Human EZF, a Krüppel-like Zinc Finger Protein, Is Expressed in Vascular Endothelial Cells and Contains Transcriptional Activation and Repression Domains*

Members of the erythroid Krüppel-like factor (EKLF) multigene family contain three C-terminal zinc fingers, and they are typically expressed in a limited number of tissues. EKLF, the founding member, transactivates the β-globin promoter by binding to the CACCC motif. EKLF is essential for expression of the β-globin gene as demonstrated by gene deletion experiments in mice. Using a DNA probe from the zinc finger region of EKLF, we cloned a cDNA encoding a member of this family from a human vascular endothelial cell cDNA library. Sequence analysis indicated that our clone, hEZF, is the human homologue of the recently reported mouse EZF and GKLF. hEZF is a single-copy gene that maps to chromosome 9q31. By gel mobility shift analysis, purified recombinant hEZF protein bound specifically to a probe containing the CACCC core sequence. In co-transfection experiments, we found that sense but not antisense hEZF decreased the activity of a reporter plasmid containing the CACCC sequence upstream of the thymidine kinase promoter by 6-fold. In contrast, EKLF increased the activity of the reporter plasmid by 3-fold. By fusing hEZF to the DNA-binding domain of GAL4, we mapped a repression domain in hEZF to amino acids 181–388. We also found that amino acids 91–117 of hEZF confer an activation function on the GAL4 DNA-binding domain.

It has been estimated that 10% of the proteins within a cell are DNA-binding transcription factors that regulate important cellular processes such as cell lineage determination, cell growth and differentiation, and temporal or cell type-specific gene expression (1)(2)(3). After binding to cognate cis-acting elements, these transcription factors either activate or repress initiation of transcription (4,5). Transcription factors are grouped into several classes, which include the helix-loop-helix, leucine zipper, homeodomain, and zinc finger protein families (2).
The zinc finger transcription factors can be classified further into subfamilies on the basis of the sequence and position of amino acid residues important for zinc binding (Cys 2 -His 2 , Cys 4 , or Cys 3 -His 1 ), the spacing between the zinc-binding amino acids, and the transcription activation or repression domains (glutamine-rich, acidic, or proline-rich domains) (6 -10). A new zinc finger subfamily was identified recently whose members are characterized by a highly conserved C-terminal region containing three Cys 2 -His 2 zinc fingers and a proline rich N-terminal domain (8, 10 -13). Members of this subfamily include the erythroid (EKLF), 1 lung (LKLF), and basic (BKLF) Krü ppel-like factors, and BTEB2 (or placental Krü ppel-like factor). All four factors transactivate gene expression after binding to DNA.
The founder of this family, EKLF, was originally isolated as an erythroid cell-specific factor by subtractive cloning (8). It binds and transactivates via the CACCC site of the ␤-globin gene promoter (8,9,14). In vitro, EKLF plays an important role in human ␥-globin to ␤-globin gene switching (11). This observation is consistent with data showing that disruption of the EKLF gene by homologous recombination in mice results in defective hematopoiesis in the fetal liver and lethal ␤-thalassemia (15,16).
The other members of the EKLF family, LKLF, BKLF, and BTEB2, were isolated by homology screening with the zinc finger regions of EKLF, Sp1, and BTEB (a GC box-binding zinc finger protein) (10,12). LKLF is expressed highly in the lung and the spleen and transactivates the ␤-globin gene via the CACCC site (10). Although BKLF is also expressed in hematopoietic precursor cells, its expression is less restricted than that of EKLF (13). Also, even though BTEB2 was isolated from a placental library with a BTEB probe, the BTEB2 zinc finger region is more homologous to the zinc finger region of EKLF than it is to that of BTEB or Sp1 (12).
To identify new members of the EKLF family that may be involved in the regulation of vascular endothelial cell function, we used the zinc finger region of EKLF to screen a human vascular endothelial cell cDNA library. We isolated a member of the EKLF family and found it to be the human homologue of mouse EZF and GKLF (17,18). Mouse EZF/GKLF has been shown to be a nuclear protein. Its mRNA is expressed highly in * This work was supported by a grant from the Bristol-Myers Squibb Pharmaceutical Research Institute and by National Institutes of Health Grants HL03194 (to M. A. P.) and GM53249 (to M.-E. L.). 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) AF022184.
