Sonic Hedgehog-induced Activation of the Gli1Promoter Is Mediated by GLI3*

Drosophila transcription factorcubitus interruptus (Ci) and its co-activator CRE (cAMP response element)-binding protein (CBP) activate a group of target genes on the anterior-posterior border in response to hedgehog protein (Hh) signaling. In the anterior region, in contrast, the carboxyl-truncated form of Ci generated by protein processing represses Hh expression. In vertebrates, three Ci-related transcription factors (glioblastoma gene products (GLIs) 1, 2, and 3) were identified, but their functional difference in Hh signal transduction is unknown. Here, we report distinct roles for GLI1 and GLI3 in Sonic hedgehog (Shh) signaling. GLI3 containing both repression and activation domains acts both as an activator and a repressor, as does Ci, whereas GLI1 contains only the activation domain. Consistent with this, GLI3, but not GLI1, is processed to generate the repressor form. Transcriptional co-activator CBP binds to GLI3, but not to GLI1. Thetrans-activating capacity of GLI3 is positively and negatively regulated by Shh and cAMP-dependent protein kinase, respectively, through a specific region of GLI3, which contains the CBP-binding domain and the phosphorylation sites of cAMP-dependent protein kinase. GLI3 directly binds to theGli1 promoter and induces Gli1 transcription in response to Shh. Thus, GLI3 may act as a mediator of Shh signaling in the activation of the target gene Gli1.

Hedgehog protein (Hh) 1 plays an important role in the pattern formation of many species including insects and vertebrates (for review, see Refs. 1 and 2). The hh gene was originally identified as a segment polarity gene in Drosophila. Hh is a secreted and diffusible protein and is a critical signaling molecule for the pattern formation of the anterior-posterior (A-P) axis. Hh is expressed only in the posterior compartment and acts locally at the compartment border to induce the expression of border-specific genes in anterior cells, such as decapentaplegic (dpp), wingless (wg), and gooseberry (gsb) (3)(4)(5)(6)(7)(8).
The zinc finger-containing protein (Ci) encoded by the segment polarity gene cubitus interruptus (ci), which has a homology with the vertebrate family of glioblastoma (Gli) transcription factors (9), mediates Hh-induced transcriptional activation (10). On the other hand, in the anterior region, located far from the A-P border, Ci is processed into a C-terminal truncated form that functions as a repressor of hh expression (11,12). Hh signaling suppresses this processing on the A-P border. Also on the A-P border, extracellular Hh binds to the transmembrane protein Patched (Ptc), preventing its normal inhibition of Smoothened (Smo), another transmembrane protein of the receptor complex (13)(14)(15)(16). This allows Smo to transduce a signal to Ci through the positive regulator Fused (Fu), which is a serine-threonine kinase (17). Hh signaling also negates the inhibitory effects of Costal2 (Cos2), a kinesin-related protein (18,19), cAMP-dependent protein kinase (PKA) (20 -24), and Suppressor of fused (Su(fu)) (25). Ci, Fu, Su(fu), and Cos2 are normally found in a complex that docks at microtubules, and tight binding of this complex to microtubules is prevented by Hh signaling (18,19,25) These facts suggest that Hh signaling prevents association of the complex with the cytoskeleton and cleavage of Ci into its repressor form.
The most studied member of the hh multigene family in vertebrates is Sonic hedgehog (Shh). Targeted disruption of mouse Shh leads to multiple defects in embryonic tissues, including the notochord, floor plate, and limb structures (26). The mechanism of Shh signaling is thought to resemble to that of Drosophila hh. For instance, the negative regulation of Shh signaling by PKA was also observed in vertebrate embryos (27). It was further demonstrated that suppression of PKA activity is sufficient to activate targets of the Shh signaling pathway in the vertebrate central nervous system (28). The three vertebrate genes Gli1, Gli2, and Gli3 have a homology with ci (29,30). All three GLI proteins bind to the consensus sequence 5Ј-TGGGTGGTC-3Ј through their metal finger regions (31,32). Gli1 is downstream of Shh, and ectopic Shh signaling induces its expression, whereas expression of Gli3 is down-regulated by ectopic Shh signaling (33)(34)(35). In fact, Gli1 and Ptc, both of which are downstream of Shh, are expressed in similar domains in diverse regions of developing mouse embryo (36), and ectopic expression of Gli1 in Xenopus and mouse embryos induced the expression of midline neural plate markers, such as Hnf-3␤ (34,35), indicating that GLI1 is a positive regulator of Shh signaling. As in the case of Ci, which acts as a repressor in the anterior compartment, GLI3 also represses Shh expression in the anterior region of the limb bud (37). Moreover, GLI1 is an activator of transcription, whereas GLI3 represses transcription of an Hnf-3␤ enhancer element (38,39). Although these facts suggest functional differences between GLI1 and GLI3 proteins, their precise roles in Shh signaling remain unclear.
Translocations and mutations of the human Gli3 gene cause abnormal pattern formation through haploinsufficiency in humans and mice. The associated human syndromes are known as Greig cephalopolysyndactyly syndrome (GCPS) (40) and Pallister-Hall syndrome (PHS) (41), and the mouse mutants such as extra-toe (Xt) have the mutations in the Gli3 gene (42). Mutations or deletions of the human Cbp gene at chromosome 16p13.3 also cause the Rubinstein-Taybi syndrome (RTS) through haploinsufficiency (43). RTS consists of a wide variety of developmental defects, including craniofacial malformations, broad thumbs, broad big toes, and mental retardation (44). Disruption of one copy of the mouse Cbp gene by gene targeting causes skeletal abnormality partially resembling that of RTS (45). CBP was originally identified as a transcriptional coactivator of CREB (46,47) and has recently been found to be required for many other transcription factors (for review, see Ref. 48). RTS and GCPS are distinct syndromes, but both include preaxial limb anomalies and craniofacial features. A recent genetic study of Drosophila CBP mutants indicated that CBP is used as a co-activator of Ci and plays an important role in hh signaling (49). The structural homology of Ci and GLI3 supports the idea that CBP acts as a co-activator of GLI3 in vertebrates. As described above, however, GLI3 was reported to be a repressor of Shh transcription, not an activator of Shh targets.
