It has recently been demonstrated that the transcription factor genes Emx2 and Pax6, expressed in the developing cerebral cortex along two complementary tangential gradients, are essential for the shaping of the cortical areal profile at late developmental ages, when cortical neuronogenesis is almost completed. In this study we addressed the question of whether cortical regionalization is already affected in Emx2 and Pax6 loss of function mutants at the beginning of neuronogenesis. By comparing expression patterns of selected molecular markers in these mutants at this age, we found that: (i) Emx2 and Pax6 are necessary for the establishment of their own specific expression profiles and are able to down-regulate each other; and (ii) absence of functional EMX2 or PAX6 proteins results in reduction of caudal–medial and rostral–lateral cortical regions, respectively, as well as in impairment of the WNT signalling center at the medial–caudal edge of the cortical field, crucial for cortical growth. These results suggest that pre-neuronogenic cortical regionalization may rely on mutual interactions between these two transcription factors and that the late areal phenotype of Emx2−/− and Pax6−/− mutants may possibly arise from both misconfiguration of the cortical molecular protomap and distortion of the cortical growth profile.
In the past, mechanisms controlling arealization of the developing cerebral cortex have been the subject of a long debate (Rakic, 1988; O'Leary, 1989). Today there is a general consensus that this process largely relies on the interplay between factors intrinsic to the cortical primordium and influences coming from subcortical structures. In particular, a large body of experimental evidences suggests that cortex-autonomous molecular cues drive the early phases of arealization, independent of information delivered later by thalamic afferents. This evidence includes: (i) early patterned expression of transcription factors in the cortical primordium (Walther and Gruss, 1991; Gulisano et al., 1996; Mallamaci et al., 1998); (ii) establishment of EphAs areal expression profiles and regional commitment of late areal markers prior to the arrival of thalamic fibers (Barbe and Levitt, 1991, 1992; Arimatsu et al., 1992; Ferri and Levitt, 1993; Donoghue and Rakic, 1999; Gitton et al., 1999); and (iii) areal restriction of Cadherins gene products in the cortex of mutants lacking thalamo-cortical connections (Miyashita-Lin et al., 1999;Nakagawa et al., 1999). However, little is known about mechanisms controlling the development of the early cortical molecular protomap.
It has been demonstrated recently that the transcription factor genes Emx2 and Pax6, expressed in the cortical pseudostratified ventricular epithelium (PVE) along two complementary tangential gradients (Walther and Gruss, 1991; Gulisano et al., 1996; Mallamaci et al., 1998), are essential for the proper shaping of the cortical areal profile. In the absence of Emx2, a dramatic reduction of cortical regions with caudal–medial identity takes place, together with an enlargement of those with rostral–lateral identity (Bishop et al., 2000; Mallamaci et al., 2000). Complementary phenotype has been described in Pax6 homozygous (Sey/Sey) mutants (Bishop et al., 2000). Assessments of areal profiles in these mutants were made by a variety of experimental approaches, but only after the end of primary neurogenesis; mechanisms underlying the genesis of the areal phenotypes were not extensively addressed. In particular, it was not established whether the Emx2−/− and Pax6−/− phenotypes are simply due to a distorted tangential growth of the PVE after its areal commitment, or whether they reflect an altered regionalization of the PVE prior to this commitment. These two hypotheses are not mutually exclusive.
