Pax6 is a member of an evolutionarily conserved family of transcription factors. It is developmentally regulated and is required for the normal embryonic development of the central nervous system, eye and pancreas. Pax6 mutations in the mouse result in the Small eye (Sey) phenotype. Heterozygous mice have eye defects and homozygotes die immediately after birth lacking eyes, nasal cavities and with severe brain abnormalities, including a malformed cerebral cortex. Recent work has established that there are changes in expression of cell adhesion molecules and these may underlie at least a part of the Pax6Sey/Sey phenotype. Here we used cell transplants and explant cultures to investigate the role of Pax6 in cell adhesion. Pax6Sey/Sey embryonic cortical cells transplanted into wild-type embryonic cortex were observed to segregate from wild-type cells and form dense clusters. Cells migrating from explants of Pax6Sey/Sey embryonic cortex clustered to a greater extent than cells migrating from wild-type controls. These new data support the hypothesis that Pax6 exerts a cell-autonomous effect on the adhesiveness of cortical cells.
Pax6 is a member of a family of transcription factors characterized by the presence of an N-terminal 128 amino acid DNA binding domain, the paired box. This domain is divided into two helical sub-domains called PAI and RED that can each bind DNA independently (Jun and Desplan, 1996). At least nine murine and human paired box genes have been identified to date (Callaerts et al., 1997). Separated from the paired box by a 78 amino acid glycine-rich linker sequence is a second 60 amino acid DNA binding domain, the homeobox. These two domains can interact independently and cooperatively with DNA. At the C-terminal of Pax6 is a 153 amino acid proline–serine–threonine rich domain that is thought to be the transcriptional regulatory element.
Pax6 is expressed in the retina, lens and cornea of the developing vertebrate eye (Walther and Gruss, 1991; Grindley et al., 1995). It is also expressed at a range of developmental stages in regions of the forebrain, hindbrain, cerebellum, the ventral neural tube and the pancreatic islet cells (Walther and Gruss, 1991; Stoykova and Gruss, 1994; Grindley et al., 1995; St-Onge et al., 1997; Warren and Price, 1997; Kioussi et al., 1999). Humans heterozygous for mutations in PAX6 suffer from aniridia (iris hypoplasia), which is associated with cataracts, lens dislocation, foveal dysplasia, optic nerve hypoplasia and nystagmus (Jordan et al., 1992; Glaser et al., 1994). The vast majority (92%) of known PAX6 mutations in humans are nonsense mutations (Hanson et al., 1999). Interestingly, there are a few cases where the mutation has been found several hundred kilobases away from the PAX6 transcriptional start site (Fantes et al., 1995; Lauderdale et al., 2000), demonstrating the presence of distant regulatory domains. A rare case of an infant with a compound heterozygous mutation in PAX6 suffered severe craniofacial and central nervous system defects, had no eyes, no adrenal glands, and died neonatally, a phenotype similar to the homozygous null mutation in mice (Glaser et al., 1994).
Pax6 mutation in the mouse results in the Small eye (Sey) phenotype. At least eight alleles of Pax6 have been identified in the mouse so far and all are similar loss-of-function mutants (Glaser et al., 1990; Hill et al., 1991; St-Onge et al., 1997; Lyon et al., 2000). It is unclear whether these alleles are complete nulls but they will be referred to here as Pax6–/–. A premature stop codon in the linker domain generates the Pax6SeyEd allele (Hill et al., 1991). Heterozygous mice have a reduced eye size, iris hypoplasia, corneal opacification, and cataracts. Homozygotes die immediately after birth with no eyes, no nasal structures and severe brain abnormalities, including malformed cerebral cortex (Hogan et al., 1986; Hill et al., 1991; Schmahl et al., 1993; Caric et al., 1997). The diencephalic equivalent is reduced in size, is not differentiated to a normal extent (Stoykova et al., 1996; Warren and Price, 1997), and fails to innervate the cortex (Pratt et al., 2000). As in the human, distant regulatory modules have been shown to be essential for the correct expression pattern of Pax6 (Kleinjan et al., 2001).
A Role for Pax6 in Cortical Development
In normal development of the mouse, neurogenesis occurs from embryonic day 12 (E12) to E18 (Gillies and Price, 1993; Levers et al., 2001). At these ages, the lateral ventricle is lined by a population of cells in a region called the ventricular zone, which gives rise to most neurones and glial cells of the mammalian cortex. These cortical precursor cells are not a homogeneous population and there is mounting evidence that different precursor cells generate different differentiated cell types (Grove et al., 1993; Luskin et al., 1993; Reid et al., 1995; Tan et al., 1998; Heins et al., 2002). The mechanisms controlling the fates of these precursor cells are not yet elucidated.
