Projection neurons destined for the cortical plate are generated sequentially from the proliferative ventricular and subventricular zones (VZ/SVZ) of the pallium. However, the respective contribution of both proliferative zones to the generation of cortical plate neurons is better established in humans and non-human primates than in rodents. We identified Cux2 as a new marker for murine cortical subpopulations and used it to provide new insights to the development of the mouse cortex. Cux2 is an orthologue of the Drosophila cut gene, which encodes a homeodomain protein involved in neuronal specification. During cortical development Cux2 identifies two subpopulations with different spatial origins, migratory behaviours and phenotypic characteristics: (i) a population of interneurons, which invades the pallium from the subpallium; and (ii) a neuronal population produced in the pallium around embryonic day 11.5, which divides in the SVZ and accumulates in the intermediate zone (IZ). Subsequently, Cux2 is a marker of upper cortical layers. Using different molecular markers and Pax6-deficient mice, we provide data that suggest a relationship between the early-determined Cux2-positive neuronal precursors in the SVZ/IZ and upper layer neurons. This suggests that laminar determination of upper cortical layer neurons occurs during the earliest stages of corticogenesis.
The telencephalon derives from the anterior-most part of the neural tube. It mainly gives rise to the basal ganglia and the cortex. Although the mouse cortex is derived from the dorsal telencephalon (pallium), its neuronal populations have two embryonic origins. Projection neurons arise from the pallium (Iwasato et al., 2000; Gorski et al., 2002), while cortical interneurons are generated in the ventral telencephalon (subpallium) and invade the cortex by tangential migration (De Carlos et al., 1996; Anderson et al., 1997a, 1999, 2001; Tamamaki et al., 1997; Lavdas et al., 1999; Wichterle et al., 1999; Parnavelas, 2000; Marin and Rubenstein, 2001).
Orderly generation and radial migration of cortical neurons leads to the formation of five layers in an ‘inside-out’ pattern, with deep layers (DL: V–VI) composed of early-born neurons, and upper layers (UL: II–IV) of late-born neurons (Angevine and Sidman, 1961; Sidman and Rakic, 1973; Rakic, 1974). Two germinal zones have been implicated in corticogenesis, the pseudostratified epithelium of the ventricular zone (VZ) and the underlying subventricular zone (SVZ) (Boulder Committee, 1970). In humans and non-human primates, the SVZ is a major source of cortical projection neurons and appears to contribute particularly to the generation of UL neurons (Sidman and Rakic, 1973; Smart et al., 2002). Such a developmental process has also been suggested in mouse, with generation of DL by the VZ and UL by the SVZ (Smart and McSherry, 1982; Tarabykin et al., 2001). During the latest stages of embryogenesis the SVZ generates massively glial cells (Sidman and Rakic, 1973; Levitt and Rakic, 1980; Smart and McSherry, 1982; Bayer and Altman, 1991; Parnavelas, 1999).
An important question concerns the molecular determinants that define layer identity in the cortex (McConnell, 1995). For example, the T-box transcription factor, Tbr1, is essential for the correct differentiation of the Cajal–Retzius cells, the subplate, as well as layer VI neurons (Hevner et al., 2001). Otx1, which encodes for a POU homeodomain protein, is expressed early in the VZ, later by DL (Frantz et al., 1994) and postnatally plays an essential role in axonal refinement of layer V neurons (Weimann et al., 1999). Thus, as in other regions of the central nervous system (CNS), sequential and/or combinatorial transcriptional regulation is strongly implicated in the generation of neuronal subtypes (Jessell, 2000).
In this study, we identified Cux2 as a candidate factor for involvement in the determination of layer identity in the cortex. The Cux2 gene encodes a homeodomain transcription factor, which is an orthologue of the Drosophila cut gene. Cut specifies the identity of a subtype of sensory organs in the peripheral nervous system (PNS) (Bodmer et al., 1987), but is also expressed in the developing CNS (Blochlinger et al., 1990). In addition, the level of Cut protein regulates the patterning of dendritic branching in multidentritic neurons (Grueber et al., 2003).
Two vertebrate homologues of cut have been identified, Cux1 and Cux2 (Nepveu, 2001). Although both genes are expressed in many tissues, Cux2 is particularly strongly expressed in the nervous system (Valarche et al., 1993; Ellis et al., 2001; Iulianella et al., 2003).
We conducted a detailed analysis of the Cux2 expression pattern during development of the mouse dorsal telencephalon. This analysis shows that Cux2 is expressed in tangentially migrating interneurons that invade the pallium from the subpallium. Moreover, we show that a subpopulation of dividing precursors in the SVZ as well as UL projection neurons express Cux2. Combining Cux2 expression studies with the use of Pax6 mutants, we provide data that argue for a lineage relationship between SVZ precursors, appearing as early as embryonic day (E) 11.5, and UL neurons.
