J Biol Chem, Vol. 275, Issue 17, 12879-12888, April 28, 2000

The Zinc Finger Repressor, ZBP-89, Binds to the Silencer Element of the Human Vimentin Gene and Complexes with the Transcriptional Activator, Sp1^*

Elzbieta Wieczorek^§, Zhili Lin, E. Brent Perkins, David J. Law, Juanita L. Merchant^¶**, and Zendra E. Zehner

From the  Department of Biochemistry and Molecular Biophysics and the Massey Cancer Center, Medical College of Virginia Campus/Virginia Commonwealth University, Richmond, Virginia 23298 and the ^¶ Departments of Internal Medicine and Physiology and  Howard Hughes Medical Institute, University of Michigan, Ann Arbor, Michigan 48109

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Vimentin is a component of the eukaryotic cytoskeleton belonging to the family of intermediate filament proteins. It exhibits a complex pattern of tissue- and development-specific expression. It is also a marker of the metastatic potential of many tumor cells. Previously, the human vimentin promoter was shown to contain several regions for the binding of positive and negative acting regulatory factors. Until now, the silencer element, which shuts down vimentin synthesis in selected tissues during development, was not precisely localized; nor was its binding protein known. In vivo DMS footprinting by ligation-mediated PCR delineated the position of guanine residues important to vimentin expression. Transient transfection assays in HeLa cells of various vimentin 5'-end promoter sequences and mutants thereof precisely defined two regulatory elements, a negative element and an adjoining positive acting element. Band shift assays, UV cross-linking, and Southwestern blot analysis confirm that the silencer element specifically binds a protein. Several lines of evidence show that ZBP-89, a zinc finger, Kruppel-like repressor protein is vimentin's silencer element binding factor. Co-immunoprecipitation and DNA affinity chromatography prove that Sp1 heterodimerizes with ZBP-89 when bound to the silencer element to yield a DNA-protein complex whose mobility is indistinguishable from that displayed by HeLa nuclear extract in band shift assays.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The cytoskeleton, an important complex and dynamic cell component, is composed of three networks: the microtubules, the microfilaments, and the intermediate filaments (IFs).¹ IFs comprise a heterogeneous family of nearly 50 different proteins, which are expressed in a tissue- and differentiation-dependent manner (1). IFs include such proteins as keratins, vimentin, glial fibrillary acidic protein, desmin, peripherin, nestin, internexin, neurofilaments, and lamins (2, 3). The IF system is a highly regulated, dynamic structure (4). So far, efforts to clearly define the physiological functions of IFs have not resulted in any unequivocal answers. It is known, for example, that cells can grow, divide, and differentiate in culture without any known cytoplasmic IFs being present (5). However, there is an enormous amount of data concerning the origin of tumors, and IF markers have been useful in identifying tumor types (6-8). Although vimentin does not seem to be essential for growth, cell division, or development, it does contribute to the metastatic potential of melanoma and mammary tumors (9, 10). We hope that an understanding of the mechanism of vimentin gene regulation will ultimately contribute to controlling the invasiveness of some tumor cells.

Initially, vimentin is widely expressed in the embryo but becomes progressively restricted to fewer cell types during development (1). It is postulated that negative regulation is involved in the cellular extinction of vimentin during embryogenesis. To date, the following elements have been identified in the promoter region of the vimentin gene: a TATA-box, several positive regulatory elements including eight GC-boxes (11), a NF-B site (12), a PEA3 site (13), and two AP-1 binding sites (14) plus silencer elements (15, 16). Previously, we have shown that of the eight GC-boxes, GC-box 1 is indispensable for vimentin gene expression (17). Another interesting feature of the vimentin promoter is the presence of an upstream antisilencer element, which was recently identified in the human as well as the chicken gene (16, 18). The human antisilencer element can overcome the minimal activity of a distal silencer (DS) element as well as the repression exerted by the proximal silencer (PS) element described in this paper (16).

