A novel zinc finger transcription factor with two isoforms that are differentially repressed by estrogen receptor-alpha.

Estrogen receptor-alpha (ERalpha) can induce the expression of genes in response to estrogen by binding to estrogen response elements in the promoters of target genes. There is growing evidence that ERalpha can alter patterns of gene expression in response to ligand by regulating the activity of other factors through a direct protein-protein interaction. To identify other factors that are regulated by ERalpha, a yeast two-hybrid screen was performed that identified a novel Cys(2)His(2) zinc finger protein named ZER6. The ZER6 protein contains a Kruppel-associated box domain and six Cys(2)His(2) zinc fingers. Transcripts from the ZER6 gene can have alternate 5' exons and encode either a p71 or p52 isoform. The p52-ZER6 protein interacts strongly with ERalpha in the presence of 17beta-estradiol, whereas the p71-ZER6 isoform has a HUB-1 amino-terminal domain that inhibits the interaction with ERalpha. A consensus ZER6 binding element was defined using PCR-assisted binding site selection. In COS-1 cells, both the p52 and p71 isoforms can activate transcription through the ZER6 binding element; however, in the presence of ERalpha, transactivation by the p52 isoform is specifically repressed. Overexpression of the p52 isoform was able to abrogate activation by p71-ZER6. Expression of ZER6 was largely restricted to the mammary gland with a lower level of expression in the kidney. We conclude that ZER6 is a novel zinc finger transcription factor in which regulation of transcription in hormone-responsive cells can be controlled by the relative level of expression of two distinct isoforms.

Estrogen receptor-␣ (ER␣) can induce the expression of genes in response to estrogen by binding to estrogen response elements in the promoters of target genes. There is growing evidence that ER␣ can alter patterns of gene expression in response to ligand by regulating the activity of other factors through a direct proteinprotein interaction. To identify other factors that are regulated by ER␣, a yeast two-hybrid screen was performed that identified a novel Cys 2 His 2 zinc finger protein named ZER6. The ZER6 protein contains a Kruppelassociated box domain and six Cys 2 His 2 zinc fingers. Transcripts from the ZER6 gene can have alternate 5 exons and encode either a p71 or p52 isoform. The p52-ZER6 protein interacts strongly with ER␣ in the presence of 17␤-estradiol, whereas the p71-ZER6 isoform has a HUB-1 amino-terminal domain that inhibits the interaction with ER␣. A consensus ZER6 binding element was defined using PCR-assisted binding site selection. In COS-1 cells, both the p52 and p71 isoforms can activate transcription through the ZER6 binding element; however, in the presence of ER␣, transactivation by the p52 isoform is specifically repressed. Overexpression of the p52 isoform was able to abrogate activation by p71-ZER6. Expression of ZER6 was largely restricted to the mammary gland with a lower level of expression in the kidney. We conclude that ZER6 is a novel zinc finger transcription factor in which regulation of transcription in hormone-responsive cells can be controlled by the relative level of expression of two distinct isoforms.
Two human estrogen receptors (ERs) 1 have been identified: ER␣ and ER␤ (1)(2)(3)(4). These nuclear receptors are members of the steroid-thyroid-retinoic acid superfamily of transcription factors (5). In the classic model of transactivation by the recep-tor, ligand-activated ER␣ forms a homodimer, which is able to bind specific DNA regulatory sequences in the promoters of ER␣ target genes called estrogen response elements (6). This mechanism of transactivation by ER␣ induces the expression of a set of target genes in hormone-responsive tissues and tumors. Several ER␣ target genes have been described in hormoneresponsive breast tumors including progesterone receptor (7), pS2 (8), transforming growth factor-␣ (9), cathepsin D (10), HSP27 (11), and GREB1 (12). The promoters of these genes are directly activated by ER␣, and induction of target gene expression is dependent upon the ability of ER␣ to bind to DNA.
Over the past several years, data have been accumulating demonstrating that ER␣ can alter the expression of genes independent of direct DNA binding. One mechanism that has been proposed is through the ability of ER␣ to regulate the activity of other nuclear transcription factors by mechanisms involving direct protein-protein interactions. In many cases the interactions between ER␣ and other nuclear factors have been shown to be ligand-dependent. One example of this alternate mechanism of gene regulation is the effect of ER␣ on expression of genes regulated by AP1 (13). ER␣ and ER␤ have been shown to interact with AP1 with differential ligand activation. In experiments in HeLa cells, ER␣ stimulated an AP1 reporter plasmid in the presence of estrogen or anti-estrogens. However, ER␤ demonstrated activation of transcription mediated by AP1 only in the presence of anti-estrogens. In another report, ligand-activated ER␣ was shown to repress transactivation by members of the pro-apoptotic forkhead transcription factor family related to FKHR (14). In ER␣-positive MCF7 cells, alterations in cell cycle induced by FKHRL1 were abrogated by estradiol. Another study found that steroid receptors could regulate cell cycle and apoptosis through a small fraction of receptors that may be present in the cell membrane (15). In this model of regulation, ligand-activated ER␣, ER␤, or androgen receptor were able to attenuate apoptosis through activation of the Src/Shc/extracellular signal-regulated kinase pathway. These studies indicate that ER␣ can modulate the expression of genes controlled by transcription factors that are not thought to be influenced by steroid hormones. Through direct proteinprotein interactions, ER␣ has the potential of altering the function of a wide range of nuclear factors.
