The Role of Foxg1 and Dorsal Midline Signaling in the Generation of Cajal-Retzius Subtypes

Cajal-Retizus cells of both cortical hem- and PSB-derived character are expanded in the absence of Foxg1

To elucidate the molecular basis of CR cell development, we first assessed whether Foxg1 represses CR cell production through antagonizing dorsomedial identity. We have therefore examined in detail whether the increased production of CR cells in Foxg1^−/− nulls correlates with the expansion of the dorsomedial signaling center. In Foxg1^−/− mutants, the onset of production of CR neurons is normal. However, the period during which CR neurons are produced in these mutants is prolonged, resulting in the supernumerary production of CR cells in the cortex by late embryogenesis (Hanashima et al., 2004). Foxg1 is known to antagonize BMPs (Dou et al., 1999). At E11.5, Foxg1 and BMP exhibit a complementary expression pattern within the telencephalon (Fig. 1). Bmp (Bmp4, Bmp6) and Wnt (Wnt3a, Wnt2b) expressing cells are normally restricted to the dorsomedial wall of the telencephalon demarcating the cortical hem (Fig. 1A–D), with the exception of Wnt8b whose expression expands farther into the lateral pallium (Fig. 1E). In the absence of Foxg1, the size of the telencephalon is smaller as a result of reduced proliferation of cortical progenitors (Xuan et al., 1995). However, both Bmp and Wnt expression are upregulated and expanded in the dorsomedial pallium (Fig. 1F–J). As a result, the expression of these genes in this mutant extends farther anteriorly into the rostral telencephalon but does not cover the entire pallium (Figs. 1F–J, 2F) (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Expansion of the dorsomedial signaling center in Foxg1^−/− mutants. Top, Complementary expression of Foxg1 and Wnt3a, and BMP4 in E11.5 Foxg1^+/− control embryos. Arrows indicate the boundary between the cortical hem and the isocortex. Scale bar, 200 μm. Bottom, Expression of multiple BMPs (BMP4, BMP6) and Wnts (Wnt3a, Wnt2b, Wnt8b) mRNA in E11.5 whole-mount embryos (dorsal view). A–E shows expression of these genes in the Foxg1^+/− control embryos. F–J shows expression of the same genes in Foxg1^−/− mutants. Figures in the insets show the lateral view of the telencephalon of each corresponding embryos (rostral is to the left). BMPs and Wnts are both expanded laterally and rostrally in the Foxg1^−/− telencephalon. Scale bar, 1 mm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

A, B, D, E, Both cortical hem- and PSB-derived CR cells are overproduced in the absence of Foxg1. Top, Expression of p73 (red) and Reelin (green) in E11.5 Foxg1^+/− control (A, B) and Foxg1^−/− (D, E) embryos. a′–d‴ show the enlarged view of the boxed region of medial (a′, d′), lateral (a″, d″), and ventral (a‴, d‴) pallium indicated in A and D, respectively. The ventricle is to the bottom, and the pial surface is to the top. The p73-expressing cells are detected in both the ventricular zone and postmitotic cells at E11.5. p73 marks CR cells before their expression of Reelin (a′). Reelin-expressing cells in the ventral pallium do not express p73 (a‴). C, F, Wnt3a mRNA expression in sections adjacent to those indicated in B and E. Arrowheads demarcate the boundary of p73-expressing cells at this stage. Scale bars: A–F, 200 μm; a′–d‴, 100 μm. Bottom, Reelin (green) and p73 (red) expression in E13.5 Foxg1^+/− control (G, Ia–IIc) and Foxg1^−/− (H, IIIa–IVc) embryos. Ia–Ic, IIIa–IIIc show the dorsal pallium, and IIa–IIc, IVa–IVc show the ventral pallium. Arrowheads in Ic indicate double p73/Reelin-positive cells in the marginal zone in Foxg1^+/− controls. In Foxg1^+/− cortex, p73 expressing CR cells are distributed in the lateral cortex (I), and many of them coexpress Reelin (Ib, Ic). In the Foxg1^−/− cortex, many Reelin⁺ cells express p73 in the lateral pallium (IIIa–IIIc), but Reelin⁺ cells are p73 negative in the ventral pallium (IVa–IVc). Scale bar, 100 μm.

