Dorsoventral patterning is established in the telencephalon of mutants lacking both Gli3 and Hedgehog signaling

The forebrain arises from the most anterior part of the neural plate and is subsequently divided into the telencephalon and the diencephalon(Fishell, 1997;Rubenstein and Beachy, 1998;Shimamura et al., 1995). After neural tube closure (mouse E9), proliferation within the ventral regions of the telencephalon results in the emergence of three eminences: first, the medial ganglionic eminence (MGE) and later, the lateral and caudal ganglionic eminences (LGE and CGE, respectively)(Fishell, 1997) (S. Nery, G. F. and J. G. C., unpublished). Collectively, these eminences make up the anlage of much of the telencephalic basal ganglia. These structures are thought to give rise to the globus pallidus, the striatum, and large populations of oligodendrocytes and cortical interneurons(Wichterle et al., 2001) (S. Nery, G. F. and J. G. C., unpublished). All three eminences appear to be absent in sonic hedgehog (Shh) null mutants(Chiang et al., 1996) (J. G. C., unpublished).

Shh has been shown to play a fundamental role in the establishment of ventral identity throughout the neural tube (reviewed byEricsson et al., 1995a;Litingtung and Chiang, 2000b;Patten and Placzek, 2000). As a result, much work has been performed to elucidate the mechanisms by which Shh generates cell diversity within the ventral neural tube, especially in the spinal cord. A considerable body of evidence favors the idea that Shh acts as a morphogen, forming a gradient within the ventral neural tube to which cells respond in a concentration-dependent manner (reviewed byBriscoe and Ericsson, 2001;Jessell, 2000). Collectively,in vitro explant studies using recombinant Shh(Ericsson et al., 1996;Ericsson et al., 1997;Roelink et al., 1995) and in vivo electroporation studies with Shh(Agarwala et al., 2001) or a dominant-negative form of patched (Briscoe et al., 2001) strongly support this hypothesis in the midbrain and spinal cord, respectively.

Although Shh is also clearly required in forebrain regions(Chiang et al., 1996;Ericsson et al., 1995b), to date less effort has been focused on its role in patterning the telencephalon. Similar to the spinal cord, embryos that lack Shh fail to form ventral telencephalic structures and to express corresponding markers, while dorsal marker expression is expanded ventrally(Chiang et al., 1996). Furthermore, ectopic expression of Shh is sufficient to induce ventral telencephalic markers in dorsal regions of the telencephalon both in vitro (Ericsson et al., 1995b;Kohtz et al., 1998;Shimamura and Rubenstein,1997) and in vivo (Barth and Wilson, 1995; Corbin et al.,2000; Gaiano et al.,1999; Hauptmann and Gerster,1996). However, when and how Shh is required for ventral telencephalic patterning and whether it acts to differentially specify the MGE, the LGE and the CGE is not presently clear. Beginning around E9.0, Shh is expressed within the ventral telencephalon, primarily in the MGE, preoptic area and prospective amygdala (Nery et al., 2001; Sussel et al.,1999). Notably, the specific loss of Shh expression within the telencephalon, which occurs in Nkx2.1 (Titf1 -Mouse Genome Informatics) (Sussel et al.,1999) and BF1 (Foxg1 - Mouse Genome Informatics)(Huh et al., 1999) null mutants, does not recapitulate the phenotype observed in Shh mutants. This suggests that earlier expression of Shh, probably from non-neural tissues such as the dorsal foregut (i.e. prechordal plate)(Shimamura and Rubenstein,1997) or Hensen's node(Gunhaga et al., 2000), must be responsible for the more extensive telencephalic defects seen inShh mutants.

The Ci/Gli family of zinc-finger transcription factors has been implicated as required transducers of the hedgehog signaling pathway (reviewed byAza-Blanc and Kornberg, 1999;Ruiz i Altaba et al., 2002). Gli proteins, like Ci, display both activator and repressor activities that are regulated post-translationally(Aza-Blanc et al., 1997;Ingham and McMahon, 2001). Different regions of the Gli proteins encode distinct functions: N-terminal regions encode a repressor function, whereas C-terminal regions are required for positive activity. Analysis of mice mutants for each of the three murine Gli genes has shown that Gli3 is the only Gli gene that plays a significant role in patterning the telencephalon. Specifically, mice that lackGli1 and Gli2 show minor defects in telencephalic patterning(Park et al., 2000), whereasGli3 mutant mice have strong abnormalities in the dorsal telencephalon (Grove et al.,1998; Theil et al.,1999; Tole et al.,2000). Gli3 protein appears to function primarily as a repressor(reviewed by Ingham and McMahon,2001) and its activity seems to be negatively regulated by Shh(Marigo et al., 1996;Wang et al., 2000). Shh signaling, both in vitro and in the limb, has been shown to inhibit Gli3 processing into an N-terminal fragment that carries a repressor activity. In agreement with this, analysis of mice lacking both Shh andGli3 revealed that, in the spinal cord, Shh is required to inhibit Gli3 function in order to form motoneurons(Litingtung and Chiang,2000a).

