Sonic hedgehog regulates Gli activator and repressor functions with spatial and temporal precision in the mid/hindbrain region

The formation of many organs involves distinct organizing centers acting in concert to pattern the tissue in three dimensions. In the central nervous system (CNS), local organizers pattern different regions along the anteroposterior (AP) and dorsoventral (DV) axes by defining a series of molecular identities that direct differentiation of diverse neuronal cell types (Rubenstein et al.,1994). The mesencephalon (mes) and rhombomere 1 (r1) of the anterior hindbrain are a powerful model system for studying these events, as the key molecules that direct AP and DV patterning are known, and each region forms distinct structures. Dorsally, the mes gives rise to the tectum consisting of the laminated superior and inferior colliculi, while r1 becomes the foliated and layered cerebellum. Ventrally, the mes and r1 are organized into distinct clusters of neurons. The isthmic organizer, positioned at the mes/r1 boundary, directs AP patterning of the mes and r1 via the secreted factor Fgf8 (Wurst and Bally-Cuif,2001; Zervas et al.,2005). DV patterning of the neural tube involves morphogens that are secreted from dorsal and ventral sources, with sonic hedgehog (Shh)expressed ventrally in the floor plate and underlying notochord(Ingham and McMahon, 2001). Although the molecular events underlying AP patterning in the mes/r1 have been well defined (Wurst and Bally-Cuif,2001; Zervas et al.,2005), the extent to which Shh regulates DV patterning in the mes/r1 and the involvement of Shh-downstream effectors is poorly understood.

The few studies to date addressing Shh signaling in the mes/r1 have uncovered a role for Shh in inducing ventral cell types, in addition to a general role in proliferation and cell survival in the mes(Zervas et al., 2005). Miss-expression studies in the chick dorsal mes have demonstrated that Shh can induce in a concentration-dependent manner a series of transcription factors normally found in a nested pattern in the ventral mes(Agarwala et al., 2001). Similarly, in vitro explant studies and ectopic expression of Shh in vivo have shown that Shh can induce ventral mes/r1 cell types, including dopaminergic,serotonergic and motor neurons (Agarwala and Ragsdale, 2002; Fedtsova and Turner, 2001; Ye et al.,1998). Both dopaminergic and serotonergic neurons also require Fgfs for their induction (Ye et al.,1998). Analysis of Shh-null mutant mice has been less informative as Shh is required for survival and proliferation of mes precursors within 1 day (E9.0) of initiation of Shh expression(Agarwala et al., 2001; Chiang et al., 1996; Echelard et al., 1993; Fedtsova and Turner, 2001; Ishibashi and McMahon, 2002). In addition, the remaining mes is completely dorsalized(Fedtsova and Turner, 2001),precluding analysis of potential requirements for Shh signaling at subsequent stages of mes development. Although the only established role for Shh in early patterning of r1 is the induction of ventral serotonergic neurons(Ye et al., 1998), conditional gene inactivation has demonstrated that Shh signaling plays a later role(after E16.5) in regulating granule cell precursor proliferation in the cerebellum (Corrales et al.,2004; Lewis et al.,2004; Corrales et al.,2006). Given the severity of the early mes defects in Shh-null mutants and the largely unresolved role of Shh signaling in early r1 patterning, the precise in vivo requirement for Shh signaling in embryonic mes/r1 development remains to be determined. In particular, it has not been addressed whether Shh signaling plays a role in mes/r1 development after E9.0, and whether and how Shh might regulate patterning of dorsal mes/r1 structures.

The downstream components of Shh signaling have been dissected in detail in the spinal cord and forebrain (Jacob and Briscoe, 2003; Ingham and McMahon, 2001). Shh signaling is transduced through the transmembrane receptors patched (Ptch1) and smoothened (Smo). The inhibition of Smo by Ptch1 is relieved by Shh, thus allowing for transcription of downstream target genes via the Gli zinc-finger transcription factors. In mouse, the three Gli proteins have distinct biochemical functions and in vivo requirements. Whereas Gli1 is largely dispensable for normal murine development, Gli2 and Gli3 mutants die at birth(Bai et al., 2002; Johnson, 1967; Mo et al., 1997; Park et al., 2000). Mouse Gli3 protein is primarily cleaved into an N-terminal repressor form (Gli3R), but Shh signaling counteracts this processing and high levels of Shh signaling can induce a weak Gli3 activator (Bai et al.,2004; Wang et al.,2000). By contrast, Gli2 is efficiently converted into a transcriptional activator (Gli2A) by strong Shh signaling and Gli1 is a constitutive activator. Notably, Gli1 is a transcriptional target of Shh signaling (Bai et al.,2002; Bai et al.,2004), providing a precise readout of Gli2A-mediated Shh signaling. Gli3 transcription also appears to be negatively regulated by Shh signaling, but probably by an indirect mechanism(Marigo et al., 1996). Thus,Shh signaling can be divided into two basic signaling functions: Shh signaling that acts primarily via induction of Gli2A (Gli2A-mediated Shh signaling), and Shh signaling that inhibits the processing of Gli3 into a repressor(Gli3R-mediated Shh signaling).