The increase in CR cells in the lateral pallium of Foxg1^−/− mutants (Hanashima et al., 2004), a region beyond the territory of BMP and Wnt expression, led to the question of whether this region gives rise to a distinct subtype of CR cells. Notably, CR neuron subtypes can be molecularly distinguished by distinct gene expression patterns (Meyer et al., 2004; Yamazaki et al., 2004; Bielle et al., 2005). For instance, p73 is an early marker for CR neurons derived from the cortical hem, and its expression is observed in these cells before the initiation of reelin expression during development. At E11.5, CR cells from the dorsal midline are p73 positive (p73⁺)/reelin negative (reelin⁻) (Fig. 2a′), although reelin is already highly expressed in CR cells arising from the PSB at this age (Fig. 2a‴). In mutants that lack p73, cortical hem-derived CR cells are lost but the PSB-derived CR population is unaffected (Meyer et al., 2004). Indeed, whereas the majority of reelin-positive CR cells in the lateral pallium coexpress p73 (Fig. 2Ia–Ic) at E13.5, p73 expression is excluded from the PSB-derived CR cells populating the ventral pallium at this stage (Fig. 2IIa–IIc). Conversely, these CR cells express Ebf2, which is specifically upregulated in the ventral pallial CR cell population (supplemental Fig. S1, available at www.jneurosci.org as supplemental material) (Yamazaki et al., 2004).

The ability to distinguish between CR cells arising from the dorsal midline versus the PSB prompted us to investigate whether the ectopic and supernumerary CR cells in the Foxg1^−/− nulls were all cortical hem derivatives that coexpress p73 and reelin. As in control animals, the expression of p73⁺ cells did not overlap with that of reelin⁺ cells at this early stage (Fig. 2D,d′–d‴). At E11.5, both the p73⁻/reelin⁺ CR cells within the lateral pallium (Fig. 2D,d″) and the p73⁺/reelin⁻ CR cells within the dorsomedial pallium (Fig. 2E,d′) are present. However, it should be noted that, because of the loss of ventral regions in Foxg1^−/− mutants, dorsomedial areas are extended rostroventrally at this stage (Figs. 1F–J, 2d‴). As a result, the expression of p73⁺ cells appear discontinuous when cut in coronal section as demonstrated in Figure 2, E and F. Either whole-mount analysis or sagittal sections reveals the p73 expression domain to be continuous along the rostral pole (data not shown). Notably, cortical hem-derived (p73⁺/reelin⁻) CR cells are expanded in a manner that correlates with the expanded Wnt3a expression observed in these mutants (Fig. 2B,E, arrowheads, C,F). Importantly the overproduction of CR cells only becomes evident in Foxg1^−/− during mid-neurogenesis. At E13.5, both the number of PSB-derived and cortical hem-derived CR cells was expanded in the Foxg1^−/− mutants (Fig. 2IIIc,IVc). In the lateral pallium, the majority of reelin⁺ cells coexpressed p73 (Fig. 2IIIa–IIIc), whereas in the ventral pallium, CR cells only expressed reelin and did not express p73 (Fig. 2IVa–IVc). Thus, the expanded dorsomedial signaling center in Foxg1^−/− directly correlates with cortical hem-derived CR cell overproduction. Moreover, loss of Foxg1 results in expansion of both cortical hem- and PSB-derived CR cell character.

