Cdk5rap2 regulates centrosome function and chromosome segregation in neuronal progenitors

The cerebral cortex that provides humans with our unique cognitive abilities consists of six morphologically and functionally distinct neuronal layers (Rakic, 2009). All cortical neurons are produced from a neuroepithelium lining the ventricular cavity of the developing brain (Noctor et al., 2001). During neurogenesis, postmitotic neurons differentiate and migrate away from the ventricular zone (VZ) towards their final destination in the cerebral cortex (Rakic, 2009). Earlier born neurons localize to deeper layers of the cortex, whereas later born neurons populate progressively more superficial layers (Angevine and Sidman, 1961; Rakic, 1974). Hence, proper development of the cerebral cortex requires the coordination of cell proliferation, fate determination and migration. Defects in these processes can cause neurodevelopmental disorders such as microcephaly, a disease associated with mental retardation (Mochida, 2008). Primary microcephaly is defined as decreased head circumference and a disproportionately reduced cerebral cortex, but no other major developmental malformations (Woods et al., 2005).

The centrosome plays key roles in neuronal progenitor proliferation and fate determination, as well as neuronal differentiation and migration (Higginbotham and Gleeson, 2007). Recent work suggests that asymmetric inheritance of daughter and mother centrosomes regulates cell fate determination in cortical progenitors (Wang et al., 2009). Moreover, mutations in several genes that encode centrosomal proteins cause human primary microcephaly (Woods et al., 2005). Among them is CDK5RAP2 (cyclin-dependent kinase 5 related activator protein 2 or CDK5 regulatory subunit-associated protein 2) (Bond et al., 2005), which was originally identified in mammals by its interaction with p35^Nck5a (p35; Cdk5r1), an activator of the neuronal cyclin-dependent kinase 5 (CDK5) (Ching et al., 2000). CDK5RAP2 encodes a 1893 amino acid, 215 kDa centrosomal protein. In somatic cells, CDK5RAP2 promotes centrosomal cohesion (Graser et al., 2007) and recruits the γ-tubulin ring complex (γ-TuRC) – the microtubule nucleator – to the centrosome (Fong et al., 2008). However, the role of CDK5RAP2 in neuronal development remains poorly understood.

Hertwig's anemia (an) arose in the progeny of a heavily irradiated male mouse (Hertwig, 1942). Homozygous mutant animals have a macrocytic, hypoproliferative anemia and leukopenia attributable to intrinsic defects that are progressively more severe in the more mature hematopoietic progenitors (Barker and Bernstein, 1983; Barker et al., 2005; Eppig and Barker, 1989). Male animals are infertile due to severe deficiencies in germ cells, whereas female animals have less severe defects in oocytes, but cannot deliver pups (Russell et al., 1985). There is a high level of spontaneous aneuploidy in primary cultures of hematopoietic fetal liver, bone marrow and kidney epithelial cells derived from mutant animals (Eppig and Barker, 1984). Both the severity of the hematological disease and a coexisting predisposition to hematopoietic tumors are highly dependent upon the genetic background of the animal (Barker et al., 2005). Here, we show by positional cloning that an is due to a mutation in Cdk5rap2, and that an mutants show additional severe neurological defects that were not previously recognized.

Mutations in human CDK5RAP2 cause severely reduced brain size with relatively well-preserved cortical patterning (Bond et al., 2005), but without anemia or tumor predisposition, the hallmarks of an. Although CDK5RAP2 is one of several centrosomal proteins implicated in human microcephaly (Bond et al., 2002; Bond et al., 2005), we present the first analysis of any of these centrosomal proteins in mutant mice. Our analysis reveals important roles for Cdk5rap2 in mitotic progression and mitotic spindle orientation. Our data also suggest that Cdk5rap2 is essential for cortical progenitor proliferation and survival. These results illustrate the fundamental role of centrosome function in disease mechanisms associated with human primary microcephaly.

