Pax6 is required to regulate the cell cycle and the rate of progression from symmetrical to asymmetrical division in mammalian cortical progenitors

The adult neocortex (hereafter called the cortex) is composed of six radial layers of neurones, each defined by cellular morphology, cell density and the formation of distinct axonal connections. During development, cortical cells are generated in the pseudostratified epithelium that forms the ventricular zone (VZ), a layer of dividing progenitor cells at the inner (ventricular) surface of the dorsal telencephalon (Sauer, 1935; Sidman et al., 1959; Bayer and Altman, 1991; Takahashi et al., 1993; Takahashi et al., 1995a; Takahashi et al. 1995b). As progenitor cells progress through the cell cycle their nuclei undergo dynamic intracellular migration, termed interkinetic nuclear migration. During this process, nuclei move away from the apical surface during G1, occupy the basal half of the VZ during S phase and return apically in G2 so that mitosis occurs at the ventricular surface (Sidman et al., 1959; Fujita, 1964).

Retroviral lineage tracing experiments have revealed multipotent progenitors in the mammalian VZ (Luskin et al., 1988; Price and Thurlow, 1988; Walsh and Cepko, 1992; Grove et al., 1993). At least some multipotent progenitors are thought to be self-renewing stem cells (Temple and Davis, 1994; Reid et al., 1995; Williams and Price, 1995; Qian et al., 1998) that predominate early in development (Williams and Price, 1995; Qian et al., 1998). It has been hypothesised that they are produced by progenitors dividing symmetrically, in a plane roughly at right angles to the ventricular surface (also known as vertical cleavage) to produce two identical daughter cells that remain within the VZ, thus expanding the proliferative population. As cortical neurogenesis progresses, asymmetrical division, which occurs in a plane roughly parallel to the ventricular surface (also known as horizontal cleavage), starts to predominate. It has been hypothesised that asymmetrical division produces two different daughter cells, one that remains within the VZ as a progenitor and one that migrates away from the VZ to differentiate (Kornack and Rakic, 1995; Chenn and McConnell, 1995; Doe, 1996; Mione et al., 1997; Lu et al., 2000; Matsuzaki, 2000). Although these hypotheses are attractive, there is no direct proof that the symmetry of division predicts the fates of the daughter cells in this way.

In the mouse cortex, neurones are generated between embryonic day (E) 12.5 and E17.5 (Gillies and Price, 1993) and the VZ is eventually lost at around birth. Cortical progenitors undergo a maximum of 11 cell cycles during neurogenesis (Takahashi et al., 1995b) and the laminar identity of cortical neurones is closely correlated to the cell cycle in which the cell is born. Thus, neurones of layers 6 and 5 arise from cycles 1-8 in mice, whereas neurones destined for layers 4 and 3/2 are born in cycles 9-11 (Takahashi et al., 1996). The layers of the cortex are, therefore, formed in an inside-first, outside-last sequence by migration of neurones from the VZ.

The mechanisms that regulate the cell cycle of cortical progenitors and the transition from predominantly symmetrical to predominantly asymmetrical divisions in the VZ as cortical neurogenesis proceeds are poorly understood. We present evidence that the transcription factor Pax6 is involved in the regulation of this progression. Pax6 encodes two DNA-binding motifs, a paired domain (Bopp et al., 1986; Treisman et al., 1991) and a paired-like homeodomain (Frigerio et al., 1986). Pax6 expression in the mouse embryo begins at E8.0 in the anterior surface ectoderm and neuroepithelium of the closing neural tube in the presumptive spinal cord, forebrain and hindbrain (Walther and Gruss, 1991; Stoykova and Gruss, 1994; Grindley et al., 1995). As the forebrain develops, Pax6 expression becomes restricted to specific telencephalic regions, including the VZ of the developing cortex, where it persists throughout corticogenesis, and specific diencephalic regions (Walther and Gruss, 1991; Stoykova and Gruss, 1994; Grindley et al., 1995; Grindley et al., 1997; Stoykova et al., 1996; Caric et al., 1997; Mastick et al., 1997; Warren and Price, 1997). Homozygous Pax6^Sey/Sey embryos, lacking any functional Pax6 protein, fail to develop eyes and nasal cavities, exhibit brain abnormalities and die soon after birth (Hogan et al., 1986; Hill et al., 1991; Schmahl et al., 1993; Stoykova et al., 1996; Warren and Price, 1997; Grindley et al., 1997; Mastick et al., 1997). Abnormalities of the forebrain arise as early as E9.5 when the region of the prosencephalon giving rise to the cerebral vesicles shows altered morphology (Mastick et al., 1997). Previous single-pulse bromodeoxyuridine (BrdU) labelling studies have suggested that defects in the cell cycle of Pax6^Sey/Sey cortical and diencephalic progenitors might underlie some of these abnormalities (Warren and Price, 1997; Gotz et al., 1998; Warren et al., 1999).

