Genetic control of dorsal-ventral identity in the telencephalon: opposing roles for Pax6 and Gsh2

In the embryonic telencephalon, the cerebral cortex and striatum develop largely from two adjacent neurogenic territories along the dorsal-ventral axis. The cortex derives from the pallium (Bayer and Altman, 1991), while the more ventrally situated lateral ganglionic eminence (LGE) gives rise to the majority of striatal neurons (Deacon et al., 1994; Olsson et al., 1995, 1998). A number of developmental control genes are specifically expressed by progenitors in either the cortex or LGE. For instance, progenitor cells in the developing cortex express the homeobox gene Emx1 (Simeone et al., 1992; Briata et al., 1996) as well as the basic helix-loop-helix (bHLH) genes neurogenin 1 (Neurod3 – Mouse Genome Informatics) and neurogenin 2 (Atoh4 – Mouse Genome Informatics) (Sommer et al., 1996). Progenitors in the LGE express homeobox genes of the Dlx family (Liu et al., 1997) and Vax1 (Hallonet et al., 1998) as well as the bHLH gene Mash1 (Ascl1 – Mouse Genome Informatics) (Guillemot et al., 1993; Porteus et al., 1994). Each of these genes has been shown to be required for the normal development of their respective telencephalic compartments (Yoshida et al., 1997; Anderson et al., 1997b; Casarosa et al., 1999; Horton et al., 1999; Fode et al., 2000). Although the molecular and cellular mechanisms underlying the establishment of these expression domains along the dorsal-ventral axis of the embryonic telencephalon is not completely clear, it appears that there are many similarities with more caudal parts of the neural tube. Dorsal-ventral patterning in the developing spinal cord and hindbrain is known to be regulated by the actions of bone morphogenetic proteins (BMPs) and sonic hedgehog (SHH) emanating from the dorsal and ventral midline, respectively (Tanabe and Jessell, 1996). These signals establish domains of regulatory gene expression along the dorsal-ventral axis that subsequently control the generation of specific cell types (Ericson et al., 1997; Liem et al., 1997; Briscoe et al., 1999). SHH is also capable of inducing ventral gene expression in the telencephalon (Ericson et al., 1995; Kohtz et al., 1998; Gaiano et.al., 1999). BMPs, conversely, have been shown to regulate the expression of certain dorsal telencephalic markers in explant assays (Furuta et al., 1997). These signalling molecules, however, remain confined to the ventromedial (Shimamura et al., 1995) and dorsomedial aspects (Furuta et al., 1997), respectively, despite a dramatic expansion of the telencephalic neuroepithelium. Thus, cell-and/or region-intrinsic factors must maintain the molecular specification of telencephalic progenitors located at different positions along the dorsal-ventral axis throughout neurogenesis.

In this respect, it is interesting to consider the homeobox genes Pax6 and Gsh2, which show complementary expression in progenitors of the cortex and LGE, respectively. Small eye (Sey) mice, containing a point mutation in the Pax6 gene (Hill et al., 1991), show severe disruptions in the development of the cortex (Schmahl et al., 1993; Caric et al., 1997). These abnormalities include a progressive dorsal spread of Dlx1 and Vax1 expression, in cortical cells (Stoykova et al., 1996; Hallonet et al., 1998). Conversely, Gsh2 mutants show a reduction in the size of the LGE and a loss of Dlx2 expression in the LGE but not the medial ganglionic eminence (MGE) (Szucsik et al., 1997). These findings implicate Pax6 and Gsh2 in the establishment and/or maintenance of dorsal-ventral identity. In this study we have examined the respective roles of Pax6 and Gsh2 in the specification of progenitors in the cortex and LGE. By studying single and double loss-of-function mouse mutants, we demonstrate that Pax6 and Gsh2 genetically oppose each other in the maintenance of Mash1, Dlx and neurogenin gene expression in telencephalic progenitors. Alterations in the expression domains of these genes in Sey/Sey, (i.e. Pax6 mutant) and Gsh2^−/− embryos, leads to abnormal development of the cortex and striatum, respectively.