Understanding the mechanisms through which ER␣ regulates gene expression is an important step toward developing new therapies for diseases of hormone-responsive tissues. Using a yeast two-hybrid screen, we have identified additional proteins that interact with ligand-activated ER␣. One of these genes was found to be a novel Cys 2 His 2 zinc finger protein called ZER6, which has a Kruppel-associated box (KRAB) domain and six zinc finger motifs. Transcripts from the ZER6 gene have alternative 5Ј exons, which encode two related iso-forms: p71-ZER6 and p52-ZER6. The p71 isoform contains a HUB-1 domain, which was previously identified to be important for repression of long terminal repeat-mediated transactivation (16). The p52 isoform of ZER6 interacts strongly with ER␣ in the presence of 17␤-estradiol, whereas the p71 isoform demonstrated no binding to ER␣. PCR-assisted binding site selection was used to define a consensus ZER6 binding element (ZBE). Although both isoforms were able to transactivate through the ZBE, activation by p52-ZER6 was repressed in cells expressing ER␣. In ER␣-positive cells, overexpression of the p52 isoform was able to abrogate activation by p71-ZER6. The pattern of expression of ZER6 was analyzed by hybridization to an array of RNA from different human tissues. The expression of ZER6 was largely restricted to the mammary gland with lower levels of expression also noted in the kidney. We conclude that ZER6 is a novel mammary-specific zinc finger transcription factor with two isoforms, one of which is repressed through a direct interaction with ER␣.

EXPERIMENTAL PROCEDURES
Cell Lines-The cell lines COS-1 and MCF7 were obtained from ATCC and maintained as described previously (17).
Two-hybrid Screen-The two-hybrid screen was performed as described previously (14). Briefly, the full-length ER␣ cDNA was amplified using the oligonucleotides 5ЈERNde (5Ј-GGGGAATTCCATATGA-CCATGACCCTCCACACCAAAGCATCAGGG-3Ј) and 3ЈERBam (3Ј-G-CCAGGGGATCCTCAGACTGTGGCAGGGAAACCCTC-3Ј) and was cloned into the NdeI and BamHI site of pGBKT7 (CLONTECH, Palo Alto, CA). This vector encoding GAL4 DNA binding domain-ER␣ fusion (DNA-BD/ER) was co-transformed with the human mammary gland MATCHMAKER cDNA library (CLONTECH, Palo Alto, CA) into AH109 yeast. The yeast were selected following the instructions of the manufacturer with the exception that the plates were treated with water (carrier), 17␤-estradiol (final concentration on the plate 100 nM), or tamoxifen (final concentration 1 M). The plasmids encoding the 5Ј end of the ZER6 fragment (AD/ZER6 -5Ј) and the 3Ј end of ZER6 (AD/ZER6 -3Ј) were created by PCR amplification using clone 6 (ZER6 clone) as template and the oligonucleotides 5ЈZER269 (5Ј-GCCGAATT-CAAGAATCTCAGCCAAGACATGTTG-3Ј) and 3ЈZER411 (5Ј-CCCCTC-GAGTTAGAAGTCAATGCCACACTGAGGG-3Ј) for the 5Ј end and the primers 5ЈZER411 (5Ј-CGGGAATTCTTCAACGGCCACTCGGCCCTG-ATCCGC-3Ј) and 3ЈZER561 (5Ј-CCCCTCGAGTTACAAAACTCCCCCT-CCACTCCCT-3Ј) for the 3Ј end. The PCR products were cut with EcoRI and XhoI and were cloned into the EcoRI and XhoI sites of pGADT7 (CLONTECH Laboratories, Inc., Palo Alto, CA). The plasmids were sequenced to confirm that the ZER6 fragments were inserted in-frame with the GAL4 activation domain.
Screening the MCF7 cDNA Library-The cDNA libraries prepared from MCF7 mRNA using oligo(dT) (library C52) or random primers (library C662) have been described previously (18 -20). Approximately 20,000 phage from each library were screened as described previously (20). The probe used for hybridization to plaque lifts was prepared from the ZER6 clone identified in the yeast two-hybrid screen. The yeast plasmid was restricted with EcoRI and XhoI, and the DNA fragment was excised and labeled using an [␣-32 P]CTP and High Prime DNA labeling kit (Roche Molecular Biochemicals). Secondary and tertiary screens were performed to purify phage with ZER6 cDNA inserts. DNA was recovered as phagemids, and the plasmids were sequenced.