Materials and Methods
All animals were treated according to protocols approved by the French Ethical Committee. CD1 mice (Iffa-Credo, France) were used to analyze the Cux2 expression pattern and to perform BrdU injection studies. Mash1 mutant mice (Guillemot et al., 1993) were maintained on a C57BL/6J background and mice with a defective allele for Reelin (Reeler-Orleans mice, A. Goffinet), on a BalbC background. Brains of Pax6 mutants (Sey mice) and control were kindly provided by V. Tarabykin (E13.5, E15.5) and D.J. Kleinjan (E17.5). In all cases, mutant animals were compared with wildtype littermates. The day of appearance of the vaginal plug was considered E0.5, and the day of the birth termed postnatal (P) 0. Pregnant mice were killed by cervical dislocation. Embryos were harvested in cold phosphate-buffered saline (PBS), fixed in freshly prepared 4% paraformaldehyde/PBS (PFA), cryoprotected in 15% sucrose/0.12 M, pH 7.2 phosphate buffer (PB), embedded in 7.5% gelatine/15% sucrose/PB at 65°C, frozen in isopentane at −40°C and cut at 10 or 12 μm on a cryostat. Brains were dissected before fixation from E12.5 to E18.5. To collect P14 and adult brains, animals were anaesthetized with a mixture of Rompun/Imalgen500 and intracardiacally perfused with 4% PFA. Tissues were processed as described above for freezing except in Fig. 1GG′ and Fig. 3I–N, in which the adult brains were frozen in isopentane without being embedded.
In Situ Hybridization and Immunohistochemistry
To synthesize the Cux2 antisense riboprobe, a NotI–SacII DNA fragment corresponding to the 5′ part of the Cux2 full-length cDNA (Quaggin et al., 1996) was subcloned into pSK. This vector was linearized by NotI and the probe was produced using T3 RNA polymerase. The other probes used were: Dlx5 (Liu et al., 1997), Emx1 (Simeone et al., 1992), ER81 (Lin et al., 1998), Otx1 (Frantz et al., 1994). All antisense RNA probes were synthesized with DIG-RNA labelling kit (Roche) following the manufacturer's instructions. In situ hybridization and combined in situ hybridization with immunochemistry were performed as described previously (Tiveron et al., 1996; Hirsch et al., 1998; Dubreuil et al., 2000) for all probes, the pan-DLX (Dll-DLX) polyclonal rabbit IgG (1:100, (Panganiban et al., 1995) and the phosphohistoneH3 (PHH3) polyclonal rabbit IgG (1:200, #06-570, Upstate). In some cases, modifications were added after in situ hybridization: to detect proliferating cell nuclear antigen (PCNA) using a monoclonal mouse IgG (1:100, #MAB424, Chemicon), sections were treated with 70% ethanol at −20°C for 15 min. To detect NeuN using a monoclonal mouse IgG (1:100, #MAB377, Chemicon), sections were treated with 0.1 M, pH 6, boiling citric acid for 60 s or 3 min, respectively. Appropriate Elite ABC Vectastain kits (Vector) were used for the detection of the immunological signal. In Figure 5, adjacent sections were processed for Nissl staining. Pictures were taken with a Nikon DXM1200 camera using the ACT-1 software and assembled using Adobe Photoshop and Freehand.
BrdU Injection and Staining
Pregnant mice were injected intraperitoneally with a sterile solution of BrdU (100 mg/kg in PBS, Sigma) at 12.5 days post coitum and they were killed 6 h later. Brains were cut in serial coronal sections (10 μm thick) with a cryostat.
For BrdU detection, slides were post-fixed in 4% PFA for 10 min after in situ hybridization and washed with PBS. They were then incubated in HCl, 2N/0.1% Triton X-100/PBS, for 30 min at 37°C and washed with 0.1 M, pH 8.5 boric acid (Sigma) three times for 8 min at room temperature. Slides were then rinsed in PBS and incubated with anti-BrdU mouse monoclonal IgG (1:100, #674401, Dako) in PBS/10% FCS/0.1% Triton X-100 ON at 4°C. After three washes of 5 min with the blocking solution, sections were processed with the Elite ABC Vectastain kit (Vector).
Developmental Dynamics of Cux2 Expression in the Cortex
First expression of Cux2 was detected in the mouse telencephalon as early as E10.5, although at low levels (data not shown). At E11.5 and E12.5 (Fig. 1A,A′), Cux2 mRNA was found in both the ventral telencephalon (subpallium), which is the primordium of the basal ganglia, and the dorsal telencephalon (pallium), which gives rise to the future neocortex as well as the hippocampus (Rallu et al., 2002; Campbell, 2003). In the subpallium, Cux2 expression was mainly localized in the Medial Ganglionic Eminence (MGE), which generates neurons of the globus pallidus as well as dorsal interneurons (Fig. 1A,B) (Lavdas et al., 1999; Wichterle et al., 2001; Anderson et al., 2001, 2002). Furthermore, individual Cux2 expressing (Cux2+) cells were present in the Lateral Ganglionic Eminence (LGE), which gives rise to the striatum and to olfactory bulb interneurons (Fig. 1A,B) (Deacon et al., 1994; Lois and Alvarez-Buylla, 1994; Olsson et al., 1995; Wichterle et al., 1999, 2001; Stenman et al., 2003a).