Sey/Sey mutants were identified by the lack of eyes. At early stages (e.g. E12.5), wild types were obtained from non-littermates, while in the later stage litters (E14 and E16.5) genotyping for wild types and heterozygotes was also carried out based on eye morphology as described by Hill et al. (1991). Gsh2^−/− mutants were genotyped using PCR as previously described (Szucsik et al., 1997). Double Sey/Gsh2 homozygotes were first identified by the absence of eyes and then with PCR for the mutant and wild type Gsh2 alleles. All genotyping was confirmed by immunostaining for GSH2 or PAX6 (or both in the case of the Sey/Gsh2 double mutants).

All embryos were fixed overnight in 4% paraformaldehyde at 4°C and subsequently cryoprotected in 30% sucrose before sectioning at 10-12 μm on a cryostat. Immunohistochemistry was performed as described previously for DLX and RBP1 (Toresson et al., 1999). The other primary antibodies were used at the following concentration; rabbit anti-PAX6 at 1:400 (provided by S. Wilson), rabbit anti-ISL1/2 at 1:400 (provided by T. Edlund), mouse anti-MASH1 at 1:5 (provided by D. J. Anderson), rabbit anti-SCIP at 1:100 (provided by G. Lemke), rabbit anti-Ki67 at 1:100 (Dianova) and rabbit anti-EMX1 used at 1:500 (provided by G. Corti). The GSH2 antibody was raised in rabbits against the C-terminal peptide, ANEDKEISPL (Hsieh-Li et al., 1995). This antibody was used at a 1:5000 dilution. In situ hybridization was performed as described (Toresson et al., 1999), using digoxigenin-labeled cRNA probes for Gsh2, Mash1, neurogenin 1, neurogenin 2 and Pax6. TUNEL staining was carried out on embryos at E10.5, E11.5, E12.5, E14.5 and E16.5 according to the protocol provided by Boehringer Mannheim, and detection was made by fluorescence microscopy.

The expression patterns of PAX6 and GSH2 in the embryonic telencephalon suggests a role for these molecules in the regulation of dorsal-ventral identity. Indeed, already at E10.5, when GSH2 is first detected in the ventral telencephalon (Fig. 1A), its expression was largely complementary to that of the cortically expressed PAX6 (Fig. 1B). By E12.5, when both the MGE and LGE were morphologically distinct, GSH2 was found in progenitors of both the MGE and LGE ventricular zone (VZ) (Fig. 1C). At this stage, PAX6 was expressed at high levels throughout the cortical VZ and into the lateral-most part of the LGE, forming a boundary with GSH2 (Fig. 1D). Beyond this limit, PAX6 was expressed at low levels in the LGE VZ (Fig. 1D and Sussel et al., 1999). The expression level of both PAX6 and GSH2 was graded along the dorsal-ventral axis, with the highest levels present in progenitor cells close to the boundary. In addition to the VZ expression, PAX6 was also observed in a stream of cells emanating from the boundary of PAX6 and GSH2 expression in the VZ (Fig. 1D). These cells have previously been suggested to form portions of the claustrum and amygdala (Puelles et al., 1999). The expression patterns of GSH2 and PAX6 were maintained throughout embryogenesis, marking the boundary between the LGE and cortical VZ (Fig. 1E,F).

In addition to the reported abnormalities in Dlx2 expression (Szucsik et al., 1997), we show here that Gsh2 mutants exhibit altered expression patterns of several developmental control genes in the LGE. At E11.5, when the LGE is first morphologically distinct, DLX proteins (including DLX1,2,5 and 6; Fig. 2A) and MASH1 (Fig. 2B), which are normally expressed at high levels in progenitors of the LGE were missing from a large portion of the Gsh2 mutant LGE (Fig. 2G,H). Interestingly, however, the targeted Gsh2 transcript could still be detected in its normal expression domain (compare Fig. 2I with 2C). The loss of DLX and MASH1 expression correlated well with ectopic expression of cortical markers in progenitors of the mutant LGE. Specifically, PAX6 (Fig. 2D) was dramatically upregulated in the mutant LGE (Fig. 2J), while the bHLH genes neurogenin 1 (not shown) and neurogenin 2 (Fig. 2E) were found ectopically in mutant LGE cells (Fig. 2K and data not shown).

