Division of Developmental Neuroscience, Center for Translational and Advanced Animal Research, Tohoku University School of Medicine, Sendai, Japan
Division of Developmental Neuroscience, Center for Translational and Advanced Animal Research, Tohoku University School of Medicine, 2-1, Seiryo-machi, Aoba-ku, Sendai 980-8575, Japan. Telephone: 81-22-717-8201; Fax: 81-22-717-8205
Pax6 is a highly conserved transcription factor among vertebrates and is important in various developmental processes in the central nervous system (CNS), including patterning of the neural tube, migration of neurons, and formation of neural circuits. In this review, we focus on the role of Pax6 in embryonic and postnatal neurogenesis, namely, production of new neurons from neural stem/progenitor cells, because Pax6 is intensely expressed in these cells from the initial stage of CNS development and in neurogenic niches (the subgranular zone of the hippocampal dentate gyrus and the subventricular zone of the lateral ventricle) throughout life. Pax6 is a multifunctional player regulating proliferation and differentiation through the control of expression of different downstream molecules in a highly context-dependent manner.
Disclosure of potential conflicts of interest is found at the end of this article.
Author contributions: N.O.: conception and design, financial support, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript; H.S.: collection and/or assembly of data, data analysis and interpretation, manuscript writing; K.N.-T.: data analysis and interpretation, manuscript writing; M.M.: collection and/or assembly of data, data analysis and interpretation.
Pax6 was first identified as a paired box (Pax) family member and cloned on the basis of its homology to the Drosophila gene paired . Pax6 is also responsible for a human eye disease called aniridia and for the mouse mutant Small eye [2, 3]. Pax6 is a highly conserved transcription factor containing two DNA binding domains, a paired domain (PD) and a paired-type homeodomain (HD) (Fig. 1A; a more detailed explanation is given below), and is crucial for the development of the central nervous system (CNS), eyes, nose, pancreas, and pituitary gland [4, –6]. Pax6 is expressed in specific spatiotemporal patterns during mammalian brain development and thus is thought to be upstream of gene networks involved in brain patterning, neuronal migration, and neural circuit formation (reviewed in [7, –9]).
The function of Pax6 in development has been studied by examining the phenotypes of several mutant lines of mice and rats, including mouse Sey and SeyNeu  and rat rSey  and rSey2 , all of which have mutations in the Pax6 gene. Although Pax6 heterozygous mutant mice/rats show a small-eye phenotype with various malformations in the eye [12, , –15], there is no apparent difference in gross brain anatomy between Pax6 heterozygous mutants and the wild-type (WT) littermates during the embryonic and neonatal stages. Homozygous mutant embryos lack the eyes and nose and die soon after birth [16, 17]. These mutant phenotypes are similar to an artificially targeted Pax6-deficient mouse (Pax6−/−) . The Sey mouse mutant contains a single base substitution that truncates the C-terminal half of Pax6 , and the rSey2 rat mutant has a single base inserted into the coding region adjacent to the paired domain of Pax6, leading to a frame shift and premature stop codon . The mRNA of Pax6 is expressed in both Sey mice and rSey2 rat homozygous mutants, whereas truncated Pax6 protein is not detected in the embryonic brain of rSey2 rat but is detected in that of Sey mouse . Since Pax6Sey mutation results in production of a truncated protein that lacks a DNA-binding homeodomain and the C-terminal activation domain, the truncated Pax6 is considered functionally inactive. Therefore, both Sey mouse and rSey2 rat homozygous mutants are widely used as Pax6-null mutants.
In this review, we focus on the role of Pax6 in neurogenesis, that is, the production of new neurons from neural stem/progenitor cells, especially in the embryonic telencephalon and postnatal hippocampus. Pax6 is crucial for maintaining these neural stem/progenitor cells, but, interestingly, it can also promote neuronal differentiation in context-dependent manners.
