Identification of Glis1, a novel Gli-related, Kruppel-like zinc finger protein containing transactivation and repressor functions.

In this study, we describe the identification and characterization of a novel Krüppel-like protein named Gli-similar 1 (Glis1). The Glis1 gene encodes an 84.3-kDa proline-rich protein. Its five tandem zinc finger motifs exhibit highest homology with those of members of the Gli and Zic subfamilies of Krüppel-like proteins. Glis1 was mapped to mouse chromosome 4C6. Northern blot analysis showed that expression of the 3.3-kb Glis1 mRNA is most abundant in placenta and adult kidney and expressed at lower levels in testis. Whole mount in situ hybridization on mouse embryos demonstrated that Glis1 is expressed in a temporal and spatial manner during development; expression was most prominent in several defined structures of mesodermal lineage, including craniofacial regions, branchial arches, somites, vibrissal and hair follicles, limb buds, and myotomes. Confocal microscopic analysis showed that Glis1 is localized to the nucleus. The zinc finger region plays an important role in the nuclear localization of Glis1. Electrophoretic mobility shift assays demonstrated that Glis1 is able to bind oligonucleotides containing the Gli-binding site consensus sequence GACCACCCAC. Although monohybrid analysis showed that in several cell types Glis1 was unable to induce transcription of a reporter, deletion mutant analysis revealed the presence of a strong activation function at the carboxyl terminus of Glis1. The activation through this activation function was totally suppressed by a repressor domain at its amino terminus. Constitutively active Ca(2+)-dependent calmodulin kinase IV enhanced Glis1-mediated transcriptional activation about 4-fold and may be mediated by phosphorylation/activation of a co-activator. Our results suggest that Glis1 may play a critical role in the control of gene expression during specific stages of embryonic development.

subfamilies based on the number of zinc finger motifs, sequence homology between the zinc-fingers, and the presence of specific repressor and activation domains (3)(4)(5)(6).
In this study, we describe the cloning of a cDNA encoding a novel member of the Krü ppel-like zinc finger family not previously described. We named this protein Gli-similar 1 (Glis1) 1 based on its relationship to Gli proteins. Glis1 contains five tandem zinc finger motifs that show high homology with those of Gli and Zic proteins. The zinc finger region of Glis1 exhibits the highest identity (77%) with the Drosophila Gli-like protein Lame duck (Lmd, also named gleeful or glf) (40,41). In adult mouse tissues, Glis1 mRNA is particularly abundant in placenta, kidney, and testis. Analysis of its expression during embryonic development demonstrated that Glis1 is expressed in both a temporal and spatial manner in the frontal nasal region, branchial arches, somites, vibrissal and hair follicles, * 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 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF486579.
¶ To whom correspondence should be addressed. Tel.: 919-541-2768; Fax: 919-541-4133; E-mail: jetten@niehs.nih.gov. and limb buds. Its nuclear localization and binding to oligonucleotides containing the Gli-binding site suggested that Glis1 might modulate gene transcription. This conclusion was supported by studies examining the transcriptional activity of Glis1 by monohybrid and deletion analysis. This analysis revealed that Glis1 contains both transactivation and repressor domains indicating that it may function as an activator and repressor of transcription. These observations suggest that Glis1 functions as a transcription factor that regulates specific stages of embryonic development.

EXPERIMENTAL PROCEDURES
Cell Culture-Green monkey kidney fibroblast CV-1 and COS-7, Chinese hamster ovary, and human kidney 293 cells were obtained from ATCC. Human kidney epithelial PK-1 cells were obtained from Dr. Bonventre (Harvard Medical School). Cells were routinely maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum except for Chinese hamster ovary, which was cultured in Ham's F-12.
Yeast Two-hybrid Analysis-Yeast two-hybrid analysis was performed using the ligand-binding domain of the nuclear orphan receptor ROR␥ (42) as bait and a pACT mouse lymphoma MATCHMAKER cDNA library following the manufacturer's protocol (CLONTECH). The cDNA fragment encoding the ligand-binding domain of the nuclear receptor ROR␥ was cloned into pGBT9 (CLONTECH). One of the positive cDNA clones contained a 1.4-kb insert encoding the carboxylterminal region of Glis1.