Only PSB-derived CR cells are overproduced during conditional inactivation of Foxg1

The removal of Foxg1 throughout the entire cortical progenitor domain from the onset of CR neurogenesis resulted in expansion of both cortical hem- and PSB-derived CR subtypes. Therefore, we next tested whether we can specifically bias the overproduction toward the PSB-derived CR subtype by expanding the competence window of CR cell production without altering dorsomedial signaling. To achieve this, we used a conditional approach (Hanashima et al., 2004) to remove Foxg1 function in vivo at later development stages (E13), a time point when the cortical hem structure is already well established. Removal of Foxg1 function at this stage results in resumption of CR cell production (Fig 3J,j′), with ectopic and supernumerary CR cells being produced throughout the mediolateral extent of the cortex. To assess whether dorsomedial signaling is altered in these mutants, we examined the expression of BMPs as well as a downstream target of BMP proteins, Msx1, and the cortical hem marker Wnt3a. After inactivation of Foxg1 at E13, multiple dorsomedial ligands, Wnt3a, BMP4, and BMP6, as well as Msx1, expression were restricted to their normal distribution within the cortical hem (Fig. 3C,D) (supplemental Fig. S2, available at www.jneurosci.org as supplemental material), choroid plexus (Msx1 expression in Fig. 3d′), and fimbria (Wnt3a and Msx1 expression in Fig. 3H,I), when examined at early (E14.5) and late (E16.5) time points. Furthermore, the graded expression of both Emx1 and Emx2 (medial^high to lateral^low) were maintained in these mutants (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). Hence, ablation of Foxg1 during this period, unlike null mutants, does not result in an obvious respecification of the lateral cortical progenitor domain to a dorsomedial fate. These results demonstrate that, on removal of Foxg1 at E13, cortical progenitors are competent to produce supernumerary numbers of CR cells in the absence of any apparent expansion of dorsomedial signals.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Dorsomedial signaling is not expanded in Foxg1 E13 conditional mutants. A–D, Top, Coronal sections showing Wnt3a and Msx1 expression in E14.5 Foxg1^tTA/+ heterozygote control (A, B) and Foxg1 conditional (removed at E13) mutants (C, D). Arrows indicate that the expression of both genes is restricted to the cortical hem and the choroid plexus. a′–d′ show an enlarged view of the region indicated in A–D. Middle, Wnt3a, Msx1, and reelin expression at E16.5. E–G show Foxg1^tTA/+ control embryos. In H–J, Foxg1 conditional (removed at E13) mutants are shown. Insets in E, F, H, and I show an enlarged view of the region indicated by the arrows. The increased production of CR cells is indicated by the expansion in reelin expression in both the medial and lateral cortex of Foxg1 conditional (removed at E13) mutants (J), whereas expression is confined to the marginal zone in the lateral cortex in Foxg1^tTA/+ control mice (G). g′ and j′ show an enlarged view of the boxed region in G and J, respectively. Bottom, Expression of Ebf2 (K, O), Reelin (L, P), and p73 (M, Q) in Foxg1^tTA/+ (K–N) and Foxg1 E13 conditional mutants (O–R). Pia is to the top. The ectopic and supernumerary CR cells in the cortical plate express reelin but do not coexpress p73 in the conditional mutants (R). Note the background staining in the pia in L–N, which do not represent CR cells. cp, Cortical plate; iz, intermediate zone. Scale bars: A–D, E–J, 0.5 mm; a′–d′, g′, j′, K–R, 100 μm.

We next examined whether the expanded CR population in Foxg1 E13 conditional mutants represent a biased production of CR subtypes. We used p73 as the marker for cortical hem-derived subtype, as well as Ebf2, which is specifically expressed in the ventral pallial CR cell population (supplemental Fig. S1, available at www.jneurosci.org as supplemental material) (Takiguchi-Hayashi et al., 2004). In control animals at E16.5, both p73 and reelin expression was confined to CR cells restricted to the marginal zone (Fig. 3L–N). Notably, the number of p73-expressing CR cells within the marginal zone of Foxg1 conditional mice was comparable with that seen in controls (Fig. 3Q,R). In contrast, in E13 Foxg1 conditional mutants, the ectopic and supernumerary CR cells in the cortical plate expressed reelin (Fig. 3P) but did not coexpress p73. Furthermore, Ebf2 expression was upregulated in newly-born neurons after Foxg1 removal in the E13 mutant cortices (Fig. 3O) (supplemental Fig. S3, available at www.jneurosci.org as supplemental material), whereas its expression was restricted to the marginal zone in controls (Fig. 3K). These results demonstrate that, in the absence of overt expansion in dorsomedial signaling, inactivation of Foxg1 function selectively elongates the competence window for PSB-derived (but not hem-derived) CR cell production.

