Three basic aspects of cerebral cortex development — specification of cortical versus ganglionic identity, regionalization of the early cortical primordium and arealization of the developing cortex — were the main subject of our recent investigations. We previously demonstrated that the two homeobox genes Emx2 and Pax6 promote development of caudal–medial and rostral–lateral cortex, respectively, by properly shaping the early cortical protomap and possibly modulating the tangential growth ratio between medial and lateral cortical anlagen. More recently, by analyzing the brains of embryos bearing mutations for Emx2 and Pax6 in different combinations, we found that both genes are necessary and sufficient for a more basic developmental choice, i.e. the specification of neuroblasts in the dorsal telencephalon as cortical versus ganglionic neuroblasts. Finally, we explored the possible roles of the Emx2 paralog, Emx1, in these processes. By looking at embryos mutant for Emx1, Emx2 and Pax6 in various combinations, we did not get any evidence of Emx1 involvement in the process of cortical specification; conversely, this gene appeared to be involved to some extent in the process of regionalization of the cortical primordium along the medial–lateral axis, as a promoter of medial fates.
Emx2, Pax6 and Cortical Arealization
The specification of area identities in the cerebral cortex is a complex process, beginning at mid-gestational ages and completed after birth. Early phases of this process occur before axons coming from the thalamus reach the cortex, on the basis of cortex-autonomous cues; late phases occur after the arrival of the first thalamo-cortical projections and are partly influenced by them (Ragsdale and Grove, 2001; O’Leary and Nakagawa, 2002). The graded and complementary expression of the homeogenes Emx2 and Pax6 in the ventricular zone of the cerebral cortex since early stages of its development (Walther and Gruss, 1991; Gulisano et al., 1996; Mallamaci et al., 1998) (Fig. 1A) suggested these genes could control early, thalamus-independent phases of cortical arealization, in an antagonistic way. By using a variety of experimental approaches (in situ hybridization-immunohistochemistry analysis of early area-specific markers, histochemistry of the area-specific β-galactosidase activity encoded by the transgene H2Z1, DiI profiling of thalamo-cortical connections, study of area-specific BrdU uptake patterns), we previously found that, in the absence of Emx2, the normal spectrum of areal identities is still encoded, but a relevant reduction of cortical areas with more caudal– medial identities, together with an expansion of anterior–lateral territories takes place (Mallamaci et al., 2000) (Fig. 1B). Similar findings were reported by the O’Leary group, which also detected a complementary areal phenotype in Pax6Sey/Sey mutants, hereafter reported as Pax6–/– mutants (Bishop et al., 2000, 2002) (Fig. 1B). This analysis left three relevant questions open. First, the original characterization of Emx2 and Pax6 null areal phenotypes was performed at late gestational stages, i.e. ∼5/6 days after areal commitment of cortical neuroblasts (Barbe and Levitt, 1991; Arimatsu et al., 1992; Gitton et al., 1999). Therefore, the question arose as to whether these phenotypes reflect a coherent distorsion of the early regional profile of the cortical primordium, prior to its areal commitment, or are consequences of a distorsion of its tangential growth profile, after this commitment; obviously, these two mechanisms are not mutually exclusive. Secondly, it remained to clarify which functional relationship, if any, does take place between Emx2 and Pax6 in arealization mechanisms. Thirdly, do Emx2 and Pax6 do all the job or does any other functional pathway exist that is relevant for cortical arealization but does not involve Emx2 and Pax6?
To address the first question, we systematically studied the expression profiles of a panel of regional markers in the early cortical anlage of wild-type, Emx2–/– and Pax6–/– mouse embryonic brains (Muzio et al., 2002a). In Emx2 null brains, the Wnt8b+/Emx1+ archicortical anlage and the Wnt8b–/Emx1+ neocortical anlage were reduced and the paleocortical, Emx1–/Tbr2+ ventricular zone (VZ) was enlarged (Fig. 1C); on horizontal sections of the same brains, the Coup-tf1–/FgfR3– rostral–medial domain was enlarged, the Coup-tf1+/FgfR3+ intermediate domain caudally displaced and the Coup-tf1–/FgfR3– caudal–medial domain suppressed (Fig. 1D). Complementary changes in medial–lateral regional profile could be detected in Pax6 null mutants (Fig. 1C). This study showed that late areal phenotypes of these mutants may be due to mispatterning of their cortical primordium prior to areal commitment. However, down- and up-regulation of Wnt3a and Wnt8b, two Wnt genes promoting proliferation of cortical neuroblasts (Lee et al., 2000; Fukuchi-Shimogori and Grove, 2001; Mc Laughlin et al., 2000) in the medial cortical field of Emx2 and Pax6 null mutants, respectively (Fig. 1E), also suggested that late areal phenotypes of these mutants could be worsened by unbalanced growth of their medial and lateral cortical anlagen.