To investigate the role of CBP in vertebrate pattern formation, we have examined whether CBP directly interacts with human GLI3 and GLI1. Our results indicate that GLI3 and GLI1 have multiple differences in their domain structures including the CBP-binding domain. CBP binds only to GLI3. Furthermore, a specific region in the GLI3 protein that includes the CBP-binding domain can mediate Shh-induced trans-activation. We have also demonstrated that GLI3 directly binds to the mouse Gli1 promoter and mediates Shhinduced activation of the Gli1 promoter in Shh-responsive multipotential neural stem (MNS)-70 cells. Thus, our results demonstrate distinct roles for GLI3 and GLI1 in Shh signaling.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The human Gli1 and Gli3 cDNA was kindly provided by Drs. Kenneth W. Kinzler and Bert Vogelstein. The plasmids used for in vitro transcription/translation of the various forms of human GLI3 were generated by inserting various fragments of GLI3 cDNA, which were prepared by the polymerase chain reaction (PCR)based method, into the NcoI-XbaI site of pSPUTK (Stratagene). For all constructs generated by PCR, it was confirmed by sequencing that they do not contain mutations. The pA10CAT6GBS reporter plasmid containing the GLI-binding sites was constructed by inserting six tandem copies of the GLI-binding site (5Ј-GCGTGGACCACCCAAGAC-GAAATT-3Ј) (32) into the BglII site of pA10CAT2, in which the SV40 early promoter was linked to the CAT gene (50). The plasmids pSR␣-GLI3 and pSR␣-GLI1 to express human GLI3 and GLI1 in cultured cells were made by placing the Gli3 and Gli1 cDNA downstream of the SR␣ promoter, respectively (51). To generate the plasmids pact-Flag-GLI3 and pact-Flag-GLI1 encoding the Flag-linked GLI3 and GLI1, the synthetic oligonucleotides containing one copy of the Flag sequence were inserted downstream of the chicken cytoplasmic ␤-actin promoter and fused in frame to the N terminus of GLI3 and GLI1, respectively. The plasmids to express Gal4-GLI3 and Gal4-GLI1, in which various portions of GLI3 and GLI1 were fused to the DNA-binding domain of Gal4 (amino acids 1-147), were constructed by the PCR-based method using the CMV promoter-containing vector. The two Gal4-GLI3 fusions containing the N-terminal region or the CBP-binding domain contained amino acids 1-397 and 827-1132 of GLI3, respectively. The luciferase reporter plasmid containing the mouse Gli1 promoter was generated by inserting the 3.5-kilobase HindIII-EcoRI fragment containing the Gli1 promoter upstream of the luciferase gene in the plasmid pGL3-basic vector (Promega). The Gli1 promoter fragment 2 was kindly provided by Drs. Heidi Park and Alexandra L. Joyner. The mutant Gli1 promoter, in which the eight GLI3-binding sites were disrupted, was constructed by the PCR-based method. The plasmids to express CBP and E1A were described previously (52). The Shh expression plasmid pJT4/Shh and the Gal4 site-containing luciferase reporter, in which three tandem repeats of Gal4 sites were linked to the thymidine kinase promoter, was kindly provided by Drs. Sumihare Noji, Kazuhiko Umesono, and Ronald M. Evans, respectively.
GST Pull-down Assay and Co-immunoprecipitation-The GST pulldown assay using the GST-CBP and the in vitro translated GLI3 was performed as described (52).
For co-immunoprecipitation, a mixture of 5 g of the CBP expression plasmid pSR␣-CBP and 5 g of the Flag-linked GLI3 expression plasmid pact-Flag-GLI3 or the Flag-linked GLI1 expression plasmid pact-Flag-GLI1 were transfected into 293T cells. Forty hours after transfection, cells were lysed in the lysis buffer (50 mM Hepes, pH 7.5, 250 mM NaCl, 0.2 mM EDTA, 10 mM NaF, 0.5% Nonidet P-40), and whole cell lysates were prepared. Lysates were immunoprecipitated using anti-CBP antibodies CT (Upstate Biotechnology) and A-22 (Santa Cruz), and the immune complexes were separated on 7% SDS-polyacrylamide gels and analyzed by Western blotting using anti-Flag antibodies (Upstate Biotechnology) and ECL detection reagents (Amersham Pharmacia Biotech).
Transient Co-transfection Assays-The amount of each plasmid DNA used for transfection is described in the legends to the figures. The co-transfection experiments and CAT assays were performed as described (52). Dual luciferase assays were done as described by the supplier.
Examination of Proteolysis of GLI1 and GLI3 Proteins-Extracts were prepared from mouse embryos at 10.5 days post-coitus as described (45), and separated on 7 and 10% gels to detect GLI3 and GLI1, respectively. Western blotting was performed using anti-GLI3 and anti-GLI1 antibodies (Santa Cruz). The anti-GLI3 antibody that recognizes the N-terminal region of GLI3 was further purified by using the GST-GLI3 affinity resin.
To examine the processing of GLI3 in cultured cells, 4 g of the N-terminal Flag-linked GLI3 expression plasmid, pact-Flag-GLI3, or the GLI1 expression plasmid, pSR␣-GLI1, was transfected into 293T cells with or without 1 g of the Shh expression plasmid and 1 g of the plasmid to express the catalytic subunit of PKA. Forty hours after transfection, cells were lysed in the lysis buffer (50 mM Hepes, pH 7.5, 250 mM NaCl, 0.2 mM EDTA, 10 mM NaF, 0.5% Nonidet P-40), and whole cell lysates were prepared. Lysates were separated on 7% SDSpolyacrylamide gels, and analyzed by Western blotting using anti-Flag antibodies (Upstate Biotechnology), anti-GLI1 antibodies (Santa Cruz), and ECL detection reagents (Amersham Pharmacia Biotech).
Induction of Endogenous Gli1 Gene Expression by Shh in MNS-70 Cells-MNS-70 cells were transfected with a mixture of 4 g of the Shh expression plasmid pJT4/Shh and 4 g of the plasmid to express the GLI3-binding domain of CBP (amino acids 454 -718) or GLI3, or the control plasmid lacking the cDNA to be expressed. Forty hours after transfection, total RNA was prepared using TRIZOL (Life Technologies, Inc.), and reverse transcription-PCR was performed as described by Sasaki et al. (38) to measure Gli1 mRNA and cytoplasmic ␤-actin mRNA as a control.

