Morpholinos for splice modificatio

Morpholinos for splice modification



The Shh signaling pathway is required in many mammalian tissues for embryonic patterning, cell proliferation and differentiation. In addition, inappropriate activation of the pathway has been implicated in many human tumors. Based on transfection assays and gain-of-function studies in frog and mouse, the transcription factor Gli1 has been proposed to be a major mediator of Shh signaling. To address whether this is the case in mouse, we generated a Gli1 null allele expressing lacZ. Strikingly, Gli1 is not required for mouse development or viability. Of relevance, we show that all transcription of Gli1 in the nervous system and limbs is dependent on Shh and, consequently, Gli1 protein is normally not present to transduce initial Shh signaling. To determine whether Gli1 contributes to the defects seen when the Shh pathway is inappropriately activated and Gli1 transcription is induced, Gli1;Ptc double mutants were generated. We show that Gli1 is not required for the ectopic activation of the Shh signaling pathway or to the early embryonic lethal phenotype in Ptc null mutants. Of significance, we found instead that Gli2 is required for mediating some of the inappropriate Shh signaling in Ptc mutants. Our studies demonstrate that, in mammals, Gli1 is not required for Shh signaling and that Gli2 mediates inappropriate activation of the pathway due to loss of the negative regulator Ptc.


GLI1, which encodes a member of the Gli-Kruppel family of transcription factors, was initially identified as an amplified gene and potential oncogene in a human glioblastoma (Kinzler et al., 1987). A family of Gli genes that includes Gli1, Gli2 and Gli3 was cloned from human and mouse, and found to be expressed in many organs during mouse development (Ruppert et al., 1988; Hui et al., 1994). The fly homolog of Gli, Cubitus interruptus (Ci), is involved in mediating all Hedgehog (Hh) signaling (Methot and Basler, 2001). Given that Hh proteins are critical for many developmental processes in vertebrates, considerable effort has been made in the past decade to elucidate the function of each of the Gli genes in the Hh pathway (for reviews, see Matise and Joyner, 1999; Ingham and McMahon, 2001). In particular, the roles of the Gli proteins downstream of Sonic Hedgehog (Shh) signaling have been studied in ventral patterning of the nervous system.

In fly, Hh regulates Ci function by inhibiting processing of Ci into a repressor protein, and at the same time potentiating the full-length activator protein (Aza-Blanc et al., 1997; Ohlmeyer and Kalderon, 1998; Methot and Basler, 1999). Gli2 and Gli3, but not Gli1, have been similarly found to have an N-terminal repressor domain and be cleaved into a repressor form (Dai et al., 1999; Sasaki et al., 1999; Aza-Blanc et al., 2000; Wang et al., 2000). Gli2 and Gli3 also have been found to be required for development and Shh signaling. Loss of Gli2 function results in defective Shh signaling in the floorplate of the neural tube and other tissues (Mo et al., 1997; Ding et al., 1998; Matise et al., 1998), indicating Gli2 is an activator in the Hh pathway. By contrast, loss of mouse Gli3 results in dorsal brain defects and limb polydactyly that are associated with ectopic activation of the Shh pathway (Hui and Joyner, 1993; Masuya et al., 1995; Buscher et al., 1997). The Gli3 mutant phenotype suggests that Gli3 functions primarily as a repressor in the Shh pathway. Indeed, biochemical studies have shown that Shh functions to inhibit the formation of the repressor form of Gli3 (Wang et al., 2000), and removal of Gli3 function in Shh mutants largely rescues the Shh mutant defects, showing that part of the Shh mutant phenotype is due to an excess of Gli3 repressor (Litingtung and Chiang, 2000).

Interestingly, gain-of-function studies in mouse and frog embryos have shown that Gli1, but not Gli2 or Gli3, can mimic Shh function by inducing proliferation and activating Shh target genes, including Hnf3b in the dorsal CNS (Hynes et al., 1997; Lee et al., 1997; Park et al., 2000). In addition, Gli1 is always expressed near Shh and can be transcriptionally activated by Shh (Grindley et al., 1997; Hynes et al., 1997; Lee et al., 1997). Based on these findings, it was proposed that Gli1 is the key transcription factor acting in the Shh pathway (Hynes et al., 1997; Ruiz i Altaba, 1997). In support of this assertion, GLI1 is expressed in human tumors thought to be caused by elevated Shh signaling, such as basal cell carcinomas (Dahmane et al., 1997; Reifenberger et al., 1998). Furthermore, endogenous Gli1 in frog was recently shown to be required for a transiently supplied exogenous human GLI1 protein to induce hyperproliferation in the nervous system (Dahmane et al., 2001).