We maintained the Hertwig's anemia (an) mutation on the WB/ReJ (WB) and C57BL/6J (B6) inbred backgrounds, as homozygous mutant mice are poorly viable, particularly on the C57BL/6J background. WB.Cg an/+ × B6.Cg an/+ F₁ (WBB6F1) animals are nearly fully viable and were used for studies in adults. To identify the gene underlying the an phenotype, we performed an intersubspecific intercross between B6.Cg an/+ and the +/+ Mus musculus castaneus (CAST/Ei) inbred strain to generate second filial generation (F₂) intercross animals segregating the an allele. F₂ animals were phenotyped by complete blood count at 3 weeks of age. We reconfirmed linkage to the brown locus on chromosome 4 in 46 consecutive F₂ animals at 3 weeks of age, and further delineated the genetic interval by genotyping an additional 734 F₂ animals and correlating the phenotype with crossover events within the originally defined locus. Genotyping and breeding of additional novel crossover events in the B6.Cg an/+ colony further narrowed the critical region.

Subsequent to mutation identification, animals were genotyped using standard PCR reactions specific for the wild-type (1040 bp, primers 5′-TCACTGAGCTGAAGAAGGAGAA-3′ and 5′-GCAATCACTAAAATGTCCGATT-3′) and mutant (507 bp, primers 5′-GCAATCACTAAAATGTCCGATT-3′ and 5′-TGTCTTTCTGCCCTGACAGT-3′) alleles.

Total RNA from Cdkd5rap2^an/an and Cdk5rap2^+/+ tissue was extracted and purified as described (Baelde et al., 2001). A ³²P-labeled Cdk5rap2 probe (500 bp) was generated by PCR amplification of exon 24 from mouse genomic DNA (129S6.Tac) with primers 5′-GCCTTATTACCAGCATGCAA-3′ and 5′-TCACCGAAAAGTTCCAAGTTC-3′. A commercial mouse multi-tissue northern blot (Origene) and the Cdk5rap2^an/an and Cdk5rap2^+/+ blot were run, transferred, hybridized and processed as previously described (Lee et al., 2007).

Cdk5rap2^+/+ and Cdk5rap2^an/an tissues were homogenized and protein extracted in RIPA buffer (50 mM Tris-HCl pH 7.2, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% Nonidet P40) with Complete Protease Inhibitors (Roche). Protein (30 μg per sample) was loaded onto a 7% Novex Tris-acetate gel (Invitrogen) and run according to the manufacturer's instructions. Gels were transferred to nitrocellulose membranes using a semi-dry transfer system (Hoeffer) and analyzed with rabbit anti-Cdk5rap2 antibody at 1:1500 (Bethyl Laboratories).

Embryonic and postnatal day 0 (P0) brains were dissected in ice-cold 1×PBS, fixed in 4% paraformaldehyde by immersion or cardiac perfusion, embedded in paraffin and sectioned at 5 μm. Immunohistochemistry was performed as previously described (Feng and Walsh, 2004). Animals were handled according to protocols approved by the IACUC of the Beth Israel Deaconess Medical Center and the Children's Hospital Boston.

The following antisera were used at the specified dilutions: rabbit anti-Cdk5rap2 (1:1000) (Bethyl laboratories, BL2320); rabbit anti-Ki67 (1:500) (Vector Labs, VP-RM04); rabbit anti-Dcx (1:500) (Covance); goat anti-Cux1 (1:10) (Santa Cruz, SC6327); rat anti-Ctip2 (1:500) (Abcam, ab18465); mouse anti-aurora kinase A (1:200) (Abcam, ab13824); rabbit anti-Tbr2 (1:300) (Abcam, ab23345); rabbit anti-phospho-histone H3 (1:500) (Upstate, 07-492); rabbit anti-Sox2 (1:200) (Novus-Biologicals, NB110-372335C); rat anti-BrdU (1:300) (abdSerotec, MCA2060); rabbit anti-Tbr1 (1:500) (gift from R. Hevner, University of Washington, Seattle, WA, USA); goat anti-Brn1 (1:200) (Santa Cruz, SC-6028); mouse anti-aurora kinase B (1:200) (BD biosciences, 611083); rabbit anti-Par3 (1:300) (Upstate, 07-330); and rabbit anti-activated cleaved caspase 3 (1:50) (Cell Signaling, 9661). Secondary detection reagents (1:500) included Alexa Fluor 488-, 594- or 546-conjugated anti-rabbit, mouse or goat IgG (Invitrogen); Cy5-conjugated anti-rabbit, mouse or goat IgG, Cy3-conjugated or unconjugated Fab fragments of anti-rabbit IgG (Jackson ImmunoLaboratories); and Cy3-conjugated Streptavidin reagent.