In the pallium, Cux2 expression increased from E11.5 to E12.5 (Fig. 1A,B), appearing first in a layer underlying the VZ, which we define already as the SVZ to be in agreement with our further results (Fig. 1A′,B′) (Ishii et al., 2000). The Cux2 expression pattern followed the dorsomedial cortical maturation gradient and delineated the pallio–subpallial boundary (Fig. 1A,B). The VZ was mainly devoid of Cux2, whereas the epithelium of the cortical hem, which is implicated in the dorsal patterning of the pallium, showed strong Cux2 expression (Fig. 1A,B) (Monuki et al., 2001; Ragsdale and Grove, 2001). At E13.5 and E14.5, Cux2 labelling covered the enlarged SVZ and the intermediate zone (IZ; Fig. 1C,C′,D,D′), while single Cux2+ cells were visible in the developing cortical plate (CP). At this time point, a row of Cux2+ cells appeared in the marginal zone (MZ), which contains in particular Cajal–Retzius cells as well as interneurons invading the cortex from the MGE via tangential migration routes (Fig. 1C,C′,D,D′; Meyer et al., 2000; Zecevic and Rakic, 2001; Hevner et al., 2003a; Rakic and Zecevic, 2003). At E16.5, Cux2 labelling in the SVZ was clearly stronger than in the IZ (Fig. 1E,E′). In addition, large numbers of individual Cux2+ cells appeared in the already formed DL of the CP, while at the same time, Cux2 staining accumulated superficially in the CP. Two days later (E18.5), Cux2 labelling in the SVZ/IZ had strikingly diminished (Fig. 1F,F′). In contrast, the majority of the Cux2+ population was now positioned in the outer most part of the CP, which gives rise to the upper cortical layers (II–IV layers). In the postnatal and adult brain, Cux2+ cells were almost exclusively restricted to the upper layers of the cortex. Only few isolated cells were visible in the deep part (Fig. 1G,G′). At all developmental stages, the Cux2 expression pattern was homogeneous along the rostro-caudal axis except for minor differences due to the maturation gradient in the developing cortex (data not shown; McSherry and Smart, 1986; Bayer and Altman, 1991).
In conclusion, Cux2 expression in the developing telencephalon is highly dynamic and includes pallial and subpallial cell populations. During early stages it is mainly localized in the SVZ/IZ, while later in development Cux2 labels the outermost part of the cortex.
Cortical Interneurons Are One of the Cux2+ Subpopulations
We then aimed at identifying the Cux2+ cell populations in the developing telencephalon. The spatio-temporal expression pattern of Cux2 in the SVZ/IZ and also in the MZ was reminiscent of the tangential migration streams of GABAergic interneurons from the subpallium to the pallium (De Carlos et al., 1996; Anderson et al., 1997a, 1999, 2001; Tamamaki et al., 1997; Lavdas et al., 1999; Wichterle et al., 1999; Parnavelas, 2000; Marin and Rubenstein, 2001).
These interneurons are born as early as E11.5 in the subpallium, more precisely in the MGE, which seems to be the main source of cortical interneurons in mice (Lavdas et al., 1999; Anderson et al., 2001; Wichterle et al., 2001; Polleux et al., 2002; Nery et al., 2003), while in humans and non-human primates, GABAergic neurons are generated mainly in the pallium and only a small fraction is subpallial-derived (Zecevic and Rakic, 2001; Letinic et al., 2002; Rakic and Zecevic, 2003). After their generation in the ventral telencephalon, they migrate dorsally to populate all cortical layers as well as the hippocampus (Pleasure et al., 2000; Ang et al., 2003; Valcanis and Tan, 2003).
From E14.5 to E18.5, the comparison of Cux2 expression with that of Dlx5, a marker for cortical interneurons (Eisenstat et al., 1999; Stuhmer et al., 2002a), suggested that migrating interneurons expressed Cux2 (Fig. 2A–F). Particularly in the MZ, staining for Cux2 was highly comparable to that of Dlx5, while in the SVZ/IZ, the size of Cux2 population by far exceeded that of the Dlx5+ population. This suggested that at least two different cell populations expressed Cux2 in the SVZ/IZ.
We then used a pan-DLX antibody to clearly discriminate the cortical interneuron population in the developing cortex (Panganiban et al., 1995; He et al., 2001; Stenman et al., 2003a). At E15.5, all DLX-labelled cells localized in the pallium were also positive for Cux2 (Fig. 2G–I). These Cux2+/DLX+ cells were mostly confined to the SVZ/IZ and the MZ, although individual Cux2+/DLX+ cells were also detected in the VZ (Fig. 2I, arrows in the VZ). Nevertheless, the majority of the Cux2+ cells in the SVZ/IZ at this age were DLX negative. These cells showed in general a slightly lower intensity of Cux2 labelling than the double positive cells (Fig. 2H).
