Negative Regulation of β Enolase Gene Transcription in Embryonic Muscle Is Dependent upon a Zinc Finger Factor That Binds to the G-rich Box within the Muscle-specific Enhancer*

We have previously identified a muscle-specific enhancer within the first intron of the human β enolase gene. Present in this enhancer are an A/T-rich box that binds MEF-2 protein(s) and a G-rich box (AGTGGGGGAGGGGGCTGCG) that interacts with ubiquitously expressed factors. Both elements are required for tissue-specific expression of the gene in skeletal muscle cells. Here, we report the identification and characterization of a Kruppel-like zinc finger protein, termed β enolase repressor factor 1, that binds in a sequence-specific manner to the G-rich box and functions as a repressor of the β enolase gene transcription in transient transfection assays. Using fusion polypeptides of β enolase repressor factor 1 and the yeast GAL4 DNA-binding domain, we have identified an amino-terminal region responsible for the transcriptional repression activity, whereas a carboxyl-terminal region was shown to contain a potential transcriptional activation domain. The expression of this protein decreases in developing skeletal muscles, correlating with lack of binding activity in nuclear extract from adult skeletal tissue, in which novel binding activities have been detected. These results suggest that in addition to the identified factor, which functionally acts as a negative regulator and is enriched in embryonic muscle, the G-rich box binds other factors, presumably exerting a positive control on transcription. The interplay between factors that repress or activate transcription may constitute a developmentally regulated mechanism that modulates β enolase gene expression in skeletal muscle.

primarily expressed in the cardiac and skeletal muscles, where, during embryonic development, the ␤ isoform progressively replaces the nearly ubiquitous ␣ isoform (5). Indeed, the message for the ␤ enolase is first detectable in the cardiac tube and in the myotome (11); expression remains extremely low in skeletal primary fibers, whereas a striking increase occurs in the second generation of fibers at the fetal stage of development (6,11), and a further increase is observed after birth. In the adult, ␤ enolase is expressed in both cardiac and skeletal muscles, with higher levels of expression detected in fast-twitch fibers than in slow-twitch fibers (11). In vitro studies on myogenic cell lines and primary myoblasts have shown that the level of ␤ enolase expression increases with terminal differentiation but, at variance with the majority of the muscle-specific genes, expression already occurs in proliferating myoblasts (6 -8). Interestingly, ␤ enolase, like desmin (9), belongs to a relatively small group of muscle-specific genes expressed in proliferating myoblasts, as well as in differentiated myotubes, and it has been suggested that it might be a marker of adult satellite cells in humans (8,10).
From these data, it can be presumed that regulation of the ␤ enolase gene expression may take place at multiple levels and involve complex molecular mechanisms, making the gene a suitable model to investigate various aspects of muscle-gene transcriptional control.
In the last few years, a great number of transcription factors and DNA regulatory elements have been identified as contributors to the activation of the muscle differentiation program. In skeletal muscle, the MyoD family of basic helix-loop-helix proteins (MyoD, myogenin, Myf-5, and MRF4), the ectopic expression of which can activate skeletal muscle gene expression in a wide range of non-muscle cell types, plays a pivotal role during development (12). The MEF-2 family of MADS-box transcription factors, which bind an A/T-rich element found in the promoters and enhancer of the majority of skeletal and cardiac muscle genes, is involved both in a direct and indirect mechanism of transcriptional activation (13,14).
Recently, the scenario has become more complex, because several reports have outlined the importance of ubiquitously expressed factors in association with tissue-restricted factors to maintain tissue-specific expression. Functional cooperation between elements that bind ubiquitous factors and tissue-restricted factors has been demonstrated for the regulatory regions of both cardiac and skeletal muscle genes, and in almost all cases, these sequences are located in relatively close proximity (less than 60 bp), suggesting that protein-protein interactions might be involved in the cooperation (reviewed in Ref. 15). Furthermore, it has recently been reported that an apparently ubiquitous binding activity consists itself of a complex * This work was supported in part by Telethon-Italia (projects 416 and 943 to A. G.) and MURST to S. F. and G. C. We are indebted to the UILDM-Palermo and the Fondazione Telethon for the excellent work in administering the Telethon grants. 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  composed of a ubiquitous factor and a tissue-restricted cofactor, thus expanding the number of potential regulatory combinations required for the control of tissue-specific transcription during myogenesis (16).
In our previous studies, we identified several distinct regulatory regions in the human ␤ enolase gene and characterized a muscle-specific enhancer present in the first intron of the gene (15). This element interacts through an A/T-rich box with members of the MEF-2 family of transcription factors and through a G-rich box, AGTGGGGGAGGGGGCTGCG, with a ubiquitous factor(s). Mutation of either the G-rich box, termed ␤ enolase element 1 (BEE-1), 1 or the A/T-rich box resulted in a significant reduced activity of the enhancer in transient-transfection assays, indicating that MEF-2 and BEE-1 binding factors are each necessary for tissue-specific expression of the ␤ enolase gene in skeletal muscle cells. Furthermore, sequences homologous to the BEE-1 site in association to a MEF-2 binding site are present in the transcriptional regulatory regions of several skeletal and cardiac muscle-specific genes (15), suggesting the existence of a widespread pathway of muscle-gene transcriptional control.
This article reports the isolation and characterization of a protein that binds in a sequence-specific manner to the BEE-1 element. The deduced amino acid sequence revealed the presence of four C2H2 zinc finger domains, which identify the protein as belonging to the family of transcription factors resembling the Kruppel segmentation gene product of Drosophila (17), and high similarity with a human zinc finger protein, which has been shown to bind the promoter of the gene for the V␤8.1 chain of the T-cell receptor (18). Analysis of both transcripts and proteins expression in myogenic cells and tissues, coexpression of the protein with several ␤ enolase reporter constructs, and the identification of a transferable repression domain indicated that the zinc finger factor acts as a repressor of ␤ enolase gene transcription.
Furthermore, the data reported indicate that in addition to the identified zinc finger protein that we have designated ␤ enolase repressor factor 1 (BERF-1) and the expression of which is enriched in embryonic muscle, other factors bind to the BEE-1 element in adult muscle. These results support the hypothesis of regulatory pathway involving positive and negative regulators that bind to the same site or overlapping binding sites within the muscle-specific enhancer of the ␤ enolase gene.

