Aspm sustains postnatal cerebellar neurogenesis and medulloblastoma growth in mice

Microcephaly syndromes may provide clues for identifying genetic targets for brain tumor therapy. During brain development, proliferation of transient-amplifying cells generates appropriately sized populations of different types of neurons. In primary microcephaly, impaired progenitor amplification produces a developmental outcome whereby the overall brain size is markedly reduced, but the diversity of cell types is preserved (Woods et al., 2005). Mutations found in microcephaly may indicate genes required by transient-amplifying cells. Because specific brain tumors grow as transient-amplifying cells, disrupting microcephaly genes may produce a clinically useful anti-tumor effect.

The microtubule-associated centrosomal protein ASPM (abnormal spindle microcephaly-associated) has been implicated in both primary microcephaly and cancer. Mutations in human ASPM occur in familial microcephaly (Bond et al., 2002, 2003; Woods et al., 2005), and studies in model organisms have shown that ASPM is expressed by neural stem cells and promotes brain growth (Bond et al., 2002; Pulvers et al., 2010; Marinaro et al., 2011; Rujano et al., 2013). ASPM has also been implicated in the growth of diverse cancers, including medulloblastomas (Vulcani-Freitas et al., 2011), gliomas (Horvath et al., 2006; Bikeye et al., 2010), hepatocellular carcinomas (Lin et al., 2008), ovarian cancers (Brüning-Richardson et al., 2011) and pancreatic cancers (Wang et al., 2013). Thus, loss of Aspm function is associated with growth failure, and aberrant ASPM is associated with excessive growth.

Various functions of ASPM have been identified that may support proliferation. In Drosophila, asp maintains mitotic spindle orientation during both mitosis and meiosis (Zhong et al., 2005; van der Voet et al., 2009; Kaindl et al., 2010; Rujano et al., 2013). In mice, Aspm is required to maintain spindle organization and positioning and acts as a microtubule-organizing center in telencephalic neuroepithelial progenitors (Fish et al., 2006; Higgins et al., 2010). Asp and Aspm are expressed at centrosomes and spindle midbody microtubules, and their loss causes defects in spindle assembly and mitotic progression (Saunders et al., 1997; Wakefield et al., 2001; Fish et al., 2006; van der Voet et al., 2009; Pulvers et al., 2010; Novorol et al., 2013). Aspm might also play a role in DNA double-strand break repair through non-homologous end joining (Kato et al., 2011). The net effect of these diverse functions is promotion of proliferation, but it remains unclear how these functions converge.

The postnatal cerebellum presents an opportunity to study Aspm function in both developmental brain growth and tumorigenesis. Cerebellar growth is typically reduced in primary microcephaly (Adachi et al., 2014), whereas medulloblastoma, the most common malignant pediatric brain tumor, is a neoplastic overgrowth of the cerebellum. Postnatal CGNP proliferation drives cerebellar growth. This proliferation lasts through the first 2 weeks of life in mice or first year of life in humans, generating the largest population of neurons in the brain. During this period, CGNPs proliferate in the external granule layer (EGL) of the cerebellum. Actively cycling CGNPs are PCNA⁺ and typically reside in the outer zone (oEGL), whereas postmitotic CGNPs express Cdkn1b (p27) and accumulate in the inner zone (iEGL). From the iEGL, postmitotic CGNPs migrate through the molecular layer (ML) into the internal granule layer (IGL) where they complete differentiation to become granule neurons (GNs) (Hatten et al., 1982; Roussel and Hatten, 2011; Hibi and Shimizu, 2012).

Postnatal CGNP proliferation depends on mitogenic signaling from locally secreted Shh (Dahmane and Ruiz-i-Altaba, 1999; Wechsler-Reya and Scott, 1999; Kenney and Rowitch, 2000). Shh pathway activation is found in 25% of medulloblastoma patients including 65% of patients younger than 2 or older than 16 years (Zurawel et al., 2000a,b; Ellison et al., 2011; Northcott et al., 2011). In genetically engineered mice, Shh pathway hyperactivation in CGNPs produces malignant cerebellar tumors that resemble human SHH subtype medulloblastomas, providing in vivo models of medulloblastoma tumorigenesis (Zurawel et al., 2000a,b; Hallahan et al., 2004; Schüller et al., 2008; Yang et al., 2008) and metastasis (Hatton et al., 2008).

Here, we show that Aspm plays an essential role in CGNP proliferation, differentiation and survival in mice. Genetic deletion of Aspm reveals important functions in mitotic progression, maintenance of CGNP position and orientation during mitosis, and mitigation of proliferation-dependent DNA damage. Deletion of Aspm increases DNA damage and induces progenitor apoptosis that depends on Bax and p53. Aspm expression persists in Shh-driven medulloblastomas in transgenic mice, where these functions support tumor growth.