We analysed the distribution and cleavage orientation of mitotic cells and the production of postmitotic cells in the Pax6^Sey/Sey cortex during neurogenesis. We found that proportions of asymmetrically dividing cells increased more rapidly than normal from E12.5 to E15.5 and that newborn cells expressed neural-specific markers sooner after their generation. To assess how these observations relate to progression through the cell cycle, we performed cumulative BrdU labelling. At E12.5 the length of the cell cycle in the mutant was approximately 60% of that in wild-type littermates. To test whether this was an intrinsic property of mutant cortical cells, we cultured E12.5 dissociated Pax6^Sey/Sey cortical cells in isolation. These cells displayed increased proliferation rates compared with wild-type cells and an accumulation of the cyclin-dependent kinase inhibitor protein p27^Kip1. This protein has been shown to be necessary for the timing of cell-cycle withdrawal that precedes terminal differentiation (Casaccia-Bonnefil et al., 1997; Durand et al., 1997; Durand et al., 1998; Borriello et al., 2000; Levine et al., 2000; Miyazawa et al., 2000; Sasaki et al., 2000). However, although the cell cycle was shortened early in corticogenesis in Pax6^Sey/Sey embryos, we found that, later in development, proliferation was slowed. Analysis of progenitors in older cortex (E15.5) showed that the length of the cell cycle increased to over 140% of that in wild-type littermates, with a large increase in the length of S phase in the mutant cortex.

All mouse embryos were derived from Pax6^SeyEd heterozygote crosses (referred to as Pax6^Sey; also called Sey^Ed) (Roberts, 1967) maintained on an outbred CD1 background. The mutation in Pax6^SeyEd is G194X, which is predicted to result in a truncated nonfunctional protein lacking the homeodomain and C-terminal sequences (Hill et al., 1991). The day of the vaginal plug following mating was designated E0.5. Pregnant females were killed by cervical dislocation.

To obtain a complete representation of cell division patterns throughout the developing cortex, a comprehensive count of cells in telophase was undertaken. Experiments were performed at three ages: E10.5, E12.5 and E15.5. Whole embryos at E10.5 and E12.5 and brains at E15.5 were dissected into cold phosphate buffer, fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight and washed 3×5 minutes in 4% sucrose in PBS at room temperature. Tissues were embedded in melted 1.5% agar, 5% sucrose in PBS, transferred to 30% sucrose in PBS and left at 4°C until the blocks sank (usually 2 days). Blocks were rapidly frozen on dry ice and 10 μm coronal cryostat sections were cut, defrosted at room temperature for at least 30 minutes and processed for immunohistochemistry as described in Zhong et al. (Zhong et al., 1996). Antibodies used were against Numblike (Zhong et al., 1997) (rabbit polyclonal, 1:1000 dilution; generous gifts from W. Zhong and Y. N. Jan), phosphorylated histone H3 (Hendzel et al., 1997) (rabbit polyclonal, 1:2000 dilution, Upstate Biotechnology, NY) and Pax6 (Kawakami et al., 1997) (mouse monoclonal, 1:100, Developmental Studies Hybridoma Bank (DSHB), University of Iowa). After secondary labelling with fluorescent conjugated antibodies (Jackson Laboratories, PA), slides were washed three times in PBS and mounted in Vectashield (Vector Laboratories, UK) containing 1 μg/ml 4,6-diamidino-2-phenylindole (DAPI, Molecular Probes, Eugene). Images were taken with a Xillix Microimager 1400 CCD camera on a Zeiss Axioplan II microscope using IPLab (Scanalytics). For the analysis of cell division patterns, two wild-type and two Pax6^Sey/Sey embryos for each age were sectioned. At least 180 cells in the proliferative zones in each brain were assessed as shown in Fig. 1; numbers of cells undergoing ectopic division, symmetrical division, asymmetrical division or intermediate division were counted. Each data set was shown to comply with the assumptions necessary for ANOVA.

