Gli3 coordinates three-dimensional patterning and growth of the tectum and cerebellum by integrating Shh and Fgf8 signaling

A fundamental question in developmental neuroscience is how the simple mammalian neural tube is transformed into the complex and intricate structures that constitute the adult brain. Patterning of tissues in three dimensions requires precise coordination of a number of basic developmental processes including cell proliferation, specification and migration. An ideal system in which to study patterning events is the brain region consisting of the mesencephalon (mes) and rhombomere 1 (r1). The mes and r1 are the primordia for the tectum [superior colliculus (SC) and inferior colliculus (IC)] and cerebellum dorsally, respectively, and for the tegmentum ventrally. Organizing centers that govern mes/r1 growth and patterning along the dorsal-ventral (DV)and anterior-posterior (AP) axes have been identified. AP patterning of the mes/r1 is mediated by fibroblast growth factor 8 (Fgf8), which is secreted by the isthmic organizer located at the mes/r1 boundary(Wurst and Bally-Cuif, 2001; Zervas et al., 2005). Sonic hedgehog (Shh), expressed in the notochord and the ventral midline (floor plate) of the neural tube, controls specification of ventral cell types in the tegmentum and formation of normal dorsal structures, as well as cell proliferation and survival (Agarwala et al., 2001; Bayly et al.,2007; Blaess et al.,2006; Corrales et al.,2006; Fedtsova and Turner,2001; Ishibashi and McMahon,2002). It remains to be addressed how these two signaling centers are integrated to control growth and patterning in three dimensions.

The zinc-finger transcription factor Gli3 influences multiple signaling pathways in the mes/r1. Gli3 is an essential downstream component of the Shh signaling pathway, and an increase in Gli3 repressor (Gli3R) levels is the primary cause of the dorsal mes/r1 defects in Shh signaling mutants(Blaess et al., 2006). Furthermore, Gli3 plays a role in regulating the Fgf8 expression domain at the mes/r1 boundary (Aoto et al.,2002). Therefore, Gli3 could play a key role in coordinating three-dimensional patterning of the mes/r1 by affecting these two signaling pathways.

Only one mutation in mouse that is thought to represent a Gli3-null allele has been characterized. The lethal Gli3 extratoe(Xt) mutation causes multiple defects, including a background-dependent occurrence of exencephaly(Hui and Joyner, 1993; Maynard et al., 2002). In non-excencephalic embryos, however, tectum- and cerebellum-like structures develop, but are morphologically highly abnormal(Aoto et al., 2002; Blaess et al., 2006). The analysis of Gli3^Xt/Xt mutants has so far provided little insight into Gli3 function in the mes/r1 and it remains unclear whether the defects result from changes in the expression of key DV patterning genes,abnormal growth and/or changes in Fgf8 signaling in the isthmic organizer(Aoto et al., 2002).

Gli3 transcription and protein activity are regulated at many levels by Shh signaling. In the absence of Shh activity, Gli3 is processed into an N-terminal repressor form that suppresses Shh target genes(Hu et al., 2006; Wang et al., 2000). Shh signaling attenuates the level of Gli3R, and full-length Gli3 can act as a weak activator (Gli3A) when the concentration of Shh is high(Bai et al., 2004; Dai et al., 1999; Wang et al., 2007). Gli3,however, is also transcriptionally downregulated in cells receiving high levels of Shh. Consequently, Gli3 expression, which is initially found throughout the neural plate, is restricted to the intermediate and dorsal neural tube after E8.5 (Bai et al.,2002; Blaess et al.,2006; Hui et al.,1994; Marigo et al.,1996).

Activation of Shh target genes is primarily mediated by Gli1 and Gli2(Fuccillo et al., 2006; Jacob and Briscoe, 2003). Gli2 requires Shh signaling to act as an activator (Gli2A), whereas Gli1 is a transcriptional target of Gli2/3A and is itself a constitutive activator(Fuccillo et al., 2006; Jacob and Briscoe, 2003). We recently showed that Gli2A-mediated Shh signaling is the key regulator of the initial specification of ventral neurons in the embryonic mes/r1 before E11,and of granule cell precursor proliferation in the postnatal cerebellum(Blaess et al., 2006; Corrales et al., 2006; Corrales et al., 2004). It is unclear, however, whether Gli3A is also required for the development of these two regions.

Analysis of the severely abnormal Gli3^Xt/Xt null mutants can only provide insight into the first crucial early developmental function of Gli3. To study the underlying mechanisms of Gli3 function we combined the analysis of null and time-specific conditional mutants by generating a Gli3 conditional allele and inactivating Gli3in the mes/r1 at either E9.0 or E11.5. We demonstrate that Gli3 is required to pattern dorsal mes/r1 into distinct structures before E9.0,continues to regulate the growth of the tectum and the cerebellum as well as cerebellar foliation between E9.0 and E11.0, and plays a role beyond E11.0 in regulating growth of the isthmus, SC and IC. We further show that the role of Gli3 in isthmic and cerebellar development, but not tectal patterning, is largely mediated through the repression of Fgf8 expression in r1.