cDNA Library Screening and 5Ј SMART-RACE-To obtain the remaining sequence of mGlis1, a mouse kidney cDNA library made in pcDNA3.1 (Invitrogen) was screened by PCR using two Glis1-specific primers. This screening was performed by IncyteGenomics (St. Louis, MO) and yielded one clone containing the full-length coding region of Glis1. To obtain the 5Ј-untranslated region, 5Ј SMART-RACE was performed following the manufacturer's protocol (CLONTECH). cDNA was prepared using RNA from mouse placenta as template, Moloney murine leukemia virus-reverse transcriptase, and SMARTII and 5Ј-RACE cDNA synthesis primers (CLONTECH). The obtained cDNA was PCR amplified with the universal primer (5Ј-CTAATACGACTCACTA-TAGGGCAAGCAGTGGTAACAACGCAGAGT) and Glis1-specific primer GSP5 (5Ј-GCCGGTCTTGGACACACCATAGTCTCGAGG). A second PCR reaction was performed using nested primers, a universal (5Ј-AAGCAGTGGTAACAACGCAGAGT) and Glis1-specific primer GSP4 (5Ј-TGGCACCGGTAGCCGAGGCAGCAGGTCTAG). PCR products were then subcloned into pGemT vector (Promega) and sequenced.
DNA Sequencing-Plasmids were purified using Wizard miniprep or midiprep kits from Promega. Automatic sequencing was carried out using a Dynamic ET Terminator Cycle Sequencing Ready reaction kit (PerkinElmer Life Sciences) and an ABI Prism 377 automatic sequencer. DNA and deduced protein sequences were analyzed by the seqWEB and MacVector sequence analysis software packages.
Northern Blot and RT-PCR Analysis-A multitissue blot containing total RNA (25 g) from 14 different mouse tissues was purchased from Seegene (Seoul, Korea). The expression of the mGlis1 gene was also examined by RT-PCR using total RNA (1 g) from different mouse tissues as template and two Glis1-specific primers. The RT-PCR was carried out at the following conditions: 45 min at 48°C and 2 min at 94°C (1 cycle), followed by 0.5 min at 94°C, 1 min at 60°C, and 2 min at 68°C (25 cycles). Finally, after a 7-min incubation at 68°C, samples were analyzed on a TBE gel (0.8%) and transferred to a HyBond-N ϩ membrane. The blots were hybridized to a 32 P-labeled probe for mGlis1. Hybridizations were performed at 68°C for 1 h, the membranes were then washed twice with 2ϫ SSC, 0.1% SDS at room temperature for 30 min and 0.1% SSC, 0.1% SDS at 58°C for 30 min. Autoradiography was carried out with Hyperfilm-MP (Amersham Biosciences) at Ϫ70°C.
Fluorescence in Situ Hybridization-The regional chromosomal localization was determined by fluorescence in situ hybridization using a Glis1 genomic fragment as a probe. Genomic clones containing the Glis1 gene were obtained by screening a library of BAC1 vectors containing 100 -150-kb fragments of the mouse genome using a radiolabeled Glis1 cDNA as a probe. The identity of the clones was verified by restriction mapping using different cDNA fragments of Glis1 as probes. Genomic DNA derived from the Glis1 BAC clone was labeled with digoxigenin dUTP by nick translation. Labeled probe was combined with sheared mouse DNA and hybridized to normal metaphase chromosomes derived from mouse embryo fibroblasts in a solution of 50% formaldehyde, 10% dextran sulfate, and 2ϫ SSC. Specific hybridization signals were detected by incubating the hybridized slides in fluoresceinated anti-digoxigenin antibodies followed by counterstaining with 4,6diamidino-2-phenylindole. The initial experiment resulted in specific labeling of chromosome 4 on the basis of 4,6-diamidino-2-phenylindole staining. The latter was confirmed when a probe specific for the centromeric region of chromosome 4 was co-hybridized with the Glis1 probe.
Whole Mount in Situ Hybridization-Whole mount in situ hybridization studies were performed according to Haramis and Carrasco (43). Mouse embryos from 7.75 to 14.5 days postcoitus (E7.75-E14.5) were fixed overnight in 4% paraformaldehyde in phosphate-buffered saline, washed twice for 10 min in phosphate-buffered saline ϩ 0.1% Tween 20 (PBST), and then dehydrated with a series of 10 min methanol/PBT washes (25, 50, and 75, and twice at 100%). Embryos were stored at Ϫ20°C until probed. The Glis1 probe, encoding the region from nucleotide 502 to 1069, was generated by PCR using primers containing either EcoRI or an HindIII site. The PCR product was then cloned into pGEM3Zf(ϩ). For labeling, the construct was linearized either with HindIII for T7-generated sense transcripts or with EcoRI for Sp6generated antisense transcripts, and riboprobes were produced using digoxigenin-substituted UTP (Lofstrand Labs, Ltd.).