RESULTS
Binding of CBP to GLI3 but not to GLI1-We first examined whether human GLI3 and GLI1 could interact directly with mouse CBP in vitro (Fig. 1A). The full-length form of human GLI3 was in vitro translated and mixed with the GST-CBP resin containing the 265-amino acids region of CBP, which is responsible for binding to multiple transcription factors, including phospho-CREB, c-Jun, and c-Myb (46,52,53). GLI3 efficiently bound to GST-CBP but not to the GST resin alone (Fig. 1A, left panel). PKA treatment of GLI3 did not increase the efficiency of binding of GLI3 to CBP. To examine the possibility that the in vitro translated GLI3 is already phosphorylated by some kinase(s), we treated the in vitro translated GLI3 with potato acid phosphatase. The phosphatase treatment did not affect the binding of GLI3 to CBP. In contrast to GLI3, in vitro translated GLI1 protein did not bind to CBP in the GST pull-down assay using the same GST-CBP resin (Fig.  1A, right panel). GLI1 also did not bind to the GST-CBP fusions containing various other regions of CBP, which covered the entire CBP molecule (data not shown). These results indicate that CBP binds to GLI3 in a phosphorylation-independent manner, but not to GLI1.
A possible in vivo interaction between GLI3 and CBP was investigated by co-immunoprecipitation (Fig. 1B). Whole cell lysates were prepared from 293T cells transfected with the two plasmids to express CBP and Flag-linked GLI3, and incubated with anti-CBP antibodies or a control antibody, anti-␤-galactosidase. The immunoprecipitates were separated on an SDSpolyacrylamide gel and the N-terminal Flag-linked GLI3 was detected by Western blotting using the anti-Flag antibody. A significant amount of Flag-GLI3 was co-precipitated with the CBP C-terminal region-specific antibody (CT), whereas the CBP N-terminal region-specific antibody (A-22) co-precipitated only a small amount of Flag-GLI3. The low efficiency of coimmunoprecipitation with the A-22 antibody could be due to the proximity of the epitope for this antibody with the GLI3binding domain in the CBP molecule. The control antibody, anti-␤-galactosidase, did not co-immunoprecipitate Flag-GLI3.
To investigate which domain of GLI3 interacts with CBP, we used a series of deletion mutants of GLI3 in the in vitro binding assay (Fig. 1C). Various forms of GLI3 were synthesized using an in vitro translation system, mixed with the GST-CBP affinity resin, and the bound proteins were analyzed. The results showed that the 306-amino acid region between amino acids 827 and 1132 in the C-terminal half of GLI3 bound to CBP with almost the same efficiency as full-length GLI3.
By using a series of GST-CBP proteins containing various portions of the CBP molecule, we confirmed that regions other than the N-terminal region containing the CREB-binding domain (KIX: the region in CBP molecule that binds to the kinase-inducible activation domain (KID) of CREB) did not bind to GLI3 (data not shown). This region of CBP is known to directly interact with multiple transcription factors, such as CREB and c-Myb (52). To map the GLI3-binding domain more precisely, we used three additional GST-CBP resins for the GST-pull down assay (Fig. 1D). This revealed that the region GLI3 to CBP was similarly analyzed with (ϩAPase) or without (ϪAPase) treatment of [ 35 S]GLI3 by potato acid phosphatase. The amount of GLI3 protein bound to GST-CBP was about 18% of the input protein, whereas the GST control resin bound to less than 1% of the input protein (data not shown). B, co-immunoprecipitation. Whole cell lysates were prepared from 293T cells transfected with the two plasmids to express CBP or the Flag-linked GLI3, and used for immunoprecipitation with the anti-CBP antibody CT, A-22, or the control antibody anti-␤-galactosidase. The immune complexes were analyzed by Western blotting using the anti-Flag antibody. In the left lane (Western), an aliquot of whole cell lysate was directly used for Western blotting. C, identification of the CBP-binding domain in GLI3. The structures of the various forms of GLI3 used are shown below. The GST pull-down assays were performed as described above, and the results of binding assays shown below are indicated on the right. The binding activities of the mutants are designated ϩ and Ϫ, which indicate the binding of 15-20% and 1% of the input proteins, respectively. In the input lanes, various forms of GLI3 indicated above each lane were synthesized in vitro and analyzed by 10% SDS-PAGE. Various forms of GLI3 synthesized were mixed with the GST-CBP resin, and the bound proteins are shown on the right half (Bound to CBP) of the bottom panel. The amount of protein in the input lanes was 10% of that used for the binding assay. The lanes for the 827/1132 protein were exposed for a longer time than the other lanes, because the efficiency of in vitro translation of this protein was relatively low. D, identification of the GLI3-binding domain in CBP. The domains of CBP are indicated at the top: C/H, cysteine-and histidine-rich domain; KIX, CBP-binding domain; Bromo, bromo domain. The structures of the various forms of CBP that were reported previously (52) are shown below. The GST pull-down assays were performed as described above, and the results of the binding assays shown below are indicated on the right. The binding activities of the mutants are designated ϩ and Ϫ, which indicate the binding of 15-20% and 1% of the input proteins, respectively. The results of binding assays using CREB and c-Myb to each GST-CBP resin, which were previously reported (52), are also indicated. GLI3 Has the Activation and Repression Domains-To examine whether CBP acts as a co-activator of GLI3, we then performed CAT co-transfection experiments using mouse fibroblast NIH3T3 cells (Fig. 2, A-C). The plasmid pA10CAT6GBS, in which six tandem repeats of the GLI-binding site were linked to the SV40 early promoter, was used as a reporter. Co-transfection of this reporter with a GLI3 expression vector into NIH3T3 cells increased the level of CAT activity in a dose-dependent manner ( Fig. 2A). Increasing amounts of the CBP expression plasmid were then co-transfected with the reporter in the presence or absence of 6 g of the GLI3 expression plasmid (Fig. 2B). In the absence of CBP expression plasmid, 6 g of the GLI3 expression plasmid activated promoter activity 2.5-fold. Co-transfection of increasing amounts of the CBP expression plasmid increased the degree of trans-activation up to a maximum of 8.2-fold (obtained with 16 g of the CBP expression plasmid in the presence of the GLI3 expression plasmid). In the control experiment without GLI3 expression plasmid, addition of the CBP expression plasmid did not affect the level of CAT activity at all. Thus, CBP potentiates GLI3induced trans-activation. Consistent with this, adenovirus wild type 12SE1A and its mutant form ⌬121-150, which is unable to interact with any retinoblastoma-family protein but retains the capacity to bind to CBP, inhibited GLI3-induced trans-activation, but another mutant ⌬30 -85, which is unable to interact with CBP/p300, did not (Fig. 2C).