A 5.8 kb PstI-MfeI genomic fragment of Gli3containing intron 8 was inserted into a targeting vector upstream of a neo cassette flanked with Frt sites, with one loxP site 3′ to the neo cassette and a thymidine kinase (TK) cassette. Another loxP site containing an EcoRV site was inserted into a NheI site within intron 9 in a 4.8 kb MfeI-KpnI genomic fragment of Gli3. This fragment was inserted 3′ to the loxP site in the neo-TK vector.

Three targeted W4 (Auerbach et al.,2000) ES cell clones were identified(Matise et al., 2000) by Southern blot analysis using SphI and a 1 kb external probe 3′of the SphI site, or using KpnI and an internal probe containing neo. ES cell chimeras were generated through injection of C57BL/6 blastocysts (Skirball Transgenic Facility)(Papaioannou and Johnson,2000). Chimeras were bred to C57BL/6 mice and heterozygous Gli3^flox-neo/+ offspring to SV129 ACTB-Flpe mice(Rodriguez et al., 2000) to produce Gli3^flox/+ heterozygotes lacking neo. Gli3^rec/+ mice carrying a loxP-mediated deletion were intercrossed or crossed with Gli3^Xt/+ mice(Hui and Joyner, 1993). Both Gli3^rec/rec and Gli3^rec/Xt E18.5 embryos displayed the same brain and limb phenotypes described in Gli3^Xt/Xt embryos (see Fig. S1D-F in the supplementary material) (Hui and Joyner,1993). Other phenotypes observed in Gli3^Xt/Xtembryos, such as lethality at birth and exencephaly in some mutants, were also observed (data not shown). The Gli3^flox allele was genotyped using the following primers: S1,5′-CTGGATGAACCAAGCTTTCCATC-3′ and AS3,5′-CTGCTCAGTGCTCTGGGCTCC-3′. For detecting the recombined allele,primers were S1 (as above) and AS2,5′-CAGTAGTAGCCTGGTTACAG-3′.

Other alleles were genotyped as described: Cre, Smo^flox,Smo^rec (Blaess et al.,2006), Fgf8 null (Chi et al., 2003) and Gli3^Xt null(Maynard et al., 2002). Gli3^Xt/+ mice were maintained on a C57/BL6 background; all other mouse lines were maintained on an outbred Swiss Webster background. Noon of the day a vaginal plug was observed was designated as E0.5; the day of birth was designated as P0.

Embryos/brains were fixed in 4% paraformaldehyde (PFA) at 4°C,processed for paraffin- or cryosectioning and sectioned at 7-12 μm. Antibody staining, BrdU labeling and RNA in situ hybridization were performed using standard methods. Primary antibodies: rabbit anti-TH (Chemicon, 1:500);rabbit anti-Ki67 (NovoCastra, 1:500); mouse anti-BrdU (BD-Bioscience, 1:100);rabbit anti-phosphohistone H3 (PH3) (1:500, Cell Signaling Technology); rabbit anti-caspase 3 (Cell Signaling Technology, 1:200); mouse anti-calbindin(Sigma, 1:1000); rabbit anti-calbindin (Swant, 1:5000); rabbit anti-Pax2(Zymed, 1:500); rabbit anti-neurogranin (Chemicon, 1:500); mouse anti-Nkx6.1[(Pedersen et al., 2006)Developmental Studies Hybridoma Bank, 1:100]. Apoptosis was quantified by immunostaining for cleaved caspase 3. Proliferation and/or cell cycle exit was quantified by PH3 immunostaining or BrdU pulse labeling (1 hour or 24 hours)followed by BrdU, BrdU/Ki67 immunostaining. Coronal sections of wild-type and conditional mutants (≥3 embryos for each genotype and stage) were used. In situ hybridization and/or immunostaining for region-specific markers (En1 for mes/r1 at E9.5; Otx2 for mes; Pax7 for dorsal mes/r1; Pax6 for diencephalon)was performed on adjacent sections. Cells were counted in the ventral and dorsal mes and r1 (≥3 sections) at E9.5 and E10.5 and either normalized for ventricle length (cleaved caspase 3, PH3) or total cell number (BrdU). At E12.5, cells in the dorsal mes (≥3 sections) and in r1 and ventral mes (1-3 sections) were counted and normalized for ventricle length (PH3) or cell number (BrdU⁺, Ki67⁺/BrdU⁺). Cell counts were performed using ImageJ.

Detailed protocols are available at http://www.mskcc.org/mskcc/html/77387.cfm.