To further assess Cux2 as a marker for cortical interneurons we analysed its expression pattern in Mash1 knockout mice (Guillemot et al., 1993). These animals show a severe reduction of neuronal progenitors in the ventral telencephalon, and more specifically in the MGE (Casarosa et al., 1999; Horton et al., 1999). As a consequence, this leads to the diminution of cortical interneurons as identified by markers like GAD67, Dlx1 and Dlx5 (Casarosa et al., 1999; Horton et al., 1999). We found considerably less Cux2+/DLX+ in the mutant pallium (Fig. 2J) compared with the wildtype (Fig. 2I). However, the large number of Cux2+/DLX− cells was not affected in Mash1 mutants (Fig. 2I,J), suggesting that they are dorsally derived (Nakagawa et al., 1999). This cell population was mainly confined to the SVZ/IZ at E15.5 (Fig. 2I, but see also Fig. 2A at E14.5). At E16.5, Cux2+ individual cells are found in the innermost part of the CP, whereas Cux2 staining accumulated in the outermost part of the CP (Fig. 2B). At E18.5, the majority of the Cux2+ population was now located below the MZ in the position of the future upper cortical layers (Fig. 2C).
Altogether these results show that Cux2 is expressed in cortical interneurons that invade the pallium from the subpallium via tangential migration routes. Our data also strongly suggest that the Cux2+/DLX− population represents dorsally derived cells, which are known to migrate radially during later stages of corticogenesis, and to differentiate into cortical projection neurons (Tan et al., 1998; Nadarajah and Parnavelas, 2002).
Cux2 is a New Marker for Upper Cortical Layer Neurons
To characterize the dorsally derived Cux2+ population, we compared Cux2 expression to that of Otx1 and Emx1, two markers for neuronal subpopulations of the dorsal telencephalon (Simeone et al., 1992; Frantz et al., 1994; Cecchi et al., 2000; Chan et al., 2001).
Comparison of Cux2 expression (Fig. 3A) with that of Otx1 (Fig. 3B) at E13.5 demonstrated that Cux2 was strongly expressed in SVZ/IZ, whereas Otx1 was restricted to the VZ (Frantz et al., 1994; Zhang et al., 2002). Individual Cux2+ cells in the VZ corresponded to cortical interneurons (see arrows Fig. 2I). Both the Otx1- and the Cux2-positive cell populations were included in the expression domain of Emx1, a marker for most cells of the cerebral cortex in the process of proliferation, migration and differentiation (Simeone et al., 1992; Cecchi and Boncinelli, 2000; Chan et al., 2001). At E18.5, Otx1 is, in addition to the VZ, also expressed in the deeper part of the CP, which subsequently differentiates into the deep cortical layers V and VI (Fig. 3E). At this stage, single Cux2+ cells were visible within the Otx1+ domain. However, most Cux2+ cells accumulated superficially to the Otx1-labelled area, in the outermost part of the CP, which eventually gives rise to the II–IV upper cortical layers (Fig. 3D,E) (Frantz et al., 1994; Zhang et al., 2002). Again Cux2+ cells were included in the Emx1+ domain, which in addition to proliferative and migrating cells now contains differentiated cortical projection neurons (Cecchi and Boncinelli, 2000; Chan et al., 2001). In the postnatal cortex (P14), comparison of Cux2 with ER81, a marker for layer V neurons (Fig. 3G,H) (Xu et al., 2000; Nery et al., 2002), showed that Cux2+ cells were distinct from ER81+ cells and corresponded to the II–IV upper cortical layers. Their neuronal identity was demonstrated by double labelling using the pan-neuronal marker NeuN (Mullen et al., 1992). In the adult cortex Cux2 remained expressed in upper layer neurons (Fig. 3I,J) as well as in individual neurons distributed in the DL (Fig. 3K). These cells probably corresponded to cortical interneurons, since they expressed both Cux2 and DLX (Doetsch et al., 2002; Stuhmer et al., 2002b) and were distributed throughout all cortical layers (Fig. 3L–N).
To show that Cux2 is expressed in projection neurons of the upper cortical layers, we analysed its expression pattern in Reeler mice. Reeler is a mutant mouse strain in which the lack of the secreted molecule Reelin leads to sev ere alterations in cortical layering, namely an inversion of the normal ‘inside-out’ pattern (Caviness, 1982; Goffinet, 1984; D'Arcangelo et al., 1995; Ogawa et al., 1995; Rice and Curran, 2001; Magdaleno et al., 2002). During late phases of corticogenesis, at E18.5, Cux2+ neurons, which normally accumulated in the upper most part of the CP (Fig. 3O), remained in deep positions and appeared more diffuse and dispersed in the Reeler cortex (Fig. 3P). Cux2-positive cells localized in superficial positions in the Reeler cortex coexpressed DLX and thus represent cortical interneurons (Fig. 3Q).
The comparison of the Cux2 expression pattern with that of ER81 confirmed the inversion of cortical layering and highlighted the general loss of organization of the mispositioned neurons (Fig. 3O,P,R,S) (Hoffarth et al., 1995; Polleux et al., 1998). This showed that even in an inverted and disorganized cortex, Cux2 still identified the same neuronal populations.
Thus, in the telencephalon Cux2 is expressed by two subpopulations that present different spatial origins, migratory behaviours and phenotypic characteristics: a subpopulation of cortical interneurons born in the subpallium, which migrates tangentially into the pallium and a subpopulation of cortical projection neurons born in the dorsal telencephalon. These cells seem to accumulate in the SVZ/IZ before their radial migration to superficial positions, where they differentiate into upper cortical layer neurons (Tan et al., 1998; Nadarajah and Parnavelas, 2002).