Plasmids-The reporter plasmid pFR-LUC, containing five copies of the Gal4 upstream activating sequence (UAS) upstream and referred to as (UAS) 5 -LUC, was obtained from Stratagene. The vectors pM and VP16, and the ␤-galactosidase reporter plasmid pCMV␤ were purchased from CLONTECH. pM-Glis1 deletion mutants were created by placing different Glis1 cDNA fragments at the 5Ј-end of Gal4(DBD). These fragments were generated by PCR using Glis1-specific 5Ј-and 3Ј-primers that included either EcoRI or BamHI restriction sites, respectively, to allow the PCR fragments to be subcloned into the EcoRI or BamHI sites of the pM vector. Details on the length of each deletion are described in the text and figure legends.
Reporter Gene Assays-Cells were plated in 6-well dishes at 2 ϫ 10 5 cells/well and 20 h later co-transfected (1-2 g of total DNA in 1 ml) with 0.25 g of (UAS) 5 -LUC, 0 -1.0 g of pM-Glis1 plasmid containing various Glis1 deletions as indicated, 1.0 -0 g of pBSK, and 0.25 g of pCMV␤, which served as an internal control to monitor transfection efficiency. Cells were transfected in Opti-MEM (Invitrogen) and 3-6 l of FuGENE 6 transfection reagent (Roche Molecular Biochemicals). Cells were incubated for 48 h and then assayed for ␤-galactosidase and luciferase activity. Luciferase activity was assayed with a luciferase kit (Promega). The level of ␤-galactosidase activity was determined using a luminescent ␤-galactosidase detection kit (CLONTECH) according to the manufacturer's instructions. Transfections were performed in triplicate and each experiment was repeated at least twice.

RESULTS
Cloning of Full-length Glis1 cDNA-Yeast two-hybrid analysis using the LBD of the nuclear orphan receptor ROR␥ (42) as bait pulled out several cDNAs encoding fragments of proteins interacting with ROR␥. One of these cDNAs encoded part of a novel protein, referred to as Gli-similar 1 (Glis1) because of its relationship to Gli proteins as will be discussed further below. Although Glis1 interacted with ROR␥ in the yeast two-hybrid assay, we have not yet been able to observe any interaction between these two proteins in mammalian two-hybrid or pulldown analysis. A cDNA containing the full-length coding region of Glis1 was obtained after screening a mouse kidney cDNA library. A part of the 5Ј-untranslated region was obtained by 5Ј SMART-RACE. The nucleotide and the deduced amino acid sequences of mouse Glis1 are shown in Fig. 1. The putative translation initiation sequence GCGCCATGC was, except for two nucleotides, identical to the Kozak consensus sequence CC(A/G)CCATGG (46). A putative polyadenylation signal was found at the end of the 3Ј-untranslated region. The open reading frame encodes for a protein of 789 amino acids with a calculated mass of 84.3 kDa. Analysis of its amino acid sequence showed that Glis1 was a neutral protein with a theoretical pI of 8.26. Glis1 was relatively proline-rich with 13.6% of its residues consisting of proline and contains a proline-rich region between Pro 632 and Pro 680 . Motif scanner analysis identified a zinc finger domain (ZFD) between Cys 368 and His 517 composed of five tandem C 2 H 2 -type zinc finger motifs with the consensus Cys-X 4 -Cys-X 12,15 -His-X 3,4 -His that are separated by sequences showing homology to the consensus sequence (T/S)GEKP(Y/F)X typically found in Krü ppel-like zinc finger proteins (47).
Comparison of the amino acid sequence of Glis1 with those in the GenBank TM data base revealed that its ZFD exhibited high homology with those of the Krü ppel-like zinc finger proteins of the Gli and Zic subfamily (7,9,18,48,49), the recently described Glis2 (50), and with the ZFD of the Drosophila Gli-like zinc finger protein glf/Lmd (40,41) (Fig. 2). Although these proteins share a highly conserved five zinc finger repeat, they exhibit little homology in regions outside their ZFD. The ZFD of Glis1 exhibits highest identity (77%) with the ZFD of glf/Lmd. Their 2nd, 3rd, 4th, and 5th zinc fingers were most highly conserved (83-88%). The ZFD of Glis1 exhibits 52-68% homology with those of the Gli and Zic subfamily, along with Glis2, and was most homologous (68%) to that of Gli2. The 3rd, 4th, and 5th zinc finger motifs of Glis1 and Gli2 were the most similar, sharing an identity of 87, 80, and 84%, respectively. Analysis of its secondary structure using the GOR3 secondary structure prediction method indicated that the second half of each zinc finger motif of Glis1 consist of a ␣-helix (Fig. 2C).
