Combinatorial function of the homeodomain proteins Nkx2.1 and Gsh2 in ventral telencephalic patterning

Development of the vertebrate telencephalon is a complex process that is dependent on the coordinate interactions of extrinsic and intrinsic cues to regulate the growth, patterning, fate specification and migration of cells(reviewed by Wilson and Rubenstein,2000; Schuurmans and Guillemot, 2002; Rallu et al.,2002b; Campbell,2003). Foremost among the extrinsic cues that act to establish dorsal (pallial) and ventral (subpallial) domains within the neuraxis is sonic hedgehog (Shh) (reviewed by Ho and Scott,2002). Although in the telencephalon it is now clear that cues other than Shh must contribute to establishing regional identity(Rallu et al., 2002a),extrinsic signals appear to converge in their induction of a common set of transcription factors. These intrinsic determinants apparently form a transcriptional code, which acts directly to establish regional identity within the telencephalon. Among the factors that contribute to this code are Pax6, Gsh2 and Nkx2.1 (Titf1 – Mouse Genome Informatics), three homeodomain containing transcription factors (reviewed by Wilson and Rubenstein, 2000; Schuurmans and Guillemot,2002; Rallu et al.,2002b; Campbell,2003). Recent analysis indicates that all domains of the telencephalon require at least one of these proteins for the proper development of regional pattern.

In the E10.0 mouse telencephalon the expression of Pax6, Gsh2 and Nkx2.1 is complementary, although, as shown here, this situation exists only transiently. Nonetheless, this pattern of gene expression provides a reliable indication of where these genes are required for regional patterning. For example, Pax6, whose expression at E10.0 is restricted to the dorsal telencephalon (pallium), regulates many aspects of cortical development,including specification of progenitor populations(Stoykova et al., 1996; Caric et al., 1997; Götz et al., 1998; Heins et al., 2002). Similarly, Nkx2.1 and Gsh2, whose expression at this stage is confined to the medial and lateral telencephalic domains, respectively, are required for the proper patterning of each of these regions. In Nkx2.1^–/– mutants, the medial ganglionic eminence (MGE) acquires a lateral ganglionic eminence (LGE) character(Sussel et al., 1999). Moreover, the loss of Gsh2 results in the ectopic expression of cortical genes throughout much of the LGE(Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001).

Both the means by which these three genes regulate one another's expression and their role in establishing regional telencephalic pattern are of considerable interest. In this regard, it is notable that the Drosophila gene vnd (an ortholog of Nkx2.1)functions to repress expansion of lateral fate into the ventral domain of the Drosophila nerve cord (Chu et al., 1998; McDonald et al.,1998). Furthermore, the Drosophila ind gene (an ortholog of Gsh2) is essential for repressing dorsal character within the lateral domain of the fly nerve cord(Weiss et al., 1998). Recent work indicates that similar mechanisms may regulate telencephalic patterning in mice. For example, the primary function of Pax6 and Gsh2,with regard to regional patterning of the telencephalon, is to cross-repress one another. This is evident both from the complementary expansion of Gsh2 into the normal Pax6 domain in Pax6^–/– mutants(Toresson et al., 2000; Yun et al., 2001), and from the expansion of Pax6 into the normal territory of Gsh2expression in Gsh2^–/– mice(Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001). Importantly, telencephalic patterning is largely normal in Gsh2^–/–;Pax6^–/–double mutants (Toresson et al.,2000).

In this study we have analyzed how the complementary patterns of Pax6, Gsh2 and Nkx2.1 expression observed at E10.0 are generated. Interestingly, we find that prior to the onset of Gsh2 expression in the telencephalon, Nkx2.1 expression transiently abuts the Pax6 expression domain at E9.5. Prior to the present study, it was not known whether Nkx2.1 functions to repress either Pax6 or Gsh2 at the stages of development in which their expression patterns are complementary, although such interactions might have been predicted based on the interactions of the orthologs of these genes in the Drosophilanerve cord (Chu et al., 1998; McDonald et al., 1998; Weiss et al., 1998). Surprisingly, we find that these genes do not function cross-repressively to establish distinct progenitor domains within the telencephalon. Furthermore,unlike Gsh2^–/–;Pax6^–/–double mutants, in which the defects observed in the single mutants are rescued in the double mutant, regional patterning is further perturbed in double-mutant mice lacking both Nkx2.1 and Gsh2 gene function. Indeed, in many aspects, the phenotype observed in Nkx2.1^–/–;Gsh2^–/–double mutants resembles that observed in Shh^–/– animals. This indicates that these two genes are primary downstream effectors of the extrinsic signals that act to establish ventral identity in the telencephalon. Hence, although a small ventral telencephalic domain persists in Nkx2.1^–/–;Gsh2^–/–double mutants, it is clear that patterning in this region is largely dependent on the combined function of these genes.