Aspm is expressed at major sites of postnatal neurogenesis in the brain. At postnatal day (P) 7, the peak of CGNP proliferation, Aspm mRNA localized predominantly to the oEGL (Fig. 1A). We mapped Aspm expression and postnatal neurogenesis by injecting P6 Aspm-GFP reporter mice with 5-ethynyl-2′-deoxyuridine (EdU) and harvesting brains 24 h later. GFP expression corresponded with sites of proliferation indicated by EdU, including the EGL and subventricular zone (SVZ; Fig. 1B). Widefield (Fig. 1C) and confocal (Fig. 1D) microscopy demonstrated GFP throughout the EGL in P7 Aspm-GFP mice, including differentiating iEGL cells. In light of our in situ hybridization data, we attribute the persistent GFP signal in the iEGL to stability of the GFP protein beyond its period of transcription. Like Aspm mRNA, GFP reporter fluorescence waned by P16 (Fig. 1E), following the temporal pattern of CGNP proliferation (Fujita, 1967). The relationship between Aspm expression and CGNP proliferation was further demonstrated by real-time, quantitative PCR (RT-qPCR). Aspm expression was relatively high at the peak of neurogenesis (P7) and waned by P16 (Fig. 1F), mirroring the expression of cyclin D2 (Ccnd2), a previously identified marker of CGNP proliferation (Kenney and Rowitch, 2000).

We examined the relationship between Shh signaling and Aspm expression in isolated, freshly explanted CGNPs. Shh-supplemented CGNPs in vitro expressed >4-fold more Aspm mRNA compared with Shh-deprived CGNPs (Fig. 1G). Shh signaling is transduced by Smoothened (Smo) and can be blocked by the Smo inhibitor vismodegib (Rubin and de Sauvage, 2006). CGNPs treated with Shh+vismodegib expressed Aspm similarly to Shh-deprived CGNPs (Fig. 1G). CGNPs from Aspm-GFP mice expressed robust GFP when cultured with Shh and downregulated GFP when Shh-deprived (Fig. 1H). These findings show that Aspm expression is integral to the Shh-driven CGNP proliferation program.

To determine whether Shh hyperactivation induces Aspm in vivo, we crossed Aspm-GFP reporter mice with the transgenic, medulloblastoma-prone ND2:SmoA1 mouse line, which expresses a constitutively active allele of Smo under control of the NeuroD2 promoter. CGNPs of ND2:SmoA1 mice show cell-autonomous Shh-pathway activation, prolonged CGNP proliferation, and predisposition to medulloblastoma (Hallahan et al., 2004). In Aspm-GFP;ND2:SmoA1 mice, CGNPs proliferated beyond the typical 15-day period, and expressed Aspm-GFP uniformly throughout the aberrantly persistent EGL (Fig. 1I). In spontaneous medulloblastomas, GFP was consistently and homogeneously expressed (n>15; Fig. 1J,K). Although not specifically implicating components of the Shh pathway as transcriptional regulators of Aspm, these data show that Shh signaling through Smo was necessary to maintain Aspm expression in vitro and is sufficient to drive Aspm expression in vivo, as an integral part of the Shh-driven proliferative program in both persistent CGNPs and Shh-driven medulloblastomas in ND2:SmoA1 mice.

To examine the fate of Aspm-expressing postnatal progenitors, we generated Aspm-CreER;Rosa26-Lox-STOP-Lox-tdTomato mice, which express a tamoxifen-inducible red fluorescent reporter controlled by the Aspm promoter (Marinaro et al., 2011). Mice were harvested at P11 following tamoxifen injection at P4 and P7. Red fluorescence was observed in neurogenic regions throughout the brain (Fig. 2A), including neural progenitors in the rostral migratory stream (Fig. 2B), hippocampus (Fig. 2C; Fig. S1D,E) and EGL (Fig. 2D,E). TdTomato also labeled differentiated neurons in the cerebellum, including interneurons (Fig. 2E and Fig. S1B, yellow arrowheads) and GNs, identified by their position in the IGL and colocalization with NeuN (Rbfox3 – Mouse Genome Informatics) (Fig. 2F; Fig. S1A,B). In the ML, tdTomato and NeuN co-labeled differentiating CGNPs migrating to the IGL (Fig. 2G and Fig. S1A, red arrowheads). In addition, tdTomato⁺ tubular structures throughout the brain expressed the endothelial marker CD31 (Pecam1 – Mouse Genome Informatics), identifying them as capillaries (Fig. 2E and Fig. S1A‴,C, blue arrowheads). No tdTomato⁺ cells expressed the glial marker GFAP, which labels Bergmann glia and astrocytes (Fig. 2H; Fig. S2F). Thus, Aspm⁺ progenitors in the postnatal brain gave rise to CGNPs (identified by their position in the EGL), differentiated neurons, and capillary endothelial cells, with minimal glial progeny.