In the present study, we have used both gain- and loss-of-function methods to study the importance of hedgehog signaling in general and Shh in particular in the establishment of dorsoventral patterning in the telencephalon. We find that virally mediated ectopic sonic hedgehog expression induces different regional markers, depending on the dorsoventral or anteroposterior position of the infection. Conversely, some aspects of ventral telencephalic patterning persist in the absence of Shh. In addition, we observe that dorsoventral patterning in the telencephalon appears to be largely restored inShh/Gli3 and Smo/Gli3 double homozygous mutants. Our data reveal that Shh and Gli3 antagonize each other's function in patterning the telencephalon. Furthermore, our findings suggest that a hedgehog-independent mechanism, such as possibly the BMP-signaling pathway(Liem et al., 2000), is capable of acting in parallel to Shh in establishing dorsoventral pattern in the telencephalon.

Mice used in these studies were maintained according to protocols approved by the Institutional Animal Care and Use Committee at New York University School of Medicine. For staging of embryos, noon on the day the vaginal plug was considered as embryonic day 0.5 (E0.5). Virus preparation and ultrasound surgery were both performed as previously described(Gaiano et al., 1999). Stocks of Shh-expressing or ActSmo-expressing viruses were injected at titers of 5×10⁷ and 5×10⁸ cfu/ml respectively. Mouse embryos were injected from E8.5 to E10.5 and sacrificed 4 days later (E12.5 to E14.5). Viral expression usually begins 12 hours after infection (E9.0 at the earliest in our case).

Shh (Chiang et al.,1996) and Gli3 (Xt^J from Jackson Laboratory) mutant mice were maintained on a C57/B16 and C3HeB/FeJ background,respectively. Different mutant combinations were generated by intercrossingShh^+/-;Gli3^+/- orSmo^+/-;Gli3^+/- heterozygous animals. Embryos were genotyped by PCR as described previously(Litingtung and Chiang, 2000a;Maynard et al., 2002;Zhang et al., 2001).

Whole heads or embryos were fixed at 4°C for 1-4 hours in 4%paraformaldehyde, then cryoprotected overnight in 30% sucrose in PBS, embedded in HistoPrep (Fisher Scientific) and frozen. All tissue was sectioned serially at 12 to 20 μm and processed for immunohistochemistry or RNA in situ hybridization.

Infected cells were identified by the expression of the alkaline phosphatase reporter (PLAP) as described previously(Gaiano et al., 1999). Analysis of gene induction was performed by RNA in situ hybridization on adjacent 20 μm sections. Note that the thickness of the sections and the incompatibility of histochemistry and in situ hybridization on the same sections sometimes made it hard to correlate the infected cells with the cells that display ectopic expression (e.g. Fig. 2J-K).

Immunofluorescence was performed as described previously(Corbin et al., 2000). The following primary antibodies were used: mouse anti-Pax6 (1:1000, gift of A. Kawakami) and rabbit anti-Gsh2 (1:3500, gift of K. Campbell).

In situ hybridization was performed as described previously(Schaeren-Wiemers and Gerfin-Moser,1993; Wilkinson and Nieto,1993) using non-radioactive digoxigenin-labeled probes forNkx2.1 (Shimamura et al.,1995), Dlx2 (Porteus et al., 1991), Gsh2(Hsieh-Li et al., 1995),Mash1 (Ascl1 - Mouse Genome Informatics)(Guillemot and Joyner, 1993),Gli1, Gli3 (Hui et al.,1994) and Ptch(Goodrich et al., 1997).

Dlx2 mRNA expression was also determined by X-Gal staining of embryos from a strain of mice in which the taulacZ gene has been targeted to the Dlx2 locus(Corbin et al., 2000) (S. Nery, G. F. and J. G. C., unpublished). X-Gal staining was performed as described previously (Corbin et al.,2000).