At the onset of the primary neuronogenesis, a number of genes, including the homeobox genes Emx2 (Simeone et al., 1992; Gulisano et al., 1996; Mallamaci et al., 1998), Pax6 (Walther and Gruss, 1991; Stoykova and Gruss, 1994) and Emx1 (Simeone et al., 1992; Briata et al., 1996; Gulisano et al., 1996), the T-box gene Tbr2 (Bulfone et al., 1999; Kimura et al., 1999), the secreted factor genes Wnt3a (Roelink and Nusse, 1991; Grove et al., 1998), Wnt8b (Grove et al., 1998; Lee et al., 2000) and Sfrp2 (Ragsdale et al., 2000), and the FGF receptor gene Fgfr3 (Ragsdale et al., 2000), are expressed in the cortical primordium in graded ways along the tangential axes. It can reasonably be assumed that specific combinations of these genes, expressed by neuroblasts in different regions of this primordium, provide us with valuable information about latent areal values characterizing these regions. Moreover, special functional significance has to be attributed to the expression profiles of the Wnt genes, proven to play crucial roles in controlling growth of the cortical neuroepithelium (Lee et al., 2000; McLaughlin et al., 2000). Therefore, in order to cast light onto mechanisms leading to the late areal phenotypes of Emx2 and Pax6 null mutants, we systematically studied the expression profiles of the above-mentioned markers in the cortical anlage of wild-type, Emx2−/− and Pax6−/− mouse embryonic brains at the beginning of neuronogenesis. We found that both EMX2 and PAX6 proteins are necessary to allow the graded expression of the corresponding genes and to down-regulate the activity of each other. In the absence of functional alleles of Emx2 or Pax6, size reductions of regions with caudal–medial and rostral–lateral molecular identities, respectively, take place, accompanied by complementary changes in the cortical expression patterns of WNT signalling genes. This suggests that the proper setting up of the early cortical molecular protomap largely relies on the mutual interaction between these two transcription factors, and that the late areal phenotype of Emx2−/− and Pax6−/− embryos possibly reflects a misconfiguration of this protomap. Furthermore, the results imply that the late Emx2−/− and Pax6−/− areal phenotypes might be additionally worsened by a distorted cortical growth, as a result of an impairment of the WNT signalling center in the mutant cortex.
Materials and Methods
Animal Husbandry, Recovery of Embryos and Tissue Sampling
Pax6 mutant embryos were obtained by intercrossing viable heterozygous Small eye mutants, Sey allele (Roberts, 1967). Sey/Sey (hereafter called Pax6−/−) embryos were recognized by the absence of eyes, Sey/+ embryos by ocular abnormalities. The point mutation in the Pax6 gene results in the generation of truncated nonfunctional protein (Hill et al., 1991), whereas the transcription is not affected, thus allowing study of the activity of the gene in the affected brain regions. Emx2 mutant embryos were obtained by intercrossing viable Emx2−/+ mutants. Embryos were genotyped by Southern analysis as described previously (Pellegrini et al., 1996).
Pregnant females were anesthetized by CO2 and killed by cervical dislocation, in compliance with European laws [European Communities Council Directive of November 24, 1986 (86/609/EEC)] and according to guidelines of H. San Raffaele Institutional Animal Care and Use Committee. Embryos were fixed in 4% paraformaldehyde–phosphate-buffered saline (PBS) overnight at +4°C and then washed, dehydrated and embedded in wax according to standard protocols. Embryonic brains were subsequently cut at 10 μm and sections were mounted on Fischer SuperFrost Plus slides. Samples were subsequently dewaxed by xylene, rehydrated in descending ethanol series and processed for in situ hybridization or immunohistochemistry. Unless otherwise stated, for each experiment at least three embryos per mutant genotype were analysed, together with three littermate wild-type controls.
In situ Hybridization
Sense and anti-sense riboprobes were prepared by appropriately linearizing and transcribing Emx2, Pax6, Mash1, Tbr2, Wnt8b, Wnt3a, Sfrp2 and Fgfr3 cDNA clones. The Emx2 cDNA clone contained a 0.8 kbp fragment, encompassing 0.3 kbp of 5′-UTR plus 0.5 kbp of cds. The Pax6 cDNA clone contained a 3′-UTR 965 bp fragment, amplified by PCR and corresponding to nt 1485–2450 of Genbank file NM-013627. The Mash1 cDNA clone contained a 796 bp fragment encompassing 582 bp of 5′-UTR plus 214 bp of cds, amplified by PCR on the basis of sequence data assembled from Genbank files NM008553 and MMU68534. The Tbr2 cDNA clone contained a 2.7 kbp fragment, encompassing 5′-UTR, cds and 3′-UTR, and was obtained from Dr Alessandro Bulfone (Milan). The Wnt8b cDNA clone contained a 1370 bp fragment, encompassing the last 488 bp of cds plus the first 882 bp of the 3′-UTR, amplified by PCR on the basis of sequence data from Genbank files NM-01172, AW488375 and AA874401. The Wnt3a cDNA clone contained a 1421 bp fragment of 3′-UTR, amplified by PCR and corresponding to nt 1213–2634 of Genbank file NM-009522.1. The Sfrp2 cDNA clone contained a 798 bp fragment of 3′-UTR, amplified by PCR and corresponding to nt 1127–1925 of Genbank file MMU88567. The Fgfr3 cDNA clone contained a 897 bp fragment of 3′-UTR, amplified by PCR and corresponding to nt 829–1726 of Genbank file L42132. Non-radioactive in situ hybridization was performed as described in Bovolenta et al. (Bovolenta et al., 1997).