Concerning the second question, analysis of null mutants showed that Emx2 and Pax6 pathways inhibit each other. Moreover, the EMX2 protein sustains Emx2 expression in the medial cortical field and the PAX6 protein is necessary to achieve the Emx2-dependent repression of Pax6 in the same area (Muzio et al., 2002a) (Fig. 1F). Molecular details of these interactions are currently under analysis in our laboratory.
Concerning the third question, two other genes, Fgf8 and Coup-tf1, have recently been shown to be involved in cortical arealization. Ectopic expression of Fgf8 in the rostral cortical primordium results in size-reduction of caudal areas and inhibition of Fgf signaling into a complementary phenotype (Fukuchi-Shimogori and Grove, 2001). Absence of Coup-tf1, leads to uniform rostro-caudal expression of regional markers Cad8, ROR-b and Id2, as well as to improper connection of caudal cortex to the somatosensory thalamic ventro-basal complex (Zhou et al., 2001). However, both Fgf8 and Coup-tf1 are likely to act along the same pathway of Emx2, upstream (Crossley et al., 2001; Fukuchi-Shimogori and Grove, 2001) and possibly downstream of it (Zhou et al., 2001) (Fig. 1D), respectively. To see if there exist other arealization pathways, not involving either Emx2 or Pax6, we generated mice double-knockout for both two these genes and looked for possible residual signs of regionalization in their cortical primordium.
Emx2, Pax6 and Cortical Specification
Unfortunately, analysis of these double mutants did not allow us to answer the question, because the specification itself of cortical identity was abolished in their brains. However, this study was fruitful because it provided us with intriguing results about molecular mechanisms underlying the early partition of the telencephalic wall into its main subdivisions, namely cortical hem, pallium, medial and lateral ganglionic eminences (Muzio et al., 2002b). Results of this study can be summarized here as follows.
At E14.5, the ‘cortex’ of Emx2–Pax6 double knockouts, size-reduced and not showing any sign of lamination, did not express a set of six transcription factor genes peculiar to the developing pallium (Emx1, Emx2, Pax6, Tbr2, Ngn1, Ngn2; Fig. 2A), with the exception of a few marginally located Emx1+ presumptive neurons. This cortex expressed a set of basal forebrain markers (Vax1, Gsh2, Islet1, Gad65/67, Calbindin, Ebf1 among them; Fig. 2A) along the normal ventricular-to-marginal progression they exhibit in the lateral ganglionic eminence, suggesting it acquired properties of this structure (Fig. 2B). Expression profiles of some of the genes demarcating the boundaries among choroid plexus, cortical hem and pallial field properly called (Msx1, Otx2, Bf1 and Id3, but not Lhx2, Ttr and Wnt8b; Fig. 2A), were also altered in these mutants, suggesting that some spreading of cortical hem identity into their pallial field took place place as well (Fig. 2B). Moreover, the lateral ganglionic eminence was converted into medial ganglionic eminence (Fig. 2B), as indicated by the confinement of Ebf1 to dorsal telencephalon and the spreading of Nkx2.1 up to the cortico-striatal notch. Remarkably, only one of either of the Emx2 or Pax6 functional alleles was able to support cortical specification, in sectors of dorsal telencephalon where the spared allele is more intensely expressed. Monitoring molecular features of Emx2–/–Pax6–/– telencephalons at E11.5 revealed changes similar to E14.5. The three pallial markers Emx1, Tbr2 and Pax6 were dramatically, even if not completely, down-regulated and the double mutant ‘cortex’ expressed the five pan-basal markers Mash1, Dlx1, Dlx2, Islet1 and Gad65/67 (Fig. 2C), indicating that, at this age, dorsal-to-ventral transformation of this structure was already in progress (Fig. 2D). Boundaries between ganglionic eminences and between cortical hem and pallial field were, conversely, not misplaced (Fig. 2D), as suggested by expression patterns of Shh, Nkx2.1, Wnt3a and Wnt8b (Fig. 2C). Thus, in the absence of both Emx2 and Pax6, between E11.5 and E14.5, boundaries among all the main telencephalic morphogenetic fields are not fixed, like the pallial–subpallial boundary in Gsh2 null mutants at similar gestational ages (Corbin et al., 2000). This is particularly remarkable, because it could be symptomatic of latent, unexpected plasticity of the system, becoming apparent in the absence of specific molecular constraints.
