The transcription factor Foxg1 regulates the competence of telencephalic cells to adopt subpallial fates in mice

Two major components of the mammalian forebrain, the cerebral cortex and the basal ganglia, develop from the dorsal and ventral parts of the embryonic telencephalon. The telencephalon arises at the rostral end of the neural tube and its dorsal and ventral domains are patterned by the activities of a number of secreted molecules produced by surrounding signalling centres (Danesin et al., 2009; Hebert, 2005; Hebert and Fishell, 2008; Rallu et al., 2002a; Sur and Rubenstein, 2005; Wilson and Houart, 2004). Sonic hedgehog (Shh) is produced ventrally, fibroblast growth factor 8 (Fgf8) is produced rostrally and a number of bone morphogenetic proteins (Bmps) and Wnt proteins are produced caudo-medially. These signalling molecules activate the graded expression of transcription factors that control many aspects of the subsequent development of telencephalic cells, including their molecular and cellular identities.

The transcription factor Foxg1 is essential for normal telencephalic development. Foxg1^−/− mutant mice die perinatally with a severely hypoplastic telencephalon comprising dorsal, but not ventral, tissue (Martynoga et al., 2005; Xuan et al., 1995). Previous studies have focused mainly on defects in the residual dorsal telencephalon, revealing defects of progenitor proliferation, cell differentiation, progenitor patterning, signalling molecule expression and specification of neuronal subtypes (Ahlgren et al., 2003; Bourguignon et al., 1998; Dou et al., 1999; Hanashima et al., 2007; Hanashima et al., 2004; Hanashima et al., 2002; Hardcastle and Papalopulu, 2000; Hebert and Fishell, 2008; Muzio and Mallamaci, 2005; Xuan et al., 1995). Examining possible causes of the failure of ventral telencephalic development is much harder because ventral telencephalon is not detectable even at very early ages (Martynoga et al., 2005; Xuan et al., 1995).

Foxg1 is one of the earliest transcription factors to be expressed in the part of the neural plate from which telencephalon develops. One hypothesis proposed to explain the requirement of Foxg1 in ventral telencephalic development is that Foxg1 is needed for normal production of Fgf8 (Hebert and Fishell, 2008). This suggestion is based on research showing that Fgf signalling is required for normal ventral telencephalic development (Gutin et al., 2006; Storm et al., 2006) coupled with the observation that rostral telencephalic Fgf8 expression is reduced in Foxg1^−/− embryos (Martynoga et al., 2005). In addition, Shh expression, an important player in ventral telencephalic development (Fuccillo et al., 2006a; Fuccillo et al., 2004), is reduced in Foxg1^−/− telencephalon (Huh et al., 1999). Although it is reasonable to suppose that reduced levels of telencephalic Fgf8 and Shh would selectively impair ventral telencephalic development, the extent to which they account for the phenotype of mouse embryos lacking Foxg1 is unknown. The likelihood that Foxg1 has other actions essential for normal mammalian ventral telencephalic development is strongly supported by a recent study in zebrafish indicating that Foxg1 is required downstream of Hedgehog activity for the acquisition of ventral telencephalic identity (Danesin et al., 2009).

We began by examining the ability of telencephalic cells from Foxg1^−/− mouse embryos to respond to Shh and Fgf8. We studied their expression of genes known to require Shh or Fgf8 for their activation and tested the responses of cultured Foxg1^−/− telencephalic cells to Shh and Fgf8. We found that the expression of many genes that are targets of Shh and Fgf8 signalling could be induced in Foxg1^−/− cells but we could not induce the expression of other ventral telencephalic marker genes. We explored the development of Foxg1^−/− cells in Foxg1^−/− Foxg1^+/+ chimeric embryos in which the wild-type cells were patterned normally. We found that Foxg1^−/− cells contributed to the chimeric ventral telencephalon, but the mutant cells were specified abnormally, expressing dorsal rather than ventral telencephalic markers. These findings indicate that, in addition to regulating production of ventralising cues, Foxg1 has a crucial cell-autonomous function, giving telencephalic cells the competence to adopt ventral identities.

Foxg1^−/− embryos were generated by intercrossing Foxg1^+/− mice (Martynoga et al., 2005). Foxg1^−/−;Gli3^+/− embryos were made by crossing Foxg1^+/− and Foxg1^+/−;Gli3^+/− mice, whereas Foxg1^−/−;Pax6^−/− embryos were made by intercrossing Foxg1^+/−;Pax6^+/− mice.