All animals used in these studies were maintained according to protocols approved by the Institutional Animal Care and Use Committee at the NYU School of Medicine. Wild-type, heterozygous and homozygous Gsh2(Szucsik et al., 1997), Nkx2.1 (Sussel et al.,1999), Pax6 (Sey)(Hill et al., 1991) and Shh (Chiang et al.,1996) mutant embryos were obtained from intercrosses of Gsh2^+/–, Nkx2.1^+/–, Pax6 (Sey)^+/– or Shh^+/– mice. Double Nkx2.1^–/–;Gsh2^–/–mutant mice were obtained by intercrosses of Gsh2^+/–and Nkx2.1^+/– mice. For staging of embryos, midday of the vaginal plug was considered as embryonic day 0.5 (E0.5). For viral injection studies, Swiss Webster mice (Taconic Farms, Germantown, New York)were used. Virus preparation and ultrasound surgery were both performed as previously described (Gaiano et al.,1999). CLEG (CLE virus expressing GSH2) was injected into the developing forebrain at titers of 2 to 5×10⁸ cfu/ml.

The Nkx2.1 and Gsh2 alleles were genotyped by PCR using primers previously described (Nery et al.,2001; Szucsik et al.,1997), with the exception that the PCR reaction was enhanced by the use of GC-rich PCR reagents (Roche). Pax6 (Sey) mice were identified based on eye morphology(Hill et al., 1991) and confirmed by PCR as previously described(Grindley et al., 1995). Shh mutants were readily identified based on their severe holoprosencephalic phenotype (Chiang et al., 1996). All genomic DNA for PCR was isolated using a QIAamp Genomic DNA isolation kit (Qiagen).

Whole heads (E9.5-E12.5) or isolated brains (E18.5) were fixed at 4°C in 4% paraformaldehyde for 1-4 hours, rinsed in phosphate-buffered saline(PBS), cryoprotected overnight in 30% sucrose in PBS and embedded in HistoPrep(Fisher Scientific). Embedded tissues were sectioned on a cryostat between 12-20 μm. Section RNA in situ hybridization was performed as described(Schaeren-Wiemers and Gerfin-Moser,1993; Wilkinson and Nieto,1993) using non-radioactive DIG-labelled probes. The following probes were used in this study: Gad67(Behar et al., 1994), Mash1 (Guillemot and Joyner,1993), Ebf1 (Garel et al., 1997), Ngn2(Gradwohl et al., 1996), Gsh1 (Valerius et al.,1995), Lhx6(Grigoriou et al., 1998), Olig2 (Lu et al.,2000), Pdgrfa(Mercola et al., 1990) and Plp/DM20 (Timsit et al.,1995). Localization of Dlx2 expression was achieved by X-gal staining (Corbin et al.,2000) of Dlx2-tau-lacZ heterozygous animals(Corbin et al., 2000; Nery et al., 2002; Nery et al., 2003), or by in situ hybridization using a probe to Dlx2 mRNA(Porteus et al., 1991).

Tissue was processed for immunohistochemistry as described above. The following antibodies were used for immunofluorescence: mouse anti-Pax6(1:1000, gift of A. Kawakami), rabbit anti-Pax6 (1:1000, Covance, CA, USA),mouse anti-Nkx2.1 (1:2000, DAKO, CA, USA), rabbit anti-Nkx2.1 (1:150, Biopat,Italy), mouse anti-5E1 (anti-Shh, 1:2000, Developmental Studies Hybridoma Bank, IA, USA), rabbit anti-Crbp1 (1:400, gift of U. Eriksson, Stockholm,Sweden), rabbit anti-Gsh2 (1:2000, gift of K. Campbell, Cincinnati, OH, USA),rabbit anti-GABA (1:1000, Sigma, MO, USA), rabbit anti-Calbindin (1:1000,Calbiochem, CA, USA), rabbit anti-PLAP (1:100, Accurate Chemical, NY, USA). Secondary antibodies used were: FITC-conjugated donkey anti-rabbit,Cy3-conjugated donkey anti-rabbit, Cy3-conjugated donkey anti-mouse,FITC-conjugated donkey anti-mouse (all from Jackson ImmunoResearch, West Grove, PA, USA). Sections were washed in PBS, blocked for 1 hour with PBS containing 10% donkey serum and 0.2% Triton X-100. Sections were incubated in primary antibodies diluted in block (with 10% serum) overnight at 4°C,then washed three times in PBS and incubated with secondary antibodies diluted in PBS containing 1% donkey serum and 0.2% Triton X-100 for 1-2 hours at room temperature in the dark. Fluorescent images were obtained using either a cooled-CCD camera (Princeton Instruments) and Meta-morph software (Universal Imaging, West Chester, Pennsylvania) or a confocal microscope (Leica).