We show in this report that the human homologue (hEZF) is expressed in vascular endothelial cells of an endodermal origin, in contrast to the ectodermal origin of the mouse homologue in epithelial cells. We also demonstrate that purified, recombinant full-length hEZF protein binds specifically to a probe containing the CACCC core sequence in gel mobility shift assays. In contrast to other members of the family, which are transcriptional activators, hEZF functions as a transcriptional repressor, as demonstrated by its ability to repress reporter gene activity in transient transfection assays. By gene fusion experiments, we identified both the activation domain and the repression domain within hEZF.

EXPERIMENTAL PROCEDURES
Cloning of hEZF-A cDNA probe encoding the C-terminal zinc finger region of EKLF (bp 895-1146) was generated by reverse transcription polymerase chain reaction (PCR) (19,20). The forward primer (5Ј-GAACTTTGGCACCTAAGAGGCAG-3Ј) and reverse primer (5Ј-ACGCTTCATGTGCAGAGCTAAGTG-3Ј) were designed according to the published sequence (8). The DNA fragment was labeled by random priming (Stratagene, La Jolla, CA) and used as a probe to screen a human umbilical vein endothelial cell cDNA library. Approximately 1.6 million phages were plated, transferred to nitrocellulose, and screened according to standard techniques with minor modification (20). The filters were washed initially with 0.5 ϫ SSC (75 mM sodium chloride, 7.5 mM sodium citrate) and 0.1% SDS (sodium dodecyl sulfate) at 37°C and then more stringently with 0.2 ϫ SSC and 0.1% SDS at 65°C. More than 40 clones were obtained that hybridized differentially. Six were isolated, three were sequenced, and one was characterized further. It included the entire coding region of hEZF. The cDNA was mapped by restriction digestion and sequenced from both orientations by the dideoxy chain termination method with Sequenase version 2 (Amersham, Arlington Heights, IL) or on an automated DNA Sequencer (Licor, Lincoln, NE) according to the manufacturer's instructions. Sequence analysis was performed using the GCG software package (Genetics Computer Group, Madison, WI).
Southern Blot Analysis and Chromosomal Localization of hEZF-High molecular weight genomic DNA was prepared from cultured human aortic endothelial cells (21). Genomic DNA (10 g) was digested with several restriction enzymes, fractionated on 0.8% agarose gels, and transferred to nylon membranes. The membranes were then hybridized with a random-primed hEZF cDNA probe. The final membrane wash was in 0.1 ϫ SSC and 0.1% SDS at 65°C for 30 min, after which the membranes were exposed to Kodak X-AR film at Ϫ80°C. To localize the hEZF gene, we performed PCR-based radiation hybrid panel mapping (Research Genetics, Huntsville, AL). Two oligonucleotide primers specific to the hEZF cDNA sequence (5Ј-CCACCTGGCGAGTCTGACAT-3Ј and 5Ј-CACCGTGTCCTCGTCAGCGT-3Ј) were used to amplify genomic DNA by PCR. The PCR products were separated on 1.2% agarose gels, and the results were analyzed on the worldwide web server at the Whitehead Institute/MIT Center for Genome Research (URL: http://www-genome.wi.mit.edu/cgi-bin/contig/rhmapper.pl).
Cell Culture-Bovine aortic endothelial cells (BAEC) were isolated and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics as described (22). BAEC were passaged every 2-3 days, and cells from passages 5 to 8 were used in all experiments. We used BAEC because they are easy to transfect.
Construction of Plasmids-A cDNA fragment containing bp 411 to 1873 of hEZF was amplified by PCR with Pfu DNA polymerase (Stratagene). The product was then digested at the BamHI sites that had been added to the primers. The fragment was ligated into the BamHI site of the eukaryotic expression plasmid pcDNA3 (Invitrogen) in the sense (pcDNA3-hEZF) and antisense (pcDNA3-hEZF(AS)) orientations. The open reading frame was confirmed by sequencing and by in vitro transcription and translation in a reticulocyte lysate system (Promega, Madison, WI) according to the manufacturer's instructions. The pcDNA3-EKLF plasmid was constructed by cloning the EKLF EcoRI (filled in)-BamHI fragment of pSG5/EKLF (8) into the HindIII (filled in)-BamHI sites of pcDNA3.
For the GAL4-hEZF fusion constructs, hEZF fragments were fused in-frame to the C terminus of the GAL4 DNA-binding domain (amino acids 1-147) of plasmid pSG424. Segments of hEZF were generated by PCR with Pfu DNA polymerase and 5Ј-and 3Ј-primers containing the BamHI and XbaI sites, respectively. The PCR products were digested with BamHI and XbaI and ligated into the corresponding sites of pSG424. The authenticity of the fusion constructs was verified by dideoxy chain termination sequencing.