To get a more detailed comprehension of the roles Emx2 and Pax6 in developmental choices taking place in the early telencephalic primordium, we addressed mechanisms leading to the huge accumulation of Islet1+Gad65/67+ presumptive neurons which characterizes the Emx2–Pax6 double mutant ‘cortex’. In particular, we tried to define presumptive place of birth of these neurons as well as to cast light onto the origin of their progenitors. Results of short-term BrdU pulse-chase experiments and the almost complete absence in dorsal telencephalon of transcripts of Lhx6, Lhx7 and Nrp1, specific to interneurons born in basal forebrain and migrating to more dorsal structures (not shown), suggested that these neurons were largely born in the cortex itself (Fig. 1E, route a) and not in the basal telencephalon (Fig. 1E, route b). Remarkably, neuroblasts in dorsal telencephalon proliferated and to underwent apoptosis at rates comparable to those of the adjacent basal telencephalon (not shown), which makes total replacement of them by infiltrating basal neuroblasts (Fig. 1F, case b) unlikely. Moreover, a large fraction of dorsal neuroblasts coexpressed both Emx1 and the ganglionic marker Gsh2 (Fig. 3), a phenomenon peculiar to Emx2/Pax6 double knockouts, reasonably interpretable as an index of dorsal-to-ventral respecification in progress. All this suggests that a substantial fraction of progenitors of cortical Islet1+Gad65/67+ neurons are autochthonous neuroblasts, undergoing cortical-to-ganglionic respecification (Fig. 1F, case a).
In brief, in the absence of both Emx2 and Pax6, dorsal telencephalon neuroblasts are converted early into ganglionic neuroblasts and this is followed by huge accumulation of basal-type neurons in the double mutant ‘cortex’, which gives rise to an aberrant structure, somehow resembling the adjacent striatum. Moreover, one of either of the Emx2 or Pax6 functional alleles is necessary and sufficient stably to promote cerebral cortex morphogenesis in the dorsal telencephalon, indicating that these two genes act in parallel, as master genes for cerebral cortex morphogenesis. Actually, ventralization of cerebral cortex is not peculiar to Emx2–/–Pax6–/– mutants. Distorted molecular patterning at the pallial–subpallial border has been described in simple Pax6–/– mutants (Stoykova et al., 1996, 2000). Moreover, ventralization of cerebral cortex is also detectable in different genetic environments, such as upon cortical overexpression of the subpallial proneural gene Mash1 and/or inactivation of pallial proneural genes Ngn1 and Ngn2, as well as in mice naturally knockout for the zinc finger gene Gli3. However, only Emx2 and Pax6 and neither Ngns nor Gli3 act as proper ‘master genes’ for cerebral cortex morphogenesis. In fact, in Mash1 and Ngn mutants, cortical ventralization, as assessed by morphological and molecular criteria, is much less dramatic than in Emx2/Pax6 double knockouts; moreover, in the same mutants, cortical expression of Emx1, Emx2 and Pax6 is not impaired (Fode et al., 2000), whereas Ngn and Mash1 patterns are deeply perturbed in Emx2/Pax6 double knockouts. This suggests that Emx2 and Pax6, as general activators of corticogenesis, act at a higher hierarchical level as compared to Ngn1 and Ngn2, which would control more limited aspects of cortical morphogenetic programs. In Gli3 loss-of-function mutants, Emx2, Pax6 and Ngn2 are still expressed, even if at lower levels and residual signs of cortical specification can be detected in the dorsal telencephalon (Theil et al., 1999; Tole et al., 2000). This suggests that Gli3 does not select cortical fates, but rather promotes cortical morphogenesis, by positively modulating expression of Emx2 and Pax6.