To study Gli3 function, we generated mice carrying a conditional Gli3 mutant allele, Gli3^flox, in which two loxP sites flank exon 8 (see Fig. S1 in the supplementary material). Deletion of exon 8 and splicing from exon 7 to 9 result in a frameshift mutation upstream of the DNA-binding domain (Gli3^rec). Since we previously showed that a short period of gene function can be sufficient to rescue the most severe defects observed in null mutants(Blaess et al., 2006; Sgaier et al., 2007), En1-Cre was used to remove Gli3 specifically in the mes/r1 by E9.0, about 36 hours after the onset of Gli3 expression(Hui et al., 1994; Kimmel et al., 2000; Li et al., 2002). To assess whether Gli3 is required for mes/r1 development after midgestation, Nestin-Cre was used to remove Gli3 from the entire neural tube around E11.5 (Blaess et al.,2006; Graus-Porta et al.,2001; Tronche et al.,1999). Indeed, inactivation of Gli3 by E9(Gli3^flox/Xt;En1-Cre mice, referred to as En1-Gli3 cko) or E11.5 (Gli3^flox/Xt;Nestin-Cremice, referred to as Nes-Gli3 cko) resulted in rescue of the early postnatal lethality of Gli3-null mutants and the conditional knockout mice survived for several months.

The Gli3 conditional mutants were first compared with Gli3-null mutants at E18.5, as the latter die at birth. Gli3-null mutants (Gli3^Xt/Xt,Gli3^Xt/rec or Gli3^rec/rec) showed a phenotype of a variable nature in the midbrain and cerebellum at E18.5 (see Fig. S2A-J in the supplementary material), but the variability was independent of the allele (rec or Xt). Therefore, all three genotypes were used to represent the Gli3-null (Gli3^-/-)mutant phenotype. At the gross morphological level, the phenotype of E18.5 Gli3^-/- mutants included: (1) a poorly foliated cerebellum that was not clearly separated from the isthmus; (2) an expanded isthmus-like region with ectopic cell clusters; (3) an overgrown tectum; (4) loss of the distinct morphology that normally defines the isthmus, IC and SC (8/11)(Fig. 1A,B,E,F; see Fig. S2A-J in the supplementary material). By contrast, the tegmentum of Gli3^-/- mutants was morphologically unaffected and the size of the ventral nuclei comprising tyrosine hydroxylase (TH)-positive dopaminergic neurons or Isl1-positive motoneurons appeared similar to that of wild-type (WT) mice (Fig. 1I,Jand data not shown).

Similar to Gli3^-/- mutants, E18.5 En1-Gli3 cko mutants had a normal tegmentum, and a misshapen and enlarged tectum and isthmus (Fig. 1C,G,K; see Fig. S2K,L in the supplementary material). In contrast to Gli3^-/- mutants, however, the isthmus, IC and SC each appeared as morphologically distinct structures in the majority of En1-Gli3 cko mutants (9/16). Furthermore, the cerebellum was well developed, appeared to have a normal cytoarchitecture and had begun to foliate, although the foliation pattern was abnormal(Fig. 1C,G; Fig. 3O-Q; see Fig. S2K,L in the supplementary material). When Gli3 function was left intact until E11.5 (Nes-Gli3 cko mutants), all the morphological defects in the cerebellum, isthmus and tectum were rescued at E18.5(Fig. 1A,D,E,H), but the size of the isthmus and of the entire tectum was significantly increased(Fig. 1A,D,E,H,L; see Fig. S2M,N in the supplementary material).

In summary, histological analysis indicates that Gli3 is required for the dorsal mes/r1 primordium to form distinct brain structures before E9.0, and to establish the cerebellar foliation pattern and the normal size and shape of the isthmus and tectum between E9 and E11. Furthermore, Gli3 regulates growth of the tectum and isthmus beyond E11.5.

To assess when the mes/r1 phenotypes arise in Gli3 mutants, we analyzed E10.5 (Gli3^-/- and En1-Gli3 cko) and E12.5 (all mutants) embryos. We observed severe morphological defects in the mes/r1 of Gli3^-/- and En1-Gli3 cko mutant embryos, which in general were more pronounced in Gli3^-/-mutants (Fig. 1M-O and data not shown): (1) the mesencephalic ventricle was expanded; (2) dorsal posterior mes, isthmus and r1 were not morphologically distinct from each other and the isthmic flexure was less prominent; (3) the thickness of the ventricular zone of the posterior dorsal mes, isthmus and r1 region was increased.

The increased growth of the mes/r1 in Gli3 mutant embryos could be caused by a decrease in cell death or an increase in proliferation. A previous whole-mount analysis suggested that cell death is decreased in the mes/r1 at E8.5 and in the mes dorsal midline at E9.5 in Gli3^-/-mutants (Aoto et al., 2002). To investigate whether decreased cell death could underlie the expansion of dorsal mes/r1 in Gli3 mutant embryos, we analyzed the number of cleaved caspase-3-positive cells in the mes/r1 of E9.5 and E10.5 En1-Gli3 cko mutants. There was a trend toward decreased cell death in En1-Gli3 cko mutants as compared with WT, but only in the E10.5 dorsal mes/r1 (see Fig. S3D,E in the supplementary material). However, because these results indicate that cell death is not dramatically reduced in the absence of Gli3, reduced cell death is unlikely to be the sole cause of mes/r1 overgrowth in the mutants.

