer as Abl/Arg CNS-specific double conditional knock-out (CNS-DKO) mice hereafter.

Abl/Arg deficiency in the brain results in severe cerebellar lamination defects and motor coordination deficits

Abl/Arg CNS-DKO mice were viable and recovered at a frequency (20.5%) slightly less than the expected Mendelian ratio (supplemental Table 1, available at www.jneurosci.org as supplemental material). Whole-mount analysis of the CNS-DKO young adult brain revealed profound irregularities in the cerebellum, while the rest of brain appeared normal (Fig. 2A). The phenotype was observed in all DKO mice examined, but not in their control littermates. The brains from those mice that still retained at least one functional copy of either Abl or Arg gene (e. g. Abl^F/+;Arg^−/−;Nestin-Cre) were normal upon histological examination, indicating that the functions of the two genes are largely redundant. The DKO cerebellum was smaller in size compared with control. The anterior–posterior extent of the mutant cerebellum, with the disappearance of lobules IV and V in the anterior vermis (Fig. 2A, arrows), was markedly reduced. Furthermore, both the cerebellar vermis and the two lateral hemispheres lacked clear fissures on the surface.

Histological analysis of the DKO brain revealed proper lamination of the cerebral cortex and hippocampus (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). However, sagittal and coronal sections of the DKO cerebellum not only confirmed that its length along the anterior–posterior axis was dramatically reduced, but also revealed several other striking abnormalities (Fig. 2B–G). First, the basic laminar structure of the anterior cerebellum, encompassing lobules I–V, was completely lost in the DKO mice. As a result, patches of granule cells were randomly scattered throughout the white matter. Second, although the basic foliation pattern of the hemispheres and the posterior cerebellum seemed to be preserved (supplemental Fig. S3, available at www.jneurosci.org as supplemental material), adjacent lobules, such as lobules VIII and IX, were fused and the depth of the fissures was also significantly decreased (Fig. 2F,G). Moreover, while a rudimentary IGL formed, many granule cells failed to migrate and formed ectopias at the cerebellar surface and along the fusion lines of adjacent lobules (Fig. 2F,G, arrows). Together, the histological analysis demonstrates that the development of cerebellar cortex is specifically compromised in the absence of Abl and Arg, leading to anterior cerebellar morphogenesis defects and cortical lamination abnormalities.

Considering the profound morphological defects in the cerebella of Abl/Arg CNS-DKO mice, we examined motor coordination function in these mutants. The DKO mice did not display obvious ataxia in an open field test. However, when placed on an elevated balance beam, differences between the DKO mice and their control littermates were evident (supplemental movies S1, S2, available at www.jneurosci.org as supplemental material): DKO mice took approximately twice as long to cross the beam (Fig. 2H) and also demonstrated a dramatic increase in the number of foot slips (Fig. 2I). Thus, the DKO mice exhibit substantial motor coordination deficits that do not affect normal locomotor function but becomes apparent upon challenge with more complicated tasks. Like the cerebellar malformation, poor motor coordination was also observed only in the DKO mice, suggesting that the impairment of motor function is probably caused by cerebellar defects.

Disruption of the BM during embryonic development leads to morphogenesis defects in the anterior cerebellum

To determine the developmental stage at which the malformation of the anterior cerebellum arises, we examined Abl/Arg CNS-DKO mice at younger ages (Figs. 3, 4). Sagittal sections of P7 and P1 cerebella showed obvious defects in the organization of anterior lobules I–V in the DKO mice (Fig. 3A–D). Immunostaining of P1 sections with markers for granule cells (Pax6) and Purkinje cells (calbindin) revealed that both cell types were mislocalized: many GCPs were ectopically embedded within the cortex and Purkinje cells were loosely aligned, with some abnormally distributing at the cerebellar surface (Fig. 3E,F). At E17, while RORα-positive Purkinje cells formed a normal multilayer below the EGL (Fig. 3G), GCPs appeared more dispersed in the anterior DKO cerebellar primordium when compared with control (Fig. 3G). A close examination revealed that laminin-labeled BM was fragmented, and scores of GCPs protruded into the adjacent subarachnoid space between the midbrain and cerebellum in the DKO embryos (Fig. 3H, arrows). These observations suggest that the mislocalization of Purkinje cells (supplemental Fig. S4, available at www.jneurosci.org as supplemental material, Fig. 3F) in postnatal DKO mice is likely a consequence of the anterior cerebellar malformation involving defective BM and abnormally distributed granule cells.