Antigens were unmasked by boiling samples in 10 mM sodium citrate, pH 6.0, for 5 min and allowing them to cool down slowly, prior to applying conventional immunohistochemical protocols (Mallamaci et al., 1998). The following primary antibodies were used: a-EMX1, rabbit polyclonal (Briata et al., 1996), 1:400; α-neuron-specific class III β-tubulin, mouse monoclonal (clone TuJ1; BabCo, Richmond, CA), 1:100.
Photography and Editing
Micrographs of the immunohistochemical preparations were take using an SV Micro CV3000 digital microscope camera (Taunton, MA). Electronic files were processed on a MacIntoshG3 computer using Adobe Photoshop 5.0 software (San Jose, CA).
EMX1, Tbr2 and Wnt8b Morphometry
Wax embedded brains were cut at 10 μm in the coronal plane. Every seventh section was processed for Wnt8b in situ hybridization and single sections immediately after in the series were processed for α-EMX1 immunohistochemistry, Mash1 in situ hybridization, Tbr2 in situ hybridization and TuJ1 immunohistochemistry. After taking photographs of the sections, the Mash1 expression profiles were electronically copied onto EMX1 images, the TuJ1 and the EMX1 profiles onto Tbr2 images, and the EMX1 profiles onto Wnt8b pictures. Medial–lateral extensions of the cortical PVE subdomains expressing EMX1, Tbr2 and Wnt8b were measured on each section by two independent investigators blind to genotype. In the case of EMX1 and Wnt8b, the lateral boundaries of their expression domains were placed on the computer screen using Adobe Photoshop 5.0 (San Jose, CA) software and always referring to given sets of chromatic parameters. In the case of Tbr2, in order clearly to distinguish lateral PVE, where Tbr2 positive cells are tightly clustered, from medial PVE, where only a few Tbr2 positive cells can be found, PVE was partitioned into equally spaced radial sectors, each extending for 50μm of the pallial ventricular boundary, Tbr2 positive cells were counted in each sector and those bearing five or more cells (corresponding to ≥1 cell/10 μm) were classified as ‘high density’, those bearing fewer cells as ‘low density’. This analysis was performed in parallel on five Emx2−/−and three wild-type littermates, as well as on five Pax6−/− and three wild-type littermates, at E11. Data from animals of each genotype, relative to rostral, intermediate and caudal thirds of the cortical primordium (EMX1 and Tbr2) or to the whole primordium (Wnt8b), were averaged and ratios between extensions of lateral and medial subfields were calculated; statistical significance of differences was evaluated by Student's t-test and results synthesized in Figures 2H,P and 4E.
Fixed and wax-embedded brains were cut at 10 μm in the coronal plane. Every fourth section was processed for Wnt3a in situ hybridization and photographed as described above. The rostro-caudal extension of the cortico-diencephalic boundary and Wnt3a expression domain were determined by inspecting all micrographs; the relative rostro-caudal extension of the Wnt3a expression domain along this boundary was subsequently calculated. This analysis was performed in parallel on five Emx2−/−and three wild-type littermates, as well as on five Pax6−/− and three wild-type littermates, at E11. Data from animals of each genotype were averaged; statistical significance of differences was evaluated by Student's ttest and results finally synthesized in Figure 4P.