To determine whether Aspm function is required for CGNP proliferation, we generated and analyzed Aspm^SA mice, which harbor a splice-acceptor and lacZ-neomycin cassette into the intron between exons 6 and 7 (see supplementary Materials and Methods; Fig. S2A). We verified Aspm deficiency in Aspm^SA/SA mice by RT-qPCR, which demonstrated an absence of full-length Aspm mRNA in Aspm^SA/SA cerebella, compared with Aspm^SA/+ littermates (Fig. 3A). Aspm^SA/SA males were sterile, like the Aspm-deficient males previously described (Pulvers et al., 2010). Because of the complex geometry of the brain, we used brain weight to detect changes in size. Whole brains and dissected cerebella of Aspm^SA/SA mice were significantly smaller than brains and cerebella of Aspm^SA/+ or Aspm^+/+ littermates (Fig. 3B,C). These results show that the Aspm^SA/SA mouse line phenocopies previously reported Aspm mutant mice (Pulvers et al., 2010), and that Aspm deficiency reduced the size of the cerebellum along with the rest of the brain.

To determine whether Aspm is specifically required by CGNPs, we generated an Aspm conditional allele (Aspm^f; Fig. S2B) by crossing Aspm^SA/SA and Rosa26-FLPe (Farley et al., 2000) mouse lines. We then used Cre-mediated recombination to delete Aspm conditionally in the Math1 (Atoh1 – Mouse Genome Informatics) lineage (Fig. S2B,C), which in the cerebellum is restricted to CGNPs (Helms et al., 2000; Machold and Fishell, 2005; Matei et al., 2005). The resulting Math1-Cre;Aspm^f/f mice (Aspm cKO), expressed less Aspm mRNA in cerebella, compared with Aspm^f/f littermates without Cre (Fig. 3A). Although overall brain weight was preserved in Aspm cKO mice owing to limited expression of Math1-Cre outside the cerebellum (Fig. 3D), cerebellum weight was significantly reduced compared with littermate controls (Fig. 3E). We also measured the volume of intact, unprocessed brains using magnetic resonance imaging (MRI), and used manual segmentation to determine the volume of the cerebellum (Fig. 3F-I; Movies 1, 2). Consistent with weight measurements, MRIs showed reduced brain and cerebellum volume in Aspm^SA/SA mutants, and reduced cerebellar volume in Aspm cKOs. The smaller cerebellum in the cKO demonstrates a cell-autonomous requirement for Aspm in CGNPs to maintain normal cerebellar growth.

We next sought to determine the onset of growth restriction observed in Aspm cKOs. The EGL typically thins between P7 and P15 as proliferation wanes, whereas the IGL grows as GNs populate it. Measurements of EGL and IGL area at P7 and P11 showed that both regions were significantly smaller in Aspm cKOs, with the most significant difference between genotypes occurring in the IGL at P11 (Fig. 3J). Thus, in Aspm cKOs, the thinning of the EGL was accelerated and the growth of the IGL reduced.

The organization of the EGL was largely preserved in cerebella of Aspm cKOs, with the PCNA⁺ oEGL and post-mitotic, p27⁺ iEGL present at similar ratios as in controls (Fig. 3K,L). In littermate controls at P7, the ratio of the area of the PCNA⁺ region to the p27⁺ region was 1.031±0.032 (mean±s.e.m.), decreasing to 0.838±0.043 by P11 as proliferation wanes. In Aspm cKOs from P7 to P11, the PCNA:p27 ratio decreased more rapidly, from 1.135±0.12 to 0.771±0.054 (Fig. 3M). Comparison of the thickness of the oEGL (PCNA⁺) and iEGL (p27⁺) layers between cKOs and controls showed that each layer grew progressively thinner in cKOs, with the most significant difference occurring in the iEGL at P11 (Fig. 3N). These data suggest a defect in the differentiating population of CGNPs in the Aspm cKOs that becomes progressively more evident over time.

To determine whether Aspm deficiency decreased the mitogenic response of CGNPs to Shh, we compared proliferation in CGNPs isolated from either Aspm cKO mice or Aspm^f/f littermates, cultured with or without Shh for 48 h by counting mitotic figures, identified by immunofluorescence labeling for phospho-histone H3 (pHH3). This comparison showed that although Shh increased the mitotic rate of both genotypes, cKO CGNPs demonstrated significantly more mitoses (Fig. 4A,B). Thus, in vitro, Shh-induced proliferation was increased, rather than decreased, by Aspm deficiency.

To compare the proliferation rate of CGNPs in vivo, we counted mitoses along the entire length of the EGL in midline sagittal sections of P5 Aspm cKO mice and Aspm^f/f littermate controls (n=6 per genotype). We identified mitotic cells by co-immunofluorescence for pHH3, which marks prophase-early anaphase, and survivin (Birc5 – Mouse Genome Informatics), which localizes to the midbody region and cleavage furrow during anaphase-telophase (Caldas et al., 2005; Williams et al., 2011). To control for differences in EGL thickness between controls and cKOs, we normalized mitotic cell counts to EGL area to determine the mitotic rate across the entire EGL population. Using these criteria, we noted a significant increase in overall mitotic rates in the Aspm cKO EGL (Fig. 4C,D), consistent with our in vitro data and with previous observations in zebrafish aspm mutants (Kim et al., 2011; Novorol et al., 2013).