Chromosomal Localization of the Glis1 Gene-The chromosomal localization was determined by fluorescence in situ hybridization analysis using a genomic fragment of the mouse Glis1 gene as a probe. DNA was labeled with digoxigenin-dUTP by nick-translation and hybridized to normal metaphase chromosomes derived from mouse embryo fibroblasts. This resulted in the specific labeling of the middle region of chromosome 4. Labeling with a specific marker for the centromere of chromosome 4 confirmed its localization (not shown). Measurements of 10 specifically labeled chromosomes indicated that Glis1 was located at a position that was 62% of the distance from the heterochromatic-euchromatic boundary to the telomere of mouse chromosome 4, an area that corresponds to band 4C6 (Fig. 3).
Tissue-specific Expression-To examine in which tissues Glis1 was expressed, we performed Northern blot analysis using RNA from mouse placenta and 13 adult tissues. Fig. 4A shows that the radiolabeled Glis1 probe hybridized to a single 3.3-kb transcript that was most highly expressed in placenta and kidney. A lower level of Glis1 expression was observed in testis. Northern blot analysis showed little Glis1 expression in the other tissues analyzed. Glis1 expression was also examined by RT-PCR (Fig. 4B). These results confirmed the high expression of Glis1 in kidney and indicated low levels of mRNA expression in brain, colon, brown fat tissue, testis, and thymus. Glis1 mRNA was undetectable in lung, spleen, liver, pancreas, and muscle.
Whole Mount in Situ Hybridization-To determine the expression pattern of Glis1 during mouse development we performed whole mount in situ hybridization on embryos from stages E7.75 through E14.5. At early headfold stages Glis1 transcripts were detected in extraembryonic mesoderm at E7.75 (Fig. 5A) and in lateral mesoderm at E8.0 (Fig. 5B). No Glis1 transcripts were detected at E8.5 (data not shown). At E10.5, Glis1 expression was detected in the branchial arches, somites, proximal genital tubercle, and ventral mesenchyme of the tail (Fig. 5, C and D). Sectioning revealed that transcripts in branchial arches were localized distally to the mesenchyme and were most prominent in the posterior aspect of the arch (data not shown). Transverse sections of E10.5 embryos revealed that Glis1 transcripts extended throughout the presumptive dermomyotome in the interlimb region ( Fig. 5E) but were restricted to lateral mesoderm (apparent hypaxial myo-tome) in more caudal regions (Fig. 5, C and F). At E10.5 a stripe of expression was apparent in head mesenchyme below the telencephalic vesicles (Fig. 5C) and by E11.5 this domain extended posteriorly and was continuous with expression in the fused nasal and maxillary processes but excluded the nasal capsules (Fig. 5G). Glis1 expression in the frontal nasal region was evident through E14.5 (Fig. 5, H and I, and data not shown). Expression was distinct in all vibrissal (tactile) hairs of the face and nasal regions at E13.5 (data not shown) and in other hair follicles at E14.5 (Fig. 5I). Glis1 expression was detected during limb development starting at E9.5 through E14.5 in a similar pattern in both fore-and hindlimbs (with the exception of Fig. 5J in which only forelimb data are shown). During early limb bud outgrowth at E9.5 Glis1 expression was first observed in the anterior-proximal mesenchyme (Fig. 5, J  and K). At E10.5 this anterior mesenchymal expression extended along 2 ⁄3 of the proximal limb bud and expression was also detected in the posterior proximal region (Fig. 5L). At E11.5 the anterior expression domain of Glis1 extended along the entire limb, continuous with expression in the mesenchyme underlying the AER of the limb bud paddle (the "progress zone") ( Fig. 5M). Additionally, two expression domains in the anterior and posterior proximal base of the paddle/hand plate first appeared at E11.5 and were present through E13.5 as other limb expression domains faded (Fig. 5, M-O). At E14.5 Glis1 transcripts were detected in the footpads on the ventral limb surface and the joint interzone of the digits (Fig. 5,  P and Q).