To address whether increased proliferation contributes to the increase in mes/r1 size, we analyzed ventral and dorsal mes/r1 of WT and cko mutant embryos for BrdU incorporation (1-hour pulse, S phase of the cell cycle) or for expression of phosphohistone H3 (PH3) (G2-M phase of the cell cycle). Quantitative assessment of proliferating cells indicated a slight increase in proliferation in the dorsal mes/r1 of En1-Gli3 cko mutants at E9.5 and E10.5 and in the dorsal mes of E12.5 Nes-Gli3 cko mutants as compared with WT (see Fig. S3A-C in the supplementary material). Importantly,in the ventral mes/r1, proliferation appeared to be similar in cko mutants and WT, consistent with the lack of a ventral phenotype at E18.5. Since an increased proliferation rate could be caused by a shortening of cell cycle length or by a delay in differentiation, we analyzed the cell cycle exit rate in E12.5 WT and Nes-Gli3 cko mutants(Chenn and Walsh, 2002). We quantified the percentage of differentiating cells (BrdU⁺Ki67^- cells/BrdU⁺ cells) 24 hours after BrdU administration and did not observe an obvious changes between WT and mutant in ventral or dorsal mes or r1 (see Fig. S3F in the supplementary material). Thus, the increase in proliferation is not due to a decreased ability of cells to leave the cell cycle. In summary, these data demonstrate that Gli3is an important regulator of the growth of dorsal r1 (up to E11.0) and mes(beyond E11.0), modulating both cell proliferation (attenuation) and cell death (augmentation).

To investigate whether the abnormal tectal morphology in Gli3mutants is due to the observed mes overgrowth or to a direct requirement for Gli3 in tectal patterning, we analyzed the expression of SC and IC markers at E18.5. We used En1 as a marker for the IC because En1 is normally expressed strongly throughout the IC and at lower levels in the posterior SC (Fig. 2A). Otx2 was used as an SC marker because its expression in the superficial layers of the SC is clearly distinguishable from the Otx2 expression in the IC ventricular zone(Fig. 2E). In Gli3^-/- mutants, En1 expression was very weak and was confined to the most posterior tectum (3/3), even in mutants in which the tectum appeared to be organized into an SC and IC at a morphological level(2/3) (Fig. 2B). In En1-Gli3 cko mutants, En1 expression was stronger and broader than in Gli3^-/- mutants, but was restricted to a more posterior tectal region than in WT brain (3/3)(Fig. 2C). These data indicate that the IC was not properly specified in the mutants. This was also evident by Otx2 expression, which was not restricted to the ventricular layer of the posterior tectum in Gli3^-/- and En1-Gli3cko mutants. In addition, the thickness of the Otx2-positive layer was increased in the SC of Gli3^-/- and En1-Gli3cko mutants, consistent with the tectal overgrowth in these mutants. As expected, given the normal tectal morphology of E18.5 Nes-Gli3 cko mutants, the En1 and Otx2 expression domains were comparable to WT (Fig. 2A,E,D,H).

Despite the abnormal En1 and Otx2 expression in the posterior tectum, this region was not transformed into SC in the mutants, as the AP extent of the Otx2 expression domain in the superficial layers of the SC in Gli3^-/- and En1-Gli3 cko mutants was comparable to WT (Fig. 2F,G). Furthermore, neurogranin, which is expressed in differentiated cells in the lateral IC at E18.5 and in the entire IC at P16(Fig. 2I,L), was found to be expressed in the lateral tectum of E18.5 Gli3^-/- mutants. The neurogranin expression domain, however, was severely reduced and shifted anteriorly (3/3) (Fig. 2J). In P16 En1-Gli3 cko mutants, neurogranin immunostaining revealed a more elongated IC domain than in WT (Fig. 2L,M). By contrast, the shape of the IC in P16 Nes-Gli3cko mutants was comparable to WT, even though the size of the IC was increased(Fig. 2L,N). These data indicate that Gli3 is required for proper establishment of the IC before E9.0, regulates normal IC morphology between E9.0 and E11.0 and growth beyond E11.0. By contrast, Gli3 is not required for establishing the SC, although it plays a prolonged role in controlling normal SC growth.

To study the cellular phenotype of the isthmus-cerebellum-like region in Gli3^-/- mutants, we analyzed markers for cerebellar and isthmic cell types. In the E18.5/P0 WT cerebellum, Purkinje cells are organized in a layer underlying Math1 (Atoh1)-positive granule cell precursors in the external granule cell layer (Fig. 3A,C,F,I). At this stage, all Purkinje cells express the retinoic acid receptor-related orphan receptor alpha (RORα), while a large subset expresses calbindin and inositol 1,4,5-trisphosphate receptor 1 (IP3R1; Itpr1- Mouse Genome Informatics) (Fig. 3F,I and data not shown). Pax2 is expressed throughout the isthmus and deeper cerebellum where it marks (among other cell types) a subset of interneuron precursors (Fig. 3L) (Maricich and Herrup,1999).