Interactions between Emx2 and Pax6
In order to examine whether an interplay between Emx2 and Pax6 might be involved in the control of early cortical regionalization, we compared expression patterns of these two genes in the cortical primordium of wild-type, Emx2 null and Pax6 null mutant mice, at the beginning of neocortical neuronogenesis. Using normal embryos, we confirmed previous reports (Walther and Gruss, 1991; Gulisano et al., 1996; Mallamaci et al., 1998) on the existence of two complementary medial–lateral expression gradients, an Emx2 gradient increasing medially and a Pax6 gradient showing increased expression levels laterally (Fig.1B,E). In both Emx2 and Pax6 null mutants, dramatic changes of Emx2 and Pax6 expression profiles were detectable. Emx2 expression was down-regulated in the medial cortex of Emx2 mutants (Fig. 1A), thus flattening the normal medial-high to lateral-low expression gradient. Emx2 expression spread more laterally in Pax6 null mutants, almost abolishing the normal expression gradient (Fig. 1C). Pax6 expression was enhanced in the medial cortical anlage of Pax6 mutants, thus flattening the normal medial-low to lateral-high expression gradient (Fig. 1F). Pax6 expression was also up-regulated in the entire cortical field of Emx2 null mutants, which conversely still displayed a properly oriented expression gradient (Fig. 1D). Thus, the generation of the opposite Emx2 and Pax6 expression gradients appears to depend on the availability of functional EMX2 and PAX6 proteins, respectively, and the average expression levels of Emx2 and Pax6 arise in the absence of functional PAX6 and EMX2 proteins, respectively.
Emx1 and Tbr2 Expression Patterns in Emx2 and Pax6 Null Brains
In order to obtain information about areal values characterizing different regions of the early cortical primordium in the absence of Emx2 or Pax6, we selected two transcription factor genes whose expression normally displays specific regional restrictions in this primordium, the homeobox gene Emx1 (Simeone et al., 1992; Briata et al., 1996; Gulisano et al., 1996) and the T-box gene Tbr2 (Bulfone et al., 1999; Kimura et al., 1999) and compared their expression profiles in E11 wild-type, Emx2−/− and Pax6−/− brains.
EMX1 immunoreactivity could be detected in two cortical domains, in the primordial plexiform layer (PPL) as well as in the PVE (Fig. 2B,G). These domains could be easily distinguished by comparing EMX1 immunoreactivity (Fig. 2B) with immunoreactivity against class-III-neuron-specific β-tubulin, specifically detectable in post-mitotic neurons of the PPL (Fig. 2M). Emx1 products were abundant in the medial PVE, but fell below detectability threshold in the lateral part (Fig. 2B,G); thus, at E11, the cortical anlage could be divided into two subfields — a medial subfield, where the PVE is strongly immunoreactive for EMX1 (EMX1+ subfield) and a lateral one, where the PVE is EMX1 negative (EMX1− subfield). To estimate medial–lateral extensions of these subfields, it was necessary to define borders of the entire cortical field; for this purpose, it seemed reasonable to choose the medial boundary of EMX1 domain as medial border of the cortical field (ME, Fig. 2B,G) and the dorsal boundary of the Mash1 expression domain as the lateral border (LM, Fig. 2E,G). By superimposing EMX1 and Mash1 expression profiles from adjacent sections, it was possible to evaluate transverse extensions of EMX1− and EMX1+ cortical subfields at different rostro-caudal locations. Data relative to rostral, intermediate and caudal thirds of the cortical primordium were averaged and ratios between extensions of lateral EMX1− and medial EMX1+ subfields were calculated for these three different axial levels. This analysis was performed in parallel on wild-type (Fig. 2B,E), Emx2−/− (Fig. 2A,D) and Pax6−/− (Fig. 2C,F) embryos and results were synthesized in Figure 2H. We found that the relative extension of the lateral cortical subfield was significantly increased in Emx2 null mutants and this phenomenon was much more pronounced in caudal as compared with rostral sections. On the other hand, the lateral subfield was dramatically reduced or even disappeared in Pax6 null brains, at all rostro-caudal levels.