Nuclear Localization-To examine the cellular localization of the Glis1 protein, CV-1 cells were transfected with plasmid DNA encoding different EGFP-Glis1 fusion proteins. Thirty h later the cellular distribution was analyzed by confocal microscopy. As shown in Fig. 6A, EGFP-Glis1 was detected in the nucleus, whereas little fluorescence was observed within the nucleoli or cytoplasm. The spacial distribution of Glis1 in the nucleus occurred in a speckled pattern as observed for many transcription factors (51). These observations indicate that Glis1 functions as a nuclear protein. InterPro Scan and Profile Scan analysis identified a potential bipartite nuclear localization signal (NLS) between Arg 511 and Lys 527 that overlaps with the 5th zinc finger motif. PSort II analysis identified two additional putative NLS motifs at Lys 12 and Lys 289 . Deletion of the last two NLS motifs had little influence on the nuclear localization of Glis1 (Fig. 6, B and C). Further deletion up to Ser 544 localized Glis1 to the cytoplasm suggesting that the region containing the zinc finger domain and bipartite NLS was important for nuclear localization of Glis1. The latter was supported by analysis of the localization of EGFP-Glis1(⌬ZFD) in which the zinc finger region between Arg 292 and Ser 544 was deleted (Fig. 6E).
Interaction of Glis1 with DNA-Because the zinc finger regions were highly conserved among Glis1, Gli, and Zic proteins, we examined by EMSA whether Glis1 and GLI1 were able to interact with similar DNA elements. As His 5 -GLI1(ZFD), the GST-Glis1(ZFD) fusion protein consisting of GST and the zinc finger domain of Glis1, was able to bind to an oligonucleotide containing the consensus Gli binding sequence (GBS) GAC-CACCCA (Fig. 7A). Increasing concentrations of unlabeled GBS oligonucleotide competed for GST-Glis1(ZFD) binding, whereas GST itself did not show any GBS binding. The specificity of the interaction of Glis1 with the oligonucleotide GBS was further examined by competition with three different mutant oligonucleotides, Mt1-3, containing several point mutations within the GBS. As shown in Fig. 7B, oligonucleotide Mt1 competed well with GBS for Glis1 binding, whereas Mt2 and Mt3 competed poorly. The latter further confirms the specific nature of the interaction of Glis1 and GBS.

Identification of Transactivation and Repressor Functions-
Mammalian monohybrid analysis was performed to assess the transcriptional activity of Glis1. For this purpose CV-1 cells were co-transfected with (UAS) 5 -CAT reporter and pM-Glis1 expression plasmid DNA encoding the Gal4(DBD)-Glis1 fusion protein. Fig. 8 shows that full-length Glis1 was unable to induce transcription of the reporter gene effectively.
Next, we examined the effect of a series of amino-and car- boxyl-terminal deletions on the transcriptional activity of Glis1. As shown in Fig. 8, carboxyl-terminal deletions had little effect on the transactivation activity of Glis1. In contrast, deletions at the amino terminus had a major impact on Glis1 activity. Amino-terminal deletions up to Leu 98 had little effect on Glis1 activity. However, deletion of the amino terminus up to Lys 317 (Glis1(⌬N317)) caused a 30 -40-fold induction of Glis1dependent transactivation, whereas an additional 4 -5-fold increase was observed when the region up to Ser 544 was deleted (Glis1(⌬N544)). Although the level of transactivation decreased 70% upon further deletion (as for Glis1(⌬N697) and Glis1(⌬N757)), these mutant Glis1 proteins were still able to substantially activate transcription of the reporter. These results suggest that Glis1 contains a strong activation domain at its carboxyl-terminal region and that the activity of this acti-vation function was suppressed by a repressor domain at its amino-terminal half.
To further map the amino-terminal repressor domain(s), the effect of several additional amino-terminal deletions was examined. As shown in Fig. 9A, analysis of several Glis1(⌬N) mutants demonstrated that deletion of the amino terminus up to Phe 150 has little effect, whereas deletion up to Arg 200 induced transcriptional activation about 6-fold. An additional 12and 3-fold increase in transactivation was observed when regions up to Lys 317 and Ser 544 were deleted. These results appear to indicate the presence of a major repressor function within the region between Phe 150 and Lys 317 , whereas an additional repressor function was associated with the region between Gly 459 and Ser 544 .
The location of the transactivation function was further mapped by examining the effect of a series of carboxyl-terminal deletions on the transactivation activity of Glis1(⌬N544) that lacks the repressor function. As demonstrated in Fig. 9B, Glis1(⌬N544) induced transcriptional activation of the LUC reporter about 100-fold. Deletion of the first 15 amino acids at the carboxyl terminus reduced reporter activity by more than 90%, indicating that this region was essential for Glis1-mediated transactivation (Fig. 9A). The results in Figs. 8 and 9B suggest that the region between Ala 618 and Thr 789 of Glis1 was required for optimal transactivation activity.