Chimeras were produced by morula aggregation. Eight-cell embryos were obtained by crossing Foxg1^lacZ/+ females with Foxg1^+/Cre males. Embryos obtained from this cross were: Foxg1^+/+, Foxg1^lacZ/+, Foxg1^+/Cre and Foxg1^lacZ/Cre (Foxg1^−/−). Foxg1^+/+ embryos for aggregation were from (BALB/c × A/J) F₂ intercrosses. Embryos were collected from superovulated females and aggregated (West and Flockhart, 1994). The two morulae for aggregation differed at the Gpi1 locus (encoding glucose phosphate isomerase). Aggregated embryos were cultured overnight, transferred to recipient pseudopregnant females and recovered at embryonic day 12.5 (E12.5). The genotype of each chimera was determined by PCR (Martynoga et al., 2005). The use of two null-mutant Foxg1 alleles allowed distinction between chimeras containing Foxg1^lacZ/Cre (Foxg1^−/−) compound heterozygous cells and those containing heterozygous (Foxg1^lacZ/+ or Foxg1^+/Cre) or wild-type (Foxg1^+/+) cells. The global contribution of each morula to the chimera was estimated using Gpi1 electrophoresis (West and Flockhart, 1994).

E10.5 embryos were dissected in ice-cold Earle's balanced saline solution (EBSS). Foxg1^−/− embryos were distinguished by their eye and telencephalic defects. Explants were dissected and cultured on Falcon inserts (0.4 μm pore size, Becton Dickinson) in wells containing defined culture medium with 10% foetal bovine serum. For dose-response experiments, a minimum of three explants were pooled and cultured with 0, 5, 20 or 50 nM Shh (recombinant mouse Shh-N, R&D Systems) or with 0, 0.42 or 4.2 nM Fgf8 (recombinant mouse Fgf8b, R&D Systems). After 24 hours, explants were harvested and RNA was extracted for qRT-PCR analysis.

For induction with Shh-coated beads, 1 hour after placement on an insert, 3-6 Affi-gel blue beads (100-200 mesh, Bio-rad) soaked either in 25 μM Shh-N or 0.1% BSA were placed on the explants, which were then cultured for 24-36 hours. Explants were fixed in 4% paraformaldehyde.

Antisense RNA probes for Ptc1, Fgfr1, Fgfr3, Sprouty2 (Spry2 — Mouse Genome Informatics), Mest and Etv5 were digoxigenin-labelled. In situ hybridisations were on 10 μm paraffin sections (Christoffels et al., 2000).

Immunohistochemistry was carried out as previously described (Martynoga et al., 2005). The primary antibodies used were: anti-Pax6 [Developmental Studies Hybridoma Bank (DSHB)], anti-Ngn2 (Lo et al., 2002), anti-beta-galactosidase (β-gal) (Molecular Probes), anti-Gsh2 (Toresson et al., 2000), anti-Islet1 (DSHB), anti-Mash1 (BD Bioscience), anti-Nkx2.1 (Abcam), anti-calbindin (Sigma), anti-Olig2 (Ligon et al., 2004), anti-Tbr1 and anti-Tbr2 (Englund et al., 2005), anti-pan-Dlx (Hevner et al., 2004), anti-cyclin D1 (Novocastra) and anti-p21^cip1 (BD Pharmingen).

Western blot analysis of Gli3 expression in the telencephalon of E12.5 control, Foxg1^−/− and Gli3^−/− embryos was performed as in Fotaki et al. (Fotaki et al., 2006).

RNA was extracted from embryonic tissues (RNEasy Mini Kit, Qiagen). cDNA was synthesised from the RNA template using M-MLV (Promega) or Sensiscript (Qiagen) reverse transcriptase and random hexanucleotide primers. PCR on the cDNA template used gene-specific primers (see Table S1 in the supplementary material). ‘No RT’ and ‘no cDNA’ controls were included where appropriate and in all cases were negative.

For qRT-PCR, Quantitect SYBR Green PCR reagents (Qiagen) were used on an Opticon single wavelength PCR machine (MJ Research). All primers spanned introns. Expression levels of genes of interest were normalised to the expression of β-actin or Gapdh. Following normalisation, expression levels for each gene were expressed as a ratio to normalised levels of a control group (wild-type telencephalon for expression analyses or untreated samples for Shh and Fgf8 induction analyses). For expression analyses, each qRT-PCR reaction was run in triplicate on cDNA from at least three independent RNA extracts. For morphogen dose-response analyses, each qRT-PCR was performed on a minimum of three samples from independent cultures (11-15 control and 6-8 Foxg1^−/− samples for Shh induction analysis; four control and three Foxg1^−/− samples for Fgf8 induction analysis)