Role of Pax6 in Embryonic Neurogenesis
Pax6 Expression in CNS Development
Pax6 expression starts from the earliest stage of CNS development, that is, at embryonic day 8 (E8) in the mouse (corresponding to E10 in the rat), when the neural plate (i.e., the primordium of the CNS) consists entirely of proliferating neuroepithelial cells. At E10 in mice, when the neural tube is regionalized, Pax6 is expressed in the forebrain, hindbrain, and spinal cord (Fig. 1B) . In this review, we focus on the dorsal telencephalon, that is, the primordium of the cerebral cortex. Pax6 is expressed in the ventricular (germinal) zone (VZ) that lines the wall of the lateral ventricle but not in the small number of neurons in the marginal zone near the brain surface (Fig. 1C, 1D). These early-born neurons are Cajal-Retzius cells that are derived from the medial and lateral edges of the dorsal telencephalon (reviewed in ). The thickness of the VZ layer increases up to E12 (Fig. 1D), when production of glutamatergic pyramidal neurons reaches peak levels in the dorsal telencephalon. The cortical primordium also receives γ-aminobutyric acid (GABA)ergic inhibitory interneurons that originate from the ventral telencephalon [22, 23]. During these stages of embryonic neurogenesis, Pax6 expression persists in the VZ of the dorsal telencephalon and the boundary region between the dorsal and ventral telencephalon . At later stages, the VZ gradually becomes thinner and finally forms the ependymal layer that lines the wall of the lateral ventricle. The ependymal layer also expresses Pax6 (unpublished data).
Neural Stem/Progenitor Cells in Embryonic Neurogenesis
Histologically, the VZ predominantly contains “nuclei” of neuroepithelial cells (Fig. 1C, 1D), but at the cellular level, the neuroepithelial cells themselves have long apical and basal processes stretching between the ventricular surface and the basement membrane (Fig. 2A). As formation of new neurons (i.e., neurogenesis) proceeds, the neuroepithelial cells become longer and longer. These cells are also called “radial glia” because their processes locate radially within the cortical primordium and serve as a scaffold for newly born neurons to climb up toward the pial surface. The terminology of the radial glia is somewhat confusing; they are actually not simple glial cells but are rather currently recognized as neural stem/progenitor cells that produce neurons [25, 26]. However, interestingly, these radial glial cells exhibit astroglial properties at the molecular level [27, –29]. These neural stem/progenitor cells (i.e., neuroepithelial cells and radial glial cells) undergo mitosis at the apical ventricular surface [29, 30]. Each of these cells divides symmetrically to produce two neural stem/progenitor cells or asymmetrically to produce a daughter cell fated to become a neuron (or neurons) and a neural stem/progenitor cell [31, –33]. These neuroepithelial cells and radial glial cells that divide at the ventricular surface strongly express Pax6 (Fig. 1D–1F) . There is another type of neuronal progenitor cells that divide in the subventricular zone (SVZ) to produce two neurons. These are called “basal progenitors” or “intermediate progenitors” and originate from mitosis of radial glial cells that have lost the radial processes [31, –33]. Interestingly, the latter basal/intermediate progenitor cells are negative for Pax6 (Fig. 1F) but positive for other transcription factors (described below) [29, 35].
Transcriptional Regulation of Embryonic Neurogenesis
Many transcription factors are expressed in the cortical primordium [36, –38]. Among them, Neurogenin2 (Ngn2), a basic helix-loop-helix (bHLH) transcription factor, is expressed in the neuronal progenitor cells and strongly induces the production of glutamatergic neurons . Pax6 expression overlaps with that of Ngn2 in a small number of neuronal progenitor cells (Fig. 1E, inset), and Pax6 may regulate transcription of Ngn2 (; also described below). There is some overlapped expression of Pax6 with T-box brain gene (Tbr) 2, a T-domain transcription factor in the developing brain, but Tbr2 is expressed mainly in basal/intermediate progenitors (Fig. 1F) . Although sequential expression of the transcription factors is difficult to demonstrate directly in a single cell, there may be a transition of expression of the transcription factors from Pax6 → Ngn2 → Tbr2 → neurogenic differentiation (NeuroD) → Tbr1 . NeuroD is another bHLH transcription factor expressed in the neocortex and hippocampus and necessary for proliferation and postnatal differentiation of dentate gyrus (DG) neurons [41, 42]. Tbr1, another member of T-domain transcription factors, is expressed in all cortical pyramidal-projection neurons and regulates layer-related differentiation, cell migration, and axon pathfinding in the neocortex . Considering this sequential expression of transcription factors, Pax6 seems to be crucial in the earliest steps of cortical neurogenesis.
