Molecular dissection of Pax6 function: the specific roles of the paired domain and homeodomain in brain development

The transcription factor Pax6 is conserved from invertebrates to vertebrates and regulates key developmental processes in several organs throughout phylogeny (Chi and Epstein,2002). In all of these organs, Pax6 acts as a potent cell fate determinant, and often affects cell proliferation, yet it is not well understood how Pax6 coordinates these diverse effects at the molecular level in vivo. Pax6 contains two DNA-binding domains, the paired domain (PD) and a paired-like homeodomain (HD) linked via a glycine-rich domain, and acts as an activator of gene expression by a proline-, serine- and threonine-rich (PST) domain at the C-terminus(Bouchard et al., 2003). Both the PD and HD bind to specific DNA target sites(Chi and Epstein, 2002) and may also influence each other's binding to the DNA or cooperate on DNA binding(Duncan et al., 1998; Jun and Desplan, 1996; Mikkola et al., 2001; Mishra et al., 2002; Singh et al., 2000). Several splice variants have been discovered for the Pax6 gene(Anderson et al., 2002; Duncan et al., 2000; Epstein et al., 1994; Kozmik et al., 1997; Mishra et al., 2002) resulting predominantly in a truncated form lacking the paired domain (`paired-less')and a form with a 14 amino acid (AA) insert into the PD [Pax6(5a), see Fig. 1]. The canonical PD(without the 14 AA insert) binds via its N-terminal PAI domain(Epstein et al., 1994) to DNA,while the insertion of the 14 AA from exon 5a (now designated as exon 5, but here still referred to as 5a) into the PAI domain [PD(5a)] abolishes its DNA-binding and favors DNA-binding of the C-terminal RED domain(Epstein et al., 1994)(Fig. 1A). Distinct consensus sites have been identified for these domains, with the N-terminal PAI domain binding to both consensus sites (P6CON; 5aCON), while the C-terminal RED domain can only activate targets with the 5aCON site(Epstein et al., 1994)(Fig. 1A). Notably, the Pax6(5a) form is required for aspects of postnatal differentiation in the eye(Singh et al., 2002), and an important role of the same modification of PD-mediated DNA binding was recently discovered in the developing compound eye of Drosophila(Dominguez et al., 2004).

Despite the crucial function of Pax6 in the developing brain, nothing is known about the specific roles of its DNA-binding domains, since analysis has so far been performed exclusively with the functional null alleles Pax6^Sey, Pax6^Sey-Neu in mice(Hill et al., 1991; Schmahl et al., 1993) and the Pax6^Sey rat (Matsuo et al., 1993), all of which contain a truncated Pax6 protein lacking the transactivator domain (TA) (Fig. 1E). The strongest effects of Pax6 mutation in the brain are manifested in the forebrain (Schmahl et al., 1993), where Pax6 expression starts at embryonic day (E) 9 in the mouse (Walther and Gruss,1991). In the telencephalon, Pax6 is restricted to the dorsal region, and this region is misspecified in the Pax6 mutants. Severe defects in patterning are indicated by the expansion of ventral transcription factors into the dorsal telencephalon or the loss of boundary structures restricting cell migration between these regions of Pax6^Sey–/– mice(Chapouton et al., 1999; Heins et al., 2002; Stoykova et al., 1996; Stoykova et al., 1997; Stoykova et al., 2000; Toresson et al., 2000; Yun et al., 2001). Notably these Pax6 mutations also affect patterning in the hindbrain, spinal cord and the ventral diencephalon (Briscoe et al., 1999; Grindley et al.,1997; Osumi, 2001; Osumi et al., 1997; Stoykova et al., 1996)implying a widespread role of Pax6 in patterning.