Analysis of cell-cycle kinetics in the cortex of E12.5 and E15.5 wild-type and Pax6^Sey/Sey embryos was performed as described by Nowakowski et al. (Nowakowski et al., 1989). In brief, BrdU pulses (70 μg/g body weight, i.p.) were given to pregnant dams every 2 hours for up to 8 hours (at E12.5) or for up to 12 hours (at E15.5). Dams at age E12.5 were killed 0.5, 2.5, 4.5 and 8.5 hours after the first injection. Dams at E15.5 were killed 0.5, 4.5, 8.5 and 12.5 hours after the first injection. The embryos were fixed for 3 hours in 4% paraformaldehyde, wax-embedded and processed to reveal BrdU immunoreactivity as described elsewhere (Gillies and Price, 1993). Parasagittal sections, 10 μm, were cut. The proportions of cells in the proliferative zone that were BrdU labelled were counted in 100 μm wide strips through the depth of the cortex at the rostral pole, the caudal pole and centrally in three nonadjacent sections that cut through the full rostrocaudal extent of the cortex. We found no consistent differences in the proportions of labelled cells (termed the labelling index, LI) rostrally, caudally and centrally and so data from the three sampling points were combined to give mean LIs and standard errors. For each age, we obtained data from two embryos for each timepoint and plotted data for each animal separately against the length of exposure to BrdU.

For double-labelling studies, after BrdU detection the sections were washed in PBS and incubated overnight in neurone-specific anti-β tubulin isotope III (TuJ1; mouse monoclonal, dilution 1:400, Sigma). The second antibody was detected using an alkaline phosphatase conjugated rabbit mouse IgG (Sigma) and 4-nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphate (Roche). Five hundred rostral, 500 caudal and 500 centrally positioned cells (1500 in total) from nonadjacent sections were counted. Again, we found no consistent differences in the counts from each location and data from the three regions in each animal were combined.

Cells from E12.5 cortices were dissociated by incubation for 45 minutes at 37°C in Earle’s buffered salts solution containing papain (20 units/ml; Papain Dissociation System, Worthington Biochemical, UK). Cells in suspension were plated onto poly-L-lysine-coated wells (LabTek II™; Nunc Life Technologies, UK) at a density of 1.5×10⁵ cells/cm² and cultured in defined serum-free medium (Romijn et al., 1994) at 37°C and 5% CO₂. Cultures were analysed at various times after plating (12.75-24 hours, see below) and each time-point was repeated in three separate experiments. Viable cells were identified on the basis of morphology and exclusion of Trypan Blue. For cell counts, between 800 and 1500 viable cells per culture were assessed in five or six randomly selected microscope fields.

Twelve hours after plating, BrdU was added to the culture medium (final concentration 10 μM) and, after 45 minutes, BrdU was removed and fresh culture medium was added. Cells were fixed in 4% paraformaldehyde either immediately after the pulse of BrdU, i.e. 12.75 hours after plating, or at 15, 18, 21 and 24 hours after plating, washed in PBS, and processed for BrdU incorporation as described previously (Shetty and Turner, 1998). Monoclonal anti-BrdU (clone BU 33, Sigma, UK) was used at 1:1000 dilution.

Other immunocytochemical reactions were carried out as follows. All steps were at room temperature. Cultures were treated with 90% methanol prior to incubation in blocking solution (PBS containing 0.5% Triton X-100 and 2.5% normal rabbit serum). Cells were then exposed to one of the following monoclonal antibodies for 18 hours: anti-p27^Kip1 (mouse monoclonal, 1:1000, Transduction Laboratories, NJ); anti-cyclin A (mouse monoclonal, 1:1000, Sigma, UK); anti-MAP2 (microtubule associated protein 2; mouse monoclonal, 1:400, Sigma, UK); TuJ1 (mouse monoclonal, 1:400, Sigma, UK); glial cell marker RC2 (mouse monoclonal, 1:50, DSHB, University of Iowa); anti-nestin rat 401 (mouse monoclonal, 1:100, DSHB, University of Iowa); anti-GFAP (glial fibrillary acidic protein; mouse monoclonal, 1:400, Sigma, UK); and anti-CNPase (2-′,3′-cyclic nucleotide-3′-phosphodiesterase; mouse monoclonal, 1:500, Sigma, UK). Subsequent processing was with biotin conjugated rabbit anti-mouse Ig (IgM for RC2) (1:400, both from Dako, UK) for 2 hours and ExtrAvidin®-peroxidase (1:500 in PBS, Sigma, UK) for 20 minutes. The reactions were developed with 3,3′- diaminobenzidine (DAB, Sigma, UK). All antibodies were diluted in blocking solution. Omission of the primary antibody resulted in no detectable staining.