To Proliferate or to Differentiate, That Is the Question
In the above-mentioned Pax6 mutants, the development of the cortex is severely impaired; production of neurons is dramatically reduced, leading to a thinner cortical plate [43, , , , –48]. This finding has lead to the idea that Pax6 promotes “neuronal differentiation.” Supportive data for this model are provided by Götz et al.  and Heins et al. , who assayed the effects of transfected Pax6 in cultured cells isolated from the developing cortex. Introduction of the Pax6-expression vector induced more neurons than in control.
Pax6 mutant cortical primordia simultaneously show a paradoxical phenotype: a thinner VZ, with reduced expression of PCNA, a marker for proliferating cells . From a detailed analysis of cell-cycle kinetics, Estivill-Torrus et al. suggested that the absence of Pax6 accelerated cortical neurogenesis in vivo . More recently, Quinn et al. used chimera techniques to show that underproduction of neurons in the Sey/Sey cortex is caused by early depletion of the progenitor pool via abnormally high proportions of newly divided cells exiting the cell cycle . They have proposed a model in which Pax6 functions early in cortical development to prevent precocious neuronal differentiation and depletion of the progenitor pool and to induce normal development of basal/intermediate progenitor cells by regulating the expression of Tbr2 . Our previous finding that the formation of later-born neurons is more affected in rSey2/rSey2 cortical primordium  is in accordance with this model. This effect on later-born neurons in Pax6 mutants may be due to depletion of the neurogenic progenitor pool during the preceding days, as suggested by Quinn et al. .
Similarly, a study on the development of the retina, which expresses Pax6 and is considered a part of the forebrain, reported precocious neuronal differentiation with upregulation of Mash1, another proneural bHLH factor, in Sey/Sey embryos . In adult mammalian eyes, Pax6 is required for the proliferation and expansion of retinal stem cells . Another report, however, showed that neuroepithelial progenitors are more abundant than normal in the Pax6 mutant retina , conflicting with the data of Philips et al.  and Xu et al. . This intriguing issue of Pax6 as a binary player promoting cell proliferation or cell differentiation in a context-dependent manner is discussed in the next section.
Molecular Mechanisms Underlying Multiple Functions of Pax6
Regulation Based on Expression Levels and Isoforms of Pax6
The mammalian Pax6 locus encodes at least three isoforms: the “canonical” Pax6, Pax6(5a), and Pax6(ΔPD). The canonical Pax6 protein contains a PD near the N terminus, a paired-type HD in the middle, and a proline/serine/threonine-rich (PST) domain at the C terminus (Fig. 1A). Both the PD and the HD have DNA binding activities, whereas the PST domain possesses a transactivation function [55, , –58]. Pax6(5a) is produced through alternative splicing retaining exon 5a and contains a 14-residue insertion in the N-terminal subdomain of PD, which alters the DNA binding specificity [57, 59] (Fig. 1A). Both Pax6 and Pax6(5a) mRNAs are expressed in the telencephalon, diencephalon, and hindbrain, and interestingly, the expression levels of Pax6 in these tissues are 6–10 times higher than that of Pax6(5a) during early neurogenesis . The third Pax6 isoform, which lacks PD (Fig. 1A), arises from an alternative promoter between exons 4 and 5 and an alternative translation start codon in exon 7 [61, 62]. Although Pax6(ΔPD) mRNA is also detected in the embryonic forebrain, its expression level is much lower than that of Pax6 . Overexpression of Pax6(ΔPD) causes a severe microphthalmic phenotype that is due to apoptotic cell death in the lens during embryonic development [62, 64], whereas the role of Pax6(ΔPD) in brain development, if any, is unknown at present. Haubst et al. showed that the PD of Pax6 is required for regulation of neurogenesis, cell proliferation, and patterning by comparing the telencephalic development in several lines of mutant mice harboring mutations in distinct DNA binding domains of Pax6 . Furthermore, retroviral overexpression of Pax6(5a) inhibited cell proliferation but not cell fate, whereas Pax6 affected both proliferation and cell fate . Therefore, the PD of Pax6 is crucial in normal brain development.