To better understand the mechanisms by which Shh patterns the telencephalon, we chose to investigate the effects of ectopic Shh signaling in vivo. We used a gain-of-function approach that combines high-titer retroviral vectors and ultrasound imaging (Gaiano et al., 1999). Retroviruses expressing the full-length Shh protein were injected into the embryonic forebrain at different stages of development,the earliest of which (mouse E8.5) preceded the establishment of overt dorsoventral patterning in the telencephalon. The injection of Shh virus at these early time points had a dramatic effect on the morphology of the brains,with the vesicles being substantially enlarged compared with wild-type embryos(Fig. 1E,F). These morphological defects may be the result of an abnormal proliferation induced by Shh. Indeed, the role of Shh in cell proliferation within the nervous system has been shown in the cerebellum(Dahmane and Ruiz-i-Altaba,1999; Wechsler-Reya and Scott,1999), the retina (Jensen and Wallace, 1997) and the dorsal neural tube(Rowitch et al., 1999). The morphological abnormalities in Shh-infected brains made any analysis of dorsoventral patterning in these animals difficult. Thus, we constructed a retrovirus encoding a constitutively active form of the Smoothened protein(ActSmo) (Hynes et al., 2000;Xie et al., 1998). Smoothened is a transmembrane protein that, in the absence of Shh, is repressed by the Shh receptor Patched (Ptch) (reviewed byIngham and McMahon, 2001;Kalderon, 2000;Murone et al., 1999). This repression is thought to be relieved by the binding of Shh to Patched,allowing Smoothened to activate downstream elements of the Shh signaling pathway cell-autonomously. Activated Smoothened has been shown to mimic the patterning effects of ectopic Shh in the dorsal neural tube(Hynes et al., 2000). Furthermore, the activity of ActSmo is likely to be cell-autonomous(Hynes et al., 2000), allowing us to study the effects of ectopic hedgehog signaling without perturbing the local environment of the infected cells. We found that infections with a virus expressing ActSmo did not result in the gross morphological deformations observed with Shh (Fig. 1C,D)but was indistinguishable from Shh in its ability to alter regional gene expression.

ActSmo- and Shh-expressing viruses were injected at various times ranging from E8.5 to E10.5 and embryos were sacrificed 4 days later for analysis(E12.5 to E14.5). Notably, we observed that the ectopic expression of ActSmo(or Shh itself) resulted in the same modifications of telencephalic patterning regardless of the age of the infection (data not shown). This suggests that the competence of the telencephalic tissue to respond to ectopic induction of the Shh pathway is not temporally regulated within this window of embryonic development.

Ventral patterning by Shh in the spinal cord is in part mediated through the inhibition of dorsally expressed transcription factors(Briscoe et al., 2000;Ericsson et al., 1997). Similarly in the telencephalon, in vitro exposure to Shh can inhibit the expression of dorsal markers (Kohtz et al., 1998). Moreover, dorsal gene expression expands ventrally in the absence of Shh (Chiang et al.,1996). We found that the expression of Pax6, a dorsal telencephalic marker, is indeed repressed in the infected cells(Fig. 2A-C), in accordance with previous results in spinal cord (Ericsson et al., 1996). A similar result was obtained for the dorsally expressed proneural genes such as Ngn2, Math2 and NeuroD(data not shown).

We next sought to determine the extent to which ventral markers characteristic of the MGE (Nkx2.1) and MGE/LGE (pan-ventral genesGsh2, Dlx2) could be induced ectopically by virally mediated activation of Shh signaling. We observed that the identity of the markers induced, and therefore the cell types, was mainly dependent on the position of the infection. Along the dorsoventral axis, the induction of Nkx2.1expression in response to ectopic Shh signaling was restricted to the LGE and lateral neocortex (Fig. 2G,H),whereas Gsh2 and Dlx2 could be induced throughout the neocortex (Fig. 2I,J and data not shown). Interestingly, we noted that the induction of pan-ventral gene expression (Gsh2, Dlx2) was weaker in the medial regions of the telencephalon (Fig. 2I,J). Sagittal sections of ActSmo infected brains revealed the rostrocaudal extent of ventral gene induction: while Nkx2.1 induction was restricted to the LGE and septum (Fig. 2D,Fand data not shown), pan-ventral gene induction within the cortex was only excluded from the more posterior regions(Fig. 2D,E). These differences could be accounted for by regional variations in the ability of the telencephalon to activate downstream components of the Shh pathway. To address this issue we analyzed patched (Ptch) and Gli1 gene expression, which are believed to be generic downstream response genes in the Shh signaling pathway. We found that the expression of either of these genes was upregulated at all positions along the dorsoventral or anteroposterior axis (Fig. 2K-O). Based on these criteria, this suggests that all regions of the telencephalon are equally competent to respond to Shh signaling.

These gain-of-function experiments show that regional identity can be altered by Shh signaling through the early phases of neurogenesis. However, we found that this plasticity is somewhat limited: the ventral identities that Shh and ActSmo can induce are dependent on the dorsoventral or anteroposterior position of the mis-expression. This suggests that an intrinsic pattern, which is either Shh dependent or independent, has been established within the telencephalon by the time the Shh pathway is activated ectopically (E9.0 at the earliest, i.e. 12 hours after infection).