For dual-labelling experiments, after developing in DAB, cultures were washed and exposed to 90% methanol to inhibit the peroxidase activity due to the first immunostaining. Wells were washed and processed for the second staining using the procedure described above. However, to distinguish the brown product from first reaction, the second staining was visualised with a blue reaction product using tetramethylbenzidine as substrate (HistoMark TrueBlue®, Kirkegaard and Perry Laboratories, UK).

For both sectioned material and cell culture experiments, data sets were compared by ANOVA (statistical significance between data sets was assigned when P<0.05).

In vivo analyses were performed at E10.5, E12.5 and E15.5, which correlate with the onset of cortical neurogenesis, early neurogenesis and mid-neurogenesis (Gillies and Price, 1993; Caviness et al., 1995; Price et al., 1997). Cells in metaphase were visualized using an antibody directed against phosphorylated histone H3. This antibody binds weakly within pericentromeric heterochromatin during G2 and then spreads in an ordered fashion coincident with mitotic chromosome condensation, becoming strongest during metaphase (Hendzel et al., 1997). In wild-type cortex at all three ages metaphase cells were largely restricted to the ventricular surface, as expected (Fig. 1C). This was also the case in Pax6^Sey/Sey cortex at E10.5 and E12.5, but at E15.5 there appeared to be a significant proportion of metaphase cells located ectopically, away from the ventricular edge (Fig. 1D).

Quantification gave a more comprehensive representation of cell-division patterns throughout the developing cortex. Sections were stained with DAPI, which clearly identifies cells in late anaphase and telophase when condensed chromosomes are separating, and the number of cells in these stages were counted. In each section (n=15 per brain), cells throughout the depth of the cortical proliferative zone were simultaneously classified by position, either at the ventricular surface or ectopic (within the proliferative zone but more than three cell diameters from the ventricular surface) (Fig. 1C,D); or by orientation of division, either symmetrical (plane of division at 90°±30° to the ventricular surface), asymmetrical (parallel, 0°±30°, to ventricular surface) or intermediate (30-60° or 120-150°) (Fig. 1E-G). The spindles of cortical progenitors rotate throughout metaphase but stop at the start of anaphase (Adams, 1996) and so this method provides an accurate measure of the plane of division. It has been shown that the distributions of cleavage orientations observed by staining fixed sections are identical to those observed by following living cells in cortical slices (Chenn and McConnell, 1995). The cortex exhibits temporal gradients of development; for example, anterior cortex is more mature than posterior cortex at any given stage of corticogenesis (Bayer and Altman, 1991). To test whether cell division patterns were influenced by the anterior-posterior gradient of development, counts were carried out on five adjacent sections at three different anterior-posterior positions in the cortex: 25%, 50% and 75% through the telencephalon (where 0% is the most rostral part and 100% is the most caudal: Fig. 1A,B). Position was found to have no significant effect on the incidence of either ectopic or asymmetrical division and data from the different regions were combined in the final analyses.

ANOVA revealed that the proportions of dividing cells located ectopically was significantly higher in Pax6^Sey/Sey cortex than in wild-type cortex specifically at E15.5 (Fig. 2A). At E10.5, 3% of wild-type and 7% of Pax6^Sey/Sey divisions were ectopic. This proportion rose slightly at E12.5 to 12% in wild-type and 17% in Pax6^Sey/Sey cortex. At E15.5, only 13% of divisions occurred ectopically in wild-type compared with 47% in Pax6^Sey/Sey embryos (P<0.005). These data strongly suggest that interkinetic nuclear migration is disrupted by mid-neurogenesis, perhaps because cells enter mitosis before or after their nuclei undergo apical relocation to the ventricular surface.

The analysis of cleavage orientations revealed that the proportions of cells undergoing asymmetrical division in the Pax6^Sey/Sey cortex progressed to higher levels more rapidly than in wild-type cortex (Fig. 2B,C). In the E10.5 and E12.5 wild-type cortex (Fig. 2B), the majority of divisions (almost 80%) were symmetrical; the rest were almost equally divided between asymmetrical and intermediate. By E15.5, the proportion of symmetrical divisions had fallen by about 10%, whereas the proportion of asymmetrical divisions had risen by a similar amount (Fig. 2B). These data are consistent with a predicted increase in proportions of asymmetrical divisions as neurogenesis progresses (Chenn and McConnell, 1995). In the Pax6^Sey/Sey cortex at E10.5, the distribution of cleavage orientations was close to that observed in wild-type littermates, although the proportion of symmetrical divisions was slightly lower and that of asymmetrical divisions was correspondingly higher than in wild-types (Fig. 2C). This difference from wild-type cortex became progressively more pronounced with age. By E12.5 the proportion of asymmetrical divisions was up to 17±2% (compared with 8±2% in wild-types; P<0.05); by E15.5, 36±2% of divisions occurred asymmetrically (compared with 17±2% in wild-types; P<0.0005) and only 48±2% were dividing symmetrically (Fig. 2C). The proportions of intermediate cells remained at similar levels throughout in both genotypes (Fig. 2B,C). These data indicate that the change towards asymmetrical division is accelerated in the Pax6^Sey/Sey cortex as compared with wild-type cortex.