In normal mouse embryogenesis, Shh signalling through the Hh/Gli pathway is essential for the development of the ventral telencephalon: ventral telencephalic cell types are lost in embryos lacking Shh or its obligate cell surface transducer smoothened (Smo) (Chiang et al., 1996; Corbin et al., 2003; Ericson et al., 1995; Fuccillo et al., 2004; Fuccillo et al., 2006b; Hebert and Fishell, 2008; Ohkubo et al., 2002; Rallu et al., 2002b; Zaki et al., 2005). In Foxg1^−/− embryos, telencephalic exposure to Shh is likely to be abnormally low (Huh et al., 1999), but the impact of this and the extent to which Foxg1^−/− telencephalic cells can respond to Shh are unknown. We tested whether the Shh receptors Ptc1 and Smo are still expressed in Foxg1^−/− telencephalon. Fig. 1A shows that Ptc1 and Smo are expressed in Foxg1^−/− telencephalon; in situ hybridisation showed that Ptc1 is expressed in the ventral half of both the wild-type and Foxg1^−/− telencephalon (Fig. 1D,E). As Ptc1 expression is dependent on Hh signalling (Goodrich et al., 1996; Marigo and Tabin, 1996; Wijgerde et al., 2002), this finding indicates that at least some ventral Hh signalling is retained in Foxg1^−/− embryos, although the level of expression of Ptc1 determined by qRT-PCR is lower than in wild-type ventral telencephalon (Fig. 1B). Gli1, whose expression is also activated by and requires Hh signalling (Bai et al., 2002; Bai et al., 2004; Lee et al., 1997), is expressed in Foxg1^−/− mutant telencephalon (Fig. 1A). The Hh target Nkx2.1 (Ruiz i Altaba, 1998; Vokes et al., 2007) is absent in Foxg1^−/− telencephalon (Fig. 1A). These results indicate that many components of the Hh signalling pathway are active in the Foxg1^−/− telencephalon, albeit with evidence of impaired activation, whereas others such as Nkx2.1 are lost.

As the telencephalon is likely to be exposed to reduced levels of Shh (Huh et al., 1999), which might explain the reduced expression of Ptc1 in the Foxg1^−/− telencephalon (Fig. 1B), we predicted alterations in the expression and processing of the transcription factor Gli3. A primary action of Hh signalling in patterning the ventral neural tube involves antagonism of the ability of Gli3 to repress ventral gene expression. This is achieved in two ways. First, exposure to Shh can decrease transcription of Gli3 (Marigo et al., 1996; Ruiz i Altaba, 1998; Schweitzer et al., 2000). Second, Shh inhibits the post-translational cleavage of the Gli3 protein (Bastida et al., 2004; Wang et al., 2000): the cleaved form (Gli3-CL) is thought to act as a repressor required for development of dorsal telencephalon (Jacob and Briscoe, 2003; Quinn et al., 2009; Rash and Grove, 2007; Wang et al., 2000). Results from wild-type embryos in Fig. 1C,F confirmed lower levels of Gli3 transcript and Gli3 protein in ventral than in dorsal embryonic telencephalon, supporting previous data (Fotaki et al., 2006; Nery et al., 2001; Sussel et al., 1999), and the abundance of Gli3-CL relative to that of full-length Gli3 (Gli3-FL) was lower in ventral telencephalon (average ratio=1.33:1, n=8) than in dorsal telencephalon (average ratio=2.75:1, n=7). The level of Gli3 mRNA in Foxg1^−/− telencephalon was similar to that in the dorsal part of wild-type telencephalon (Fig. 1C) with an overabundance of Gli3-CL relative to Gli3-FL (average ratio=11.04:1, n=4) (Fig. 1F). We went on to test the significance of this for the Foxg1^−/− telencephalon.

Previous work has shown that the removal of just one copy of Gli3 in Shh^−/− and Smo^−/− embryos is sufficient for a high level of recovery of ventral telencephalic morphology and gene expression. The recovery includes restoration of ventral domains of Mash1, Gsh2 and Nkx2.1 expression; these domains are cleared of abnormally high levels of Pax6 expression, which becomes restricted to a dorsal domain as in wild-type embryos (Rallu et al., 2002b). Given that the post-translational processing of Gli3 in Foxg1^−/− telencephalon showed abnormalities that are associated with loss of Shh signalling, we tested whether loss of one copy of Gli3 restored any aspects of ventral telencephalic development in Foxg1^−/− embryos. We observed no obvious restoration of ventral telencephalic morphology and little evidence of rescue of ventral patterns of gene expression in Foxg1^−/−;Gli3^+/− compared with Foxg1^−/−;Gli3^+/+ embryos: there was weak re-expression of Mash1 and a very few Islet1-expressing cells were found in the ventral part of the Foxg1^−/−;Gli3^+/− telencephalon (see Fig. S1 in the supplementary material). As loss of both copies of Gli3 in Foxg1^−/− embryos leads to loss of the telencephalon (Hanashima et al., 2007; Hebert and Fishell, 2008), Foxg1^−/−;Gli3^−/− embryos were unlikely to be informative in the present context. Our findings suggest that defective antagonism of Gli3-CL is unlikely to account for most of the ventral defects in Foxg1^−/− telencephalon.