To probe whether loss of the BM integrity is caused by aberrant tangential migration of GCPs, we analyzed sections from earlier embryos. At E15, while the migrating GCPs had not yet arrived at the anterior cerebellum, there were already multiple small discontinuities in the BM in the DKO mice (Fig. 4B,D′, arrows), suggesting that the BM is not disrupted by migrating GCPs. During early CNS development, radial glial processes extend from the ventricular zone to the cortical surface, where their endfeet are the preferred site of BM assembly (Halfter et al., 2000). Therefore, the failure of radial glia to maintain the BM may potentially cause its breach. Indeed, we found that the BM in the anterior cerebellum at E15 serves as the primary anchorage site for dense brain lipid-binding protein (BLBP)-positive radial glial processes (Fig. 4C,C′). In the DKO cerebellum, radial glial processes were highly disorganized, with some of their endfeet protruding into the meninges (Fig. 4D,D′, asterisk). Since the Nestin-Cre transgenic mice did not induce recombination in the ECM-secreting meningeal fibroblasts (Fig. 1E), these data provide strong evidence that Abl family kinases in radial glia is likely required for BM integrity. In conclusion, our time course analyses demonstrate that the abnormalities in the BM in Abl/Arg CNS-DKO embryos, likely caused by altered radial glia-BM interactions, leads to mislocalization of tangentially migrating GCPs and subsequent foliation defects in the anterior cerebellum.

Abnormal Bergmann glial network and defective BM maintenance in the postnatal cerebellum

During postnatal cerebellar development, GCPs in the EGL proliferate and give rise to differentiated granule cells that migrate radially along the Bergmann glial fibers and then through the PCL to form the IGL (Hatten, 1999; Sillitoe and Joyner, 2007). By P21, the migration of granule cells is complete and the EGL is no longer present, leaving behind the molecular layer (ML). Staining of a postmitotic neuronal marker NeuN showed frequent granule cell ectopias in the posterior regions of adult Abl/Arg CNS-DKO cerebella, where the basic foliation pattern was generally intact (Fig. 5A–D). Ectopic granule cells were predominately located at the cerebellar surface and along the fused fissures (Fig. 5B, arrow). This distribution correlated with regions in which the loss of the pial BM starting from around P8 (Fig. 5F, arrows). In control cerebellum, GFAP-labeled Bergmann glia extended processes across the ML, and their endfeet were anchored along the BM (Fig. 5G,I). In contrast, the glial fibers in the DKO mice were highly disorganized, with many random protrusions at the cerebellar surface (Fig. 5H,J). Since the interaction with the endfeet of Bergmann glial fibers has been suggested to be critical for maintenance of the BM in the postnatal cerebellum (Graus-Porta et al., 2001; Belvindrah et al., 2006), our results indicates that, similar to the defect in the DKO cerebellar primordium, the disruption of the BM in the postnatal cerebellum may likely also result from the abnormal Bergmann glial network.

GCP proliferation and differentiation in Abl/Arg CNS-DKO mice

With the formation and growth of the lobules, the size of the postnatal mouse cerebellum increases dramatically, which is largely caused by the proliferation of GCPs. The smaller cerebella with reduced foliation in the posterior cerebellum in the DKO mice (Fig. 2C) indicate a deficiency in GCP proliferation. To directly assess proliferative defects, we performed 1 h BrdU pulse-labeling experiments at P8 during the peak stage of GCP proliferation (supplemental Fig. S5, available at www.jneurosci.org as supplemental material, Fig. 6A,B). Because the secondary fissure is the major fissure in the posterior cerebellum, the initial formation of which is normal in the DKO mice (supplemental Fig. S5, available at www.jneurosci.org as supplemental material and Fig. 6B), we quantified the number of proliferating GCPs in it and found a significant reduction in the DKO mice (311 ± 17 BrdU-positive cells/mm; n = 3 mice for each genotype) compared with controls (405 ± 25 BrdU-positive cells/mm). Furthermore, loss of proliferating GCPs seemed to be concentrated in areas where the BM was disrupted (Fig. 6B, arrows). Since an intact BM is important for maintaining the proliferative capacity of GCPs (Blaess et al., 2004; Belvindrah et al., 2006), these data suggest that impaired granule cell proliferation may not be a cell-autonomous defect. To test this hypothesis, we isolated granule cells from control and DKO pups and cultured them for 48 h in medium with or without 3 μg/ml Shh, which is a potent mitogen for GCPs and regulates the growth and elaboration of the cerebellar lobules (Corrales et al., 2006). Before fixation, cells were labeled with BrdU and the prevalence of BrdU-positive cells was determined (Fig. 6C–G). Consistent with previous studies (Wechsler-Reya and Scott, 1999), the vast majority of GCPs derived from control mice stopped proliferating in the absence of Shh, and the addition of Shh strongly promoted GCP proliferation. A similar response in proliferation of GCPs from the DKO mice was also observed (Fig. 6G), indicating that Abl kinases are not required for the proliferation of GCPs in response to Shh in vitro. This result further suggests that the decreased granule cell proliferation in the DKO mice is likely secondary to defects in the BM.