Cloning ZER6 Transcripts with 5Ј-RACE-The 5Ј ends of ZER6 transcripts were identified using 5Ј-RACE and was performed using the 5Ј-RACE System for Rapid Identification of cDNA Ends, version 2.0 (Invitrogen) according to the instructions of the manufacturer. In the first strand synthesis, the primer ZER6-GSP1 (5Ј-CTGTGGGAATCT-CACTTTCC-3Ј) was used. For PCR amplification of dC-tailed cDNA, the oligonucleotide ZER6-GSP2 (5Ј-GGCCCTGCCTGGTCCTCTGT-3Ј) was used. Nested amplification was performed using the primer ZER6-GSP3 (5Ј-GAGACTCGTAGTTGCCCTTCATG-3Ј), and the PCR products were cloned using the TA cloning kit (Invitrogen, Carlsbad, CA).
To create an expression plasmid for p71-ZER6, the vector pcDNA3.1 was restricted with ApaI, treated with S1 nuclease, and re-ligated to destroy the ApaI site. pcDNA3.1-HT-p52-ZER6 was cut with EcoRI and XhoI and the insert moved into the pcDNA3.1 vector lacking the ApaI site. The 5Ј end of the p71-ZER6 cDNA was amplified by RT-PCR using MCF7 mRNA as template with the primers ZER6alt5 (5Ј-GGGGAAT-TCGCCGCCACCATGCATCATCATCATCATCATGCTGAGGCGGCC-CCGGCCCCGACATCTGAA-3Ј) and ZER6short3 (5Ј-GATGACAAGCT-CTTCATCAGCATAAGT-3Ј). The PCR product was cloned using the TA cloning kit, sequenced, and excised using EcoRI and ApaI. This fragment was used to replace the 5Ј end of the HT-p52-ZER6 clone, and the resulting plasmid was called pcDNA3.1-HT-p71-ZER6.
GST Pull-down Assay-The GST fusion protein with the ligand binding domain of ER␣(246 -595aa) was expressed in Escherichia coli strain DH5␣ and purified by glutathione-Sepharose beads according to manufacturer's instructions. The p71 and p52 isoforms of ZER6 were produced using in vitro transcription/translation from the pcDNA3.1-HT-p71-ZER6 and pcDNA3.1-HT-p52-ZER6 vectors in the presence of [ 35 S]methionine using T7 polymerase and the coupled transcription/translation kit (Promega, Madison, WI). In experiments using mixed plasmids, equal amounts of the two expression plasmids were used in the reactions. Equal amount of GST-ER fusion protein was incubated with labeled ZER6 proteins in binding buffer (50 mM Hepes, 100 mM NaCl, 20 mM Tris-Cl, pH 8.0, 0.1% Tween 20, 10% glycerol, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 0.3 mM sodium vanadate, 1 mM NaF, 5 g/ml each of leupeptin and aprotinin, and 50 M ZnCl 2 ) with or without 17␤-estradiol and incubated at 4°C for 3 h. The beads were then washed with binding buffer containing 0.5% Nonidet P-40, resuspended in SDS-PAGE buffer, boiled for 5 min, and resolved on 10% SDS-PAGE, followed by autoradiography. Identification of ZBE-PCR-assisted binding site selection was performed as described previously (17). A GST-HT-p52-ZER6 fusion protein was created by cloning the HT-p52-ZER6 cDNA as an EcoRI/XhoI fragment into the vector pGEX-4T-1 (Amersham Biosciences, Inc.). The protein was bacterially produced as described previously with the exception that zinc chloride was added to the medium at a final concentration of 100 M. In each round of selection, bound oligonucleotide was recovered from the bound complex on the dried gel-shift blot after autoradiography. After four rounds of binding selection, the PCR amplification products were cloned into the TA cloning vector pCR2.1 (Invitrogen Inc., Carlsbad, CA) and 100 cloned inserts were sequenced.
The ZBE binding probe was created by annealing the two oligonucleotides 5Ј-TCGAGGGGTGGGGGTGGGC-3Ј and 5Ј-TCGAGCCCTCC-CCCACCCC-3Ј. A nonbinding probe, which had G nucleotides changed to A and T nucleotides to G was created by annealing the two oligonucleotides 5Ј-AAAAAAAAGAAAAAGAAAT-3Ј and 5Ј-AAAAATTTCTTT-TTCTTTT-3Ј. Gel shift was performed with the bacterially produced protein as described previously. The anti-HT antibody used to supershift the ZER6 complex was SC-8036X (Santa Cruz Biotechnology).