For analysis of S-phase cells, E10.5 and E14.5 pregnant females were injected intra-peritoneally with 70 μg BrdU/g body weight and sacrificed 30 minutes after injection. The percentage of cells in S phase was calculated by counting cells stained with BrdU as a function of total nuclei. For analysis of cell cycle exit, E14.5 pregnant females were injected with BrdU as above, sacrificed 24 hours after injection, and analyzed as previously described (Chenn and Walsh, 2002).

Images were acquired and analyzed as previously described (Sepp et al., 2008). For spindle orientation analysis, images were acquired as 5 μm z-stacks using a Zeiss confocal microscope and angles were measured in three-dimensional (3D) reconstructed projections using ImageJ (NIH). At least three coronal brain serial sections from at least three different animals were analyzed per genotype per experiment. Unless otherwise noted, brain sections were chosen at the mid-anterior level of the brain, and micrographs were taken of the dorsal and dorsal-lateral cortex region. Cell counts were performed as a percentage of total nuclei per high-magnification field per coronal section. P-values were calculated with an ANOVA single-factor test.

We mapped Hertwig's anemia (an) to a 290 kb genomic interval between 68.91 and 69.94 Mb (genomic assembly NCBI m35, Dec 2005; www.ensembl.org/mus_musculus) on mouse chromosome 4, which includes the 5′ end of Cdk5rap2 and the multiple epidermal growth factor-like domains 9 (Megf9) gene. Sequencing of the entire coding regions of Megf9 and Cdk5rap2 revealed no exonic or splice junction mutations. However, RT-PCR of Cdk5rap2 from WBB6F1 an/an animals demonstrated an in-frame 111 bp deletion resulting from loss of exon 4 from the mRNA (see Fig. S1A in the supplementary material). This change predicts a 37 amino acid deletion of Cdk5rap2, but as the mutant mRNA maintains the reading frame a shortened protein product could be formed (see Fig. S1B in the supplementary material). Northern analysis demonstrated that exon 4 was uniformly omitted from the mutant mRNA (Fig. 1A). Amplification of genomic DNA indicated that the basis of this exon skipping was an inversion of exon 4 (Fig. 1B). Furthermore, B6.Cg an/+ females bred to male animals chimeric for a Cdk5rap2 gene-trap allele yielded F₁ compound heterozygous animals with leukopenia and macrocytic anemia, similar to the an mutants (see Table S1 in the supplementary material), confirming Cdk5rap2 as the Hertwig's anemia gene and allowing the designation of the mutation as Cdk5rap2^an.

Cdk5rap2^an/an animals had defects in multiple organs, including brain, thymus and testis. Using a probe to exon 24, we found that the predominant Cdk5rap2 transcript differs from tissue to tissue, with a 5.8 kb major transcript in testis and thymus, and a 2.3 kb transcript most abundant in heart and most other tissues (Fig. 1C). A unique 1.4 kb transcript was also present in testis. Although the mRNA was undetectable in adult brain using a probe to exon 24, the GENSAT in situ hybridization reference database (www.ncbi.nlm.nih.gov/projects/gensat/) and published work (Bond et al., 2005) show that Cdk5rap2 is highly expressed in embryonic neural progenitors.

Immunoblotting of Cdk5rap2 protein in the mutant indicated that deletion of exon 4 does not substantially destabilize the protein. In most tissues, a large (∼230 kDa) species was seen to predominate. A smaller isoform (∼105 kDa) was present in testis and thymus and predominated in adult brain, but was barely detectable in fetal brain or other tissues (Fig. 1D and data not shown). This suggests that the 230 kDa isoform disrupted in Cdk5rap2^an/an animals is the predominant functional species in fetal brain. The major protein isoforms might correspond to the two largest of the three mRNAs. Furthermore, as in the northern analysis (Fig. 1C), there was an additional testis-specific protein isoform (∼70 kDa). All of these proteins isoforms are too large to permit the resolution of a shortened an mutant protein by western analysis.