Nuclei of ventricular progenitor cells undergo dynamic intracellular migration during the cell cycle. Nuclei move away from the apical surface during G1, occupy the outer half of the ventricular zone during S phase and return apically in G2 so that mitosis occurs at the ventricular surface (Sidman et al., 1959; Fujita, 1964). Neurons exit the cell cycle in contact with radial glial fibres to migrate into more superficial positions. When neurons reach the top of the cortical plate they detach and associate into layers with cohorts of a similar birth date. This results in the cortex being formed in an ‘inside-out’ laminar fashion. After neural production has finished, astrocytes and oligodendrocytes are produced in large numbers from precursors in the subventricular zone (Gleeson and Walsh, 2000; Morrison, 2000).
In Pax6–/– mice both the cortical ventricular zone and the subventricular zone are enlarged (Schmahl et al., 1993; Stoykova et al., 1996; Caric et al., 1997). In addition, the cortical plate is thinner and within the intermediate zone (i.e. between the subventricular zone and cortical plate) there are large collections of cells characteristic of those in the subventricular zone. Cumulative labelling with bromodeoxyuridine (BrdU) has revealed that proliferative rates in the early Pax6–/– embryonic cortex increase (Estivill-Torrus et al., 2002). In addition, proliferating cells in S phase are found scattered throughout the ventricular zone, suggesting either a failure in interkinetic nuclear migration or asynchronous cycling of precursor cells in the mutant cortex (Götz et al., 1998; Estivill-Torrus et al., 2002). Birthdating studies with BrdU in vivo show that many later-born neurons fail to migrate to the cortical plate and accumulate in the subventricular zone (Caric et al., 1997). Immunohistochemical analysis of neuron-specific class III β-tubulin isotype (TuJ1), an early marker for postmitotic neurons (Lee et al., 1990), has shown that cells in Pax6–/– mutant cortices that fail to migrate do begin neuronal differentiation (Caric et al., 1997). There is a similar defect in Small eye rats (rSey): the E20 cortices have an abnormal clustering of cells in the ventricular and intermediate zones of the cortex (Fukuda et al., 2000).
Not all cells in the ventricular zone express Pax6; rather, expression appears to be localized to a subset of radial glial cells (Götz et al., 1998). In Pax6–/– embryos, the morphology of radial glial cells is altered. At E15.5, wild-type radial glia have straight processes running towards the pial surface whereas mutant radial glial processes appear wavy and have frequent small extrusions and branches (Götz et al., 1998). Co-culture experiments mixing E13.5 Pax6–/– cortical cells and wild-type cells failed to rescue the phenotype of mutant radial glial cells, suggesting that the defect may be cell-autonomous (Götz et al., 1998). Work over the past few years has shown that radial glial cells are able to generate not only glial cells but also neurons (Campbell and Götz, 2002). In cultures of Pax6–/– radial glial cells, less neural clones and more non-neural clones were produced than in cultures of wild-type radial glial cells (Heins et al., 2002). Furthermore, in vivo quantification showed a 50% reduction of radial glial-derived neurons in the Pax6–/– cortex at E14 and E16 (Heins et al., 2002). Infecting cells from E14 Pax6–/– cortex with a retroviral vector containing full length Pax6 cDNA increased the number of differentiated neurons and appeared to reduce proliferation (Heins et al., 2002). These findings suggest that Pax6 may play a cell-autonomous role driving radial glial cells to produce cells of a neuronal fate.
Some defects in the developing central nervous system of Pax6–/– embryos are not due to a direct cell-autonomous requirement for Pax6 in the affected process. Transplantation of Pax6–/– cortical precursors into a wild-type cortical environment can rescue their migrational defect, suggesting that it may be secondary to defects of other cells such as radial glia which normally guide migration (Caric et al., 1997). Abnormally high levels of cell death among late-embryonic Pax6–/– dorsal thalamic cells are most likely secondary to the inability of these cells to obtain trophic support from the cerebral cortex, to which they do not connect (Lotto et al., 2001).
Here, we present new data on experiments to test whether Pax6 has a cell-autonomous role regulating cell adhesion in the developing cortex. We examined the behaviour of Pax6–/– cortical cells when they were either transplanted into wild-type cortex or cultured as explants. The rationale behind the first approach was that embedding the mutant cells in a wild-type environment would reveal their cell-autonomous defects and ameliorate any defects that might arise in mutant embryos as a secondary consequence of abnormalities in other cells. The explant approach provided a means of testing the adhesive properties of mutant cortical cells in isolation from other cell types or from wild-type cells.