Glis1-mediated Transactivation Is Cell-type Dependent-To determine whether the transcriptional activity by Glis1 was limited to CV-1, we compared the transactivating activity of Glis1(⌬N317) and Glis1(⌬N544) in several different cell lines using monohybrid analysis. As shown in Fig. 10, Glis1(⌬N317) and Glis1(⌬N544) induced transcriptional activation of the LUC reporter gene in all cell lines tested, however, the magnitude of the induction differed greatly. Glis1(⌬N544) induced a 270-fold increase in reporter activity in 293 cells and a 120-and 65-fold increase in CV-1 and COS-7, respectively, compared with a 6-and 15-fold induction in Chinese hamster ovary and PK-1 cells. These differences in the degree of transactivation were not because of the fact that Glis1 was not transported to the nucleus because in all cell lines tested Glis1 localized to the nucleus (not shown). Alternatively, these differences in transcriptional activation among cell lines may be related to different levels of expression or activation of co-activators able to interact with Glis1. Stimulation of Glis1-dependent Transcriptional Activation by CaMKIV-The activity of transcription factors was often regulated by specific protein kinase signaling pathways. Fig.  11A demonstrates that Glis1(⌬N317)-dependent transcriptional activation can be modulated by Ca 2ϩ -dependent calmodulin kinases. Co-transfection with a plasmid encoding a constitutively active CaMKIV (CaMKIV*) caused a 4 -7-fold enhancement in Glis1(⌬N317)-dependent transactivation. CaMKI* increased reporter activity about 1.5-2-fold, whereas CaMKII* had little effect. Addition of the CaMK inhibitor KN93 reduced the CaMKIV*-induced stimulation in a concentration-dependent manner (Fig. 11B). As shown in Fig. 11C, CaMKIV* increased the transcriptional activity of Glis1 (⌬N317), Glis1(⌬N544), and Glis1(⌬N757) but had little effect on the transactivation activity of full-length Glis1, Glis1(⌬C594), or Glis1(⌬462). These results indicate that the transactivation enhancing effect of CaMKIV does not involve the repressor domain of Glis1 but only depends on the carboxylterminal region containing the transactivation function. Sequence analysis of the carboxyl terminus did not reveal any potential CaMKIV phosphorylation sites suggesting that the stimulation by CaMKIV may not be because of phosphorylation of Glis1 itself. DISCUSSION In this study, we describe the cloning and sequence of a cDNA encoding a novel Krü ppel-like zinc finger protein that was named Glis1 based on its relationship to Gli proteins. As Gli proteins, Glis1 contains a ZFD that comprises five tandem zinc finger motifs with the consensus Cys-X 2,4 -Cys-X 12,15 -His-X 3,4 -His. The motifs are separated by sequences homologous to the consensus sequence (T/S)GEKP(Y/F)X, a typical feature in Krü ppel-like "zinc finger" proteins. The ZFD of Glis1 exhibits highest homology with members of the Gli and Zic subfamily of Krü ppel-like zinc finger proteins, the Gli-related proteins Glis2, and Drosophila glf/Lmd. These proteins exhibit little homology outside their ZFDs.
ZFDs have been implicated in many functions, including DNA recognition, protein/protein interactions, transcriptional repression, and nuclear localization (52)(53)(54)(55)(56). Examination of the secondary structure of the Glis1 zinc finger domain indicated that the second half of each zinc finger consists of an ␣-helix. Crystal structure analysis of DNA-bound GLI1 revealed that the ␣-helices in the 4th and 5th zinc finger motifs in particular are involved in making DNA contacts and therefore in the recognition of specific DNA elements (57,58). Because these ␣-helices exhibit a 90% or more identity among Gli, Zic, and Glis1 proteins, one may predict that these proteins interact with very similar DNA response elements. Both Gli and Zic proteins have been demonstrated to bind DNA elements with the consensus sequence GACCACCCA (52, 56). EMSA (Fig. 7) showed that Glis1 was also able to interact with oligonucleotides containing this consensus sequence in a specific manner.