Downregulation of Foxg1 in the absence of dorsomedial signaling center results in overproduction of PSB-derived CR cells and the loss of the cortical hem-derived CR subtype

The biased overproduction of PSB-derived CR cells in E13 Foxg1 conditional mutants suggests that Foxg1 is responsible for suppressing the competence of CR neuron production in general, regardless of their origin. However, in Foxg1 conditional mutants, the dorsomedial signaling center still persists, and whether this contributes to the supernumerary production of CR cells is unclear. We therefore examined whether downregulation of Foxg1 expression can specifically expand the number of PSB-derived CR cells in the absence of dorsomedial signaling. This also allowed us to assess whether the dorsomedial signaling center is itself dispensable for the production of CR cells. We addressed this question by using the extra-toes^J (Xt^J) mutant, which carries an intragenic deletion of the Gli3 gene (Hui and Joyner, 1993; Rallu et al., 2002). Specifically, Gli3 homozygotes fail to establish the entire BMP- and Wnt-expressing dorsomedial signaling center (Grove et al., 1998; Tole et al., 2000). Unexpectedly, however, the complete removal of Gli3 in the absence of Foxg1 (i.e., Foxg1^−/−Gli3^−/− mice) results in loss of telencephalic identity (supplemental Fig S4, available at www.jneurosci.org as supplemental material). This at least partially appears to be attributable to a requirement for Gli3 for the maintenance of Foxg1 expression. In these mice, cells expressing Foxg1 (as assessed through Cre expression in Foxg1^Cre/Cre;Gli3^−/− mice) acquire a mixed telencephalic–diencephalic character. This was revealed by the overlap of Cre expression driven from the Foxg1 locus and Foxd1 expression, a gene normally restricted to the diencephalon (Hatini et al., 1994; Herrera et al., 2004). These results suggest that Gli3 is required for the maintenance of telencephalic identity in the absence of Foxg1. Consistent with this, the level of dorsal Foxg1 expression is affected by the level of Gli3. In Gli3 mutants, Foxg1 and Cre transcripts were reduced specifically in the dorsal but not ventral telencephalon in both Foxg1^+/+ (Foxg1 transcript) and Foxg1^−/− null backgrounds (Cre transcript). As shown in supplemental Figure S5 (available at www.jneurosci.org as supplemental material), removing one copy of Gli3 significantly affected the dorsal expression of the Cre transcript in the Foxg1^Cre/Cre null background, and loss of both copies of Gli3 resulted in a reduction of the expression of Foxg1 in a Foxg1^+/+ wild-type background. We therefore assessed whether the observed downregulation of Foxg1 expression in Gli3^−/− cortices alters the production and specification of CR cells in these mutants.