Materials and Methods
All mouse embryos were derived from Pax6SeyEd heterozygote crosses and were genotyped as previously described (Hogan et al., 1986; Hill et al., 1991; Caric et al., 1997). Wild-type Long–Evans hooded rats were obtained from external suppliers. The day of the vaginal plug following mating was designated E0.5.
Pax6–/– and wild-type embryos were obtained from pregnant mice that had been injected on E15.5 with BrdU (70 μg/g in sterile saline i.p.) 1 h prior to death by cervical dislocation. The embryonic cerebral neocortices were isolated and dissociated as described by Caric et al. (Caric et al., 1997). Viabilities of dissociated cells were assessed by trypan blue exclusion and were ∼95%. E15.5 pregnant rats were anaesthetized with ketamine (60 mg/kg i.m.) and xylazine (6 mg/kg i.m.), and dissociated mouse cells were injected into the telencephalic vesicles using methods described before (Caric et al., 1997). The rats recovered and gave birth as normal. Their young were deeply anaesthetized with sodium pento-barbitone (1 mg i.p.) on postnatal day 7 (P7) and perfused transcardially with 4% paraformaldehyde. Wax sections were cut and reacted to reveal BrdU labelling, as described previously (Gillies and Price, 1993).
E13.5 or E15.5 wild-type and Pax6–/– neocortex was obtained as described above, sectioned parasagittally and divided into anterior, middle and posterior thirds. Anterior (A) and posterior (P) thirds were cut into pieces and placed in 9 × 9 mm wells on chambered coverglass slides coated with poly(L)-lysine and laminin in serum-free medium (Lotto and Price, 1999). Wild-type and Pax6–/– diencephalon from brains of the same age were dissected and added to the wells such that tissue was co-cultured in the following combinations. (i) Both wild-type A and P cortex were cultured with either wild-type or mutant diencephalon (four combinations); (ii) both mutant A and P cortex were cultured with either wild-type or mutant diencephalon (four combinations). Diencephalic explants were included since it is known that factors from this tissue, which interacts with the cortex in vivo, are required to ensure the survival, growth and migration of cortical cells (Lotto and Price, 1996; Price and Lotto, 1996; Lotto et al., 1999; Edgar and Price, 2001). Cortical explants in contact with diencephalic tissue, or which became innervated by processes that grew from the diencephalon, were excluded from the analyses. For each age, the explant cultures were set up using three separate wild-type and Pax6–/– brains (in three independent experiments). After 24 h in culture, digital images of five randomly selected explants in each culture well were recorded. Explants were fixed in 4% paraformaldehyde and immunostained with antibodies using standard techniques.
Analysis of Clumping In Vivo and In Vitro
We observed differences in the degree of clumping of mutant and wild-type cells that had integrated into the wild-type cortex or migrated out of cultured explants. These differences were analysed quantitatively.
For analysis of the distributions of transplanted cells in vivo, a series of 1-in-10 sections through the cortex were examined and every BrdU labelled cell was scored for whether it had another BrdU labelled cell within 1, 2, 3, 4 or >4 nuclear diameters of it. The percentages of cells within each of the five categories were calculated from six transplants of mutant or eight transplants of wild-type cells.
To quantify clumping of cells migrating from cortical explants in culture, explant images were analysed using an IPLab script (Scanalytics Inc., Fairfax, VA). Measurements were made by defining a series of concentric rings of equal width surrounding the explant, calculating their area and counting the numbers of touching and non-touching cells in each ring. A measure of the degree to which cells clump together, IC, independent of cell density, was obtained. IC is calculated based on the expected proportion of isolated cells (cells not touching another cell) under the null hypothesis that cells are distributed randomly around an explant. A scale of clumping is produced where IC = 0 if all cells are isolated (not touching another cell), IC = ∞ if all cells are touching one or more other cells, and IC = 1 if the proportion of isolated cells is equal to that expected under the null hypothesis that cells are randomly distributed. The model was subjected to tests for validity and shown to produce a measure of clumping that fulfilled the criteria for analysis of variance [ANOVA; A. Carothers, described by Pearson (Pearson, 1999)].