Of the four genes we examined, Pax6 is expressed earliest and is first detected at 5 somites throughout the anterior neural plate(Inoue et al., 2000; Bell et al., 2001). By 8 somites, expression of Pax6 in the telencephalon is excluded from the ventral-most regions. Nkx2.1 expression is initially observed slightly later at 10-12 somites, followed shortly by the nested expression of Shh in the ventral telencephalon(Shimamura et al., 1995; Sussel et al., 1999). Gsh2 expression is the last to be observed, beginning at about E10.0(Corbin et al., 2000). Currently, how the expression of these genes changes relative to one another during early telencephalic development has not been well characterized. We therefore performed an analysis of the expression patterns of the proteins encoded by these genes.

At 23 somites (∼E9.5), before the onset of Gsh2 expression, Nkx2.1 and Pax6 expression is complementary and forms a clear boundary in the subpallium(Fig. 1A,C). At this stage, Shh is expressed in the ventral-most telencephalon nested within the Nkx2.1 positive domain (Fig. 1B,C; see Fig. S1A). By 27 somites (∼E10.0), a small gap appears between Nkx2.1 and Pax6 expression in the lateral telencephalon(Fig. 1D). At this stage, Gsh2 expression is first observed in this lateral region in a pattern complementary to Pax6, and largely complementary to Nkx2.1 expression(Fig. 1E,F). By E10.5, a larger gap between Nkx2.1 and Pax6 expression is observed(Fig. 1G), and Gsh2 expression expands ventrally into the Nkx2.1 positive domain(Fig. 1H,I). At E10.5,expression of Gsh2 and Pax6 partially overlaps(Fig. 1H)(Toresson et al., 2000);however, by E12.5 this border is more sharply defined(Rallu et al., 2002a) (data not shown).

By 36-38 somites (E10.25-E10.5), Gsh2-expressing cells are no longer restricted laterally, and begin to appear more medially in a posterior to anterior gradient (Fig. 2A-F). At anterior levels, Gsh2 expression remains restricted to the ventrolateral region (Fig. 2A). Further posterior, at the level of the septum (Fig. 2B) and anterior MGE (Fig. 2C), Gsh2 expression remains mostly laterally restricted, but a few Gsh2-positive cells are observed more ventrally(Fig. 2B). However, in the more caudal MGE and LGE more Gsh2-positive cells are present within the Nkx2.1 positive domain (Fig. 2D,E). At the most posterior region of the telencephalon, Gsh2 is highly expressed in the caudal ganglionic eminence (CGE) (Fig. 2F). At this level, Nkx2.1 expression is only observed in the diencephalon. In summary, these data reveal that between E9.5 and E10.5 expression of Nkx2.1, Gsh2 and Pax6 in the telencephalon is highly dynamic,and during a brief period (∼1/2 day) these three proteins are expressed in a complementary pattern.

The complementary expression of Nkx2.1 and Pax6 at 23 somites, and of Gsh2 and Nkx2.1 at 27 somites, suggests that these proteins may be acting cross-repressively to generate distinct domains within the telencephalon. To explore this possibility, we analyzed mice lacking the genes encoding these proteins to determine whether there were changes in gene expression consistent with such cross-repressive interactions (i.e. expansion of gene expression across boundaries). In the absence of Pax6 gene function (in Sey/Sey mutant mice), expression of Nkx2.1 does not expand dorsally(Fig. 3A,B). Conversely, in Nkx2.1^–/– mutant mice, expression of Pax6 does not expand ventrally (Fig. 3C,D). Hence, at ∼E9.5, when the expression of Pax6 and Nkx2.1 form a boundary, Pax6 and Nkx2.1 do not cross repress one another. Notably, Nkx2.1^–/– mutants display a loss of telencephalic Shh (Sussel et al.,1999) (see Fig. S1A,B),indicating that telencephalic Shh is not required to repress Pax6 expression from the subpallium. Thus, Shh-dependent repression of Pax6 in the subpallium(Chiang et al., 1996) is mediated by non-telencephalic sources of Shh (M. Fuccillo and G.F.,unpublished). Moreover, Pax6 does not regulate the domain of ventral telencephalic Shh expression because the expression of Shh is unaltered in Sey/Sey mutant mice (see Fig. S1C,D).