Cux2 is Expressed by Cells that Divide in the SVZ
Two proliferative zones have been described in the developing telencephalon. The pseudostratified epithelium of the VZ and the SVZ that shows no apparent stratification (Boulder Committee, 1970). It has been reported that the murine SVZ is established around E13.5 (Smart and McSherry, 1982; Bayer and Altman, 1991), although mitotic figures have been identified under the VZ as early as E10.5 (Ishii et al., 2000). In the dorsal telencephalon, we found low Cux2 mRNA levels overlying the VZ as early as E10.5 (not shown) and considerable expression at E11.5 (Fig. 1A,A′). This expression in the SVZ/IZ increased until E16.5, when the Cux2 labelling started to shift to superficial positions (Fig. 1A–E).
To determine whether Cux2 was expressed by dividing cells in the SVZ, we used different cell cycle markers. First, we used PCNA, which is mainly expressed from G1-phase through S-phase of the cell cycle (Takahashi and Caviness, 1993; Bolton et al., 1994; Dehay et al., 2001). At E12.5, many Cux2+ cells in the SVZ were also labelled with PCNA (Fig. 4A,B). In contrast, cortical interneurons expressing Dlx5 were always negative for PCNA (Fig. 4C,D) not only at E12.5, but also at later developmental stages (E14.5, E16.5, data not shown) in agreement with previous observations (Anderson et al., 2001; Polleux et al., 2002; Xu et al., 2003). This suggested that at least a fraction of the Cux2+/PCNA− cells corresponded to postmitotic migrating interneurons and confirmed that the main Cux2+ population in the SVZ correspond to another cell type (see Fig. 2). We further assessed the presence of mitotically active Cux2+ cells in the SVZ using Phosphohistone H3 (PHH3) (Fig. 4E,F), which is expressed mainly from late G2-phase through the M-phase of the cell cycle (Weissman et al., 2003). Finally, we identified dividing cells via BrdU incorporation during the S-phase of the cell cycle (Takahashi et al., 1992). After a 2 h pulse, most cells in the VZ were localized in the basal part and were still in S-phase, whereas Cux2+/BrdU+ cells in the SVZ were undetectable, even at later developmental stages (data not shown). However, after 6 h of incorporation (Gotz et al., 1998) a subfraction of Cux2-positive cells in the SVZ showed BrdU labelling (Fig. 4G,H). Thus, Cux2 seems to be expressed by cells during the late G2-phase and/or the M-phase (Cux2+/PHH3+) as well as the G1 phase of the cell cycle (Cux2+/PCNA+ and Cux2−/BrdU+ only 6 h after BrdU pulse; see also discussion).
In conclusion, these data show that at E12.5 an early pool of dividing neuronal progenitors in the SVZ expresses Cux2. Moreover, the observation that Cux2+/PCNA+ as well as Cux2+/PHH3+ cells in the SVZ could be identified at later developmental stages (E14.5, E16.5; data not shown) suggests that Cux2 is expressed by cells that divide in the SVZ throughout cortical development.
Analysis of the Pax6 Mutant Cortex Suggests a Relationship between Cux2+ SVZ/IZ Pool and UL Neurons
Our findings that Cux2 is expressed by a subpopulation of cells that divide in the SVZ, combined with the dynamics of Cux2 expression and its identification as a marker for upper cortical layers, raised the issue whether the SVZ-Cux2+ subpopulation gives rise to upper layers. Indeed, such a lineage relationship between SVZ cells and upper layers has previously been proposed in mouse (Smart and McSherry, 1982; Tarabykin et al., 2001) as well as in humans and non-human primates (Sidman and Rakic, 1973; Smart et al., 2002).
To address this point, we used Sey mice, in which the mutation of the Pax6 gene has been suggested to induce a massive accumulation of UL neurons within the enlarged SVZ, while DL neurons appear to migrate relatively normally (Caric et al., 1997; Tyas et al., 2003). To discriminate Cux2+ interneurons from Cux2+ UL neurons, in situ hybridization for Cux2 was again combined with pan-DLX immunolabelling.
At E13.5, the Cux2-positive population in the SVZ was severely diminished in the mutant, whereas the DLX-positive population was augmented (Fig. 5C–F). At this stage DL neurons are migrating to the CP in wildtype as well as in Pax6 mutants. Thus, the reduction of Cux2 labeling in mutants demonstrates that its expression in the wildtype is restricted to a neuronal subpopulation and does not represent a maturation step realized by all cortical projection neurons.
At E15.5, the Cux2+/DLX+ population of cortical interneurons in the MZ was increased in Pax6 mutants (Fig. 5B,D,E) (Chapouton et al., 1999), while we observed a loss of these interneurons in the SVZ/IZ (see Fig. 2). The latter ones might have joined the MZ migration pathway as proposed previously (Stoykova et al., 2003). At the same time point the Cux2+/DLX− population in the SVZ/IZ was also strongly reduced (Fig. 5B,D, white arrow in K).