The results of co-transfection experiments using various GLI3 deletion mutants with the GLI site-containing reporter suggested that the N-terminal region of GLI3 contains the transcriptional repression domain, whereas the CBP-binding domain mediates transcriptional activation (data not shown). To confirm this, we made the two Gal4-GLI3 fusions, which consisted of the Gal4 DNA-binding domain and the N-terminal region or CBP-binding domain of GLI3, and examined their capacity to modulate luciferase expression from the Gal4 sitecontaining luciferase reporter (Fig. 2D). The Gal4 fusion containing the N-terminal 397 amino acids (Gal4-N) repressed luciferase expression to 35% of the level of the control Gal4 DNA-binding domain alone (Fig. 2D, cf. lanes 1 and 2), whereas the Gal4 fusions containing the CBP-binding domain (Gal4-CBD) stimulated luciferase expression 5-fold. The expression levels of these two fusion proteins were similar (data not shown). Co-expression of CBP enhanced trans-activation by the CBP-binding domain (Fig. 2D, cf. lanes 5 and 6) but did not affect the function of the N-terminal repression domain (cf. lanes 2 and 3). E1A inhibited CBD-dependent trans-activation (cf. lanes 5 and 7). These results indicated that GLI3 has both FIG. 2. Presence of the activation and repression domains in GLI3. A, transcriptional activation by GLI3. trans-Activation by GLI3 was examined by co-transfection assays in NIH3T3 cells using the CAT reporter containing the GLI-binding sites. A mixture of 0, 3, 6, 9, 12, or 15 g of the GLI3 expression plasmid pSR␣-GLI3, 6 g of the CAT reporter plasmid pA10CAT6GBS, and 1 g of the internal control plasmid pact-␤-galactosidase was transfected into NIH3T3 cells. The total amount of plasmid DNA was adjusted to 22 g by adding the control plasmid DNA pSR␣0 lacking the GLI3-coding region. CAT assays were done as described under "Experimental Procedures." The experiments were repeated three times, and the differences between the experiments were no more than 20%. Typical results are indicated by a bar graph. The shaded bar shows the data obtained in the presence of the GLI3 expression plasmid. B, potentiation of GLI3-induced transactivation by CBP. The effect of CBP on GLI3-induced trans-activation was investigated by further addition of increasing amounts of the CBP expression plasmid. A mixture of 0, 4, 8, 12, or 16 g of the CBP expression plasmid pSR␣-CBP, 5 g of the CAT-reporter plasmid pA10CAT6GBS, 0 or 6 g of the GLI3 expression plasmid pSR␣-GLI3, and 1 g of the internal control plasmid pact-␤-galactosidase was transfected into NIH3T3 cells, and CAT assays were done. The total amount of plasmid DNA was adjusted to 27 g by adding the control plasmid DNA pSR␣0. The experiments were repeated three times, and the average is shown as the relative CAT activity, with S.D. as indicated. C, repression of GLI3-dependent transcriptional activation by E1A. A mixture of 6 g of the CAT-reporter plasmid pA10CAT6GBS, 9 g of the GLI3 expression plasmid pSR␣-GLI3, 1 g of the plasmid to express wild type 12SE1A, ⌬121-150, or ⌬30 -85 mutant, and 1 g of the internal control plasmid pact-␤-galactosidase was transfected into NIH3T3 cells, and CAT assays were done. The total amount of plasmid DNA was adjusted to 20 g by adding the control plasmid DNA pSR␣0. D, activation and repression domains in GLI3. The two Gal4 fusion proteins containing the N-terminal region (Gal4-N) and the CBP-binding domain (Gal4-CBD) of GLI3 are shown below. Co-transfection assays using the plasmid to express these Gal4-GLI3 fusions were done using the luciferase reporter containing Gal4-binding sites. A mixture of 4 g of the Gal4 site-containing reporter plasmid pGal-TK-luc, 0.2 g of the plasmid to express various forms of Gal4-GLI3 fusion or Gal4 DNA-binding domain alone as a control, 3 g of the CBP expression plasmid pSR␣-CBP, 1 g of the E1A expression plasmid, and 1 g of the internal control plasmid pRL-CMV, in which the sea pansy luciferase gene is linked to the CMV promoter, was transfected into NIH3T3 cells, and luciferase assays were done. The total amount of plasmid DNA was adjusted to 10 g by adding the control plasmid DNA. The shaded bar shows the data obtained with the Gal4-GLI3 expression plasmid. E, Shh-responsiveness of GLI3. A mixture of 0, 1, 2, 3, or 4 g of the Shh expression plasmid pJT4/Shh, 4 g of the Gal4 site-containing reporter plasmid pGal4-TK-luc, 0.2 g of the plasmid to express Ga4-GLI3 fusion, which contains the full-length GLI3 or the CBP-binding domain, or Gal4 DNA-binding domain alone as a control, 3 g of the PKA catalytic subunit expression plasmid pSR␣-PKA, and 1 g of the internal control plasmid pRL-CMV was transfected into MNS-70 cells, and the luciferase assay was performed. The total amount of plasmid DNA was adjusted to 12.2 g by adding the control plasmid DNA. The experiments were repeated three times, and the average is shown with S.E. The data indicating the positive regulation by Shh is shown by a shaded bar. the repression and activation domains.