To investigate the function of Gli1 in mouse development, we previously generated a Gli1 mutant allele lacking the exons that encode zinc fingers 2-5 (zinc-finger deletion or Gli1zfd) (Park et al., 2000). In contrast to the predictions based on gain-of-function assays, mice homozygous for the Gli1zfd mutation are viable and do not have obvious defects. Further analysis, however, revealed an alternatively spliced transcript produced by the mutant allele that could encode a Gli1 protein lacking only the zinc fingers. It is therefore possible that the Gli1zfd allele is hypomorphic, and that Gli1 may be required for mouse development.

To address definitively the requirement for Gli1 in Shh signaling, as well as in mouse development, we generated a new Gli1-null allele in which the coding sequences of Gli1 are replaced with lacZ (Gli1lz). Using this allele, we show that Shh is required for the initial transcriptional activation of Gli1, and thus Gli1 cannot transduce the earliest Shh signaling in tissues. Consistent with this, Gli1-null mutants develop properly and adults appear normal. We also tested the requirement for Gli1 and Gli2 in patched (Ptc; Ptch – Mouse Genome Informatics) mutants in which the Shh pathway is ectopically activated. Our results show that ectopic activation of the Shh pathway via loss of the negative regulator Ptc is not dependent on Gli1, but significantly is dependent on Gli2.


Generation of Gli1 null mutant knock-in mice

W4 ES cells (Auerbach et al., 2000) were electroporated with a Gli1lz targeting construct followed by gancyclovir and G418 double selections according to Matise et al. (Matise et al., 2000). Targeted ES cell clones were identified by restriction enzyme digestion and Southern blot analysis of ES cell DNA using 5′ and 3′ external probes (see Fig. 1). The 5′ probe identified a 9 kb fragment in the targeted allele and a >20 kb fragment in the wild-type allele following EcoRV digestion. The 3′ probe identified a 4.5 kb fragment in the targeted allele and a 9 kb fragment in the wild-type allele with XbaI digestion. The targeting frequency was one in four ES cells surviving double selections. Three targeted cell lines were then injected into C57BL/6 blastocysts to generate chimeras (Papaioannou and Johnson, 2000). Chimeras were then bred with 129SvEv and Black Swiss mice (Taconic) to establish F1 heterozygotes and three independent germline transmitting mouse lines were established. The floxed neo gene was removed by breeding with TK-Cre transgenic mice (W. A., unpublished). PCR analysis was used for routine genotyping with the following primers (see Fig. 1):

Fig. 1.

Gli1 gene targeting strategy and ES cell screening. (A, top) The Gli1 genomic locus and targeting construct. White boxes indicate Gli1 exons. Black boxes indicate Gli1 zinc-finger domains 1-5. B, BamHI; RV, EcoRV; H, HindIII; Hs, Hsp921; Xb, XbaI. (A, middle) Targeted allele with neo cassette. The N-terminal and zinc-finger domains are replaced by nuclear lacZ and a neo cassette. The neo cassette is in the opposite orientation relative to Gli1 transcription. (A, bottom) Targeted allele without neo allele. Cre recombinase was used to remove the loxP-neo-loxP cassette. (B) ES cell Southern blot hybridization using 5′ and 3′ external probes indicated in A following EcoRV and XbaI digestion, respectively. The 5′ probe identifies a 9 kb mutant fragment and a >20 kb wild-type fragment, whereas the 3′ probe identifies a 4.5 kb mutant fragment and a 9 kb wild-type fragment.





P1 and P2 identify the wild-type allele, whereas P3 and P4 (against lacZ) identify the targeted allele.

Breeding and genotyping of Gli2lzki, Gli2zfd and Ptc mutants were as described (Mo et al., 1997; Goodrich et al., 1997; Bai and Joyner, 2001). All mice were kept and analyzed on an outbred Swiss Webster background.

Immunohistochemistry and RNA in situ hybridization

Embryos were fixed in 4% paraformaldehyde for 20 minutes at 4°C before embedding in OCT. Frozen sections were cut at 12 μm. Immunohistochemistry was performed as previously described (Matise et al., 1998) using the following monoclonal antibodies: Shh, Nkx2-2, Isl1/2 and Pax6 (Ericson et al., 1996). Cy3-conjugated anti-mouse secondary antibody (Jackson ImmunoResearch) was used at a dilution of 1:500. Images were captured via a Hamamatsu cooled CCD camera and colors assigned offline. X-gal staining and whole-mount RNA in situ hybridization were performed as previously described (Bai and Joyner, 2001).