Impaired Shh signalling in Foxg1^−/− embryonic telencephalon might occur because cells are exposed to reduced concentrations of Shh from ventral sources and/or because cells have an abnormal response to Shh. To test the second of these possibilities, we cultured telencephalic explants from E10.5 Foxg1^−/− embryos with 5, 20 or 50 nM Shh in solution. Explants were harvested after 24 hours in culture and analysed by qRT-PCR (Fig. 2). We used explants from the dorsal telencephalon of wild-type embryos as controls (Fig. 2A,B): Shh application resulted in upregulation of Ptc1, Gli1 and Olig2 [a gene expressed in ventral telencephalic progenitors that is known to be induced by Shh (Lu et al., 2000)], caused downregulation of Gli3 and had no effect on expression of Mash1 [a proneural gene normally expressed at relatively low levels in dorsal telencephalon compared with high levels in ventral telencephalon (Fode et al., 2000)]. In Foxg1^−/− telencephalic explants, Shh application upregulated Ptc1, Gli1 and Olig2 and downregulated Gli3, as in wild-type explants, but significantly downregulated Mash1 (Fig. 2C,D). Using this experimental paradigm we were unable to induce consistent, statistically significant upregulation of the ventral telencephalic marker Dlx2 in either wild-type or Foxg1^−/− explants (data not shown).

We did succeed in inducing Dlx2 and Mash1 expression in wild-type explants when Shh was delivered for 24-36 hours from beads soaked in a 25 μM solution of recombinant Shh (Fig. 3A,D; 17/18 explants showed induction), as has been demonstrated previously (Kohtz et al., 1998; Kuschel et al., 2003). Induction was seen around the beads (Fig. 3D), suggesting that successful induction of these genes in wild-types requires a localised source of Shh delivering a particular concentration or perhaps a concentration gradient. Delivery of Shh in this way failed to induce Mash1 and Dlx expression in Foxg1^−/− explants (Fig. 3E; 0/16 explants showed induction). In control experiments, no Dlx- and Mash1-positive cells were found in wild-type or Foxg1^−/− explants treated with BSA-soaked beads (Fig. 3B,C; 0/34 explants showed expression). We used qRT-PCR to confirm that, as in experiments with Shh added in solution, Ptc1 and Gli1 were induced in Foxg1^−/− explants exposed to Shh-soaked beads (Fig. 3F; Student's one sample t-test, P<0.05, n=4 independent experiments in each case). CyclinD1, which is also a target of Shh signalling in neural progenitors (Kenney and Rowitch, 2000; Oliver et al., 2003), was significantly upregulated by Shh-soaked beads in Foxg1^−/− explants (Fig. 3F).

Overall, these in vitro experiments suggested that loss of Foxg1 results in a selective impairment in the response of telencephalic cells to Shh: whereas many Shh targets are upregulated (Ptc1, Gli1, Olig2, CyclinD1) or downregulated (Gli3) as in wild-types, we were unable to upregulate expression of the ventral telencephalic markers Mash1 and Dlx2 in Foxg1^−/− explants using a paradigm that did upregulate these markers in wild-type explants. One experiment suggested that Shh might even suppress Mash1 expression under some conditions when Foxg1 is absent.

We considered the possibility that persistent expression of Pax6 throughout the Foxg1^−/− telencephalon might mediate suppression of ventral telencephalic marker expression, as Pax6 is known to suppress ventral marker expression in normal dorsal telencephalon (Kroll and O'Leary, 2005; Quinn et al., 2007; Stoykova et al., 2000; Toresson et al., 2000; Yun et al., 2001). We generated Foxg1^−/−;Pax6^−/− embryos in which the coding sequence of Foxg1 is replaced by a lacZ reporter cassette (Xuan et al., 1995). The Foxg1-expressing telencephalon labelled by β-galactosidase immunostaining was smaller in Pax6^−/−;Foxg1^−/− than in Foxg1^−/− mutants (Fig. 4), consistent with the known role of Pax6 in the regulation of cortical progenitor proliferation (Estivill-Torrus et al., 2002; Quinn et al., 2007). We found stronger Mash1 staining in the Pax6^−/−;Foxg1^−/− telencephalon compared with Foxg1^−/− telencephalon and Gsh2 and Olig2 expression was restored in a small number of cells, well below the numbers of Gsh2 and Olig2 expressing cells in wild-type ventral telencephalon (Fig. 4E-O). There was no rescue of Islet1, Nkx2.1 or Dlx2 in Pax6^−/−;Foxg1^−/− telencephalon (data not shown). The dorsal markers Tbr2 (Eomes — Mouse Genome Informatics) and reelin were expressed throughout the telencephalon in both Pax6^−/−;Foxg1^−/− and Foxg1^−/− mutants (Fig. 4P-R). These findings indicate that some defects of ventral telencephalic gene expression in Foxg1^−/− embryos, most notably the loss of Mash1 expression, might be mediated in part by persistent ventral telencephalic expression of Pax6.