Materials and Methods
Animal husbandy, recovery of embryos, tissue sampling, immunohistochemistry, immunofluorescence, in situ hybridization, photography and editing were performed as described previously (Mallamaci et al., 2000; Muzio et al., 2002a,b). For Coup-tf1 in situ hybridization, the plasmid COUP-TF1, a kind gift of M. Studer, was used. Experiments were carried out in compliance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) as well as in agreement with the San Raffaele Institutional Animal Care and Use Committee (HSR-IACUC) guidelines.
New Results and Discussion
What about Emx1?
Thus, Emx2 and Pax6 control two very basic aspects of cerebral cortex development, commitment of dorsal telencephalon neuroblasts to cortical fates (Muzio et al, 2002b) and execution by these neuroblasts of antero-lateral or caudo-medial regional– areal morphogenetic programs, cooperating and antagonizing each other, respectively (Bishop et al., 2000, 2002; Mallamaci et al., 2000; Muzio et al., 2002a,b). We wondered whether the Emx2 paralog Emx1 would also be involved in these processes, maybe in a way similar to Emx2. In fact, Emx1 products are specifically detectable in presumptive pallium, since ages at which ganglionar versus cortical commitment is going on (Simeone et al., 1992; Briata et al., 1996); graded expression of Emx1 in the pallial ventricular zone is highly similar to that of Emx2 (Briata et al., 1996; Mallamaci et al., 1998); Emx1 and Emx2 are structurally very similar (Simeone et al., 1992). Actually, Emx1 has been described as controlling specific cortical morphogenetic processes, such as early neocortical lamination, development of corticofugal, cortico-cortical and thalamo-cortical projections and migration of GABAerging interneurons from basal forebrain (Yoshida et al., 1997; Shinozaki et al., 2002). However, no impairment of cortical committment and arealization has been found in mice knockout for this gene (Yoshida et al., 1997). It is reasonable to hypothesize that, if such phenomena occurred, they could be very subtle. Therefore, we re-addressed the problem on sensitized genetic backgrounds by generating mouse embryos of different ages, mutant for Emx1, Emx2 and Pax6 in different combinations and looking at the expression of selected molecular markers in their telencephalons. Results can be summarized as follows.
To assess a possible role of Emx1 in regionalization of the early cortical primordium, we compared medial–lateral extent of Tbr2 and Wnt8b expression domains in telencephalons of wild-type, Emx2–/– and Emx1–/–Emx2–/– E11.5 embryos. The Tbr2 ventricular domain, normally corresponding to ventral and lateral pallium and enlarged in the absence of Emx2 (Muzio et al., 2002a), coincided with the entire cortical neuroepithelium in Emx1–/–Emx2–/– mutants (Fig. 4A–C). The Wnt8b domain, normally corresponding to cortical hem and medial pallium and restricted to the cortical hem in the absence of Emx2 (Muzio et al., 2002a), was further shrunken in Emx1–/–Emx2–/– mutants (Fig. 4D–F). To see if this early enlargement of latero-ventral pallium at expenses of dorso-medial pallium would result in a coherent change of late cortical areal profile, we monitored the cerebral cortex of E18.5 wild-type, Emx2–/– and Emx1–/–Emx2–/– brains for the expression of two regionally restricted genes, Cad6 and Lamp. The paleocortical Lamp expression domain, medially shifted in Emx2–/– embryos (Mallamaci et al., 2000), was further shifted in the same direction in Emx1–/–Emx2–/– mutants (Fig. 4G–I). The archicortical Lamp expression domain, size-reduced in Emx2–/– embryos (Mallamaci et al., 2000), was undetectable in Emx1/Emx2 double knockouts (Fig. 4G–I, arrows). The main Cad6 pallial expression domain, normally spanning both sides of the cortico-striatal notch and medially shifted in Emx2–/– embryos (Mallamaci et al., 2000), was further shifted in the same direction in Emx1–/–Emx2–/– mutants (Fig. 4J–L). The subicular Cad6 expression domain, barely detectable in Emx2 knockouts (Mallamaci et al., 2000), was indistinguishable in Emx1/Emx2 double knockouts (Fig. 4J–L, arrows). Thus, early changes in pallial regionalization were followed, at perinatal stages, by