Emx1 and Mash1 expression profiles were also studied in telencephalons of E10 embryos, where cortical neuronogenesis is just beginning, and results similar to E11 were obtained. In Emx2−/− mutants, the cortical EMX1+ domain was reduced in its absolute and relative medial–lateral extension, more dramatically in the caudal half of the telencephalic vesicle (Fig. 3A,B). An absolute shrinking of the EMX1+ domain could also be noted in Pax6-/- mutants, where, however, relative extension of the EMX1+ domain in the Mash1− cortical field did not decrease (Fig. 3C,D). Remarkably, the EMX1−/Mash1− presumptive lateral cortical field was significantly enlarged in the caudal telencephalon of Emx2−/− mutants (Fig. 3A,B,E,F) and almost suppressed in the rostral telencephalon of Pax6−/− mutants (Fig. 3C,D,G,H).
At E11, Tbr2 products were detectable in a large subset of neurons in the PPL, as well as in a relatively small number of cells in the PVE (Figs 2J, 6B). Tbr2 expression in the PVE was not uniform along the medial–lateral axis; cells expressing Tbr2 were tightly clustered in the lateral part of the PVE and were much rarer in the medial portion (Fig. 2J,O). Thus, two cortical subfields could be distinguished: a lateral one (Tbr2+ subfield), where >1 Tbr2+cell/10 μm of free ventricular profile could be counted and a medial one (Tbr2− subfield), where the linear density of Tbr2+ cells in the PVE abruptly fell far below this threshold. Assuming again the medial boundary of the Emx1 as the medial edge of the cortical field, the transverse extensions of the these two cortical subfields could be measured. Data relative to rostral, intermediate and caudal thirds of the cortical primordium were averaged and ratios between extensions of the lateral and medial subfields were calculated for these three different axial levels. This analysis was performed in parallel on wild-type (Fig. 2J,M), Emx2−/−Fig. 2I,L) and Pax6−/− (Fig. 2K,N) embryos and the results are summarized in Figure 2P. We found that the relative extension of the lateral Tbr2+ subfield was significantly increased in the caudal third of Emx2 null mutants. Conversely, the lateral Tbr2+ subfield was dramatically reduced in Pax6 null embryos at all rostro-caudal levels. If we assume that activation of Tbr2 in cells in the ventricular zone at a spatial frequency >1 cell/10 μm of free ventricular profile is a regional property of the lateral pallial field, the almost complete absence of such cells in Pax6 mutants can be interpreted as an index of shifting of the lateral pallial field toward a different, more medial, identity. The distribution of Tbr2 transcripts along the rostro-caudal axis was also examined on horizontal sections, taken at an intermediate dorso-ventral level. In wild-type embryos, numerous Tbr2 positive cells were seen in the PPL along the entire rostro-caudal axis (Fig. 6B, delimited by arrowheads), except for the caudal–medial-most part of the cortical primordium (Fig. 6B, asterisk). In both Emx2−/− and Pax6−/− mutant brains, the medial-most sector of the caudal cortical primordium was, conversely, populated by Tbr2+ cells (Fig. 6A,C).
WNT Signalling Genes in Emx2 and Pax6 Null Mutants
Several genes involved in WNT signalling are expressed in the early cortical primordium in a highly patterned way. Wnt expression domains, including those of Wnt3a and Wnt8b, are hemi-nested around the caudal–medial edge of the cortical vesicle, or cortical hem (Roelink and Nusse, 1991; Grove et al., 1998; Lee et al., 2000). Sfrp2, encoding a putative secreted WNT antagonist, is specifically expressed near the cortico-striatal boundary, thus forming a lateral ‘mirror-image’ to the WNT-rich medial hem (Ragsdale et al., 2000). Moreover, WNT signalling is crucial for growth of the cerebral cortex (Lee et al., 2000; McLaughlin et al., 2000; Galceran et al., 2000). Therefore, scoring expression profiles of WNT signalling genes in wildtype, Emx2 and Pax6 null mutants should provide us with information about distributions of pre-areal values in the early cortical primordium of these mice; in addition, it could cast light onto molecular mechanisms controlling cortical growth. For these reasons, we systematically analysed Wnt3a, Wnt8b and Sfrp2 expression domains in telencephalons of wild-type, Emx2−/− and Pax6−/− mice, at E11 (Wnt3a and Wnt8b) and E12.5 (Sfrp2).
