Molecular Properties of Zic Proteins as Transcriptional Regulators and Their Relationship to GLI Proteins*

Zic family genes encode zinc finger proteins, which play important roles in vertebrate development. The zinc finger domains are highly conserved between Zic proteins and show a notable homology to those of Gli family proteins. In this study, we investigated the func-tional properties of Zic proteins and their relationship to the GLI proteins. We first established an optimal binding sequence for Zic1, Zic2, and Zic3 proteins by electrophoretic mobility shift assay-based target selection and mutational analysis. The selected sequence was almost identical to the GLI binding sequence. However, the binding affinity was lower than that of GLI. Consistent results were obtained in reporter assays, in which transcriptional activation by Zic proteins was less dependent on the GLI binding sequence than GLI1. More-over, Zic proteins activated a wide range of promoters irrespective of the presence of a GLI binding sequence. When Zic and GLI proteins were cotransfected into cul-turedcells,Zicproteinsenhancedorsuppressedsequence-dependent,GLI-mediatedtransactivationdependingoncelltype.Takentogether,theseresultssuggestthatZicproteinsmayactastranscriptionalcoactivatorsandthattheirfunctionmaybemodulatedbytheGLIpro-teinsandpossiblybyothercelltype-specificcofactors.

A Zic1-deficient mouse shows cerebellar abnormalities and marked ataxia (9). Mutations in human ZIC2 or mouse Zic2 lead to a congenital brain anomaly known as holoprosencephaly (10 -12). The human ZIC3 mutations result in a disturbance of the left-right body axis (13). In Xenopus, Zic genes play a role in the initial phase of neural and neural crest development (5,7). The zinc finger domain was highly conserved not only in the vertebrate Zic family but also in Drosophila odd-paired, which regulates segmentation and midgut development (14,15). All of these findings indicate that Zic genes exert an essential role in multiple aspects of development.
In addition to Zic orthologues in other species, the zinc finger domains of the Zic proteins, which consist of five tandemly repeated C2H2 motifs, show a notable homology to those of Gli family proteins (1). At present, Gli1, Gli2, and Gli3 have been reported in human, mouse, and frog (16 -21) in addition to their nonvertebrate homologues Drosophila Cubitus interruptus (Ci; Ref. 22) and Caenorhabditis elegans Tra1 (23). Gli proteins are also known to regulate events in vertebrate development. In particular, it has been reported that Gli and Cubitus interruptus proteins function downstream of the Sonic hedgehog-Patched signaling pathway as transcriptional regulators (24 -26). For example, in response to the Sonic hedgehog signal, activated Gli proteins bind directly to the enhancer element of a winged helix transcription factor, hepatocyte nuclear factor-3␤, the expression of which is critical for floor plate development (27,28).
The sequence-specific DNA binding of the GLI proteins has been studied in detail. A consensus nonamer target site, 5Ј-TGGGTGGTC-3Ј (GLI-BS), 1 has been identified for human GLI1 (29). A crystallographic study showed that GLI1 bound DNA through the last four fingers via extensive base contacts, which are highly conserved between GLI and Zic families (2,30). Tra1 protein can also bind the GLI-BS possibly through the last three zinc fingers, because elimination of the first two fingers had little effect on its binding to DNA (23). Furthermore, the GLI3 zinc finger domain (GLI3-ZF) can also bind the GLI-BS (31). The target has been optimized to an extended 16-nucleotide sequence including the GLI-BS (GLI3-BS).
The fact that Zic proteins are nuclear proteins and have zinc finger domains highly similar to those of Gli proteins led us to infer that the Zic proteins might act as transcription factors in a manner similar to that of Gli proteins. However, little is known about the regulation of gene expression by the Zic proteins. The only finding in support of this to date is that the zinc finger domain of Zic1 (Zic1-ZF) can bind the GLI-BS (1). Therefore, the Zic proteins may regulate the expression of target genes by interacting with a GLI-BS-like sequence. Nevertheless, no methodical analysis of the DNA recognition sequence for Zic family proteins has been performed to date. We therefore undertook the task of empirically determining the Zic binding sequence and characterizing its binding properties. Furthermore, we examined transcriptional regulation by Zic proteins and tested whether the Zic proteins affected the GLI-* This work was supported by the Special Coordination Fund for Promoting Science and Technology and grants from the Japanese Ministry of Education, Science and Culture, the Takeda Science Foundation, the Naito Foundation, and the Senri Life Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  DNA interaction and GLI-mediated transcriptional activation. This study should pave the way for a better understanding of the molecular mechanism of Zic-mediated transcriptional regulation.