Examination of potential interactions between Gsh2 and Nkx2.1 prior to when Gsh2 expression partially expands into the Nkx2.1 domain also provided no evidence for cross-repression between these genes (or gene products). In Nkx2.1^–/– mice,Gsh2 expression does not expand (Fig. 3E,F). Conversely, in Gsh2^–/–mice, Nkx2.1 expression remains restricted to its normal ventral domain(Fig. 3G,H). Interestingly however, retroviral-mediated ectopic expression of Gsh2 at in the MGE at E9.5,is sufficient to repress Nkx2.1 expression(Fig. 4A-F). Notably,Gsh2-expressing retrovirus was delivered at E9.5, a time that is before Gsh2-expressing cells are normally observed in the MGE (∼E10.5). Thus,Gsh2 gain- and loss-of-function experiments reveal that although Gsh2 is sufficient to repress Nkx2.1 expression in the medial domain prior to its normal onset of expression, it is not necessary to repress Nkx2.1 expression in the lateral domain. Previous studies have also revealed that Nkx2.1 expression does not expand dorsally in the absence of either Gsh1 alone, or in the combined absence of both Gsh1 and Gsh2, indicating that neither of the known Gsh genes repress Nkx2.1 gene function in the LGE(Toresson and Campbell, 2001). Hence, although Pax6 and Gsh2 function to cross-repress one another(Toresson et al., 2000; Yun et al., 2001) (data not shown) during the developmental periods when Pax6 and Nkx2.1, and Nkx2.1 and Gsh2, maintain complementary patterns of expression, they do not function cross-repressively.

Nkx2.1^–/– and Gsh2^–/– mutant mice display defective patterning of the MGE and LGE, respectively. In Nkx2.1^–/– mutant mice there is a conversion of the MGE to an LGE fate (Sussel et al.,1999). Conversely, the loss of Gsh2 in mice results in severe patterning defects in the LGE. In Gsh2^–/– mutants, there is a loss in the expression of pan-ventral telencephalic genes (e.g. Dlx1/2, Mash1),in all but the ventral-most domain of the LGE (Suczick et al., 1997; Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001), that is combined with an expansion of pallial markers, such as Pax6 and Ngn2. These results indicate that Nkx2.1 and Gsh2may act in combination to pattern both the MGE and LGE. To investigate the effect of the combined loss of these genes on early ventral telencephalic patterning, Nkx2.1^–/–;Gsh2^–/–compound mutant mice were generated and the expression of a variety of subpallial markers (Dlx2, Mash1, Gsh1, Lhx6) was analyzed. As loss or reduction of Nkx2.1 and Gsh2 gene expression are two of the more prominent abnormalities seen in the telencephalon of Shh^–/– mutants(Corbin et al., 2000; Pabst et al., 2000; Rallu et al., 2002a),patterning defects in Shh^–/– mutant mice were also compared with that of double Nkx2.1^–/–;Gsh2^–/–mutant mice.

At E12.5, the patterning genes Dlx2(Fig. 5A) and Mash1(Fig. 5F) are expressed pan-ventrally. In the absence of Gsh2, expression of both Dlx2 and Mash1 is lost in the dorsal-most two thirds of the LGE (Fig. 5B,G)(Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001). In Nkx2.1^–/– mutant mice, the domain of Dlx2 and Mash1 expression is unaffected(Fig. 5C,H)(Sussel et al., 1999). In Nkx2.1^–/–;Gsh2^–/–double-mutant mice, Dlx2 and Mash1 expression is significantly reduced (Fig. 5D,I). A similar persistence of Dlx2 and Mash1expression is observed in the ventral-most region of the mutant telencephalon of less severely affected Shh mutant mice(Fig. 5E,J)(Rallu et al., 2002a) (K. Campbell, personal communication).

Gsh1, the other known member of the Gsh family, is expressed in the MGE and the ventral-most LGE (Fig. 5K) (Toresson and Campbell,2001). Functioning in combination with Gsh2, Gsh1 plays a key role in both patterning of the LGE and generation of early born striatal cells (Toresson and Campbell,2001). In Gsh2^–/– mutant mice, Gsh1 expression expands dorsally into the LGE(Fig. 5L)(Toresson and Campbell, 2001). In Nkx2.1^–/– mutant mice, a small domain of Gsh1 expression persists in the subpallium(Fig. 5M). A similar level of Gsh1 expression is also observed in Nkx2.1^–/–;Gsh2^–/–double mutant mice (Fig. 5N). By contrast, Gsh1 expression is completely lost in all Shh^–/– mutants analyzed (n=9),regardless of their severity (Fig. 5O). At E12.5, Lhx6, a LIM homeodomain-containing transcription factor gene, is expressed in the MGE and MGE-derived cells that migrate to the LGE and cortex (Fig. 5P) (Lavdas et al.,1999). Expression of Lhx6 is unaffected in Gsh2^–/– mutants(Fig. 5Q), but is completely lost in Nkx2.1^–/– mutants(Fig. 5R)(Sussel et al., 1999). In Nkx2.1^–/–;Gsh2^–/–double mutants (Fig. 5S), and in all Shh^–/– mutants analyzed(Fig. 5T) (K. Campbell,personal communication), Lhx6 expression is also lost. Taken together, these analyses reveal that the loss of both Nkx2.1 and Gsh2 results in more severe patterning defects than the loss of either Nkx2.1 or Gsh2 alone. Furthermore, specific changes in subpallial gene expression (Dlx2, Mash1, Lhx6) mimic that found in Shh^–/– mutant mice.