In the spinal cord, Gli2A is required to induce the most ventral cell types and Gli3R only partially regulates patterning of intermediate regions(Bai and Joyner, 2001; Bai et al., 2004; Chiang et al., 1996; Ding et al., 1998; Matise et al., 1998; Persson et al., 2002). By contrast, in the telencephalon Shh is necessary dorsally and ventrally to generate a Gli3R gradient and Gli2A is not required(Chiang et al., 1996; Fuccillo et al., 2004; Park et al., 2000; Rallu et al., 2002). Given that Gli2A and Gli3R are used to different extents in the spinal cord and forebrain, it is important to address the contribution of Gli2A- and Gli3R-mediated Shh signaling to mes/r1 development. Furthermore, as the temporal requirement for Shh signaling is distinct for ventral verses dorsal forebrain patterning (Fuccillo et al.,2004), it is necessary to explore the temporal contributions of the two signaling functions in the mes/r1.

To determine how and when Gli2A- and Gli3R-mediated Shh signaling regulates mes/r1 development, we analyzed the phenotypes of Shh-, Shh;Gli3 and Gli2-null mutants in detail and compared them with mutants in which Smo or Gli2 were ablated at E9.0 or at E11.5. This allowed us to distinguish the sequential requirements for total(induction of Gli2A and inhibition of Gli3R) verses Gli2A-mediated Shh signaling. Unlike the spinal cord or telencephalon, we found that both Gli2 and Gli3 play crucial roles in mes/r1 development. Strikingly, Shh is required to regulate the level of Gli3R for both patterning of dorsal structures and overall growth of the mes/r1. Furthermore, Gli3R continues to regulate dorsal patterning after E11, at least in part by controlling expression of Fgf8 in the isthmic organizer. By contrast, Gli2A-mediated Shh signaling is required primarily before E11.5 to sequentially induce distinct ventral neurons from medial to lateral. Therefore, a balance between Gli2A-and Gi3R-mediated Shh signaling is instrumental in controlling the size and intricate morphology of all mes/r1-derived structures.

The age of embryos was determined by designating noon of the day a vaginal plug was detected as E0.5. The day of birth was designated as P0. The generation of the Gli2 floxed allele is described elsewhere(Corrales et al., 2006). The other mutant alleles and genotyping are as described: Smo floxed(Long et al., 2001), Smo^recombined (Zhang et al., 2001), R26R lacZ(Soriano, 1999), Shh-null (Chiang et al.,1996), Gli2^zfd-null(Mo et al., 1997), Gli3^Xt-null (Maynard et al., 2002), En1-Cre(Kimmel et al., 2000; Li et al., 2002) and Nestin-Cre allele (Tronche et al., 1999). Gli3^Xt was maintained on a C57/BL6 background, other mouse lines on an outbred Swiss Webster background.

Dissected brains were immersion fixed in 4% paraformaldehyde (PFA) for 30 minutes, cryoprotected in 15% and 30% sucrose and embedded in OCT(Tissue-Tek). β-gal activity was detected in 12 μm frozen sections by incubation in X-gal solution at 37°C overnight. For detailed protocols,see http://saturn.med.nyu.edu/research/dg/joynerlab/protocols.html

Embryos or postnatal brains were fixed in 4% paraformaldehyde or Carnoy,respectively. Paraffin sections (7 μm) were processed for TUNEL assay[ApopTag, Apoptosis detection Kit (Chemicon)], BrdU pulse assays(Graus-Porta et al., 2001),standard antibody staining or RNA in situ hybridization(http://saturn.med.nyu.edu/research/dg/joynerlab/protocols.html). Primary antibodies were: anti-calbindin (1:4000, Sigma/1:2000, Chemicon),anti-BrdU (1:500, Becton Dickinson), anti-Th (1:500, Chemicon) and anti-5-HT(1:500, ImmunoStar). Secondary antibodies were: biotinylated, FITC- or Cy3-conjugated goat-anti-mouse or anti-rabbit (Jackson ImmunoResearch); or Alexa-488-conjugated donkey anti-rabbit (Molecular Probes).

The brain (Shh-null mutants) or mes/r1 (controls) of E12.5 embryos was extracted in RIPA buffer/0.1 mg/ml PMSF. The extract was run on a 4-15%gradient SDS gel (BioRad) and analyzed by western blot(http://saturn.med.nyu.edu/research/dg/joynerlab/protocols.html). The Gli3 antibody was kindly provided by Baolin Wang (Cornell University,Weill Medical College, NY).