Similar alterations in the expression of PAX6 (Fig. 3B), neurogenin 1 and neurogenin 2 (data not shown), DLX (data not shown), and MASH1 (Fig. 3D) were also observed in the E12.5 LGE of the Gsh2 mutant. The mis-specification of striatal progenitors at this stage, however, is less severe than at E11.5 with almost half of the LGE expressing MASH1 and DLX. In addition to the mis-specification of progenitors in the LGE VZ, the size of the proliferative subventricular zone (SVZ) in the Gsh2 mutant LGE, as marked by the cell cycle marker Ki67 (Schlüter et al., 1993), is reduced (Fig. 3F) when compared with the wild type (Fig. 3E). Furthermore, the stream of PAX6-positive cells in the lateral LGE, which normally emanates from the boundary between the cortical and striatal VZ (Fig. 1D and 3A) was severely reduced in the Gsh2 mutant (Fig. 3B).

As noted above, the targeted Gsh2 transcript in the mutant LGE could still be detected in its normal expression domain (Fig. 2I), indicating that only a partial mis-specification of LGE progenitors has occurred. Indeed, cellular retinol binding protein 1 (RBP1), which is normally expressed in LGE VZ cells (Toresson et al., 1999 and Fig. 3C), was still expressed in its normal domain in the mutant LGE at E12.5 (Fig. 3D). Moreover, expression of the cortically restricted homeobox protein, EMX1 (Fig. 2F) was not altered in the Gsh2 mutant (Fig. 2L). Gsh2, therefore, appears to act in parallel with other genetic pathways to control the molecular specification of early LGE progenitors.

The abnormalities observed in LGE development at earlier stages in Gsh2 mutants were less pronounced by E16.5. Although the mutant LGE is notably smaller than the wild type, numerous Ki67-positive cells (i.e. proliferating progenitors) were observed in the SVZ of the mutant LGE (data not shown). Furthermore, MASH1 (Fig. 4B) and DLX proteins (data not shown) were expressed in progenitors throughout the LGE up to the LGE/cortex boundary, while PAX6 (Fig. 4B) and neurogenin 2 (data not shown) were no longer ectopically expressed in LGE progenitors.

Since the LGE is known to generate the vast majority of striatal neurons (Deacon et al., 1994; Olsson et al., 1995, 1998) the early mis-specification of VZ progenitors combined with the reduction in the proliferative SVZ in the Gsh2 mutant is likely to have effects on striatal development. In fact, at E16.5, the size of the mutant striatal complex (including the dorsal striatum, nucleus accumbens and the olfactory tubercle), as marked by the LIM-homeodomain protein, islet 1 (ISL1) was notably reduced when compared with wild type (Fig. 4C-F). Moreover, ISL1 staining in the olfactory tubercle was missing in Gsh2 mutants (Fig. 4D). Cell death, as detected by TUNEL staining, did not appear to contribute to this phenotype, since no differences in the number of apoptotic profiles were found between the mutant or wild-type LGE at any stage examined (data not shown). Interestingly, olfactory tubercle neurons, which are deficient in the Gsh2 mutant, are largely generated at early time points in striatal neurogenesis (Bayer and Altman, 1995). Thus, these defects in striatal development correlate well with the early molecular mis-specification of striatal progenitors in the LGE.

The above results demonstrate that Gsh2 is required to maintain the correct molecular identity of LGE precursor cells. In its absence, the dorsal genes, Pax6, neurogenin 1 and neurogenin 2 are ectopic, and ventral genes, like Dlx, and Mash1 are lost. Interestingly, Sey/Sey mice (i.e. Pax6 mutants (Hill et al., 1991) display ectopic ventral gene expression (Dlx1 and Vax1) in cells of the developing cortex (Stoykova et al., 1996; Hallonet et al., 1998). We were therefore interested to determine if Gsh2 is similarly deregulated in Sey/Sey cortical progenitors. Indeed, at E12.5, the expression of GSH2 was expanded throughout the lateral LGE and even into the ventrolateral cortical VZ of the Sey/Sey telencephalon (Fig. 5F). The expression level of the mutated Pax6 transcript appears to be downregulated specifically within the domain of ectopic GSH2 (Fig. 5E). In addition, both Mash1 (Fig. 5H) and DLX proteins (data not shown), whose correct LGE expression depends on Gsh2 at this stage, were also ectopic in the germinal zone of the Sey/Sey cortex. The expression of the dorsally restricted neurogenin 1 and neurogenin 2 genes was significantly downregulated throughout the cortex and lost in the lateral-most edge of the LGE and ventrolateral cortex (Fig. 5G and data not shown).