The current model of eukaryotic gene regulation suggests that many positive and negative acting factors control its expression. Over the years factors that enhance gene expression have been isolated, and their mechanisms have been studied in great detail. More recently, mechanisms that repress gene activity have come under scrutiny, and it is now apparent that there are several different ways to silence gene activity (19-22). One level of gene repression is achieved by the highly organized and condensed structure of chromatin itself (21, 23). In this case, the unraveling of chromatin by acetylation or methylation has been shown to affect transcription, and it has been proven that some repressors work by stimulating such processes as histone deacetylation (23, 24). An example of a second class of repressors is Dr1 (25). These proteins do not bind to any specific sequence in DNA but are proposed to act directly at the level of transcription initiation by blocking the formation of the preinitiation complex. The third class of repressors involves proteins that do recognize specific sequences in DNA (22). These proteins can be further divided into two subgroups depending on their proposed mechanism of action. One group recognizes a DNA binding site, which overlaps with that of a positive factor. Therefore, the negative factor works by competing with an activator for DNA binding. Other repressors bind DNA and inhibit gene expression directly by interacting with a specific component of the general transcription complex presumably by protein-protein interaction.

Since vimentin synthesis is practically eliminated in certain stages of tissue development, we were interested in identifying its major silencer element and binding protein. In this paper, we have used ligation-mediated PCR (LMPCR) to first determine which guanine residues are important to vimentin expression in vivo. Transient transfection analyses with wild type and mutants thereof confirmed the precise localization of the major PS element. Moreover, we have identified the repressor, which binds to this element. We also present evidence for its interaction with the transcriptional activator, Sp1. These results suggest a mechanism by which the vimentin gene is regulated.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

In Vivo DNA Footprinting by Ligation-mediated PCR-- HeLa, MCF-7, and MDA-MB-231 (MDA) cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, L-glutamine, and 1% penicillin/streptomycin (Life Technologies, Inc.). In vivo dimethyl sulfate (DMS)-treated genomic DNA was isolated from five dishes of each cell type using established protocols (26). As a control, genomic DNA isolated from untreated cells was exposed to DMS in vitro. All DNA samples were treated with piperidine prior to analysis by LMPCR.

To cover the entire promoter region (900 to +41) of the human vimentin gene, multiple oligonucleotides were synthesized for use as primers. Each region required a unique set of primers designated PR3, PR4, and PR5. Five such sets covered the entire promoter region (Table I). The position of annealing to the 3'-end of the sequencing primers, PR5, is shown. The oligonucleotides PR1 and PR2 were annealed to produce the unidirectional (staggered) linker (27). The sequence of the PR3 primer was chosen to anneal to a known sequence 3' of the region to be analyzed. Ultimately, LMPCR allows the clear visualization of any fragment containing a cleavage site due to DMS methylation of guanosine residues within approximately 200 bases of the 3'-end of PR3. Footprinting reactions were performed as described (26, 27). Briefly, in step 1 the DNA (2.5 µg) is denatured and PR3 (0.3 pmol) is annealed and then extended with Vent DNA polymerase (0.5 units, New England BioLabs Inc., catalog no. 245L). In step 2, the unidirectional, staggered linker fragment (5 µM), with one blunt end, is ligated to the blunt end of the DNA fragment generated in step 1. In step 3, the new DNA fragment is denatured and PR4 is annealed and extended. In step 4, this fragment is PCR-amplified in the presence of PR4 and PR2 (both at 10 pmol). The sequence of the amplified fragment is visualized on a 6% polyacrylamide sequencing gel by a final round of PCR in the presence of ³²P-end-labeled PR5 (2.3 pmol). Primers (PR3, PR4, and PR5) were designed to possess a high T_m (>59 °C) to reduce nonspecific binding and display increasing values during the various steps of LMPCR, since initial primers are still present in subsequent PCR steps when annealing to the newly added primer is preferred. The temperature of the annealing step in each reaction was chosen to equal the T_m of that featured primer. Replicate experiments using the same DNA and primers produced reproducible cleavage patterns.

Plasmids-- Various 5'-deletions of the human vimentin promoter were constructed by PCR and fused to the reporter gene CAT in the vector, p8CAT (28). The expression vector pEcoCAT containing 1416 bp of 5'-promoter sequence from the human vimentin gene was used as the template (11). The 5'-primers were designed to contain an EcoRI restriction site at the 5'-end and anneal to various 5'-promoter sequences beginning at positions 261, 319, 339, and 353 as noted in Fig. 2B. In each case, the 3' primer (5'-ACGGAATTCCGGATGAGCATTC-3') was complementary to the sequence upstream of the EcoRI site inside the CAT gene itself. All PCR products were digested with EcoRI and subsequently inserted into the EcoRI restriction site of p8CAT.