Shh is expressed in the mes/r1 starting at E8.0(Echelard et al., 1993) and we found that it is maintained at least up to E18.5 (see Fig. S1 in the supplementary material and Fig. 5A,E), suggesting that Shh signaling could regulate mes/r1 development throughout embryogenesis. To assess the temporal requirements for Shh signaling, we analyzed mes/r1 development in Shh-null mutants and in conditional mutants in which the Shh receptor Smo(Long et al., 2001) was inactivated using two distinct Cre-driver lines. En1-Cre was used to remove Smo specifically in the mes/r1 between E8.5 and E9.0(Kimmel et al., 2000; Li et al., 2002) and Nestin-Cre to remove Smo in the entire neural tube after E11.5 (see Fig. S2 in the supplementary material)(Graus-Porta et al., 2001; Tronche et al., 1999; Corrales et al., 2006).

Consistent with previous studies showing that growth of the mes is impaired by E9.0 in Shh-null mutants(Ishibashi and McMahon, 2002),we found that the size of both the mes and r1 in E12.5 Shh-null embryos was greatly reduced compared with wild type and that the ventral and dorsal neural tube was fused at the isthmic constriction(Fig. 1M,N). By E18.5, all ventral structures were absent in the mutants, and the cerebellum and superior and inferior colliculi (anterior and posterior dorsal midbrain, respectively)were not recognizable by morphology (Fig. 1A,B,G,H). However, we detected calbindin-positive Purkinje cells and Math1-positive granule cell precursors, revealing that a cerebellar structure with a layered cytoarchitecture was generated(Fig. 1H, inset and data not shown).

When Shh signaling was removed in the mes/r1 1 day after the initial onset of Shh expression (Smo^fl/-;En1-Cre mice; referred to as Smo-En1 cko for Smo conditional knock-out driven by En1-Cre), a partial rescue of the Shh null mutant mes/r1 phenotype was observed (compare Fig. 1B,H,N with Fig. 1C,I,O), although the Smo-En1 cko mutants died a few hours after birth. A combination of histological and fate-map analysis of Smo-En1 cko brains revealed that dorsal and ventral mes/r1-derived structures were clearly discernible at E18.5/P0, but severely reduced in size. In addition, the cerebellum was not foliated and the tectum did not appear to be divided into superior and inferior colliculi(Fig. 1A,C,G,I; see Fig. S3A,B in the supplementary material). Consistent with the latter, analysis of Dmbx1, a homeodomain transcription factor expressed specifically in the superficial layer of the superior colliculus(Gogoi et al., 2002) (see Fig. S3C in the supplementary material) showed superficial expression throughout the remaining tectum in Smo-En1 cko mutants, indicating that the inferior colliculus was absent (Fig. S3D). By contrast, Dmbx1 was not detected in Shh-null mutants (data not shown).

Inactivation of Shh signaling 2 days later in Smo^fl/-;Nestin-Cre mice (referred to as Smo-Nescko) resulted in a much milder phenotype and the mutants lived up to 4 weeks(data not shown) (Machold et al.,2003). The main mes/r1 phenotype of Smo-Nes cko mutants was a truncated inferior colliculus and a small, unfoliated cerebellum(Fig. 1D,J). The lack of foliation in the cerebellum of the two Smo cko mutants is probably due to the late role of Shh signaling in granule cell precursor proliferation(Corrales et al., 2004; Lewis et al., 2004).

To determine when the phenotypes first arise in the Smo cko embryos, we analyzed sagittal sections of E9.5 (Smo-En1 cko) and E12.5 (both Smo cko) wild-type and mutant embryos. In Smo-En1 cko embryos the size of the mes vesicle was slightly reduced within 12 hours of inactivation of Shh signaling(Fig. 3; data not shown). By E12.5, the mes/r1 area of Smo-En1 cko embryos was reduced severely along both the DV and AP axes, and ventral and dorsal aspects of the isthmus were closely opposed (Fig. 1M,O). In addition, dorsal r1 (the cerebellar anlage) was thinner than in wild-type embryos, and the dorsal posterior mes was truncated. Consistent with the phenotype of Smo-Nes cko embryos at P0, the mes/r1 phenotype at E12.5 consisted of a slight reduction of dorsal tissue,primarily in the most medial regions (Fig. 1P). By E16.5, there was a clear truncation of the posterior midbrain and a slight reduction in the size of the cerebellum, but the phenotype was milder than in Smo-En1 cko embryos (data not shown). In summary, these data demonstrate that Shh signaling is crucial for development of mes/r1 structures throughout the entire DV axis from E8.0 to after E11.5.