To investigate further the kinetics of the cell cycle in the ventricular zone of mutant and wild-type embryos, E12.5 and E15.5 pregnant dams were given repeated injections of BrdU to label cumulatively the proliferating population. The proportions of BrdU-labelled cells in the proliferating zone after different times of exposure to BrdU are plotted in Fig. 3A,B. Using the model described by Nowakowski et al. (Nowakowski et al., 1989), and used since by others (Takahashi et al., 1993; Takahashi et al., 1995a; Takahashi et al., 1995b; Takahashi et al., 1996), these data can be interpreted as summarised in Fig. 3C to give the proportion of cells in the proliferative zone that are actually proliferating (i.e. the growth fraction, GF), the length of the cell cycle (Tc) and the length of S phase (Ts).

Growth fractions were almost 1.0 at E12.5 and approximately 0.9 at E15.5, irrespective of genotype (Fig. 3A,B). The length of the cell cycle in E12.5 mutants was 60% that of wild-type littermates (7.5 hours for Pax6^Sey/Sey, 12.5 hours for wild-type). The length of S-phase was approximately 3 hours in mutants and approximately 4.5 hours in wild-types; these values represent 40% and 36% of the lengths of their respective cell cycles. Our values for Tc and Ts in E12.5 wild-type telenlencephalon were very close to those found before in E12-E13 mouse cortex (Takahashi et al., 1995b). In E15.5 wild-type cortex, the length of the cell cycle had increased, as described previously (Takahashi et al., 1995b). However, although the cell cycle was around 14 hours in wild-type embryos, it had increased to over 20 hours in mutant brains. Even more surprisingly, we found that, although Ts in the wild-type cortex remained almost constant (4.5 hours at E12.5 and 5 hours at E15.5, i.e. constantly 36% of the total cell cycle), Ts in the mutant increased dramatically from 3 hours (40% of the total cell cycle) at E12.5 to 11.5 hours (58% of the total cell cycle) at E15.5. Thus, although at E12.5 mutant cell cycles are much shorter, both wild-type and mutant cells spend similar proportions of time in S phase. By contrast, by E15.5 mutant cell cycles are longer and mutant cells spend a much greater proportion of time in S phase (summarized in Fig. 3D).

The cell-division analysis described above revealed that the proportion of cells undergoing asymmetrical division in the Pax6^Sey/Sey cortex progressed to higher levels more rapidly than their wild-type littermates. It has been suggested that asymmetrical divisions generate postmitotic neurones (Chenn and McConnell, 1995), although this hypothesis has not been proved directly. We tested whether cells switch on neural-specific markers earlier in the mutant cortex than in the wild-type cortex. We performed double labelling for BrdU and TuJ1 expression in sections from pregnant dams aged E12.5 and E15.5 that had been injected with BrdU at two hourly intervals for up to 8 and 12 hours, respectively. These animals were killed at 2.5 hours (E12.5 dams only), 4.5 hours (E12.5 and E15.5 dams), 8.5 hours (E12.5 and E15.5 dams) and 12.5 (E15.5 dams only) after the initial administration of BrdU. The percentages of the BrdU-labelled cells that were also labelled for TuJ1 were counted in the VZ. Within each of the four groups of animals defined according to age and genotype (E12.5 wild-type; E12.5 mutant; E15.5 wild-type; E15.5 mutant), the proportions of BrdU-labelled cells within the VZ that were also TuJ1 positive remained comparatively constant with increasing length of exposure to BrdU. This was presumably because the generation of new TuJ1-positive cells would be counterbalanced by the migration of older TuJ1-positive cells out of the VZ, as it has previously been observed that migration from the mutant VZ occurs at a normal rate at the ages studied here (Caric et al., 1997). There were, however, major differences between the groups such that only 2.5±0.6% of BrdU-labelled cells within E12.5 wild-type VZ were double-labelled cells (Fig. 4A), whereas 21.2±2.7% of BrdU-positive VZ cells were double labelled in the mutant (Fig. 4B). At E15.5, 6.4±3.9% of BrdU-labelled cells were double labelled in the wild-type VZ, compared with 51.2±9.3% in the mutant VZ. These data indicate that the time between S phase and the onset of expression of neural specific markers is significantly shorter than normal in the Pax6^Sey/Sey cortex.