During normal differentiation, granule cells exit the cell cycle, enter the inner EGL, and spontaneously express differentiation markers such as the adhesion molecules L1 and TAG1 (Fig. 6H,J). In the DKO mice, however, with focal GCPs losing contact with the BM, they exited the cell cycle (Fig. 6B) and ectopically expressed L1 and TAG1 at the cerebellar surface (Fig. 6I,K). This might result in a depletion of the precursor pool within the EGL and likely contribute to the cerebellar hypoplasia in the DKO mice.

Granule cell migration in Abl/Arg CNS-DKO mice

The observation that granule cell ectopias formed at sites of broken BM in the postnatal DKO cerebellum (Fig. 5) raises the possibility that the phenotype is secondary to abnormalities in the BM and is not directly caused by intrinsic migration defects of Abl/Arg-deficient granule cells. To evaluate the rate of granule cell migration, we labeled proliferating GCPs with a single BrdU injection at P6 and analyzed their extent of migration at P10 before the extensive loss of the BM in the DKO mice (Fig. 7A,B). As expected, the majority of BrdU-labeled cells (90.9 ± 2.7%; n = 3 mice for each genotype) were found in the IGL in lobule VIII in control mice. A similar percentage (85.3 ± 4.0%) was also observed in the same lobule in the DKO cerebella, except in few regions where the integrity of the BM was already destroyed and clusters of BrdU-labeled granule cells remained at the cerebellar surface (data not shown). This provides strong evidence that the intrinsic program for cerebellar granule cell migration is not perturbed in the DKO mice.

To investigate this further, we examined granule cell–glia interactions both in vivo and in vitro. The attachment of migrating granule cells to Bergmann glial fibers was readily detected in both control and DKO cerebellar sections (Fig. 7C,D). The close interaction was also evident in cultures derived from both control and DKO mice in which many of the granule cell bodies assumed an elongated shape, indicating that they were migrating (Fig. 7E,F). Together, these results demonstrate that Abl family kinases are not essential in either granule cell or glia for their interactions.

Granule cell-specific deletion of Abl family kinases is not sufficient to induce a cerebellar lamination defect

To further determine whether Abl family kinases in granule cells are essential for cerebellar lamination and their proliferation, we generated conditional knock-out mice in which Abl was specifically deleted from granule cells using a Math1-Cre transgenic line (Matei et al., 2005). Math1-Cre was characterized by crossing with a ROSA26lacZ reporter line (Soriano, 1999). As shown by lacZ staining, Math1-Cre drives efficient recombination in all GCPs except those in the most posterior part of the cerebellum (Fig. 8A,B). Western blot analysis of cerebellar extracts from Abl^F/F;Math1-Cre pups showed that the level of Abl protein was markedly decreased when compared with that in control mice (Fig. 8C). The residual Abl protein detected probably reflects the protein present in other Cre-negative cells such as Purkinje cells and glia in the cerebellar tissue. When Abl^F/F;Math1-Cre mice were crossed into the Arg knock-out background, the cerebella of the resulting mutant mice were strikingly normal, with no malformation or granule cell ectopia observed (Fig. 8D,E). Although we cannot rule out the possibility that Abl protein is not completely deleted in granule cells by Math1-Cre, this result provides further genetic evidence that granule cells are not the primary cause of the cerebellar lamination defects observed in Abl/Arg CNS-DKO mice.

Analysis of signaling pathways regulated by Abl family kinases

In fibroblasts, integrin-mediated attachment on the ECM triggers the tyrosine phosphorylation of intracellular signaling proteins, including Abl family kinases and the adaptor protein CrkII, and leads to the rearrangement of the actin cytoskeleton (Bradley and Koleske, 2009). Interestingly, among other defects, mice lacking Crk family members in the brain also showed similar granule cell ectopias as observed in Abl/Arg CNS-DKO mice (Park and Curran, 2008). CrkII binds to the PxxP motifs of Abl family kinases through its SH3 domain (Antoku et al., 2008) and is phosphorylated by them at residue Y221 (Feller et al., 1994). When Y221 is phosphorylated, CrkII forms an intramolecular interaction with its SH2 domain, which blocks its activity as an adaptor protein and prevents its interaction with a key scaffold molecule, p130Cas (Kain and Klemke, 2001). While it was unchanged in Arg^−/− newborn brain extracts, phosphorylation of CrkII Y221 was significantly reduced in Abl^F/F;Nestin-Cre extracts and almost completely abolished in Abl/Arg CNS-DKO extracts (Fig. 9A). This indicates that CrkII phosphorylation at Y221 in the brain is primarily dependent on Abl and, to a lesser extent, on Arg. The diminished phosphorylation of CrkII Y221 also correlated with the increased association of p130Cas with Crk in the DKO extracts (Fig. 9B). Although how Abl-mediated CrkII phosphorylation affects Rac1 activity is dependent on different biological contexts (Bradley and Koleske, 2009), the overall level of active Rac1 GTPase was elevated in the DKO brain extracts (Fig. 9B).