We next determined the temporal requirement for Shh in generating specific ventral neuronal cell types. In Shh-null mutant embryos, no dopaminergic [tyrosine hydroxylase (Th) positive] or serotonergic[5-hydroxytryptamine (5-HT) positive] neurons were detected (data not shown),consistent with an early loss of all ventral mes/r1 progenitors. Interestingly, in Smo-En1 cko embryos in which Shh signaling is removed at E9.0, dopaminergic and serotonergic neurons were generated, but were severely depleted (Fig. 2A-D). Moreover, the removal of Shh signaling 2 days later in Smo-Nes cko mutants did not affect the induction of these neurons(Fig. 2E,F). Thus, Shh signaling is required for the specification of dopaminergic and serotonergic neurons before and shortly after E9.0, but not after E11.5.

To determine whether more laterally generated neurons are also differentially affected by sequential inactivation of Shh signaling, we analyzed expression of the transcription factors Islet1 (Isl1) (ventral motoneurons) and Nkx2.2 (ventral-laterally derived neurons). In E18.5 Shh-null mutant embryos (data not shown) and in E10.5 and E12.5 Smo-En1 cko embryos, neither transcription factor was detected in the ventral mes/r1 (Fig. 2I-L and data not shown). Dorsal Isl1-positive cells were, however, maintained in Smo-En1 cko mutants (Fig. 2J,L). By comparison, E12.5 Smo-Nes cko mutants had both ventral Isl1- and Nkx2.2-expressing neurons, although they were slightly reduced in number compared with wild-type embryos(Fig. 2M-P). Thus, neuronal subpopulations derived from distinct ventral to lateral domains in the mes/r1 are dependent on Shh signaling in a sequential manner.

To determine the extent to which Gli2A-mediated Shh signaling is used downstream of Shh for development of ventral and dorsal mes/r1-derived structures, we analyzed the requirement for Gli2A at different time points using Gli2-null mutants and a Gli2 conditional allele(Corrales et al., 2006), in which Gli2 was inactivated at E9.0 (Gli2-En1 cko) or E11.5(Gli2-Nes cko). In E12.5 and E18.5 Gli2-null mutant brains,the ventral midbrain and r1 were reduced in size, but not absent(Fig. 1E,K,Q). Consistent with this, a small number of dopaminergic and serotonergic neurons are present in E12.5 Gli2-null mutants (Matise et al., 1998). Furthermore, although the dorsal midbrain was thinner probably because of hydrocephaly at E18.5(Corrales et al., 2004; Palma and Ruiz i Altaba,2004), there was no obvious truncation of the inferior colliculus(Fig. 1K,Q). The E18.5 cerebellum was unfoliated and smaller than normal, owing to the requirement for Gli2 in granule cell precursor proliferation after E16.5(Corrales et al., 2004). Comparison of the phenotypes of wild-type, Gli2 and Shh-null mutants shows that the contribution of Gli2A signaling to overall embryonic mes/r1 development is restricted to ventral regions(Fig. 1G,H,K,M-O).

Removal of Gli2 at E9.0 in Gli2-En1 cko mutants resulted in a surprisingly mild phenotype. The size of the ventral mes/r1 appeared grossly normal at E12.5 and E18.5 (Fig. 1L,R). Dopaminergic and serotonergic neurons were present in reduced numbers at E18.5, but were not as severely depleted as in Smo-En1 cko mutants (Fig. 2C,D,G,H). By contrast, the more laterally derived Isl1-and Nkx2.2-expressing neurons were not present in Gli2-En1cko mutants (data not shown). As expected, the cerebellar phenotype at E18.5 was similar to Gli2-null mutants and the tectum appeared normal(Fig. 1K,L)(Corrales et al., 2004). To address whether Gli2-mediated signaling plays a role after E11.5 in mes/r1 development, we analyzed sections of Gli2-Nes cko brains. Histological and marker analysis for Isl1 and Nkx2.2 at E12.5 or Th at P0 did not show any obvious reduction in the number of ventral neurons (data not shown). Thus, Gli2A-mediated Shh signaling is only required for induction of ventral mes/r1 structures and primarily before E11.5.

To address the cellular mechanism underlying the rapid tissue loss when all Shh signaling is removed after E9.0, we analyzed Smo-cko mutants for cell death and proliferation, as both are affected in Shh null mutants by E9.0 (Ishibashi and McMahon,2002). Analysis of apoptosis using a TUNEL assay revealed a massive increase in cell death in Smo-En1 cko mutants at E9.5, and to a lesser extent at E10.5, throughout the DV axis except in the floor plate. Along the AP axis, cell death was highest near the isthmus(Fig. 3A-D; data not shown). By contrast, no obvious difference in the percentage of proliferating cells in the remaining mes/r1 ventricular zone was observed with BrdU pulse labeling of E9.5 Smo-En1 cko embryos (data not shown). Consistent with the near normal size of Smo-Nes cko brains, no changes in proliferation or apoptosis were detected at E12.5 in these mutants(Fig. 3E,F; data not shown). These results demonstrate that Shh signaling continues to regulate expansion of the mes/r1 between E9.0 and E11.0, but primarily through prevention of cell death.

