The correct size of the different areas composing the mature cerebral cortex depends on the proper early allocation of cortical progenitors to their distinctive areal fates, as well as on appropriate subsequent tuning of their area-specific proliferation–differentiation profiles. Whereas much is known about the genetics of the former process, the molecular mechanisms regulating proliferation and differentiation rates within distinctive cortical proto-areas are still largely obscure. Here we show that a mutual stimulating loop, involving Emx2 and canonical Wnt signalling, specifically promotes expansion of the occipito-hippocampal anlage. Collapse of this loop occurring in Emx2−/− mutants leads progenitors within this region to slow down DNA synthesis and exit prematurely from the cell cycle, due to misregulation of cell cycle-, proneural- and lateral inhibition-molecular machineries, and eventually results in dramatic and selective size-reduction of occipital cortex and hippocampus. Reactivation of canonical Wnt signalling in the same mutants rescues a subset of molecular abnormalities and corrects differentiation rates of occipito-hippocampal progenitors.
Development of the murine cerebral cortex and proper growth of the distinctive areas composing it rely largely on the fine kinetic tuning of proliferating progenitors populating the embryonic psudostratified ventricular epithelium (PVE). Within this layer, between embryonic day (E) 11 and E17, both the cell cycle length and the fraction of cells leaving the cell cycle and differentiating into neurons increase progressively. At each age, these parameters are regulated regionally, being higher in rostral–lateral than in caudal–medial cortex (reviewed in Caviness et al., 2003). Moreover, orientation of mitotic spindles also varies, with radial mitoses becoming more frequent at later ages (Chenn and McConnell, 1995; Estivill-Torrus et al., 2002). Finally, apoptosis rates progressively increase, even if there is poor agreement about their actual values (Blaschke et al., 1996; Thomaidou et al., 1997). Genetic control of these processes is extremely complex and the information available to date is still fragmentary. Cortical activation pattern of the general core machinery regulating cell cycle progression has been carefully described (Zindy et al., 1997; Van Lookeren Campagne and Gill, 1998; Delalle et al., 1999; Coskun and Luskin, 2001) and it has been shown that the progressive increase in cdk-inhibitor Kip1p27 gene expression in the E11.5–E14.5 PVE can be crucial for progression of the cycling time in such a structure (Mitsuhashi et al., 2001). It has been demonstrated that pallium-restricted basic domain-belix-loop-belia (bHLH) proneural genes of the Ngn family promote neuronal differentiation (Fode et al., 2000), whereas belix-loop-belix (HLH) antineural Id genes inhibit it (Lyden et al., 1999) and stimulate cell cycle progression (Zebedee and Hara, 2001). It has been shown that, during embryonic corticogenesis, Dllon newborn neurons can signal to Notchon radial glial cells (i.e. proliferating neuroblasts), forcing them to switch Hes genes on and to retain their state, not differentiating into neurons (Ishibashi et al., 1994; Chambers et al., 2001; Ohtsuka et al., 2001; Gaiano and Fishell, 2002). It is commonly accepted that the activity of these ‘general purpose’, cell cycle-, proneural/antineural- and lateral inhibition-machineries can be finely modulated within the cortical field by a variety of stimuli, including signalling machineries active around the borders of this field as well as patterned transcription factor genes expressed within it, both potentially able to regulate sizes of distinctive cortical areas. However, how this is achieved is still poorly known. Canonical Wnt signalling, active at early developmental stages around the caudal–medial cortical hem, (grove et al., 1998, and Lee et al., 2000) forces cortical neuroblasts to keep proliferating (Chenn and Walsh, 2002; Zechner et al., 2003). Later, the same pathway promotes their neuronal differentiation (Hirabayashi et al., 2004). The transcription factor (TF) gene Foxg1, expressed within the cortical field along a caudal–mediallow-to-rostral–lateralhigh gradient, promotes cell cycle progression and stops neuronogenesis (Hanashima et al., 2002), by inhibiting the cdki gene Cip1p21 (Seoane et al., 2004) and enhancing lateral inhibition (Yao et al., 2001). The TF gene Pax6, bearing similar tangential restrictions, exerts complex effects on cell cycle progression (Estivill-Torrus et al., 2002), promotes conversion of VZ- into SVZ-progenitors and stimulates neuronogenesis (Heins et al., 2002), possibly by directly transactivating the proneural gene Ngn2 in the lateral cortical field (Scardigli et al., 2003). Emx2, encoding for another TF gene, expressed within the PVE along a caudal–medialhigh-to-rostral–laterallow and earlyhigh-to-latelow spatio-temporal gradient (Simeone et al., 1992; Mallamaci et al., 1998; Leingärtner et al., 2003), is also an ideal candidate gene for controlling spatio-temporal progression of cortical population kinetics, possibly as a promoter of cell cycle progression and a repressor of cell cycle exit and neuronogenesis. A few years ago, O'Leary's group and ours independently found