In E10.5 and E13.5 Gli3^+/− heterozygous embryos, as in wild types, the expression of multiple Wnt genes (Wnt3a, Wnt2b, Wnt5a) is observed within the medial region of the cortical hem, (Fig. 4A,B) (Grove et al., 1998) In contrast, in Gli3^−/− homozygous embryos, the expression of Wnt3a, Wnt2b, and Wnt5b is undetectable at either of these ages (Fig. 4E,F and data not shown) (Grove et al., 1998). The only Wnt gene that persists in the cortex of Gli3^−/− homozygotes is Wnt8b (Fig. 4G), whose expression normally extends beyond the cortical hem (Fig. 4C). As expected, CR cells are detected throughout the pial surface in Gli3^+/− heterozygote control animals (Fig. 4D). Surprisingly, in the Gli3^−/− homozygous mutants, although the architecture of the dorsal telencephalon is perturbed because of the lack of the dorsal signaling center, reelin-positive CR cells were detected throughout the surface of the telencephalon (Fig. 4H). Hence, although their distribution is irregular because of the general perturbation of the cortex in these mutants, CR neurons were observed in these mutants (Fig. 4H). Because the Gli3 mutants lack the dorsomedial signaling center, we assessed whether these CR neurons represent a specific CR subtype. Indeed, this CR cell population did not express p73, suggesting that they do not possess cortical hem-derived CR character (Fig. 4j′–j‴). We therefore examined whether this population represents the PSB-derived subtype. In E10.5 Gli3 homozygous mice, the pallial–subpallial boundary as judged by Dbx1 expression appears to be normal (supplemental Fig. 4S, available at www.jneurosci.org as supplemental material), and, in these mice, the earliest reelin-positive neurons are detected in the domain flanking this region as observed in Gli3^+/− control animals (Fig. 4R,T). Furthermore, by E14.5, Dbx1 expression in the progenitor cells within the ventricular zone (Fig. 4X) and Ebf2 expression in early differentiating cells (Fig. 4Y) were specifically expanded in the dorsal telencephalon in the Gli3^−/− mutants, and upregulated reelin expression in postmitotic neurons indicated production of CR cells in this region (Fig. 4Z). These results suggest that downregulation of Foxg1 expression seen in Gli3^−/− mutants specifically results in the expansion of PSB-derived CR cells. Together, these results support the hypothesis that Foxg1 is required to suppress CR cells possessing a distinct regional character, whereas the acquisition of the cortical hem-derived subtype is solely dependent on the presence of dorsomedial signaling.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Supernumerary production of Cajal-Retzius cells in the absence of the cortical hem. A–G, Coronal sections showing the expression of Wnt3a, Wnt2b (cortical hem Wnts), and Wnt8b in E13.5 Gli3^+/− control embryos (A–C) and Gli3^−/− mutants (E–G). Expression of Wnt3a and Wnt2b expression is absent in the dorsal midline of Gli3^−/− mutants (E, F), whereas Wnt8b expression is restricted the medial most dorsal telencephalon (G). D and H show reelin expression, indicating the presence of CR cells in both control (D) and Gli3^−/− mutant mice (H). However, reelin-expressing CR cells are clustered rather than being evenly distributed throughout the cortex in Gli3^−/− mutants (shown at arrowheads). I, J, Expression of Reelin and p73 in control and Gli3^−/− cortex. i′–i‴ and j′-j‴ show an enlarged view of the boxed region indicated in I and J (pial surface is to the top). Many Reelin⁺ are detected in the dorsal pallium in Gli3^−/− mutants (J, j′), and most of these do not express p73 (j″, j‴). Scale bars: A–J, 200 μm; i′–j‴, 100 μm. Q–T, PSB is preserved in Gli3^−/− mutants, as indicated by the normal expression of Dbx1 (Q, S). CR cells are present in both wild-type and mutant embryos as indicated by reelin expression (R, T) in Gli3^+/− control (Q, R) and Gli3^−/− (S, T) E10.5 embryos. Dbx1 is expressed in the ventrolateral telencephalon (shown at asterisk), which is not perturbed in the Gli3^−/− mutants (S). The earliest reelin-expressing CR cells were detected in the ventral telencephalon at this stage in both control and Gli3^−/− mice. R and T show sections that are situated anterior to the Dbx1-expressing regions shown in Q and S. Insets show an enlarged view of reelin-expressing cells as indicated with arrowheads. U–Z, Expression of Dbx1, Ebf2, and reelin in E14.5 control (U–W) and Gli3^−/− (X–Z) mice. Arrowheads in U indicate the restricted expression of Dbx1 within the corticostriatal boundary in controls, whereas Dbx1 expression is expanded in Gli3^−/− nulls (X). Asterisk indicates the cluster of supernumerary CR cells in the Gli3^−/− cortices. Scale bars: Q–T, 100 μm; U–Z, 0.5 mm.