As Gli2A-mediated Shh signaling is not required for mes/r1 growth or dorsal patterning, it is likely that precise regulation of the level of Gli3R is crucial for both processes. If this is the case, then the absence of Gli3R in Gli3^Xt-null mutants(Hui and Joyner, 1993; Johnson, 1967) should result in cerebellum and dorsal midbrain defects. Indeed, the dorsal mes area was found to be truncated and the isthmus and r1 enlarged in E12.5 Gli3^Xt-null mutants (compare Fig. 4B with Fig. 1M). At E18.5, the remaining tectum and isthmus were thicker than in wild type and the cerebellum was abnormally shaped in mutants without exencephaly (compare Fig. 4A with Fig. 1G).

We next tested whether the Shh-null phenotype could be partially rescued by removing one copy of Gli3 in Shh null mutants. Indeed, we found that Shh^-/-;Gli3^Xt/+ mutant embryos had a considerably milder phenotype than Shh-null mutants(compare Fig. 4C,D with Fig. 1H,N). At E12.5, the mes vesicle was almost normal in size and ventral structures were present, but the dorsal and ventral neural tube was fused at the isthmus(Fig. 4D). In contrast to Shh-null mutants, the midbrain was recognizable in E18.5 double mutants and appeared to be divided into inferior and superior colliculi dorsally (Fig. 4C). A more normally structured, but still abnormally shaped cerebellum with calbindin-positive Purkinje cells was also present(Fig. 4C, inset).

To determine whether Gli3R levels are increased relative to full-length Gli3 in the absence of Shh signaling, we performed western blot analysis for Gli3 protein in wild-type and Shh mutant brain extracts(Fig. 4E)(Wang et al., 2000). In E12.5 Shh-null mutant brains, the N-terminal cleavage product was clearly increased compared with wild type and the full-length product was reduced(Fig. 4E). A slight reduction in the Gli3 full-length form was even observed in the mes/r1 of Shh^+/- embryos. Hence, the upregulation of Gli3R levels in the absence of Shh signaling is indeed the cause for the lack of overall growth of the mes/r1 and abnormal patterning of dorsal structures.

It has previously been shown that markers for dorsal neurons and progenitors (Pax7) are extended ventrally in the mes of Shh null mutants by E11.5 (Fedtsova and Turner,2001). To determine the onset of this phenotype and the role of Shh and Gli2 in both setting up and maintaining ventral and dorsal mes/r1 gene expression over time, we analyzed the expression of marker genes that define four DV domains in the mes and r1 from E8.5 to E12.5: Shh expression in the ventral midline, Gli1 expression demarcating the adjacent cells responding to Gli2A-mediated Shh signaling, Gli3 expression comprising the lateral and dorsal mes/r1, and Pax7 expression defining the dorsal plate (Fig. 5A,F,K,P). In E8.5 and E9.5 Shh-null mutants, Shh and Gli1 expression was absent in the mes/r1 (data not shown). The induction of Shh expression by Shh signaling from the underlying notochord in the mes/r1 must be dependent on Gli2A-mediated signaling, as Shh is not detected in the mes/r1 of Gli2-null mutants (Fig. 5C)(Matise et al., 1998). Interestingly, expression of Shh in the ventral midline was not lost when Shh signaling was ablated at E9.0 (Smo-En1 or Gli2-En1cko mutants) or E11.5 (Smo-Nes and Gli2-Nes cko mutants)(Fig. 5B,D,E; data not shown). Thus, Gli2A-mediated Shh signaling is required only to initiate, but not to maintain, Shh expression.

By contrast, Gli1 was initiated but not maintained in Smo-En1 and Smo-Nes cko mutants once Smo was inactivated (Fig. 5F,G,J; data not shown). Unlike Shh, Gli1 was expressed weakly in Gli2-null mutants (Fig. 5H) (Bai et al.,2004) and also in Gli2-En1 cko and Gli2-Nes cko mutants (Fig. 5I; data not shown). This residual expression of Gli1 reflects a weak activator function of Gli3 in cells close to the source of Shh(Bai et al., 2004). The source of Shh protein in Gli2 null mutants is probably the ventral midline of the forebrain (Fig. 5C,inset) (Matise et al., 1998),whereas Shh is provided by the ventral midline of the mes/r1 in Gli2cko mutants (Fig. 5D). In summary, a low level of Gli1 is induced in the absence of Gli2 in the mes/r1, whereas a high level requires Gli2A-mediated Shh signaling.