GLI3 Mediates the Shh-induced Transcriptional Activation-The identification of the CBP-binding domain of GLI3 allowed us to examine the effect of Shh and PKA on the capacity of this domain to regulate transcription. For this purpose, we used a neural stem cell line, MNS-70, that is able to express different sets of ventral-specific genes including Isl-1, Nkx-2.1, and Nkx-2.2 in response to Shh (54). The Gal4 site-containing reporter and the plasmid to express the Gal4 fusion protein containing GLI3 were co-transfected into MNS-70 cells with increasing amounts of the Shh expression plasmid in the presence or absence of the PKA catalytic subunit expression plasmid (Fig. 2E). We first examined the effect of Shh and PKA on trans-activation by the Gal4 fusion containing full-length GLI3 (Fig. 2E, left panel). The Gal4-fusion containing full-length GLI3 repressed luciferase expression to 65% of the level of the control Gal4 DNA-binding domain. Co-expression of Shh increased luciferase expression in a dose-dependent manner, whereas PKA suppressed trans-activating capacity. We then performed a similar experiment using a Gal4 fusion with the CBP-binding domain (Fig. 2E, right panel). In the absence of PKA, Shh only slightly enhanced trans-activation by the CBPbinding domain. PKA negatively regulated this trans-activation, and co-expression of Shh restored this trans-activation in a dose-dependent manner in the presence of PKA. Thus, the CBP-binding domain mediates the antagonistic actions of PKA and Shh. The Shh responsiveness of full-length GLI3 in the absence of PKA may suggest that Shh may enhance GLI3 activity through not only the CBP-binding domain but also other region. On the other hand, the trans-repression mediated by Gal4 fusion containing the N-terminal repression domain was affected neither by Shh nor by PKA (data not shown).
Negative Regulation of GLI1 Activity by Shh-To compare the functional domains of GLI3 and GLI1, we next examined the functional domains of GLI1. In the CAT co-transfection experiments using the GLI site-containing reporter and NIH3T3 cells, GLI1 increased the level of CAT activity (Fig.  3A). Consistent with the results of the GST pull-down assays, co-expression of CBP did not potentiate GLI1-induced transactivation (Fig. 3B). To identify the transcriptional activation domain in GLI1, three different portions of GLI1 were fused to the Gal4 DNA-binding domain, and their capacity to regulate transcription was investigated by co-transfection assays using the Gal4 site-containing reporter in MNS-70 cells (Fig. 3C). The expression levels of these three fusion proteins were similar (data not shown). Full-length GLI1 fusion and the Gal4 fusion containing the C-terminal 394 amino acids enhanced luciferase expression (Fig. 3C, left panel). These results indicated that GLI1 has the activation domain in its C-terminal region and does not have a repression domain. The Shh and PKA responsiveness of the Gal4 fusion containing full-length GLI1 was investigated (Fig. 3C, right panel). The full-length GLI1 activity was suppressed by Shh, whereas PKA did not affect GLI1 activity.
Processing of GLI3 but not GLI1-Our results indicated that GLI3 has both activation and repression domains, as does the Drosophila homolog Ci, whereas GLI1 has only the activation domain. Because Ci is processed into a repressor form lacking the activation domain in the anterior compartment (11), we examined GLI3 for similar processing (Fig. 4). Extracts were prepared at 11.5 days post-coitus from mouse embryos, which express both GLI3 and GLI1, and were used for Western blotting (Fig. 4A). An antibody that can recognize the C-terminal region (amino acids 1577-1596) of mouse GLI3 detected 190and 110-kDa proteins. Analysis of the short, 110-kDa form of GLI3 by SDS-PAGE using two in vitro translated N-truncated FIG. 3. Domain structure and Shh responsiveness of GLI1. A, transcriptional activation by GLI1. trans-Activation by GLI1 was examined by co-transfection assays in NIH3T3 cells using the CAT reporter containing the GLI-binding sites. A mixture of 0, 4, or 8 g of the GLI1 expression plasmid pSR␣-GLI1, 4 g of the CAT-reporter plasmid pA10CAT6GBS, and 1 g of the internal control plasmid pact-␤-galactosidase was transfected into MNS-70 cells, and CAT assays were performed. The total amount of plasmid DNA was adjusted to 13 g by adding the control plasmid DNA pSR␣0. The experiments were repeated three times, and the differences between the experiments were no more than 20%. Typical results are indicated by a bar graph. The shaded bars show the data obtained in the presence of the GLI expression vector. B, there was no effect of CBP on GLI1-induced transactivation. The effect of CBP on GLI1-induced trans-activation was investigated by the addition of the increasing amounts of the CBP expression plasmid. A mixture of 0, 4, or 8 g of the CBP expression plasmid pSR␣-CBP, 5 g of the CAT-reporter plasmid pA10CAT6GBS, 0 or 5 g of the GLI1 expression plasmid pSR␣-GLI1, and 1 g of the internal control plasmid pact-␤-galactosidase was transfected into MNS-70 cells, and CAT assays were done. The total amount of plasmid DNA was adjusted to 21 g by adding the control plasmid DNA. The experiments were repeated three times, and the average is shown as the relative CAT activity, with S.D. indicated. C, regulation of GLI1 activity by Shh and PKA. The structure of Gal4-GLI1 fusions used is indicated, and the trans-activating capacity of each construct is shown to the right. Left panel, identification of the activation domain. A mixture of 2 g of the Gal4 site-containing reporter plasmid pGal-TK-luc, 0.3 g of the plasmid to express various forms of Gal4-GLI1 fusion or Gal4 DNAbinding domain alone, and 0.5 g of the internal control plasmid pRL-CMV was transfected into MNS-70 cells, and luciferase assays were done. The total amount of plasmid DNA was adjusted to 5 g by adding the control plasmid DNA. Right panel, inhibition of full-length GLI1 activity by Shh. A mixture of 0, 0.5, 1, or 2 g of the Shh expression plasmid pJT4/Shh, 2 g of the Gal4 site-containing reporter plasmid pGal4-TK-luc, 0.5 g of the plasmid to express Ga4-full-length GLI1 fusion or Gal4 DNA-binding domain alone as a control, 1 g of the PKA catalytic subunit expression plasmid pSR␣-PKA, and 0.5 g of the internal control plasmid pRL-CMV was transfected into MNS-70 cells, and the luciferase assay was performed. The total amount of plasmid DNA was adjusted to 6 g by adding the control plasmid DNA. mutants as markers suggested that this short form was generated probably by processing at a site between amino acids 650 and 750, corresponding to the region between the zinc finger region and the CBP-binding domain. Another fragment generated by this processing should have the N-terminal repressor domain and the zinc finger. In fact, an antibody recognizing the N-terminal region (amino acids 2-20) of GLI3 detected a 100-kDa protein in addition to the full-length 190-, 60-, and 50-kDa proteins. In contrast, the GLI1-specific antibody recognizing the N-terminal region (amino acids 2-17) of GLI1 detected only a 150-kDa protein, which corresponds to the predicted size of full-length GLI1.