EXPERIMENTAL PROCEDURES
Southwestern (DNA-protein) Screening of the Skeletal Muscle cDNA Expression Library-A Zap II expression cDNA library was prepared with mRNA from limbs of 12-day mouse embryos using a cDNA synthesis kit (Stratagene). The amplified expression library was screened by the method of Vinson et al. (19) using as a probe double-stranded BEE-1 oligonucleotides (sense, 5Ј-AGCTGTTCTGAGTGGGGGAGGG-GGCTGCGCCTGC-3Ј) that had been end-labeled and ligated into concatamers. One positive clone out of 2.0 ϫ 10 6 plaques survived three rounds of plaque purification. The pBluescript phagemid was excised from the Zap expression vector by helper phage coinfection (R408 strain) according to the instructions of the manufacturer (Stratagene). The isolation of additional cDNAs was carried out by standard methods using the cDNA identified by Southwestern screening as a probe.
DNA Sequencing and Expression of the Cloned cDNAs by in Vitro Transcription and Translation-The nucleotide sequence of the cDNA clones were determined on both the sense and antisense strands by the dideoxynucleotide chain-termination method using modified T7 DNA polymerase (Sequenase; United States Biochemicals). In vitro transcription of the isolated clones was carried out with 2 g of linearized pBluescript plasmid using an mRNA capping kit (Stratagene). In vitro translation was performed with a commercially available rabbit reticulocyte lysate system according to the instructions of the manufacturer (Promega), and when needed, [ 35 S]methionine was added to the translation mixture.
RNA Isolation and Northern Blot Analysis-Total RNA was extracted from cultured cells and from limbs and hearts of adult mice or embryos isolated at 12, 14, and 16 days postcoitum (dpc) by the guanidine isothiocyanate method (20). Mouse multiple tissue Northern blot was purchased by CLONTECH. Fifteen micrograms of total RNA were fractionated by electrophoresis on denaturing agarose gel, transferred to nylon membranes, and hybridyzed as described previously (6). A 2.3-kb (kilobase) BamHI fragment containing the almost entire coding region was isolated from the A22 cDNA and used as a probe. As a control of the amount of RNA loaded per lane and to check differentiation of myogenic cells in cultures, filters were rehybridized with a chicken ␤ actin cDNA (6) and/or with a human glyceraldehyde-3-phosphate dehydrogenase (GAPD) cDNA (ATCC 5701) Generation of Polyclonal Antibodies and Immunoblot Analysis-DNA fragments encoding different portions of the ZF22 protein were subcloned into the bacterial expression vector pGEX-2T (Pharmacia). A 344-bp BstXI-AccI fragment and a 884-bp ClaI-BamHI fragment were isolated from the A22 cDNA, whereas a 2-kb EcoRV-BamHI fragment was excised from the A21 cDNA encoding the truncated ZF21 polypeptide. The three glutathione S-transferase-ZF22 fusion proteins, which consisted of the amino-terminal region from amino acid 26 to amino acid 111, the zinc finger region from amino acid 76 to amino acid 319 and the carboxyl-terminal region from amino acid 430 to amino acid 740, respectively, were overexpressed in Escherichia coli and affinity purified by binding to glutathione-linked Sepharose beads (21). Rabbit anti-ZF22 polyclonal antisera were raised against the purified fusion proteins, and antibodies were immunopurified on columns containing the respective glutathione S-transferase-fusion protein used as immunogen according to established procedures (22). For immunoblot analysis, nuclear proteins (5 g), prepared as described in the following section, or total cell lysates (30 g) obtained by extraction in 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycolate, 0.1% SDS, 1 mM dithiothreitol, 0.4 mM phenylmethylsulfonyl fluoride, 0.5 g/ml aprotinin, 1 g/ml leupeptin, 0.5 g/ml pepstatin, were resolved by electrophoresis on an SDS-7% polyacrylamide gel and electroblotted to a nitrocellulose membrane (Hybond-C, Amersham Corp.). The membrane was incubated with affinity purified anti-ZF22 antibodies (0.5 g/ml) and then with a secondary antibody conjugated to horseradish peroxidase (Pel-Freez) (1:5000).The antigen-antibody complexes were visualized by enhanced chemiluminescence (ECL kit, Amersham Corp.).
EMSAs were performed by incubating end-labeled probes (0.1 ng, about 40,000 cpm) with nuclear extracts (4 -8 g) or in vitro-translated proteins (3 l) as described previously (15). When the E-boxL was used as a probe, incubation was carried out in 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 15% glycerol as described by Ferrari et al. (26). DNA-protein complexes were resolved by electrophoresis on a 5% polyacrylamide gel in 25 mM TBE (25 mM Tris, pH 8.3, 20 mM boric acid, 0.5 mM EDTA) and visualized by autoradiography. For antibody interference, EMSAs were performed under the conditions described above except for the addition to the reaction mixture of 1 l of antiserum. Anti-Sp1 and anti-Sp3 antibodies (Santa Cruz Biotechnology) were kindly provided by Dr. H. Schoeler (European Molecular Biology Laboratory, Heidelberg, Germany), and anti-MNF serum (27) was a generous gift of Dr. R. Bassel-Duby (University of Texas Southwestern Medical Center, Dallas, TX).
Nuclear extract from cells transfected with constructs expressing various GAL4-ZF22 fusion proteins were prepared according to Hoppe-Seyler et al. (28), and DNA binding activity of the fusions was examined by EMSA using a double-stranded oligonucleotide containing one GAL4 binding site (5Ј-CTAGAGGTCGGAGTACTGTCCTCCGACT-3Ј) as a probe (29).