Confocal microscopic analysis demonstrated that Glis1 was localized primarily to the nucleus where it was distributed in a speckled-like pattern. The latter suggests that Glis1 was associated with a larger protein complex. Deletion mutant analysis showed that absence of the amino terminus containing the first two putative NLS motifs had little effect on nuclear localization and demonstrated that the region containing the ZFD and the bipartite NLS was essential for nuclear localization of Glis1. In transcription factors, NLS motifs have been demonstrated to often overlap with the DNA-binding domain (59), and zinc finger motifs themselves can be involved in nuclear translocation as has been demonstrated for Gli and Zic proteins (55). The nuclear localization of Gli proteins was regulated at multiple levels that involves Sonic hedgehog signaling, nuclear import and export signals, and interactions with Fused (FU) and Suppressor of Fused (SUFUH) (60 -62). In addition, the nuclear localization of Gli proteins can be facilitated through heterodimerization with Zic proteins (56). Future studies have to ascertain the precise mechanisms that determine the nuclear transport of Glis1.
The demonstration that Glis1 is a nuclear protein and can bind the consensus GBS suggested that Glis1 functions as a transcription factor. Monohybrid analysis in several cell lines using the full-length Glis1 showed that Glis1 was not a very effective inducer of transcription. However, deletion of the amino terminus converted Glis1 into a strong transcriptional activator suggesting the presence of a repressor domain at the amino terminus. The amino-terminal region does not exhibit any resemblance with the Krü ppel-associated box, an evolutionarily conserved repressor domain found in approximately one-third of the Krü ppel-like zinc finger proteins (63, 64), and does not contain a SCAN box, a repressor domain identified in several Krü ppel-like zinc finger proteins (6). The repressor domain in Glis1 appears to be novel and does not exhibit homology with the repressor domains recently identified in Glis2, a protein related to Glis1 (50). Repressor functions are often mediated through an interaction of the repressor domain with nuclear co-repressors. For example, the repression by the Krü ppel-associated box is mediated through interaction with the co-repressor TIF1 (4, 5, 65). As mentioned above, the speckled-like distribution of Glis1 in the nucleus suggests association with a larger protein complex and possibly a co-repressor complex. Two-hybrid analysis using the Glis1 repressor domain as bait may help to identify (novel) co-repressors interacting with Glis1.
Deletion analysis demonstrated that the region from Ala 618 and Thr 789 at the carboxyl terminus of Glis1 was important for its transactivation function. The region has only a weak resemblance to the TAF II 31 sequence identified in Gli proteins (66). Because full-length Glis1 is not a very effective inducer of transcription, its function as a transcriptional activator will likely require activation through a specific mechanism that results in the release of a co-repressor, and/or association with a co-activator. Although little is known about the signaling pathways that regulate the activity of Zic proteins, the Sonic hedgehog-Patched-Smoothened signaling pathway has been linked to activation of Gli proteins (8,11), which in the case of Gli2 and Gli3 involves removal of an amino-terminal repressor domain by proteolytic cleavage (14). This might be a potential mechanism of Glis1 activation because deletion of the aminoterminal region converts Glis1 into an effective inducer of transcription. As reported for Gli/Cubitus interruptus proteins (11,62,67), activation of Glis1 could involve many other mechanisms, including (de)phosphorylation of Glis1. Alternatively, phosphorylation of a co-repressor or co-activator might cause, respectively, its release from or its association with Glis1. The observed increase in Glis1-mediated transcriptional activation by CaMKIV appears to be one such mechanism. CaMKIV has been reported to preferentially phosphorylate substrates with the consensus motif ⌽RXX(S/T), in which ⌽ and X are, respectively, a hydrophobic and any amino acid (68). Based on this consensus motif, Glis1 contains three potential CaMKIV phosphorylation sites at Ser 72 , Ser 187 , and Thr 458 . Because CaMKIV is also able to enhance transcription by the deletion mutant Glis1(⌬N544) that does not contain any potential CaMKIV phosphorylation sites, it appears unlikely that the increase in transactivation by CaMKIV involves phosphorylation of Glis1 itself. Relief of repression of the transcriptional factor MEF2 induced by CaMKIV has been reported to be because of phosphorylation of histone deacetylases and their subsequent transport to the cytoplasm (45). Because the Glis1(⌬N544) mutant lacks the repressor domain, the transcriptional stimulation by CaMKIV does not appear to involve the repressor domain of Glis1, or the phosphorylation and release of co-repressors. Alternatively, the induction by CaMKIV may be because of phosphorylation and activation of a co-activator(s) resulting in an enhanced interaction with Glis1. It is interesting to note that CaMKIV has been reported to enhance the activity of the cAMP-response element-binding protein that serves as a co-activator for a number of different transcription factors (69).
Recent studies have indicated that members of the BMP and Wnt families of proteins, in addition of being downstream targets of Gli proteins, may also influence the expression or activity of Gli proteins (22)(23)(24). In light of the induction of Glis1- mediated transactivation by CaMKs, it is interesting to note that certain Wnt signals cause an increase in intracellular Ca 2ϩ and activation of calcium-dependent CaMKs (70). Therefore, it will be interesting to investigate whether Wnt signaling pathways act upstream from Glis1 activation.