Transient Transfections-Transient transfection assays in BAEC were performed with LipofectAMINE according to the manufacturer's instructions (Life Technologies). Cells were plated at a density of 300,000 per 60-mm dish on the day before transfection. BAEC were transfected with a total of 3 g of reporter plasmid and expression plasmid. To correct for differences in transfection efficiency, we cotransfected 0.5 g of pCMV-␤gal in all experiments. Each construct was transfected at least three times, and each transfection was performed in triplicate. Cell extracts were prepared by a detergent lysis method (Promega) 48 h after transfection, and chloramphenicol acetyltransferase (CAT) activity was assayed by a modified two-phase fluor diffusion method (22). ␤-Galactosidase activity was assayed as described (22). The ratio of CAT activity to ␤-galactosidase activity in each sample served as a measure of normalized CAT activity.

RESULTS
Isolation and Characterization of the hEZF cDNA-To identify additional members of the EKLF family that may be involved in the regulation of vascular endothelial cell function, we screened a human umbilical vein endothelial cell cDNA library using a DNA probe containing the zinc finger region of EKLF under low-stringency conditions. One of the cDNAs isolated contained 1876 nucleotides and a deduced open reading frame coding for a 470-amino acid protein with an estimated pI of 9.2. Analysis of the amino acid sequence revealed three Cys 2 -His 2 Krü ppel-type fingers at the C terminus, a prolineand serine-rich N terminus, and a potential nuclear localization signal at amino acids 371-377 (Fig. 1). A single transcript of 3.5 kilobases was detected by Northern blot analysis with this 1876-bp cDNA used as a probe in total RNA from both human aortic endothelial cells and human umbilical vein endothelial cells (data not shown). By a comparison with sequences in the GenBank™ data base, we found that our cDNA is the human homologue of the recently described mouse EZF and GKLF cDNAs (17,18). We refer to the human gene as hEZF because of its expression in endothelial and epithelial cells. A comparison of the human and mouse EZF sequences revealed 91% identity at the amino acid level (Fig. 1). The three tandem zinc finger motifs (Fig. 1, boxed) are conserved com-pletely in the human and mouse sequences.
Chromosomal Localization of the hEZF Gene-Hybridization of an hEZF cDNA probe with human genomic DNA that had been digested with BamHI, EcoRI, and PstI revealed a simple pattern of hybridization, indicating that hEZF is a single-copy gene in the human genome. To map the chromosomal location of hEZF, we carried out genomic PCR analysis against a Gene-Bridge 4 radiation hybrid panel with specific primers from the hEZF cDNA sequence. The results from the genomic PCR experiments were analyzed against a human genome data base of sequence-tagged sites at the Whitehead Institute/MIT Center for Genome Research worldwide web site. The human EZF gene mapped to chromosome 9q31. Thioredoxin and the disease locus TAL2 (T-cell acute lymphocytic leukemia-2) have been mapped to the same locus.
Binding of Recombinant hEZF to the CACCC Site of the ␤-Globin Gene-The high degree of sequence conservation among the zinc finger regions of EZF, EKLF, and LKLF suggests that hEZF may also bind to the CACCC sequence. Gel mobility shift analysis was performed with the purified recombinant full-length hEZF protein and an oligonucleotide probe encoding a CACCC site derived from the ␤-globin gene (8). Incubation of hEZF with the probe resulted in a DNA-protein complex (Fig. 2). This complex was specific because it was competed away by an unlabeled identical probe but not by an unrelated probe. Mutation of the core CACCC sequence to CACCG has been shown to obliterate the binding and transactivation of EKLF (9). In our analysis (Fig. 2), an unlabeled probe with this single base mutation failed to compete for binding, indicating that hEZF binds specifically to the CACCC site.
hEZF Represses Transcription in Transient Transfection Experiments-All members of the EKLF family identified before hEZF function as transcriptional activators. In particular, EKLF, LKLF, and BKLF have been shown to transactivate reporter plasmids via the CACCC site (8 -10, 13). Because hEZF bound to the CACCC site, we decided to determine the effect of hEZF on a CAT reporter plasmid (pCAC-tkCAT) that contains a single copy of the ␤-globin CACCC site upstream of the minimal thymidine kinase promoter (8). Cotransfection of pcDNA3-hEZF decreased the promoter activity of pCAC-tkCAT in a dose-dependent manner in BAEC (Fig. 3A). A 10 to 1 expression plasmid to reporter plasmid ratio resulted in a 6-fold repression. This repression was specific because cotransfection of the antisense plasmid pcDNA3-hEZF(AS) had no effect on activity. In contrast, cotransfection of pcDNA3-EKLF increased CAT activity by 3-fold in BAEC (Fig. 3B). These results demonstrate that hEZF functions as a transcriptional repressor in our transient transfection system.