With regard to the expression of Pax6 locus, at least six enhancer regions and three promoters (P0, P1, and Pα) have been identified (reviewed in ). Both canonical Pax6 and Pax6(5a) mRNAs are produced by transcription from both P0 and P1 promoters, whereas the transcript from Pα promoter encodes Pax6(ΔPD) [61, 62]. However, little is known about the molecular mechanisms that control transcriptional levels of distinct Pax6 isoforms in CNS development. Interestingly, several putative Pax6 binding sites are often present in different species in a position equivalent to that of the Pax6 gene . Within the head surface ectoderm-specific enhancer of Pax6 gene, it has been demonstrated that Pax6 directly interacts with Pax6-responsive element and positively regulates its own gene expression . On the other hand, crossing Pax6 yeast artificial chromosome (YAC) transgenic mice with a Pax6-tau-GFP reporter YAC transgenic line showed that increased expression of Pax6 is limited by negative autoregulation . These studies suggest that Pax6 levels are stabilized by positive and negative autoregulation in vivo. It is reported that proliferation of the cortical progenitors is sensitive to altered Pax6 levels, whereas cortical regionalization is not . Activated expression of either Pax6 or Pax6(5a) inhibits progenitor proliferation in the developing cortex . Upon activation of transgenic Pax6, specific progenitor pools with distinct endogenous Pax6 expression levels at different developmental stages show defects in cell cycle progression and in the acquisition of apoptotic or neuronal cell fate . In the mouse optic vesicle, Pax6 overexpression also represses proliferation of neuroepithelial cells . Taken together, the isoforms of Pax6 and their expression levels are important for the proliferation of neuroepithelial progenitors in the cortical and retinal primordia.
Partners of Pax6 in Transcriptional Regulation
The complexity of Pax6 function could be based not only on the internal DNA-binding domain usages in individual isoforms but also on the capability of Pax6 to interact with various transcription factors and to synergistically regulate target gene expression. In lens development, transcriptional regulation of several crystallin genes by Pax6 is coordinately achieved with other transcription factors, such as Sox2 and Maf (reviewed by Cvekl et al.  and Kondoh et al. ). For example, Pax6 forms a co-DNA-binding complex with Sox2, a high-mobility group domain-containing transcription factor, for transcriptional activation of the δ-crystallin gene . In case of the above-mentioned positive autoregulation of Pax6 gene, Sox2 and Sox3 activate the enhancer synergistically with Pax6 and form a complex with Pax6 also . Synergistic activation of the glucagon gene promoter by Pax6 and Cdx2 and by Pax6 and Maf transcription factors is also reported in the pancreas [73, 74]. Thus, it is highly likely that Pax6 functions together with cobinding partners during neurogenesis. A good candidate is Sox2, since it is expressed in the developing mouse CNS from an early stage  and regulates the expression of fibroblast growth factor 4 (Fgf4) and nestin, which are important in maintaining neural stem cells . POU transcription factors, which are expressed during neurogenesis, control the expression of two CNS stem cell-specific genes, nestin and brain-type fatty acid binding protein (B-Fabp, or Fabp7; described below) . Sox2 and POU transcription factors synergistically activate Fgf4 and nestin [78, 79]. Therefore, it would be interesting to know whether there is also interplay between Pax6 and Sox2 or Pax6 and POU factors in neurogenesis. Expression of Tlx, a forebrain-restricted transcription factor, also overlaps with that of Pax6 , and genetic data suggest that Pax6 and Tlx cooperate in the formation of the dorsoventral boundary in the telencephalon  and in specification of later-born (upper-layer) neurons . It will be important to elucidate their mutual target genes in different phases of corticogenesis. Furthermore, cooperative transcriptional regulation by Pax6 with these factors should be elucidated in more detail in the future.