Previous analysis of Shh-null animals revealed that, in addition to the lack of any obvious ventral telencephalic structures, the domain of dorsal gene expression spreads ventrally to include the whole telencephalon(Chiang et al., 1996;Pabst et al., 2000). However,when we investigated the expression of pan-ventral genes in these mutants, we were surprised to observe that some ventral telencephalic patterning remains in the absence of Shh (Fig. 3).Dlx2 expression was observed in the remaining telencephalon of E10.5 to E12.5 Shh mutants as visualized by in situ hybridization or by using a Dlx2^taulacZ allele(Fig. 3A-D). Similarly,Gsh2 expression (another pan-ventral gene) persisted in Shhmutants (Fig. 3E,F) (H. Toresson and K. Campbell, personal communication). Notably, in these mutants the domain of expression of both Dlx2 and Gsh2 was restricted to a region adjacent to the ventral midline. As Shhmutants are extremely dysmorphic, it is impossible to know with certainty which ventral telencephalic domains persist in these mutants. Nonetheless, the fact that Nkx2.1 expression was missing in Shh^-/-embryos (data not shown) (Pabst et al.,2000) while Dlx2 and Gsh2 expression persisted suggests that the most ventral domain (i.e. MGE) is absent inShh^-/- mutants, while the lateral regions (i.e. LGE) are displaced ventrally. In this regard, this result is likely to be analogous to that in spinal cord, where the laterally positioned V0 and V1 interneurons,which are characteristic of lateral cell fates, persist in Shh^-/-embryos but are displaced toward the midline(Pierani et al., 1999). The observation that ventral gene expression can occur in the absence of Shh suggests the existence of a Shh-independent pathway. This prompted us to examine other genes that could play a role in establishing dorsoventral patterning in the telencephalon.

Shh has been shown to be required to antagonize Gli3 function in ventral regions of the spinal cord (Litingtung and Chiang, 2000a). In the telencephalon, Gli3 is expressed at high levels in dorsal and lateral regions (cortex and LGE) and at low levels ventrally (in the MGE). Interestingly, extra-toes mice(Xt^J), which carry a deletion encompassing theGli3 gene (Buscher and Ruther,1998; Hui and Joyner,1993; Schimmang et al.,1992; Vortkamp et al.,1992), show defects opposite to those observed in Shhmutants. In Gli3 mutants the morphology of the forebrain is perturbed dorsally: the cortex is highly reduced in size, and the hippocampus and the choroid plexus are absent (Grove et al.,1998; Theil et al.,1999; Tole et al.,2000). Furthermore, the expression of a number of pan-ventral genes, such as, Dlx2 (Fig. 3G-H) (Tole et al.,2000), Gsh2 (Fig. 3I-J) and the bHLH gene Mash1 (data not shown), spreads precociously into cortical areas. Even though ventral markers were ectopically expressed, the cortex of Gli3 mutants retained its dorsal character,as shown by the persistence of dorsal gene expression, including Pax6(Fig. 3K,L)(Tole et al., 2000). Interestingly, Gsh2 and Pax6 proteins appeared to be co-expressed in the cortex of Gli3 mutants, as shown by double immunostaining(Fig. 3L). Gsh2 andPax6 expression domains generally abut at the boundary between the cortex and the LGE (i.e. the corticostriatal boundary)(Fig. 3K) and these proteins are normally mutually repressive (Toresson et al., 2000; Yun et al.,2001). Our results showed that Gsh2 expression no longer respects this boundary in the telencephalon of Gli3 mutants,suggesting that the repression of Gsh2 by Pax6 is dependent on Gli3.

In order to analyze the extent to which dorsoventral patterning can occur in the absence of Shh and Gli3, we analyzed the telencephalic phenotype of Shh/Gli3 compound mutants. Interestingly,Shh^-/-;Gli3^+/- embryos displayed a remarkable rescue of brain morphology when compared with Shh^-/-embryos. Although the telencephalon of Shh^-/- embryos at E12.5 was greatly reduced in size and composed of a single vesicle fused at the midline (Fig. 4B,B′)(Chiang et al., 1996),Shh^-/-;Gli3^+/- embryos had a much bigger telencephalon in which two vesicles were clearly discernable(Fig. 4C,C′, arrows). Moreover, these embryos had a reduced proboscis (arrowheads) and two relatively well-formed eyes, albeit fused together and located at the midline(Fig. 4D-S, arrows).

As these embryos showed a significant morphological rescue of theShh-null phenotype, we analyzed in detail the status of dorsoventral patterning in the telencephalon. We observed that Dlx2 expression,which was shifted either ventrally or dorsally in Shh^-/-and Gli3^-/- mutants, respectively, was restored inShh^-/-;Gli3^+/- embryos to a level and distribution which closely resembled the wild-type pattern(Fig. 4D,E). A similar rescue was observed for other pan-ventral genes such as Gsh2 andMash1 (Fig. 4J-M). The restoration of the expression of pan-ventral genes to their normal regions was further demonstrated by the existence of a sharp boundary between theGsh2 and Pax6 expression domains, similar to what is observed at the corticostriatal boundary of wild-type brains(Fig. 4N,O). Notably, while the telencephalic morphology of the Shh^-/-;Gli3^+/-embryos was affected, a thickening in the lateral wall of the telencephalon was observed ventrally to the Pax6 expression domain in a region where pan-ventral genes were strongly expressed(Fig. 4D-S; arrowheads). Thus,it seems that the LGE is properly specified in these mutants.