that, in the cerebral cortex of late gestation Emx2 null mutants, caudal–medial regions are size-reduced and rostral–lateral ones relatively enlarged (Bishop et al., 2000; Mallamaci et al., 2000). Although this phenotype is prefigured by abnormal regionalisation of the early cortical primordium prior to its areal commitment (Muzio et al., 2002a), the possibility stands that early patterning abnormalities of these mutants could be later worsened by selective reduction of tangential expansion rates of caudal–medial regions, upon their areal commitment. Actually it has been shown that Emx2 influences kinetics of cortical proliferating populations (Gangemi et al., 2001; Heins et al., 2001; Galli et al., 2002). However, these studies were run in vitro, on cell cultures, and, moreover, their results were apparently controversial. So they can be hardly extrapolated to decipher the role of Emx2 in the control of cortical proto-areas growth in vivo. To address this topic, we selected a set of three key parameters describing the rates of basic processes underlying tangential expansion of the cortical PVE, we compared their values in vivo, in the presence or in the absence of Emx2, and thus we could define the basic kinetic abnormalities leading, in Emx2−/− mice, to hypoplasia of the caudal–medial cortex. We explored molecular correlates of these abnormalities and discovered that they could stem from misregulation of at least four distinctive molecular pathways. Interestingly, pharmacological reactivation of one of them, i.e. the canonical Wnt signalling pathway, partially rescued the others and restored at some extent the normal kinetic profile.
Materials and Methods
Animal Husbandry, Embryo Genotyping and Lithium Administration
Mutant embryos were generated starting from Emx2 null (Pellegrini et al., 1996) and Bat-gal (Maretto et al., 2003) founder mice and genotyped by PCR (see supplementary material available on-line). When appropriate, they were exposed to chronic lithium, by i.P. injection of pregnant dams (see supplementary material available on-line).
Population Kinetics Profiling
Total cell cycle length, TC, and S-phase length, TS, were determined via the cumulative S-phase labeling method (Takahashi et al., 1993). The fractions of neuroblasts exiting the cell cycle and differentiating into neurons, fDN, were determined by anti-bromodeoxyuridine (BrdU)/anti-class III-neurospecific-betatubulin immunofluorescences (see supplementary material available on-line).
Gene Expression Profiling
In situ hybridization, immunofluorescence/immunohistochemistry, Western blotting and quantitative retrotranscription polymerase chain reaction (qRT-PCR) were performed according to standard techniques, as described elsewhere (Briata et al., 1996; Muzio et al., 2002b; see also supplementary material available on-line).
The Early Caudal–Medial Cortical Primordium Specifically Requires Emx2 for its Proper Expansion
At first approximation, the tangential expansion rate of the cortical PVE can be considered a function of a three parameters, describing the basic kinetic behaviour of its components: cycling time (TC), neuronal differentiation rate (fND) and apoptosis rate. To assess if and how Emx2 controls tangential expansion of the PVE, these parameters were carefully compared in wild type and Emx2−/− mutant embryos, at different tangential locations as well as at different stages of cortical morphogenesis. As assessed by the cumulative S-phase labeling method (Takahashi et al., 1993), TC, normally 8.6–8.7 h at E10.5, increased to 12.0 and 10.0 h in the caudal and rostral cortex of Emx2−/− mutants, respectively, largely due to slowing down of DNA synthesis [TS(wt) = 3.3–3.4 h; TS(ko/caudal) = 7.3 h; TS(ko/rostral) = 5.8 h]. A similar increase of TC was also detectable at E13.5, even if less dramatic and restricted to caudal medial cortex [TC(ko,caudal–medial) = 14.0; TC(wt,caudal–medial) = 12.0 h] (Fig. 1a). Neuronal differentiation rates, fND, measured by labelling E11 neuroblasts in S-phase with BrdU and calculating, 10 h later, the fraction of them expressing the early post-mitotic neuronal marker β3-tubulin, were more than doubled in the Emx2 null caudal cortex (0.084 versus 0.040) and increased by about one-third in the rostral cortex (0.060 versus 0.045). Noticeably, the BrdU–β3-tubulin double labeling assay gave similar results also at E13; however, in this case, changes of fND were less dramatic (0.087 versus 0.053 and 0.074 versus 0.060 in caudal and rostral cortices, respectively) (Fig. 1c). Finally, apoptosis rates, assayed by anti-activated-caspase3 immunohistochemistry, did not display any statistically relevant change: at E12.5, activated-caspase3+ cells were 54 400 ± 23 200 and 50 100 ± 19 600 per mm3, in mutant and wild type brains, respectively, with n = 6 and P > 0.5. Coherent results were obtained by TUNEL, at E13.5 (not shown). In synthesis, in the absence of Emx2, tangential expansion rates of the cortical primordium were reduced, due to the slowing-down of DNA synthesis and the exaggerated exit of neuroblasts from the cell cycle. Moreover, these phenomena were much more pronounced in the earlier stages, as well as around the caudal–medial pole of the telencephalic vesicle, i.e. just when and where Emx2 is expressed most intensely.