Although Pax6 is primarily a transcriptional activator, the molecular mechanisms of “negative” transcriptional regulation by Pax6 with some other transcription factors have recently been identified. The MAD homology 1 domain of Smad3, which is a direct mediator of TGF-β signaling, represses autoregulation of the Pax6 P1 promoter by interaction with the PD of Pax6, thereby releasing Pax6 from the promoter element . In a glucagon-producing pancreatic islet cell line, peroxisome proliferator-activated receptor γ binds to the transactivation domain of Pax6 in a ligand- and retinoid X receptor-dependent but DNA-binding-independent manner and blocks glucagon gene transcription by Pax6 . NK homeobox 6.1 (Nkx6.1), a β-cell-specific transcription factor, inhibits glucagon gene transcription in a different manner, by competing for its Pax6-binding site . Thus, Pax6 could play a role in inhibition/repression of transcription under circumstances where the interaction between Pax6 and another factor interferes with the DNA-binding ability or transcriptional activity, also during neurogenesis.
In spinal cord development, loss of Pax6 results in premature specification of oligodendrocytes and astrocytes due to ectopic expression of oligodendrocyte transcription factor 1 (Olig1), Olig2, Nkx2.2, and glutamine synthase . There seems to be mutual repression between Pax6 and these transcription factors, but the molecular mechanisms responsible for this repression have not yet been elucidated.
Downstream Effectors of Pax6 in Embryonic Neurogenesis
Because Pax6 has multiple roles in brain development, many groups, including ours, have looked for genes downstream of Pax6 in the developing CNS. However, there are only few examples that have been strictly proven as direct Pax6 target genes. In this section, we introduce Pax6 downstream molecules, including both direct and indirect targets, in CNS development (summarized in Fig. 2B).
Using microarray analysis, our group identified Fabp7 as a downstream gene of Pax6 in the E12.5 rat cortex . FABP7 (B-FABP/brain lipid binding protein) is a member of the fatty acid-binding protein family and is widely used as a marker for neural stem/progenitor cells (i.e., neuroepithelial cells or radial glia) [77, 88, 89]. Inhibition of FABP7 function by RNA interference in the developing cortex dramatically reduced cell proliferation and conversely increased production of immature neurons , indicating that FABP7 is required for the maintenance of proliferating neuroepithelial cells. Expression of Fabp7 in early mouse embryos is drastically different from that of rat embryos; Fabp7 is more intensely expressed in the ventral than the dorsal telencephalon . Transcriptional regulation of the mouse Fabp7 gene is governed by the Notch pathway, which is critical for maintaining stem cells . Therefore, although regulation of Fabp7 expression is different in mice and rats, at least in the early stage of development, FABP7 could generally function to keep neural stem cells in the CNS, which is controlled by Pax6 in the rat cortex primordium. We are now trying to demonstrate that Pax6 directly transactivates rat Fabp7 gene in neural stem/progenitor cells by in vivo and in vitro assays.
It is known that the Ngn2 gene, coding an already-mentioned proneural bHLH factor, is a direct target of Pax6 [40, 91]. Ngn2 E1 enhancer activity in the spinal cord is almost completely abolished in E1-LacZ transgenic Sey/Sey mouse embryos , and a Pax6 binding element has been identified by electrophoretic mobility shift assay (EMSA) within the E1 enhancer . In addition, the retina-specific E2 enhancer of Ngn2 gene also contains Pax6-responsive element activated in multipotent retinal progenitor cells . When Pax6 is electroporated into the dorsal telencephalon of cultured rat embryos, induction of Ngn2 is observed 3 hours later but disappears by 18 hours, whereas Fabp7 shows continuous induction . This may be due to differential levels of Pax6 protein needed for these two genes to be activated; induction of Ngn2 expression may need a higher Pax6 protein level, which cannot be maintained for 18 hours, whereas a lower Pax6 protein level may be sufficient to induce Fabp7. Another possibility is that transcription of Ngn2 may be induced by Pax6 binding, but Ngn2 protein may subsequently downregulate Pax6 expression. In the posterior neural tube, neuronal precursors maintaining Pax6 expression fail to differentiate, whereas turning off Pax6 function in these precursors is sufficient to provoke premature differentiation . Importantly, Pax6 expression involves negative feedback regulation by Ngn2, and this repression is critical for the proneural activity of Ngn2 . Immunostaining of the mouse cortex primordium using Pax6- and Ngn2-specific antibodies showed that neuroepithelial progenitors that strongly expressed Ngn2 were negative for Pax6 (Fig. 1E, arrowheads), supporting the finding that Ngn2 negatively regulates Pax6. There may be a similar repression of Pax6 by Tbr2, as these proteins show no overlapping expression .