Figure 1A–C shows examples of the distributions of wild-type and Pax6–/– BrdU labelled (on E15.5) cortical cells in P7 wild-type cortex following transplants at E15.5. As shown before, both mutant and wild-type cells migrated preferentially to the superficial layers of the cortex, a location that was appropriate for their birthdate (Caric et al., 1997). Genotype had a clear effect on their tangential distribution. Wild-type cells were scattered throughout the cortex (Fig. 1A) whereas mutant cells were found in a small number of very dense clusters (Fig. 1B,C). These clusters were found throughout the rostrocaudal extent of the cortex. Combining quantitative data from eight wild-type and six mutant transplants showed that the vast majority of mutant cells were found within a single nuclear diameter of another mutant cell, whereas most transplanted wild-type cells were separated by much greater distances from other transplanted wild-type cells (Fig. 1D). The total numbers of transplanted mutant and wild-type cells identified in the analysis were comparable and represented tiny proportions of the overall numbers of cells in the recipient cortices. Analysis of transplanted wild-type and Pax6–/– cells labelled with fluorescent dyes revealed that they adopted neuronal morphologies and appeared viable (Caric et al., 1997).
After 24 h in culture, both wild-type and Pax6–/– cortical explants had extended numerous processes and cells had migrated out from the explant body (Fig. 2A,B). A striking difference in behaviour between cells from the wild-type and mutant tissue was observed. Whereas wild-type cells migrated individually or in association with one or two other cells (Fig. 2A), many mutant cells migrated together in streams and formed distinct clusters away from the body of the explant (Fig. 2B). ANOVA on values of IC revealed that, at both E13.5 and E15.5, clumping was significantly greater in cells migrating from mutant than from wild-type cortical explants (P < 0.001) (Fig. 2C,D). The analysis revealed no significant difference between co-culturing E13.5 or E15.5 explants (whether wild-type or mutant) with wild-type or mutant diencephalic tissue (Fig. 2C,D). This indicates that the clumping phenotype of Pax6–/– cortical cells is regulated independently of thalamic factors. Cellular clumping was not significantly different between anterior and posterior cortical explants at either age examined (data were combined for Fig. 2C,D). There were no differences in rates of cell death in explants of different genotypes (counts of dead cells were made on the basis of nuclear morphology after staining with fluorescent nuclear stains).
Immunostaining of cells migrating from the explants was performed to evaluate the cell types involved. Antibody TuJ1 stains the earliest born postmitotic neurons. Its use revealed that many cells migrating from both wild-type and mutant explants were neuronal and that some, but not all, cellular clumps contained neurons (Fig. 3A,B). Staining with an antibody against phosphorylated histone H3 marks metaphase cells (Estivill-Torrus et al., 2002). This antibody revealed that most cell division occurred in the body of both wild-type and Pax6–/– explants and very rarely within clumped cells, indicating that the clumps were not due to cell division after migration (Fig. 3C,D). Finally, RC2 antibody revealed radial glial cells extending from both wild-type and mutant explants and, in both strains, some cells appeared to be migrating along them (Fig. 3E,F).
The new data presented here indicate that Pax6 regulates the adhesive properties of cortical neurons. The transplant experiments show that cortical cells lacking Pax6 segregate from wild-type cells and form dense clusters, pointing to a difference in the cell surface properties of mutant and wild-type cells. The in vitro experiments indicate that Pax6–/– cortical cells have an increased tendency to aggregate with each other even in the absence of wild-type cells. This suggests that the cell-surface molecules whose expression is affected by Pax6 include cell adhesion molecules that regulate the absolute adhesiveness of cortical cells. The role of Pax6 in controlling these cellular properties appears to be cell-autonomous. Given the fundamental importance of cell adhesion in developmental processes, a defect of cell adhesion in Pax6–/– cells is likely to underlie many of the defective processes in mutant embryos.
Regulation of Cell Adhesion by Pax6
Other studies have indicated abnormalities in the adhesive properties of Pax6–/– cortical cells. Cortical cells express Pax6 whereas striatal cells do not and when cells from E12.5 to E14.5 wild-type cortex and striatum are co-cultured in a short-term assay they segregate strongly from each other (Stoykova et al., 1997; Götz et al., 1996). Pax6–/– cortical and striatal cells, however, segregate only weakly from each other (Stoykova et al., 1997). Pax6–/– cortical cells segregate from wild-type cortical cells but Pax6–/– striatal cells mix with wild-type striatal cells (Stoykova et al., 1997). These results suggest that Pax6 regulates the adhesive properties of cells in the telencephalic region where it is expressed, i.e. the cerebral cortex.