The resultant cDNA fragments were cut with BamHI and SalI, and inserted into pGEX4T2 (Amersham Pharmacia Biotech) or pQE-30 (Qiagen). The constructs were checked by DNA sequencing. Bacterially expressed proteins were prepared and purified according to the manufacturer's instructions. Glutathione S-transferase (GST) fusion proteins were purified by glutathione-Sepharose 4B (Amersham Pharmacia Biotech) without denaturation. The histidine-tagged proteins were affinitypurified by nickel-chelate affinity (Qiagen), denatured by urea, and subsequently renatured in the purification process. The purified proteins were finally dialyzed against a solution consisting of 25 mM HEPES, pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 10 M ZnSO4, 1 mM dithiothreitol, 0.1% Nonidet P-40, and 12% glycerol and stored at Ϫ80°C in small aliquots.
Binding Site Selection and Mutational Analysis-Initially, DNA fragments consisting of random oligonucleotides flanked on each side by a known sequence were generated by PCR amplification of R62AB (5Ј-AGACGGATCCATTCAG(N)30TGTCAGGAATTCGGAC-3Ј) using primers A1 (5Ј-AGACGGATCCATTCAG-3Ј) and B1 (5Ј-GTCCGAATTC-CTGACA-3Ј) by 25 cycles of PCR, in which each cycle consisted of denaturation at 94°C (30 s), annealing at 52°C (30 s), and extension at 72°C (1 min). The PCR products were purified on a CHROMA SPIN TE 30 column (CLONTECH). A fraction of the PCR product was 32 P-labeled with T4 polynucleotide kinase. Unincorporated nucleotides were removed by column chromatography over a TE Micro Select-D G-25 microcentrifuge spin column (5 Prime 3 3 Prime Inc., Boulder, CO).
After the first EMSA reaction, the bands corresponding to the protein-DNA complexes, which were identified by autoradiography, were excised from the dried gel and eluted into water for 1 h at room temperature. The eluted DNA was amplified by PCR with the A1 and B1 primers and purified by the same procedure as described above. Four additional cycles of EMSA, elution, and PCR amplification were performed. After the fifth cycle, the PCR products were digested with BamHI and EcoRI and subcloned into pBluescript (Stratagene). Insertcontaining clones were sequenced, and sequences were analyzed by DNASIS (Hitachi Software).
For mutational analysis, double-stranded oligonucleotides were prepared by annealing of complementary oligonucleotides, and an EMSA was performed using 80 ng of GST-Zics-ZF or 8 ng of GST-GLI3-ZF protein.
Calculation of the Equilibrium Binding Constant-The GLI3-BS probe, which contains the entire optimized target sequence of GLI3 (31), was used to determine the K d . The nucleotide sequence of the probe was 5Ј -AGACGGATCCATTGCATCTGTGATTTTCGTCTTGGGTGGTCTC-CCTCCTGTAGGAATTCGGAC-3Ј, where the underline indicates the optimized target sequence flanked by PCR primer sequences. The procedures for the PCR amplification and radiolabeling of the probe were the same as those described above. The following equilibrium equation was used to calculate K d 2 from the midpoint of the curves: B ϭ B max ‫ء‬ F/(K d ϩ F), where B and F are the concentration of the protein-DNA complex and unbound probe, respectively. B max is the total concentration of the protein.
Plasmid Construction-A luciferase reporter plasmid driven by a thymidine kinase (TK) promoter, TK-0GBS-Luc, was constructed by ligating a herpes simplex virus TK promoter derived from plasmid pRL-TK (Promega) into the HindIII and BglII sites of the pGL2-Basic vector (Promega). Six tandem copies of the GLI-BS (32) were inserted upstream of the TK promoter in TK-0GBS-Luc, giving rise to TK-6GBS-Luc. A luciferase reporter plasmid driven by a major late promoter of adenovirus, MLP-Luc, was constructed by removing the GAL4 binding sites from pG5Luc (Promega), and a luciferase reporter plasmid driven by an SV40 promoter, SV40-Luc (pGL2 promoter vector), was purchased from Promega. A luciferase reporter plasmid driven by a cytomegalovirus (CMV) promoter, CMV-Luc, was constructed by inserting a luciferase gene downstream of the CMV promoter region of the pCS2ϩ vector (33,34), and a luciferase reporter plasmid driven by a Zic1 promoter, Zic1-Prom-Luc, was constructed as described previously (SmaI (Ϫ148); Ref. 35). The internal standard, pRL-EF, was constructed by inserting the renilla luciferase gene of a pRL-TK vector (Promega) into a pEF-BOS vector (36).