To further compare the phenotype observed in Nkx2.1^–/– and Gsh2^–/– single versus compound mutants, we analyzed the expression of LGE (Crbp1, Ebf1) and pallial(Ngn2, Pax6) specific markers in mice bearing these genotypes. Expression of the pallial markers Ngn2 and Pax6 normally extends just across the cortical-striatal sulcus into the dorsolateral LGE(Fig. 6A,E,M). In Gsh2^–/– mutants, expression of both Ngn2 and Pax6 extends ectopically into all but the ventral third of the LGE (Fig. 6B,F,N)(Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001). Expression of Ngn2 and Pax6 in Nkx2.1^–/– mutants(Fig. 6C,G,O) resembles that observed in control embryos. In Nkx2.1^–/–;Gsh2^–/–double mutants (Fig. 6D,H,P)the expansion of Ngn2 and Pax6 into the subpallium is similar to that observed in Gsh2^–/– mutants. As in the Gsh2^–/– single mutants, this expansion extends to the level of remnant subpallial gene expression(Fig. 5D,I,N).

Consistent with the MGE adopting LGE character in Nkx2.1^–/– mutants(Sussel et al., 1999)(Fig. 6K,O), the LGE-specific marker Crbp1 expands ventrally into the MGE in Nkx2.1^–/–;Gsh2^–/–double mutants (Fig. 6L,P). Similar changes are observed in the expression of the helix-loop-helix transcription factor gene Ebf1, an early marker of striatal(LGE-derived) projection neurons (Fig. 6Q) (Garel et al.,1997; Garel et al.,1999). In Gsh2^–/– mutants, Ebf1 expression is significantly reduced(Fig. 6R)(Corbin et al., 2000), and in Nkx2.1^–/– mutants, expression of Ebf1expands ventrally (Fig. 6S). Although the level of expression is reduced, Ebf1 expression also expands ventrally in Nkx2.1^–/–;Gsh2^–/–double mutants (Fig. 6T). Notably, Ebf1 expression is never observed in Shh^–/– mutant mice (n=9) (data not shown). In summary, these data demonstrate that the defects observed in Nkx2.1^–/–;Gsh2^–/–double mutant mice resemble a combination of the phenotypes observed in single Nkx2.1^–/– and Gsh2^–/– mutants. Similar to Nkx2.1^–/– mutants, Nkx2.1^–/–;Gsh2^–/–double mutants display a conversion of the MGE to an LGE fate, and, similar to Gsh2 mutants, expression of LGE genes is supplanted by expression of pallial markers throughout the dorsal two-thirds of the LGE.

Significant numbers of interneurons and oligodendrocytes arise from the subpallium, and populate the pallium via tangential migration (reviewed by Corbin et al., 2001; Marin and Rubenstein, 2001). Furthermore, a number of genes, most prominently Dlx1/2 and Mash1, have been hypothesized to be involved in the specification of both cell types (reviewed by Bertrand et al., 2002). Reduction in the expression of these genes in mutant mice lacking both Nkx2.1 and Gsh2 gene function(Fig. 5) indicates that specification of oligodendrocytes and interneurons in ventral regions may also be affected. Gad67, the precursor enzyme for formation of GABA(Behar et al., 1994), marks developing ventral interneuron populations, many of which subsequently undergo long range tangential migration to the developing cortex. In Nkx2.1^–/–;Gsh2^–/–double mutants, Gad67 expression is significantly reduced(Fig. 7A-D). Interestingly, in less severely affected Shh^–/– mutants,expression of Gad67 persists (Fig. 7E). As Gad67 also marks developing striatal projection neurons, the status of cortical interneurons in Nkx2.1^–/–;Gsh2^–/–double mutant mice was determined by the analysis of calbindin and GABA expression at E18.5. At a level similar to the cortical interneuron defect in single Nkx2.1^–/– mutants(Sussel et al., 1999),generation of cortical interneurons in Nkx2.1^–/–;Gsh2^–/–double mutants is markedly reduced (Fig. 8A-D).