Although it has previously been shown that Pax7 expression is expanded into the ventral mes in Shh-null mutants (analysis at E11.5)(Fedtsova and Turner, 2001),it is not known whether Pax7 downregulation depends on Gli2,nor whether Shh signaling is required to sustain this regulation at later stages. Interestingly, whereas in the Shh-null and Smo-En1cko mutants the Pax7 expression domain was expanded ventrally, the Pax7-positive domain was not extended into ventral areas in Gli2 null or cko mutants (Fig. 5P-S; data not shown). This suggests that in Gli2mutants, the residual GliA-mediated Shh signaling through Gli3A is sufficient to restrict Pax7 expression dorsally(Fig. 5H,I,R,S). Alternatively,only inhibition of Gli3R by Shh is required to repress Pax7 ventrally.

In contrast to Pax7, Gli3 expression was extended ventrally in the absence of all (Shh null, Smo-En1 cko mutants) and Gli2A-mediated (Gli2 null, Gli2-En1 cko mutants) Shh signaling at E9.5 (Fig. 5K-N). However, when Shh signaling was inactivated after E11.5 in the Smo-Nes cko or Gli2-Nes cko mutants, Gli3 and Pax7 expression continued to be excluded from the ventral mes/r1 similar to wild-type embryos (Fig. 5K,O,P,T; data not shown). Thus, Shh signaling is required only before E11.5 to repress Gli3 and Pax7 in the ventral mes/r1.

The defects in the inferior colliculus and cerebellum in Smo-En1cko and Smo-Nes cko mutants at P0(Fig. 1I,J) are similar to defects seen when the isthmic organizer is disrupted(Liu and Joyner, 2001; Zervas et al., 2005). Indeed,the expression domain of Fgf8 was shown to be smaller in Shh-null mutants and larger in Gli3-null mutants by E10.5(Aoto et al., 2002). This could, however, be secondary to loss or expansion of the isthmus in Shh- or Gli3-null mutants, respectively(Fig. 1M,N; Fig. 4B)(Ishibashi and McMahon, 2002). To explore further how and when Shh and Gli3 regulate isthmic gene expression,we analyzed the expression of a number of genes involved in isthmic organizer function in Smo and Gli2 mutant embryos. In Smo-En1cko embryos, Fgf8 expression in the isthmus was found to be slightly reduced at E9.5 (data not shown). By E10.5, Fgf8 expression was severely reduced and the domain was shifted posteriorly. Importantly, the relative reduction in Fgf8 expression was greater than the overall size reduction of the isthmic region (Fig. 6A,B). Consistent with the reduction of Fgf8, expression of sprouty 1 (Spry1) (a direct target of Fgf8)(Liu and Joyner, 2001), was restricted to an area encompassing only the isthmus and the anterior part of r1 (Fig. 6C,D). Similarly, the expression of Fgf17, another Fgf contributing to AP patterning of the mes/r1 (Xu et al., 2000) was reduced in the Smo-En1 cko mutants (data not shown). Although Fgf signaling was reduced, expression of Wnt1 in the mes was expanded posteriorly into the isthmus of Smo-En1 cko mutants(Fig. 6E,F). Further supporting a direct role for Shh signaling in maintaining Fgf expression in the isthmus, we observed a slight reduction in the expression domains of Fgf8,Fgf17 and Spr1 in E12.5 Smo-Nes cko embryos(Fig. 6G-J; data not shown). Importantly, in Gli2-null and Gli2-En1 cko mutants no obvious alterations in Fgf8 or Spry1 expression were observed at E12.5 (data not shown). Therefore, Shh signaling, through regulation of Gli3R, plays an important role in maintaining normal Fgf8 expression in the isthmic organizer from E9.5-E12.5.

Comparative analysis of the developmental progression of mes/r1 defects in mice with conditional and null mutations in components of the Shh signaling pathway has allowed us to determine the temporal contributions of Gli2A and Gli3R to Shh signaling in mes and r1 development(Fig. 7). In general, high levels of Shh induce Gli2A-mediated Shh signaling and this regulates ventral cell type specification, whereas negative regulation of Gli3R dictates patterning of dorsal structures as well as overall growth in the mes/r1.

The combined results of our analysis and previous studies show that before E9.0, Gli2A-mediated Shh signaling from the notochord induces Shh and Gli1 transcription in the ventral midline(Fig. 5C; data not shown)(Bai et al., 2004; Matise et al., 1998). Simultaneously, Shh signaling is required to downregulate ventral Gli3 and Pax7 expression at the transcriptional level(Fig. 5M; data not shown)(Fedtsova and Turner, 2001). We now demonstrate that Gli3, but not Pax7 downregulation requires a high level of Gli2A-mediated Shh signaling. Although, dorsally, Shh signaling does not regulate Gli3 transcription, a gradient of Shh determines the degree to which full-length Gli3 is processed into Gli3R. Whereas the level of Gli3R is involved in patterning of the tectum and cerebellum,Gli2A-mediated signaling induces a first wave of ventral neurons medially. Finally, Shh signaling regulates cell survival and proliferation before E9.0(Ishibashi and McMahon,2002).