The identification of Cdk5rap2 as the an gene prompted us to examine the brain of an mutants, which had not previously been evaluated. WBB6F1 hybrid Cdk5rap2^an/an adult animals showed moderate cortical hypoplasia visualized by shortening of the forebrain, which did not extend fully over the superior colliculus of the midbrain, as it does in normal animals (see Fig. S1C in the supplementary material). However, inbred B6.Cg Cdk5rap2^an/an animals showed a dramatic reduction in brain size and rarely survived beyond 1 week of age. They were runted, with disproportionately smaller brains both prenatally and at birth (Fig. 1E,F). The mutants exhibited a significant decrease in the ratio of brain to body weight as compared with controls (7.5±1.5 versus 9.5±0.9 mg/g, P<1×10^–4) (Fig. 1G). The brain phenotype had a variable expressivity, being more severe in some individuals than others (Fig. 1E,F), including variably increased ventricular size and variably decreased cortical thickness (Fig. 1H,I). Other brain structures, including the hippocampus, were also considerably smaller than normal (Fig. 1J). Together, these data demonstrate that Cdk5rap2 is essential for normal brain development.

The cortical neuronal layers are formed in an `inside-out' fashion, whereby the deepest layers form first and the most superficial layers form last (Rakic, 2009). In newborn (P0) Cdk5rap2^an/an animals, immunostaining for Tbr1 and Ctip2 (Bcl11b – Mouse Genome Informatics) (layers V-VI) (Hevner et al., 2001; Leid et al., 2004), Foxp1 (layers III-V) (Ferland et al., 2003), Brn1 (Pou3f3) and Cux1 (layers II-IV) (McEvilly et al., 2002; Nieto et al., 2004) showed profound defects in neurogenesis but relatively preserved cortical layering (Fig. 2A-E), suggesting that neuronal migration was generally normal. However, Cdk5rap2^an/an mice showed consistent thinning of the cerebral cortex (Fig. 2A and see Fig. S2A,B in the supplementary material), accompanied by preferential reduction of the superficial, later-born neuronal layers (Fig. 2D-F), as demonstrated by a significant decrease in Brn1-positive cells compared with controls (17.8±1.4% versus 28.6±2.43%, P<1×10^–9). The preferential decrease of superficial neurons in mutant animals was confirmed by Cux1 immunostaining (Fig. 2E). The dramatic decrease in superficial layers resulted in fewer total cells and was associated with correspondingly increased early-born, Tbr1-positive cells in mutants as compared with controls (43.4±2.2% versus 36.7±1.9%, P<0.001) (Fig. 2B,C,F). Thus, Cdk5rap2^an/an animals have fewer total neurons and the last-born superficial neurons are particularly reduced.

Our observations in newborn animals prompted us to investigate the effect of the Cdk5rap2^an/an mutation during embryonic development. We found a significant decrease in overall cortical thickness at E13.5 (Fig. 3A,B), indicating a defect at early stages of neurogenesis. Immunostaining for Dcx, which marks immature neurons (Gleeson et al., 1999), showed that E15.5 Cdk5rap2^an/an embryos had a thinner neuronal layer than controls, expressed as the ratio of the thickness of Dcx-immunolabeled neuronal layers over total cortical thickness (43.9±1.6% versus 58.3±2.4%, P<0.001) (Fig. 3C,D). The decrease in cortical neurons was associated with a reduction in the precursor population, as revealed by immunoreactivity for Ki67 (Mki67), a marker of dividing cells, which was diminished in E16.5 Cdk5rap2^an/an embryos compared with controls (141.1±9 versus 185±5, P<0.01) (Fig. 3E,F). In addition, short pulses of BrdU at E10.5 (when the developing cortex is primarily composed of neuronal progenitors) and at E14.5 (when most of the deeper layers have already formed), showed that Cdk5rap2^an/an embryos have a progressive decrease in the density of S-phase cells (see Fig. S3A-C in the supplementary material), suggesting that progenitor cells decrease as development proceeds.