Construction of Reporter Expression Vectors and Plasmids Expressing Wild-type ZF22 and Its Deletion
Mutants-Details about the construction of the ␤ enolase-chloramphenicol acetyltransferase gene (CAT) expression vectors used in this study have been published (15). Reporter constructs containing multiple copies of the BEE-1 binding site or a mutated consensus site upstream of the ␤ enolase promoter, termed pB3-BEE-1w4X and pB3-BEE-1m4X, were obtained by bluntend ligation into the HindIII site of the pB3-CAT of BEE-1w or BEE-1m double-stranded oligonucleotides (four copies), respectively. Orientation and proper insertion of the four BEE-1 consensus sites was determined by polymerase chain reaction (PCR) analyses with appropriate primers and confirmed by nucleotide sequencing. A NarI-BamHI fragment containing the four wild-type or mutated consensus sites and the ␤ enolase promoter was isolated from the above described constructs and used to replace the homologous fragment present in the pB3SV-CAT plasmid to generate pB3SV-BEE-1w4X and pB3SV-BEE-1m4X, respectively. For cotransfection experiments, ZF21/22 expression vectors encoding the short or the long form of the factor, were generated by inserting a 2.3-kb BamHI fragment, isolated from the A21 cDNA, and a 2.5-kb NciI fragment, isolated from the A22 cDNA, into the BamHI and EcoRV sites, respectively, of the cytomegalovirus promoter-directed expression vector pCDNAI (Invitrogen). The resulting constructs are indicated as pCDNAI-ZF21 and pCDNAI-ZF22. Plasmids expressing fusion polypeptides of ZF22 and the yeast GAL4 DNA-binding domain were obtained by inserting into the EcoRI site of the expression vector pSG424 (30) DNA fragments encoding different portions of ZF22. Briefly, an EcoRI site with the same reading frame as the EcoRI site in pSG424 was introduced at the 5Ј end of the ZF22 coding region, using the following pair of oligonucleotide primers: forward, 5Ј -CCGGAAT-TCAACATTGACGACAAACTGGA-3Ј (oligomer A); reverse, 5Ј-ATCAT-GGGATATCATATCCTG-3Ј (oligomer B). The 221-bp EcoRI-EcoRV DNA fragment obtained by PCR was used to replace the EcoRI-EcoRV fragment of the A22 cDNA, which contains the 5Ј-untranslated sequence, and the full-length coding sequence was then isolated by restriction with EcoRI and XhoI (the latter is the 3Ј-cloning site) and used to generate GAL4-ZF22 (2-794) (see Fig. 9). Constructs c, d, f, and g (shown in Fig. 9B), which encode carboxyl-terminal deletion mutants, were generated from the plasmid containing the full-length coding region by excision of DNA fragments up to the SacI, EcoRI, AccI, and EcoRV sites, respectively. Construct e, GAl4-ZF22 (2-184), was obtained by PCR using oligomer A as forward primer and as reverse primer, oligomer C (5Ј-CCGGAATTCATAGTTCGTTCTAAAGGCAG-3Ј), which contains an EcoRI cloning site. Similarly, construct k, Gal4-ZF22 (136 -184), was obtained by PCR using oligomers D (5Ј-CCG-GAATTCGAGCCAGTAGACTTACAGAA-3Ј) and C as forward and reverse primers, respectively. To generate construct j, GAL4-ZF22 (444 -794), which encodes the carboxyl-half portion of the protein, the ClaI-XhoI DNA fragment was first subcloned into pBluescript/ET-3a (Novagen) to add supplementary cloning sites in frame with the pSG424 polylinker. Finally, constructs h and i were obtained from construct j by excision of DNA fragments up to the EcoRI and SacI sites, respectively (see Fig. 1A for a partial restriction map). Nucleotide sequence of ZF22 DNA fragments generated by PCR was confirmed by DNA sequencing, whereas proper expression and stability of the GAL4-ZF22 fusion proteins was tested by immunoblot analysis on extracts of transiently transfected COS 7 cells, using a rabbit antiserum against the DNA-binding domain of GAL4 (Santa Cruz Biotechnology) (data not shown). In the case of cotransfection experiments with the various GAL4-ZF22 mutants, the reporter construct used was pG5TKCAT. This construct was obtained by inserting a HindIII-XbaI fragment, excised from the pG5E1bTATACAT plasmid and containing five GAL4-binding sites (31), upstream of the thymidine kinase gene (TK) promoter in plasmid pBL2CAT (32).
Cell Culture, Transfection, and CAT Assays-C2C12 myogenic cells (33) were cultured as proliferating myoblasts in a growth factor-rich medium consisting of Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 20% fetal calf serum (Life Technologies, Inc.). Differentiation into myotubes was induced by exposure of subconfluent cultures to a medium containing 5% horse serum (Life Technologies, Inc.) and 5 g/ml of insulin (Sigma) for 48 -72 h. Embryonic and fetal myoblasts were cultured from limbs of mouse embryos (11 dpc) and fetuses (16 dpc) as described previously (6,34). PAF human fibroblasts, CV1 and COS7 monkey kidney cells, and C3H10T1/2 mouse fibroblasts from laboratory stocks were maintained in Dulbecco's mod-ified Eagle's medium with 10% fetal calf serum. Cells were transfected by the calcium phosphate method (35) as described previously (15). Briefly, 10 g of recombinant CAT plasmid, 2 g of the ␤-galactosidase expression plasmid pON1 (36) (used to monitor transfection efficiency), and 0.5-5 g of BERF-1 expression vectors (either pCDNAI-ZF22/21 or GAL4-ZF22 constructs) were used to transfect 3-4 ϫ 10 5 cells in a 5.8-cm-diameter culture dish. Cells were harvested 48 h later and subjected to ␤-galactosidase and CAT assays as described previously (15). All transfections were performed on multiple sets of cultures with at least two different DNA preparations for each plasmid.

RESULTS
Cloning of BEE-1-binding Proteins: Isolation of cDNAs Encoding the Zinc Finger Polypeptides ZF21 and ZF22-The BEE-1 binding site is a G-rich element required to support muscle-specific expression of the ␤ enolase gene together with an adjacent MEF-2 site. For isolation of BEE-1 binding factors, an embryonic muscle cDNA expression library was screened by the Southwestern method using a concatenated BEE-1 probe. One single clone (A21) that displayed sequence-specific binding to the BEE-1 probe was isolated (Fig. 1A). The nucleotide sequence of the cDNA insert revealed the presence of an open reading frame (ORF) of 960 bp, potentially encoding a polypeptide of 320 amino acid residues with a calculated molecular mass of 35 kDa (Fig. 1B, ZF21). Additional clones were isolated by screening the cDNA library with 5Ј and 3Ј fragments of the cDNA originally identified by Southwestern screening. One cDNA (A22) was sequenced and appeared to be identical to the one encoding ZF21 except for the presence of a longer 5Јuntranslated region and a one-nucleotide deletion. This deletion would result in a frameshift (Fig. 1A) and the creation of an ORF of 2382 bp encoding a 794-amino acid polypeptide with a predicted molecular mass of 89 kDa (Fig. 1B, ZF22). The same deletion was found in six independent clones, suggesting that the originally isolated cDNA encoding ZF21 might be representative of a rare message or be the result of a reversetranscriptase error that occurred during the preparation of the library.