In contrast, the recently discovered potent neurogenic role of Pax6 appears more restricted to the telencephalon, and seemingly contributes to the maintenance of neurogenesis into adulthood exclusively in this part of the brain (Alvarez-Buylla et al.,2001; Gage, 2002; Hack et al., 2004; Heins et al., 2002; Nakatomi et al., 2002). The absence of functional Pax6 not only severely impairs neurogenesis in the dorsal telencephalon, its overexpression is also sufficient to drive embryonic precursors, postnatal cortical astrocytes and adult neural stem cells towards neurogenesis (Hack et al.,2004; Heins et al.,2002). In addition, Pax6 is involved in the reduction of proliferation in the telencephalon, since its loss-of-function results in an increased number of precursors(Estivill-Torrus et al., 2002; Götz et al., 1998; Heins et al., 2002), while its gain-of-function reduces the number of progeny generated by a single progenitor cell (Heins et al.,2002). To gain a better understanding of potentially separate roles exerted by distinct DNA-binding domains of the Pax6 transcription factor, we examined here the function of its three DNA binding domains, the PD, PD(5a) and HD in the developing forebrain.

Pax6^Sey mice were maintained as heterozygotes on a mixed C57BL/6JxDBA/2J background, Pax6^Aey18 mice (J.G.,unpublished) and Pax6^4Neu mice were maintained on a C3HeB/FeJ background). Pax6^tm1Gfs mice are referred here as Pax6(5a)–/– mice(Singh et al., 2002) and were maintained as homozygotes on a C57BL/6 background. Analysis was performed with embryos from embryonic day (E)10-16.

Embryonic brains were fixed in 4% paraformaldehyde in PBS (PFA) and sections were cut with a vibratome or with a cryostat after cryoprotection. Sections were processed for immunohistochemistry(Götz et al., 1998; Hartfuss et al., 2001), using the primary antibodies against the phosphorylated form of Histone H3 (PH3,rabbit, Biomol, 1:200), NeuN (mouse IgG1, 1:50, Chemicon), Mash1 (mouse IgG1,1:2, kindly provided by F. Guillemot), Gsh2 (rabbit, 1:1000, kindly provided by K. Campbell), Ngn2 (mouse IgG2a, 1:10, kindly provided by D. Anderson),Olig2 (rabbit, 1:1000, kindly provided by D. Rowitch), Ki67 (rat Tec-3, 1:50,Dako), β-galactosidase (rabbit, 1:300, Cappel; mouse IgG2a, 1:500,Promega), O4 (mouse IgM, 1:1000, kindly provided by J. Price), GFAP (mouse IgG1, 1:200, Sigma), RC2 (mouse IgM, 1:500, kindly provided by P. Leprince),Pax6 (rabbit, 1:300, Babco), reticulon1 (mAb 9-4, 1:10, kindly provided by T. Hirata) and nestin (mouse IgG1, 1:4, Dev. Hybridoma Bank). The respective secondary antibodies were used from Jackson Immunoresearch, Inc. and Southern Biotechnology Associates, Inc. Specimens were mounted in Aqua Poly/Mount(Polysciences, Northampton, UK) and analyzed at a Confocal Microscope (Leica TCS 4NT; Leica Microsystems; Heidelberg). Digoxigenin-labeled RNA probe for SFRP2 was made and used as described(Chapouton et al., 2001).

The entire coding sequence of Pax6 (1873bp fragment) and Pax6(5a) (1915 bp fragment) was cloned in sense orientation into the BglII unique restriction site of the retroviral vector 1704 between the upstream LTR and the EMC IRES sequence (gift of J. E. Majors)(Ghattas et al., 1991). BOSC23 helper-free packaging cells (Pear et al.,1993) were used for viral production by means of transient transfection with the respective viral plasmid and resulted in a typical titre of 1×10⁵/ml. Primary cells from cerebral cortex were isolated and cultured as described previously(Heins et al., 2002), and infected 2 hours after plating at a concentration giving no more than 50 clones per coverslip and analyzed after 1 week in vitro.

The quantification of neurons was performed in sections of the cerebral cortex at E14 from wild-type (WT) and homozygous mutants at corresponding rostral, intermediate and caudal levels stained for NeuN. The thickness of the NeuN-positive cortical plate was measured in confocal pictures taken in a defined (lateral) area of the cortex by first drawing a line at right angles to the ventricular surface (VS) from the VS to the pial surface. The length of this line was determined by the ImageJ program and served as measure of the total cortical thickness. A second line was drawn from the apical and basal side of the NeuN-positive band, and its length served as a measure for the thickness of the band of neurons, the cortical plate. The width of the cortical plate was calculated as the proportion of the overall thickness of the cerebral cortex (Table 1).