Pax6 is expressed in neuroepithelial cells (neural stem/progenitor cells) of the optic cup (i.e., the primordium of the retina). In the mouse optic vesicle of Pax6 mutant embryos, the number of neuroepithelial progenitors was higher than normal . In vitro, Pax6-null retinal spheres overproliferate and display reduced expression levels of several negative regulators of the cell cycle, such as p16Ink4a, p19Arf, p27kip1, p57kip2, and p21cip1. Pax6 overexpression repressed cellular proliferation and secondary sphere formation, suggesting that Pax6 acts cell-autonomously on the cell cycles of neuroepithelial progenitors. The presence of aberrant numbers of mitotic cells and low expression levels of p27kip1, p57kip2, and p21cip1 were verified in the primitive forebrain of Pax6 mutants. In a chromatin immunoprecipitation (ChIP) assay, Pax6 binds to the genomic DNA upstream of the p27kip1 gene encoding a member of the cyclin-dependent kinase (cdk)-interacting protein/kinase inhibitor protein (Cip/Kip) family of the cdk inhibitors that negatively regulate G1 cell cycle progression . Therefore, Pax6 controls not only the proliferation rate of neuroepithelial progenitors in the optic vesicle but also that in the cerebral cortex through a direct control of cell cycle regulators such as p27kip1.
Lewis X, FucT9, and Wnt7b.
Other interesting molecules related to neural stem/progenitor proliferation/differentiation include fucosyl transferase IX (FucT9) and CD15/Lewis X glycosyl moiety, which is synthesized by FucT. Lewis X is another marker for some but not all of the population of neural stem cells . We previously reported that the expression of FucT9 and localization of Lewis X were decreased in the cortical primordium of rSey2/rSey2 rat embryos , although it is not clearly discerned yet whether Pax6 directly regulates FucT9 gene expression. Interestingly, FACS cells with high Lewis X expression contain more stem cells as assayed in neurospheres [96, 97]. Moreover, expression of Wnt7b is decreased in the developing cortex, hindbrain, and spinal cord of Pax6 mutant mice and rats , although the roles of Wnt factors on proliferation and differentiation of neural stem/progenitor cells are still debated [98, , , –102]. Taken together, Pax6 may comprehensively regulate factors that are important for proliferation of neural stem/progenitor cells.
Cell Adhesion Molecules.
Pax6 also controls cell adhesion. Transplantation analyses have revealed that cells isolated from Sey/Sey cortex were segregated from the WT host cells and formed dense clusters . L1 gene, coding a cell adhesion molecule of the Ig superfamily, was the first example of a direct target of Pax6 in the CNS. Pax6 binds to an element within the bone morphogenetic protein-responsive region in the first intron, called the homeodomain and paired domain binding site, and transactivates L1 gene . Optimedin (olfactmedin 3) gene is also a direct downstream target of Pax6 . Optimedin A variant is expressed in the eye and brain, and Pax6 binding to the Optimedin A promoter was confirmed by ChIP assay in vivo . Optimedin is reported to induce the expression of N-cadherin, β-catenin, α-catenin, and occludin and to stimulate Ca2+-dependent aggregation of nerve growth factor-stimulated PC12 cells , suggesting that Pax6 regulates cell-cell adhesion and/or cell attachment to the extracellular matrix indirectly via optimedin gene expression. In the dorsal telencephalon, expression of R-cadherin, one of the Ca2+-dependent cell-cell adhesion molecules expressed in the mouse fetal brain, is markedly reduced [107, 108], thereby regulating boundary formation in the developing forebrain. However, it is still unknown whether transcriptional regulation of R-cadherin by Pax6 is direct or indirect. Expression of δ-catenin, a neural-specific member of the Armadillo-domain subfamily  and involved in neuroepithelial cell adhesiveness, was also decreased in the Sey/Sey telencephalon in a microarray analysis conducted by Duparc et al. . Binding of Pax6 to a conserved site on the δ-catenin gene was confirmed by EMSA . Since isolated single cells easily differentiate into neurons by default, regulation of adhesiveness in neuroepithelial cells may be also important to maintain proliferation. Moreover, in Pax6-deficient mice, expression of Tenascin C, an extracellular matrix glycoprotein, is lost in the cortex primordium . Tenascin C is widely present in the vertebrate central nervous system during development and repair, and it functions in both neural progenitor proliferation and differentiation of neurons and oligodendrocytes [111, 112]. Furthermore, Pax6 acts as a modulator of alternative splicing and differentially regulates the expression of 20 Tenascin C isoforms . In the transgenic mice that overexpress Pax6(5a) using αA-crystallin promoter, the expression of α5β1 integrin, paxillin, and p120ctn in the lens is upregulated . Furthermore, putative Pax6/Pax6(5a) binding sites in the human α5- and β1-integrin promoters were identified by EMSA . Thus, Pax6 directly or indirectly regulates several aspects of cell-cell and cell-substrate adhesion during development.