To further investigate to what extent dorsoventral patterning was restored in Shh^-/-;Gli3^+/- embryos, we looked at the expression of the homeobox gene Nkx2.1, which is missing inShh^-/- embryos. Surprisingly, we observed that low levels of Nkx2.1 expression were rescued in these embryos(Fig. 4F,G). In these animalsNkx2.1 expression was specifically located in the most ventral part of the telencephalon and was nested within the domains of expression ofDlx2 (Fig. 3H,I),Gsh2 and Mash1. Thus, although not very prominent, it appeared that a small MGE-like structure can also form in these embryos in absence of Shh (Fig. 4D-S,asterisks). The high levels of recovery observed in Shh mutants where only one copy of Gli3 gene is removed prompted us to investigate the extent of recovery in animals lacking both Shh and Gli3 gene function.

Telencephalic analysis of Shh/Gli3 double homozygous mutants was extremely difficult as a result of a high rate of exencephaly. Most of theShh/Gli3 double homozygous mutants analyzed were exencephalic(11/12=91%): the neural tube failed to close at various levels of the forebrain or midbrain and a massive overgrowth of the tissue disrupted telencephalic morphology. We found thatShh^-/-;Gli3^-/- embryos had a higher proportion of exencephaly than did Gli3^-/- (2/12=16.6%) orGli3^-/-;Shh^+/- (10/32=31%) embryos. Although exencephalic mutants could not be analyzed for dorsal patterning defects,ventral patterning was analyzed in parallel to non-exencephalic mutants. As the exencephalic defects were more severe at E12.5 than at E10.5, we chose to analyze the double homozygous mutants at E10.5.

In a Shh^-/-;Gli3^-/- double mutant embryo that was not exencephalic, the telencephalon appeared to have a relatively normal morphology, although dorsal midline structures were missing(Fig. 5; arrows). The level and distribution of expression of pan-ventral genes such as Dlx2 andGsh2 was similar in Shh^-/-;Gli3^-/-embryos and in their wild-type littermates(Fig. 5A-D). Remarkably, in contrast to Shh^-/-;Gli3^+/- andShh^-/- embryos, double homozygous mutants had a complete rescue of Nkx2.1 expression (Fig. 5E,F). Similar to wild-type animals, Nkx2.1 expression was observed in an area nested within the wider expression domain ofDlx2 (Fig. 5G,H,asterisk) and Gsh2 (Fig. 5C,D). This nested expression pattern suggested that both the MGE and the LGE were specified in Shh^-/-;Gli3^-/-mutants. However, in the absence of a more refined analysis of gene expression patterns within the ventral telencephalon, we cannot exclude the possibility that some aspects of dorsoventral patterning (such as the ventral midline) are not restored in these mutants. Nonetheless, our data suggest that a key role of Shh in patterning the ventral telencephalon is to inhibit Gli3function, thereby indirectly allowing ventral gene expression. Complementary to the restoration of ventral patterning defects observed inShh^-/- mutants, the dorsal telencephalic defects present in Gli3^-/- mutants were also rescued inShh^-/-;Gli3^-/- double mutants. This argues for a requirement for Gli3 in antagonizing Shh function in the dorsal telencephalon, a role apparently not played by Gli3 in the spinal cord(Litingtung and Chiang, 2001a).

The persistence of dorsoventral patterning in the absence of Shhand Gli3 gene function could be the result of compensation by other members of the hedgehog family or the actions of a hedgehog-independent pathway. To differentiate between these possibilities, we analyzed ventral telencephalic gene expression in Smoothened/Gli3 double mutants, in which all hedgehog signaling is abrogated(Zhang et al., 2001). Although, as in the Shh/Gli3 mutants, these animals suffer from a high rate of exencephaly, ventral aspects of the Smo/Gli3telencephalon can still be analyzed. In these mutants both pan-ventral markers such as Dlx2, as well as the MGE-specific marker Nkx2.1 were expressed in their normal distribution(Fig. 5M,N). Furthermore, the fact that Nkx2.1 expression was nested within the Dlx2expression domain suggested that ventral patterning is largely rescued in these compound mutants. Thus, our results reveal that, with the exception of the dorsal midline, a hedgehog-independent signaling pathway is able to establish at least basic elements of dorsoventral patterning in the telencephalon in the absence of both Gli3 and Shh or Smogene function.