Transactivation Assays-The IL2-LUC minimal reporter has been described previously (17). ZBE-IL2-LUC was created by inserting a double-stranded ZBE oligonucleotide into the XhoI site of IL2-LUC. DNA was introduced into either MCF-7 or COS-1 cells by lipid-mediated transfection using FuGENE-6 (Roche Molecular Biochemicals) according to the manufacturer's instructions. For COS-1 cells, the cells were plated at 2.5 ϫ 10 5 cells/well in six-well plates 24 h prior to transfection and were 80% confluent at the time of transfection. Transfection medium contained zinc chloride to 100 M. Each transfection contained 200 ng of luciferase reporter plasmid, 50 ng of p␤Gal-Control vector, and 750 ng of a ZER6 expression vector or plasmid DNA (control) as indicated. All transfections were performed in triplicate, and cells were harvested 48 h after transfection. Cell extracts for luciferase or ␤-galactosidase assays were prepared using a luciferase assay system (Promega, Madison, WI). Luciferase assays were performed with 10 l of cell extract and 100 l of luciferase assay buffer. The enzyme activity was measured for 10 s in a luminometer (Analytical Luminiscence Laboratory, San Diego, CA). ␤-Galactosidase activity in cell extracts was assayed using a Galacto-Light system (Applied Biosystems, Bedford, MA).
For MCF7 cell transfections, cells were seeded at 5 ϫ 10 5 /well in six-well plates in growth medium the day before transfection. On the day of transfection, the medium was replaced with fresh growth medium. Generally, 250 ng each of reporter plasmid, 1.5 g of ZER6 expression plasmid, and 250 ng of p␤gal-control plasmid were used per sample. After 24 h of incubation in the presence of the DNA-lipid complexes, the transfection mixture was removed, the cells were washed twice with phosphate-buffered saline, and the transfection mixture was replaced with minimal essential medium with 100 M zinc chloride, 10 nM 17␤-estradiol, and 10% FBS. The cells were replaced at 37°C, 5% CO 2 until harvest. Luciferase and ␤-galactosidase activity was assayed as described above. For each sample the luciferase relative light units were normalized with the ␤-galactosidase relative light units.
Tissue-specific Expression of ZER6 -The cDNA for the ZER6 gene was hybridized to a Human RNA Master Blot (CLONTECH, Palo Alto, CA) according to the instructions of the manufacturer.

Ligand-dependent Interaction of ER␣ with a Novel Zinc
Finger Protein-The yeast two-hybrid system was used to identify proteins with ligand-dependent interaction with ER␣. Fulllength ER␣ was cloned as a fusion protein with the DNA binding domain of GAL4. A normal human mammary epithelial cell cDNA library cloned to create a fusion protein with the GAL4 activation domain was used in yeast co-transformation. Yeast transformants were screened on SD LeuϪ/TrpϪ medium to select for double transformants. To identify proteins that interact with ER␣ in a ligand-specific fashion, yeast colonies were selected on SD LeuϪ/TrpϪ/HisϪ/AdeϪ medium supplemented with 17␤-estradiol, tamoxifen, or no ER␣ ligand. Approximately 300 yeast colonies underwent a secondary screening procedure. Twenty-five colonies were identified that demonstrated a ligand-dependent growth phenotype, and four colonies were isolated that demonstrated growth only with 17␤-estradiol (data not shown). The plasmids encoding the activation domain fusion proteins were recovered from these four yeast transformants, and the inserts were sequenced. One of these inserts (clone 6) was found to contain an insert of 993 bp that encoded an in-frame protein fragment of 293 amino acids that was terminated by a stop codon. Protein analysis of the predicted amino acid sequence indicated that part of a cDNA encoding a novel zinc finger protein had been cloned. This protein has been tentatively called ZER6 (zinc fingerestrogen receptor interaction, clone 6).
The 5Ј and 3Ј regions of the ZER6 protein fragment were subcloned into pGADT7. The 5Ј region contained the first 143 amino acids, and the 3Ј region contained the carboxyl-terminal 151 amino acids. These two plasmids were transformed into AH109 yeast with the full-length clone in parallel and assayed for growth on selected media with various ER␣ ligands. As seen in Fig. 1, all transformants grew on SD LeuϪ/TrpϪ medium; however, on SD LeuϪ/TrpϪ/HisϪ/AdeϪ medium, the fulllength ZER6 insert and the 5Ј region grew only in the presence of 17␤-estradiol. No growth was observed on selective media with the plasmid encoding an activation domain fusion protein with the carboxyl-terminal 151 amino acids of ZER6. Yeast transformed with expression plasmids encoding a GAL4 DNA binding domain/p53 and GAL4 activation domain/T antigen were included as a positive control, which grew independent of ER␣ ligand. This result confirmed the ligand-dependent interaction of ER␣ and ZER6 and localized the interaction region to 143 amino acids of the ZER6 protein.