Expression vectors were constructed by inserting the cDNA fragments containing the entire open reading frame of mouse Zic1, Zic2, and Zic3, human GLI1, or GLI3 into the XbaI sites of pEF-BOS (36). The expression vector Flag-GLI3 was constructed by placing the Flag sequence in frame to the N terminus of human GLI3. In addition to the full-length constructs, Zic1-FS and Zic3-FS were made, which included the entire cDNA fragments of mouse Zic1 and Zic3 with frameshift mutations to produce truncated proteins lacking all five zinc fingers and C-terminal regions (Zic1-FS, amino acids 1-159; Zic3-FS, amino acids 1-153). Transfection-293T and C3H10T1/2 cells were maintained in Dulbecco's modified Eagle's medium and basal medium Eagle's medium, respectively, each containing 10% fetal bovine serum. At 60 -70% confluence, the cells in a 24-well dish were transfected with Effectene (Qiagen) or LipofectAMINE Plus (Life Technologies, Inc.) transfection reagent according to the manufacturer's instructions. Cells were harvested 24 h after transfection and processed for reporter assay and Western blot analysis.
Reporter Assay-For 293T cells, cotransfection experiments were performed with 90 ng each of the luciferase reporters and 100 ng each of the expression constructs, together with 5 ng of pRL-EF as an internal standard. C3H10T1/2 cells were cotransfected with 180 ng of the luciferase reporters, 200 ng of the expression constructs, and 10 ng of pRL-EF. When the expression vectors for Zic and GLI were cotransfected, the total amounts of expression vectors were kept constant by adding an empty vector (pEF-BOS). Luciferase activities were measured according to the manufacturer's recommendation (Promega) using a Minilumat LB 9506 luminometer (Berthold). The firefly luciferase activity was normalized to the renilla luciferase activity obtained by cotransfection. The relative fold-activation was presented as the ratio of the normalized value to that from empty vector transfectant.

Binding Site Selection of Zic Family
Proteins-To determine the DNA sequence to which Zic proteins bind, we synthesized oligonucleotides containing 30 contiguous random nucleotides flanked on either side by a known sequence to facilitate PCR amplification and subsequent cloning of the selected DNA sequence. The radiolabeled random oligonucleotides were incubated with purified GST fusion protein containing the zinc finger domain of three mouse Zic proteins (Zic1-ZF, Zic2-ZF, and Zic3-ZF), DNA-protein complexes were separated on native polyacrylamide gels, and the bands shifted by the proteins were excised. After elution, the DNA was amplified by PCR and subjected to four additional rounds of DNA-protein binding, elution, and PCR amplification to enrich the Zic binding sequences. After a total of five rounds of selection, the bound oligonucleotides were cloned and sequenced (Fig. 1). The core sequence most favored by Zic1-ZF, Zic2-ZF, and Zic3-ZF was nearly identical to the 9-nucleotides GLI consensus sequence (GLI-BS), i.e. 5Ј-TGGGTGGTC-3Ј. The relative frequency of the appearance of each nucleotide in the 9 positions indicated that all 5 guanosine residues, 2 thymidine residues at the fifth and eighth positions, and a cytidine residue at the 3Ј-most position were preferred in common, whereas a thymidine residue at the 5Ј-most position was not.
However, there seemed to be no absolute requirement for any specific nucleotide position. So we asked whether a different target sequence might exist in strongly bound sequences. To test this, we selected target sequences that gave stronger shifted bands than GLI3-BS. Among 13 sequences with high binding affinity to Zic1-ZF, there was no conserved sequence other than GLI-BS (data not shown). This result indicated that the GLI-BS-like sequence was uniquely preferred by the Zics-ZF.