The spatial pattern of mitoses was disrupted by Aspm deficiency. Because all mitotic cells are expected to be PCNA/Ki67⁺, we used spatial criteria, rather than PCNA and p27 (as in Fig. 3) to define EGL layers. We considered the superficial EGL (EGL_S) to extend from the pial surface inwards to 50% of EGL depth, and the deep EGL (EGL_D) to extend from the midway point to the inner EGL margin. At the ages examined (P5-P7), this demarcation approximates but does not exactly correspond with the border of the oEGL, defined by expression of PCNA or Ki67 (Mki67 – Mouse Genome Informatics), both of which extend over half the thickness of the EGL (Fig. S3F,G).

We measured the position of all pHH3⁺ cells in the EGL at P5 relative to the pia mater (Fig. S3A,B). Mitotic figures tended to occur further from the pia in cKOs compared with Aspm^f/f littermates (Fig. S3C; n>3000 cells/genotype). In controls, ∼80% of mitoses occurred in the EGL_S, consistent with previous observations (79.6%) in rat EGL (Zagon and McLaughlin, 1987). By contrast, in cKOs, the proportion of mitoses occurring in the EGL_S was reduced to ∼67% (Fig. 4E). Normalizing cell counts to either EGL area (Fig. 4F) or length (Fig. S3D,E), we found similar mitotic frequency in the EGL_S of both genotypes, but significantly increased mitotic frequency in the EGL_D in cKOs. Co-labeling with pHH3⁺ and either Ki67 or PCNA to mark the proliferative oEGL, showed that >90% of mitoses occurred within the Ki67/PCNA⁺ domains, with no statistically significant difference between genotypes (Fig. S3F-H). Importantly, all mitotic cells outside of these domains (Fig. S3F,G, arrowheads) were PCNA⁺ or Ki67⁺ as expected. Thus, spatial measurement revealed differences in the position of mitoses that could not be determined by analysis of marker expression. In the independent cohort used for co-labeling with Ki67 or PCNA, we again observed a significant increase in pHH3⁺ figures in the cKO compared with controls (Fig. S3I), predominantly localized to the EGL_D (Fig. 4F; Fig. S3J). As shown in Fig. 4E (arrows), the greatest increase in mitotic figures in cKOs was observed in the 50th-70th percentile of EGL thickness near the EGL_S/EGL_D boundary.

Because Asp/Aspm have been reported to play a role in mitotic spindle assembly and function (Wakefield et al., 2001; Riparbelli et al., 2002; Paramasivam et al., 2007), we used chromosome morphology and pHH3/survivin co-immunofluorescence to determine the distribution of mitotic stages found in wild-type and Aspm cKO CGNPs (Fig. 4G). We noted a decrease in the proportion of prophase cells and an increase in the proportion of prometaphase cells in cKOs relative to Aspm^f/f controls (Fig. 4H). This altered distribution at early mitosis suggests that Aspm loss impairs mitotic progression at the time the spindle is forming. We also noted that Aspm-deficient CGNPs displayed abnormal rounded DNA during anaphase, suggestive of defects in organizing the spindle poles, whereas Aspm^f/f control CGNPs showed a more typical bar shape (Fig. 4I,J). The median distance between anaphase chromosome pairs was also significantly shorter in Aspm cKOs compared with wild-type controls (Fig. 4K; Mann–Whitney test, P=0.0157). Moreover, whereas wild-type cells showed a Gaussian distribution of inter-chromosome distances (Shapiro–Wilk normality test, P=0.6824), this distribution was not maintained in cKOs (P=0.0241). A disproportionately high percentage of Aspm cKO chromosome pairs were separated by 7.5-8.0 µm (Fig. 4L; n=21/46 cells examined compared with 9/37 controls). This altered distribution suggests a delay in the transition between anaphase-A, during which chromosome separation is driven by destruction of sister-chromatid adhesion complexes, and anaphase-B, during which the mitotic spindle is elongated as the spindle poles pull the chromosomes apart. Overall, these data support a role for Aspm in promoting mitotic spindle assembly and progression through metaphase into anaphase.