The quantification of PH3-labeled cells was performed by placing a 150-μm-wide square covering the entire cortical thickness parallel to the VS. All PH3-positive cells in this square were counted separately for cells lining the ventricle (VZ cells) and cells at abventricular positions (SVZ cells defined as PH3+ cells, five or more cell diameters away from the VS)(see Smart, 1976)(Table 1).

For the clonal analysis in vitro, all clones per coverslip were assessed for their cell type or size (Fig. 5) and the mean was calculated per coverslip (excluding coverslips with more than 50 clones; Fig. 5C,D). Group comparisons were made with the unpaired t-test and P-values smaller than 0.05 were considered significant (^*), P-values smaller than 0.001 were considered highly significant (^***).

To assess the role of the PD of Pax6 we analyzed the Pax6^Aey18 mutant mice, which are characterized by the loss of a splice acceptor site in front of exon 5a such that exon 5a and 6 are excluded and a large part of the PD is deleted, whereas the HD and TA remain unaffected (Fig. 1B) (J.G.,unpublished). To examine specifically the role of the Pax6(5a) isoform we used the Pax6^tm1Gfs mice [here referred to as Pax6(5a)–/–] carrying a targeted deletion of the exon 5a(Fig. 1C)(Singh et al., 2002). The Pax6^4Neu mutation carries a point mutation in the third helix of the HD that abolishes its DNA-binding(Fig. 1D)(Favor et al., 2001). These mutants were compared to the Pax6^Sey–/– mice,characterized by a truncated form of Pax6 lacking the TA domain(Fig. 1E), which has been identified as a functional null by comparison to the Pax6–/– (e.g. Götz et al., 1998; Briscoe et al., 1999).

Neurogenesis was examined in these mutants by immunohistochemistry for neuronal markers, such as NeuN (Fig. 2) or βIII-tubulin (data not shown). Neurons accumulate in the cortical plate (CP) forming a distinct band underneath the pial surface(Fig. 2A) at midneurogenesis,embryonic day (E) 14, in the WT cortex. The NeuN-positive band is distinctly smaller in the cortex of homozygous Pax6^Aey18–/– mouse mutants(Fig. 2A,B), reflecting a significant decrease in the number of neurons generated(Table 1). Since the neurogenesis phenotype of Pax6^Aey18–/– mice was reminiscent of the functional null allele Pax6^Sey(Fig. 2E), we quantified the thickness of the cortical plate relative to the total cortical thickness (see Materials and methods) in the lateral cortex at rostral, intermediate and caudal levels (Table 1). Notably the significant reduction of the thickness of the NeuN-positive band compared to WT in the Pax6^Aey18–/– cortex(P<0.001) and the comparable reduction observed in Pax6^Sey–/– cortex(Table 1), suggests that Pax6^Aey18 is a functional null allele for cortical neurogenesis. In contrast, we found no reduction in neurogenesis in Pax6(5a)–/– mice(Table 1; Fig. 2C) or Pax6^4Neu–/– mice(Table 1; Fig. 2A,D). Thus, the PD, but not the HD of Pax6 is required for the regulation of neurogenesis in the cerebral cortex.