Neuronal progenitors form two groups that undergo mitosis either at the apical surface of the ventricle or on the basal side away from the VZ (Gotz and Huttner, 2005), and both progenitor types were affected in Cdk5rap2^an/an mutants. Apical progenitors give rise to deep layer neurons and basal progenitors, whereas basal progenitors primarily produce superficial layer neurons (Pontious et al., 2008). Radial glial cells, a subset of apical progenitors, produce most CNS neurons and provide a scaffold for newly formed neurons to migrate (Misson et al., 1988). We immunostained E15.5 Cdk5rap2^an/an mice for Glast (Slc1a3 – Mouse Genome Informatics), a radial glial marker (Shibata et al., 1997), and found no apparent changes in their morphology (data not shown), although the decreased Ki67 and Sox2 staining (Fig. 3E-J) suggested that their number was reduced. In addition, E14.5 and E16.5 Cdk5rap2^an/an embryos showed a marked loss of basal progenitors upon immunostaining for the basal progenitor marker Tbr2 (Eomes – Mouse Genome Informatics) (Englund et al., 2005) and for Sox2, a marker for apical progenitors and a subset of basal progenitors (Bani-Yaghoub et al., 2006) (Fig. 3G-I). The ratio of Tbr2-positive cells to total progenitors was decreased in mutants at both E14.5 (28.9±1.8% versus 40.8±2.2%, P<5×10^–4) and E16.5 (34.5±1.2% versus 43.8±1.1%, P<1×10^–5) (Fig. 3I), demonstrating that the decrease in superficial layer neurons correlates with a reduction in basal progenitors. Interestingly, at mid-neurogenesis, a fraction of these Tbr2-positive cells normally localizes to the VZ, whereas in mutant animals fewer Tbr2-positive cells were located within the VZ. Whereas E15.5 control embryos had roughly an equal distribution of Tbr2-positive cells in the VZ (93.2±17.8) and subventricular zone (SVZ) (101.7±11.3), Cdk5rap2^an/an embryos had 59.8% fewer Tbr2-positive cells within the VZ (38.4±5.9) compared with the SVZ (64.2±2.6) (Fig. 3J). This profound reduction of basal progenitors at later stages in neurogenesis might reflect a more generalized loss of all progenitors over the course of neurogenesis, although a more specific defect in basal progenitors cannot be ruled out.

Since Cdk5rap2 is a centrosomal protein, we hypothesized that the neuronal progenitor deficiency in Cdk5rap2^an/an animals reflects underlying mitotic defects. Using the mitotic (M)-phase marker phospho-histone H3, we analyzed E12.5 (see Fig. S4 in the supplementary material), E14.5 and E16.5 embryos (Fig. 4A,B). We found 1.8-fold more M-phase cells in mutants compared with controls at E16.5 (61.75±6.24 versus 35.26±4.8, P<1×10^–7). This increase in mitotic cells reflected increases at the luminal VZ (45.95±5.57 versus 28.57±4.23, P<1×10^–5) and outside the VZ (13.07±1.81 versus 5.78±1.27, P<1×10^–7) (Fig. 4C). Since the overall progenitor number decreases, this surprising increase in M-phase cells suggests that Cdk5rap2 deficiency leads to a delay in mitotic progression. To determine the cause of this delay, we examined aurora kinase A immunoreactivity (Katayama et al., 2003) in neuronal precursors. Aurora kinase A labels the mitotic spindle poles as two distinct dots at the focused ends of a bipolar spindle structure. Cdk5rap2^an/an embryos showed a 2.8-fold increase in pro-metaphase and metaphase precursor cells with mono-, tri- and other aneupolar spindle poles labeled with aurora kinase A, as compared with controls (12.95±3.26% versus 4.62±1.69%, P<1×10^–5) (Fig. 4D,E). These abnormal spindles confirm previous data indicating that atypical meioses are readily apparent in Cdk5rap2^an/an embryonic germ cells and suggests a mechanism for the predisposition to chromosomal aneuploidy seen in Cdk5rap2^an/an primary cells (Eppig and Barker, 1984).

During neurogenesis, the orientation of the mitotic division plane relative to the ventricular surface (VS) has been proposed to influence cell fate determination (Zhong and Chia, 2008). The disruption of proteins that regulate mitotic spindle orientation at the VZ correlates with premature and excessive generation of neurons and early depletion of cortical progenitors (Feng and Walsh, 2004; Fish et al., 2006; Sanada and Tsai, 2005), although the relevance of this mechanism to human microcephaly has been uncertain. To investigate mitotic orientation during neurogenesis in Cdk5rap2^an/an mice we focused on anaphase and telophase spindles, as metaphase spindles undergo dynamic rotations (Haydar et al., 2003). We defined cleavage plane orientations in relation to the VS as horizontal (0-30°, i.e. mitotic spindle aligned parallel to the VS), oblique (30-60°) or vertical (60-90°, i.e. mitotic spindle aligned along the apical-basal axis of the VS) (Chenn and McConnell, 1995). During cortical development, most cleavage planes are oriented perpendicular to the VS, producing side-by-side daughter cells (Zhong and Chia, 2008). By co-immunostaining for Par3 (Pard3 – Mouse Genome Informatics) to label the VS (Bultje et al., 2009) and aurora kinase B to label the central spindle (anaphase) and the midbody (telophase) (Vader and Lens, 2008), we found increased horizontal cleavage planes in Cdk5rap2^an/an mice compared with controls at E14.5 (15.6±3.2% versus 7.7±4.0%, P<0.05) (see Fig. S5 in the supplementary material). E16.5 Cdk5rap2^an/an embryos had increased horizontal cleavage planes (21.2±4.5% versus 10.0±2.6%, P<0.01) and reduced vertical cleavage planes (37.2±7.1% versus 56.1±5.8 %, P<0.01) (Fig. 4F). Thus, Cdk5rap2 is important for mitotic spindle orientation in cortical progenitors.