At E17.5, the SVZ/IZ of Pax6 mutants showed a severe reduction of Cux2 expression, comparable to the loss seen for the Svet1 marker (Tarabykin et al., 2001). However, this region contained now large amounts of DLX+ cells that did not express Cux2 (Fig. 5G,J). At the same time the CP showed only few condensed Cux2+ UL neurons (Fig. 5H,K).
These findings are in contradiction to the previous interpretation, that a migration defect of UL neurons is the cause of the enlarged SVZ (Caric et al., 1997). However, our results are in agreement with a ventralization of the Pax6 mutant cortex including a dorsal shift of the pallio–subpallial boundary (Kim et al., 2001; Assimacopoulos et al., 2003; Stenman et al., 2003b). This shift induces a change of fate of the dorsal telencephalic neurons which now adopt a striatal identity as shown by their DLX+/Cux2− phenotype (Fig. 5D,E,J) (Anderson et al., 1997b).
In conclusion, these analyses show that the reduction of the Cux2+ population in the Pax6 mutant SVZ/IZ is correlated to the loss of UL neurons. Our data demonstrate that Cux2 expression is not linked to a maturation step of all cortical projection neurons, but that it identifies a subpopulation of projection neurons that is derived from the pallium. Furthermore, they suggest a relationship between the Cux2+ precursor pool in the SVZ and UL neurons. Finally, they argue that Pax6 acts upstream of Cux2 and is specifically implicated in the determination of the UL fate, since DL seem to differentiate relatively normally (Fig. 5L,O) (Caric et al., 1997).
A Model for Mouse Cortical Neurogenesis
Taken together, our data suggest that Pax6-dependent precursors localized in the SVZ generate UL neurons (Smart and McSherry, 1982; Tarabykin et al., 2001). They also argue that Cux2-expressing cells in the SVZ are fated to become UL neurons. As in humans and non-human primates, the SVZ may be the source of UL neurons in the mouse cortex (Sidman and Rakic, 1973; Smart et al., 2002). This leads us to refine the previously proposed models for the genesis of DL and UL neurons in the mouse cortex (Fig. 6) (Smart and McSherry, 1982; Tarabykin et al., 2001). In agreement with previous transplantation and lineage studies, DL and UL precursors might form two separate pools (Fishell et al., 1990; Krushel et al., 1993; Kornack and Rakic, 1995) that arise from the same multipotent progenitor pool at around E10.5 (Walsh and Cepko, 1988; McConnell and Kaznowski, 1991; Reid et al., 1995, 1997; Tan et al., 1998). In the following, two populations are produced in the VZ: a post-mitotic population of DL neurons and a population of UL neuron precursors that divides in the SVZ (E12.5 in Fig. 6). Based on the dynamics of Cux2 expression (Fig. 1), we propose that these newly generated UL neurons migrate from the SVZ into the IZ, which shows the major increment in thickness in the mouse cortex at E14.5 (Takahashi et al., 1993). Post-mitotic UL neurons may stay in this ‘sojourn zone’ (Bayer and Altman, 1991) for some time, until they receive the appropriate signal(s) to migrate radially through the CP and reach its outer most part (E16.5 in Fig. 6). The process of division of upper layer progenitors in the SVZ as well as their staging in the IZ appears to be continuous throughout corticogenesis (from E12.5 to E18.5) until the terminal differentiation of UL neurons (P14 in Fig. 6).
We show here that Cux2 expression identifies two subpopulations of cortical neurons that present different spatial origins, migratory behaviours and phenotypic characteristics: (i) GABAergic interneurons, born in the subpallium, which migrate tangentially into the pallium; and (ii) a neuronal population produced in the pallium around E11.5, which divides in the SVZ and accumulates in the intermediate zone (IZ). Later, Cux2 is a marker of upper cortical layer neurons. Moreover, Cux2 seems to be expressed from birth to terminal differentiation of UL neurons, thereby providing a new tool to gain deeper insights into the complex process of corticogenesis (cf. Fig. 6).
Cux2, an Early Molecular Determinant for UL Neurons
The large number of different cell types that build the cortex and the establishment of complex neuronal networks implicate a perfect orchestration of these processes during development. It turns out that in the mouse, the cortex has elaborated precise strategies to perform this task in only 8 days. One of these strategies is the use of different sources for neuron generation, as it is the case for the pallial origin of cortical projection neurons and the subpallial origin of interneurons (Iwasato et al., 2000; Wichterle et al., 2001; Gorski et al., 2002). Furthermore, the generation of glial cells takes place in a second proliferative zone under the pseudostratified epithelium of the VZ, the SVZ (Smart and McSherry, 1982; Bayer and Altman, 1991).