Reduction of neurogenesis in the developing telencephalon is accompanied by an increase in cell proliferation in mutants bearing functional null alleles of Pax6 (Estivill-Torrus et al.,2002; Götz et al.,1998). The number of precursors was determined by immunostaining against the antigen Ki67 labeling actively dividing precursors in all phases of the cell cycle (Gerlach et al.,1997) and the phosphorylated form of histone H3 (PH3)(Hendzel et al., 1997). Since PH3 labels all precursors in late G2 and M-phase(Fig. 2), we could distinguish between precursors located in the VZ, that undergo interkinetic nuclear migration and perform M-phase at the ventricular surface (VS; arrow in Fig. 2A), from precursors located in the subventricular zone (SVZ) that undergo mitosis at abventricular positions (arrowhead in Fig. 2A) (Haubensak et al.,2004; Miyata et al.,2004; Noctor et al.,2004; Smart,1976). In the Pax6^Aey18–/– mutant cortex the increase in the number of PH3-positive cells was obvious(Fig. 2A,B) and highly statistically significant (Table 1). Interestingly, the number of precursors in the VZ was not changed in the Pax6^Aey18–/– cortex, but the mutation results in a specific increase in SVZ precursors(Fig. 2A,B; Table 1). This phenotype is identical to the cortex of the Pax6^Sey–/– mice(Fig. 2E) that also exhibited a marked increase in SVZ precursor cells [2.7-fold increase in SVZ precursors in Pax6^Aey18–/– cortex compared to 2.4-fold increase in Pax6^Sey–/– cortex; Table 1(Estivill-Torrus et al., 2002; Götz et al., 1998)],again supporting a functional null phenotype in the PD mutant Pax6^Aey18–/– mice. In contrast, no changes in proliferation, either in the VZ or SVZ, were detectable in mice solely lacking exon 5a (Fig. 2A,C; Table 1) or in the HD mutant Pax6^4Neu–/–(Fig. 2A,D; Table 1). Taken together, it appears that regulation of neurogenesis and proliferation does not involve the HD, but depends solely on PD function.

Next we examined the role of the PD and HD in mediating dorso-ventral patterning in the forebrain. Previous analysis of the functional null Pax6^Sey–/– described an almost complete lack of neurogenin (Ngn) 2 expression in the dorsal telencephalon, while expression of the transcription factors Mash1, Gsh2 and Olig2, that is normally restricted to the ventral telencephalon, expands into the dorsal telencephalon in the absence of Pax6 function (Fig. 3D,D′, and data not shown) (see Stoykova et al., 2000;Torresson et al., 2000; Yun et al.,2001). Precisely this phenotype was observed in the cortex of Pax6^Aey18–/– mice(Fig. 3A,A′; see Fig. S1 in supplementary material) while no changes in dorso-ventral expression patterns of these transcription factors were observed in the cortex of the Pax6(5a)–/– (Fig. 3B,B′) and the HD mutant Pax6^4Neu–/–(Fig. 3C,C′; Fig. S1). Since Ngn2 is a direct target of the canonical form of the PD(Scardigli et al., 2003),these findings further support the crucial role of the canonical PD of Pax6 in its regulation. Consistent with the gradient of Pax6 in the cortex, Ngn2 expression was still visible in the caudal-most and medial regions of the cortex in the Pax6^Sey–/– and Pax6^Aey18–/–, where Pax6 expression is lowest(see Fig. S2A,B,E in supplementary material).

The telencephalic region where these two domains of transcription factors normally abut differentiates various features of a boundary(Lumsden and Krumlauf, 1996)with formation of a radial glia fascicle and expression of specific diffusible molecules thereby inhibiting most, but not all, cells from crossing between the dorsal and the ventral telencephalon(Chapouton et al., 1999; Chapouton et al., 2001; Fishell et al., 1993). One characteristic marker is SFRP2, the soluble frizzled related protein 2, expressed specifically at the ventral-most region of the Pax6-positive territory(Chapouton et al., 2001; Kim et al., 2001). SFRP2 expression failed to appear in the Pax6^Aey18–/–(Fig. 4A,A′), the Pax6^4Neu–/–(Fig. 4C,C′) and Pax6^Sey–/– telencephalon(Kim et al., 2001)(Fig. 4D,D′), while it was normally expressed in the Pax6(5a)–/– telencephalon(Fig. 4B,B′). In contrast the boundary marker protein reticulon-1(Hirata et al., 2002) was absent in the Pax6^Aey18–/– as in the Pax6^Sey–/–, but was still detectable in the Pax6(5a)–/– and in the HD mutant Pax6^4Neu–/– (data not shown). Taken together,these results suggest that the HD and the PD cooperate on some, but not all aspects of differentiation of the dorso-ventral telencephalic boundary, while the other effects of Pax6 on dorso-ventral patterning mediated by region-specific expression of transcription factors are predominantly regulated by the PD.