In the neural tube, Pax6 regulates the expression of secreted frizzled-related protein 2 (sFRP2) and T-cell factor 4 (Tcf4), both of which are involved in Wnt signaling . sFRP2 is an extracellular inhibitor of Wnt signaling, and Tcf4 is a member of the Tcf/lymphoid enhancer-binding factor family of transcriptional mediators of canonical Wnt signaling. Pax6 represses Nkx2.2 expression in the ventral spinal cord by regulating the spatially restricted expression of sFRP2 and Tcf4 . Therefore, Pax6 controls ventral patterning by regulating Wnt signaling in neuronal progenitors. In contrast, homeobox (Hox) transcription factors control anteroposterior patterning of the neural tube. ChIP assays revealed that Pax6 associates with Hoxd4 gene enhancer and regulates the expression of Hoxd4 in the mouse embryo . Moreover, Hoxd4 expression is decreased throughout the CNS in Pax6-deficient mouse embryos . Therefore, Pax6 can also regulate anteroposterior-restricted Hoxd4 gene expression. Thus, Pax6 is also important for regulation of both dorsoventral and anteroposterior patterning in the CNS.
Other Transcription Factors.
Moreover, in microarray analyses of tissues from Pax6-dependent (dorsal) or -independent (ventral) telencephalon at E12 and E15, Holm et al. reported the finding of lower expression of various transcription factors involved in neurogenesis (such as satb2, Nfia, AP-2γ, NeuroD6, Ngn2, Tbr2, and Bhlhb5) and of a retinoic acid signaling molecule, Rlbp1, in the dorsal telencephalon of Pax6-deficient mice than in that of the WT . This analysis elucidated time- and region-specific differences in Pax6-mediated transcription, explaining differential functions of Pax6 at early and later stages of neurogenesis. Therefore, Pax6 is a multifunctional player in embryonic neurogenesis by regulating, directly and/or indirectly, the expression of various molecules.
Altogether, Pax6 is a binary player in embryonic neurogenesis, promoting cell division and cell differentiation in a highly context-dependent and complex manner in terms of different protein expression levels, different isoforms, different copartners, different direct target genes, and different indirect downstream effectors.
Pax6 in Postnatal Neurogenesis
Postnatally, strong expression of Pax6 is detected in neurons in various brain regions, including the olfactory bulb, amygdala, thalamus, and cerebellum [119, , , –123]. Modest expression is seen in the subgranular zone (SGZ) of the hippocampal DG [124, 125] and in the ependymal layer and the SVZ of the lateral ventricle [126, 127], the two areas in which neurogenesis takes place throughout life [128, 129].