A search in the GenBank TM data base revealed a high similarity (79% at the nucleotide level and about 89% at the amino acid level) with a human cDNA encoding a CACCC-box-binding protein called ht␤, previously identified as one of the factors binding to the promoter of the gene for the V␤8.1 chain of the T-cell receptor (18). The cDNA encoding ht␤ contains an ORF of 1362 nucleotides, predicting a protein of 49 kDa, and the sequence showed several transitions and two insertions that would shift the frame and create several stop codons not present in the cDNA encoding ZF22. Recently, while this manuscript was in preparation, the isolation of the rat homologue of ZF22 was reported, and the factor, termed ZBP-89, was shown to bind the human gastrin promoter (37); independently, other authors reported the isolation of a mouse cDNA encoding an amino-terminal shorter form of ZF22, named BFCOL1, which binds the proximal promoters of the two mouse type I collagen genes (38). In this last report, the isolation of a partial cDNA for human ht␤ was also described, the sequence of which shows a continuous ORF and suggests that the major form of the factor in human cells might be a polypeptide with a predicted molecular mass of 89 kDa, as has been found for mouse and rat. The amino acid sequence is extremely well conserved among mouse, rat, and human proteins (the mouse and rat sequences are more than 99% similar, and the amino-terminal 400 amino acids of ht␤ share about 95% similarity with the other sequences) and displays several distinctive features (Fig. 1B). These include the presence of four C2H2 Kruppel-like zinc finger motifs, an amino-terminal acidic domain, and two basic domains (both containing a nuclear localization signal (39), located upstream and downstream of the zinc finger cluster.
The amino acid sequences of ZF22, BFCOL1, and ZBP-89 also show the presence of a PEST domain (NSSDVPEVTQSE) that has been identified in a number of eukaryotic proteins characterized by a short half-life (40) (Fig. 1B). Alignment of the nucleotide sequences encoding ZF22, ZBP-89, ht␤, and BFCOL1 showed that the latter is shorter at the 5Ј-end, where three in-frame AUG codons are present in close proximity. According to a recent report (41) no one of these putative translation initiation codons is enclosed in a full consensus sequence. The AUG codon encoding the third methionine in ZF22, ZBP-89, and ht␤ has been indicated as the translation initiation site of BFCOL1; however, the nucleotide sequence preceding this methionine residue is conserved up to the codon specifying the lysine in position 6, and the first 25 nucleotides, which diverge completely, display a continuous ORF (Fig. 1C). Additional differences between ZF22 and BFCOL1 are due to four A 3 C transitions, which result in three amino acid substitutions at residues 283 (lysine to asparagine), 344 (lysine to proline), and 349 (lysine to glutamine). These differences may represent sequence polymorphisms in the mouse genome.
Identification of ZF22 as the Major BEE-1-binding Protein in C2C12 Nuclear Extracts-To confirm the predicted ORF of both A21 and A22 cDNAs and to check the ability of the encoded polypeptides to bind the BEE-1 element, the pBluescript plasmids carrying the two cDNAs were in vitro transcribed and translated. The major product of the ZF22-encoding plasmid was a polypeptide migrating with an apparent molecular mass of about 110 -120 kDa, whereas two major polypeptides of about 50 and 43 kDa were detected in the translation mixture programmed with the ZF21-encoding mRNA, likely resulting from the utilization of more than one translation start site (data not shown). Consistent with results previously reported for BFCOL1 (38), both ZF22 and ZF21 migrate in SDS-poly-acrylamide gel electrophoresis more slowly than expected from their predicted molecular masses (89 and 35 kDa, respectively). In gel shift experiments, using as a probe the same oligonucleotide utilized in the Southwestern screening, the in vitrotranslated ZF22 gave rise to a major DNA-protein complex comigrating with the major endogenous binding activity detected in C2C12 nuclear extracts ( Fig. 2A, compare lanes 1 and 2), whereas two faster migrating DNA-protein complexes were resolved with ZF21 ( Fig. 2A, lane 5). This result was consistent with the heterogeneity of the translation products. The specificity of the binding was confirmed by addition of a molar excess of the oligonucleotide containing either a wild-type BEE-1 consensus or a mutated consensus site ( Fig. 2A, lanes 3 and 6 and  lanes 4 and 7, respectively). These results indicated that both ZF22 and ZF21 contain a DNA-binding domain, presumably the four zinc fingers, and suggested that ZF22 corresponds to the predominant BEE-1-binding protein detected in nuclear extracts by EMSA.
To confirm that the ZF22 zinc finger protein represents a component of the endogenous BEE-1 binding activity in skeletal muscle cells, antibodies were raised against three different glutathione S-transferase-ZF22 fusion proteins and subsequently examined for their ability to interfere with the BEE-1 binding activity in C2C12 nuclear extracts. All three antisera, directed against the amino-terminal region of the protein, the zinc finger cluster, and the carboxyl-terminal region, respectively, when added to the myotubes nuclear extract, diminished the major BEE-1 binding activity and gave rise to a supershifted complex (Fig. 2B, lanes 2, 4, and 6); preimmune sera had no detectable effect (Fig. 2B, lanes 3, 5, and 7). Because the core sequence of the BEE-1 element, GGGAGG, has been shown to be an Sp1-like site (24), we decided to test whether Sp1 and Sp3 proteins bind to the ␤ enolase G-rich box using specific Sp1 and Sp3 antibodies in EMSA experiments followed by separation of the DNA-protein complexes on a more discriminating polyacrylamide gel (compare Fig. 2B and 2C). Fig. 2C shows that three closely migrating complexes and a fourth, faster mobility complex (a) were resolved. The most abundant complex was due to binding of ZF22 as confirmed by the addition of specific antibodies and the consequent appearance of a supershifted complex at the top of the gel (Fig. 2C, lane 3). Similarly, the fastest migrating complex (a), the intensity of which varied among different experiments, was supershifted in the presence of anti-ZF22 antibodies (Fig. 2C, lanes 3, 6, 8, and  9), indicating that a related protein or a breakdown product of the zinc finger factor might be part of the complex. The other two minor complexes, one slightly more retarded than the BEE-1-ZF22 complex, the other migrating relatively faster, were not reduced in the presence of the anti-ZF22-serum. Addition of specific anti-Sp1 antibody in the binding reaction resulted in reduction of the most retarded complex (Fig. 2C,  lane 4), whereas addition of Sp3 antibody resulted in the absence of the faster migrating complex (lane 5). Preincubation with both Sp1 and Sp3 antibodies resulted in the absence of both complexes (lane 7). In agreement with data reported for Sp1 and SP1-like consensus site identified in other genes (43,44), the faster migrating complex is due to Sp3 binding, whereas the most retarded complex is indeed composed of two very closely retarded bands due to specific interaction with Sp1 and Sp3. The concurrent addition of the three antisera, anti-Sp1, anti-Sp3, and anti-ZF22, resulted in the change of mobility of all the DNA-protein complexes (Fig. 2C, lane 9). When a canonical Sp1 consensus site was used as a probe, no DNA binding activity due to ZF22 was detected, as assessed by inclusion of anti-ZF22 antibodies in the binding reaction (data not shown). Taken together, these results confirmed that ZF22 is the major component of the endogenous BEE-1 binding activity in C2C12 myotubes.