Mutation of Nde1, which encodes another spindle pole protein, causes premature cell cycle exit of neuronal progenitors, depleting the progenitor pool prematurely and reducing brain size (Feng and Walsh, 2004). To investigate whether similar processes cause progenitor cell deficits in Cdk5rap2^an/an animals, we labeled progenitors with BrdU at E14.5 and examined the fraction remaining in the cell cycle at E15.5 by double immunostaining for BrdU and the proliferation marker Ki67 (Chenn and Walsh, 2002). The `exit fraction' (Caviness et al., 2003), representing the ratio of BrdU-positive Ki67-negative cells over total BrdU-positive cells, was increased by 25% in mutants compared with controls (P<1×10^–4) (Fig. 5A,B), and this may contribute to the premature decrease in progenitors seen in Cdk5rap2^an/an embryos.

Defects in spindle orientation also correlate with increased cell death during neuroepithelial stem cell divisions (Yingling et al., 2008). We analyzed cells undergoing apoptosis by immunostaining for activated cleaved caspase 3 (caspase 3). Prior to the onset of neurogenesis (E10.5) we found 4.5-fold more apoptotic cells in mutants compared with control littermates (12.7±4.4% versus 2.8±1.6%, P<1×10^–5), suggesting that these increased dying cells represent apical progenitors, the predominant cell type at this stage. At E12.5 we found a 6.8-fold increase in cell death in Cdk5rap2^an/an animals compared with controls (16.3±3.1% versus 2.4±1.5%, P<1×10^–9) (Fig. 5C-E). To investigate whether a particular cell type was targeted for cell death, we first analyzed the distribution of apoptotic cells in apical (alongside the ventricle), basal (away from the ventricle) and cortical plate (CP) regions at E12.5 (Fig. 5F,G). Cdk5rap2^an/an animals had more dying cells than controls in both apical (12.3±3.1 versus 1.6±0.7, P<1×10^–5) and basal (40.5±10.9 versus 2.8±1.1, P<1×10^–5) regions, as well as in the CP (20.4±1.0 versus 3.8±2.4, P<1×10^–5) (Fig. 5G). Double immunostaining for Tuj1 (Tubb3 – Mouse Genome Informatics) and caspase 3 showed increased neuronal cell death in the mutants (Fig. 5F,G). Similarly, immunoreactivity of Tbr2 and caspase 3 showed increased cell death of basal progenitors in Cdk5rap2^an/an animals compared with controls (46.0±3.7% versus 8.7±4.9%, P<1×10^–9) (Fig. 5H,I). Together, these data suggest that both the precursor and neuronal populations are probably susceptible to cell death in the Cdk5rap2 mutants.

Here, we show that Cdk5rap2 is mutated in the Hertwig's anemia mouse, and we demonstrate that Cdk5rap2 is essential for normal progenitor proliferation and survival in the cerebral cortex. Cdk5rap2^an/an animals have smaller cerebral cortices that result from an overall reduction of the neuronal layers caused by a decrease in progenitor numbers. Increases in both cell death and premature cell cycle exit reduce the cortical precursor population. Moreover, the decrease of neuronal progenitors in Cdk5rap2^an/an animals correlates, paradoxically, with an increased mitotic index, suggesting delays in mitotic progression. We found altered mitotic orientation as well as increased abnormal mitotic figures, with aneupolar spindles, implicating essential roles for Cdk5rap2 in spindle and centrosome function during neurogenesis. These data suggest that CDK5RAP2 mutations in humans could cause microcephaly by mechanisms that include not only mitotic arrest and cell death, but also may include defects in cell fate determination. It is puzzling that anemia has not been reported in the few known human patients with CDK5RAP2 mutations, as it is the defining feature of Cdk5rap2^an/an mutants. However, both the microcephaly and anemia in Cdk5rap2^an/an mice are variable and strain dependent, and so similar modifiers might affect penetrance in humans.