To determine whether the restoration of dorsoventral patterning inShh^-/- mutants when Gli3 is removed(Shh^-/-;Gli3^-/-) was due to an activation of downstream elements of the Shh pathway independent of Shh protein, we examined the levels of expression of Gli1 and Ptch, two targets of Shh signaling. Gli1 and Ptch are normally expressed in the ventral telencephalon in close proximity to the sites of Shh expression(Platt et al., 1997).Gli1 expression in the telencephalon is restricted to a small region at the border of the MGE and the LGE (Fig. 4P, Fig. 5I;brackets). As in Shh^-/- mutants (data not shown) this expression was absent in Shh^-/-;Gli3^+/- andShh^-/-;Gli3^-/- embryos(Fig. 4Q,Fig. 5J), consistent with the suggestion that Gli1 activation is strictly dependent on Shh. In wild-type animals Ptch is expressed in a broader region thanGli1, encompassing the entire extent of the MGE(Fig. 4R,Fig. 5K; brackets). WhereasPtch expression is absent in Shh^-/- andShh^-/-;Gli3^+/- mutants (data not shown andFig. 4S), low levels ofPtch expression were detected throughout the telencephalon inShh^-/-;Gli3^-/- mutants, with higher levels ventrally (Fig. 5L, brackets). Higher levels of expression ventrally were also observed in double homozygous mutants that were exencephalic (data not shown). This suggests thatPtch expression is dependent on the removal of Gli3 repressor activity in ventral areas and that the loss of Ptc1 expression observed in Shh^-/- mutants may be due to a lack of inhibition of Gli3 function.

In this paper, we present evidence for the existence of a pathway that can establish elements of dorsoventral patterning in the telencephalon in the absence of both hedgehog signaling and Gli3 gene function. We find that ectopic activation of the Shh pathway at E8.5/E9.0 results in ventralization of the telencephalon. Importantly, it appears that the ventral identity induced upon ectopic activation of the Shh pathway is determined by the dorsoventral position of Shh misexpression. Second, we find that some remnants of ventral gene expression persist in Shh^-/-mutant animals and, in a complementary fashion, these same ventral genes are expanded in Gli3^-/- mutants. Third, analysis ofShh^-/-;Gli3^-/- andSmo^-/-;Gli3^-/- mutants shows that aspects of dorsoventral pattern are established in the telencephalon relatively normally in the absence of hedgehog signaling and Gli3 gene function. Nevertheless, we cannot rule out that a more refined analysis of ventral patterning in the telencephalon in these compound mutants will not reveal further defects. Regardless of this caveat, this analysis suggests that the primary role of Shh in the telencephalon is to repress Gli3 function ventrally, and that the primary requirement for Gli3 is to repress Shh function in dorsal areas. The notable exception to this relationship is in the dorsal midline, which is disrupted in Shh, Gli3 and Shh/Gli3double homozygous mutants. Indeed, the more penetrant exencephalic phenotype in Shh^-/-;Gli3^-/- double mutants argues that these genes may actually cooperate in this region.

In the present study, we undertook an in vivo gain-of-function approach to address the role of Shh in telencephalic patterning. By infecting the mouse telencephalon at different time points (E8.5 to E10.5), we showed that the differential induction of MGE and LGE/MGE markers by Shh signaling was determined by the intrinsic character of the infected tissues, within both the dorsoventral and anteroposterior axes. Although it could be argued that the failure to induce Nkx2.1 in dorsal regions of the telencephalon orGsh2 in dorsomedial telencephalon was the result of our viral vectors producing insufficient levels of Shh signaling, results from our previous studies argue against this (Kohtz et al.,1998). Indeed, Nkx2.1 expression could not be induced in vitro by Shh in E11.5 rat dorsal telencephalic explants (E9.5 in mouse), even when they were exposed to extremely high levels of recombinant protein (i.e. 300 nM). Furthermore, the differential induction we observed with our viral vectors did not appear to be the result of a differential activation of the Shh pathway in dorsal and lateral regions, as Gli1 and Ptchcould be induced ectopically in the telencephalon, regardless of where or when the infection occurred. Thus, our gain-of-function data suggest that by E8.5/E9.0 the telencephalon has been patterned along both the dorsoventral and anteroposterior axis. However, these experiments cannot distinguish whether this apparent prepattern is the result of hedgehog independent signaling or patterning by Shh prior to E9.0. In support of the latter suggestion, work by the Edlund laboratory suggests that the telencephalic expression ofNkx2.1 results from Shh signaling in the node(Gunhaga et al., 2000). Similarly, we found that rat explants at headfold stage (E9.5 in rat, E8.0 in mouse) could uniformly induce Nkx2.1 expression in response to recombinant Shh protein (Kohtz et al.,1998), suggesting that the restriction observed in this paper may not exist in slightly younger embryos (before E9.0).