In Gli3^-/- mutants, the cerebellar external granule cell layer did not extend into the anterior area of the medial isthmus-cerebellum-like region, suggesting that the anterior region might be transformed into isthmus (Fig. 1B; Fig. 3B,D). This anterior area, however, did not contain Pax2-positive cells, but instead had clusters of RORα- and IP3R1-positive Purkinje cells, indicating that this region had some characteristics of the cerebellum rather than of the isthmus (Fig. 3G,J,M and data not shown). The organization of the posterior isthmus-cerebellum-like region had some similarities to WT cerebellum, with a relatively normal distribution of Pax2-positive cells and RORα- and IP3R1-positive Purkinje cells underlying a Math1-positive external granule cell layer(Fig. 3B,D,G,J,M and data not shown). The Purkinje cell layer, however, was thicker than normal and did not consistently extend into the most posterior area. In addition, Purkinje cell axons that form bundles in WT were highly disorganized in Gli3^-/- mutants (Fig. 3F,G and data not shown). This disorganization of Purkinje cells was equally severe in more-lateral regions of the cerebellum (data not shown). These results indicate that the anterior isthmic-cerebellar region is not properly specified in Gli3^-/- mutants and the posterior isthmus-cerebellum area is not organized into a normal cerebellar cytoarchitecture.

In P0 En1-Gli3 cko mutants, the defects in cerebellar cytoarchitecture were partially rescued, based on marker analysis(Fig. 3E,H,K,N and data not shown). Purkinje cell axons projected aberrantly into the posterior isthmus and some Purkinje cells formed clusters in the anterior isthmic region(Fig. 3H,K), suggesting that the isthmus-cerebellum boundary was not properly established in En1-Gli3 cko mutants. Furthermore, at E18.5 and all the postnatal stages analyzed (P2, 5, 8, 16 and 30), the cerebellar foliation pattern was clearly abnormal and varied between mutants(Fig. 3O-Q; see Fig. S2K,L in the supplementary material). Interestingly, these defects in foliation were not associated with major changes in AP gene expression domains, as the expression of Otx2 (posterior region) and of Runx1 (central region) in the external granule cell layer were comparable to WT (data not shown). Consistent with the histological analysis, the cytoarchitecture,morphology and foliation pattern of E18.5 and postnatal Nes-Gli3 cko cerebella were similar to those of WT (Fig. 1A,D and data not shown). In conclusion, Gli3 is required primarily before E9.0 for the proper specification of the isthmus and cerebellum, including the formation of normal cerebellar cytoarchitecture, and between E9.0 and E11.0 for establishing the stereotypic cerebellar foliation pattern.

Since Shh signaling induces ventral structures in the neural tube, one possible reason for the abnormal patterning of the IC, isthmus and cerebellum in Gli3^-/- and En1-Gli3 cko mutants is a Shh-induced ventralization of dorsal structures. To address this, we investigated whether the expression of ventrally (Gli1, Nkx2.2 and Nkx6.1) and dorsally [Pax7, Gbx2, ephrin A5(Efna5)] restricted genes is altered in the mes/r1 of Gli3mutants. Surprisingly, no dorsal expansion of Gli1, Nkx2.2 or Nkx6.1 was observed in Gli3^-/- and En1-Gli3 cko mutants (Fig. 4D-I and data not shown). Furthermore, in both Gli3^-/- and En1-Gli3 cko mutants, Pax7encompassed a similar dorsal domain as in WT embryos, indicating further that the dorsal mes/r1 was not ventralized in Gli3 mutants(Fig. 4A-C). In addition,expression of Gbx2 in dorsal r1 and of Efna5 in dorsal-posterior mes were maintained in Gli3^-/- and En1-Gli3 cko mutant embryos (data not shown). In summary, these results show that Gli3 does not play a major role in establishing DV spatial molecular patterning in the mes/r1.

The unaltered DV gene expression domains in Gli3 mutants indicate that there is no ectopic GliA-mediated Shh signaling activity in the dorsal mes/r1 in the absence of Gli3. To definitively demonstrate that GliA-mediated Shh signaling does not contribute to the patterning and growth defects in Gli3 mutants, we generated double cko mutants for Smo and Gli3 (En1-Gli3;Smo cko), as removal of Smo results in the absence of GliA activity. Indeed, we found that E12.5 and P0 En1-Gli3;Smo cko mutant embryos had a tectal and isthmus phenotype very similar to that of En1-Gli3 cko mutants(Fig. 4K-M; compare Fig. 4J with Fig. 1C). The morphology and overgrowth of r1 in En1-Gli3;Smo cko mutants was also similar to that shown by En1-Gli3 cko mutants at E12.5. At P0, the overall size of the cerebellum and thickness of the external granule cell layer, however, were reduced (Fig. 4J and data not shown), resembling the En1-Smo cko phenotype that results from severely decreased granule cell precursor proliferation after E16.5(Blaess et al., 2006; Corrales et al., 2006). This indicates that the overgrowth in the En1-Gli3;Smo cko cerebellar anlage is initially caused by loss of Gli3R (before E16.5), whereas after E16.5 the loss of Gli1/2A-mediated Shh signaling downstream of Smo results in reduced cerebellum growth owing to decreased granule cell precursor proliferation. In summary, these data demonstrate that the initial phenotypes in the dorsal mes/r1 of Gli3 mutants result from a loss of Gli3R activity, rather than from ectopic GliA-mediated Shh signaling.