By analyzing conditional mutants, we have determined that between E9.0 and E11.5, Shh signaling is no longer required for maintaining Shhexpression in the ventral midline of the mes/r1, but continues to be required to maintain Gli1 expression and a Pax7/Gli3 negative domain ventrolaterally. Furthermore, it is Gli2A-mediated Shh signaling that induces a high level of Gli1 expression and ventral downregulation of Gli3, but not Pax7 transcription. In addition,Gli2A-mediated Shh signaling continues to promote generation of ventral neurons both medially and ventrolaterally. Simultaneously, regulation of Gli3R levels by Shh signaling antagonizes cell death both dorsally and ventrally,and also controls the normal development of dorsal mes/r1 structures, most prominently the inferior colliculus and cerebellum.

After E11.5, we found that Shh signaling is still required to induce Gli1 expression, but not to restrict Pax7 and Gli3expression dorsally. Similarly, Gli2A-mediated Shh signaling is no longer required to generate the ventral neuronal subpopulations we analyzed. Shh signaling via inhibition of the Gli3R does, however, continue to influence development of the inferior colliculus and cerebellar anlage, but ceases to be necessary for promoting cell survival.

Our analysis demonstrates that Gli2A- and Gli3R-mediated Shh signaling is used in a distinct manner in the mes/r1 compared with the forebrain or spinal cord. In the forebrain, although Shh signaling regulates Gli3R production throughout the DV axis, it is only required before E9.0 for normal development of dorsal structures (Fuccillo et al.,2004; Park et al.,2000; Rallu et al.,2002). Furthermore, Gli2A-mediated Shh signaling has only a very minor, if any, role in patterning the telencephalon(Fuccillo et al., 2004; Park et al., 2000; Rallu et al., 2002). In the spinal cord, both Gli2A- and Gli3R-mediated Shh signaling is required, but is restricted to the ventral and intermediate zones, respectively(Bai et al., 2004; Jacob and Briscoe, 2003). By contrast, we have found in the mes/r1, inhibition of Gli3R by Shh plays a sustained role in dorsal mes/r1 structures and Gli2A-mediated Shh signaling induces only ventral cell types.

We identified abnormal patterning of dorsal mes/r1-derived structures in Smo-En1 and Smo-Nes cko mutants at P0, but not in Gli2 mutants. In addition, we found that Shh signaling continues to be required for overall cell survival between E9.0 and E11.0, and that this function does not depend on Gli2. Consistent with a primary role for Gli3R-mediated Shh signaling in mes/r1 growth and dorsal patterning, we showed that the level of Gli3R is increased in Shh-null mutants and that both growth and dorsal patterning are partially rescued in Shh-null mutants when the level of Gli3 is reduced. Thus, Gli3R is probably the main downstream effector for these two processes and continues to regulate dorsal patterning after E11.5. However, as Fgf8 signaling is downregulated in the absence of Shh signaling and it has previously been shown that complete removal of Fgf8 in the isthmic organizer results in massive cell death(Chi et al., 2003), we cannot exclude that Gli3R acts at least in part indirectly through regulation of Fgf8. A major effect of decreased Fgf signaling on cell death is unlikely, however, as apoptosis occurred primarily at E9.5 in Smo-En1cko mutants, a stage when we did not observe a severe reduction in Fgf8 expression.

Interestingly, even in the absence of a direct source of Shh within the mes/r1 of Gli2-null mutants, owing to the lack of Shh expression in the ventral midline, Shh secreted from the ventral forebrain is sufficient to promote dorsal development and general growth. It is perhaps surprising that Shh signals can normally reach the dorsal mes/r1, especially after E11.5 when the mes/r1 has undergone substantial growth. Of likely relevance, the dorsal defects in Smo cko mutants are most prominent in the region close to the isthmus. It is therefore possible that, as the neural tube is smaller at the isthmic constriction, Shh protein can travel the shorter distance to reach dorsal cells.