Production of a Gli1-lacZ marker allele in mice

In order to determine whether Gli1 is indeed required during development and to have mice expressing lacZ as a read-out of Gli1 activity, we generated a new mutant Gli1-null allele by replacing the genomic fragment that encodes the entire Gli1 N-terminal and zinc-finger domains with the lacZ gene (Fig. 1). This replacement results in a loss of Gli1 function, even if alternative splicing occurs from the 5′ untranslated exon to an exon downstream of the lacZ insertion. Three independent lines of mice derived from independently targeted ES cells were established from chimeras (Gli1lz). Analysis of β-galactosidase activity in all three lines showed identical lacZ expression patterns at various embryonic stages (E7.5-E12.5). Furthermore, the lacZ expression pattern was indistinguishable from Gli1 mRNA expression (Hui et al., 1994; Platt et al., 1997), showing that the knock-in allele recapitulates Gli1 expression (data not shown and see Fig. 3).

Both Gli1lz and Gli1zfd are null alleles not required for development

To determine whether Gli1 is required for mouse development and Shh signaling, Gli1lz/lz homozygous mutants were generated. Homozygous Gli1lz mutant mice were found to be viable with no obvious phenotypes. Thus, Gli1 is not required for development or survival of mice.

Although our previous studies showed that homozygous Gli1zfd mutant mice are phenotypically normal, reducing the level of Gli2 in Gli1 homozygous mutant mice (Gli1zfd/zfd;Gli2zfd/+) results in multiple defects including a partial loss of the floor plate, failure of notochord regression, aberrant lung development and lethality (Park et al., 2000). We compared Gli1lz/lz;Gli2zfd/+ embryos with Gli1zfd/zfd;Gli2zfd/+ embryos to determine if there are phenotypic differences between the Gli1lz and Gli1zfd alleles. In Gli1lz/lz;Gli2zfd/+ embryos, as in Gli1zfd/zfd;Gli2zfd/+ embryos, some floor plate cells were found to be lost at E10.5. In the most severe embryos, HNF3β-expressing cells are not present (Fig. 2A,E), Nkx2.2 interneurons are greatly reduced (Fig. 2B,F), and motor neurons occupy a more ventral position (Fig. 2C,D,G,H). At E12.5, the notochord failed to regress (data not shown). Furthermore, similar to Gli1zfd;Gli2zfd double homozygous mutants, Gli1lz/lz;Gli2zfd/zfd embryos displayed smaller lungs at E12.5 and E18.5 compared with Gli2zfd/zfd embryos (Fig. 2I and data not shown), and an extra postaxial nubbin was present in the forelimbs at E18.5 (Fig. 2J,K). Given that the Gli1lz and Gli1zfd alleles show similar phenotypes as single mutants or combined with Gli2 mutant alleles, our results show that both Gli1lz and Gli1zfd are null alleles.

Fig. 2.

Gli1lz/lz;Gli2zfd/+ embryos have defects similar to Gli1zfd/zfd;Gli2zfd/+ embryos. In E10.5 Gli1lz/lz;Gli2zfd/+ embryos, there is a variable loss of floor-plate cells. In the most extreme mutant embryos, HNF3β- and Nkx2-2-expressing cells are greatly reduced (compare E and F with A and B), such that Isl1/2-expressing cells occupy the ventral midline of the spinal cord (compare G and H with C and D). (C,G) Sections at forelimb level; (D,H) sections at hindlimb level. Arrowhead indicates ventral midline of spinal cord. (I) At E12.5, Gli1lz/lz;Gli2zfd/+ lungs are smaller than wild-type lungs, in addition, Gli1lz/lz;Gli2zfd/zfd lungs have only two lobes and are much smaller than wild-type lungs. At E18.5, a postaxial nubbin was found in the limbs of Gli1lz/lz;Gli2zfd/zfd embryos (K), similar to Gli1zfd/zfd;Gli2zfd/zfd embryos. (J) E18.5 wild-type limb. Scale bar: 0.1 mm in A-H.

Shh is required for transcription of Gli1

One possible reason why Gli1 is not required during development is that it is not expressed when the Shh pathway is first activated, because Gli1 transcription requires Shh signaling. This would be consistent with the finding that Shh signaling can activate Gli1 transcription. To examine whether Shh could be required for the initial transcriptional activation of endogenous Gli1, we compared the expression patterns of Shh mRNA with Gli1-lacZ.

Expression of Shh starts at E7.5 in the midline mesoderm of the head process (Echelard et al., 1993). Shortly thereafter, Shh expression expands posteriorly along the midline into the node. By the headfold stage, Shh expression is initiated in the anterior midbrain of the neural ectoderm. As somites form, Shh expression expands anteriorly into the ventral forebrain and posteriorly along the spinal cord. By E9.5, Shh expression can be detected in both the notochord and floor plate throughout the anterior/posterior (AP) axis, along the sulcus limitans in the forebrain, and in the gut endoderm (Fig. 3A-D) (Echelard et al., 1993). Expression of Gli1-lacZ was found to follow closely the expression pattern of Shh. Gli1-lacZ expression can first be detected weakly at the late streak stage (E7.5), close to the midline in the head process (Fig. 3E). It then followed Shh expression by extending into the area surrounding the node (Fig. 3F). When Shh is expressed throughout the AP axis in the notochord and floor plate, Gli1-lacZ was also detected in mesodermal cells surrounding the notochord and in cells next to the floor plate, as well as in the gut (Fig. 3G,H).