To further confirm the processing of GLI3 and also to exam-ine whether Shh signaling prevents the processing of GLI3 as reported in the case of Ci of Drosophila, we tested the processing of GLI3 in cultured cells. The plasmid to express the Nterminal Flag-linked GLI3 protein was transfected into the highly transfectable 293T cells, and the GLI3 proteins expressed from the transfected DNA was analyzed by Western blotting using anti-Flag antibody (Fig. 4B, left panel). When the Flag-GLI3 expression plasmid alone was transfected, only the full-length form of GLI3 was detected, indicating that the processing did not occur. However, when the plasmid to express the catalytic subunit of PKA was co-expressed with Flag-GLI3, the 95-kDa GLI3 protein was generated. The size of this protein is close to but slightly smaller than that (100 kDa) of one of the GLI3 proteins detected in mouse embryonic lysates. This small difference in the molecular mass could be due to retardation of mobility because of high amount of proteins loaded in Fig. 4A, and appears to be within error. The 60-and 50-kDa GLI3 proteins detected in mouse embryonic lysates may be artifactual degradation products, because the similar size of proteins were not detected in the transfected cells. These results suggest that PKA stimulates the processing of GLI3, as in the case of Ci. When the Shh is co-expressed with Flag-GLI3 and PKA, this processing was inhibited. In the similar experiment with the GLI1 expression plasmid and the PKA expression plasmid, only the full-length form of GLI1 was detected, and the processing of GLI1 was not observed (Fig. 4B, right panel). Thus, consistent with their domain structures, GLI3, but not GLI1, is processed to generate a repressor form.
Shh-induced Activation of the Gli1 Promoter Is Mediated by GLI3-It was recently reported that the transcription of the Gli1 gene, but not the Gli3 gene, is induced by ectopic Shh signaling (15,34,35). These reports and our finding that GLI3, but not GLI1, has an Shh-induced activation domain suggested that GLI3 could mediate Shh-induced activation of the Gli1 promoter. To investigate this possibility, we examined whether GLI3 directly binds to the mouse Gli1 promoter region by gel mobility shift assays using the GST-GLI3 fusion protein containing the metal finger region of GLI3 (Fig. 5A). Among six DNA fragments covering the mouse Gli1 promoter region (55), 2 only two DNA fragments, the BamHI-XhoI 210-base pair fragment (fragment A) and the adjacent XhoI-EcoRI 800-base pair fragment (fragment B) containing the two exons that encode the 5Ј-untranslated region of Gli1 mRNA, bound to the GST-GLI3 recombinant protein and generated the specifically retarded bands (Fig. 5A, left panel). Increasing the amount of GST-GLI3 protein generated multiple retarded bands (Fig. 5A,  middle and right panels), suggesting the presence of multiple GLI3-binding sites in these two fragments. In fact, the DNA sequences of both fragments indicated the presence of three and five putative GLI-binding sites in fragments A and B, respectively (Fig. 5B, left panel). A comparison of these DNA sequences with the previously reported consensus sequence (5Ј-TGGGTGGTC) for the GLI-binding site (32) indicated that these eight DNA sequences have fewer than three mismatches with the reported consensus sequence. To confirm that these sites are really responsible for GL13 binding, the sixth G residue of all of these sites was mutated to an A residue, and the mutated fragments were used for the gel retardation assays. The introduction of this mutation into these putative GLI3binding sites abolished GST-GLI3 binding (Fig. 5B, right  panel).
When either of fragment A or fragment B was inserted upstream of the luciferase gene, the resulting constructs expressed a significant level of luciferase in transfected NIH3T3 and MNS-70 cells, indicating that both of these fragments have promoter activity (data not shown). We also constructed a FIG. 4. Processing of GLI3 but not GLI1. A, analysis of GLI3 in mouse embryo. Extracts were prepared from the 11.5-day post-coitus mouse embryos and separated on 7 and 10% SDS-polyacrylamide gels. Western blotting was performed using the anti-GLI1, which recognizes amino acids 2-17 of GLI1, or anti-GLI3, which recognizes amino acids 1577-1596 (anti-GLI3C) or 1-19 (anti-GLI3N) of GLI3. The short form of GLI3, which may be generated by protein processing, is indicated by an arrow. In the lower right panel, the two N-truncated mutants of GLI3 were in vitro translated and analyzed on a 7% SDS-polyacrylamide gel. By using these two in vitro translated proteins as markers, the approximate position of GLI3 protein processing was estimated and is shown at the top. B, Shh inhibits the processing of GLI3. The plasmid to express Flag-GLI3 (left panel) or GLI1 (right panel) was transfected with (ϩ) or without (Ϫ) the PKA or Shh expression plasmid into 293T cells. Whole cell lysates were prepared and subjected to SDS-PAGE followed by Western blotting using anti-Flag (left panel) or anti-GLI1 antibody (right panel). luciferase reporter construct that has the 3.5-kilobase HindIII-EcoRI fragment containing both fragments A and B and used it in the following experiments. Co-transfection of the GLI3 expression plasmid with this reporter into NIH3T3 cells enhanced the luciferase activity in a dose-dependent manner (Fig.  5C). To investigate the role of GLI3 in Shh-induced activation of the Gli1 promoter, increasing amounts of the Shh expression plasmid were co-transfected into MNS-70 cells with the Gli1 promoter-containing luciferase reporter in the presence or absence of GLI3 and the PKA expression plasmid (Fig. 5D). Neither Shh nor PKA affected Gli1 promoter activity in the absence of exogenous GLI3. In the presence of exogenous GLI3, PKA suppressed Gli1 promoter activity, whereas Shh enhanced Gli1 promoter activity in a dose-dependent manner. This antagonistic regulation by Shh and PKA was not observed with the mutant promoter in which the eight GLI3-binding sites of the Gli1 promoter were disrupted (Fig. 5E).