The extent of ectopic GSH2 expression in Sey/Sey cortical progenitors progresses during development so that by E14 most of the lateral cortical VZ was covered (Fig. 6H). This progression was paralleled also by ectopic DLX and Mash1 expression (Fig. 6F,G). While many of the cortical cells ectopically expressing DLX were found in the SVZ and intermediate zone, Mash1 expression was confined to progenitor cells in the VZ. Again, despite the fact that the mutated Pax6 transcript was expressed throughout its normal domain it was severely downregulated specifically within the domain of ectopic GSH2 expression (Fig. 6I). Furthermore, neurogenin 2 expression was completely extinguished in the dorsolateral cortical VZ where GSH2 is ectopic (Fig. 6J). These results indicate that Pax6 is required for the maintenance of molecular identity within cortical progenitors throughout neurogenesis. Moreover, these findings implicate the ectopic expression of ventral genes (Gsh2, Dlx and Mash1) in the manifestation of cortical defects in Sey/Sey mutants.

As was the case in the Gsh2 mutants, RBP1 (Fig. 6K) and EMX1 (Fig. 6L) expression was not altered in Sey/Sey mutants. These findings indicate that Pax6 and Gsh2 do not regulate all aspects of dorsal-ventral identity but rather a specific subset of developmental regulators, including Mash1, Dlx and neurogenin genes.

The present results indicate that the loss of Pax6 or Gsh2 in their respective mutants leads to an imbalance in Mash1, Dlx and neurogenin gene expression and subsequent misspecification of either cortical or striatal progenitor cells (Fig. 6M). Given this reciprocal regulation, we were interested to study telencephalic development in the absence of both Pax6 and Gsh2. To do this, we generated Sey/Gsh2 double homozygous mutants. The gross morphology of these mutants is similar to Sey/Sey embryos in that they lack eyes and nasal structures (data not shown). At E12.5 in the double mutant, expression of the mutated Pax6 transcript was seen throughout the LGE (Fig. 7A), similar to that in the Gsh2 mutant. Furthermore, the reduction in Pax6 expression observed in the cortical VZ of Sey/Sey mice (Figs 5E, 6I and 7G), appeared restored to wild-type levels in the double mutants (Fig. 7A,H). These findings demonstrate that Gsh2 is required to repress Pax6 gene expression in the LGE and also when ectopic in the Sey/Sey cortical VZ. Expression of the targeted Gsh2 allele in Sey/Gsh2 double mutants appeared rather normal at E12.5 (Fig. 7B). At later stages, weak expression was found in the cortical VZ (data not shown), suggesting that functional GSH2 is required for the full ectopic expression of Gsh2 in the Sey/Sey cortex.

Unlike the case in Gsh2 mutants, at E12.5 in Sey/Gsh2 double homozygous mutants, DLX (data not shown) and MASH1 (Fig. 7C) are expressed throughout the LGE and even into the ventrolateral cortex, similar to that seen in Sey/Sey embryos (Fig. 5H). Furthermore, at E16.5 DLX (Fig. 7F) and MASH1 (data not shown) remain ectopic in progenitors of the double mutant cortex but at a somewhat reduced level than in the Sey/Sey cortex (Fig. 7E). Moreover, proliferation in the double mutant SVZ (as marked by Ki67 expression) appeared to be improved over that in the Gsh2 mutants but clearly not restored to wild-type levels (data not shown).

In addition to the improvements in DLX and MASH1 expression, neurogenin 2 was only found in a small number of LGE progenitors of the Sey/Gsh2 double mutant (Fig. 7D). Furthermore, neurogenin 2 expression in the double mutants was similar to that in Sey/Sey (Figs 5F and 6J), being notably downregulated in the cortex at E12.5 (Fig. 7D) and lost in the dorsolateral cortex by E16.5 (data not shown).