Since defects in the roof plate contribute to the telencephalic phenotype in Gli3^-/- mutants(Grove et al., 1998; Theil et al., 2002), we assessed whether alterations in signaling from the mes/r1 roof plate cause the dorsal mes/r1 phenotypes of Gli3 mutants. The expression of two secreted factors involved in the organizing function of the roof plate, Wnt1 and Gdf7, was maintained in the roof plate of Gli3^-/- and En1-Gli3 cko embryos at E9.5 and E10.5 (Fig. 5A-F and data not shown). In addition, Msx1, which is a downstream target of BMP signaling (Alder et al., 1999; Bei and Maas, 1998), was present in the roof plate in Gli3^-/- and En1-Gli3cko mutants (Fig. 5G,H and data not shown). Finally, expression of Axin2, a target of Wnt signaling(Jho et al., 2002), was not grossly altered in E10.5 Gli3^-/- or En1-Gli3 cko mutant embryos (Fig. 5I,J and data not shown). In summary, Gli3 does not appear to play a major role in establishing or maintaining the mes roof plate.

Since changes in DV patterning or roof plate signaling do not seem to account for the dorsal mes/r1 phenotype in Gli3 mutants, we next examined whether alterations in the Fgf8 signaling pathway, which is a primary regulator of AP patterning, could contribute to the dorsal mes/r1 defects in Gli3 mutants. Based on whole-mount analysis, it has been reported that Fgf8 expression is expanded in E9.5 and E10.5 Gli3^-/- mutants, whereas the Fgf8 domain is reduced when Shh signaling is decreased and Gli3R is increased(Aoto et al., 2002; Blaess et al., 2006). We first analyzed Fgf8 expression in spatial and temporal detail on sagittal sections of Gli3 mutant embryos. Indeed, we observed ectopic Fgf8 expression, but only in medial r1 in both E9.5 and E10.5 Gli3^-/- (Fig. 6G-I and data not shown) and En1-Gli3 cko (data not shown) mutants. By E12.5, shortly before the normal termination of Fgf8 expression, ectopic Fgf8 expression was restricted to the posterior-most part of medial r1 in Gli3^-/- (data not shown) and En1-Gli3 cko (Fig. 6J,K) mutants. Interestingly, this region corresponds to the rhombic lip in the WT and normally expresses Wnt1 and Math1(Fig. 6A,B,D,E and data not shown). Furthermore, Fgf17, which also plays a role in AP patterning of the mes/r1 and is expressed in a broader domain than Fgf8 in the WT (Xu et al., 2000), and sprouty 1 (Spry1), a direct target of Fgf8 signaling(Liu et al., 2003), were expanded posteriorly in both Gli3 mutants, but were not altered in the mes (Fig. 6M-Q,S-W). To address whether Gli3R rather than Shh signaling is required for Fgf8suppression in r1, we analyzed E12.5 En1-Gli3;Smo cko mutants and found that similar to En1-Gli3 cko mutants, Fgf8, Fgf17 and Spry1 were expressed ectopically in dorsal-posterior r1 (data not shown).

Consistent with the normal tectal and cerebellar morphology in Nes-Gli3 cko mutants, no changes in Fgf8, Fgf17 or Spry1 expression were observed in these mutants at E12.5(Fig. 6F,L,R,X). In addition,in all Gli3 mutants the expression domains of Wnt1 and Otx2 in the mes appeared normal (anterior to Fgf8expression) and Gbx2 expression in r1 was maintained(Fig. 6A-F and data not shown). These results indicate that changes in gene expression are specific to the Fgf pathway and that the boundary between the mes and r1 is intact in Gli3 mutants. Thus, Gli3R is required to inhibit Fgf8 and Fgf17 expression beyond E9.0, but only in dorsal-medial r1.

To test whether the expansion of Fgf expression in r1 is responsible for any of the dorsal defects in Gli3 mutants, we removed one copy of Fgf8 in Gli3^-/- mutants. Histological analysis of Gli3^-/-;Fgf8^+/- mutants (n=4) at E18.5 or P0 showed a striking partial rescue of the defects seen in the isthmus-cerebellum-like region, but not tectum, of Gli3^-/-mutants (Fig. 7; see Fig. S4 in the supplementary material). Unlike any of the Gli3^-/-littermates, the cerebellum of all Gli3^-/-;Fgf8^+/- mutants had begun to foliate at P0 and the external granule cell layer extended along the AP length of the medial cerebellum (Fig. 7A,E;for a direct comparison of littermates see Fig. S4A-H in the supplementary material). In addition, the isthmus appeared to have formed, based on morphology and the expression of Pax2 (Fig. 7A,H; see Fig. S4A-H in the supplementary material). Marker analysis further revealed a more normal organization of Purkinje cells,although cell clusters remained in and near the isthmus region, showing that a separation of isthmus and cerebellum was not fully established(Fig. 7F,G; see Fig. S4I,J in the supplementary material).