Emx2 Stimulates Tangential Expansion of the Caudal–Medial Cortical Primordium by Promoting Canonical Wnt Signalling
To clarify molecular mechanisms leading to these abnormalities, we looked at the expression of four gene panels encoding for the molecular machineries involved in the control of population kinetics, i.e. cyclin/cdk/cki-, proneural/antineural-, lateral inhibition- and canonical Wnt signalling-machineries. Among genes regulating early-to-late G1 progression, cyclinD1 and cyclinD2, normally expressed within the cortical ventricular zone along a latero-ventralhigh-to-dorso-mediallow gradient, were up-regulated (Fig. 2a–d, Supplementary Fig. 4), Cdk4, Ink4p18 and Ink4p19 were unaffected (Supplementary Fig. 1a–e). Kip1p27 and Kip2p57, both inhibiting G1–S transition and S-phase progression and normally expressed along a latero-ventralhigh-to-dorso-mediallow gradient, were strongly up-regulated in the dorso-medial cortex (Fig. 2e–h, Supplementary Fig. 5), possibly accounting for slowing down of DNA synthesis and exaggerated neuronogenesis occurring in this region. Other genes involved in the control of the same processes were unaffected (Cdk2, cyclinE, Cip1p21 and cyclinA1, Supplementary Fig. 1f–l,o,p) or slightly reinforced (cyclinA2, Supplementary Fig. 1m,n). Concerning proneural genes, Ngn1 and Ngn2 were both up-regulated within the medial pallium (Fig. 2i–l, arrowheads). In particular, on 10-μm-thick sections, within the caudal–medial pallium of E11.5 Emx2−/− and wild type embryos, we could count 13.8 ± 2.8 and 8.3±2.9 Ngn1on cells/100μm of ventricular profile, respectively (n = 6 + 6 and P < 0.025). Among antineural genes, Id3, normally confined to this region, collapsed and its paralog Id4, normally displaying a complementary expression pattern, was down-regulated as well, but less dramatically (Fig. 2n–q). As a consequence, within the medial pallium, the proneural/antineural products ratio specifically increased, thus possibly over-boosting neuronogenesis in this region. Relevant changes were also found in the lateral inhibition machinery, supposed to be activated by newborn neurons expressing Delta and to inhibit differentiation of adjacent neuroblasts to neurons. Whereas Notch1 was not affected (not shown), the frequency of Delta1 expressing cells specifically dropped within presumptive archicortex (0.195 ± 0.003 versus 0.119 ± 0.004, n = 4 + 4, P < 0.005) (Fig. 2r,s) and the dorsomedial Hes5 expression subdomain disappeared (Fig. 2t,u), possibly contributing to exaggerated neuronogenesis taking place in this region. Finally, dramatic changes occurred to canonical Wnt signalling, also playing a pivotal role in regulating the balance between proliferation and differentiation in the developing cerebral cortex (Chenn and Walsh, 2002). As suggested by the expression pattern of the nuclear beta-catenin-responsive Bat-gal transgene (Maretto et al., 2003), this pathway, normally active within the pallium along a caudal–medialhigh-to-rostral–laterallow gradient, almost collapsed (Fig. 3a,b). This phenomenon possibly arose from mis-regulation of distinct classes of molecules involved in Wnt signalling, including ligands, membrane receptors, cytoplasmic modulators and nuclear modulators (Fig. 3q). Four ligand genes, Wnt3a, Wnt8b, Wnt5a and Wnt2b, normally expressed in the cortical hem as well