The restoration of molecular identity in the LGE of Sey/Gsh2 double homozygous mutants correlated with significant improvements in striatal development. In the double mutant embryos, the striatal complex was larger than in Gsh2 mutants and the olfactory tubercle (as marked by ISL1) was present in each case (Fig. 8A,B). Thus, ectopic Pax6 in the Gsh2 mutant LGE appears to be largely responsible for the striatal phenotype observed in these mutants. Unlike the progenitors in the LGE, cortical progenitors in the Sey/Gsh2 double mutant remain mis-specified with respect to Mash1, Dlx, neurogenin 1 and neurogenin 2 gene expression. Similar to that observed in the Sey/Sey cortex (Fig. 8E), the number of proliferating cells in the double mutant cortex, as marked by the expression of Ki67, was increased when compared with wild types (Fig. 8D,E). This increase in progenitors resulted in an enlargement of VZ/SVZ in the cortex of both Sey/Gsh2 and Sey/Sey mutants (Fig. 8G,H). Despite these facts, however, development of the cortex in Sey/Gsh2 double mutants was considerably improved over that in Sey/Sey embryos. The intermediate zone and cortical plate, which are severely deficient in Sey/Sey embryos (Fig. 8H and Schmahl et al., 1993; Caric et al., 1997), were restored in the double mutants, similar to wild types (Fig. 8F,G). Moreover, expression of the POU-domain transcription factor SCIP (POU3F1 – Mouse Genome Informatics), which marks migrating cortical neurons (Frantz et al., 1994), was more similar to that in wild types than in Sey/Sey mutants (Fig. 8I-K). Thus, ectopic Gsh2 in the Sey/Sey cortex impedes the normal formation of the cortical plate.

The results of this study demonstrate that Pax6 and Gsh2 genetically oppose each other in the regulation of molecular identity within cortical and striatal progenitors, respectively. In this respect, Gsh2 gene function is required to repress Pax6 expression both in the wild-type LGE and when ectopic in the progenitors of the Sey/Sey cortex. Conversely, Pax6 appears to be necessary for the repression of Gsh2 in cortical progenitors but when ectopic in the LGE, as in the Gsh2 mutant, no effect on the expression of the targeted Gsh2 transcript is observed. This finding suggests that other factors, which are confined to the cortical VZ in the Gsh2 mutants are required to co-operate with Pax6 to repress Gsh2 in cortical progenitors. In addition to their opposing regulation of each other, Pax6 and Gsh2 also reciprocally control the expression of Mash1, Dlx and neurogenin genes in telencephalic progenitors. Pax6 appears to be required to maintain the expression of neurogenin 1 and neurogenin 2, as well as to repress Mash1 and Dlx gene expression in cortical progenitors. Recently, Fode et al. (2000) demonstrated that neurogenins were required to repress Mash1 and Dlx genes in the developing cerebral cortex. It is, therefore, likely that the apparent repression of these genes by Pax6 is actually mediated via the neurogenins (Fig. 9). Gsh2, however, is not required for the expression of Mash1 and Dlx genes or the repression of neurogenin 1 and neurogenin 2. Thus, an essential role for Gsh2 in telencephalic development is to repress Pax6 and its dorsalizing program in the LGE and thereby protect ventral identity (Fig. 9).

As is the case in Sey/Sey mutants, Mash1 and Dlx genes are also found ectopically in the cortex of neurogenin 1 and neurogenin 2 mutants (Fode et al., 2000). The temporal profile of this phenotype, however, is different in the two mutants. While Mash1 and Dlx genes are already ectopic throughout the cortex in the neurogenin 1 and neurogenin 2 mutants at E12.5 (Fode et al., 2000), their misexpression in the Sey/Sey cortex is only beginning at E12.5 but occupies most of the dorsolateral germinal zone by E16.5. This progressive phenotype is likely due, at least in part, to the fact that neurogenins are induced independently of Pax6 and are thus present in the early Sey/Sey cortex but disappear at later stages. It should be mentioned that the progressive spread of ventral gene expression in the Sey/Sey cortex has previously been suggested to result from the increased tangential migration of neurons, taking origin from the ventral telencephalon (Stoykova et al., 1997). In fact, Chapouton et al. (1999) have recently demonstrated that increased numbers of ventrally generated neurons do cross the cortical-striatal boundary in Sey/Sey mutants and take up residence in the cortex, particularly in the ventrolateral regions. Our results, however, demonstrate that in addition to increased tangential migration of ventrally derived neurons, the loss of Pax6 results in a progressive mis-specification of cortical progenitors. In support of this, Fode et al. (2000) showed that the ectopic expression of Dlx1 in the cortex of neurogenin 2 mutants was a direct consequence of the mis-specification of cortical progenitors in the absence of neurogenin function and not due to increased migration of ventrally derived neurons.