enlargement of paleocortex at the expense of medial neocortex and archicortex. However, archicortical fates were not completely abolished in Emx1/Emx2 knockouts, as suggested by the absence of Cad6 transcripts in the dorsomedial-most double mutant cortex (Fig. 4L), which, conversely, still expressed the dentate gyrus–fimbria marker Id3 (not shown). Put in other words, in the absence of Emx2, Emx1 can functionally replace it — to a partial extent — as a promoter of archicortical morphogenetic programs. What is not clear is whether this is its normal functional role. The absence of obvious areal abnormalities in Emx1–/– mutants would suggest that it is not so. However, subtle changes in areal profiles could also have been taking place in these mutants, falling under the detectability threshold of our analysis. Remarkably, during the preparation of this manuscript, Bishop et al. reported that abnormal rostro-caudal distribution of Cad6, p75, EphrinA5 and EphA7 transcripts occurring in Emx2–/– mutants is not worsened in the absence of both Emx1 and Emx2 (Bishop et al., 2002). Together with our findings, this suggests that, unlike Emx2 and Pax6, area-profiling activity of Emx1 takes place along only one tangential axis — the medial–lateral one — and not along its perpendicular.
To assess a possible Emx2-like role of Emx1 as a promoter of cortical versus other competing morphogenetic programs, we generated E14.5 embryos, mutant for Emx1 and Pax6 in different combinations and analyzed expression patterns of pallial (Tbr2 and Ngn2), basal ganglia (Gsh2) and cortical hem (Otx2) markers in their telencephalons. Dorso-medial shift of the pallial–subpallial boundary, between Ngn2/Tbr2 and Gsh2 expression domains, occurring in Emx1–/+Pax6–/– (Fig. 5A,B,D, E,G,H) as well as in Pax6–/– mutants [not shown (Toresson et al., 2000)], was not more pronounced in Emx1/Pax6 double knockouts (Fig. 5C,F,I), suggesting that Emx1 is not essential for proper placement of this boundary. On the other hand, no obvious enlargement of the Otx2 expression domain was detectable in either Emx1–/+Pax6–/– or Emx1–/–Pax6–/– mutants (Fig. 5J–L), further suggesting that Emx1 is not involved in medial confinement of cortical hem fates. However, the fact that in the Emx2–/–Pax6–/– cortical primordium, cortical-to-striatal respecification is not complete around E11, when substantial amounts of EMX1 can be still found in the dorso-medial cortical field, raises the possibility that, in the absence of Emx2 and Pax6, Emx1 could early and transiently promote corticogenesis, against morphogenesis of basal ganglia (Muzio et al., 2002b). To test this hypothesis, we simultaneously inactivated all of all three genes, Emx1, Emx2 and Pax6, and examined the resulting embryonic brain at E11. Triple knockout neither abolished residual expression of Tbr2, still detectable in the dorso-medial pallium of Emx2/Pax6 double knockouts, nor led to early activation of Gsh2 in this area (Fig. 5M–R). So, again, we did not obtain any experimental evidence of Emx1 involvement in cortical versus ganglionic specification.
In sum, our analysis indicates that Emx1 is marginally involved in high-level patterning choices, occurring at the beginning of cortical development (essentially, it seems involved only in medial–lateral profiling of the pallial primordium). Conversely, it reveals that at least one other unknown gene, different from Emx1, Emx2 and Pax6, is able independently to activate corticogenesis in a transient way (Muzio et al., 2002b). Identifying this gene is subject of our current research.
Studies on cortical arealization, regionalization and specification in mice mutants for Emx2 and Pax6 were performed in collaboration with the groups of John Parnavelas and Peter Gruss/Anastassia Stoykova; early phases of this analysis took place when the corresponding author was still in the laboratory of Dado Boncinelli. These studies and recent work on Emx1 mutants were funded by the EU (QLG3-CT-2000-00158; QLG3-CT-2000-01625; HPRN-CT-2000-00097), the Italian Ministery of Health (CS030.5/RF00.73), the Italian National Research Council (PF01.00026.PF49) and the University Excellence Center on Physiopathology of the Cell.