Even if Shh normally acts to prepattern the telencephalon prior to E9.0,our results show that some ventral pattern is established in the telencephalon in absence of Shh. Specifically, we found that Dlx2 and Gsh2expression within the ventral telencephalon is reduced but not absent inShh^-/- mutant embryos. This result is reminiscent of the persistence of V0 and V1 cell fates in the spinal cord ofShh^-/- mutants(Pierani et al., 1999). Given that in some contexts members of the hedgehog family can partially compensate for the loss of Shh (Pathi et al., 2001), perhaps the persistence of ventrolateral markers indicates the presence of other hedgehog ligands. Indeed, Indian and sonic hedgehog are both expressed in the early foregut(Bitgood and McMahon, 1995;Ingham and McMahon, 2001). Unfortunately, this question cannot be addressed directly asSmo^-/- mutants die before any telencephalic patterning is molecularly evident (Zhang et al.,2001). However, based on the analysis of Smo/Gli3 double homozygous mutants (see below), such compensation appears unlikely.

To futher investigate the cross-repressive interaction between Shh and Gli3 in the telencephalon, we analyzed several Shh and Gli3mutant combinations. The development of the telencephalon is grossly abnormal in Shh^-/- (Chiang et al., 1996) and to a lesser extent in Gli3^-/-mutants (Grove et al., 1998;Theil et al., 1999). We found that telencephalic morphology is largely restored inShh^-/-;Gli3^+/- mutants. Consistent with their rescued morphology, most ventrolateral patterning is established inShh mutants lacking a single copy of Gli3(Shh^-/-;Gli3^+/-). Furthermore, barring the loss of dorsal midline structures, mutants lacking both Shh andGli3 function possess sets of markers characteristic of MGE, LGE and cortex, showing that all these structures are at least partially specified in the absence of Shh and Gli3. The same result was obtained inSmo^-/-;Gli3^-/- mutants in which all hedgehog signaling is abolished. Therefore, many aspects of dorsoventral telencephalic patterning can be established in the complete absence of hedgehog provided that Gli3 gene function is also abolished.

Despite this, the interplay between Shh and Gli3 is crucial to the normal positioning of the different domains of ventral telencephalic gene expression. For example, our data suggests that MGE and LGE markers are differentially sensitive to the dose of Gli3 protein or activity. Although the expression of pan-ventral genes characteristic of the LGE is fully restored inShh^-/- animals with one copy of Gli3 removed, the MGE marker Nkx2.1 is only rescued in the complete absence ofGli3 gene function. Consistent with this, unlike more broadly expressed ventral markers, Nkx2.1 cannot be induced by ectopic Shh signaling in the cortex, where Gli3 is highly expressed. Similarly,Nkx2.1 expression does not spread dorsally inGli3^-/- mutants, whereas pan-ventral gene expression does. One explanation for these findings may be that Gli2, which is expressed in a pattern similar to Gli3 in the telencephalon,possesses weaker repressor activity than Gli3 and acts in a complementary fashion. In this scenario, Gli2 prevents the expansion of Nkx2.1 expression in Gli3^-/- mutants but is insufficient to repress pan-ventral genes, such as Dlx2 andMash1. Indeed, the greater sensitivity of MGE markers to repression by Gli repressors may be one mechanism that underlies the differential specification of the MGE and LGE.

In the limb, the processing of Gli3 into an N-terminal repressor form is negatively regulated by Shh (Wang et al.,2000). As a result, Shh and the N-terminal Gli3 repressor are thought to form opposing gradients in the limb, akin to that suggested by our genetic analysis of the telencephalon. Our results also in many respects parallel the phenotype found in the spinal cord of Shh/Gli3 double mutants (Litingtung and Chiang,2000a). Similar to the telencephalon, ventrolateral cell fates in the spinal cord, including motoneurons and V0-V2 interneurons, were restored to wild-type levels in Shh/Gli3 double homozygous mutants. However,unlike the telencephalon, the ventral-most fates fail to form in the spinal cord of these same animals (as indicated by the loss of Nkx2.2 at the floorplate). Moreover, only partial rescue of ventrolateral genes expression was observed in the spinal cord of these mutants, while ventrolateral telencephalic gene expression appeared normal inShh^-/-;Gli3^+/- mutants. Furthermore,in the telencephalon the antagonism between Shh and Gli3 is reciprocal. By contrast, there is no evidence that Gli3 may antagonize Shh function in the spinal cord, as ventral gene expression does not expand dorsally in this tissue in the absence of Gli3 gene function.

In light of the Shh/Gli3 and Smo/Gli3 double mutants phenotypes, it is worth considering how other signaling pathways might contribute to dorsoventral telencephalic patterning. It is possible that hedgehog-dependent and -independent pathways act in parallel and are functionally redundant, each able to specify ventral, lateral and dorsal cell fates in isolation. When hedgehog signaling is missing, as it is inShh^-/-;Gli3^-/- andSmo^-/-;Gli3^-/- mutants, such redundancy would account for the observed rescue of dorsoventral patterning(Fig. 6). The use of two pathways that complement each other is reminiscent of the mechanism of anterior patterning in Drosophila. In flies, opposing anterior and posterior gradients specify the segmented body pattern. Maternal nanos protein acts as a repressor of hunchback in the posterior region, allowing abdominal patterning to occur; and the absence of both genes results in normal embryos(Hulskamp et al., 1989;Irish et al., 1989;Struhl, 1989). In parallel,bicoid specifies anterior patterning and hunchback can substitute for it in thorax and abdomen (Simpson-Brose et al.,1994; Wimmer et al.,2000).