Fgf proteins act downstream of Shh and Gli3 to generate ventral telencephalic cell types (Gutin et al., 2006) and Fgf8 has a prominent role in this process (Storm et al., 2006). Simultaneous deletion of Fgf receptors 1 and 2 (Fgfr1 and Fgfr2) results in a loss of ventral cell types even though Shh remains expressed and, unlike the Shh^−/− and Smo^−/− phenotypes, the Fgfr1^−/−;Fgfr2^−/− phenotype is not rescued by removal of Gli3 (Gutin et al., 2006). We tested whether defects in Fgf responsiveness might contribute to the loss of ventral cell types in Foxg1^−/− telencephalon. We first examined the expression of Fgf8 signalling pathway components (Fig. 5). We examined the expression of Fgfr1, which is thought to be the main Fgf8 receptor in the early telencephalon (Gutin et al., 2006), and that of Fgfr3 as it is the highest affinity receptor for Fgf8 (Chellaiah et al., 1999; Ornitz et al., 1996), although its function in the early telencephalon is unclear (Gutin et al., 2006); we found that Fgfr3 (Fig. 5A,B) and Fgfr1(Fig. 5G,H) are expressed throughout both wild-type and Foxg1^−/− telencephalon, with higher levels expressed dorsally for Fgfr3 (Fig. 5A,B).

We examined the expression of known targets of Fgf8 signalling. The sprouty 2 gene (Spry2) is induced by Fgf8 signalling (Chambers and Mason, 2000) and is expressed in the early telencephalon (Zhang et al., 2001), where its expression requires Fgf8 signalling (Storm et al., 2003). We found that Sprouty2 is expressed throughout the telencephalon of Foxg1^−/− embryos as in wild-types (Fig. 5I,J). Mest (Peg1) encodes a highly conserved enzyme of unknown function whose telencephalic expression is induced by and depends on Fgf8 signalling (Sansom et al., 2005). In wild-type telencephalon, Mest is expressed in the dorsal lateral ganglionic eminence (dLGE); in Foxg1^−/− mutants, Mest expression is retained in the ventral-most part of the telencephalon (Fig. 5E,F, arrows). The transcription factor Etv5 is another target of Fgf8 signalling whose telencephalic expression is strongly down-regulated in Fgfr1^−/− embryos (Sansom et al., 2005). In wild-type telencephalon, Etv5 is expressed in the dLGE; in Foxg1^−/− embryos, Etv5 expression is retained in the ventral-most part of the telencephalon (Fig. 5C,D, arrows). The retention of Spry2, Etv5 and Mest expression domains in the ventral part of the Foxg1^−/− telencephalon suggests that the Fgf8 signalling pathway is active in this region.

To test whether the Fgf8 pathway can be activated in Foxg1^−/− cells we cultured telencephalic explants from Foxg1^−/− embryos with 0.42 or 4.2 nM Fgf8 in solution. The explants were harvested after 24 hours in culture and analysed by qRT-PCR (Fig. 6). Explants from the dorsal telencephalon of wild-type embryos were used as controls. Exposure to 4.2 nM Fgf8 resulted in upregulation of Etv5 and Spry2 mRNA levels in control and in Foxg1^−/− explants (Fig. 6). These findings, together with our in vivo observations on expression of Fgf receptors and targets, indicate that Foxg1^−/− telencephalic cells are able to respond to Fgf8.

Our results so far indicated that Foxg1^−/− telencephalic cells might have a profound cell autonomous inability to adopt ventral fates, even though they are exposed to and show some responses to ventralising signals in Foxg1^−/− embryos. We pursued this possibility further using an in vivo approach in which Foxg1^−/− cells and wild-type cells in an individual embryo are exposed to the same ventralising signals. To do this, we generated chimeric embryos comprising Foxg1^−/− and wild-type cells.

We confirmed that Foxg1^−/− cells can contribute to both dorsal and ventral telencephalon in chimeras (see Fig. S2 in the supplementary material). In Foxg1^−/− Foxg1^+/+ chimeras, the telencephalon developed its characteristic morphology, including distinct ganglionic eminences ventrally. Many Foxg1^−/− cells (recognised by their expression of β-galactosidase) segregated into clusters at all rostrocaudal levels in both dorsal and ventral telencephalon (see Fig. S2A-C in the supplementary material), whereas in Foxg1^+/− Foxg1^+/+ control chimeras the two cell types were intermingled throughout the telencephalon even in cases where one cell type dominated (see Fig. S2D in the supplementary material). The segregation in experimental chimeras suggested a difference in the identities of the two cell types.