For constructs containing the various mutations between 339 and 261, different cloning strategies were used. Seven 5' PCR primers were synthesized, with lowercase letters designating the mutated bases as follows: mut1, 5'-TACGAATTCCTCTGGTTCAGTCCCAGGCGGACCCCCCCCTCACaatatGACCC-3'; mut2, 5'-TACGAATTCCTCTGGTTCAGTCCCAGGCGGACCaattaCTCAC-3'; mut3, 5'-TACGAATTCCTCTGGTTCAGTCCCAGatatACCCC-3'; mut4, 5'-TACGAATTCCTCTGGTTCAGTataAGGCGG-3'; mut5, 5'-TACGAATTCCTtTaaTTCAGTCCC-3'; mut6, 5'-TACGAATTCTTGGatTaaTGCCACC-3'; mut7, 5'-TACGAATTCTTGGCGTGGTGCCAaaaGACC-3'. Italics show the EcoRI site (plus 3 bases to increase restriction enzyme efficiency) used for cloning. The 3' CAT primer mentioned above was used. The expression vector pEcoCAT was used as the template for PCR, and an expression vector, pH15, a derivative of pEcoCAT containing 251 bp of the vimentin promoter fused to the CAT gene in pUC19CAT, was used as the cloning vector. A unique XhoI site present at position 150 of the pH15 construct was utilized in generating all mutant constructs. For the mut1 construct, the PCR product was digested with XhoI, and the fragment containing the sequence from 319 to 150, including the mutated bases, was gel-purified and used for subsequent cloning. PH15 was digested with HindIII, filled in using the Klenow fragment to generate a blunt end, and subsequently digested with XhoI. The PCR product from above was inserted into the blunt XhoI site of the pH15 vector. For mut2 to mut7 constructs, the PCR products were first cloned into the pGEM-T vector (Promega) and then digested with XhoI and EcoRI. The resulting fragments containing the desired 5'-promoter sequences were used for subsequent cloning. In this case, the pH15 vector was digested with XhoI and HindIII. The entire multicloning site of pUC18 was removed by EcoRI and HindIII digestion, ligated to the EcoRI end of the PCR product, and subsequently inserted into the HindIII-XhoI site of the aforementioned pH15 vector.

For constructs that contain double mutations, two 5' primers were synthesized. For the double mutant construct of the human proximal silencer element (DM-PS), the 5'-primer is 5'-AGCAAGCTTCCTCTGGTTCAGTCCCAGatatACCaattaCTCACC-3', and the mut2 construct was used as the template for PCR. For the double mutant construct of the 19 (DM-19) element, the 5'-primer was 5'-AGCAAGCTTGGatTaaTGCCAaaaGACCCC-3', and the mut6 construct was used as the template. Italics show the HindIII site (plus three bases to increase restriction enzyme efficiency) used for cloning. The 3' primer was the CAT primer described above. These PCR products were digested with HindIII and XhoI, and the desired fragment was inserted into the HindIII-XhoI digested pH15 vector. The DNA sequence of all constructs was confirmed by DNA sequencing.

Cell Culture, DNA Transfections, and CAT AssaysHeLa cells were maintained as described previously and plated in 100-mm tissue culture dishes at 30-50% confluency, 30 min prior to transfection (18, 29-31). Chimeric plasmids were transfected using the calcium phosphate precipitation method with slight modification (32). Standardization among different transfections was achieved by co-transfecting 1 µg of pCMV--galactosidase (pCMV--galactosidase) to serve as an internal control plus pUC18 as a carrier to attain a total of 20 µg of DNA per transfection (33). CAT assays were performed after standardization to a fixed -galactosidase activity and analyzed as described (34). Results were quantitated by the use of a PhosphorImager (Molecular Dynamics, Inc.).

Nuclear Extracts-- Crude nuclear extracts (NE) were prepared from HeLa, MDA-MB-231, and MCF-7 cells as described (35, 36). The protein concentration was determined by the Bradford method (37), and aliquots in Dignam buffer D were stored at 70 °C.

Recombinant His₆-ZBP-89-- The cDNA sequence encoding ZBP-89 was excised from the fusion vector GST-ZBP-89 by digestion with SalI and BamHI and subcloned into like-digested pQE32 vector (Qiagene) (38). ZBP-89 was subsequently expressed as a hexahistidine-tagged protein (His₆-ZBP-89) and purified on a nitrilotriacetic acid-Ni²⁺ column per the manufacturer's directions (Qiagene).