Cloning the Full-length ZER6 cDNA-The cDNA insert for ZER6 obtained from the yeast two-hybrid screen was used as a probe to screen a cDNA library prepared from MCF7 cells (an ER␣-positive human breast carcinoma cell line). Four cDNAs were obtained and were entirely sequenced. As seen in Fig. 2, the sequence of the four cDNA clones matched the genomic cosmid DJ0800G07 and allowed an identification of exon/intron borders. ZER6 was determined to be encoded by five exons spanning a genomic region of ϳ28,000 bp. All introns began with a GT and ended with AG. The first ATG in-frame occurred in the second exon, which was 127 bp long. The open reading frame extended beyond the 5Ј end of the longest cDNA, which suggested the possibility that the gene encoded a longer protein not completely contained in the cDNA clones.
To obtain longer cDNAs encompassing the full open reading frame, 5Ј-RACE was performed using primers at the most proximal ATG in the second exon. Fourteen RACE clones were obtained and are diagrammed in Fig. 3. As shown in Fig. 3, variable splicing results in a number of possible 5Ј ends for ZER6 transcripts. Five clones (I811-2, I815-1, I815-6, I815-11, and I815-7) utilized upstream exons which were not included in the genomic contig. Two of these (I811-2 and I815-1) encoded cDNAs that had a stop codon in-frame and indicated an aminoterminal boundary for the coding region. In mRNAs with this structure, the ATG in the 127-bp exon would be the 5Ј terminal ATG and would encode a protein with a predicted size of 52 kDa. Two of the 5Ј-RACE clones (I815-6 and I815-11) had identical sequence and demonstrated exon skipping. This cDNA structure extended the open reading frame; however, no in-frame ATG was encountered. Similarly, I815-13 utilized a unique exon sequence that also terminated the open reading frame and would utilize the same ATG as the aforementioned transcripts. Seven of the 5ЈRACE clones (I811-1, I815-5, I815-8, I 815-4, I815-9, I815-3, and I815-14) appear to be premature termination of larger cDNAs and do not provide any additional structural information.
Clone I815-10 was unique in that it utilized 41 bp of an exon that extends the open reading frame and contained an in-frame ATG. An mRNA with this structure would encode a predicted protein of ϳ71 kDa. The full-length cDNA encoding p71-ZER6 is shown in Fig. 4. The ATG beginning the translation of p52-ZER6 is also noted. There are six complete Cys 2 His 2 zinc fingers, which are aligned with the consensus sequence as shown in Fig. 5. Each of the six fingers utilizes a leucine as the hydrophobic residue before the His 2 domain. In addition, each of the linkers between the Cys 2 residues comprises two amino acids. A BLASTp search of the protein sequence unique to p71-ZER6 matched a region of a zinc finger protein that had extensive homology with a repression domain called HUB-1 (16). The HUB-1 domain was shown to be functionally important for repression of human T-cell lymphotrophic virus type I long terminal repeat-mediated expression. These findings suggest that the ZER6 proteins are involved in transcriptional regulation and that the p71-ZER6 and p52-ZER6 isoforms are likely to have important functional differences.
ZER6 Isoforms Have Differential Interaction with ER␣-Ligand-dependent interaction with ER␣ was confirmed using GST pull-down. The cDNAs encoding the p71-ZER6 and p52-ZER6 isoforms were cloned into an expression vector (pcDNA3.1) with an HT at the amino terminus. Both proteins were synthesized using in vitro transcription/translation and were reacted with ER␣-GST in the presence or absence of 17␤-estradiol. As seen in Fig. 6, p52-ZER6 interacted strongly with ER␣ in the presence of 17␤-estradiol. In the absence of estrogen, minimal interaction was detected. By contrast, the p71-ZER6 isoform did not interact with ER␣ either with or without ligand. In a separate experiment, the two ZER6 isoforms were synthesized together (labeled MIX). As seen in Fig.  6, only the p52-ZER6 isoform in the presence of 17␤-estradiol bound to ER␣. It should be noted that the region of ZER6 identified to interact with ER␣ is contained in both isoforms. Therefore, the HUB-1 domain of p71-ZER6 appears to inhibit ZER6 binding to ER␣.
Indirect immunofluorescence was used to demonstrate the FIG. 2. Schematic model of ZER6 cDNA clones. Four ZER6 cDNA clones were obtained from screening an oligo(dT) and random-primed MCF7 cDNA libraries. The four cDNA clones (I572-1, I572-8, I572-10, and I763-2) were sequenced and found to match the genomic cosmid clone DJ08000G07. Comparison of the sequence of the cDNA clones and the genomic clone identified five ZER6 exons. The open reading frame is shaded, and the 5Ј terminal ATG is indicated in the second exon. The open reading frame for ZER6 protein is indicated by shading. In each clone, the 5Ј terminal ATG is indicated by an arrowhead. Only clone I815-10 contained an ATG in frame that allowed translation of the 396 bp exon noted from nucleotides 19380 to 19775. The sequence of the 119-bp exon, which is unique to clone I815-13, was submitted to GenBank™ and was assigned the accession number AY049742. The sequence for the cDNA encoding p71-ZER6 was assigned the accession number AY049744, and the accession number for the cDNA encoding p52-ZER6 is AY049743.