In the cortical field of E11 wild-type embryos, Wnt8b mRNA is expressed along a medial–caudal high to rostral–lateral low gradient: its expression peaks in the cortical hem, is still detectable in the presumptive archicortex and fades out in the adjacent neocortical anlage (Fig. 4B). Complementary changes in the Wnt8b profile were found in Emx2 and Pax6 null brains (Fig. 4D,E). In Emx2 knock-outs, the Wnt8b cortical domain was restricted to a small area near the medial boundary of the cortical field (Fig. 4A); in Pax6 null mutants, the Wnt8b domain was enlarged along the medial–lateral axis (Fig. 4C).
In wild-type embryos of the same developmental age, Wnt3a is restricted to the cortical hem (Fig. 4G,J,M, arrowheads); more specifically, its transcripts are detectable along the caudal 90% of the boundary between alar telencephalon and diencephalon (Fig. 4O,P). In Emx2 null mice, Wnt3a was found to be severely down-regulated. Only a few cells expressing it could be detected in each frontal section, in the presumptive cortical hem (Fig. 4F,I,L, arrowheads); moreover, Wnt3a expression was restricted to only the caudal 60% of the telencephalic– diencephalic boundary (Fig. 4O,P). No major differences were found between Wnt3a expression profiles of Pax6 null mutants and wild-type embryos (Fig. 4H,K,N, arrowheads; Fig. 4O,P); only some ectopic, scattered Wnt3a+ cells were detectable in the presumptive cortical field of Pax6 mutants, more frequent in its medial–caudal part (Fig. 4H,K,N, arrows).
At E12, the presumptive WNT antagonist gene Sfrp2 is normally expressed in the caudal ganglionic eminence (Fig. 5K, CGE) as well as in the lateral part of the lateral ganglionic eminence (l-LGE) or ventral pallium (Puelles et al., 2000), in a long lateral, arc-shaped domain extending from the level of the CGE to cortical regions anterior to the foramen of Monro (Fig. 5E,H,K). This expression profile was dramatically impaired in both mutants under examination. In Emx2 knock-outs, Sfrp2 mRNA was confined to the CGE (Fig. 5J) and to the caudal-most l-LGE (Fig. 5G); no expression could be detected in intermediate and rostral l-LGE (Fig. 5A,D). Complete functional inactivation of Sfrp2 was, conversely, found in the telencephalon of Pax6 knock-outs (Fig. 5C,F,I,L).
Fgfr3 Expression Pattern in the Absence of Emx2 and Pax6
It has been reported that Fgfr3 is expressed in a graded way in the early cortical primordium, with a caudal–lateral maximum and a rostral–medial minimum (Ragsdale et al., 2000). We confirmed this report on E11 wild-type embryos. In particular, in horizontal sections taken from intermediate dorsal–ventral levels, we could distinguish a caudal–medial sector (Fig. 6E, asterisk), where no Fgfr3 expression could be detected, a lateral sector displaying a caudal-highto rostral-low-graded expression (Fig. 6E, delimited by arrowheads) and a rostral–medial sector, again free of any Fgfr3 transcript (Fig. 6E, asterisk). Interestingly, in Emx2−/− embryos, the rostral Fgfr3-free sector was substantially enlarged, the Fgfr3-positive caudal–lateral sector was reduced in its rostral–caudal extension, the caudal–medial Fgfr3-free domain was missing (Fig. 6D). No major changes in Fgfr3 profile were found in Pax6 null mutants (Fig. 6F).
EMX2 and PAX6 Proteins are Both Necessary for Proper Expression of their mRNAs
It was already known that Emx2 is normally expressed less intensely in regions of the cortical PVE where Pax6 products are more abundant (Walther and Gruss, 1991; Gulisano et al., 1996; Mallamaci et al., 1998). Here, we have shown that, in the absence of either EMX2 or PAX6 proteins, expression of mRNA encoding the other one spreads to cortical regions where it is normally less abundant; this happens not only at E11, but even at later ages (data not shown). All this suggests that in the cortical PVE a direct mutual inhibition between Emx2 and Pax6 could normally occur.