Since the canonical form of the PD can bind to both P6CON and 5aCON(Epstein et al., 1994), we were concerned whether upregulation of canonical Pax6 in the Pax6(5a)–/– may compensate for the lack of the Pax6(5a)isoform. Indeed, real-time RT-PCR analysis revealed a relative increase in Pax6 mRNA in the Pax6(5a)–/– cortex that may be sufficient to compensate for the small amount (10-20%) of the Pax6(5a) isoform during most of neurogenesis in the developing cortex (see Fig. S3 in supplementary material). We therefore chose gain-of-function experiments to directly assess the specific effect of the Pax6(5a) isoform. Replication-incompetent retroviral vectors were used to overexpress Pax6(5a)in individual precursor cells to examine the influence on the progeny of these individual precursors. As indicated in Fig. 5A, single precursor cells from the cerebral cortex isolated at E14 were infected with one of the retroviral vectors carrying the LacZ marker gene encoding for β-galactosidase (βGal)(Fig. 5B), and their progeny were detected as distinct clusters of βGal immunoreactive cells, each representing a clone, after one week in vitro(Heins et al., 2001; Heins et al., 2002; Williams et al., 1991). Besides the control construct, viral vectors containing the canonical form of Pax6 (Hack et al., 2004; Heins et al., 2002) or the Pax6(5a) isoform (see Materials and methods), an IRES sequence and the marker gene (Fig. 5B) were used. This experiment assessed the cell-autonomous effects of the two isoforms of the PD on cell fate and cell proliferation by analysing the type and number of descendants generated by a single infected precursor cell.

The cell types generated by infected precursors were identified by immunocytochemistry using anti-NeuN for neurons, O4 for oligodendrocytes,anti-GFAP for astrocytes and RC2 or anti-nestin for undifferentiated precursors. Clones were classified as pure neuronal clones when all cells of a cluster were NeuN-positive (examples in Fig. 5A), as mixed when only some cells of a cluster were NeuN-positive, and as non-neuronal when no cell of a cluster was NeuN-positive (Heins et al., 2001; Heins et al.,2002). Note that Pax6 containing the canonical PD potently increased neurogenesis and the number of neuronal clones as described previously (Heins et al.,2002), while Pax6(5a) overexpression had no effect on neuronal or glial cell fate (Fig. 5C). While the canonical form of Pax6 almost completely suppressed the generation of oligodendrocytes (0% of clones, n=81) or astrocytes (3% of clones, n=249), precursors infected with the Pax6(5a) form still generated control levels of oligodendrocyte clones (7% of clones, n=103) or astrocytes (11% of clones, n=137). One transcription factor involved in cell fate and the regulation of patterning is Mash1(Casarosa et al., 1999) that normally is reduced after Pax6 overexpression(Heins et al., 2002). Only cells infected with the virus containing Pax6 with the canonical PD downregulated Mash1 (2% of clones, control 14%) while no change in Mash1-immunoreactive cells could be observed in the Pax6(5a)-infected cells(18% of clones). The absence of any effect of Pax6(5a) on cell fate was further confirmed in rescue experiments performed in cortical cells from Pax6^Sey–/– mice(Fig. 5C), where the low degree of neurogenesis could only be rescued by introduction of the canonical form of Pax6, while the Pax6(5a) form showed no effect at all(Fig. 5C). Taken together,these data demonstrate that even after overexpression in a Pax6 functional null background, Pax6(5a) is not able to influence cell fate of cortical progenitors. This suggests that the targets important for the mediation of the potent neurogenic role of Pax6 are regulated by the consensus site that is exclusively bound by the canonical PD of Pax6, i.e. the P6CON site.

In contrast to the lack of effect on cell fate, Pax6(5a) retroviral transduction exerted a potent effect on cell proliferation. The clone size reflects the number of the progeny generated by a single infected precursor and is significantly reduced by overexpression of both forms of Pax6, the canonical PD or the PD with the 5a insert(Fig. 5D). Pax6(5a)overexpression leads to a reduction in clone size independent of their cell type, i.e. in both neuronal and glial clones. While neuronal clones were typically small (see examples in Fig. 5A) (Heins et al.,2001; Heins et al.,2002), oligodendrocyte-containing clones are much larger and had a mean clone size of 16 cells in cultures infected with the control virus but only 11 cells in Pax6(5a)-infected cultures. Thus, Pax6(5a) reduces cell proliferation independent of the cell fate, while the canonical form of Pax6 couples both effects.