Identification of the Minimal Essential Binding Sequence for the Zic Family Proteins-Although all of the selected sequences could be bound by the Zics-ZF, it was unclear whether the whole 9-nucleotide sequence was necessary for binding. To determine the minimal essential nucleotide sequence for binding, a series of mutant sequences, each of which contained a single-base substitution, were synthesized and analyzed by EMSA using GST-Zics-ZF or GST-GLI3-ZF (Fig. 2). The M1 mutant was efficiently bound by the Zics-ZF, suggesting that the mutation within the oligonucleotide was not essential for binding. In contrast, mutants M2, M4, and M5 failed to bind. M3 and M6 were bound by Zics-ZF less efficiently than the wild-type sequence. This analysis showed that the nucleotides altered in M2, M4, and M5 and in part in M3 and M6, i.e.
5Ј-GGGTGGTC-3Ј, contained the minimal essential binding sequence for the Zic proteins. This result was consistent with the fact that a thymidine residue at the 5Ј-most position was not strongly preferred by the binding site selection. The GLI3-ZF exhibited generally similar binding preference for the mutant binding sequences to the Zic proteins except for reduced dependence on the 3Ј-most cytidine residue.
Differential Binding Affinities between Zic Family Proteins and GLI3 to the Selected Binding Sequence-Although the Zics-ZF and GLI3-ZF exhibited a similar binding preference for the target nucleotide sequence, there was an apparent difference in binding affinity. To determine the relative binding affinity of the Zics-ZF and GLI3-ZF, both proteins were incubated simultaneously with the GLI3-BS probe, and the resultant complexes were analyzed by EMSA. To discriminate the two kinds of fusion proteins, different tags were used. The mixture consisting of GST-GLI3-ZF (10 nM) and His-Zic2-ZF (0 -2000 nM; Fig. 3A) or His-GLI3-ZF (20 nM) and GST-Zic2-ZF (0 -500 nM; Fig. 3B) was reacted with the GLI3-BS probe and analyzed by EMSA. Even though an excess concentration of Zic2 protein was used, the intensity of the band shifted by the GLI3 protein (the complex formed between GLI3-ZF and GLI3-BS) was not markedly changed in both experiments. In contrast, the bands shifted by the Zic proteins were almost completely removed by a lower concentration of GLI3 protein (Fig.  3C). These findings confirmed that, although the Zic family proteins specifically bound the GLI3-BS, the binding affinity was much lower than that of GLI3-ZF, and Zics-ZF do not compete with GLI3-ZF for the target sequence.
To more accurately determine the binding affinities of the Zics-ZF and GLI3-ZF, we performed EMSA with a fixed concentration of the GLI3-BS probe (1 nM) and increasing concentrations (1 nM-1 M) of the fusion proteins (Fig. 4). Further-  more, we performed EMSA with a fixed concentration (20 nM) of the fusion proteins and increasing concentrations (0.75 -200 nM) of the probe (Fig. 5). Then the intensities of the bands corresponding to the protein-DNA complexes were measured by densitometry, and the binding curves were plotted (Fig. 6). In these assays, the GLI3-ZF showed considerably higher binding affinity to the GLI3-BS (K d , 8. To test whether some of the DNA-protein complexes might dissociate during electrophoresis in the gel, we also electrophoresed the DNA-protein complexes for various times. However, there was no significant alteration in the dissociation rate during gel electrophoresis (data not shown), indicating that the different K d values reflect the differential affinities in the binding mixture.
Different Transcriptional Activities between Zic and GLI Proteins Mediated through the Selected Sequence-We next asked whether the Zic proteins were capable of activating or repressing transcription mediated by the selected binding sequence (Fig. 7). For this purpose, we prepared Zic, GLI1, and GLI3 expression vectors under the control of an EF1␣ promoter. We used the firefly luciferase genes as reporters, controlled by the major late promoter of the adenovirus (MLP-Luc), which consists of minimal promoter elements (a TATA box and a tran-scription initiation site), the Zic1 promoter (Zic1-prom-Luc), the SV40 promoter (SV40-Luc), and the CMV promoter (CMV-Luc). Luciferase activity was measured after cotransfection into 293T cells. The firefly luciferase activity was normalized to the activity of the cotransfected EF1␣-driven renilla luciferase, because the Zic and GLI expression vectors failed to transactivate the EF1␣ promoter (data not shown). All Zic proteins activated the reporter genes driven by the MLP, the Zic1 promoter, and the SV40 promoter but failed to activate that driven by the CMV promoter (Fig. 7A). On the other hand, neither GLI1 nor GLI3 proteins activated any of the promoters.