The present findings indicate that the lateral telencephalon, including the LGE and dorsolateral cortex, is predisposed to express Mash1 and Dlx genes. It is only through a Pax6 and neurogenin-mediated repression that Mash1 and Dlx genes are absent from most cortical progenitors during neurogenesis. Since Mash1 and Dlx genes are known to be required for the generation of cortical interneurons (Anderson et al., 1997a; Casarosa et al., 1999), this mechanism may represent a way to generate cellular diversity in the cerebral cortex. Although many cortical interneurons are derived from the ventral telencephalon (for a review, see Parnavelas, 2000), at least some have been shown to derive from single progenitors of the cortical VZ that are capable of generating both glutamatergic projection neurons and GABAergic interneurons (Götz et al., 1995). In fact, wild-type cortical progenitors are known to express low but detectable levels of Mash1 (Guillemot et al., 1993). Moreover, in the absence of neurogenin 2 function, ectopic expression of Mash1 and Dlx genes is accompanied by an increase the number of neurons expressing the mRNA for the GABA-synthesizing enzyme glutamic acid decarboxylase, in the mutant cortex (Fode et al., 2000).

Sey/Sey mutants exhibit multiple defects in cortical development (Schmahl et al., 1993), including the disturbed migration of late-born cortical neurons into the developing cortical plate (Caric et al., 1997). This defect appears to be non-cell autonomous, since late-born mutant cells transplanted to the wild-type cortex can migrate and differentiate normally (Caric et al., 1997). The temporal profile of this phenotype correlates well with that of ectopic Gsh2, Mash1 and Dlx genes in the cortical VZ of Sey/Sey mutants, suggesting that the molecular mis-specification of cortical progenitors plays a central role in this. When Gsh2 is removed from the Sey/Sey background (i.e. Sey/Gsh2 double mutants), cortical progenitors remain mis-specified with respect to Mash1, Dlx genes and neurogenin genes, similar to what is seen in Sey/Sey cortical progenitors. Moreover, increased proliferation remains evident in the cortical germinal zone of the double mutants when compared with wild types, which is similar to that in Sey/Sey mutants (present results; Götz et al., 1998; Warren et al., 1999). Nevertheless, the morphology of the double mutant cortex is notably improved over that in the Sey/Sey mutant. The finding that the intermediate zone and cortical plate appear normalized in the Sey/Gsh2 double mutant indicates a dramatic improvement in radial migration of cortical neurons. In support of this notion, cells in the double mutant cortex expressing the POU-domain transcription factor SCIP, which marks neurons migrating through the intermediate zone and into the forming cortical plate (Frantz et al., 1994), were distributed very similar to that in wild-type cortex. Since Mash1 and Dlx genes remain ectopic in the double mutant cortex, this improvement can largely be ascribed to the loss of Gsh2 on the Sey/Sey background. Interestingly, radial glia in the Sey/Sey cortex exhibit cell-autonomous defects in cellular morphology which has been suggested to be the cause of the cortical migration defects observed in the these mutants (Götz et al., 1998). In this respect, ectopic Gsh2 may play a role in the altered differentiation of cortical radial glia observed in Sey/Sey mutants. Therefore, the cortical phenotype in Sey/Sey mutants can be divided into at least two distinct and independent components: one that is Gsh2 independent, which includes the control of progenitor proliferation as well as Mash1, neurogenin and Dlx gene expression, and another that is Gsh2 dependent and involves radial migration of cortical neurons.

Our results demonstrate that the loss of Gsh2 function results in abnormal development of the striatal complex including a reduction in the overall size and a loss of the olfactory tubercle. These defects correlate with the loss of Mash1 and Dlx genes from most of the striatal progenitors at early stages of neurogenesis. Mash1 mutants have recently been shown to exhibit alterations in LGE/striatal differentiation, which include a significant reduction in the size of the olfactory tubercle (Casarosa et al., 1999; Horton et al., 1999). Thus, the olfactory tubercle defect observed in the Gsh2 mutants may, at least partly, be due to loss of Mash1. Double mutants for Dlx1 and Dlx2 also display abnormalities in striatal differentiation (Anderson et al., 1997b). These genes appear to be required for the correct development of the striatal neurons in the matrix compartment.