Fig. 3.

Shh is required for transcription of Gli1. (A-D) RNA in situ hybridization of Shh at E7.5, E8.0, E8.5 and E9.5 in wild-type embryos. (E-H) X-gal staining of lacZ in Gli1lz/+ embryos at E7.5, E8.0, E8.5 and E9.5. Note the brownish weak X-gal staining of Gli1-lacZ at E7.5. The embryo in E is slightly older than the embryo in A. (I-N) X-gal staining of Gli2-lacZ from E7.5 to E9.5. Note, one embryo in I is at mid-streak stage (ms), and the other embryo is at the late-streak stage (ls). (L) Spinal cord section of an E8.5 embryo showing Gli2-lacZ in the ventral midline. (M,N) Spinal cord section of an E9.5 embryo at an anterior (M) or posterior (N) position. Arrowhead indicates floorplate cells. Spinal cord is outlined by broken red lines. a, anterior; hp, head process; fb, forebrain; fg, foregut; ls, late-streak; mb, midbrain; ms, mid-streak; nd, notochord; p, posterior. (O-R) X-gal staining of lacZ in Gli1lz/+ embryos (O,Q) or Gli1lz/+;Shh–/– embryos (P,R) at E8.5 (O,P) and E10.5 (Q,R). Limb buds in Q and R are outlined by broken white lines. At E8.5, Gli1-lacZ expression can be detected in the CNS and the gut (indicated by an arrow) in wild-type embryos (O), whereas it can only be detected weakly in the gut in Shh mutant embryos (P). At E10.5, Gli1-lacZ is strongly expressed in the gut (arrow), posterior limb bud (red arrowhead) and CNS (black arrowhead) in wild-type embryos (Q). In Shh mutant embryos, the expression in the posterior limb bud and the CNS cannot be detected (R).

These expression studies showed that Gli1 expression is complementary to Shh, but it was not possible to determine whether Gli1 is actually transcribed before or after Shh. Therefore, to address whether transcription of Gli1 is dependent on Shh signaling, we examined Gli1-lacZ expression in wild-type and Shh mutant embryos at E8.5 and E10.5. In wild-type embryos at E8.5, Gli1-lacZ is strongly expressed in the ventral CNS, some mesodermal cells surrounding the notochord and in the gut (Fig. 3O). This expression pattern remains the same but becomes more intensified at E10.5 with additional expression in the forelimbs (Fig. 3Q). Significantly, in Shh mutant embryos, Gli1-lacZ was not detected at E8.5 in the ventral CNS (Fig. 3P). The expression in mesodermal cells and gut was greatly reduced, but not completely lost, which is probably due to expression of Ihh in the gut (Bitgood and McMahon, 1995). By E10.5 in Shh mutants, Gli1-lacZ could only be detected in the gut that expresses Ihh, but not in the CNS or limbs (Fig. 3R). These studies show that Gli1 transcription is absolutely dependent on Hh signaling.

In contrast to Gli1, Gli2 transcription is not dependent on Shh (Bai and Joyner, 2001). To address whether Gli2 is capable of mediating the initial Shh signaling, we examined the expression of Gli2-lacZ at early stages. Gli2-lacZ can be detected at the mid-streak stage, prior to Shh transcription (Fig. 3I). At the late-streak stage and neural plate stage (E8.0), Gli2-lacZ can be detected in a domain complimentary to Shh (Fig. 3I,J). At E8.5, Gli2-lacZ was still expressed in the ventral midline of the neural tube at the time the floorplate is being induced (Fig. 3L). At E9.5, Gli2-lacZ cannot be detected in floorplate cells in the anterior spinal cord (Fig. 3M), but can still be detected in the ventral midline of the tail region (Fig. 3N). Therefore, the temporal and spatial pattern of Gli2 expression is consistent with Gli2 mediating the initial Shh signaling in the ventral CNS. Indeed, loss of Gli2 results in defects in floorplate induction (Ding et al., 1998; Matise et al., 1998).