Our functional analysis of the distinct DNA-binding domains of Pax6 showed that the canonical form of the PD regulates coordinately, neurogenesis, cell proliferation and patterning in the developing telencephalon, while the spliced form of the PD [Pax6(5a)] affects specifically cell proliferation without any effects on cell fate and regionalization(Fig. 6). In contrast, mutation of the HD affected only subtle aspects of the boundary delineating the dorsal and ventral telencephalon. Since the canonical form of the PD binds to both Pax6 consensus sites (P6CON, 5aCON) and the Pax6(5a) isoform binds exclusively to the 5aCON site (Epstein et al.,1994), our results suggest that the regulation of neurogenesis and the forebrain patterning is mediated via target genes containing P6CON sites,whereas the regulation of proliferation is mediated by target genes containing 5aCON sites (Fig. 6).

Our analysis of the Pax6^4Neu–/– mutant,which carries a point mutation in the third helix of the HD that abolishes its DNA binding (Favor et al.,2001), revealed that DNA binding of the HD plays no role in the regulation of neurogenesis and cell proliferation in the developing forebrain,but is required together with the PD for the differentiation of one aspect of boundary formation in the forebrain, namely the expression of SFRP2. SFRP2 is a wnt-inhibitor expressed at the border between the dorsal and the ventral telencephalon. In both, the PD(Pax6^Aey18–/–) and HD(Pax6^4Neu–/–) mutant mice SFRP2 failed to be expressed at this position, while it was apparent in other parts of the brain. Other aspects of boundary differentiation, such as reticulon-1 expression or high BLBP-content in fasciculating radial glia fibers (data not shown), were normal in the HD mutant Pax6^4Neu–/–,suggesting that DNA binding of the HD is required only for some aspects of boundary specification. Since both HD and PD mutations fail to regulate SFRP2,either both DNA-binding domains may need to bind cooperatively to mediate transcription of SFRP2, or an intact PD is required for the appropriate modulation of HD DNA binding as previously demonstrated in vitro(Jun and Desplan, 1996; Singh et al., 2000; Mishra et al., 2002). These results further support the multitude of regulatory mechanisms contributing to the boundary formation at this position that may also act as an organizing centre in the developing telencephalon(Assimacopoulos et al.,2003).

The novel Pax6 mutation Pax6^Aey18 (J.G., unpublished)with a PD lacking most of its DNA-binding domain but intact HD and TA resulted in the same telencephalic phenotype as in the Pax6 functional null allele Pax6^Sey–/–(Estivill-Torrus et al., 2002; Götz et al., 1998; Stoykova et al., 1996; Stoykova et al., 2000). Even at the quantitative level, no differences in neurogenesis and cell proliferation were detectable between Pax6^Aey18–/– and Pax6^Sey–/– cortices, suggesting that the PD alone is necessary and sufficient for all of these aspects of cortical development. In addition, both genotypes expressed similar severity in the disturbance of patterning in the telencephalon. Our results therefore demonstrate that targets of the PD are necessary and sufficient to regulate cell proliferation and cell fate, comprising neurogenesis and patterning, in the developing forebrain.