The mature striatum comprises two anatomically and neurochemically distinct compartments, termed the patch and matrix (for review see, Gerfen, 1990). Neurons generated at early stages of striatal neurogenesis preferentially occupy the patch compartment, while those arising later mainly join the matrix (van der Kooy and Fishell, 1987). Since the misspecification of striatal progenitors in the Gsh2 mutant was most severe at early stages of neurogenesis, the patch compartment would be predicted to the most affected. The situation may, however, be more complex than this, since proliferation in the SVZ of the Gsh2 mutant is also disrupted. It has been suggested that the VZ represents the source of neurons that will comprise the patch compartment, while the SVZ would provide the matrix neurons (van der Kooy and Fishell, 1987; Anderson et al., 1997b). Therefore, it seems likely that both striatal compartments are affected in the Gsh2 mutant.

The fact that early striatal progenitors are most affected in the Gsh2 mutant could be taken to indicate that Gsh2 is only required for these stages of striatal neurogenesis and that other mechanisms regulate later stages. Alternatively, compensatory mechanisms may function in the absence of Gsh2. In this respect, we have observed an expansion in the expression domain of the Gsh2-related, Gsh1 gene (Valerius et al., 1995) at late stages of striatal neurogenesis, from the ventromedial LGE in wild types to encompass the entire mutant LGE VZ (H. T. and K. C., unpublished). This could indicate that the restoration of the striatal germinal zone at late stages results from an expansion of the ventromedial domain of the LGE that is not mis-specified in the Gsh2 mutant. However, the mis-specification of early LGE progenitors is only partial, since the mutated Gsh2 gene and RBP1 are expressed throughout the mutant LGE VZ. Therefore, the restoration could be due to a respecification of the initially mis-specified dorsolateral (i.e. Gsh2-, Rbp1-positive, Mash1-, Dlx-negative) domain.

In addition to the loss of ventral genes in the Gsh2 mutant LGE, the dorsal regulators, Pax6 and the neurogenins are found ectopically. Given that Pax6 and the neurogenins are required for the repression of Mash1 and Dlx genes in cortical progenitors, it is likely that they are responsible for the loss of these ventral genes in the Gsh2 mutant LGE and the resulting striatal phenotype. In support of this, removal of Pax6 from the Gsh2 mutant background (i.e. Sey/Gsh2 double mutants) results in the restoration of correct molecular identity in LGE progenitors and concomitant improvements in the development of the striatal complex, including the reappearance of an olfactory tubercle. In the portion of the Gsh2 mutant LGE that Pax6 is ectopic, proliferation in the underlying SVZ is deficient, suggesting that the Pax6-induced mis-specification of early LGE progenitors not only represses Mash1 and Dlx gene expression but also prevents the formation of a normal SVZ. This is interesting when compared with results from the Sey/Sey cortex, where the loss of Pax6 leads to an increase in progenitor proliferation, particularly in the cortical SVZ (Götz et al., 1998; Warren et al., 1999 and the present results).

The Gsh1/2 homolog, intermediate neuroblasts defective (ind), has recently been cloned in Drosophila and shown to be expressed in selected areas of the fly CNS including the intermediate region of the ventral nerve cord (Weiss et al., 1998). ind mutants show a reduction in the number of neuroblasts in this region of the ventral nerve cord. The remaining neuroblasts, by all markers tested, appear to have acquired a dorsal fate. This phenotype is reminiscent of the mouse Gsh2 mutant phenotype where the progenitors in the LGE (i.e. the intermediate region of the telencephalon) have lost the expression of certain ventral genes and instead express the dorsal genes Pax6, neurogenin 1 and neurogenin 2. These findings therefore indicate that at least some aspects of Gsh/ind function in dorsal-ventral specification of the developing nervous system have been conserved throughout evolution.

We thank A. Björklund, D. J. Epstein, J. Ericson, A. L. Joyner and M. Matise for helpful comments on the manuscript. We gratefully acknowledge H. Edlund and V. van Heyningen for the Sey mice; andD. Anderson, G. Corti, U. Eriksson, F. Guillemot, G. Lemke, G. Panganiban and S. Wilson for probes and antibodies. Thanks to K. Fogelström for excellent technical assistance. Special thanks to P. Emson for help with the generation of the GSH2 antibody. This work was supported by grants from Arbetsmarknadens Försäkringsaktiebolag (AFA), Swedish MRC (12539 and 12196) and NIH (HD29599).