For each chimera, the overall percentages of cells derived from the Foxg1^−/− embryo (in experimental chimeras) or the Foxg1^+/− embryo (in control chimeras) were estimated by quantitating the Gpi1 isozyme composition of the upper body (West and Flockhart, 1994), where Foxg1 is not expressed and its absence would be expected to have little or no effect on the contribution of either cell type. The percentage of the Gpi1b isozyme represented the contribution of the Foxg1^−/− or Foxg1^+/− cells. Both Foxg1^−/− and Foxg1^+/− cells carried one mutant allele in which the coding sequence of Foxg1 is replaced by a lacZ reporter cassette (Foxg1^lacZ; Xuan et al., 1995), allowing us to detect and count proportions of Foxg1^−/− and Foxg1^+/− cells in the telencephalon of chimeras. In control chimeras, the proportion of β-galactosidase-expressing telencephalic cells divided by the percentage of Gpi1b was, as expected, ~1 (see Fig. S3 in the supplementary material). In experimental chimeras, however, proportions of β-galactosidase-expressing Foxg1^−/− telencephalic cells were about a third of the percentages of Gpi1b (see Fig. S3 in the supplementary material), indicating that although Foxg1^−/− cells contribute to both ventral and dorsal telencephalon, they are significantly under-represented to similar extents in both.

We examined the expression of the transcription factors Olig2, Gsh2 and Mash1, ventral progenitor markers expressed in both LGE and medial ganglionic eminence (MGE), and Nkx2.1, whose expression is restricted to the MGE (Fig. 7A). Foxg1^−/− cells in chimeric telencephalon consistently failed to express any of these ventral progenitor markers, whereas surrounding wild-type cells in experimental chimeras and wild-type and Foxg1^+/− cells in control chimeras expressed them appropriately (Fig. 7B-M). We examined the expression of two markers of differentiating neurons derived from the ventral telencephalon, Islet1 and calbindin. Islet1 marks differentiating striatal projection neurons and cholinergic interneurons (Stenman et al., 2003; Wang and Liu, 2001), and calbindin marks GABAergic interneurons (Anderson et al., 1997). The vast majority of Islet1-expressing telencephalic cells in Foxg1^−/− Foxg1^+/+ experimental chimeras were wild-type (negative for β-galactosidase) (Fig. 7N-P,R-T), but a very small number of Islet1-positive Foxg1^−/− cells were seen in the MGE of mid-level telencephalic sections (Fig. 7T, arrows). Large numbers of Foxg1^+/− cells in Foxg1^+/− Foxg1^+/+ control chimeras expressed Islet1 (Fig. 7Q,U, arrows). No calbindin-positive Foxg1^−/− interneurons were ever observed (Fig. 7V-X), whereas many wild-type and Foxg1^+/− ventral telencephalic cells in experimental and control chimeras expressed calbindin (Fig. 7W-Y, arrows). The near-complete failure of Foxg1^−/− cells to express a wide range of ventral markers expressed by their surrounding wild-type cells in the chimeric telencephalon indicates a cell-autonomous requirement for Foxg1 in the acquisition of ventral telencephalic identities.

As Foxg1^−/− cells in chimeric ventral telencephalon failed to adopt ventral identities, we examined the expression of Ngn2, Pax6, Tbr1 and Tbr2 to see whether they had adopted dorsal identities instead. In wild-type embryos, dorsal telencephalic progenitors express Ngn2 and high levels of Pax6, Tbr2 is found in basal progenitors and early-born neurons of the dorsal telencephalon and Tbr1 is expressed by dorsally derived glutamatergic neurons (Englund et al., 2005; Hevner et al., 2003; Hevner et al., 2001) (Fig. 8A). Foxg1^−/− cells located in the dorsal telencephalon of experimental chimeras expressed all of these markers, as did their wild-type neighbours, consistent with their expression in the telencephalon of Foxg1^−/− embryos (data not shown). Foxg1^−/− cells in the LGE of experimental chimeras expressed Ngn2, Pax6, Tbr2 and Tbr1 (Fig. 8B-F,L,M,O,P). Foxg1^−/− cells located in the MGE of experimental chimeras expressed Pax6 (Fig. 8H-J), but did not express Ngn2 (Fig. 8B), Tbr1 or Tbr2 (data not shown). Ectopic expression of dorsal markers in the ventral telencephalon was seen only in Foxg1^−/− cells, indicating that the Foxg1^−/− cells did not alter the specification of surrounding wild-type cells in a cell-nonautonomous manner. These findings suggest that Foxg1 is required cell-autonomously for ventral telencephalic suppression of Pax6 expression and for suppression of Tbr1 and 2 and Ngn2 expression in the LGE, although not in the MGE.

We also examined expression of cyclin D1 and p21^cip1 in chimeras. Cyclin D1, an early target of Shh signalling in neural progenitors (Kenney and Rowitch, 2000; Oliver et al., 2003), is expressed in a ventral^high to dorsal^low gradient in wild-type telencephalon but was not detectable by immunohistochemistry in Foxg1^−/− telencephalon (see Fig. S4A,B in the supplementary material). In Foxg1^−/− Foxg1^+/+ chimeras, many Foxg1^−/− ventral telencephalic cells expressed cyclin D1, as did surrounding wild-type cells (see Fig. S4C-G in the supplementary material). In normal embryonic dorsal telencephalon, p21^cip1 is expressed in the mantle layer of the medial wall only, whereas in Foxg1^−/− embryos p21^cip1 is expressed more widely throughout the mantle layer of the dorsal telencephalon (Seoane et al., 2004). We found that p21^cip1 is not upregulated in Foxg1^−/− telencephalic cells in chimeras, but remains restricted to cells in the mantle layer of the medial wall only, as in wild-types (see Fig. S4H-K in the supplementary material).