Mitotic defects have been previously associated with cell fate changes (Buchman and Tsai, 2007). It has been proposed that dividing VZ progenitors with vertical cleavage planes could generate daughter cells with similar cell fates, and that small deviations from the vertical cleavage plane might be enough to disrupt these `symmetric' cell divisions and reduce the progenitor population (Zhong and Chia, 2008). The 2-fold increase in cleavage planes parallel to the VS in Cdk5rap2<sup>an/an</sup> mice would reduce such symmetric cell divisions and might correlate with reduced progenitor numbers. RNAi knockdown experiments have also provided evidence that Aspm, another gene associated with microcephaly in humans, regulates mitotic spindle orientation and cell fate in early neurogenesis in mice (Fish et al., 2006). Furthermore, Nde1 mutants have increased cell divisions with horizontal cleavage planes that correlate with premature depletion of the progenitor pool (Feng and Walsh, 2004). In these mutants, the decreased progenitor population correlates with premature and excessive generation of deep layer neurons, resulting in a larger fraction of cells leaving the cell cycle (Caviness et al., 2003). Although failure to maintain mitotic fidelity also increases the `exit fraction' in Cdk5rap2^an/an animals, unlike Nde1 mutants, Cdk5rap2^an/an mice show only modestly increased ratios of deeper versus superficial layer neurons. Lack of apparent overgeneration of deeper layer neurons in Cdk5rap2^an/an embryos could result from increased cell death early in development, in contrast to Nde1 mutants that show only modestly increased numbers of apoptotic cells (Feng and Walsh, 2004). Increased cell death in E10.5 Cdk5rap2^an/an embryos correlates with a decrease in S-phase cells, suggesting that before neurogenic divisions start, Cdk5rap2 regulates the expansion of the progenitor pool. Moreover, the distribution of caspase 3-labeled cells and their co-immunoreactivity with either neuronal or progenitor markers in E12.5 Cdk5rap2^an/an animals suggest that an overall increase in apoptosis affects both neurons and progenitors. In summary, changes in mitotic spindle orientation and premature cell cycle exit in Cdk5rap2^an/an animals seem to correlate with cell fate determination defects; however, the dramatic increase in apoptosis might mask the premature and excessive generation of neurons, an expected outcome of faulty cell fate determination.

Cdk5rap2^an/an mutants showed preferential thinning of the superficial neuronal layers. In evolutionary terms, these later-born neuronal layers only appeared after mammals diverged from reptiles (Nieuwenhuys, 1994), and the elaboration of the superficial layers has been proposed to be primarily responsible for the disproportionate expansion of the cerebral cortex relative to body size in mammals (Cheung et al., 2007). The evolutionary increase of superficial layers in mammals correlates with expansion of the SVZ, caused by increased basal progenitor numbers (Martinez-Cerdeno et al., 2006; Striedter and Charvet, 2009). In Cdk5rap2^an/an animals, the reduction of Tbr2-immunoreactive progenitors could account, in part, for the diminished superficial layers. This is interesting because the CDK5RAP2 gene shows a modest excess of non-conservative amino acid changes in the primate lineage leading to humans, suggesting that it might have been a target of evolution in primates (Evans et al., 2006). Thus, the essential roles of Cdk5rap2 in progenitor proliferation might reflect its relevance to the evolutionary expansion of the cerebral cortex (Kriegstein et al., 2006).