The Zinc Finger Factor Expression Is Down-regulated During Myogenesis-To investigate both the pattern of expression of the ZF22 mRNA in various tissues and the relationship between its expression and that of the ␤ enolase, a Northern blot analysis of RNAs from different sources was performed. Three major transcripts corresponding to 3.4, 4.1, and 7.6 kb were observed in all of the tissues and cell cultures analyzed (Figs. 3  and 4), but other, less abundant messages larger than 7.6 were also detected, as previously reported for ht␤ and ZBP-89 (18,37). These multiple bands may be explained by the existence of additional mouse cDNAs containing 3Ј-untranslated sequences of different length as result of the utilization of different polyadenylation sites (data not shown). The amount of ZF22 transcripts increased from day 12 to day 14 of mouse embryonic limb development, followed by a remarkable decrease from day 14 to day 16 (Fig. 3A, lanes 1-3) and by a further decrease in limb skeletal muscle of newborn and adult mice (Fig. 3A, lanes 4 and 5); similarly, a lower level of expression was observed in adult cardiac muscle tissue when compared with earlier stages (Fig. 3A, lanes 6 -8). RNA was also isolated from primary cultures of embryonic (11 dpc) and fetal (16 dpc) myotubes (6,34). Fig. 3B shows that ZF22 transcripts were present in a relatively large amount in embryonic myotubes, whereas they were barely detectable in fetal myotubes. ZF22 expression both in vivo (during muscle development) and in myogenic primary cultures inversely correlates with ␤ enolase expression (6). Analysis of mRNA from adult tissues showed that the ZF22 message was ubiquitously expressed; however, the relative amount varied greatly among tissues, i.e. the message was more abundant in brain and liver than in skeletal muscle and heart (Fig. 3C). Northern blot analysis of RNA extracted from C2C12 myogenic cells at various times during differentiation induced by withdrawal of growth factors (Fig. 4A) revealed that following differentiation of myoblasts to multinucleated myotubes, the level of ZF22 transcripts slightly decreased (compare lanes 2-4 with lane 1); however, the degree of such downregulation was much less significant than in vivo during muscle histogenesis (Fig. 3, A and B). Total protein lysates were prepared from C2C12 cells maintained for the same lengths of time in differentiation medium and analyzed by Western blot (Fig. 4B). All three anti-ZF22 antibodies recognized two closely migrating polypeptides, with apparent molecular masses of about 110 and 120 kDa, clearly resolved when electrophoresis was carried out on longer gels (see Fig. 6). Both polypeptides appeared to be slightly more abundant in myoblasts than in myotubes. The specificity of the antibodies was confirmed by use of preimmune sera as a negative control; furthermore, the apparent molecular weight of the two endogenous polypeptides was consistent with the size of the ZF22 protein obtained by in vitro transcription-translation. The two proteins may be the products of alternatively spliced messages or result from posttranslation modifications of a single product. Immunofluorescence analysis (not shown) confirmed that ZF22 protein is ubiquitously expressed in the mesoderm of mouse embryos but is no longer detectable in mesoderm of mouse fetuses.
Taken together, these results indicate that the expression of ZF22 is down-regulated during myogenesis in vivo and in vitro but less tightly regulated in C2C12 myogenic cells in culture.