During telencephalic development, oligodendrocyte precursors are marked by the expression of either Pdgfra or Plp/DM20(Spassky et al., 1998; Perez-Villegas et al., 1999; Nery et al., 2001). Moreover,Olig genes are necessary and sufficient for the generation of oligodendrocytes throughout the neuraxis (reviewed by Marquardt and Pfaff, 2001; Sauvageot and Stiles, 2002). Therefore, the status of oligodendrocytes in single and double Nkx2.1^–/– and Gsh2^–/– mutants was examined by the expression of Olig2, Pdgfra and Plp/DM20. Olig2 is expressed in the MGE, LGE and CGE (Fig. 7F and data not shown). In Gsh2^–/– mutants, Olig2 expression is reduced in the LGE(Fig. 7G) to a level similar to the reduction of Dlx2 and Mash1 expression(Fig. 5B,G). In Nkx2.1^–/– mutants, expression of Olig2 is unaffected (Fig. 7H). In Nkx2.1^–/–;Gsh2^–/–double mutants, expression of Olig2 at E12.5, although significantly reduced, persists in the ventral-most telencephalon(Fig. 7I). In addition,consistent with previous observations(Alberta et al., 2001), the expression of Olig2 is lost in all Shh^–/– mutant mice examined(Fig. 7J). Expression of Pdgfra is observed in the MGE(Fig. 7K), a known source of oligodendrocyte progenitors (Nery et al.,2001; Tekki-Kessaris et al.,2001). In Gsh2^–/– mutants, the normal population of Pdgfra-positive cells in the MGE is unaffected,but Pdgfra is ectopically expressed in the VZ of the MGE and in the ventral-most LGE (Fig. 7L). By contrast, Pdgfra expression is absent in Nkx2.1^–/– mutants(Fig. 7M)(Nery et al., 2001; Tekki-Kessaris et al., 2001). In Nkx2.1^–/–;Gsh2^–/–double mutants, although the ventral-most population Pdgfra is lost,there is strong expression of Pdgfra in the lateral domain, similar to that observed in the ventral LGE in Gsh2^–/–mutants. Plp/DM20 marks distinct, more caudal populations of subpallial derived oligodendrocytes that arise, in part, from the presumptive amygdaloid region of the caudal ventral telencephalon(Fig. 7P). In contrast to the Pdgfra-positive population, the generation of Plp/DM20-positive cells is not dependent on Gsh2(Fig. 7Q), Nkx2.1(Fig. 7R)(Nery et al., 2001), or the combined function of Nkx2.1 and Gsh2(Fig. 7S). However, the expression of both Pdgfra and Plp/DM20, is lost in Shh^–/– mutants(Fig. 7O,T). In summary, these data indicate that the loss of both Nkx2.1 and Gsh2 gene function has significant effects on the generation of interneurons and oligodendrocytes.

Homeodomain containing proteins play an essential role in patterning of the vertebrate telencephalon. Of these genes, Nkx2.1, Gsh2 and Pax6 are key regulators of MGE, LGE and cortical development,respectively. We demonstrate that these genes undergo dynamic changes in their expression pattern during a brief developmental time period (∼E9.5-E10.5). Although both Nkx2.1 and Pax6, and Nkx2.1 and Gsh2 have periods where their expression patterns are largely complementary, we find no evidence to suggest that these domains of gene expression are maintained through cross-repression. However, our gain-of-function data demonstrates that Gsh2 can repress Nkx2.1 when ectopically expressed in the MGE, indicating that redundant factors may prevent Nkx2.1 expansion into the LGE of Gsh2^–/– mutants. Furthermore, Nkx2.1and Gsh2 compound mutants have a more severe phenotype within the ventral telencephalon than mutants lacking either of these genes individually. This suggests that unlike Pax6 and Gsh2, whose functions appear to be largely antagonistic to one another (Toresson and Campbell, 2000; Corbin et al., 2000; Yun et al., 2001), Gsh2 and Nkx2.1 act cooperatively to pattern the ventral telencephalon. Interestingly, our gain- and loss-of-function data also indicate that the developmental lag in the expression of Gsh2 in the medial domain may have important implications for Nkx2.1-dependent early specification of oligodendrocytes.

In the developing spinal cord, distinct progenitor domains are established via cross-repressive interactions between Shh-induced class II genes(Nkx2.2, Nkx2.9, Nkx6.1 and Olig2) and Shh-repressed class I genes (Dbx1, Dbx2 and Pax6) (reviewed by Marquardt and Pfaff, 2001; Kessaris et al., 2001). To date, with the exception of the bHLH-containing Olig2 gene, these genes are members of the homeodomain-containing transcription factor family. Similarly, in the telencephalon, the homeodomain containing genes Pax6 and Gsh2 function cross-repressively in the establishment of the molecular boundary at the cortico-striatal junction(Toresson et al., 2000; Yun et al., 2001). Despite this, the present findings demonstrate that, at least in the telencephalon,the apposing expression patterns of homeodomain proteins are not always dependent on cross-repression. In the case of Nkx2.1 and Pax6 this is particularly surprising. In the ventral spinal cord, Nkx (Nkx2.2 and Nkx2.9)and Pax6 function via a cross-repressive mechanism in establishment of the P3 progenitor domain (Briscoe et al.,1999; Briscoe et al.,2000). Indeed, previous studies have indicated that these genes may function cross-repressively in the telencephalon(Sussel et al., 1999; Stoykova et al., 2000),implying that the mechanisms involved in patterning of the telencephalon and spinal cord are analogous (reviewed by Wilson and Rubenstein, 2000; Marin and Rubenstein, 2001). However, this proposal was based, in part, on the observations of the effect of Nkx2.1 and Pax6 loss-of-function mutants at later times during development (E11-E13). At this stage, Nkx2.1^–/– mutants display an expansion of Pax6 mRNA into the LGE, whereas, in Pax6^–/– (Sey/Sey) mutants, Nkx2.1-positive cells are found ectopically in the LGE in a pattern resembling an increase in tangential migration. Notably however, by this time in development, Pax6 and Nkx2.1 expression is separated by the Gsh2 positive domain and they no longer appose each other. Although these studies reveal important later functions of Nkx2.1 and Pax6 in the maintenance of regional pattern and/or cell migration pathways, our results indicate that at earlier times during development, when their expression is in apposition, they are not cross-repressive.