To examine whether the GLI-BS affected the transcriptional activation by Zic proteins, we used the firefly luciferase genes, controlled by the TK promoter with or without six copies of the GLI-BS (TK-6GBS-Luc and TK-0GBS-Luc). Zics always activated transcription driven by the TK promoter irrespective of the presence of the GLI-BS in both 293T and C3H10T1/2 cells (Fig. 7, B and C). The presence of GLI-BS enhanced the reporter gene expression to 1.5-2.0-fold in C3H10T1/2 cells, whereas in 293T cells, Zic1 or Zic2 enhanced expression to a lesser extent. In contrast, GLI1 markedly enhanced reporter expression, dependent on the presence of GLI-BS in both cell lines (14-fold (293T) or 8-fold (C3H10T1/2) elevation in comparison with TK-0GBS-Luc). GLI3 completely failed to activate the reporter gene expression mediated by the GLI-BS in either cell type. The fact that Zic proteins activated reporter gene expression independent of the GLI-BS prompted us to ask whether the Zic gene or mRNA may indirectly affect the transcriptional activation. To address this possibility, we transfected Zic1-FS and Zic3-FS, which contained the entire cDNA sequence of Zic1 and Zic3, respectively, but produced truncated proteins without all five zinc fingers. However, these constructs were not capable of activating any of the reporter genes tested, demonstrating that Zic proteins themselves could activate transcription (Fig. 7).
These findings show that the Zics have transcriptional activation capacity, but this activity is less dependent on the GLI-BS than is GLI1 activity. The different binding affinities to the target sequence can account for the differential dependence on GLI-BS between the Zic proteins and GLI1.
The Zic Family Proteins Cooperate or Interfere with the GLI in Regulating Transcription-We investigated whether Zic proteins could affect the GLI-BS-dependent transactivation by GLI1. Different relative amounts of the Zic and GLI1 expression vectors were transfected into C3H10T1/2 or 293T cells to titrate this interaction (Fig. 8). When Zics were coexpressed with GLI1 in C3H10T1/2 cells, the GLI-BS-TK-driven luciferase expression was significantly increased beyond that predicted by the summed luciferase activities induced separately by the Zics and GLI1 (Fig. 8A). Increasing amounts of cotransfected Zics enhanced reporter gene expression in a Zic dose-dependent manner. For example, cotransfection with 100 ng of Zic2 increased the GLI1-mediated transactivation by 3-fold. In contrast, when Zics and GLI1 were cotransfected into 293T cells, there was a reduction in reporter gene expression to less than that of Zic or GLI1 single transfectants, possibly because of the interaction between Zics and GLI1 (Fig. 8B).
We also carried out Zic and GLI3 cotransfection, although GLI3 failed to activate or repress the TK-6GBS-Luc reporter gene (Fig. 9). Unexpectedly, even in this case, we obtained results essentially similar to those of Zic-GLI1 cotransfection experiments. When cotransfection was performed in C3H10T1/2 cells, the luciferase activities were increased. However, the increased activities were nearly equal to the addition of the luciferase activities induced separately by the Zics and GLI3 (Fig. 9A). On the other hand, in 293T cells, luciferase activities were lower in Zic-GLI3 cotransfectants than in Zic single transfectants (Fig. 9B). At a certain ratio of expression vectors (Zics:GLI3 ϭ 10:50 ng), the activities were even lower than that from transfection of the empty vector. As a consequence, it became clear that the synergistic activation or repression of reporter gene expression is dependent on cell type for both GLI1 and GLI3. Comparable protein expression levels were confirmed by Western blot analysis for all of the transfections (Fig. 10), indicating that the observed transcriptional synergism and interference were not attributable to different efficiencies of expression of the transfected genes.

Zic Proteins as General Transcription
Activators-In the present study, we were able to establish a consensus binding sequence for Zics by EMSA-based target selection and mutational analysis. The Zic binding sequence was essentially identical to the GLI-BS, 5Ј-TGGGTGGTC -3Ј, and had a minimum consensus sequence of 5Ј-GGGTGGTC-3Ј. The binding affinities for this sequence were very similar among the three Zic proteins examined. However, the Zics-ZF bound the GLI-BS much more weakly than GLI3-ZF, as shown by competition experiments and the calculated binding constant. The K d values of Zics were much higher than those of other transcription factors that function in a sequence-specific manner (37,38). Therefore, it is unlikely that Zic proteins compete with GLI for the GLI-BS.