ER␣-ZER6 interaction in intact cells. COS-1 cells were cotransfected with expression vectors for ER␣ and either p52-ZER6 or p71-ZER6. The ZER6 proteins were expressed with an HT epitope. A representative cell staining pattern is shown in Fig. 7. Expression of both proteins was nuclear with faint cytoplasmic staining seen with the anti-HT antibody. In transfections with p52-ZER6 (Fig. 7, top panel), there was co-localization of ZER6 and ER␣ evident by a yellow staining color within the nucleus. Notably, there was minimal red nuclear staining when examined by confocal microscopy indicating that the nuclear p52-ZER6 was in complexes containing ER␣. By contrast, transfections with p71-ZER6 demonstrated separate red and green nuclear immunofluorescence, which indicated that ER␣ did not co-localize with the p71 isoform.
ZER6 Is a Sequence-specific DNA-binding Protein-The finding that ZER6 contains six Cys 2 His 2 zinc fingers suggested that the protein will bind DNA. PCR assisted binding site selection was used to identify a consensus binding site for the ZER6 protein. The p52-ZER6 protein containing the His tag at the amino terminus was cloned as a GST fusion protein. The protein was expressed in bacteria and purified over glutathi-one-Sepharose. The purified HT-ZER6-GST protein was used to select for binding sequences, and, after four rounds of PCR amplification, the bound oligonucleotides were cloned and 100 inserts were sequenced. The core motif was found to be GCrich, 2 and a consensus binding site was created containing the sequence: GGGTGGGGGTGGG. As seen in Fig. 8, gel shift was used to examine DNA binding to the consensus binding site (ZBE). A complex is formed with the HT-ZER6-GST protein that is supershifted with the HT antibody. The complex can be competed with a binding oligonucleotide containing a ZBE but not a nonbinding AT-rich oligonucleotide. The conclusion from these data is that ZER6 is a sequence-specific DNA-binding protein that interacts with a GC-rich motif.
Transactivation by p52-ZER6 Is Repressed by ER␣-The results presented above suggested that ZER6 might be involved in transcriptional activation through an interaction with a ZBE sequence. To test this directly, the ZBE was linked to a minimal promoter driving expression of luciferase (ZBE-IL2-LUC). As seen in Fig. 9A, the p71 and p52 isoforms of ZER6 transactivate the reporter through the ZBE. In COS-1 cells, the two isoforms gave similar levels of transactivation. Identical results were obtained using CV-1 cells (data not shown). As seen in Fig. 9B, co-transfection with an expression vector for ER␣ specifically repressed the p52 isoform. There was a statistically significant inhibition of transactivation by p52-ZER6 when ER␣ was expressed compared with p52-ZER6 alone (p Ͻ 0.0004). There were no differences comparing transactivation by p71-ZER6 with or without ER␣ expression (p ϭ 0.51). Transactivation by the ZER6 proteins was examined in ER␣-positive MCF7 breast carcinoma cells. As seen in Fig. 9C, p71-ZER6 was able to induce expression through the ZBE, but transactivation by the p52 isoform was repressed in MCF7 cells. Because both isoforms interact with the same DNA element, it is FIG. 4. The cDNA encoding p71-ZER6. The data from cDNA clones and 5Ј-RACE were used to construct a cDNA encoding the ZER6 protein.
Utilizing the terminal 5Ј ATG will translate a protein with the predicted size of 71 kDa. As noted from 5Ј-RACE clones, the methionine encoded by the 5Ј terminal ATG utilized by the majority of cloned ZER6 transcripts is indicated by a circle and would translate a protein with predicted size of 52 kDa. The location of splice sites is indicated by arrowheads connected by a line. The six consensus zinc finger domains are indicated in bold type. The stop codon is indicated by a star, and the polyadenylation signal AATAAA is underlined. The GenBank accession number for this cDNA is AY049744. The sequence of the cDNA from MCF7 cells matched the genomic clone DJ0800G07, which is shown with the exception that nucleotide G201 was A201, nucleotide T438 was C438, and the nucleotides 2068 to 2072 (CTAGA) in the untranslated 3Ј region were deleted. None of these polymorphisms alter the protein structure. Recent BLASTp searches identified homology with partial cDNAs submitted as expressed sequence tags and listed under accession numbers BAA92577 (KIAA1339 protein), AAD45824 (similar to zinc finger proteins), and XP_032742 (KIAA1339 protein).