The lack of Gsh2 repression by Nkx2.1 is also contrary to predictions based on the genetic interactions utilized to establish neural progenitor domains in Drosophila (reviewed by Cornell and Von Ohlen, 2000). The Drosophila nerve cord is divided into three distinct neural progenitor domains, dorsal, intermediate and ventral. The ventral domain expresses the homeodomain transcription factor vnd (the Nkx ortholog), which functions in a manner analogous to Nkx2.1 by repressing intermediate character. Despite the similarity in the requirements for these genes, the failure of Gsh2 expression to expand ventrally in Nkx2.1^–/– mutants demonstrates that the conversion of the MGE to an LGE in Nkx2.1 mutants is not dependent on Gsh2. Furthermore, if Nkx2.1 acts to suppress LGE character in the MGE by blocking Gsh2 function by means other than transcriptional repression,one would predict that the MGE would be rescued in Nkx2.1^–/–;Gsh2^–/–double mutants. As shown in Fig. 6K,O, this is not the case, as even in the double mutants the MGE appears to adopt LGE character.

Although telencephalic patterning is severely affected by the loss of Shh, recent examination of ventral patterning in Shh^–/– mutants has revealed that specific aspects of ventral patterning can persist in the absence of Shh(Rallu et al., 2002a) (K. Campbell, personal communication). Taken as a whole, these data indicate that the residual ventral patterning in Shh^–/–mutants is of LGE (i.e. lateral) character. Persistence of panventrally expressed genes in Shh^–/– mutants, including Gsh2, Dlx2, Mash1, and Gad67 and Crbp1, which specifically marks LGE radial glia, supports this notion. The observation that Shh-independent processes appear to be required to establish the lateral (LGE)domain of the telencephalon is reminiscent of the specification of the lateral VO and V1 interneuron populations in the absence of Shh in the spinal cord (Pierani et al., 1999). Furthermore, although Gsh2 is a downstream target of Shhsignaling, as revealed by gain-of-function studies, in Shh^–/– mutant mice expression of this gene persists, albeit at reduced levels (Rallu et al., 2002a). The most compelling evidence for Shh-independent signaling in the establishment of ventral telencephalic pattern comes from the rescue of ventral patterning seen in Gli3;Shh or Gli3;Smomutants, including complete restoration of the normal Gsh2 and Nkx2.1 expression domains (Rallu et al., 2002a). Although the nature of this signaling at present remains unclear, Bmp, Fgf, retinoid, Wnt or Nodal signaling all represent promising candidates for mediating Shh-independent signaling within the telencephalon (Rallu et al.,2002b).

The significant reduction in ventral telencephalic patterning in the absence of Nkx2.1 and Gsh2 gene function suggests that,regardless of how Shh-dependent and independent mechanisms cooperate in the establishment of ventral telencephalic pattern, their actions must converge at the level of these two homeodomain proteins(Fig. 9). Despite the importance of these genes, we observed persistence of some ventral pattern in the Nkx2.1^–/–;Gsh2^–/–compound mutants. This residual pattern may be attributable to the persistence and expansion of Gsh1 expression in these animals. In single Gsh2^–/– mutants, Dlx2 and Mash1 expression remains in the ventral most aspect of the LGE(Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001). Strikingly,in compound mutant mice lacking both Gsh1 and Gsh2 gene function, this domain of Dlx2 and Mash1 expression is completely lost from the entire LGE at E12.5, indicating that Gsh1 in combination with Gsh2, regulates patterning in the entire LGE(Toresson and Campbell, 2001). Furthermore, as the MGE is converted to an LGE fate in the absence of Nkx2.1 gene function (Sussel et al., 1999), the remnant Gsh1 expression in Nkx2.1^–/–;Gsh2^–/–double mutants is most probably derived not from the MGE, but from the ventral LGE instead. The observation that the domain of Dlx2 and Mash1 expression closely matches that of Gsh1 expression lends further credence to the hypothesis that Gsh1 is responsible for the residual ventral patterning observed in Nkx2.1^–/–;Gsh2^–/–double mutants. Exploration of triple-mutant mice lacking Nkx2.1,Gsh1 and Gsh2 will be further required to directly address this hypothesis. Therefore, it may be that the function of Nkx and Gsh genes is all that is required to pattern the MGE and LGE, and these genes represent the convergence of Shh-dependent and Shh-independent patterning(Fig. 9).