Gli2, but not Gli1, is involved in transducing ectopic activation of the Shh pathway in Ptc mutants

Ptc is the receptor for Shh and has been found to be a negative regulator of the pathway in mice and flies (Stone et al., 1996; Marigo et al., 1996). Humans with Gorlin’s syndrome have heterozygous mutations in PTC and are born with many developmental defects. Later in life, they can develop medulloblastoma and basal cell carcinomas (Gorlin, 1987). Furthermore, in many basal cell carcinomas caused by hereditary or sporadic mutations in PTC, loss of heterozygosity of PTC is seen (Gailani et al., 1992; Unden et al., 1996; Levanat et al., 1996). Consistent with activation of the Shh pathway due to loss of a negative regulator, GLI1 is expressed in both types of tumors (Dahmane et al., 1997; Dahmane et al., 2001).

As Ptc is a negative regulator of the Shh signaling pathway, loss of Ptc in mouse embryos results in a great increase in the Hh signaling output and results in early lethality (Goodrich et al., 1997). To determine whether Gli1, or Gli2, contributes to the embryonic phenotypes seen in Ptc–/– embryos, we removed each gene from Ptc mutants. In the spinal cord of Ptc mutants, activation of the Shh signaling pathway results in an expansion of ventral CNS cell fates dorsally. lacZ inserted into the Ptc mutant allele (Goodrich et al., 1997) can be used as a read-out of Shh signaling, as Ptc can be activated by Shh and lacZ is under the control of the endogenous Ptc promoter. In wild-type embryos at E8.5, Ptc-lacZ is expressed in the ventral CNS (Fig. 4A), similar to Gli1-lacZ. As expected, in Gli1zfd/zfd;Ptc+/– embryos, the expression of Ptc-lacZ remained the same as in wild-type embryos (Fig. 4B). By contrast, in Gli2zfd/zfd;Ptc+/– embryos at E8.5, the expression of Ptc-lacZ was greatly reduced (Fig. 4C), suggesting Gli2 is normally required for most aspects of Ptc activation.

Fig. 4.

Gli2, but not Gli1, is required for Shh signaling in Ptc mutants. X-gal staining of Ptc-lacZ in six- to eight-somite E8.5 embryos. In wild-type embryos, Ptc-lacZ is expressed in the ventral CNS and somites (A). The expression of Ptc-lacZ is not altered in Gli1zfd/zfd;Ptc+/– embryos (B), but the expression is downregulated in Gli2zfd/zfd;Ptc+/– embryos (C). Loss of Ptc function results in upregulation of Ptc-lacZ throughout the embryos (D). Removal of Gli1 function in Ptc–/– embryos does not rescue the overexpression of Ptc-lacZ (E). However, removal of Gli2 function in Ptc–/– embryos (F) reduces the overexpression of Ptc-lacZ in the head (arrow) and the trunk mesoderm (arrowhead). Scale bar: 0.5 mm.

We then examined embryos lacking Ptc and a specific Gli gene. Consistent with activation of the Shh pathway in Ptc mutants, Ptc mutant embryos showed an elevated level of Ptc-lacZ expression throughout the embryo (Fig. 4D) (Goodrich et al., 1997). In addition, no Ptc–/– mutants were detected after E9.5. In Gli1zfd/zfd;Ptc–/– embryos, Ptc-lacZ was still expressed throughout the embryo, just as in Ptc mutant embryos (Fig. 4E), consistent with persistent activation of the Shh pathway. Furthermore, no Gli1zfd/zfd;Ptc–/– embryos were recovered after E9.5 when Ptc mutants die (data not shown). By contrast, in Gli2zfd/zfd;Ptc–/– embryos, the expression of Ptc-lacZ was reduced compared with Ptc mutant embryos, in particular in the forebrain and trunk mesoderm (Fig. 4F).

To better assay for a possible rescue effect of removing Gli2, we determined the frequency of all genotypes in Gli2zfd/+;Ptc+/– intercrosses at E10.5 and E11.5. As expected, no Ptc–/– embryos were recovered from two litters of 20 embryos at E10.5. By contrast, five Gli2zfd/zfd;Ptc–/– embryos were present in the two litters at E10.5, although the mutants had exencephaly (Fig. 5A-C). In 20 embryos at E11.5, no Gli2zfd/zfd;Ptc–/– embryos were recovered. Ptc-lacZ expression in the heads of the E10.5 embryos was also examined. In wild-type embryos at E10.5, Ptc-lacZ is expressed in the ventral CNS close to the source of Shh (Fig. 5D). As expected, in Gli2zfd/zfd;Ptc+/– embryos at E10.5 Ptc-lacZ expression was severely downregulated (Fig. 5E). However, although Gli2zfd/zfd can rescue the early lethality of Ptc–/– embryos to E10.5, Ptc-lacZ was nevertheless upregulated in the brains of Gli2zfd/zfd;Ptc–/– embryos. Furthermore, the dorsal brain failed to close in Gli2zfd/zfd;Ptc–/– embryos, similar to the phenotype in Gli3–/– mutants, which also have a dorsal activation of the Shh signaling pathway (Fig. 5F).