Foxg1 is required for normal embryonic development of the ventral telencephalon (Martynoga et al., 2005; Xuan et al., 1995). Previous work showed that embryos lacking Foxg1 have reduced telencephalic expression of two signalling molecules crucial for the specification of ventral telencephalic cell types, Fgf8 and Shh, and suggested that lack of these molecules might account for the lack of ventral telencephalon in Foxg1^−/− embryos (Huh et al., 1999; Martynoga et al., 2005; reviewed by Hebert and Fishell, 2008). We have now examined the competence of telencephalic cells lacking Foxg1 to develop ventral cell types. Our new data indicate that telencephalic cells lacking Foxg1 have a cell-autonomous inability to express many ventral marker genes and to repress many dorsal marker genes even though they do show some responses to Shh and Fgf8. These findings imply that the profound loss of ventral cell types in the Foxg1^−/−telencephalon is due not only to a reduced production of ventralising signals but also to an intrinsic inability of telencephalic cells to express normal ventral molecular phenotypes, as a result of defects downstream of many of the cell responses to Shh and Fgf8. Our findings strongly support recent findings in zebrafish that Foxg1 is required downstream of Shh for the development of ventral telencephalic identities (Danesin et al., 2009).

We obtained in vitro and in vivo evidence that telencephalic cells are able to respond to Shh and Fgf8 and that they do so in Foxg1^−/− embryos. First, Foxg1^−/− telencephalic cells express key receptors for both Shh and Fgf8. Second, Foxg1^−/− telencephalic cells express genes whose activation requires Shh and Fgf8 signalling. Third, exposure of Foxg1^−/− telencephalic explants to Shh or Fgf8 results in induction of known Shh and Fgf8 target genes, including Ptc1, Gli1, Olig2, cyclin D1, Spry2 and Etv5.

Previous work showed that the telencephalic actions of Shh are mediated by Gli3: Shh inhibits the cleavage of Gli3 into its repressor form (Gli3-CL) and Gli3-CL inhibits the expression of Shh targets (reviewed by Fuccillo et al., 2006; Hebert and Fishell, 2008). We found an overabundance of Gli3-CL in the Foxg1^−/− telencephalon. The most probable explanation for this is the reduced exposure of the telencephalon to Shh in Foxg1^−/− embryos (Huh et al., 1999). In wild-type embryos, a site of Shh expression develops in the ventral telencephalon by E10.5; this does not form in Foxg1^−/− embryos, although expression of Shh appears normal at other sites around the telencephalon (i.e. diencephalon, floor plate, notochord and anterior mesendoderm).

Previous work showed a reduction in the expression of Fgf8 in its rostro-ventral telencephalic domain in Foxg1^−/− embryos aged E9.5 and older (Martynoga et al., 2005). Fgf signalling is essential for the generation of ventral telencephalic cell types and there is evidence that it patterns the ventral telencephalon in a dose-dependent manner (Gutin et al., 2006; Shanmugalingam et al., 2000; Shinya et al., 2001; Storm et al., 2006; Walshe and Mason, 2003). Our new results indicate that in Foxg1^−/− embryos, residual Fgf8 expression is sufficient to activate ventrally-expressed Fgf8 target genes Spry2, Mest and Etv5 that depend on Fgf8 for their expression (Sansom et al., 2005; Storm et al., 2003). Overall, these findings indicate that the levels of Fgf8 and Shh available to telencephalic cells in Foxg1^−/− embryos are sufficient to activate many of their target genes.

We tested the possibility that elevated levels of Gli3-CL in Foxg1^−/− embryos block the expression of ventral telencephalic marker genes by genetically deleting one allele of Gli3 in Foxg1^−/− embryos. Unlike in Shh^−/− and Smo^−/− telencephalons, where loss of one copy of Gli3 leads to widespread reactivation of ventral gene expression (Rallu et al., 2002b), Foxg1^−/−;Gli3^+/− embryos showed no restoration of Nkx2.1, Olig2 and Gsh2, no loss of ectopic Pax6 and Lhx2/9 in ventral telencephalon and only a minor restoration of expression of Mash1 and Islet1. This suggests that loss of Shh signalling is insufficient to explain the majority of the ventral telencephalic defects in Foxg1^−/− embryos.