Electrophoretic Mobility Shift Assays (EMSA)-- Samples of crude HeLa NE (10 µg), affinity-purified, recombinant His₆-ZBP-89 (10-200 ng), and/or purified Sp1 (1 footprinting unit; Promega) were preincubated in reaction buffer (20 mM HEPES, pH 7.6, 5 mM MgCl₂, 1 mM dithiothreitol, and 10% glycerol) with competitor DNA (1 µg of sonicated salmon sperm DNA and 0.5 µg of poly(dI-dC) for NE versus 0.25 µg of salmon sperm DNA and 0.25 µg of poly(dI-dC) for purified proteins) for 15 min on ice. Thirty fmol of a ³²P-labeled DNA fragment (radiospecific activity of 1 × 10¹² cpm/µmol) was added. The samples were incubated for 10 min at room temperature and applied to a 5% nondenaturing polyacrylamide gel in 0.5× TBE and electrophoresed at 4 °C at 170 V, dried, and subjected to autoradiography. For competition assays, DNA fragments at a 50- or 250-fold excess were added to the preincubation mixture, prior to the addition of ³²P-labeled DNA. For supershift experiments, the following antibodies were used: 1) anti-ZBP-89, a purified rabbit polyclonal antibody against the N-terminal part (amino acids 1-521) of bacterially expressed rat ZBP (38); 2) anti-KAP-1, a purified mouse polyclonal antibody against hexahistidine-tagged KAP-1 (amino acid 423-584) (39); 3) anti-Ku-86, a goat polyclonal IgG to the carboxyl terminus of the 86-kDa subunit of the mouse Ku protein (Santa Cruz Biotechnology, Inc., Santa Cruz, CA); and 4) anti-Sp1, a goat polyclonal IgG supplied as TransCruz Gel Supershift reagent (Santa Cruz Biotechnology).

UV Cross-linking-- Protein(s) were allowed to bind to DNA in the presence of nonspecific competitor, salmon sperm DNA and poly(dI-dC) and then subjected to gel electrophoresis as discussed above for EMSA. Afterward, the wet gel was exposed to UV light (254 nm) for 60 min, to cross-link the DNA to the protein(s), and then the bands were revealed by autoradiography. The gel pieces containing the complex(es) of interest were cut out, diced, and placed into Eppendorf tubes. Elution buffer was added (100 mM Tris (pH 6.8), 200 mM dithiothreitol, and 4% SDS), and the tubes were left on a rotator for incubation overnight at 37 °C. The buffer was transferred to a new tube, and the eluted macromolecules (protein(s) with covalently attached DNA) were precipitated with 2 volumes of acetone and centrifuged. The supernatant was removed and the pellet air-dried. The DNA component of the complex was not digested with DNase I, since digestion would remove the ³²P-end-label, thus making it impossible to subsequently identify the bound protein. The precipitated macromolecules were solubilized in sample buffer and resolved on an SDS-10% polyacrylamide gel. The gel was dried and autoradiographed.

Southwestern Blot Analysis-- Crude nuclear extracts (100 µg) from HeLa, MDA-MB-231, and MCF-7 cells were separated on a SDS-10% polyacrylamide gel and transferred to a nitrocellulose membrane. Membrane-bound proteins were allowed to renature in TNED buffer (10 mM Tris (pH 7.5), 50 mM NaCl, 0.1 mM EDTA (pH 7.5), and 1 mM dithiothreitol) containing 5% milk overnight at room temperature. Afterward, the nitrocellulose was rinsed three times with TNED buffer and placed in 3 ml of the same buffer containing 20 pmol (2 × 10⁷ cpm) of a ³²P-labeled DNA fragment containing the PS-binding site. Hybridization was carried out overnight at room temperature. The membrane was washed three times (10 min each) with TNED buffer, dried, and subjected to autoradiography.

Immunoprecipitation-- Protein G-Sepharose was equilibrated with phosphate-buffered saline (PBS) and divided into three 30-µl portions. Three µg of anti-Sp1 (PEP2X, Santa Cruz) in 300 µl of PBS was added to two portions, whereas the third sample (control) was supplemented with 300 µl of PBS alone. The incubation was carried out overnight at 4 °C, and then unbound antibody was removed, and all three resins were washed once with 200 µl of PBS. Next, 100 µl of HeLa NE (740 µg of protein) and 100 µl of PBS were added to each batch of resin. Additionally, one of the resin samples, which contained anti-Sp1, was supplemented with PS DNA (final concentration 1 µM). After incubation for 3 h at 4 °C on a rotary shaker, each resin was washed three times with 400 µl of PBS. Sample buffer was added, and following boiling, resin-bound proteins were separated on a SDS-10% polyacrylamide gel. Western blotting analysis was used to confirm the presence of ZBP bound to the resin.