Fig. 5.

Gli2 can only partially rescue the Ptc mutant phenotype at E10.5. Ptc+/– (A,D), Ptc+/–;Gli2zfd/zfd(B,E) and Ptc–/–;Gli2zfd/zfd(C,F) embryos. Ptc–/–;Gli2zfd/zfd embryos survive beyond E10.5, whereas Ptc–/– embryos die by E9.5. At E10.5, even though the spinal cords are closed, Ptc–/–;Gli2zfd/zfd embryos show exencephaly (indicated by a white arrow). (D-F) X-gal staining of Ptc-lacZ. The expression of Ptc-lacZ is reduced in Ptc+/–;Gli2zfd/zfd embryos (E, indicated by an arrow), when compared with Ptc+/– embryos (D). In Ptc–/–;Gli2zfd/zfd embryos at E10.5, Ptc-lacZ expression is upregulated in the brain.

To examine the expression of Shh targets in Gli2zfd/zfd;Ptc–/– embryos, we analyzed dorsal/ventral (DV) patterning of the spinal cord in Gli2zfd/zfd;Ptc–/– embryos as the spinal cord closes normally in these mutants. In Ptc–/– embryos at E9.5, Shh, Nkx2-2 and Isl1/2 were ectopically expressed in the spinal cord, resulting in overlapping expression of the three proteins in many ventral cells (Fig. 6A-D). In wild-type embryos at E10.5, Shh is expressed in the floorplate cells, and Nkx2-2 is expressed in adjacent V3 interneurons (Fig. 6E,F). More laterally, Isl1/2 is expressed in motoneurons and Pax6 is expressed at highest levels in the intermediate spinal cord (Fig. 6G,H). By contrast, in Gli2zfd/zfd embryos at E10.5, Shh is not expressed in the ventral spinal cord, Nkx2-2 is greatly reduced and Isl1/2-expressing motoneurons occupy the ventral spinal cord (see Fig. 7A-D) (Matise et al., 1998). In the rescued Ptc–/– embryos at E10.5 in which Gli2 was removed, Shh was no longer expressed in the ventral spinal cord, even though it was still detected in the notochord (Fig. 6I). The number of V3 interneurons was greatly reduced compared with Ptc–/– embryos, and only a small number of V3 interneurons remained specifically in the ventral midline of the spinal cord (Fig. 6J). Interestingly, Isl1/2-expressing motoneurons occupied the majority, but not all, of the ventral midline in Gli2zfd/zfd;Ptc–/– embryos (Fig. 6K). The expression of these three markers is similar to that of Gli2 mutants. However, unlike Gli2 mutants, more motoneurons were generated in the double mutants such that the motoneuron populations expanded into the dorsal half of the spinal cord. In addition, Pax6 was expressed at highest levels in the dorsal half of the spinal cord, instead of ventrally as in Gli2 mutants (Fig. 6L). Therefore, loss of Gli2 in Ptc–/– mutants results in a downregulation or loss of some Shh targets that are ectopically activated in the spinal cord in Ptc mutants. Taken together, these studies demonstrate that Gli2 is the primary transcriptional regulator of Shh signaling in Ptc mutants, whereas Gli1 is not.

Fig. 6.

The ventral spinal cord phenotype of Ptc mutants can be partially rescued by removing Gli2 function. In Ptc mutants that survive to E9.5, ventral spinal cord markers expand dorsally (A-D) compared with wild-type embryos a day later at E10.5 (E-H). Insert in A is a more anterior spinal cord section. In Ptc;Gli2 double homozygous mutants, Shh expression is lost in the spinal cord (I), Nkx2-2 expression is greatly reduced in the ventral midline (J), and the expansion of Isl1/2 and Pax6 seen in Ptc–/– mutants are shifted more ventrally (K,L). White arrowhead indicates ventral limit of Pax6 domain. Scale bar: 78 μm in A-D; 100 μm in E-L.

Fig. 7.

Endogenous Gli1 is not required when the Shh pathway is activated by ectopic Gli1. In Gli2 mutant spinal cords (A-D), Shh expression cannot be detected, Nkx2-2 expression is greatly reduced in the ventral midline, and the expression domains of Isl1/2 and Pax6 are shifted ventrally. When Gli1 is expressed from the endogenous Gli2 locus in Gli21ki/1ki embryos, the Gli2 mutant defects are rescued and Shh signaling is restored (E-H). Removal of endogenous Gli1 does not alter ventral spinal cord patterning in Gli1lz/lz;Gli21ki/1ki embryos (I-L). Arrowhead indicates ventral limit of Pax6 expression domain. Scale bar: 50 μm.