Although the isthmus-cerebellar phenotype was partially rescued in the Gli3^-/-;Fgf8^+/- mutants as compared with Gli3^-/- single mutants, the tectal phenotype was similar to Gli3^-/- single mutants. The tectum was enlarged, the IC was not clearly distinguishable and the range of morphological abnormalities was comparable to that of Gli3^-/- mutants(Fig. 7A; Fig. 1B; see Fig. S2A,C,E,G,I and Fig. S4 in the supplementary material). Furthermore, En1, Otx2and neurogranin were abnormally expressed in the posterior tectum, similar to in Gli3^-/- mutants(Fig. 7B-D; Fig. 2A,B,E,F). These findings provide genetic evidence that a major role of Gli3R in regulating the organization of cerebellar and isthmic cytoarchitecture is to localize Fgf8 expression to dorsal-medial r1.

By studying the sequential cellular and genetic changes underlying Gli3 function in mes/r1 development we demonstrate that Gli3R is a crucial regulator of growth and patterning of the dorsal mes/r1, whereas neither Gli3R nor Gli3A is required for development of the ventral mes/r1(summarized in Fig. 8A). We show that dorsally, Gli3 has a prolonged role in attenuating the growth of r1(before E11.0) and of the mes and isthmus (beyond E11.0) through reducing cell proliferation and promoting cell death. Importantly, analysis of Gli3;Smo double mutants demonstrates that normally it is Gli3R that attenuates growth in the mes/r1 and that factors other than Shh must stimulate mes/r1 proliferation. We further show that Gli3R has additional functions in patterning of the dorsal mes/r1. Before E9.0, Gli3R is required to establish a normal IC and a distinct isthmus and anterior cerebellum, as well as a normal cytoarchitecture. Between E9.0 and E11.0, Gli3R prevents Purkinje cells and their axons from entering the isthmus and is involved in setting up a normal foliation pattern. Finally, we demonstrate that the loss of the distinct separation between isthmus and cerebellum in Gli3-null mutants, as well as the abnormal cytoarchitecture in this region, are caused by ectopic expression of Fgf8 in r1.

We and others showed previously that Fgf8 expression in the isthmic organizer is reduced when Gli3R is upregulated (Shh/Smomutants) and increased when Gli3 is absent(Gli3^-/- mutants) (Aoto et al., 2002; Blaess et al.,2006). We built on these findings by showing that complete absence of Gli3 as well as inactivation of Gli3 at E9.0(En1-Gli3 cko mutants) result in an expanded expression domain of not only Fgf8, but also Fgf17 and the Fgf8 downstream target Spry1. Furthermore, the expansion occurs only in the dorsal-medial region of r1 and, interestingly, this region includes the rhombic lip region. Moreover, the ectopic Fgf expression in r1 in the absence of Gli3 is not diminished when Shh signaling is abolished (En1-Gli3;Smo cko mutants). Thus, Gli3R and not Shh activator signaling is required to restrict Fgf8/Fgf17 expression to the isthmus. These results raise the question whether Gli3 might regulate Fgf8 expression directly. Sequence analysis of the known Fgf8 regulatory elements(Beermann et al., 2006),however, did not reveal any canonical Gli binding sites(Vokes et al., 2007) in these conserved regions (S.B. and J. Hastings, unpublished).

Fgf8 expression in the WT mes/r1 initially encompasses the isthmus and most of dorsal r1 at E8.5, and then becomes restricted to the isthmus by E10.5 (Crossley and Martin,1995). Although there is evidence that the transcription factors Otx2 and Grg4 (Tle4 - Mouse Genome Informatics) are normally involved in repression of Fgf8 in the mes, it was unclear how Fgf8 becomes downregulated in dorsal r1 (Ye et al., 2001). Our study of Gli3 mutants indicates a role for Gli3R in the initial downregulation of Fgf8 in r1, as well as a continued requirement that extends beyond E9.0 in maintaining the dorsal restriction of Fgf8 to the isthmus. Alternatively, Gli3 might be required to maintain dorsal-medial r1 identity, and the expansion of Fgf8 expression is a secondary consequence of a mixed r1/isthmus identity. Interestingly, a similar interaction between Shh, Gli3 and Fgf8 exists in the developing telencephalon(Gutin et al., 2006; Kuschel et al., 2003; Ohkubo et al., 2002),suggesting that the interplay between these two signaling pathways constitutes a more general mechanism in the patterning of three-dimensional brain structures.