DNA Affinity Chromatography-- Affinity resins were prepared as described (65). Oligonucleotides bearing wild type and mutated PS DNA sequences were covalently bound to CNBr-activated Sepharose 4B. DNA-bound resin (two 50-µl portions, containing approximately 2 µg of DNA) was equilibrated with no salt buffer (20 mM HEPES (pH 7.6), 5% glycerol, 5 mM MgCl₂, 1 mM dithiothreitol). At the same time, two 100-µl portions of HeLa NE (700 µg protein) were diluted 10 times with no salt buffer, supplemented with 100 µg of salmon sperm DNA, and incubated on ice for 20 min. Then the extract was incubated with the resins for 4 h at 4 °C on a rotary shaker. Afterward, each resin was washed twice with 2.5 ml of no salt buffer containing 0.1% Nonidet P-40 and then once with no salt buffer without detergent. Sample buffer (50 µl) was added to each resin, and, after boiling, the proteins bound to the DNA-Sepharose were separated on an SDS-10% polyacrylamide gel. The presence of ZBP-89 and Sp1 was verified by Western blotting analysis.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

In Vivo DMS Footprinting Analyses-- In vivo DMS footprinting via LMPCR was used to determine which regions of the human vimentin promoter are important for gene expression. We chose to analyze the human vimentin gene because of the existence of established cell lines that do or do not express vimentin to varying degrees in tissue culture. MDA cells, a human metastatic breast cancer cell line, express a high level of vimentin mRNA (40). HeLa cells, originally established as a carcinoma from an epithelial cell, co-express cytokeratins and a moderate amount of vimentin (41). Finally, MCF-7 cells, a nonmetastatic breast cancer cell line, do not express vimentin mRNA as judged by Northern blot analysis (31, 40). Previous DNA transfection analyses indicated that multiple regulatory elements exist between the DS element at position 780 and GC-box 1 at position 64 (see diagram Fig. 2A) (12-14, 16, 17, 42).

Throughout most of vimentin's promoter (900 to +41), no differences could be detected in the G-specific ladders of DNA isolated from MDA, MCF-7, and HeLa cells relative to the control DNA, methylated in vitro in the absence of protein. However, two regions displayed obvious DMS-hypersensitive sites. One area centered on GC-box 1 (data not shown). The importance of this element to gene activity is reported elsewhere (17). The second area, reported here, shows four hypersensitive sites at positions 323, 313, 298, and 325 (Fig. 1, A and B). Protein binding, as detected by hypersensitivity to DMS, is specific to cell types expressing vimentin, since these sites were observed only in the MDA and HeLa cell lines but not in MCF-7 cells, which do not synthesize vimentin.

View larger version (67K): [in this window] [in a new window]   Fig. 1.   LMPCR footprinting of the promoter region (900 to +41) of the human vimentin gene. MCF-7, MDA, and HeLa cells were subjected to in vivo methylation by a 0.9% DMS solution for 2 min as described under "Experimental Procedures." Deproteinated genomic DNA was methylated in vitro as a DNA control (no protein). Both in vivo and in vitro methylated DNA were treated with piperidine to introduce strand cleavage at methylated guanosine residues. The cleaved strands were amplified and labeled by LMPCR using the primer sets identified in Table I. DNA fragments were separated on a 6% polyacrylamide sequencing gel. Guanosine residues (shown by arrows) hypersensitive to methylation are noted at position 298, 313, and 323 in A and position 325 in B.

Where Is the "True" Binding Site for the Repressor of Vimentin Gene Expression?-- The presence of a negative element (19) at position 339 to 319 had been reported earlier (15). However, these transient transfection assays were performed only with fragments of vimentin's 5'-end fused to a heterologous promoter. By in vivo DMS footprinting, we detect two additional hypersensitive sites located 3' to this region, which suggests that protein(s) bind downstream of 19. Therefore, we decided to verify the transcriptional activity of this region in a homologous system.

Vimentin's various regulatory elements within the 5'-end are shown in Fig. 2A. Relevant promoter fragments, which begin at +73 and extend to 261, 319, 339, and 353, were fused to the reporter gene CAT and referred to as p261CAT through p353CAT (Fig. 2B). The results of transient transfection assays in HeLa cells clearly show a strong repressor element (a 93% decrease) in the region between positions 319 and 261. We named this site the PS element, not to be confused with a DS element of minimal activity reported earlier (16). A previous computer analysis of this region did not reveal any homology to known regulatory elements (43). Surprisingly, transient transfection assays also indicated the presence of a positive acting element located between 339 and 319, close to the PS site. This region was described previously as a negative acting element and referred to as 19 (15). The inclusion of more upstream DNA (greater than 339) had little effect on reporter gene activity until reaching the tandem AP-1 sites at approximately 700 (14, 44).