Inhibition of Shh-induced Gli1 Expression by Overexpression of the GLI3-binding Domain of CBP-Ectopic expression of Shh in MNS-70 cells induces expression of the endogenous Gli1 gene (38), and MNS-70 cells express endogenous GLI3 at a level that may be sufficient for the induction of endogenous target genes by Shh. To confirm that GLI3 functions as a mediator of Shh induction of Gli1 mRNA, we examined the effect of overexpression of the GLI3-binding domain of CBP on the induction of endogenous Gli1 mRNA by Shh (Fig. 6). If the to the 5Ј-untranslated region of the mouse Gli1 gene is indicated at the top. B, BamHI; E, EcoRI; H, HindIII; Hi, HincII; K, KpnI; X, XhoI. The direction of transcription is from left to right. The eight putative GLI3binding sites are indicated by vertical arrows. The gel mobility shift assays were performed using the GST fusion protein containing the metal finger region of GLI3 and the six DNA fragments prepared from the indicated genomic clone as probes. Of six fragments, only two, the 210-base pair BamHI-XhoI fragment (fragment A) and the 800-base pair XhoI-EcoRI fragment (fragment B), bound to GST-GLI3. The results using these two fragments are shown below. In the left panel, the 32 P-labeled fragment A or fragment B was incubated with the GST-GLI3 fusion (200 ng), GST (200 ng), or control buffer and analyzed on a 4% nondenaturing gel. In the right two panels, increasing amounts of GST-GLI3 fusion proteins (0, 40, 80, 160, or 320 ng) were incubated with the fragment A or fragment B probe. B, multiple GLI3-binding sites in the Gli1 promoter region. The DNA sequence of eight putative GLI3-binding sites in fragments A and B are indicated on the left. The previously reported consensus sequence for the GLI-binding site (32) is shown below. On the right, the gel mobility shift assays were performed using 25 ng of GST-GLI3 protein with the wild type and mutant fragments, in which the sixth G residue in the eight putative GLI3-binding sites was changed to A. C, transcriptional activation of the Gli1 promoter by GLI3 in NIH3T3 cells. A mixture of 0, 0.2, 0.5, 1, 2, 3, 4, or 5 g of the GLI3 expression plasmid pSR␣-GLI3, 3 g of the luciferase reporter plasmid pHR-luc, which contains the 3.5-kilobase HindIII-EcoRI fragment of Gli1 promoter, and 0.5 g of the internal control plasmid pRL-SV, in which the SV40 early promoter was linked to the sea pansy luciferase gene, was transfected into NIH3T3 cells, and luciferase assays were performed. The total amount of plasmid DNA was adjusted to 8.5 g by adding the control plasmid DNA pSR␣0. The experiments were repeated three times, and the differences between the experiments were no more than 20%. Typical results are indicated by a bar graph. The shaded and open bars indicate the data with and without the GLI3 expression plasmid, respectively. D, Shh-induced activation of the Gli1 promoter by GLI3 in MNS-70 cells. A mixture of 0, 0.5, 1, or 2 g of the Shh expression plasmid pJT4/Shh, 3 g of the Gli1 wild type promoter-containing luciferase reporter plasmid pHRluc, 3 g of the GLI3 expression plasmid pSR␣-GLI3, 1 g of the PKA catalytic subunit expression plasmid pSR␣-PKA, and 0.5 g of the internal control plasmid pRL-SV was transfected into MNS-70 cells, and luciferase assays were performed. The total amount of plasmid DNA was adjusted to 8.5 g by adding the control plasmid DNA. The experiments were repeated three times, and the average is shown as the relative luciferase activity with S.D. E, requirement of GLI3-binding sites for Shh-induced activation of the Gli1 promoter. Co-transfection experiments were done as described in D using the reporter containing the mutant Gli1 promoter in which the eight GLI3-binding sites were mutated.  (18,19,25,11). Although the pathway of Shh signaling has not yet defined completely, there is evidence to suggest that it is well conserved between Drosophila and vertebrates (Fig. 7). First, GLI3, like Ci, is primarily localized in the cytosol (34). Second, GLI3 has a domain structure similar to that of Ci, with both proteins having transcriptional activation and repression domains. Third, GLI3, like Ci, appears to be cleaved into a repressor form. These facts suggest that Shh signaling blocks association of the GLI3-Cos2-Fu-Su(fu) complex to the cytoskeleton and processing of GLI3 into a repressor form.
Recently, PKA was demonstrated to directly phosphorylate the multiple sites adjacent to the dCBP-binding domain in Ci protein and to enhance the proteolysis of Ci (56). The multiple phosphorylation sites for PKA are well conserved in GLI3, and our results indicate that the small segment of GLI3 containing the CBP-binding domain and the putative PKA phosphorylation sites is sufficient for positive regulation by Shh and negative regulation by PKA (Fig. 2). In addition, PKA enhances the proteolysis of GLI3 to generate a repressor form, and Shh signaling inhibits this (Fig. 4). Phosphorylation of GLI3 by PKA does not affect binding to CBP. One possibility is that direct phosphorylation of GLI3 by PKA retains GLI3 as part of the cytoskeleton complex in the cytosol and enhances the proteolysis of GLI3. Interestingly, in the absence of PKA, Shh does not enhance trans-activation by Gal4-CBD (Fig. 2E, right  panel), suggesting that the role of Shh is to negate the action of PKA. This is consistent with the previous report that suppression of PKA activity is sufficient to activate targets of the Shh signaling pathway in the mouse central nervous system (28). However, the Shh responsiveness of full-length GLI3 in the absence of PKA (Fig. 2E) may suggest that Shh may enhance GLI3 activity through not only the CBP-binding domain but also other region. In addition, Shh and PKA enhanced the Gli1 promoter activity to the higher level than that with Shh alone (Fig. 5D). This may be consistent with the recent report by Ohlmeyer and Kalderon (57) that increased PKA activity can induce ectopic Hh target gene expression without changes in Ci protein concentration. At present, it remains unknown whether CBP is merely act as a co-activator of full-length GLI3 or is important for Shh and PKA responsiveness of GLI3. Because Ci lacks an obvious nuclear localization signal (9, 58), its movement into the nucleus could be mediated by its ability to bind to dCBP, which would carry it there. Although GLI3 has one putative nuclear localization signal in the N-terminal region, vertebrate CBP may also act as a carrier of GLI3 into the nuclei.