We also tested the possibility that the expression of other cell fate determinants normally associated with postmitotic neurones might be increased in Pax6^Sey/Sey cortex. In wild-type mice, Numblike (Nbl) is expressed in postmitotic cortical neurones but not in progenitors in the ventricular zone (Zhong et al., 1997). We examined the expression of Nbl in Pax6^Sey/Sey cortex at E10.5-E15.5 (Fig. 5).

At E10.5 the developing telencephalic wall consists almost entirely of proliferating neuroepithelial cells, and nuclear staining with Pax6 antibody was evident in the majority of these cells in wild-type embryos (Fig. 5B). The antibody did not detect Pax6 protein in Pax6^Sey/Sey mutant embryos (Fig. 5E,K). Cytoplasmic Nbl expression was detected in the emerging preplate, or primordial plexiform layer, of the cortex in both E10.5 wild-type and E10.5 Pax6^Sey/Sey telencephalon (Fig. 5C,F). The preplate is formed by the first postmitotic cells migrating from the VZ (Marin-Padilla, 1998). More Nbl-positive cells were apparent in the Pax6^Sey/Sey preplate (Fig. 5F) than in the wild-type preplate (Fig. 5C). By E15.5 the developing wild-type cortex has a well-defined laminar structure comprising proliferative layers (ventricular zone, VZ and subventricular zone, SVZ), an intermediate layer (IZ) of radially migrating neurones, and differentiating neurones and afferent and efferent axons in the cortical plate (CP). Expression of Pax6 was detected in the VZ and SVZ of the wild-type cortex (Fig. 5H). Intense Nbl immunoreactivity was seen in the nonproliferative layers of the cortex of both wild-type and Pax6^Sey/Sey embryos (Fig. 5I,L). Unlike earlier ages, it was hard to assess numbers of labelled cells at E15.5, but in the Pax6^Sey/Sey cortex there was an overall upregulation of Nbl immunostaining in the cortical plate relative to that in the IZ (Fig. 5L). These findings support the notion that neurogenesis is more advanced in the mutants at both early and intermediate stages of corticogenesis.

Because the Pax6^Sey/Sey embryo has numerous abnormalities of tissues surrounding and interacting with cortex, we tested whether alterations of the cell cycle and rates of differentiation in the cortex are an intrinsic property of mutant cortical cells by growing these cells in isolation in culture. Dissociated E12.5 cortical cells were plated and, 12 hours later, a pulse of BrdU was added for 45 minutes. Cultures were studied 12.75, 15, 18, 21 and 24 hours after plating. We characterised the behaviour of cells and their expression of molecular markers in the cultures after staining with a range of antibodies. We did not observe any significant level of cell death in either wild-type or Pax6^Sey/Sey cultures at any time-point. Although the two genotypes were plated at the same density, the mutant cells became increasingly clustered together compared with the wild-type cells (Fig. 6A,B). Cells labelled with BrdU were found in all cultures at all time-points. Fig. 6A-D shows BrdU labelling 24 hours after plating (i.e. 11.25 hours after the end of the BrdU pulse). At this stage, the latest that we studied, the wild-type BrdU-labelled cells remained either single or in small clusters containing less than 12 cells connected by a network of cellular processes (Fig. 6A,C). By contrast, Pax6^Sey/Sey cells were clustered into much larger groups of up to 50 cells (Fig. 6B,D) and their nuclei often displayed a chain-like morphology (Fig. 6D), similar to that described previously (Gotz et al., 1998).

To characterise further the cells in the cultures, we used antibodies against specific cell markers including TuJ1 (Lee et al., 1990; Easter et al., 1993) and MAP2 (Matus, 1991) for neurones, GFAP for astrocytes (Dahl, 1981), CNPase for oligodendrocytes (Sprinkle et al., 1987), RC2 for radial glia (Misson et al., 1988) and nestin for undifferentiated neural precursors (Lendahl et al., 1990). Throughout the culture period, we detected predominantly neuronal cell markers (nestin, TuJ1 and MAP2) in both mutant and wild-type cells (Fig. 6A,B,E-H). The marker RC2 was also detected in cells of both genotypes (Fig. 6I,J), but very few mutant or wild-type cells were positive for GFAP or CNPase (data not shown). Many of the RC2-positive cells co-expressed MAP2 (Fig. 6I,J), in line with recent findings that radial glial cells may generate neurones (Barres, 1999; Malatesta et al., 2000).