Aspm (Asp) has been implicated in the maintenance of cleavage plane orientation in Drosophila neural progenitors (Rujano et al., 2013) though an analogous role for Aspm in mouse is controversial (Fish et al., 2006; Pulvers et al., 2010). In the developing cerebellum, CGNPs adopt different cleavage plane orientations depending on their position within the EGL (Zagon and McLaughlin, 1987). In the EGL_S, CGNPs more frequently orient their spindle perpendicular to the pial surface, whereas CGNPs dividing in the EGL_D more often divide either parallel or transverse to the pial surface (Zagon and McLaughlin, 1987). Conditional deletion of Aspm produced significant changes in the orientation and distribution of divisions within the EGL. The incidence of transverse divisions in Aspm cKO CGNPs was increased at the expense of perpendicular divisions (Fig. 5A,B; χ² test, P<0.0001). Analysis of division orientation relative to EGL depth demonstrated a bias in control cerebella towards perpendicular divisions in the EGL_S (56.1%, n=301), with roughly equal frequencies of perpendicular, transverse and parallel divisions in the EGL_D (n=149), consistent with prior studies (Zagon and McLaughlin, 1987). This orientation bias was lost in Aspm cKOs, in which CGNP division orientation appeared randomized (Fig. 5C). These data provide the first evidence that division orientation in the cerebellum is under molecular control, regulated at least in part by Aspm.

To determine whether alterations in division orientation and mitotic progression caused by Aspm mutation affected the tendency of CGNPs to re-enter or to exit the cell cycle (e.g. self-renew or differentiate), we examined the fate of CGNPs 24 h or 48 h after injection of 5-bromo-2′-deoxyuridine (BrdU) at P5. We used co-labeling with BrdU and the differentiation marker p27 to determine the proportions of BrdU-labeled cells that either continued or exited the cell cycle (Fig. 5D-K). CGNPs in the EGL were labeled with BrdU with similar frequency at the 24 h time point in Aspm^f/f control and cKOs (Fig. 5L), suggesting that CGNPs in both genotypes enter S phase with similar frequency. However, there was a tendency towards fewer actively cycling (p27⁻) and more differentiating (p27⁺) BrdU⁺ CGNPs in the cKOs compared with controls (Fig. 5L). Further analysis showed that a greater proportion of p27⁺ CGNPs in the iEGL were labeled with BrdU in cKOs compared with controls (Fig. 5M). We quantified the cell cycle exit fraction (proportion of BrdU⁺ CGNPs in the EGL that were p27⁺) for both genotypes and found that a greater proportion of BrdU⁺ cells had exited the cell cycle in Aspm cKOs at both the 24 h and 48 h time points (Fig. 5N). In addition, the differentiation index, a measure of the number of p27⁺ cells produced per p27^– CGNP, was significantly higher in Aspm cKOs (Fig. 5O). Finally, we noted a consistent decrease in terminally differentiated p27⁺/BrdU⁺ GNs in the IGL in Aspm cKOs, and an increase in p27⁺/BrdU⁺ CGNPs in the EGL at 48 h (Fig. 5P). Together, these data demonstrate that Aspm cKO CGNPs prematurely exit the cell cycle but fail to populate the IGL as effectively as normal CGNPs.

The overproduction of p27⁺ CGNPs in the EGL coupled with underproduction of mature GNs in the IGL suggests that Aspm cKO CGNPs exhibit either delayed migration or impaired survival. We found no evidence of migration defects as similar numbers of BrdU⁺/p27⁺ neurons were found in the molecular layer in both genotypes at both time points (Fig. 5P). Therefore, we analyzed cell death in the EGL of cKOs and controls by immunohistochemistry for cleaved caspase-3 (cC3) and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining. Apoptosis is rarely detected in the EGL of wild-type mice (Garcia et al., 2012). By contrast, cC3⁺ cells were significantly increased in the Aspm^SA/SA and Aspm cKO EGL (Fig. 6A-C) and were proportional to proliferation, i.e. highest at P7 and lower at P11. Similar results were obtained with TUNEL (Fig. 6D). Although apoptotic cells were rare in controls, they were most frequently observed in the EGL_S (86.7±1.2%), whereas in Aspm cKOs the density of cC3⁺ or TUNEL⁺ apoptotic cells in the EGL_D was increased ∼5- to 7-fold over controls (Fig. 6E,F). Thus, apoptosis increased in Aspm cKOs in both the proliferating and differentiating populations of CGNPs, with the largest relative increase in the differentiating population. This increase in apoptosis specifically in the EGL_D could explain why fewer mature GNs are produced in Aspm cKOs despite the increased production of p27⁺ CGNPs.

CGNPs are highly sensitive to apoptosis in response to genotoxic stress (Chong et al., 2000; Lee et al., 2012). The mitotic abnormalities in Aspm-deficient CGNPs, and prior studies demonstrating a role for Aspm in mitotic spindle dynamics (Zhong et al., 2005; Fish et al., 2006; Higgins et al., 2010) and DNA repair (Kato et al., 2011) suggest that increased CGNP apoptosis in Aspm mutants might be caused by increased DNA damage. Immunolabeling for the DNA damage marker γH2AX demonstrated small, relatively faint γH2AX⁺ foci in the EGL of all genotypes, and larger foci that were more intense and less frequent. Co-labeling with γH2AX and BrdU, 2 h after BrdU injection, demonstrated that DNA damage was enriched in CGNPs undergoing DNA replication (Fig. 6G). In both Aspm^SA/SA and Aspm cKO mice, γH2AX⁺ CGNPs were more frequent compared with Aspm-intact controls (Fig. 6H-J). These findings implicate DNA damage as a potential mechanism of increased progenitor apoptosis in Aspm-deficient mice.