The DNA Binding Activity of the Zinc Finger Factor Is Diminished in Adult Skeletal Muscle Tissue Where Novel BEE-1 Binding Activities Are Detected-The discrepancy observed between ZF22 expression in muscle tissues and C2C12 myogenic cells prompted us to investigate the presence of BEE-1 binding activity in skeletal muscle tissues. Nuclear extracts were prepared from embryonic muscle, using as a source limbs of 12-day mouse embryos, and from skeletal muscle of adult mice; both extracts were used in EMSAs (Fig. 5). Strikingly, although the BEE-1-ZF22 complex was easily detected in nuclear extract from embryonic muscle (Fig. 5A, ME, lane 2), it was barely detectable in nuclear extract prepared from adult muscle (Fig.  5A, MA, lane 3). The DNA-ZF22 complex obtained with nuclear extract from embryonic muscle was not distinguishable from the one obtained with C2C12 nuclear extract (Fig. 5A, compare  lanes 1 and 2), as confirmed by the use of anti-ZF22 antibody (data not shown), whereas it was replaced by one major faster migrating complex and several minor complexes in adult muscle nuclear extract (Fig. 5, A and B, MA, lane 3). The newly detected complexes are specific because they were drastically reduced by addition of an excess of unlabeled probe containing a wild-type BEE-1 consensus site and were not affected by a molar excess of a mutated consensus site (Fig. 5B, lanes 2 and  3). Furthermore, the adult muscle-specific complexes were unaffected by addition of all three anti-ZF22 sera and antibody against MNF (data not shown), a factor reported to bind a CACCC-box within the muscle-specific enhancer of the myoglobin gene (27), indicating that the proteins in the complexes are not related to both factors. An independent assessment of the quality of the extracts is provided in Fig. 6, C and D, which show no significant difference in Sp1 and E-box site binding proteins between nuclear extracts of embryonic and adult muscle; the binding activity detected with the left E-box from the enhancer of the muscle creatine kinase gene was slightly higher in adult muscle than in embryonic tissue (Fig. 6D, lanes   FIG. 3. Northern blot analysis. A, expression of ZF22 in developing mouse skeletal and cardiac muscles. Total RNA (15 g) from limbs of 12-and 14-day embryos (lanes 1 and 2), 16-day fetuses (lane 3), and newborn (NB) and adult (MA) mice (lanes 4 and 5, respectively) and from hearts of 12-day embryos (lane 6), 17-day fetuses (lane 7) and adult mice (HA, lane 8) was electrophoresed, transferred to a nylon membrane, and hybridized with a 32 P-labeled ZF22 cDNA fragment as described under "Experimental Procedures." Sizes are indicated in kb. The same filter was washed and rehybridized with a GAPD cDNA to assess the presence of comparable amounts of loaded RNA and with a chicken ␤ actin cDNA (2) to monitor the pattern of expression of the muscle-specific form (␣ actin) and the ubiquitous form (␤ actin) (not shown). B, expression of ZF22 in differentiated cultures of embryonic and fetal myoblasts. Myoblasts were isolated from limbs of 11-day embryos and 16-days fetuses, and RNA was extracted after myotube formation on the 5th day of culture (E Mt and F Mt, respectively). The filter was first hybridized with the chicken ␤ actin cDNA, stripped, and then reprobed with a ZF22 cDNA fragment. C, expression of ZF22 in adult tissues. Poly(A)ϩ mRNA (2 g) was from mouse heart, brain, spleen, lung, liver, skeletal (Sk) muscle, kidney, and testis. The filter was probed with a ZF22 cDNA fragment and then with the GAPD cDNA. 1 and 2). Consistent with the results obtained by EMSA, a Western blot analysis of the same nuclear extracts confirmed that ZF22 polypeptides were present at a relative comparable level in extracts of C2C12 myotubes and embryonic skeletal muscle but were not detectable in a equal amount of nuclear extract prepared from adult muscle (Fig. 6, lanes 2-5). As observed in total lysates, ZF22 polypeptides are slightly more abundant in the nuclear extract of C2C12 myoblasts than in that of myotubes (Fig. 6, lanes 2 and 3). The proteins displayed the same apparent molecular weight in nuclear extract both from murine myogenic cells and muscle tissue, whereas two polypeptides with slightly different mobilities were detected in the extract of a human fibroblast cell line used as a control (Fig.  6, lane 1). The apparent molecular weight of the human proteins detected with the anti-ZF22 antibodies is consistent with a ORF much larger than the one reported for ht␤, supporting the data by Hasegawa et al. (38) on the isolation of a human cDNA clone with a continuous ORF and in agreement with a recent report on the expression of the factor in human cell lines and tissues (42).
These results indicate that a developmental down-regulation of the ZF22 binding activity is consistent with the pattern of expression of both ZF22 transcripts and proteins, and more interestingly, they demonstrate the presence of adult musclespecific complexes due to novel BEE-1 binding activities. These binding activities are absent or are not detectable in nuclear extracts of embryonic muscle and C2C12 myotubes.
Overexpression of the Zinc Finger Factor Results in Repression of Both Basal and Activated Transcription-To address the question of the regulatory role exerted by the zinc finger factor on ␤ enolase gene expression, both ZF22 and ZF21 proteins were overexpressed in transfection assays with CAT reporter plasmids either carrying the ␤ enolase promoter and the entire first intron (nucleotides Ϫ172 to ϩ706) with the musclespecific enhancer in its wild-type location or bearing the minimal enhancer sequence (nucleotides ϩ532 to ϩ611) upstream of the promoter (Fig. 7A, pB10-CAT and pB3-5ЈPCR1, respectively). In both cases, expression of ZF22 and ZF21 resulted in a consistent and comparable reduction of the CAT activity relative to the activity detected in cells transfected with the parental, insertless expression vector (Fig. 7B). The results shown in Fig. 7 were obtained in transiently transfected C2C12 myotubes, but a similar degree of repression was observed in CH310T1/2 and CV1 cells (data not shown). To further investigate the mechanisms involved in the observed transcriptional repression, different concentrations of the ZF22 expression plasmid were cotransfected with reporter constructs in which CAT gene expression is driven by the ␤ enolase promoter with multiple copies of a wild-type or a mutated ZF22 binding site (Fig. 8A, pB3-BEE-1w4X and pB3-BEE-1m4X) and similar constructs carrying the promiscuous SV40 enhancer downstream of the CAT transcription unit (Fig. 8A, pB3SV-BEE-1w4X and  pB3SV-BEE-1m4X). Fig. 8B shows that a dose-dependent transcriptional repression was observed in all cases; ZF22 was able to repress transcription from the reporter plasmids carrying four mutated consensus sites but was quantitatively less repressive on these plasmids than on the reporter plasmids containing four wild-type binding sites. When lower amounts of the ZF22 expression plasmid were transfected (Fig. 8B, 0.5 g) repression was greater when the reporter plasmid contained the wild-type binding sites, and this difference was more significant when constructs containing the strong SV40 enhancer were used as reporters (Fig. 8B, compare a, b, c, and d). Reporter constructs carrying heterologous promoters, (TK promoter or SV40 early promoter) behaved similarly, suggesting that the observed activity does not depend upon the promoter used (data not shown), although all are TATA-box containing promoters. The results of these transfection experiments indicated that both ZF22 and ZF21 exert a transcriptional repres-

FIG. 5. Analysis of BEE-1 binding activities in embryonic and adult skeletal muscle.