Endogenous Gli1 is not required in embryos in which Shh signaling is transduced through Gli1

The Shh signaling pathway can be artificially activated by loss of Ptc function, or it can be activated by ectopic expression of Gli1. For example, when Gli1 is ectopically expressed in transgenic mice or frog embryos, Shh target genes are ectopically activated and over proliferation occurs (Hynes et al., 1997; Lee et al., 1997; Park et al., 2000). Moreover, ectopic expression of Gli1 in postnatal mouse skin can induce various tumors, including basal cell carcinomas and trichoepitheliomas (Nilsson et al., 2000). Recently, it was shown that in frog, transient activation of the pathway using human GLI1 mRNA requires endogenous Gli1 for hyperproliferation of the CNS to be induced (Dahmane et al., 2001). We therefore tested the requirement for endogenous Gli1 in mouse embryos in which ectopic Gli1 expressed from the Gli2 locus (in place of Gli2) activates the Shh pathway.

We have previously shown that when Gli1 is expressed from the Gli2 locus at normal physiological levels, Gli1 can mediate Shh signaling in place of Gli2, and patterning of the neural tube is normal (Bai and Joyner, 2001). To test whether activation of the endogenous Gli1 gene is required for normal patterning in these mice, we removed endogenous Gli1 in these knock-in mice (Gli21ki/1ki) and analyzed patterning of the spinal cord. When Gli1 replaces endogenous Gli2 in Gli21ki/1ki embryos, the floor plate defects in Gli2zfd/zfd embryo are rescued, and Shh, HNF3β, Nkx2-2 and Isl1/2 are expressed normally (compare Fig. 7A-D with 7E-H) (Bai and Joyner, 2001). If endogenous Gli1 is required for this artificial Shh signaling, then removal of Gli1 in these embryos should disrupt patterning of the ventral spinal cord. However, we found that this is not the case, because in Gli1lz/lz;Gli21ki/1ki embryos, patterning of the ventral spinal cord remained intact, as indicated by the domains of Shh, Nkx2-2 and Isl1/2 expression (Fig. 7I-K). Overall DV patterning of the spinal cord also appears normal, as indicated by Pax6 expression (Fig. 7L).


Endogenous Gli1 is not required for normal Shh signaling or development

In the present study, we undertook a number of genetic approaches to address the normal functions of Gli1 in mouse development. First, we created a lacZ knock-in reporter allele of Gli1 that removes all Gli1 protein, and definitively show that mice lacking Gli1 function have a normal floorplate, are viable and show no obvious defects. We also compared the phenotype of the Gli1lz allele with our Gli1zfd allele that could be hypomorphic, in combination with a loss of one or two copies of Gli2. We found that these two alleles function indistinguishably and thus both act as null alleles. Second, we monitored the expression of the Shh target gene, Ptc-lacZ, in normal and Gli1 mutant embryos, and found that Gli1 is not required for Shh target gene expression. We further tested the requirement of endogenous Gli1 in an artificial situation in which normal Shh signaling is transduced through Gli1 instead of Gli2, and found that endogenous Gli1 is not required for such Shh signaling. Together, our studies demonstrate that mouse Gli1 is not required for embryonic development and normal Shh signaling, unless one copy of Gli2 is defective.

Gli1 can not transduce the initial Shh signaling in tissues

Despite our finding that Gli1 is not required in mouse development, Gli1 is capable of transducing Shh signaling. For example, Gli1 can activate a reporter construct containing tandem repeats of Gli-binding sites and an Hnf3b promoter in transfected cells (Dai et al., 1999; Sasaki et al., 1997). In addition, ectopic expression of Gli1 in the CNS of transgenic mouse embryos, or RNA injected frog embryos also induces expression of Shh targets such as Hnf3b and Ptc (Hynes et al., 1997; Lee et al., 1997; Park et al., 2000). Finally, Gli1 is able to transduce Shh signaling when it is expressed in place of Gli2 from the Gli2 locus (Bai and Joyner, 2001).

One possible reason why Gli1 is dispensable during mouse development is that Gli1 is simply a target of the Shh signaling pathway and thus acts as a read-out of Shh signaling. Alternatively, it is possible that Gli1 normally transduces some Shh signaling, but in Gli1 mutant embryos, this function is compensated for by other Gli genes. If Gli1 transduces Shh signaling, then Gli1 protein would need to be present when Shh is first expressed. We therefore examined the expression of Gli1 in the absence of Shh. Of significance, no Gli1 transcription was activated in the neural tube or limbs of E8.5 and E10.5 Shh mutant embryos. Thus, Gli1 is not normally present to transduce the initial Shh signaling. Two other Gli genes, Gli2 and/or Gli3, could transduce the initial Shh signaling, as both genes are expressed at the time Shh expression is initiated (Fig. 3I) (Hui et al., 1994). In support of Gli2 transducing the initial Shh signaling, we showed that transcription of the Shh target gene Ptc is greatly reduced in Gli2 mutant embryos. Our finding that Gli1lz/lz;Gli2zfd/+ mutant embryos have defects in Shh signaling (floorplate development), shows that Gli1 nevertheless normally contributes to propagation of Shh signaling once it is expressed.