We followed the proliferative behaviour of the cultured cells quantitatively by measuring, at a series of times after administration of BrdU, the proportions of BrdU-positive cells, the overall densities of cells in the culture, the proportions of cells containing the nuclear protein, cyclin A, which is expressed selectively during the S/G2/M phases of the cell cycle (Darzynkiewicz et al., 1996), and the proportions of cells expressing p27Kip1, a cyclin-dependent kinase inhibitor that is required for neuronal cells to exit the cell cycle and start terminal differentiation (Sherr and Roberts, 1995; Sherr and Roberts, 1999; Casaccia-Bonnefil et al., 1997; Durand et al., 1997; Durand et al., 1998; Borriello et al., 2000; Levine et al., 2000; Miyazawa et al., 2000; Sasaki et al., 2000). Results are shown in Fig. 7. Immediately after the pulse of BrdU, the proportion of BrdU-labelled mutant cells was approximately twice that of the wild-type cells (Fig. 7A), indicating that proliferation rates are increased in Pax6^Sey/Sey cortical cells in vitro as well as in vivo. In wild-type cultures, the proportions of BrdU-labelled cells increased over the following 8.25 hours, as the cells that had been in S-phase during the pulse of BrdU underwent division, and fell again over the next 3 hours, presumably as proliferating cells that had not been labelled by the BrdU pulse divided to dilute the BrdU-labelled population (Fig. 7A). The schematic in Fig. 8A illustrates this course of events. In stark contrast, the proportions of BrdU-labelled mutant cells increased steadily throughout culture (Fig. 7A). This observation, together with our finding that overall cell density increased more rapidly in the cultures of mutant cells than in the cultures of wild-type cells (Fig. 7B), suggests that mutant cells continued to proliferate more rapidly than wild-type cells for up to 24 hours in culture (Fig. 8B).

In the cultures of wild-type cells, there were no significant changes in the percentages of cells expressing cyclin A (∼20%) or p27Kip1 (∼55%) between 12.75 and 24 hours after plating (Fig. 7C,D). In the cultures of Pax6^Sey/Sey cells, the proportion of cells expressing cyclin A immediately after the pulse of BrdU was higher than in wild-type cultures (Fig. 7C), which is compatible with them having a higher proportion of BrdU-labelled cells at this time (Fig. 7A). Over the following 11.25 hours, the proportion of mutant cells expressing cyclin A decreased (Fig. 7C; P<0.01), whereas the proportion of mutant cells expressing p27Kip1 increased (Fig. 7D; P<0.01). This indicates that, unlike the wild-type cells, whose division maintains stable proportions of proliferating and postmitotic cells, division of the mutant cells is biased towards production of an ever-increasing proportion of postmitotic cells. This interpretation is illustrated in the schematic in Fig. 8.

Our main conclusions from in vivo work are that Pax6 is required in cortical development to regulate the rate of progression of progenitors through the cell cycle, the rate of progression from symmetrical to asymmetrical division and the onset of expression of neural-specific markers. In the absence of Pax6, the cell cycle and the onset of expression of neural-specific markers are both accelerated at the start of cortical neurogenesis. As neurogenesis progresses, the increase in the proportion of asymmetrical divisions occurs more rapidly than normal, and increasing proportions of progenitors divide in ectopic positions, suggesting that interkinetic nuclear migration becomes defective. By mid-neurogenesis, progenitor cell cycles in the mutant slow below the rate observed in wild-type embryos. We found that dissociated mutant cortical cells isolated in culture at E12.5 also showed more rapid proliferation and production of postmitotic cells than wild-type cells, suggesting that the changes seen in vivo are intrinsic to the cortical cells, rather than being secondary to one or more of the numerous extracortical defects in Pax6^Sey/Sey embryos.

The differentiation of multipotential progenitor cells into specific classes of postmitotic neurones or glial cells occurs at characteristic times during neural development and appears to be accompanied by a progressive restriction of potential cell fate within the progenitors (Frantz and McConnell, 1996; Desai and McConnell, 2000). The regulation of this diversity requires a complex interaction between intracellular mechanisms and environmental cues. Clonal analysis on oligodendrocyte precursor cells suggests that an intrinsic clock operates within each cell to help control when it stops dividing and differentiates (Temple and Raff, 1986). It is probable that a similar internal clock, whose mechanism may be influenced by extracellular signals, acts to determine the rate of division and differentiation of neural progenitors. To date, however, little is known about the molecular mechanism that may govern such a clock. The results we present suggest that Pax6 may be an essential component.