Unlike GLI3, GLI1 is a simple transcriptional activator encoded by a target gene of Shh signaling. It appears to contain only the transcriptional activation domain and is not cleaved. These facts suggest the following cascade between Shh and the induction of its target genes (Fig. 7). Shh first enhances GLI3 activity via the region containing the CBP-binding domain, and then the activated GLI3 directly binds to the Gli1 promoter and induces Gli1 transcription. This leads to accumulation of GLI1, which then induces a second wave of transcription involving Shh target genes such as Hnf-3␤ and possibly the Gli1 gene itself, because GLI1 can activates the Gli1 promoter in our co-transfection assays (data not shown).
We observed that the level of Gli1 mRNA in various regions of Gli3 heterozygous and homozygous mouse embryos (Xt/ϩ and Xt/Xt) was significantly lower than that of the wild type (data not shown). However, Gli1 mRNA did not completely disappear, even in the Gli3 homozygous mutant. This may be due to a redundant function of GLI3 and GLI2. Redundant functions of GLI3 and GLI2 was suggested by a study of Gli2 and Gli3 double mutant mice (59). In MNS-70 cells, Gli1 mRNA is expressed at very low level, although a significant level of Gli3 and Gli2 mRNA expression is still observed (data not shown). Because overexpression of the CBP fragment containing the GLI3-binding domain significantly lowered the Shhdependent induction of Gli1 mRNA, CBP may also act as a co-activator of GLI2. Gli3 is not thought to be expressed in the ventral midline of the central nervous system, where the floor plate is induced by Shh. In this region, GLI2 may play the same role as GLI3 to induce Gli1 expression. This is consistent with the recent study of Gli2 mutant mice, in which Gli1 is not detected ventrally at E9.5 and the floor plate does not form (60). Recently, it was reported that Gli2 and Gli3 repress the ectopic induction of frog floor plate cells by Gli1 in co-injection assays and inhibit endogenous floor plate differentiation (61). However, a large amount of repressor form of GLI3 and only a small amount of full-length activator form could be generated in this system, because Shh signaling may not act on large amount of GLI3 protein generated from injected mRNA. If this is a case, a repressor form of GLI3 may inhibit the expression of Shh target genes and antagonize Gli1 function. The antagonizing activities of Gli2 and Gli3 on Gli1 function are also not consistent with the recent loss-of function studies with Gli1 and Gli2 mutants in mouse (60).
Relationship between Gli3 Mutations and Genetic Diseases-Mutations and translocations of the human Gli3 gene are responsible for two human disorders, GCPS and PHS (40,41). The characteristics of each of these disorders overlap with one another but still remain sufficiently distinct to be classified as FIG. 6. Inhibition of Shh-induced Gli1 expression by overexpression of the GLI3-binding domain of CBP. MNS-70 cells were transfected with a mixture of 4 g of the Shh expression plasmid pJT4/Shh and 4 g of the plasmid to express GLI3 or the GLI3-binding domain of CBP or the control plasmid lacking the cDNA to be expressed. Total RNA was prepared from the transfected cells, and Gli1 expression was analyzed by reverse transcription-PCR. Cytoplasmic ␤-actin was used as a control. On the right, the degree of Gli1 expression is indicated by a bar graph. separate disorders. For instance, although both GCPS and PHS have polysyndactyly and abnormal craniofacial features, GCPS has commonly postaxial polydactyly of the hands and preaxial polydactyly of the feet, whereas PHS has typically central or postaxial polydactyly. GCPS does not cause the hypothalamic hamartoma observed in PHS. In addition, PHS is not associated with hypertelorism or broadening of the nasal root or forehead seen in GCPS. The different anomalies in these two disorders may be partly due to the generation of truncated proteins having different characteristics. In GCPS, GLI3 was found to be truncated upstream or within the zinc finger domain (40,62), whereas mutations found in PHS truncate GLI3 after the zinc finger (41). Our domain analysis indicated that the N-terminal region upstream from the zinc finger region is a repressor domain. Therefore, the truncated protein found in PHS could still retain repressor activity, unlike the protein truncated upstream or within the zinc finger domain, which was found in GCPS.
Direct Binding of CBP to GLI3: a Molecular Link between RTS and Genetic Diseases caused by Gli3 Mutations-Our results indicate that GLI3 utilizes CBP as a co-activator, as in Drosophila, in which Ci uses dCBP. Thus, the interaction of GLI3/Ci with CBP/dCBP is conserved between mammals and insects. Some of the characteristics of GCPS and PHS are similar to RTS. GCPS and RTS are caused by haploinsufficiency, and GCPS, PHS, and RTS are all associated with craniofacial, hand, and foot defects. Thus, direct interaction between GLI3 and CBP could explain the similar characteristics of these disorders. However, some characteristics of these syndromes are distinct. For instance, broad thumbs and broad halluces are associated with both GCPS and RTS, but polydactyly and syndactyly are commonly observed only in GCPS. These differences between GCPS and RTS could be explained by multiple mechanisms. In the embryos of multiple polydactylous mouse mutants, such as Xt and Hemimelic extra toes (Hx), Shh is ectopically expressed at the anterior margin of the limb buds (63). By analogy with the Ci-dependent repression of hh expression in the anterior compartment of Drosophila, the repressor form of GLI3 lacking the C-terminal activation domains may repress Shh expression. Therefore, the reduction or loss of repressor activity of GLI3 may lead to polydactyly. Although the truncated forms found in PHS still retain the N-terminal repressor domain and the zinc finger domain, these truncations may lead to increased instability of these proteins or to low repressor activity. In contrast, the deficiency of CBP should not affect the repressor activity of GLI3, and this may explain the lack of polydactyly in RTS. In addition, CBP affects the activity of many transcription factors, and some of the features seen in RTS could be explained by decreased activity of transcription factors other than GLI3. dCBP is a co-activator of Dorsal, a Drosophila homolog of NF-B, and dCBP mutations cause the loss of Dorsal-dependent expression of the twist gene (64). Because mutations of the human Twist gene are associated with the autosomal dominant Saethre-Chotzen syndrome, which is clinically analogous to RTS (65,66), some defects may be caused by decreased expression of the Twist gene. FIG. 7. Schematic representation of the role of GLI3 and GLI1 in Shh signaling. Shh signaling enhances the trans-activating capacity of GLI3 through the region containing the CBP-binding domain and the PKA phosphorylation sites and induces Gli1 transcription through the direct action of GLI3 on the Gli1 promoter. The expressed GLI1 then activates specific target genes including Hnf-3␤ and the Gli1 gene itself in an Shhindependent manner.