In humans and non-human primates, the early decrease of the VZ and the increase of the SVZ indicate that the latter is the major site of projection neuron production (Sidman and Rakic, 1973; Smart et al., 2002). Furthermore, in these species the pallial SVZ generates interneurons in addition to glial cells (Levitt and Rakic, 1980; Zecevic and Rakic, 2001; Letinic et al., 2002; Rakic and Zecevic, 2003). In rodents, the SVZ was considered as a germinal area for some neurons as well as glial cells (Boulder Committee, 1970; Smart and McSherry, 1982; Bayer and Altman, 1991); however, it has also been reported to be principally a source of glial cells late in development (Parnavelas, 1999). Hence, the time of apparition of the murine SVZ, as well as its contribution to the generation of cortical plate neurons is not well established. Smart and McSherry (1982) proposed, on the basis of histological analyses, that early detectable non-surface mitotic figures produce neurons to supply the rapid increase of the cortical plate. Recently, Haubensack et al. (2004) have demonstrated — using time-lapse video microscopic studies of a Tis21-GFP knock-in mouse — that such mitotic cells are already present at E10.5 (Haubensak et al., 2004). Furthermore, they found that these basal progenitors divide symmetrically to generate neurons as early as E12.5. Moreover, using the molecular marker Svet1, a transcript that appears in the SVZ around E13.5 and at later stages labels UL neurons, Tarabikyn et al. (2001) postulated a lineage relationship between the SVZ and UL neurons. Altogether, these findings are in agreement with the notion that the murine SVZ is a neurogenic compartment.
This notion is also in agreement with the new concept that radial glia generates cortical neurons (Malatesta et al., 2000; Hartfuss et al., 2001; Miyata et al., 2001; Noctor et al., 2001, 2002; Tamamaki et al., 2001; Rakic, 2003; Weissman et al., 2003). Indeed, using retroviral infection experiments in the E16 rat cortex, Noctor et al. (2004) recently demonstrated that, in addition to post-mitotic neurons, radial glia produces a population of ‘intermediate progenitors’ that divides symmetrically in the SVZ to generate pairs of neurons. They postulated that these neurons may correspond either to not yet identified pallial-derived interneurons as in humans (Letinic et al., 2002; Rakic and Zecevic, 2003), or to upper cortical layer neurons. Moreover, they could follow the radial migration of these pairs of neurons towards the cortical plate. This process of radial migration seems to be preferentially used by upper layer neurons since it was mainly observed at mid and late developmental stages, while soma translocation seems to be used by the earlier generated deep layer neurons (Nadarajah et al., 2001, 2003). However, interneurons seem also to migrate radially after their tangential migration to populate the cortical layers (Polleux et al., 2002). Our study shows that Cux2 is expressed by cells that divide in the SVZ from E12.5 to E16.5 (Fig. 4 and data not shown) and that do not correspond to cortical interneurons (Figs 2 and 4). The pool of SVZ/IZ Cux2 expressing cells decreases around E16.5, while the pool of Cux2 expressing cells in the outermost part of the cortical plate increases (Fig. 1). Cux2 remains expressed by UL in the adult cortex (Fig. 3I,J). We thus propose that at least a fraction of the ‘intermediate progenitors’ identified by Noctor et al. (2004) corresponds to UL neuron precursors.
Moreover, if one hypothesizes that Cux2 expressing cells in the SVZ and UL neurons share a lineage relationship, this would mean that UL precursors are determined as soon as corticogenesis starts. In this case, these UL precursors do not migrate immediately after their exit from the cell cycle, but accumulate in the IZ, thereby forming a reservoir of post-mitotic and predetermined neurons. This hypothesis is in concordance with transplantations studies in the ferret. Frantz and McConnell (1996) transplanted late VZ/SVZ progenitors in an early progenitor environment. They showed that after 8–9 days, about one-half of the transplanted cells were migrating in the CP, while the other half was still localized in the IZ and migrated to the CP only 1 week later. The authors attribute this observation to different speeds of migration within the population. However, our proposition of a reservoir of post-mitotic UL fated neurons in the IZ allows an alternative explanation, namely that heterochronically transplanted cells stop for a certain time in their normal transitory compartment, the IZ.
Thus, in addition to generating astrocytes, the embryonic mouse SVZ seems to generate at least a substantial amount of UL neuron precursors as in human and non-human primates (Sidman and Rakic, 1973; Smart and McSherry, 1982; Tarabykin et al., 2001; Smart et al., 2002). In the adult, the SVZ houses a ‘niche’ of stem cells, which have the competencies of self-renewing and generating neuronal precursors as well as astrocytes in mouse (Doetsch et al., 1999; Temple and Alvarez-Buylla, 1999; Tramontin et al., 2003).
What Could Be the Function(s) of Cux2 during Mouse Corticogenesis?
We found that Cux2 was expressed by cells in late G2 and/or M-phases of the cell cycle in the murine SVZ as shown by the double labeling for PHH3 (Fig. 4E,F). In addition, we found many Cux2+/PCNA+ cells that did not correspond to migrating interneurons (Fig. 4A–D) (Anderson et al., 2001; Polleux et al., 2002; Xu et al., 2003). PCNA is a marker for cells, which are either in G1 or S phases of the cell cycle. However, the absence of Cux2+/BrdU+ cells in the SVZ after a two-hours pulse suggests that Cux2 is not expressed during the S-phase of the cell cycle but rather during the G1-phase (data not shown; see Takahashi et al., 1992). The observation that 6 h after BrdU incorporation, Cux2+/BrdU+ cells were detected in the SVZ, could indicate that (i) these cells have a slower cell cycle or (ii) Cux2 is specifically turned on during late-G2 or M phases of the cell cycle. Since available BrdU rapidly decreases (Hayes and Nowakowski, 2000), the second hypothesis appears likely. Furthermore, time-lapse microscopic studies that followed cell divisions in the SVZ report that neuronal ‘intermediate progenitors’ undergo their last mitosis in this compartment (Haubensak et al., 2004; Noctor et al., 2004). Altogether, these data suggest that either Cux2 is expressed in particular phases of the cell cycle, or that it is turned on just before (G2) or during the last mitosis of the neuronal ‘intermediate progenitors’ of the SVZ. These cells might be unable to overcome the G1/S transition, which seems to be a critical regulator of corticogenesis (Miyama et al., 1997; Caviness et al., 1999; Delalle et al., 1999).