Cdk5rap2 contains a γ-tubulin-association domain that is found in other centrosomal proteins, including several in which mutations alter cell polarity, nuclear positioning, chromosomal alignment, spindle orientation and/or cleavage site specification (Megraw et al., 1999; Vaizel-Ohayon and Schejter, 1999; Venkatram et al., 2004; Verde et al., 2001). Cdk5rap2 and its fly ortholog Cnn (Centrosomin) recruit the γ-TuRC to the centrosome (Fong et al., 2008; Heuer et al., 1995; Zhang and Megraw, 2007; Zheng et al., 1998). The primary role of the γ-TuRC is the nucleation of microtubules (Zheng et al., 1998). However, mounting evidence suggests additional roles during cell cycle progression for γ-tubulin and other γ-TuRC subunits. When γ-tubulin fails to localize to the centrosome, some cells arrest at G2–M, followed by apoptosis, whereas cells that bypass the G2–M arrest show highly defective mitoses with increased monopolar spindles (Zimmerman et al., 2004). Furthermore, in Aspergillius nidulans, the absence of γ-tubulin arrests cells at the metaphase-to-anaphase transition (Prigozhina et al., 2004). Interestingly, in-frame deletion of exon 4 in the Cdk5rap2^an allele disrupts its γ-tubulin-association domain (Fong et al., 2008), and γ-tubulin localization to the centrosome appears decreased in Cdk5rap2^an/an mouse embryonic fibroblasts (S.P.M. and M.D.F., unpublished). Therefore, interaction of Cdk5rap2 and γ-tubulin might be required to maintain mitotic fidelity in neuronal progenitors, as has been described in other tissues (Eppig and Barker, 1984).

The abnormal number of spindle poles and consequent aneuploidy in Cdk5rap2^an/an mice might reflect defects in centrosome maturation or duplication. Centrosomes that fail to mature are less stable and prone to undergo splitting during mitosis (Fukasawa, 2007). Knockdown of CDK5RAP2 in interphase somatic cells causes abnormal centrosome splitting (Graser et al., 2007), suggesting defective centrosome maturation. Thus, failure of centrosome maturation followed by abnormal centrosome splitting would be an appealing mechanism to explain the aneupolar spindle poles observed in Cdk5rap2^an/an neuronal progenitors. Alternatively, defects in centrosome duplication could also cause abnormal numbers of spindle poles (Fukasawa, 2007). In both cases, the abnormal number of centrosomes might produce abnormal mitosis that would either result in cell death or cell cycle arrest, as we have shown in Cdk5rap2^an/an cortical progenitors. Alternatively, failure to arrest or exit the cell cycle might produce aneuploidy in Cdk5rap2^an/an embryonic germ cells (Eppig and Barker, 1984) and, potentially, in neuronal precursors.

In mammals, an emerging hypothesis suggests that the asymmetric inheritance of the mother and daughter centrosomes (by the progenitor and neuronal daughter cells, respectively) is important for stem cell renewal during neurogenesis (Wang et al., 2009). In mice, the removal of ninein, a component of the mother centriole appendages, disrupts centrosome maturation and impairs centrosome asymmetric inheritance, resulting in neuronal overproduction due to altered cell fate (Wang et al., 2009). Interestingly, mutations in Drosophila cnn randomize centrosome asymmetric inheritance and affect germ cell fate determination (Yamashita et al., 2007). We hypothesize that because Cdk5rap2^an/an animals have defects in mitotic orientation, the normal localization of centrosomes might be disrupted and could cause abnormal inheritance of the `maternal' centrosome. Alternatively, disruption of Cdk5rap2-dependent steps of centrosome maturation could impair asymmetric centrosome inheritance in Cdk5rap2^an/an mice. Either of these two mechanisms could then contribute to the premature depletion of the neuronal progenitor pool. A similar exhaustion of self-renewing progenitors might contribute to the hematological defects that originally defined the Hertwig's anemia mutation. Further analysis of centrosome maturation, duplication and inheritance in the absence of Cdk5rap2 might provide important additional insight into the relationship of the centrosome to cell fate determination and the pathology of microcephaly.

We thank S. Tzakas for technical assistance; M. Mahendroo and J. Eppig for contributing to early genetic mapping studies; members of the M.D.F. and C.A.W. laboratories for insightful discussions, especially C. Manzini, E. Morrow, N. Dwyer, E. C. Gilmore, M. Lehtinen and J. Liu; S. White (BIDMC Histology Core) for sample preparation; L. H. An and Y. Zu (BIDMC Imaging Core) and the HMS Genetics Department Imaging Core for confocal microscopy; and R. Hevner for the Tbr1 antibody. Transgenic core services were supported by NIH P30 DK049216 through the Center of Excellence in Molecular Hematology at Children's Hospital Boston. S.B.L was supported by NIH T32AG0222-13, S.P.M. by NIH K08 HL077157, M.H.H. by NIH T32 HL 076115-03 and C.A.W. by NINDS R01 NS32456. C.A.W. is an Investigator of the Howard Hughes Medical Institute. M.D.F. is supported by the Children's Hospital Pathology Foundation Wilkes Tumor Fund. Deposited in PMC for release after 6 months.