A, EMSA was performed with nuclear extracts (4 g) from C2C12 myotubes (Mt), embryonic muscle (ME), and adult skeletal muscle (MA) using a labeled oligonucleotide containing the BEE-1 element as probe. B, competition in the EMSA demonstrates the specificity of the binding. A 100-fold molar excess of unlabeled oligonucleotide containing the wild-type or the mutated consensus site (BEE-1w and BEE-1m, respectively) was added to the binding reaction mixture containing nuclear extract (8 g) from adult skeletal muscle (lanes 2 and 3). Specific major and minor adult muscle-specific complexes (large and small arrowheads, respectively) and the position of the free probe (BEE-1) are indicated. A nonspecific complex probably due to an uneven ions front is indicated by an asterisk. C and D, the same amount of nuclear extract (4 g) from embryonic muscle (ME) and adult skeletal muscle (MA) was assayed in parallel using as probe a labeled oligonucleotide containing a canonical Sp1 binding site (Sp1) or a left E-box from muscle creatine kinase enhancer (E-BoxL), respectively. sion activity on basal as well as activated transcription and suggested that the activity may reside in the amino-terminal half of the protein. The factor clearly repressed transcription more efficiently when it was targeted to the promoter, although the protein may be able to repress transcription by a mechanism that does not require DNA binding, such as interaction with a component of the basic transcriptional machinery.
ZF22 Contains a Transferable Repression Domain and a Positive Regulatory Domain-Because the BEE-1 element was identified as a positive regulatory element controlling ␤ enolase expression, we decided to conduct experiments to further substantiate the transcriptional repression activity of the isolated BEE-1 binding factor. The full coding region of ZF22 was connected in-frame to the DNA-binding domain of yeast activator GAL4 (amino acids 1-147) to generate GAL4-ZF22 (2-794), and different amounts of the construct were cotransfected with a reporter containing the CAT gene under the control of the TK promoter and five copies of the GAL4 binding site. Fig. 9A shows that repression by ZF22 is concentration-dependent, and no transcriptional activation was observed in a range of transfected expression vector from 5 to 4000 ng. These results were obtained by transfection in C2C12 myogenic cells, but comparable CAT activities were detected in CV1 cells (data not shown).
To define the boundaries of a putative repression domain and to investigate the presence of other functional domains, a preliminary dissection of ZF22 was performed, either introducing deletions in the full-length GAL4-ZF22 or fusing different por-tions of the protein to the GAL4-binding domain (Fig. 9B). The resulting chimeric proteins were tested for their abilities to repress or activate the GAL4-dependent reporter gene in C2C12 cells. As a control, the ability of each GAL4-fusion polypeptide to translocate into the nucleus and to display GAL4 DNA binding activity was evaluated by EMSA with nuclear extracts of transfected C2C12 myotubes (data not shown). Deletion of amino acids 665-794 did not affect the intrinsic repression activity (Fig. 9B, compare constructs b, c, and d), a further deletion up to amino acid 185 resulted in a 5-fold decrease of the CAT activity (construct e), whereas additional deletions from amino acid 139 to amino acid 184 and from amino acid 75 to amino acid 138 resulted in a CAT activity twice as much as the activity obtained overexpressing the GAL4 DNA-binding domain alone (Fig. 9B, compare constructs f and g with construct a). This weak activation was consistent with a slightly higher stability of the polypeptides encoded by the constructs spanning the amino-terminal region up to amino acid 138 and might be the result of a longer half-life of the expressed fusion proteins. These results indicate the presence of a strong repression domain between amino acids 138 and 184, as confirmed by the activity of a fusion protein containing only this domain (Fig. 9, construct k). Expression of fusion polypeptides spanning the carboxyl-terminal region of the pro- , and the entire first intron with the muscle-specific enhancer; in pB3-5ЈPCR1 the enhancer (␤-ENO Eh) has been cloned upstream of the ␤ enolase promoter as described under "Experimental Procedures." Effectors a, b, and c correspond to the expression vector pCDNAI and recombinant constructs pCDNAI/ZF22 and pCDNAI/ZF21, respectively. B, CAT assays of C2C12 myotubes cotransfected with 10 g of the indicated reporter plasmid and 5 g of the insertless effector plasmid (a) or plasmid expressing the long or the short form of the zinc finger factor (b and c, respectively). CAT activities, corrected for differences in transfection efficiencies, are compared with the activity observed with the insertless effector plasmid, which was arbitrarily set at 100%. The data are averages of at least three independent experiments, and the error bars represent S.D.

FIG. 8. Effect of overexpression of ZF22 on basal and activated transcription.
A, schematic representation of the reporter plasmids used in transient transfection assays. Each reporter contained four copies of a wild-type or mutated ZF22 binding site (BEE-1w4X and BEE-1m4X, respectively) upstream of the ␤ enolase promoter, and one set of constructs also contained the SV40 enhancer region (SV Eh) inserted downstream of the CAT transcription unit. B, CAT assays of C2C12 myotubes cotransfected with 10 g of the indicated reporter plasmid and different amounts (0.5-4 g) of the expression plasmid pCDNAI/ZF22. CAT activities, corrected for differences in transfection efficiencies, are compared with the activity observed with the reporter plasmid alone, which was arbitrarily set at 100%. The data are averages of at least three independent experiments, and the error bars represent S.D. tein resulted in a 6 -34-fold activation of transcription, suggesting the presence of a putative transcriptional activation domain (Fig. 9B, constructs h, i, and j), as previously reported for BFCOL1 (38). DISCUSSION The aim of this work was the identification of factors binding to the G-rich element (AGTGGGGGAGGGGGCTGCG, termed BEE-1) that are required, together with an adjacent MEF-2 site, to regulate tissue-specific and differentiation-induced expression of the ␤ enolase gene in skeletal muscle cells. One clone was isolated by screening of a phage expression library and used to isolate additional clones that resulted to encode a Kruppel-like zinc finger factor homologous to a human DNAbinding protein reported to bind a CACCC sequence (18). Northern blot analyses showed that the protein is ubiquitously expressed, but interestingly, expression decreases in limbs and hearts of mouse embryos during development; this down-regulation temporally correlates with the appearance of the secondary muscle fibers (45,46) and up-regulation of both ␤ enolase proteins and transcripts (6,11). Consistent with these results, in skeletal muscle nuclear extracts from adult mice, the binding activity due to the zinc finger factor was dramatically reduced, whereas novel binding activities were observed. On the contrary, in nuclear extracts from C2C12 myotubes, where relatively abundant levels of mRNAs and proteins are still present, a strong binding activity is detectable and presumably it does not allow detection of weaker binding activities. Cotransfection of the zinc finger factor expression vector with the native ␤ enolase promoter/enhancer-CAT fusion gene resulted in a reproducible repression of the CAT activity; therefore, we named the factor BERF-1 for ␤ enolase repressor factor 1. The previously identified human factor, ht␤, has been reported to exert a weak activation on the T cell receptor promoter (18), whereas in the recent report on the isolation of the rat homologue, ZBP-89, it was shown that the factor represses basal promoter activity and inhibits induction by EGF of the gastrin promoter (37); finally, the recently isolated amino terminus shorter form, BFCOL1, has been reported not to activate transcription and to repress the mouse pro-a2(I) collagen promoter only at high concentrations (38). Furthermore, based on the identity of the binding site (15) and on the molecular weight and the biochemical features of the purified proteins, we think that another, previously described human factor, H4TF1 (47,48), which binds the histone H4 CTCCC-box, might be related to BERF-1. This hypothesis cannot be proven since neither antibodies nor cDNAs encoding H4TF1 are available; however, the factor has been described as a positive regulator of histone H4 expression. Recently, it has been proposed that H4TF1 might bind a sequence within the distal enhancer of the human vimentin gene that acts as a positive regulatory element (49). This discrepancy among the activities may be due to differences in the reporters and the cell lines used, and in the case of BFCOL1, it might also be related to the structural difference observed; alternatively, different forms of the factor may exert different activities, as suggested by the identification of two polypeptides that differ slightly in their apparent molecular weight in both mouse and human cells.