This hypothesis is in agreement with the correlation established between the state of a progenitor in the cell cycle and its laminar fate. Hence, progenitors transplanted in an older environment show plasticity concerning their laminar fate if grafted during the S-phase of the cell cycle, but show laminar determination when grafted late in the cell cycle (McConnell and Kaznowski, 1991). These data also argue that laminar determination and cell cycle progression are closely linked (McConnell and Kaznowski, 1991; Takahashi et al., 1999; Caviness et al., 2003). However, transplantation studies of Desai and McConnell (2000) demonstrate that laminar identity is not linked to a ‘clock-like’ counting of cell division (Takahashi et al., 1999). Our data support this analysis since we show that a Cux2+ cell population is present at the onset of corticogenesis and that a subpopulation of Cux2+ neurons becomes post-mitotic at the same time as DL neurons (Figs 1 and 5). In addition, it has been shown that a neuron which becomes post-mitotic at late developmental stages can be fated to DL, provided that it expresses DL molecular determinants (Hevner et al., 2003b). Altogether, these findings are in agreement with a model in which all projection neurons are molecularly determined even during earliest stages of corticogenesis, comparable to what has been suggested for the neural populations in the spinal cord (Jessell, 2000).
The Cux2 homologue CDP/Cux1 seems to play a general role in preventing precursor cells to express genes involved in differentiation (Superti-Furga et al., 1989; Lievens et al., 1995; Pattison et al., 1997; Stunkel et al., 2000; Ellis et al., 2001; Ledford et al., 2002; Luong et al., 2002). Cux1 activity is modulated throughout cell cycle progression: it increases during the G1/S transition, but decreases during the G2/M transition (Coqueret et al., 1998; Moon et al., 2001; Santaguida et al., 2001). In the PNS of Drosophila, cut is expressed by the sensory organ precursor (SOP) as well as its entire lineage. In cut mutants, external sensory (es) organs are transformed into another type of sensory organ, the chordotonal organ, showing that cut determines the identity of es organs (Bodmer et al., 1987). Furthermore, in the mature PNS, the level of Cut protein in multidendritic neurons has been correlated to the patterning of dendritic branching (Grueber et al., 2003).
In summary, it appears possible that the duplication of the Cux genes in vertebrates has a role in cell cycle regulation, with the Cux1 gene implicated in proliferation and the Cux2 gene in specification as well as differentiation. Indeed, Cux2 is expressed by mature UL neurons and might participate in the establishment and maintenance of UL connections as it has been proposed for the molecular determinants ER81 and Otx1 (Frantz et al., 1994; Bohner et al., 1997; Weimann et al., 1999; Hevner et al., 2003b).
In conclusion, our data support the early determination of laminar identity in the cortex. Cux2 seems to be an early molecular determinant of UL cortical neurons, potentially via a role in cell cycle regulation of ‘intermediate progenitors’ in the SVZ (Haubensak et al., 2004; Noctor et al., 2004). However, their cell cycle can also be regulated by extrinsic factors (Haydar et al., 2000; Dehay et al., 2001). Altogether, this implicates that the VZ neuroblasts that give rise to these SVZ progenitors are already pre-determined. The transcription factor Pax6, which has been implicated in cell cycle regulation and radial glia differentiation (Gotz et al., 1998; Estivill-Torrus et al., 2002; Heins et al., 2002), might play this role of pre-determination of UL neuron progenitors in the VZ.
We thank F. Guillemot for Mash1 mutants and helpful discussion, V. Van Heyningen, D.J. Kleinjan and V. Tarabykin for Sey brains and A. Goffinet for Reeler mice. We thank J. Rubenstein (Dlx5), T. Jessell (ER81), V. Taylor (Emx1), R. Dono (Otx1) and G. Panganiban (Dll-DLX) for antibodies and in situ probes, and M.R. Hirsch for advice on in situ hybridization. We are grateful to P. Durbec, V. Castellani and A. Represa for helpful discussions, and to N. Dahmane and P. Carroll for critical reading of the manuscript. This work was supported by grants from Association pour la Recherche contre le Cancer (ARC) and the French Ministery of Research (A.C.I.). C.Z. was supported by a fellowship from the Fondation pour la Recherche Médicale (FRM).
1Developmental Biology Institute of Marseille, NMDA, Campus de Luminy case 907, 13288 Marseille Cedex 9, France and 2The Burnham Institute, 10901 North Torrey Pines Road, La Jolla, CA 92037, USA