The expression patterns of some adhesion molecules are altered in Pax6–/– forebrains. Expression of the extracellular matrix molecule tenascin-C (TN-C) at the cortico-striatal boundary is abolished and the expression of calcium dependent adhesion molecules, cadherins, is altered in the cortex (Stoykova et al., 1997; Bishop et al., 2000). The expression domain of the homophilic adhesion molecule R-cadherin and that of Pax6 have some overlap (Ganzler and Redies, 1995; Matsunami and Takeichi, 1995) and, in the absence of functional Pax6, expression of R-cadherin mRNA is reduced considerably in areas that normally show co-expression (Stoykova et al., 1997).
Loss of Pax6 also seems to have an effect on the cell surface in E12.5–E13.5 mouse hindbrain. The migration of post-mitotic cells from the rhombic lip seems to be controlled in part by Pax6, whose actions may be mediated by regulation of the netrin receptor Unc5h3 (Engelkamp et al., 1999), although alterations in the polarity of cytoskeletal components may also be involved (Yamasaki et al., 2001). In small eye rats there is impaired migration of midbrain neural crest cells (Matsuo et al., 1993; Nagase et al., 2001). These cells use the frontonasal epithelium as a scaffold for their migration and frontonasal epithelial cells are known to express Pax6 (Matsuo et al., 1993). Within this region the cell surface molecule HNK-1 carbohydrate epitope and the gene encoding an enzyme for the synthesis of the HNK-1 epitope are expressed ectopically in small eye rats (Nagase et al., 2001). This suggests that the impairment of migration may be due in part to the inhibitory effect of the ectopically expressed HNK-1 epitope.
Changes in the adhesive properties of other cell types that express Pax6 have also been observed. Transgenic mice with an altered ratio of expression of different splice variants of Pax6 in the lens show a change in the expression of cell adhesion molecules (Duncan et al., 2000). Expression of P120 catenin (p120ctn), a member of the armadillo family of proteins implicated in cell–cell adhesion and signal transduction, and Paxillin, a focal adhesion adapter protein implicated in integrin mediated signalling pathways, are highly elevated. Expression of N cadherin and α-catenin are both slightly elevated, α5-integrin and b1-integrin accumulate in lens although E cadherin and a6-integrin expression appears normal (Duncan et al., 2000).
Studies using microarrays of eye mRNA from various Pax6 over-expressing and null mutant cells have shown changes in the expression of ∼400 genes (Chauhan et al., 2002). These included Paralemmin and Tangerin A (Chauhan et al., 2002), which are two putative cell surface molecules thought to be important in plasma membrane dynamics and cell process formation (Kutzleb et al., 1998; Agassandian et al., 2000; Chauhan et al., 2002). Changes in these cell surface molecules could also contribute to the altered adhesion seen in Pax6 mutant cells.
Some experiments have suggested that Pax6 protein may directly interact with the regulatory elements of genes encoding adhesion molecules. The gene for neural cell adhesion molecule (N-CAM) has a Pax6 paired-domain binding region within its promoter (Holst et al., 1997). In addition, Pax6 activates the expression of L1-luciferase reporter constructs in neuroblastoma cells (Meech et al., 1999). Although this and other studies suggests an interaction between Pax6 and the cell adhesion molecule L1, which regulates axonal guidance and fasciculation during development (Chalepakis et al., 1994; Meech et al., 1999; Honig et al., 2002), the interaction is likely to be complex. For example, the expression domains of Pax6 and L1 only partially overlap and it has been shown that there is no change in L1 expression in the intermediate, ventricular and subventricular zones of E19 Pax6–/– mice (Caric et al., 1997).
The expression pattern of Pax6 in the developing mouse eye closely parallels that of the retinoic acid-responsive transcription factor (AP-2α) (Koroma et al., 1997). Genes involved in cell–cell and cell-matrix adhesion have been shown to be regulated by AP-2α in vitro (Chalepakis et al., 1994; Fini et al., 1994; Chen et al., 1997; Holst et al., 1997). For example, AP-2α is required for activation of the E-cadherin promoter in epithelial cell cultures (Behrens et al., 1991; Hennig et al., 1996). Loss of AP-2α leads to a change in Pax6 expression in the developing eye (West-Mays et al., 1999). The discovery of a number of possible binding sites for AP-2α in the Pax6 promoter has generated the suggestion that Pax6 may be a required intermediary step for AP-2α controlled cell adhesion (Plaza et al., 1995).
Overall, our results and those of others provide compelling evidence that a central role of Pax6 is to regulate cell–cell interactions and adhesion at many sites in the developing embryo, including the cerebral cortex. Further work is required to elucidate the molecular pathways by which its influence on the cell surface is mediated.
We thank Paul Perry for the IPlab script and Andrew Carothers for help with statistical analyses. D.T. is supported by the MRC. P.R. is a Lister Research Fellow.