Based on Fgf8 loss- and gain-of-function studies, the observed ectopic expression of Fgf8 in dorsal r1 of Gli3 mutants could affect both tectum and cerebellum development(Zervas et al., 2005). To test this possibility, we analyzed Gli3^-/- mutants in which one copy of Fgf8 was removed. Strikingly, we found that whereas the isthmic-cerebellar region phenotype is partially rescued in Gli3^-/-;Fgf8^+/- mutants, the defects in tectal growth and IC patterning were largely not attenuated. The lack of influence of Fgf8 on mes development is likely to be because an insufficient level of Fgf8 signaling is attained in the mes, consistent with the observation that Spry1 is not obviously upregulated in the mes of Gli3mutants.

Interestingly, even though Fgf8 expression is expanded in both Gli3^-/- and En1-Gli3 cko mutants at E10.5, the cerebellar defects are much more severe in Gli3^-/-mutants. In Gli3^-/- mutants, the cerebellar cytoarchitecture is severely altered and foliation is not initiated by P0. By contrast, inactivation of Gli3 after E9.0 results in only mild cytoarchitectural defects in the cerebellum and a normal onset of foliation,but abnormal foliation pattern. Preliminary analysis of En1-Gli3 cko mutants in which one copy of Fgf8 had been removed found that the foliation pattern is still abnormal (S.B., unpublished). One possibility is that the foliation defects in En1-Gli3 cko mutants are independent of ectopic Fgf8 and instead are directly caused by the loss of Gli3.

One striking phenotype in the Gli3-null and cko mutants is the extensive overgrowth of tectal tissue. It has previously been shown that ectopic activation of the Shh pathway by misexpression of either Shh or GliA results in a pronounced overgrowth of the spinal cord, mes/r1 and postnatal cerebellum in chick and mouse (Cayuso et al., 2006; Corrales et al.,2006; Dahmane and Ruiz-i-Altaba, 1999; Rowitch et al., 1999). The increased size of the tectum that we documented in Gli3 and Gli3/Smo mutants, however, demonstrates that removing Gli3R, without an associated ectopic activation of Gli activator function, can cause increased proliferation and overgrowth. Furthermore, Gli3R is normally required beyond E11.5 to attenuate growth of the dorsal mes.

Our present results taken together with previous studies(Blaess et al., 2006; Britto et al., 2002; Ishibashi and McMahon, 2002; Corrales et al., 2006) suggest several mechanisms by which Gli transcription factors regulate tissue growth. At high levels of signaling, Shh can simply increase proliferation by upregulating GliA (primarily Gli2) such as in cerebellar granule cell precursors (Corrales et al.,2006), and/or by downregulating Gli3R such as in the ventral midbrain (Blaess et al.,2006). At low levels of Shh signaling, such as in the dorsal mes,GliA is not induced, and Shh increases overall growth only by downregulating Gli3R levels which leads to increased proliferation and cell survival(summarized in Fig. 8B). The latter reveals that other signaling pathways must stimulate proliferation of dorsal mes progenitors and a certain level of Gli3R is required only to attenuate growth.

A possible candidate to regulate proliferation in the dorsal mes is Wnt1,as it can induce overgrowth of the posterior mes(Panhuysen et al., 2004) and recent studies have provided evidence that Gli3R inhibits canonical Wnt signaling in several different tissues and cell types(Alvarez-Medina et al., 2008; Ulloa et al., 2007). However,our analysis of Axin2 expression (a readout for canonical Wnt signaling) in the mes of Gli3 mutants provided no evidence that Gli3 is required in the dorsal mes to antagonize Wnt signaling. Furthermore, our preliminary analysis of Gli3^-/-;Wnt1^sw/sw double mutants revealed that the size of the SC is comparable to that of Gli3^-/-mutants, although the IC (and medial cerebellum) is lost in both Wnt1^sw/sw and Gli3^-/-;Wnt1^sw/sw mutants (S.B., unpublished). It is possible, however, that additional members of the Wnt family stimulate dorsal mes growth and that their action is normally attenuated by Gli3R.

In summary, we have found that the precise regulation of Gli3R levels is crucial to establish the intricate structures of the mature tectum and cerebellum. Our data provide insight into how Gli3R coordinates the function of two distinct organizer molecules in the mes/r1 by modulating their expression or downstream signaling. It will be interesting to investigate whether Gli3R plays similar complex roles in other tissues that undergo organizer-dependent three-dimensional patterning and growth.

We thank Drs F. Costantini and K. Millen for providing RNA in situ probes;the Developmental Studies Hybridoma Bank for the Nkx6.1 antibody; Frada Berenstehyn for excellent technical help; Jaden Hastings for help with sequence analysis; and Anamaria Sudarov, Charles Levine, Emilie Legue and Roy Sillitoe for critical reading of the manuscript. This work was funded by grant R01 CA128158 from NICHD. S.B. was supported by a DFG postdoctoral fellowship.