In the absence of functional EMX2 protein, the expression of Emx2 mRNA is down-regulated near the medial–caudal edge of its domain, where it is normally more intensely expressed. The fact that this down-regulation is not widespread, but can be detected just near this edge, provides clues about possible mechanisms by which Emx2 normally sustains its own expression. It seems unlikely that self-sustainment takes place via a simple intracellular autoregulatory loop, which would occur everywhere in the cortical field. More reasonably, this phenomenon could rely on an intercellular positive feedback, specifically taking place near the caudal–medial boundary of the morphogenetic field in object and involving cells with special molecular and signalling properties (Meinhardt, 1983). These cells could be those forming the cortical hem, whose development is severely perturbed in Emx2−/− mutants; WNT signalling, shown here to be deeply affected in these mutants, could be normally involved in this intercellular Emx2 autoregulatory loop.
In the absence of functional PAX6 protein, the expression of Pax6 mRNA is up-regulated near the medial–caudal edge of its expression domain, where it is normally less intensely expressed. In other words, down-regulation of Pax6 by EMX2 requires PAX6, suggesting that these two proteins could provide this function as heterodimers. That would not be surprising: as recently reported, homeodomains of diverse proteins such as CHX10, SIX3, LHX2, EN1 and PREP1 bind to PAX6 and several can enhance Pax6-mediated transcriptional modulation upon co-expression in cells (Mikkola et al., 2001).
Changes in the Cortical Molecular Protomap of Emx2 and Pax6 Null Mutants
Results of the molecular analysis reported above show that profound and coherent changes did occur in molecular regional profiles of both Emx2−/− and Pax6−/− cortical primordia at the onset of neuronogenesis.
In Emx2−/− mutants, a pronounced enlargement of regions with rostral–lateral identity at the expense of those with caudal–medial identity was clearly detectable. In E11 Emx2−/− brains, the relative transverse extension of the cortical Emx1+/ Tbr2− PVE, fated to give rise to archicortex and neocortex (Fernandez, 1998; Puelles et al., 2000; Yun et al., 2001), was reduced; remarkably, a reduction of the Emx1+ cortical domain was detectable in Emx2−/− embryos even one day earlier, before the onset of cortical neuronogenesis. Moreover, the medial-most cortical PVE, expressing Wnt8b at higher levels and fated to give rise mainly to archicortex, was shrunken, especially near the caudal pole of the telencephalic vesicle. The rostral, Wnt3a−, cortical hem was enlarged at the expense of the caudal, Wnt3a+, hem. The rostral, Fgfr3−, cortical PVE was also enlarged, at the expense of the caudal–lateral, Fgfr3+, cortical PVE. Similarly, at E12.5, the rostral l-LGE domain, normally not expressing Sfrp2, was enlarged at the expense of the caudal l-LGE domain, normally expressing Sfrp2. Noticeably, the Fgfr3− domain near the topological caudal edge of the vesicle was suppressed.
Complementary changes were found in the regionalization of the Pax6 null cortical primordium. Expression of regional markers peculiar to the lateral-most cortical field was missing. The Emx1−/Tbr2+/Mash1− l-LGE domain, presumptively fated to give rise to paleocortex, amygdala and claustrum (Fernandez, 1998; Puelles et al., 2000; Stoykova et al., 2000; Yun et al., 2001), was not detectable at E11, in keeping with findings of Toresson et al. (Toresson et al., 2000) and Yun et al. (Yun et al., 2000); remarkably, the lateral Emx1−/Mash1− domain was missing in the rostral forebrain even one day earlier, before the onset of cortical neuronogenesis. At E12.5, the expression of Sfrp2 mRNA within the l-LGE was also abolished (Kim et al., 2001). Conversely, expression domains of medial cortical markers were wider. The Wnt8b+ cortical PVE, fated to give rise mainly to archicortex, was relatively enlarged; the Wnt3a+ domain, outlining the anlage of the hippocampal hem (Grove et al., 1998), was also slightly enlarged, with ectopically localized Wnt3a positive cells within the dorsal neocortex.