FIG. 5. Amino acid sequence of the six consensus Cys 2 His 2 zinc fingers of ZER6. The six Cys 2 His 2 zinc fingers of the ZER6 protein are shown aligned to highlight the consensus amino acid structure shown at the top. X indicates any possible residue, and ⌿ denotes a hydrophobic residue, which in each case in ZER6 is a leucine. Numbers to the left indicate the amino acid residue in p71-ZER6, which denotes the location of the first amino acid of the finger shown on that line. In the numbering scheme used in the text, position Ϫ1 is the residue immediately preceding the ␣ helix, which corresponds to the second residue in the X 5 block before the ⌿ residue. For the six zinc fingers of ZER6, residue Ϫ1 of the ␣ helixes is D, L, G, R, R, and Y, respectively.
FIG. 6. GST pull-down demonstrates ligand-dependent interaction between ER␣ and p52-ZER6. The p71-ZER6 and p52-ZER6 proteins were synthesized using in vitro transcription/translation and incubated with a GST fusion protein containing the ligand binding domain of ER␣. As shown, p52-ZER6 binds to ER␣ in the presence of 17␤-estradiol. The p71 isoform does not interact with ER␣. Similarly, in an experiment where the p52 and p71 isoforms of ZER6 were synthesized together (MIX), only the p52-ZER6 protein demonstrates liganddependent interaction with ER␣. FIG. 7. Indirect immunofluorescence demonstrates co-localization of ER␣ and p52-ZER6. Indirect immunofluorescence was used to examine localization of ZER6 (red) and ER␣ (green) in COS-1 cells transfected with expression vectors for the proteins indicated. Co-transfection of p52-ZER6 and ER␣ (top panels) demonstrate nuclear staining for both proteins with slight cytoplasmic staining with the anti-HT antibody. Confocal microscopy demonstrates co-localization (yellow) of p52-ZER6 and ER␣. Co-transfection of p71-ZER6 and ER␣ (bottom panels) demonstrates predominantly nuclear staining. However, confocal microscopy to examine both proteins demonstrates separate red (p71-ZER6) and green (ER␣) nuclear immunofluorescence. possible that overexpression of the p52 isoform might interfere with transactivation by p71-ZER6. To test this hypothesis, MCF7 cells were transfected with p71-ZER6 expression vector with increasing amounts of expression vector for p52-ZER6. As seen in Fig. 9D, overexpression of p52-ZER6 abrogated transactivation by p71-ZER6. There was a statistically significant difference between transfections in which the ratio of p71 to p52 expression vector was 2:1 versus 1:2 (99.3 Ϯ 7.8 versus 69.1 Ϯ 6.7, p ϭ 0.007). These data suggest that the interaction between ER␣ and p52-ZER6 repressed ZER6-mediated trans-activation. We conclude that because the p71 isoform does not interact with ER␣, p71-ZER6 is able to activate transcription independent of ER␣ expression.
ZER6 Expression Is Restricted to the Mammary Gland-The physiologic relevance of an interaction between ER␣ and ZER6 requires that the two proteins are co-expressed in cells. To determine the pattern of ZER6 expression, a RNA dot blot with RNA from 50 different human tissues was hybridized with a ZER6 cDNA probe. These results are shown in Fig. 10. The ZER6 gene was most prominently expressed in the mammary gland. This finding agrees with the fact that the ZER6 gene was cloned from cDNA libraries derived from normal mammary gland and MCF7 breast carcinoma cells. The gene is also expressed in the kidney, although the level of expression is significantly less than in the breast. Negligible expression was detected in 41 other adult human tissues and 7 human fetal tissues. Weak hybridization signals were also detected with total human and bacterial DNA. We conclude that the expression pattern of the ZER6 gene is largely restricted to the mammary gland with relatively lower levels of expression in the kidney.

DISCUSSION
In the presence of estrogen, ER␣ forms a homodimer that is able to activate transcription by binding to estrogen response elements in the promoters of target genes (6). The specific transcriptional response to ER␣ ligands varies with the cell line and is dependent upon the repertoire of ER␣ cofactors or corepressors expressed in the cell (21)(22)(23). Over the last several years, it has become clear that ER␣ can have effects on gene expression independent of the DNA binding function of the receptor. This alternate pathway of gene regulation is dependent upon the ability of ER␣ to affect the function of other nuclear factors through direct protein-protein interactions. In this report, we described a novel Cys 2 His 2 zinc finger protein, ZER6, with two isoforms (p71-ZER6 and p52-ZER6), which demonstrate differential ligand-dependent interaction with FIG. 8. ZER6 is a sequence-specific DNA-binding protein. Gel shift was performed with a DNA probe containing the ZBE as defined by PCR-assisted binding site selection. Protein extract was from bacterially expressed p52-ZER6 protein synthesized as a GST fusion protein with an HT. The ZER6 protein forms a stable complex, which can be competed with excess cold probe sequence (ZBE) but not a nonbinding probe (NBP). Specific complexes can be supershifted with an anti-HT antibody. ER␣. Although both isoforms were able to activate transcription through a ZBE, transactivation by p52-ZER6 was repressed in the presence of ER␣. Hence, in hormone-responsive cells, regulation of transcription by ZER6 can be controlled by the relative level of expression of the two isoforms.