Because levels of not just Shh, but also Fgf8, and probably other signalling molecules, are altered in Foxg1^−/− embryos, we decided that the best way to attempt to rescue ventral telencephalic gene expression in Foxg1^−/− cells was to embed mutant cells in a wild-type environment capable of delivering combinations and levels of signals sufficient for normal dorso-ventral patterning. We generated Foxg1^−/− Foxg1^+/+ chimeras in which wild-type cells showed normal dorsoventral telencephalic patterning. Foxg1^−/− cells were present in the ventral telencephalon, but in lower proportions than in Foxg1-nonexpressing regions of the embryo. This telencephalon-specific under-representation suggests that the proliferative defects in Foxg1^−/− embryos reported previously (Martynoga et al., 2005; Xuan et al., 1995) have a significant cell-autonomous component. Foxg1^−/− cells located in the ventral telencephalon of chimeras did not express the ventral markers Nkx2.1, Gsh2, Olig2, Mash1 and calbindin, showed expression of Islet1 in only a very few medially-located cases, did not lose their expression of Pax6 throughout the LGE and MGE and did not lose their expression of Tbr1, Tbr2 and Ngn2 in the LGE. They did, however, regain expression of cyclin D1, an early target of Shh signalling, and did not show ectopic expression of p21^cip1. Because the delivery of signals sufficient to pattern surrounding wild-type cells does not restore the normal ventral telencephalic phenotype to Foxg1^−/− cells, we conclude that they have a cell-autonomous selective block to the expression of ventral telencephalic marker genes.

We examined the possibility that unchecked ventral telencephalic expression of Pax6 resulting from the absence of Foxg1 might prevent Foxg1^−/− ventral telencephalic cells from expressing ventral telencephalic genes. It is known that Pax6 normally acts cell-autonomously to repress ventral telencephalic gene expression in the dorsal telencephalon and that loss of Pax6 leads to a loss of dorsal telencephalic phenotypes at the expense of ventral telencephalic phenotypes (Kroll and O'Leary, 2005; Quinn et al., 2007; Stoykova et al., 2000; Toresson et al., 2000; Yun et al., 2001). We hypothesised that loss of Pax6 might reverse the effects of loss of Foxg1. We found that it had remarkably limited effect: it caused no ventral re-expression of Nkx2.1, Dlx2 and Islet1, only a slight re-expression of Gsh2 and Olig2 and no ventral loss of ectopic Tbr2 expression. It did, however, cause significant re-expression of Mash1. Previous work showed that Gsh2, Dlx2 and Mash1, but not Nkx2.1, are expressed in Pax6^−/− cells in the dorsal telencephalon of Pax6^−/− Pax6^+/+ chimeras and that Tbr2 expression is lost in Pax6^−/− cells in the dorsal telencephalon of Pax6^−/− Pax6^+/+ chimeras (Quinn et al., 2007). These results, together with our present findings, indicate that loss of telencephalic Mash1 expression in the absence of Foxg1 might be caused by ectopic ventral Pax6 expression. The abnormalities of other ventral telencephalic marker gene expression in Foxg1^−/− telencephalon (including loss of expression of genes such as Gsh2 and gain of expression of genes such as Tbr2) are not rescued by the loss of Pax6 from Foxg1^−/− embryos. These experiments also indicate that the ectopic dorsal upregulation of ventral genes and loss of Tbr2 observed in Pax6^−/− telencephalon (Kroll and O'Leary, 2005; Quinn et al., 2007) require Foxg1 activity.

Previous work showed that in Foxg1^−/− mouse embryos, expression of telencephalic Bmp4 and of a gene activated by Bmp4, p21^cip1, expands ventrally from the normal dorsally restricted domains (Dou et al., 1999; Hanashima et al., 2002; Seoane et al., 2004). Our finding that p21^cip1 is not upregulated in Foxg1^−/− telencephalic cells in chimeras is interesting as it suggests that the Smad3 intracellular signalling pathway is not constitutively active in these cells. This possibility was hypothesised because, in various cell lines, Foxg1 can bind to the Smad3-Smad4-FoxO complex to inhibit its transcriptional activity, including its activation of p21^cip1 (Seoane et al., 2004). This suggests that the expanded domain of p21^cip1 expression in Foxg1^−/− embryos might require higher levels of activation from overexpressed Tgfβ family members than is achieved in the chimeric telencephalon, where many cells are wild-type. Exploration of interspecies differences in the telencephalic actions of Foxg1 is likely to be fruitful as zebrafish embryos do not show telencephalic upregulation of Bmp expression in response to depletion of Foxg1 (Danesin et al., 2009).

We thank K. Campbell, D. Anderson, D. Rowitch, J. Kohtz, Y. Sasai, R. Hevner and the Developmental Studies Hybridoma Bank for provision of antibodies; J. Hebert, S. McConnell, E. Lai and L. Shen for provision of transgenic mouse lines; R.E. Hill, G. R. Martin, W. D. Richardson, S. Vainio and F. J. Livesey for in situ probes; J. Hebert and T. Theil for advice on setting up the explant culture system; K. Gillies, R. Smith, M. Molinek, L. Leon and animal house staff for expert technical assistance; and members of the Price and Mason laboratories for advice and encouragement. This work was funded by the Wellcome Trust, MRC and BBSRC. Deposited in PMC for release after 6 months.