In transfection assays using BERF-1 expression vectors and CAT reporter constructs containing the BEE-1 element, we observed a dose-dependent repression of the reporter gene ac- FIG. 9. ZF22 functions as a transcriptional repressor and contains a transferable repression domain. A, transfection assays were performed in C2C12 cells using the pG5TKCAT reporter plasmid with five copies of the GAL4 binding site and different amounts of the Gal4-ZF22 effector plasmid, which contains the entire ZF22 coding sequence fused to the yeast GAL4 DNA-binding domain (DBD). CAT activities, corrected for differences in transfection efficiencies, are compared with the activity observed with the reporter plasmid alone, which was arbitrarily set at 100%. The data are averages of at least three independent experiments, and the error bars represent S.D. B, in this set of transfections, the effector plasmids contained sequences encoding various portions of the ZF22 protein fused to the yeast GAL4-DBD. A schematic representation of the ZF22 protein is displayed at the top, with the relevant regions depicted as boxes, and schematic diagrams of the segments encoded by each effector plasmid are shown below, with numbers identifying the first and last amino acids. Transcriptional activity is relative to the basal level obtained with the GAL4 vector, which was assigned a value of 100. The data are averages Ϯ S.D. of three independent experiments.
tivity, but at the same time BERF-1 was able, though at a lower extent, to inhibit promoters lacking its binding site or bearing multiple copies of a mutated consensus site. One possible explanation for these observations is that BERF-1 is a bona fide repressor factor capable of inhibiting basal as well as activated transcription, according to the repression mechanisms proposed for the eukaryotic gene transcriptional regulation (50,51). Alternatively, the overexpression of BERF-1 results in "squelching" (52), as has been suggested for BFCOL1 (38). Several lines of evidence indicate that our data are consistent with a repression mechanism rather than with a phenomenon of squelching: (i) transcriptional activation was never observed under any experimental conditions we used, regardless the reporter, the cell type, and the concentration of the transfected BERF-1 expression vector; (ii) the amount of the expression plasmid required for repression is within the range that is normally used to measure transcriptional activity in transient transfection assays (53); (iii) the BERF-1 binding activity in vitro is reduced in skeletal muscle as differentiation proceeds, correlating well with up-regulation of the ␤ enolase gene; and (iv) using GAL4 fusion polypeptides, it has been shown that the factor contains a transferable repression domain (54). The BERF-1 repression domain does not contain alanine-, glutamine-, or proline-rich sequences, which are considered typical features of repression motifs present in suppressor factors like Kruppel and WT1 (55,56). Recently, the repertoire of the primary amino acid sequences within such domains has expanded as more transcription repressors are characterized (reviewed in Ref. 54). Several reports have indicated high charge as a common features among repression motifs (57)(58)(59), and the BERF-1 repression region, which contains a highly basic domain, may fall into this category. The identification of a putative activation domain in the serine-rich carboxyl-terminal region of BERF-1, a feature shared with BFCOL1 (38), leaves open the possibility that the factor or variant forms may behave as activator depending upon the promoter or the cell context.
The BEE-1 element was originally identified as a positive cis-regulatory element that functionally cooperates with the neighboring MEF-2 binding site in conferring muscle-specific transcription to the ␤ enolase gene. Although none of the substitution mutations in the BEE-1 element that we tested inhibited the binding of the repressor factor that we describe in this report and simultaneously result in an increase in the reporter activity in transfected C2C12 myotubes, 2 we propose that at early stage of differentiation activity of the ␤ enolase enhancer is repressed by the abundant presence of BERF-1, which precludes a potential activator from binding to the BEE-1 site or to an overlapping site. During muscle differentiation, the ratio of positive to negative regulatory binding activities changes in favor of the activator due to the developmental down-regulation of BERF-1 and/or by the increasing of different binding activities, as we observed in EMSA experiments with nuclear extracts from adult muscle. The proposed model implies that BERF-1 exerts its activity through competition for binding site occupancy, which is considered a passive transcriptional repression mechanism, but on the other hand, the factor possesses an intrinsic repressing activity that inhibits transcription initiation directly, and this is a mode of action of the active transcriptional repressors (50). However, as more studies contribute to unravelling repression mechanisms, this classification has become loose; for example, it has been recently reported that the human Cut homeodomain protein represses gene expression by both mechanisms: active repression and competition for binding site occupancy (60).
A candidate for the putative positive regulator competing for binding to the BEE-1 site might be Sp1, as has been proposed for the gastrin gERE element, which has been shown to bind both Sp1 and the rat homologue of BERF-1 (37). Although this possibility cannot be entirely excluded, our data indicate that different factors are probably involved. In nuclear extract from adult muscle where BERF-1 binding activity was dramatically reduced, we did not observe a consequent increasing of Sp1 binding activity but rather the appearance of novel DNA-protein complexes. These results are consistent with preliminary Southwestern analyses that indicate the presence of at least one muscle-enriched polypeptide that binds in a sequence-specific manner to the BEE-1 element. 2 Further studies are in progress to identify the positive regulator(s) acting on the BEE-1 element and clarify the molecular mechanisms controlling the developmentally regulated expression of the ␤ enolase gene in skeletal muscle.