BMP signaling is not required for the production of hem-derived CR neurons

Our results demonstrated that dorsomedial signaling is not required for the production of CR cells in general, but the specification of cortical hem-derived CR cells appear to depend on signals emanating from the cortical hem. We therefore tested whether BMP signaling specifically might account for the loss of specification of cortical hem-derived CR cell subtype in Gli3 mutants, by ablating BMP signaling throughout the entire cortical progenitor domain.

In the telencephalon, two type I BMP receptors are known to be expressed, Bmpr1a and Bmpr1b (Hebert et al., 2002). Although Bmpr1a is the major type I BMP receptor expressed in the telencephalon, conditional removal of Bmpr1a in the telencephalon does not have a significant effect on the development of the cortical hem structure and does not abolish all BMP signaling (Hebert et al., 2002). Moreover, expression of the Bmpr1b gene, which is restricted to the dorsal neural tube, is unaffected in Bmpr1a conditional mutants (Hebert et al., 2002). Therefore, to ablate BMP signaling throughout all cortical progenitor cells, we removed both type I BMP receptors in the telencephalon by crossing the Foxg1^Cre/+;Bmpr1a^flox/+;Bmpr1b^−/+ allele to the Bmpr1a^flox/flox; Bmpr1b^−/+ mice (Fernandes et al., 2007). Our previous fate-mapping experiments indicate that, in these mice, loss of BmprIa/Ib receptors should be complete by E8.5 (Hebert and McConnell, 2000; Fernandes et al., 2007). Given that CR cells are produced >1 d later (Hevner et al., 2001), we believe that residual BMP signaling before E8.5 is unlikely to affect CR production. Embryos that lack both Bmpr1a and Bmpr1b did not survive beyond E11.5, and the mutants displayed a reduction in the size of the telencephalon. However, reelin-expressing CR cells were detected throughout the entire telencephalon (Fig. 5B,b′). To confirm that BMP signaling is abolished in these double mutants, we assessed the expression of a downstream effector of BMP signaling, Msx1. Consistent with BMP signaling being lost in the telencephalon of these mutants, we observed extinction of Msx1 expression in the cortex (data not shown). It remains possible that, as a result of Alk2 (activin receptor-like kinase 2) function (i.e., the sole remaining BMPRI receptor in these mutants), some diminished level of BMP signaling may be retained in the cortical hem of these mutants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Cajal-Retzius cells are produced in the absence of BMP signaling. A, B, Reelin expression in both control (A) and Foxg1^Cre/+; Bmpr1B^−/−; Bmpr1A^flox/flox (BmprIa/b dKO) mice (B) at E10.5 (coronal view). Embryos carrying the Foxg1^Cre/+;Bmpr1a^flox/+;Bmpr1b^n/+ genotype are referred to as controls. a′ and b′ show the enlarged view of the boxed region indicated in A and B, respectively. vz, Ventricular zone. Scale bars: A, B, 200 μm; a′, b′, 100 μm.

We further examined whether the specification of cortical hem-derived subtype of CR cells requires BMP signaling, by delineating the subtype of CR cells produced in the Bmpr1a/b compound null mice. Although at E10.5 there is only limited differentiation of CR neurons, both p73⁺ and p73⁻ CR cells were observed in these mutants (supplemental Fig. S6, available at www.jneurosci.org as supplemental material). This result implies that signals other than BMPs are responsible for the specification of the cortical hem-derived CR subtype. Together, these data demonstrate that BMP signaling subsequent to E8.5 is not essential for the development of CR cells in general, further supporting our conclusion that the dorsomedial signaling is itself dispensable for the production of CR cells.