In the SGZ, Pax6 is expressed in neural stem/early progenitor cells that show radial glia-like morphology and are positive for glial fibrillary acidic protein (GFAP) and nestin, as well as in a small population of late progenitors that are positive for polysialylated neural cell adhesion molecule (PSA-NCAM) [124, 125]. GFAP is also a marker for astrocytes, which contribute to approximately 40% of Pax6-expressing cells in the DG (Fig. 3C). An analysis combining 5′-bromo-2′-deoxyuridine-pulse/chase labeling and immunostaining showed that Pax6-expressing cells represent less than 20% of slowly renewing neural stem cells (Fig. 3). On the basis of marker analyses, Hevner et al. have proposed that the transitions from Pax6 expression in radial glia-like progenitors to Tbr2 and NeuroD expression in intermediate stage progenitors and to NeuroD/Tbr1 expression in postmitotic neurons may represent important steps in hippocampal neurogenesis .
Expression of Pax6 is induced in the hippocampus in response to brain injury  as a result of cell proliferation of neural stem/progenitor cells caused by the lesion. Our group has shown that cell proliferation of neural stem/progenitor cells and subsequent production of neurons are considerably reduced in the hippocampus of Pax6-haploinsufficient rats  and mice . There seems to be an accelerated transition to neurogenic stages, since more cells coexpress GFAP and PSA-NCAM in rSey2/+ . Eventually, the granule cell layer in the DG becomes thinner as time passes, although the gross anatomy of the brain is not dramatically different from that of the WT.
FABP7, an important downstream molecule of Pax6 , is also expressed in cells with a radial-glial phenotype in the hippocampus . Interestingly, Pax6 seems to regulate the expression of Fabp7 in the SGZ in both the mouse and rat (M.M., unpublished data). In addition, we have shown that FABP7 indeed functions to keep neural stem/progenitor cells proliferating during hippocampal neurogenesis . Thus, regulation of Fabp7 by Pax6 is fundamental in both embryonic and hippocampal neurogenesis.
The SVZ in the lateral ventricle is another niche for neurogenesis. Astrocytes in the SVZ produce neural progenitors that transiently divide and migrate toward the olfactory bulb, where they become mature neurons within several weeks (reviewed by Alvarez-Buylla and Lim  and Lledo and Saghatelyan ). During these long-lasting processes, strong expression of Pax6 persists in migrating immature neurons as well as in mature neurons with heterogeneous character. Pax6 function is required for specification of dopaminergic periglomerular cells and for GABAergic granule cells in the superficial layer [126, 127]. The number of dopaminergic periglomerular cells are increased by overexpression of Pax6 in the adult neuronal progenitors  and decreased in Pax6 heterozygous mutant mice [127, 135]. These data suggest that a proper dose of Pax6 is essential for differentiation and/or maintenance of various olfactory interneurons . How Pax6 modulates proliferation or differentiation, through the regulation of different downstream effectors, during neurogenesis in the hippocampal SGZ and the lateral ventricular SVZ would be an issue of great interest in future studies.
In closing this section, we would like to point out, as also mentioned in the review by Hevner et al. , that there are considerable molecular similarities between embryonic and adult neurogenesis, in both of which Pax6 functions in the early step to balance proliferation of neural stem cells and differentiation into neurons.
As Scott Gilbert cited various findings related to Pax6 in his great textbook Developmental Biology , this transcription factor is one of the most intensively studied developmental molecules. We have recently reported that Pax6 is required for normal interkinetic nuclear movement (INM) in cortical neuroepithelial cells . This may lead to a breakthrough in our understanding of the molecular machinery of INM, which was discovered nearly a half century ago . Enhancer elements of Pax6 are also complex but interesting [66, 140], although elements for activity in postnatal neural stem/progenitor cells are not yet known. Finally, understanding how Pax6 switches on different target genes from neurogenesis to gliogenesis is a challenging issue for the future.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Drs. Ron McKay, Wieland Huttner, Robert Hevner, François Guillemot, Hideyuki Okano, Tetsuya Taga, Ryoichiro Kageyama, Kazunori Nakajima, and Tatsunori Hisatsune for thoughtful discussions at the Neurogenesis 2007 meeting held in Tokyo and organized by Dr. Tatsunori Hisatsune and N.O. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas “Molecular Brain Science” from Ministry of Education, Culture, Sports, Science and Technology (MEXT), the Core Research for Evolutional Science and Technology program of the Japan Science and Technology Agency, and the Global Center of Excellence program of MEXT.