hEZF Contains Transcriptional Activation and Repression Domains-To identify domains in hEZF that may mediate its transcriptional effect, we generated a series of plasmids containing various fragments of hEZF fused to the DNA-binding domain of the yeast transcription factor GAL4 (Fig. 4A). The fusion plasmids were cotransfected with a reporter construct containing five GAL4-binding sites in front of the thymidine kinase minimal promoter (pGAL4 5 tkCAT). The GAL4-hEZF plasmid containing hEZF amino acids 2-470 had little effect on reporter activity. In contrast, the plasmid coding for amino acids 2-388 (from which the three zinc fingers had been removed) increased transcription by 25-fold (Fig. 4B). These data indicate the presence of a potent activation domain between amino acids 2 and 388 of hEZF that is inhibited by the presence of the zinc finger domain. The ability to transactivate was retained when the N-terminal 90 amino acids were deleted (GAL4-hEZF(91-388)). However, the ability to transactivate was lost with deletion of a further 23 amino acids from the N terminus (GAL4-hEZF(114 -388)). The region containing 97 amino acids N-terminal of the zinc fingers (GAL4(292-388)) or the zinc finger region alone (GAL4-hEZF(386 -470)) did not affect CAT activity.
These N-terminal deletion experiments mapping the hEZF activation domain to amino acids 91-114 (Fig. 4B) are supported by a series of C-terminal deletion experiments (Fig. 4B). GAL4-hEZF constructs coding for hEZF amino acids 2-117 or 2-180 were able to increase transcription by more than 40-fold. To determine whether the hEZF activation domain was modular, we made a fusion construct containing amino acids 91-117 of hEZF and the GAL4 DNA-binding domain. GAL4-hEZF(91-117) increased transcription by more than 30-fold (Fig. 4B), indicating that a potent modular activation domain is located between amino acids 91 and 117.
The diminished transcriptional activity of GAL4-hEZF(114 -388) in comparison with that of GAL4(1-147) (Fig. 4B) suggested the presence of repression domains C-terminal of amino acid 114. Furthermore, the enhanced activation obtained with deletions GAL4-hEZF(2-180) and GAL4-hEZF(2-117) over that obtained with GAL4-hEZF(2-388) (Fig. 4B) is consistent with the loss of a domain important for repression. To identify the repression domain(s) in hEZF, we generated additional plasmids and assayed their effect on the pGAL4 5 tkCAT reporter. Plasmids containing hEZF amino acids 140 -388, 163-388, and 181-388 repressed transcription by 4 -5-fold (Fig. 5). However, plasmids containing segments C-terminal of amino acid 240 showed no repression activity. The region containing the zinc fingers alone (GAL4-EZF(386 -470)) or that containing the fingers in conjunction with the 47 amino acids N-terminal of them (GAL4-EZF(342-470)) did not repress transcription. Our data from the N-terminal deletion analysis suggest that hEZF amino acids 181-240 may contain the repression domain. To further define the C-terminal border of this repression domain, we generated GAL4-fusion constructs containing hEZF amino acids 181-325 and 178 -244. Neither construct repressed transcription (Fig. 5). Thus, the repression domain of hEZF appears to be contained within amino acids 181-388. DISCUSSION Using the zinc finger region from EKLF as a probe to screen a human endothelial cell cDNA library, we have isolated hEZF, a new member of the EKLF multigene family. hEZF maps to chromosome 9q31, close to the T-cell acute lymphocytic leukemia-2 disease locus. Further investigation will be required to determine whether hEZF is related to this disease. Although it has been shown that the zinc finger region of hEZF binds to DNA fragments containing the CACCC motif (18), our experiments are the first to show that the full-length hEZF protein binds to this sequence (Fig. 2). It has been shown that all FIG. 3. Effects of hEZF or EKLF expression plasmids on a CACCC-containing reporter. A, hEZF represses CACCC-containing reporter activity. The pCAC-tkCAT reporter plasmid was cotransfected into BAEC with empty vector (pcDNA3) or expression plasmids in the sense (hEZF) and antisense (hEZF (AS)) orientations at the indicated ratios of expression plasmid to reporter plasmid. The degree of repression (fold repression) was determined by dividing the average normalized CAT activities measured from pcDNA3-transfected samples by those measured from hEZF-or hEZF(AS)-transfected samples. Error bars show standard deviations. B, BAEC were cotransfected with 0.5 g of reporter plasmid pCAC-tkCAT and 2.5 g of empty vector (pcDNA3), the hEZF expression plasmid (pcDNA3-hEZF), or the EKLF expression plasmid (pcDNA3-EKLF). pCMV-␤gal (0.5 g) was also cotransfected for each construct to correct for differences in transfection efficiency. Fold activation/repression was determined as in A. 4. hEZF contains a potent, modular activation domain. A, GAL4-hEZF deletion constructs. Ⅺ, GAL4 DNA-binding domain; o, zinc fingers. B, transactivation by a series of GAL4-hEZF fusion plasmids harboring successive N-terminal and C-terminal deletions. BAEC were transfected transiently with 2.5 g of pGAL4 5 tkCAT and 0.5 g of the indicated expression plasmid. For each construct, 0.5 g of pCMV-␤gal plasmid was cotransfected to correct for differences in transfection efficiency. Fold activation data represent the degree of activation obtained with the hEZF fusion plasmids relative to that obtained with empty vector (GAL4(1-147)). Error bars indicate standard deviations.