To determine the contribution of genotoxic stress and progenitor death to growth restriction in Aspm-deficient mice, we investigated whether genetic deletions that interrupt the link between DNA damage and apoptosis could rescue the growth of Aspm-deficient cerebella. In neural progenitors, DNA damage activates p53, which triggers Bax-dependent apoptosis (Vousden and Lu, 2002; Fridman and Lowe, 2003; Lee and McKinnon, 2007; Crowther et al., 2013). We examined how co-deletion of either Bax or Trp53 affected the Aspm cKO phenotype. Because p53 may exert gene dosage-related effects, we selected Math1-Cre;Aspm^f/f;Trp53^+/+ (Aspm cKO) and Math1-Cre;Aspm^f/f;Trp53^f/f (Aspm;Trp53 dKO) littermates for comparison. For Bax studies, we compared littermates with genotypes Math1-Cre;Aspm^f/f;Bax^f/f (Aspm;Bax dKO), Math1-Cre;Aspm^f/+;Bax^f/+ and Aspm^f/f;Bax^f/f without Cre. Double mutants and controls were harvested at P30 for comparison of cerebellar weight, or at P7 for comparison of mitosis, DNA damage and apoptosis.

Compared with deletion of Aspm alone, cerebellum weight was significantly increased by co-deletion of Bax or Trp53 (Fig. 7A-D). Volumetric MRI confirmed that Aspm;Bax dKO cerebella did not exhibit the growth restriction observed in Aspm single mutants, and were similar in volume to Aspm-intact controls (Fig. 7E). Histological examination, however, showed focal areas of thinning in the IGL in each of the three Aspm;Bax dKO cerebella examined (Fig. 7A,B), though quantification of mean cell density over a broad area of the IGL in replicate sections did not show a statistically significant difference (Fig. 7F). These focal areas of IGL thinning were not observed in Aspm cKOs, (data not shown) or in our prior studies of Bax single mutants (Garcia et al., 2012). As in Aspm single mutants, we noted an increase in mitotic rate across the EGL in each of the double mutants (Fig. 7G). Like Aspm single mutants, Aspm;Bax and Aspm;Trp53 mutants showed an elevated percentage of cells at prometaphase and a reduction in prophase (Fig. 7H), demonstrating impaired cell cycle progression. These abnormalities might underlie the failure of Bax co-deletion to completely normalize the cerebellum.

At P7, in both Aspm;Bax and Aspm;Trp53 dKO mice, apoptosis was significantly reduced compared with Aspm cKOs (Fig. 7I,J), although DNA damage persisted (Fig. 7K,L). Thus, loss of either Bax or Trp53 led to increased tolerance of DNA damage and reduced cell death, significantly restoring cerebellar growth. The lack of cell death in Aspm;Bax dKO mice confirms that the cell death induced by Aspm deletion is apoptotic. Focal hypoplasia within the IGL of Aspm;Bax dKOs, however, demonstrates that blocking apoptosis did not completely normalize the phenotype of Aspm mutation.

Because Aspm is upregulated in medulloblastoma in both ND2:SmoA1 mice (Fig. 1I) and patient-derived samples (Vulcani-Freitas et al., 2011), we examined whether Aspm function is essential for tumor growth. We bred Aspm^f/f mice into the rapidly tumorigenic medulloblastoma model, hGFAP-Cre;SmoM2 (G-Smo). G-Smo mice express a constitutively active allele of the Shh receptor component Smo (Mao et al., 2006) in stem cells that give rise to the neurons and glia of the cerebrum and cerebellum excluding the Purkinje cells (Zhuo et al., 2001; Kuang et al., 2012), and develop medulloblastoma with 100% incidence, dying of tumor progression by P20 (Schüller et al., 2008). We investigated the effect of Aspm deletion on tumor growth by comparing G-Smo mice with Aspm^f/f (G-Smo;Aspm^KO) and Aspm^f/+ (G-Smo;Aspm^het) genotypes.