It is noteworthy that the pattern alterations observed in the cortical primordium of Emx2 and Pax6 mutants at the beginning of neuronogenesis are coherent with distortions of areal profiles previously described in brains of the same mutants at late gestational ages, after the completion of neocortical neuronogenesis (Bishop et al., 2000; Mallamaci et al., 2000). It is also remarkable that these protomap misconfigurations were detected at the same developmental ages at which primary areal commitments were described to occur, or even before (Arimatsu et al., 1992; Ferri and Levitt, 1993; Gitton et al., 1999). This strongly suggests that late areal phenotypes of Emx2 and Pax6 mutants can be the consequence of an altered areal commitment of the cortical primordium in these mutants. Further experiments, including cortical transplantations from Emx2−/− or Pax6−/− donors to wild-type hosts and vice versa, will be necessary formally to prove this suggestion.
Emx2, Pax6 and WNT Signalling
The dramatic down-regulation of Wnt3a and Wnt8b expression taking place in Emx2 null embryos is of particular interest. In these mutants, Wnt3a expression is almost abolished, being restricted to a few low-expressing cells near the caudal–medial pole of the telencephalic vescicle. This suggests a possible involvement of Wnt3a in the intercellular regulatory loop we hypothesized above to sustain Emx2 expression near the caudal–medial pole of the cortical field. Wnt8b, still expressed in the cortical hem, is selectively down-regulated in presumptive archicortical and medial neocortical fields. This suggests the existence of two independent mechanisms governing Wnt8b expression in the alar forebrain: one active in the hem and not dependent on Emx2, the other one active in the cortex properly called and requiring the function of this gene. Remarkably, down-regulation of Wnt genes is associated with caudal confinement of the presumptive WNT antagonist Sfrp2 (Rattner et al., 1997), possibly as a consequence of the shift of intermediate portions of the Emx2 null cortical primordium to more anterior regional identities. Wnt3a and Wnt8b expression alterations occurring in Emx2 knock-out embryos are very similar to those described by Lee et al. (Lee et al., 2000) in Wnt3a null mutants and provide valuable clues for a better understanding of early events leading to the abnormal late areal profile of the Emx2 null cortex. It has been demonstrated that at E11, in the absence of Wnt3a, the proliferating fraction of cortical neuroblasts is normal in the lateral cortical anlage, but is significantly reduced near the cortical hem (Lee et al., 2000).More recently, it has been reported that, after chronic administration of BrdU, the number of BrdU-labelled cells in the dentate gyrus is decreased by ∼30% in adult Wnt8b knock-out mice as compared to wild-type controls (McLaughlin et al., 2000). It is, therefore, reasonable to expect that knock-out of Emx2 could result in reduced tangential growth of cortical medial–caudal regions. This suggests a dual origin for the late areal phenotype of these mutants, possibly arising from both distortion of the early cortical molecular protomap and reduced absolute growth of cortical regions committed to caudal–medial fates.
The detected up-regulation of Wnt3a and Wnt8b in Pax6 null mice, i.e. the presence of scattered Wnt3a positive cells in the cortical field, near the hem, and the relative medial–lateral enlargement of the Wnt8b expression domain, suggests a crucial role for Pax6 in confining WNT signalling to the caudal–medial cortical field. This suggestion is strengthened indeed by the absolute requirement of the Pax6 function for the expression of the presumptive WNT antagonist Sfrp2 in the lateral-most cortical field. It is tempting to speculate that up-regulation of Wnt3a and Wnt8b occurring in Pax6 null embryos might have a functional consequence in promoting absolute tangential growth of caudal–medial regions in these mutants. In this way, as for Emx2 null mutants, a dual origin can be hypothesized for the late areal phenotype of Pax6 null mice, possibly arising from both distortion of the early cortical molecular protomap and reduced relative growth of cortical regions committed to rostral–lateral fates.
We wish to thank Alessandro Bulfone for the gift of the Tbr2 cDNA clone and John Parnavelas for critical reading of the manuscript. This work was funded by the Italian Association for the Research against Cancer, Italian Telethon, Giovanni Armenise–Harvard Foundation, University Excellence Center on Physiopathology of Cell, CNR (Italian National Research Council) and EU (QLG3-CT-2000-00158, QLG3-CT-2000-01625, TMR HPRN-CT-2000-00097).