FIG.
previously known members of the EKLF family function as transcriptional activators: EKLF, LKLF, and BKLF activate transcription via the CACCC site of the ␤-globin promoter, and BTEB2 activates transcription via the promoter's GC box (8,10,12,13). We show here that in contrast to EKLF, hEZF represses transcription when transfected into vascular endothelial cells (Fig. 3A). The ability of hEZF to function as a transcriptional repressor is similar to that of several other Cys 2 -His 2 zinc finger transcription factors, such as ZNF 174 (24), ZBP-89 (25), and Gfi-1 (26).
We next wanted to identify the functional domains important for the transcriptional activity of hEZF. By gene fusion experiments, we mapped the repression domain of hEZF to amino acids 181-388. Unlike transactivation domains, repression domains are less well characterized (4,27). A few of the known repression domains are rich in alanines (28), basic residues (29,30), and prolines (4,(31)(32)(33). The repression domain of hEZF is rich in prolines (18%). Runs of proline residues adopt a single preferred conformation, known as the polyproline II helix, that is important for protein-protein interactions (34). The repression domains of WT-1, Eve (Even-skipped, a Drosophila homeodomain protein), and Mig1 (a zinc finger protein that mediates glucose repression in yeast) are also rich in prolines (4,(31)(32)(33).
The hEZF zinc fingers had no effect on transcription when fused with GAL4(1-147). Deletion of the zinc fingers, however, revealed a potent activation domain in the rest of the hEZF molecule (Fig. 4). Further mapping localized a 27-amino acid activation domain rich in leucine, serine/threonine, and acidic residues (with an estimated pI of 3.6). The acidic nature of this activation domain is similar to that of the activation domains of GAL4, GCN4 (2), and EKLF (35). Like other members of the Cys 2 -His 2 zinc finger protein family (such as Egr-1 (27), WT-1 (36), Krü ppel (37), and EKLF (35)), hEZF contains activation as well as repression domains. The presence of activation and repression domains may allow Cys 2 -His 2 zinc finger proteins to alter their function as the situation dictates (38,39). A potential switch between a positive and negative transcriptional effect could depend on an interaction with other factors that may change the conformation of hEZF to expose either the activation or the repression domain (40 -42). For example, the thyroid hormone receptor binds a corepressor to repress transcription in the absence of thyroid hormone. Hormone binding alters the receptor's conformation and leads to the release of the bound corepressor and recruitment of a coactivator. Thus, the hormone-bound thyroid receptor acts as a transcriptional activator (41). Under conditions other than those examined here, hEZF may also act as an activator depending on its binding to other factors.

FIG. 5. Identification of a repression domain in hEZF.
A, GAL4-hEZF deletion constructs. Ⅺ, GAL4 DNA-binding domain; o, zinc fingers. B, BAEC were transfected transiently with 2.5 g of pGAL4 5 tkCAT and 0.5 g of the indicated expression plasmid. For each construct, 0.5 g of pCMV-␤gal plasmid was cotransfected to correct for differences in transfection efficiency. Fold repression data represent the degree of repression obtained with the hEZF fusion plasmids relative to that obtained with empty vector (GAL4(1-147)). Error bars indicate standard deviations.