During cortical neurogenesis in the wild-type telencephalon, the founder progenitor population undergoes a set number of 11 cell divisions (Takahashi et al., 1995b), during which there is a gradual switch from proliferative to neurogenic division (Takahashi et al., 1996). Coincident with this conversion is a decrease in the proportion of cells undergoing symmetrical divisions and an increase in the proportion undergoing asymmetrical divisions (Chenn and McConnnell, 1995). In this study, by comparing the orientations of mitotic spindles in Pax6^Sey/Sey and wild-type cortex in vivo, we demonstrated a more rapid progression to higher proportions of asymmetrical divisions in the absence of Pax6. This more rapid progression to asymmetrical division correlates with an accelerated production of cells expressing the markers Nbl and TuJ1. Both in wild-type and Pax6^Sey/Sey cortex, therefore, circumstantial evidence suggests a link between the symmetry of cleavage and the production of postmitotic versus proliferative cells. Our in vitro work indicated that E12.5 Pax6^Sey/Sey cortical cells retain the ability to proliferate more rapidly than normal and to generate increasing proportions of postmitotic cells even when dissociated and isolated in culture. This suggests that altered proliferation and differentiation in mutant cortical cells are not dependent on division taking place at an altered angle with respect to the ventricular surface, nor on the disruption to the interkinetic movement, neither of which would occur under these culture conditions. However, it cannot be ruled out that any intracellular reorganisation of cell fate determinants relative to the ventricular cycle has occurred prior to cell dissociation and that this memory is somehow retained within the dissociated cells for several cell cycles. Overall, our work does not allow us to draw firm conclusions on the nature of the relationship between cleavage plane and differentiation, i.e. whether they are causally related or independently regulated.

One parsimonious hypothesis to explain our various observations is as follows. Regulating the rate of progression through the cell cycle may be a primary function of Pax6 in early cortical progenitors. In the absence of Pax6, the cell cycle at E12.5 is shortened. If, in both wild-type and Pax6^Sey/Sey embryos, the proportion of cells undergoing terminal differentiation is related to the number of cell cycles that a progenitor has gone through, rather than to the embryonic age of the cell in days, then an increase in the proportion of cells undergoing differentiation on a given day would follow as a consequence. Progenitors from an E12.5 Pax6^Sey/Sey brain would be more advanced than those from an E12.5 wild-type brain in terms of their tendency to generate postmitotic cells, both in vivo and in vitro.

Intriguingly, examination of cell-cycle times at mid-neurogenesis (E15.5) showed that the absence of Pax6 produces a significantly slower cell cycle at this age. Even more surprisingly, the relative length of S phase has significantly increased in respect to the overall cell cycle. Previous work has indicated that, in contrast to cortical progenitors, diencephalic progenitors proliferate more slowly than normal in Pax6^Sey/Sey embryos (Warren and Price, 1997). Therefore, it may be that, although a general function of Pax6 is to influence cell cycle times, the exact nature of its effect depends on the molecular and cellular context within which it acts. This context may change as the cortex develops. Another explanation, however, is that although the cells are cycling rapidly earlier on in development, there is insufficient time for all components essential for continued division to be synthesised and replenished. By E15.5, therefore, fundamental resources within the cells may be lost and the cell cycle may subsequently become deranged.

Throughout this study we did not find any evidence that changes in cell-cycle kinetics, proportions of asymmetrical divisions and rates of differentiation were different in different regions of the mutant cortex. Bishop et al. (Bishop et al., 2000) have suggested that an anterior-posterior gradient of Pax6 expression contributes to regional differences in the differentiation of the cortex. It is possible, therefore, that Pax6 regulates rates of proliferation and differentiation independently from any subsequent effects on the region-specific nature of the differentiation.

Although it is tempting to speculate on a single primary function for Pax6 in the cortex that can account for all the Pax6^Sey/Sey cortical defects, in all probability Pax6 plays multiple roles. As a transcription factor, Pax6 may have distinct functions in different cells and at different times during development. Our data strongly suggest a cell cycle-related function, but it remains feasible that Pax6 has separate roles in regulating the cell cycle, cell differentiation and fate, cell adhesion and possibly other processes.

We thank K. Gillies, G. Grant and V. Allison for their technical help in this study, A.Carruthers and P. Teugue for help with statistics and P. Perry for help with imaging. Support was from the European Commission (BIOMED BMH4 CT97-2412 and TMR Programme, Marie Curie Research Training Grants ERB4001 GT972907) and BBSRC. G.E.-T is a Marie Curie Research Training Fellow and member of the Marie Curie Fellowship Association and P.R. is a Lister Research Fellow.