G-Smo;Aspm^KO medulloblastomas grew significantly more slowly than G-Smo;Aspm^het tumors, although the rate of tumor formation was 100% in both genotypes. Tumors could be differentiated from normal brain in Hematoxylin & Eosin (H&E)-stained sections by their characteristic histology (supplemental Materials and Methods). To compare tumor size, we measured tumor cross-sectional area, normalized to the cross-sectional area of non-tumor cerebellum. At P3, tumor size appeared to be similar in G-Smo;Aspm^KO (n=4) or G-Smo;Aspm^het (n=3) mice, and none of the mice showed hydrocephalus (Fig. 8A,B). At P13, however, tumors in G-Smo;Aspm^het mice had expanded, producing obstructive hydrocephalus, whereas tumors in G-Smo;Aspm^KO mice (n=4 both genotypes) were significantly smaller and did not develop hydrocephalus (Fig. 8C-E). Within tumors of both genotypes, γH2AX⁺ cells were too numerous and densely packed to be counted. Using western blot, however, we consistently found significantly increased γH2AX⁺ cells in Aspm-deleted tumors at P3 (Fig. 8F). By P13, γH2AX⁺ abundance was equivalent (data not shown), suggesting that tumors had developed a mechanism to compensate for Aspm deletion.

Despite decreased tumor growth and absence of hydrocephalus, cerebellar anatomy and function was severely disrupted in G-Smo;Aspm^KO mice. These mice demonstrated severe motor deficits and could not transition from nursing to autonomous feeding, failed to maintain weight on weaning, and required euthanasia to prevent animal suffering. Although G-Smo;Aspm^KO mice did not seem to die from tumor expansion, because of the motor deficits, G-Smo;Aspm^KO mice did not survive longer than G-Smo;Aspm^het controls (Fig. 8G). Although we could not detect a survival benefit in this model, Aspm deletion effectively impaired tumor growth.

Our studies demonstrate that Aspm plays an essential role in postnatal neurogenesis, and that this role is co-opted in malignant proliferation in medulloblastoma. Aspm is integral to the proliferative program induced in CGNPs by Shh signaling and remains upregulated in Shh-driven medulloblastomas. Conditional deletion of Aspm in CGNPs is sufficient to impair cerebellar growth. This growth impairment cannot be attributed to decreased proliferation, as the frequency of CGNP mitosis is actually increased in Aspm cKOs. Rather, we show that growth impairment results from a combination of mitotic defects, precocious differentiation, increased DNA damage and increased apoptosis. The increase in proliferation rates in Aspm mutants might represent compensatory proliferation, a phenomenon that has been recently described in Drosophila centrosomal mutants, which also show spindle assembly defects, elevated DNA damage and increased apoptosis (Poulton et al., 2014).

Apoptosis of Aspm-deleted CGNPs was prevented by co-deletion of Bax or Trp53, which significantly but incompletely restored cerebellar growth. The normalization of cerebellar weight and size in mice with co-deletion of Aspm and either Bax or Trp53 demonstrates that preventing apoptosis can largely rescue Aspm-dependent growth restriction. The persistent focal abnormalities in the IGL of Aspm;Bax dKO mice, however, show that apoptosis is not the only process determining growth failure.

Along with increased DNA damage and apoptosis, we note impaired mitotic progression, changes in the radial position and orientation of mitotic cells, and abnormal patterns of differentiation in Aspm cKOs. This raises the intriguing question of whether the heterogeneity in division orientation we and others (Zagon and McLaughlin, 1987) observe in the EGL reflects different modes of division. Recent work has shown that both self-renewing CGNPs and differentiating CGNPs predominantly divide symmetrically (Espinosa and Luo, 2008; Nakashima et al., 2015). Division orientation might influence whether symmetric divisions result in self-renewal or differentiation, such that perpendicular divisions, which are most common in the EGL_D, may promote continued proliferation, whereas transverse orientations, which are elevated in the Aspm cKO may promote cell cycle exit.

Human and mouse studies have shown that diverse mutations including Nde1, Cdk5rap2 and Magoh induce microcephaly through a combination of loss of progenitor self-renewal, premature differentiation, and progenitor apoptosis (Pawlisz et al., 2008; Lizarraga et al., 2010; Silver et al., 2010; Houlihan and Feng, 2014). Mutations of both Nde1 and Cenpj have been shown to induce microcephaly through p53-dependent apoptosis (Bazzi and Anderson, 2014; Houlihan and Feng, 2014). In all of these examples, the phenotype produced in mice is severe brain malformation. In both our studies and prior investigations, however, Aspm deletion produced a phenotype in mice that is much less severe than the orthologous phenotype in humans (Pulvers et al., 2010). Of interest, the microcephaly genes ASPM and CDK5RAP2 have been found to undergo positive selection in primates that correlates with increased brain size (Montgomery and Mundy, 2014); thus, Aspm appears to be particularly important in large-brained animals. We propose that Aspm loss impedes brain growth by impairing transient-amplifying cells in a manner that is proportional to the requirement for amplification. Accordingly, Aspm deletion is more deleterious in humans, in which brain size is larger and the period of neurogenesis is more prolonged. We have previously shown that a very low rate of apoptosis can exert a significant impact on cerebellar growth, detectable only by comparing growth in apoptosis-deficient, Bax-deleted mice (Garcia et al., 2012). A small increase in the rate of progenitor apoptosis may thus produce a significant reduction in brain growth that becomes more pronounced as the proliferation period is extended.