Given the predominant role of the PD in exerting the Pax6 functions in forebrain development, we further determined the specific role of the 5a splice insert into the PD that shifts DNA binding from the N-terminal to the C-terminal domain of the PD (Epstein et al., 1994), by analysis of the Pax6(5a)–/–mice (Singh et al., 2002). We did not detect any changes in neurogenesis, cell proliferation and regionalization in the Pax6(5a)–/–. This finding has to be interpreted with caution, since the mRNA of the canonical form of Pax6 was upregulated (1,4-fold) in amounts sufficient to compensate for the lack of Pax6(5a) in Pax6(5a)–/– mice. Messenger RNA of Pax6(5a) comprises about 10-20% of the total Pax6mRNA (E10-E12, see Fig. S3 in supplementary material), a ratio previously observed to be rather effective for transcriptional regulation(Chauhan et al., 2004). From our in vivo analysis we can therefore only conclude that there are no specific roles of the Pax6(5a) isoform that could not also be exerted by the canonical PD in the developing forebrain. Indeed, Pax6(5a) affects solely a subset of the functions of the canonical form of Pax6. Overexpression of Pax6(5a) in individual cortical precursor cells showed a specific and cell-autonomous effect on the number of progeny of a single precursor, without affecting cell fate even in the absence of functional Pax6 in the Pax6^Sey–/– background. This phenotype, the reduction of clone size, can be due to three mechanisms – an increase in cell death, an increase in postmitotic cells or an increase in asymmetric rather than symmetric cell divisions. We observed no difference in cell death as analyzed by DAPI staining, but discovered a significant reduction of proliferating cells in the Pax6-transduced clones already 2 days after infection (71% proliferating cells amongst all control virus infected cells compared to 53% after Pax6 overexpression). Thus, Pax6 and Pax6(5a)overexpression increases the number of cells leaving the cell cycle either due to an increase in asymmetric cell division or a lengthening of cell cycle(Calegari and Huttner, 2003). Both effects could also explain the increase in the number of precursors seen in the Pax6 loss-of-function mutations(Götz et al., 1998; Estivill-Torrus et al., 2002). Further experiments are needed to clarify the exact role of Pax6 on cell cycle length or the mode of cell division in proliferating precursors. Since the Pax6(5a) form is sufficient to induce these changes in proliferation and binds exclusively to the 5aCON site, these data demonstrate that targets of the 5aCON site are specifically involved in the regulation of cell proliferation,while targets of the P6CON site seemingly regulate neurogenesis. The latter is consistent with the role of Ngn2, a target containing the P6CON site, in neurogenesis (Bertrand et al.,2002).

The effect on cell proliferation depends on the CNS region – while Pax6 negatively regulates proliferation in the telencephalon, it promotes proliferation in the vertebrate (Marquardt et al., 2001) and invertebrate(Dominguez et al., 2004) eye. This is also the case for the Pax6(5a) isoform. The deletion of exon 5a results in reduced number of iris and lens fiber cells in the eye(Singh et al., 2002),consistent with its proliferation promoting effect. Similarly, in the Drosophila eye eyegone and the murine Pax6(5a)positively affect cell proliferation(Dominguez et al., 2004). Thus, the positive effect of the 5aCON targets on cell proliferation in the eye is widespread across vertebrates and invertebrates, while the Pax6(5a)-mediated reduction of cell proliferation observed in precursors of the developing mouse telencephalon seems to be more specific and may have evolved more recently. The HD of Pax6 plays an important role in the eye(Favor et al., 2001), but not in the developing forebrain. Favor and colleagues(Favor et al., 2001) showed that the Pax6^4Neu mutation leads to severe defects in eye formation with homozygous mice developing no eye, except a remnant of the retinal neuroepithelium (pseudo-optic cup), suggesting that targets of the HD are important for both the early role of Pax6 in the surface ectoderm for lens formation, and later processes during retina specification(Ashery-Padan and Gruss, 2001; Ashery-Padan et al., 2000; van Heyningen and Williamson,2002). Our analysis has further shown reduced proliferation and Ngn2 immunoreactivity in the remnant of the retinal neuroepithelium in the Pax6^4Neu–/– (data not shown), similar to the phenotype in the PD mutant Pax6^Aey18–/–. Both the PD and HD are thus important for the regulation of proliferation and cell fate in the retina, while the HD plays no role in these aspects in the telencephalon. In conclusion our results thus imply the selective use of PD,PD5a and HD targets as one mechanism that may contribute to the region-specific differences in Pax6 function within the CNS as well as in different organs.

We thank David Anderson, Kenneth Campbell, Francois Guillemot, Tastsumi Hirata, Pierre Leprince and David Rowitch for kindly providing antibodies and probes, Michael Hack for help with real-time RT-PCR, Yves Barde, Benedikt Berninger and Andrea Wizenmann for very helpful suggestions on the manuscript and Mücella Öcalan, Romy Müller and Marcus Körbs for excellent technical assistance. The monoclonal antibody against nestin was obtained from the Developmental Studies Hybridoma Bank. This work was supported by the DFG (MG), EU, the Max-Planck Society and the Helmholtz Society.