as in the pallium, were dramatically down-regulated (Muzio et al., 2002a; Fig. 3c–f). Receptor genes Fzd9 and Fzd10, both normally restricted to the dorso-medial cortex (Kim et al., 2001), were also strongly down-regulated (Fig. 3g–j), ubiquitous Lrp5 and Lrp6 (He et al., 2004) were not impaired and the laterally confined (Kim et al., 2001) Fzd8 was only slightly weakened (Supplementary Fig. 2a–f). Axin2, encoding for a key-component of the cytoplasmic complex promoting beta-catenin degradation (Jho et al., 2002), was almost completely switched off (not shown). Tsh3, the archicortex-restricted mouse paralog of the armadillo cofactor gene Tsh (Gallet et al., 1998), was not affected (Supplementary Fig. 2g,h). Of the genes encoding for HMG-proteins bridging beta-catenin to its DNA targets and expressed in complementary cortical domains (Galceran et al., 2000), Lef1 was dramatically down-regulated while Tcf3 was only barely affected (Fig. 3m,n and Supplementary Fig. 2i,j). Groucho, encoding for the major co-repressor competing with beta-catenin for binding to these HMG-proteins (Cavallo et al., 1998), was ectopically activated in the archicortical anlage (Fig. 3o,p). Finally, Reptin and Pontin, exerting opposite effects on beta-catenin signalling (Bauer et al., 2000), were only slightly reinforced (Supplementary Fig. 2k–n). To sum up, canonical Wnt signalling collapsed, possibly because less ligand was synthesized, less receptor was available on the surface of cortical neuroblasts and the signal was relayed from their surface to the chromatin less efficiently.
To assess functional relevance of these changes, we pharmacologically reactivated Wnt signalling in Emx2−/− embryos via chronic administration of lithium to pregnant dams, and subsequently measured the fraction of neuroblasts, which, pulse-labeled by BrdU at E12.5, 12 h later switched β3-tubulin on. In the caudal third of cerebral cortex, this fraction equalled 0.071 ± 0.008, well below the value of 0.116 ± 0.012, peculiar to untreated Emx2 null embryos, and not far from the value of 0.062 ± 0.007, found in wild type embryos. Similar results were obtained one day earlier, at E11.5. All this turned out that exaggerated exit from cell cycle peculiar to Emx2 knockouts was largely due to massive down-regulation of Wnt signalling occurring in these mutants. Moreover, it suggested that Emx2 could regulate a large body of the molecular machinery controlling this aspect of population kinetics through beta-catenin. To test this hypothesis, we scored by in situ hybridization a subset of Emx2-sensitive genes, Kip1p27, Kip2p57, Ngn1, Id3, Dll1, Wnt8b and Lef1, as well as two beta-catenin responsive genes, Bat-gal and Axin2, as positive controls, in E12.0 Emx2−/− embryos, pre-exposed in vivo to chronic lithium for 8 h. Remarkably, in these embryos, only Kip2p57, Id3, Lef1, Bat-gal and Axin2 were restored (Fig. 4a–l and Supplementary Fig. 3a–c). We were not able to detect any rescue of either Kip1p27 or Ngn1 and cells expressing Dll1 became even rarer, as compared to untreated knockout embryos, possibly as a consequence of reduced neuronogenesis (data not shown and Supplementary Fig. 3d–l). Finally, changes of Id3 and Kip2p57 expression levels were confirmed by qRT-PCR on dissociated Emx2−/− cortical neuroblasts, grown in vitro under chronic lithium (not shown).