The Cys 2 His 2 zinc finger is one of the most common motifs found in eukaryotic DNA-binding proteins (24). Zinc finger proteins containing four or more finger domains are usually able to form stable, sequence-specific DNA interactions (25,26). ZER6 contains six complete Cys 2 His 2 zinc finger domains, and it would be expected that the protein would be able to bind DNA with a core motif of 15-18 nucleotides (27). The amino acids at positions Ϫ1, 2, 3, and 6 of the ␣-helix in each finger are most critical in determining the sequence specificity of the DNA binding motif (25). The residues at Ϫ1, 3, and 6 are positioned to make contact with the primary DNA strand, whereas the residue at position 2 contacts the complimentary strand. Neighboring finger domains coordinately interact with the DNA sequence with each finger making contact with 3 nucleotides of the primary DNA strand. This arrangement offers the possibility of designing zinc finger proteins with defined DNA sequence specificity (27)(28)(29). Although the data are not yet available to make precise predictions, some amino acid residues have been defined to have preferences for specific nucleotides. In the case of ZER6, the ␣ helix of fingers F3 through F5 will create a binding pocket with the residues RREHNR at positions 6, Ϫ1, 2, 3, 6, and Ϫ1. This arrangement of residues at these key locations is very similar to the ␣ helix structure created by fingers 1-3 of Zif268 with the sequence RRDHTR (25). The glutamic acid residue at position 2 in ZER6 is very similar to aspartic acid residue in Zif268. Based on the similarity in this region, one might expect this region of ZER6 to have DNA specificity similar to that of the analogous region of Zif268. This region of Zif268 binds the core motif 5Ј-GTGGG, and this sequence motif is indeed found in the ZBE identified by PCR-assisted binding site selection. Although the ␣ helix structure of the other ZER6 fingers suggests a GC-rich binding site, data to allow a prediction of DNA interactions for the other fingers of ZER6 are incomplete. Combining the structural information of this novel zinc finger protein with the DNA binding specificity will provide additional information to help de-velop a robust model of zinc finger-DNA interactions.
A search for the ZER6 binding element was performed in the eukaryotic promoter data base (www.epd.isb-sib.ch/), which contains promoter sequence information on 1390 eukaryotic promoters from Ϫ499 to ϩ100 relative to the gene cap site. The promoter region of three genes (serum albumin, histocompatibility antigen-␥, and an early embryonic gene called H19) contained an exact match for the sequence 5Ј-GGGTGGGGGT-GGG. Because there is likely to be some degeneracy in sequence specificity, a search was performed with less stringency with a particular interest in genes known to be expressed in the breast or kidney. The insulin-like growth factor II (IGF-II) gene from human and rat contains multiple copies of sequences closely related to the ZBE, with the closest match being 5Ј-GGGTGGGGGTAGGG-3Ј. This finding is interesting because tamoxifen has been shown to stimulate expression from the IGF-II promoter (30). If the IGF-II promoter was targeted by ZER-6, it is conceivable that the ER␣-ZER6 interaction might be disrupted by tamoxifen thereby releasing active p52-ZER6. The mouse renin-1 gene promoter was found to contain the sequence 5Ј-CTGGGGGTGGAG-3Ј. Renin is synthesized in the renal juxtaglomerular apparatus, and these data are consistent with the possibility that renin expression is regulated by ZER6. It is not known if the expression of these other potential ZER6 target genes is altered by estrogen.
A BLASTp analysis of the ZER6 proteins identified a domain in the amino-terminal region, which was partially contained in p52-ZER6, with homology to the KRAB domains (31). The KRAB domains are common features of many zinc finger proteins and are associated with transcriptional regulation. The region of p71-ZER6 that is not contained in the p52 isoform demonstrated a region of striking homology with HUB-1 (16). Over a region of 139 amino acids (from amino acid 33 to 171), there is 74% identity and 84% homology with the HUB-1 repressor domain. The HUB-1 gene was cloned based on the ability for the protein to bind to the U5RE of human T-cell lymphotrophic virus type I. HUB-1 is a zinc finger protein with five finger domains that was shown to repress long terminal repeat-mediated transactivation. We hypothesize that the HUB-1 domain is involved in orchestrating protein-protein interactions involved in generating transcriptional complexes. In the case of ZER6, this domain inhibited the ability of ZER6 to interact with ER␣, which allowed the p71 isoform to escape repression in hormone responsive cells. To date, HUB-1 and ZER6 are the only two proteins with this motif and as more data are accumulated, this domain may define a subset of KRAB-related zinc finger proteins.