The binding properties were consistent with the results of the reporter assay, in which the dependence of Zic proteins on the GLI-BS for transcriptional activation was much less than that of GLI1. Instead, Zics activated transcription even in the absence of the GLI-BS via various promoters (TK promoter, adenovirus major late promoter, SV40 early promoter, and Zic1 promoter). On the basis of these facts, rather than being the transcription factors that regulate transcription by direct binding to DNA, Zic family proteins may function as transcriptional coactivators, which potentiate the activity of other transcription regulatory factors. It is possible that Zics interact with the transcription machinery or other factors that regulate tran- scriptional efficiency. An alternative possibility is that Zic proteins might function in the post-transcriptional gene expression processes. Because Zic proteins are localized in the cell nuclei (1, 3), they could be involved in RNA processing or transport from the nucleus to the cytoplasm.
Although the Zics activated a variety of promoters, they had little activity on the EF1␣ and CMV promoters. This may be related to promoter strength, because, of the six promoters we tested, these two promoters had the strongest activities (ϳ10 -1000-fold stronger than other promoters tested). The transcriptional machinery, which is required for activation by Zic proteins, might have already used by the EF1␣ or CMV promoters alone. Further study is needed to elucidate the domain required for the transcriptional activation in the Zic proteins and to identify factors that interact with that domain to understand the regulatory mechanism of gene expression by Zic proteins.
Context-dependent Regulation of GLI Function by Zic Proteins-We also examined the relationship between Zic and GLI proteins. In C3H10T1/2 cells, Zic-GLI1 or Zic-GLI3 coactivated reporter gene expression, whereas in 293T cells, coexistence of the Zic and GLI proteins had a reverse effect. These results suggest a significant regulatory relationship between Zic and GLI proteins; however, the nature of this interaction remains unclear.
The interaction between Zic and GLI proteins may be en- tirely independent of DNA binding. This direct or indirect interaction between Zic and GLI proteins may be modulated by cell type-specific cofactors. One well characterized cell typespecific cofactor is Oct-binding factor 1 (OBF-1), which is expressed in B-lymphocyte lineages and interacts with the POUhomeodomain proteins Oct-1 and Oct-2 to enhance transcriptional activation in the B-cell lineage (39 -41). Similar cell type-specific cofactors might modify the GLI-Zic interactions. It is also possible that GLI proteins may be differentially modified post-translationally depending on the presence of Zics in different cell types. Recently, it was shown that Gli3 was processed depending on cAMP-activated protein kinase to generate a phosphorylated repressor form (32,42). Zic proteins might be involved in this pathway. Alternatively, the differential binding affinities of the Zic and GLI proteins for the target sequence may underlie the regulatory relationship between these two protein families. Although Zics-ZF had much lower binding affinities to the GLI-BS, there is a DNA binding transcription factor that has a K d value similar to those of Zics (43). Moreover, the different human homeodomain proteins, despite having similar homeodomains, bind their target sequence with different affinities and thereby generate a complex regulatory network in the developmental process (44). In that case, less conserved domains other than the zinc finger may modulate binding in vivo to determine final binding specificity, because the recombinant proteins used in these experiments only included zinc finger domains. It is necessary to examine the downstream target genes in the developmental context to understand the Zic-DNA interaction in detail.
Recently, it was demonstrated that Zic1/Gli3 double mutant mice showed severe abnormalities of vertebral arches not found in single mutants (45), strongly suggesting that these two proteins act synergistically in the development of the vertebral arches. On the other hand, it was shown in Xenopus laevis that Zic2 antagonized the Gli proteins in the patterning of the neural plate (46). These findings suggest that Zic and GLI proteins may interact to variously repress or activate gene expression in vivo. Our results may be related to these devel-opmental phenomena.
In conclusion, Zic1, Zic2, and Zic3-ZF specifically recognized and bound the GLI-BS but with a much lower binding affinity than that of the GLI3-ZF. Zic proteins activated a wide range of promoters. These results suggest that Zic proteins may function as transcriptional coactivators or as factors generally involved in the gene expression process. How can such general factors regulate specific developmental processes, including the patterning of forebrain, cerebellum, axial skeleton, vasculature, and visceral organs? A clue to solving this problem may be the relationship with Gli family proteins as shown in this study. To clarify the regulatory networks under a broad range of developmental process, the relationships between Zic proteins and other molecules in the hedgehog signaling pathway and transforming growth factor ␤ superfamily, which are closely related to each other, should also be examined in both in vitro and in vivo studies.