We have shown that, in the absence of Emx2, tangential expansion rates of the embryonic cerebral cortex PVE, and of its caudal–medial sector in particular, are reduced, because cortical progenitors proliferate more slowly and more leave the cell cycle. Due to TC elongation, the proliferative pool of the mutant caudal–medial cortex ‘loses’ one cell cycle out of four/five with respect to its wild type counterpart. Moreover, because of exaggerated neuronal differentiation, it is deprived of its components at even doubled rates. Thus, the mutant occipito-hippocampal field, already shrunken at the beginning (Muzio et al., 2002a), expands between E11 and E14 by less than half, as compared with its normal conterpart, so substantially contributing to the macroscopic areal phenotype peculiar to the late Emx2 null embryo (Bishop et al., 2000; Mallamaci et al., 2000). Because caudo-medial-specific kinetic defects appear most marked at early embryonic ages, early regional disparities in the Emx2 mutant cortex could in turn be largely dependent on such defects. Whatever the case may be, molecular mechanisms leading to these phenomena look extremely complex. Concerning cell cycle progression, one would expect that the up-regulation of Kip1p27 and Kip2p57 peculiar to Emx2 null brains would depress cdk2 activity and consequently inhibit both G1–S transition and S-phase progression, so lengthening both TG1 and TS. Actually, whereas DNA synthesis slowed down, we did not find any evidence to support TG1 lengthening, possibly due to additional, faster early G1-to-late G1 progression, following (Ho and Dowdy, 2002) the rise of cyclinD1 and cyclinD2 in these mutants. On the other hand, concerning exaggerated neuronogenesis, as many as four distinctive molecular abnormalities — up-regulation of Kips, increased Ngn/Id expression ratio, depression of the Delta/Notch/Hes axis and down-regulation of canonical Wnt signalling — could be responsible for it. At the moment, we have only incomplete information about their relative functional relevance and about any possible causal relationship connecting them. We found that canonical Wnt signalling, previously shown to promote Emx2 transcription through a beta-catenin binding module located within the Emx2 telencephalic enhancer (Theil et al., 2002), is in turn promoted by Emx2, possibly thanks to concerted Emx2-dependent modulation of four distinct functional layers of the machinery mediating it: ligands (Wnt3a, Wnt8b, Wnt5a, Wnt2b), surface receptors (Fzd9, Fzd10), intracellular beta-catenin agonists (Lef1) and intracellular beta-catenin antagonists (Groucho). In this way, near the cortical hem, a positive regulatory loop establishes between Emx2 and Wnt signalling, crucial for the proper sizing of occipital cortex and hippocampus. Moreover, we confirmed the functional relevance of this loop, showing that pharmacological reactivation of the canonical Wnt pathway in Emx2−/− mutants normalizes their neuroblast-to-neuron differentiation rates. We showed that this is associated to up-regulation of Id3 and down-regulation of Kip2, suggesting that Emx2 normally controls these genes through beta-catenin. Remarkably, despite the ubiquitous availability of lithium, the effects of stabilized beta-catenin on Id3, Kip2, Lef1, Bat-gal and Axin2 were much more prominent within the dorso-medial cortex than elsewhere (Fig. 4a–l and data not shown), suggesting the existence, deeper to Fzd receptors, of a differential, medialhigh-to-laterallow, cortical responsivity profile to Wnt ligands, possibly relying on confinement of Groucho to the lateral cortex (Fig. 3o,p) and beta-catenin-dependent stimulation of Lef1 transcription (Filali et al., 2002) in the medial cortex. It has to be emphasized, however, that not all information flows from Emx2 to the final regulators of population kinetics through beta-catenin (Fig. 4m). This is the case, for example, of Kip1p27, Ngn1 and Dll1, whose normal expression rates are not restored by lithium. This is also the case of cyclinDs, whose genes, normally promoted by canonical Wnt signalling (Castelo-Branco et al., 2003; Willert et al., 2002), are up-regulated in Emx2 null embryos, notwithstanding the collapse of Wnt signalling occurring in these mutants. Much work is still necessary to disentangle the fine topology of the network connecting Emx2 to these genes; however, the present data suggest that patterned transcription factor genes such as Emx2 could have several, distinct accession points (Fig. 4m), through which they could modulate activity of general molecular machineries driving neuroblast population kinetics in complex, nonlinear ways. Addressing this issue as well as trying to clarify the deep biological meaning of these interactions in cortical development will be the main subject of our next research.
This work was funded by EU (QLG3-CT-2000-00158; QLG3-CT-2000-01625). We thank N.Guida for contributing to kinetic characterisation of Emx2 null mutants. The authors declare that they have no competing financial interests.
1Istituto Scientifico H San Raffaele, DIBIT, via Olgettina 58, 20132 Milan, Italy and 2Department of Medical Biotechnology, University of Padua, viale Colombo 3, 35100 Padua, Italy