Interestingly, we found that the time of dependence of ventral progenitors on Shh signaling correlates with their medial to lateral position in the mes/r1 ventricular zone. The Shh-expressing ventral midline cells, as well as some of the most ventrally induced neurons in the mes/r1 (dopaminergic and serotonergic) are generated in Smo-En1 and Gli2-En1 cko mutants, demonstrating that they require Shh signaling primarily before E9.0. By contrast, defects in dopaminergic or serotonergic neurons were not observed in Smo-Nes or Gli2-Nes cko mutants, suggesting that these neurons are independent of Shh signaling by E11.5. In addition, the more ventrolaterally derived Nkx2.2- and Isl1-positive neurons are not generated in Smo-En1 or Gli2-En1 cko mutants showing they require Shh signaling after E9.0. As these neurons are only slightly reduced in Smo-Nes cko embryos, Shh is required primarily between E9.0 and E11.5 for their generation. The temporal requirement for Shh signaling in induction of distinct ventral to lateral-derived neurons is consistent with genetic fate mapping studies in mice that showed that progenitors of dopaminergic and serotonergic neurons respond to GliA-mediated Shh signaling (express Gli1) between E8.0 and E10.0, whereas motoneuron progenitors respond between E9.0 and E10.0(Zervas et al., 2004) (S.B. and Emilie Dambroise, unpublished). Similarly, in vitro studies demonstrated that the generation of dopaminergic and serotonergic neurons in mes/r1 explants already requires Shh and Fgf8 at E8.0(Ye et al., 1998).

A comparison of the phenotypes of Smo and Gli2 cko mutants shows that in addition to Gli2A, Gli3A probably contributes to the induction of some ventral cell types. In particular, the number of dopaminergic and serotonergic neurons is reduced to a greater extent in Smo-En1 cko embryos compared with Gli2-En1 cko mutants. Consistent with Gli3A contributing to the generation of ventral neurons, Gli3 expression is extended ventrally and Gli1 is expressed weakly in the ventral mes/r1 of Gli2-En1 cko mutants. Alternatively,or in addition, as Shh signaling is maintained in Gli2 cko mutants,the level of Gli3R ventrally should be less in Gli2 cko than in Smo cko mutants, and this might be sufficient to allow generation of some ventral neurons.

The truncation of the inferior colliculus and the smaller cerebellum we observed in Smo cko mutants at late gestation are similar to phenotypes seen when Fgf levels are reduced in Fgf17^-/- or Fgf17^-/-; Fgf8^+/- mutants(Xu et al., 2000). The dorsal phenotype could therefore be an indirect consequence of a decrease in Fgf8 and/or Fgf17. Indeed Fgf8 was previously found to be reduced in Shh-null mutants(Aoto et al., 2002). Furthermore, the finding that Fgf8 expression is partially rescued in Shh;Gli3 double mutants and increased in Gli3-null mutants suggested that Gli3R-mediated Shh signaling regulates Fgf8expression. Our studies of Smo-cko mutants argue that Shh regulates isthmic organizer gene expression directly, rather than indirectly through loss or expansion of the isthmus in Shh- or Gli3-null mutants, respectively. We show that Shh signaling continues to be required for Fgf8 as well as Fgf17 expression after E9.0, and that the relative reduction in Fgf expression is more severe than the loss of isthmic tissue in Smo cko mutants. Furthermore, as we found that Fgf8 expression is normal in Gli2 mutants, Shh regulates Fgf8 only through antagonizing Gli3R.

Regulation of Fgf8 expression by Shh provides a mechanism by which the two key organizers in mes/r1 development can function in unison. Interestingly, interdependence of organizers that express Shh and Fgf8 has been described in the limb and forebrain(Tickle, 2003; Ohkubo et al., 2002). Our studies thus provide further support that cross-regulation of organizers represents a general mechanism to coordinate patterning along different axes during the formation of morphologically complex structures.

Our studies have revealed sequential roles for Shh in DV patterning of the entire mes/r1, and identified the processes requiring induction of Gli2/3A versus downregulation of Gli3R. We demonstrate that GliA-mediated Shh signaling (mainly via Gli2A) is limited to the induction of ventral neurons,and occurs in a medial to lateral sequential manner before E11.5. By contrast,Shh signaling through inhibition of Gli3R is required for overall expansion of the mes/r1 through regulation of cell survival up to E11.0, and for patterning of dorsal structures even after E11.5. Furthermore, the level of Gli3R is crucial in determining the size of the Fgf8 expression domain. Sustained Shh signaling thus is necessary not only for DV but also AP patterning of the mes/r1. The formation of all mes/r1-derived structures is therefore intimately linked to the level and extent of Shh signaling.

We thank Andrew McMahon for the Smo floxed mice; Daniel Stephens and Gina Rocco for technical assistance, Emilie Dambroise for fate mapping data and Gunda Schwaninger, Roy Sillitoe and Gordon Fishell for critical reading of the manuscript. We are especially grateful to Mark Zervas for his insightful comments and discussions. S.B. was supported by a post-doctoral fellowship from the DFG. This research was supported by a grant from the NICHD. A.L.J. is an HHMI investigator.