Previous work has indicated that both oligodendrocytes and interneurons are primarily specified in the ventral telencephalon, possibly by the functions of Dlx1/2 and Mash1 (reviewed by Bertrand et al., 2002). In the combined absence of both Nkx2.1 and Gsh2, the generation of both of these cell types is altered in profound and distinct ways. Although still present, interneurons in the cerebral cortex of double mutants are markedly reduced and resemble those observed in Nkx2.1 mutants(Sussel et al., 1999). Recent studies have revealed that the primary source of cortical interneurons is the MGE (Lavdas et al., 1999;Wichterle et al., 1999, 2001; Anderson et al., 2001; Nery et al., 2002). Therefore,because the combined loss of Nkx2.1 and Gsh2 results in a conversion of the MGE to an LGE fate similar to that observed in single Nkx2.1 mutants, it is not surprising that the two mutants resemble each other with regard defects in the generation of interneurons.

Our results also give a novel insight into the function of Nkx2.1and Gsh2 in oligodendrocyte development. Subpallial-derived oligodendrocytes consist of at least two distinct populations: those that express Pdgfra and those that express Plp/DM20(Spassky et al., 1998; Perez-Villegas et al., 1999; Nery et al., 2001). Our results reveal that only the PDGFRα-positive population is dependent on the function of Nkx2.1 and Gsh2; the Plp/DM20population is unaffected by the loss of these genes. As shown here and previously (Nery et al.,2001), the generation of the PDGFRα-population is positively regulated by the function of Nkx2.1. Interestingly, this population also appears to be under negative regulation by Gsh2, as indicated by the derepression of Pdgfra expression in the MGE and LGE VZ in the absence of Gsh2. Moreover, ectopic expression of Gsh2 in the medial domain early in development (E9.5), as revealed by retroviral gain-of-function experiments, results in the repression of Nkx2.1, an essential regulator of the generation of PDGFRα-positive oligodendrocytes in the MGE. Therefore, as Gsh2 may repress the generation of PDGFRα-positive oligodendrocytes in the developing MGE, specification of these cells may occur prior to the normal expansion of Gsh2 into the medial domain during development (<E10.5). In contrast to the generation of PDGFRα-positive cells, the specification of the Plp/DM20-positive cells is not dependent on Nkx2.1 or Gsh2. Plp/DM20-positive cells are generated in the presumptive amygdaloid region of the caudal subpallium. We have previously revealed that this region is a unique progenitor zone distinct from the MGE and LGE, and that it is not dependent on the function of Nkx2.1 and Gsh2 (Nery et al., 2002). Therefore, other transcription factor(s) must regulate the specification of this population. However, these factors are presumably downstream of Shh, as both PDGFRα- and Plp/DM20-positive oligodendrocyte populations are absent in Shh^–/– mutants.

Taken together with previous studies, a hierarchy of gene expression for producing interneurons and oligodendrocytes is becoming apparent. Initiating the generation of these cell types in ventral regions are extrinsic cues,including Shh. These cues result in the expression of homeodomain genes, including Nkx2.1 and Gsh2, that ensure the expression of pan-ventral transcription factors, such as Dlx1/2, Mash1 and Olig2, in the MGE and LGE. These genes, in turn, may act as key effectors in the generation of specific ventral cell types, such as interneurons, and distinct populations of oligodendrocytes.

We thank Yuan Yuan Huang for excellent technical assistance, and members of the Fishell laboratory for insightful discussions andcritical reading of the manuscript. We also thank Irene Zohn and Deborah Yelon for critical reading of the manuscript, and Hakan Torreson and Kenny Campbell for sharing unpublished data. We thank the following for probes and reagents: S. Potter (Gsh2and the Gsh2^–/– mutant mice), C. Gerfin(GAD67), J. Rubenstein (Dlx2, Lhx6), F. Guillemot(Mash1, Ngn2), P. Charnay (Ebf1), Charles Stiles(Olig2), R. Lang (Sey mice), B. Richardson (Pdgfra,Plp/DM20), A. Kawakami (mouse anti-Pax6), H. Westphal and C. Chiang(Shh^–/– mice), S. Kimura(Nkx2.1^–/– mice), U. Eriksson (anti-CRBP), and K. Campbell (anti-Gsh2, Gsh1). This work was supported by NIH grant(NS39007) to G.F. J.C. is a recipient of a NIH post-doctoral fellowship(NS10962-01).