The expression of Glis1 in adult mouse tissues was rather restricted. Glis1 was highly expressed particularly in placenta and kidney and at lower levels in testis, whereas it was barely detectable in other tissues. Although Northern analysis established high levels of Glis1 expression in adult kidney tissue, in situ hybridization studies of embryonic/fetal metanephroi showed moderate to low levels of expression and its distribution seemed relatively uniform or slightly ductular in these tissues. This suggests that Glis1 expression may not be significant to metanephric differentiation.
Studies of the embryonic expression of Glis1 show that this gene was both temporally and spatially regulated. Glis1 expression occurs first in extraembryonic tissues and lateral mesoderm during gastrulation, subsequently appears transiently in several defined structures of mesodermal lineage, including craniofacial regions, follicles, branchial arches, limb buds, somites and myotomes, genital tubercle, and tailbud. The temporal nature of Glis1 expression was demonstrated in tissues such as the limb in which expression occurred first anteriorly and posteriorly in mesenchyme at the junction of the limb and the trunk. Subsequently, this expression expanded to include mesenchyme beneath the apical ectodermal ridge at the distal end of the limb, which eventually disappeared leaving expression only at the anterior and posterior limits of the developing foot. Later, Glis1 was also up-regulated at the joint interzone of Glis1 expression in presumptive dermomyotome suggests that this factor may function in myogenesis. Consistent with this role, Glis1 expression overlapped with genes such as Myf5 (71,72), which are associated with myogenic differentiation, a target of Sonic hedgehog signaling, and positively regulated by Gli activation. However, Glis1 expression was not detected in myotome-originating skeletal muscle precursors migrating into the limb bud. Thus, it is possible that Glis1 functions tran-siently in the dermomyotome and is silenced in cells as they migrate out of the dermomyotome. In addition, because the mesenchyme of the branchial arch contributes to skeletal muscles of the face, it may also play a significant role in this process. The speculation that Glis1 may function in myogenesis is intriguing in light of the fact that the closely related Drosophila gene, glf/Lmd (Fig. 2, A and B), plays such a role during fly development (40,41).
Although clearly distinct in sequence from members of the Gli family, the expression patterns of these related transcription factors overlap with those of Glis1. Gli1 expression occurs in early extraembryonic tissues, lateral mesoderm, and subsequently in frontal nasal mesenchyme, and mesenchyme of the branchial arch, limb, and tail. Both Gli1 and -2 are expressed in the joint interzone regions (73), although we have noted spatial differences with Glis1 especially in the limb bud (19,74). Also, Gli1 was largely associated with the epidermal component of the follicle and not with mesenchyme. Expressions of Gli2 and -3, however, were reported in craniofacial structures and mesenchyme surrounding the vibrissal follicle. The expression pattern of Glis1 differs considerably to that of the recently described family member Glis2, which is found predominantly in neural tissues and somites (50,75).
One gene associated with an overlapping pattern of expression with Glis1 is dickkopf-1 (dkk-1), which encodes a soluble secreted inhibitor of Wnt signaling that is critical to proper anterior patterning (76). Similar to Glis1, dkk-1 expression has been observed in early anterior and posterior regions of limb mesenchyme just beneath the ectoderm and eventually expands into the mesenchyme underlying the AER. Furthermore, dkk-1 and Glis1 are both expressed in the first branchial arch in mesenchymal populations just beneath the surface ectodermal layer. This mesenchyme is involved in the formation of craniofacial structures, including facial muscles and vibrissal and hair follicles, where both genes are also expressed. Their expression in mesenchyme may indicate a role in the induction of surface ectoderm in the formation of structures such as the follicle.
In summary, in this study we describe the identification of the novel Krü ppel-like zinc finger protein Glis1, which is closely related to members of the Gli and Zic subfamilies and to the Drosophila glf/Lmd. Glis1 appears to function as a transcription factor that may regulate transcription of target genes through interaction with GBS-like DNA-binding sites. Induction of gene expression may be mediated through activation of the potent activation function at the carboxyl terminus of Glis1. The transactivation activity is suppressed by a repressor domain at its amino terminus and can be enhanced by CaMKIV. The temporal and spatial pattern of Glis1 expression observed during embryonic development suggests that it may play a critical role in the regulation of specific developmental programs. Both the repressor and transactivator functions of Glis1 may be involved in this control.