The restriction of tumor growth in Aspm-deleted medulloblastoma is consistent with our model in which the effect of Aspm loss is proportional to the extent of proliferation. In both cerebellar development and medulloblastoma formation, Aspm deletion increased DNA damage. In tumors, however, where proliferation is more extensive than in normal tissues, DNA damage and growth restriction with Aspm loss were most pronounced. Importantly, previous investigators have demonstrated that Aspm knockdown impairs DNA repair after ionizing radiation in vitro (Kato et al., 2011). Our findings that medulloblastomas require Aspm for typical growth suggest that Aspm could be therapeutically targeted to increase the efficacy of radiation and chemotherapy for medulloblastoma. Primary mouse medulloblastoma with Aspm deletion provides an ideal model in which to test this possibility in preclinical studies.

Aspm-GFP reporter mice [Tg(Aspm-EGFP)IH113Gsat/Mmucd] and Aspm-CreER mice are previously described (Gong et al., 2010; Marinaro et al., 2011). For lineage tracing, Aspm-CreER mice were crossed with Rosa26-Lox-STOP-Lox-TdTomato (Madisen et al., 2010) reporter mice (strain#7914; Jackson Laboratories). Tamoxifen (275 µg in 25 µl sunflower oil) was injected intraperitoneally into Aspm-CreER;tdTomato mice and TdTomato littermate controls lacking Aspm-CreER, at the indicated ages.

Aspm^SA/SA mice were generated from Aspm-targeted embryonic stem cells (line EPD0320_1_A06, Aspm^Gt(KOMP)Wtsi; KOMP Repository, Davis, CA, USA). The targeting vector is described in Fig. S2. Generation of Aspm^SA/SA, Aspm cKO and hGFAP-Cre;SmoM2;Aspm^f/f mice are detailed in supplementary Materials and Methods.

Medulloblastoma-prone NeuroD2:SmoA1 mice are previously described (Hallahan et al., 2004). Bax floxed mice were derived from the Bax^f/f;Bak^−/− strain (Takeuchi et al., 2005) crossed with C57BL/6 to recover wild-type Bak. Conditional p53 (Jonkers et al., 2001) were provided by the NCI (strain#01XC2, Frederick, MD, USA). The SmoM2, SmoA1, Bax^f/f, Trp53^f/f and all Cre-expressing lines were crossed through at least four generations into a C57BL/6 genetic background. All animal handling and protocols were carried out in accordance with established NIH practices and approved under UNC IACUC# 10-126.

CGNPs were isolated and explanted as previously described (Kenney et al., 2003) and maintained in 0.5 µg/ml Shh (R&D Systems) or vehicle (0.5% BSA in PBS).

Total RNA was isolated from cerebella or explanted CGNPs using the RNeasy kit (Qiagen), and cDNA synthesized from 1 µg total RNA using Superscript III (Invitrogen/Life Technologies). Gene expression was quantified on an ABI PRISM 7500 Sequence Detection System, using the ΔΔC_T relative quantification method. All experiments included no template controls and were performed in triplicate and repeated twice independently. Primer sequences are listed in supplementary Materials and Methods.

Mice were injected intraperitoneally at the indicated ages with 50 µl Hanks’ Balanced Salt Solution containing EdU or BrdU (250 µM, Invitrogen). After the indicated interval, brains were dissected and fixed in 4% formaldehyde in PBS for 24 h at 4°C, then processed for histology.

Mouse brains were processed and in situ hybridization and immunohistochemistry were performed as previously described (Gershon et al., 2009, 2013; Garcia et al., 2012), using the Aspm probe (Advanced Cell Diagnostics, #318711) or primary antibodies listed in supplementary Materials and Methods. Methods for quantitative analysis of mitosis, spindle orientation, BrdU/p27, cC3, γH2AX, tumor size and volumetric MRI are detailed in supplementary Materials and Methods. ITK-Snap software (www.itksnap.org) was used to isolate the cerebellar region in MR images (Yushkevich et al., 2006). TUNEL was performed using the Click-iT TUNEL Kit (Life Technologies, #C10245).

Whole cerebella were processed as previously described (Garcia et al., 2012) using primary antibodies to γH2AX and β-actin (Cell Signaling, #4970).

We thank Luca Muzio (San Raffaele Scientific Institute, Milan, Italy) for Aspm-CreER mice, GENSAT for Aspm-GFP mice, David Rowitch, (UCSF, CA, USA) and Robert Wechsler-Reya (Sanford-Burnham Medical Research Institute, La Jolla, CA, USA) for Math1-Cre mice, Eva Anton (UNC, Chapel Hill, NC, USA) for hGFAP-Cre mice, and James Olson (Fred Hutchinson Cancer Research Center, Seattle, WA, USA) for NeuroD2:SmoA1 mice. We thank the UNC Small Animal Imaging and Tissue Pathology Laboratory Core facilities for providing the MR imaging and immunohistochemistry.