Alternatively, we cannot formally exclude the possibility that the dorsoventral patterning of the telencephalon is initially established by a hedgehog-independent pathway and that the role of Shh in patterning the telencephalon is to subsequently maintain or refine the organization of the telencephalon. Perhaps Shh and Gli3, through their antagonistic interactions,act to stabilize dorsoventral telencephalic domains during mid-neurogenesis. This might occur by regulating the growth of the progenitor pools, by maintaining a pre-established pattern or by refining the regional boundaries between discrete territories, such as the MGE, LGE and cortex.

The identity of the signal(s) that lead to a normal dorsoventral patterning in the absence of Shh and Gli3 remains unknown. One possibility is that the dorsoventral patterning observed in Shh/Gli3double mutants is the result of BMP signaling, which in spinal cord acts to induce dorsal and repress ventral gene expression. Recent studies in the spinal cord have also shown that BMPs antagonizes Shh signaling and that this inhibition may be crucial to the establishment of dorsal identities(Liem et al., 2000). It seems unlikely that the late dorsal expression of BMPs contributes to hedgehog-independent patterning as their expression requires Gli3(Theil et al., 1999). However,earlier BMP7 expression in the prechordal plate or BMP4 expression in presomitic mesoderm might be playing a role(Dale et al., 1997;Dale et al., 1999), as might Nodal signaling (Rohr et al.,2001). Another alternative comes from studies in spinal cord that have implicated the retinoid signaling pathway in the generation of ventrolateral progenitors (Pierani et al.,1999). In the telencephalon, markers of retinoid synthesis are expressed in the LGE and the developing striatum(Li et al., 2000;Toresson et al., 1999) and retinoid signaling regulates striatal neuronal differentiation(Toresson et al., 1999). However, this expression occurs too late to play a role in the initial patterning of the telencephalon. If retinoids do act in the establishment of dorsoventral pattern in the telencephalon, a more likely source is the lateral cranial mesoderm, which expresses high levels of retinoids and is proximal to the telencephalon during the headfold stage ∼E7.5/E8.0(LaMantia et al., 1993). A third candidate is the Wnt pathway. Recent work has implicated an important role for Wnt inhibitors in the specification of the head (dickkopf)(Niehrs et al., 2001) as well as the telencephalon (masterblind)(Heisenberg et al., 2001;van de Water et al., 2001). Furthermore, a novel secreted Frizzled-related protein expressed in the anterior neural plate at the junction between neural and non-neural ectoderm is involved in promoting telencephalic development within the forebrain territory (Houart et al.,2002). Prior to neural tube closure, the dorsoventral axis of the telencephalon seems to be roughly translated into an anteroposterior axis(Shimamura et al., 1995). Thus, it is worth considering that, in addition to specifying the telencephalon as a whole, graded inhibition of Wnt signaling may also act to establish dorsoventral identity in the telencephalon.

Regardless of the pathway, it is clear that a complete appreciation of regional patterning within the telencephalon will require more than understanding the hedgehog signaling pathway. Determining the nature of these signals and how they act to complement the hedgehog pathway in patterning the telencephalon will no doubt prove interesting.

We thank Susana Nery, Heather Mason, Mike Rutlin, Alex Joyner, Alex Schier and Yorick Gitton for critical reading of this manuscript. We also thank the following people for the gifts of mice, probes or reagents: H. Westphal and C. Chiang (Shh mutant mice), J. Rubenstein (Dlx2), F. Guillemot(Mash1, Ngn2, Math2, NeuroD), F. Savage (actSmo M2), A. Joyner(Nkx2.1, Gli1, Gli3, Ptc1), A. Kawakami (anti-Pax6), and K. Campbell(anti-Gsh2). We are grateful to Steve Wilson and Kenny Campbell for sharing unpublished data, as well as Petra Kraus, who first observed the residualDlx2 expression in Shh-null mutants. We also acknowledge the contributions of Mike Rutlin and Yuan Yuan Huang for their technical assistance, as well as Jill McMahon for the breeding and genotyping of theGli3/Smoothened compound mutants. This work was supported by a NIH grant (NS39007), a March of Dimes basic research grant and a Children's Brain Tumor Foundation grant to G. F.; by postdoctoral grants from l'Association pour la Recherche Contre le Cancer to M. R.; by an American Cancer Society grant to N. G. (PF4473); and by a NIH grant to J. C. (NS10962-01) and R. M.(1F32NS42525-01).