Gli2 and not Gli1 is required to transduce ectopic Shh signaling in Ptc mutants

Ectopic expression of Gli1 can cause over proliferation of the dorsal neural tube and skin cancers. As many Shh target genes are activated in these situations, which resemble Shh gain-of-function phenotypes, we tested whether Gli1 or Gli2 contribute to ectopic Shh signaling in Ptc mutants. Since Ptc is a negative regulator of the Shh signaling pathway, loss of Ptc function results in ectopic activation of the Shh signaling pathway and transcriptional activation of Gli1.

We found that Ptc;Gli1 double homozygous mutants have the same phenotype as Ptc mutants, showing that Gli1 does not contribute a critical function to the ectopic Shh signaling in Ptc mutants. On the contrary, Ptc;Gli2 double mutants showed a partial rescue phenotype when compared with Ptc mutants. The partial rescue phenotype included: (1) double mutant embryos survived a day longer than Ptc–/– embryos; (2) the floorplate marker Shh, which is expressed ectopically in the ventral spinal cord of Ptc mutants, was not expressed in double mutant embryos; and (3) the ventral interneuron marker Nkx2-2, which is expressed throughout the Ptc mutant spinal cord, is greatly reduced to a small patch in the ventral midline in double mutants. As Ptc;Gli2 double mutant embryos have a ventral spinal cord phenotype similar to Gli2 single mutants, this suggests that the function of the Shh signaling pathway in the ventral spinal cord of Ptc mutants is mediated primarily by Gli2. Given that the processing of Gli2 into an N-terminal repressor form is independent of Shh signaling, at least in fly embryos (Aza-Blanc et al., 2000), and that only the activator function of Gli2 is required during mouse embryonic development (Bai and Joyner, 2001), our studies provide further evidence that a primary function of Shh in development of the ventral spinal cord is to potentiate the activator function of Gli2.

Removal of Gli2 in Ptc mutants does not completely rescue the Ptc mutant phenotypes. After an initial reduction of Ptc-lacZ expression in the brain and trunk mesoderm of six- to eight-somite embryos at E8.5 (Fig. 4F), more Ptc-lacZ expression is seen by the 10-12 somite stage (data not shown). By E10.5, Ptc-lacZ is expressed throughout the brain of double mutants (Fig. 5F). One possible explanation for this upregulation of Ptc-lacZ is the inhibition of the formation of a Gli3 repressor in Ptc mutant embryos. As one role of Shh signaling is to prevent the processing of Gli3 into a repressor form (Wang et al., 2000; Litingtung and Chiang, 2000), activation of the Shh pathway in Ptc mutants should result in loss of the Gli3 repressor. Indeed, the exencephaly phenotype of Ptc;Gli2 double mutant embryos resembles the phenotype of Gli3 mutants. In addition, contrary to our finding that some ventral cell types are not induced in Ptc mutants when Gli2 is removed, there are more motoneurons in the double mutant spinal cords than in wild-type or Gli2 mutants at E10.5. Gli3 mutants have normal spinal cords, indicating that removal of the Gli3 repressor alone does not lead to an expansion of the motoneuron pool. It is possible, however, that the combination of loss of the Gli3 repressor and persistence of full-length Gli3 protein due to ectopic activation of the Shh pathway in Ptc mutants results in motoneurons developing at more dorsal levels than normal, in addition to the exencephaly phenotype. Our results therefore suggest that altered Gli3 activity, in addition to Gli2, contributes to the Ptc–/– mutant defects.


We are grateful to Sohyun Ahn for the Shh RNA in situ hybridizations shown in Fig. 3A-C, to Rada Norinskaya for technical assistance and to the NYUSoM transgenic/ES cell chimera facilities for making the chimeric mice. We thank Sohyun Ahn and Mark Zervas for comments on the manuscript. The monoclonal antibodies developed by the Jessell laboratory (Shh, Nkx2-2, Isl1/2, Pax6) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences. This work was supported by an NIH postdoctoral fellowship to C. B. B. and NIH grants to A. L. J., who is an investigator of the Howard Hughes Medical Institute.


    • Accepted July 8, 2002.


View Abstract