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Transient transfection analysis of various human 5'-deletions in HeLa cells. A, a diagram of the 5'-end of the human vimentin gene. The following regulatory elements are noted: DS (16), antisilencer (ASE) (16), AP-1 (14), 19 (15), PS, NF-B site (12, 64), PEA3 site (13), GC-box 1 (17), and TATA-box. B, diagram of the various human vimentin promoter fragments fused to the reporter gene CAT. The length of 5'-promoter sequence that it contains defines each construct. HeLa cells were transiently transfected with the various constructs (10 µg) as described under "Experimental Procedures." CAT activity was normalized to -galactosidase as an internal standard to correct for differences in transfection efficiency. CAT activity is expressed as the percentage of acetylated 14C-chloramphenicol in relation to total 14C-chloramphenicol. Results are an average from 4-6 separate transfections with three different plasmid preparations. Error bars represent the S.E.

To ultimately confirm that the PS is a true repressor and 19 is a true activator, mutagenesis studies were conducted. Because the guanine hypersensitive sites were generally located in GC-rich regions, the exchange of AT for GC bases should either restore or destroy gene activity depending on the activity of the region under study (Fig. 3). Mutations 1-5 (mut1-5) and the double mutation DM-PS (mut2+3) were designed to target the PS and adjoining 5'-region, whereas mutations 6 and 7 plus the double mutation DM-19 (mut6 + mut7) changed selected bases within the 19 element. It should be noted that the 5'-end of all PS fragments containing mutations (mut1-5, DM-PS) began at position 318, while the fragment containing upstream mutations (mut6-7, DM-19) began at 339. This means that the 5'-end of each group of mutations is constant, thereby eliminating any additional effect on gene activity due to the artificial joining of different 5'-end sequences to the vector. Transient transfection assays showed that mut2 and mut3 recovered 42 and 23% of gene activity, respectively, and when together in DM-PS, 72% of reporter gene activity was restored. Changes in the base sequence of mut1 and mut5 had little effect on gene activity, whereas mut4 resulted in only a 12% increase, which may not represent a significant change. On the other hand, mutations within the positive element (mut6, mut7, and DM-19) resulted in a loss of activation of 30, 49, and 100%, respectively. Altogether, our data prove that a strong repressor binding site is located at 300 to 289. Moreover, the DNA fragment between 339 and 319 not only lacks a repressor site but is in fact a positive acting element.

View larger version (17K): [in this window] [in a new window]   Fig. 3.   Site-directed mutagenesis of the PS and 19 elements of the human vimentin promoter. Top, nucleotide sequence of the vimentin promoter from position 339 to 261, indicating the position and sequence of the various mutations within the PS (mut1-mut5) and 19 (mut6-mut7) regions. All PS mutants contain the DNA sequence up to position 318 or 339 for 19 mutants. DM-19 represents the double mutant containing both mut6 and mut7 sequences, whereas DM-PS contains both mut2 and mut3 sequences. DNA bases that are hypersensitive to methylation in vivo (from Fig. 1) are in boldface type, and the DNA fragment used in EMSAs is underlined. Bottom, transient transfection analysis of various mutant constructs of the PS (right panel) and 19 (left panel) elements compared with the wild type constructs, p261, p319, and p339. CAT activity is measured as described in the legend to Fig. 2. Results are an average from 4-6 separate transfections of at least two different plasmid preparations, and error bars represent the S.E.

Does the PS Element Specifically Bind Protein(s)?-- Since in vivo DMS footprinting experiments suggested binding of protein(s) to the PS element, we performed EMSA using HeLa NE and a ³²P-labeled DNA fragment containing the PS sequence from 278 to 306 (see underlined sequence in Fig. 3). Indeed, NE contains protein(s) that specifically bind to the PS element, since only an excess of wild-type, unlabeled PS DNA competes with ³²P-labeled PS DNA for protein binding (Fig. 4). DM-PS is a weak competitor, which correlates with its minimal ability to repress transcription in transient transfection assays (Fig. 3). No competition is noted with an excess of the multicloning site fragment from pUC18.

View larger version (63K): [in this window] [in a new window]   Fig. 4.   EMSAs with vimentin's PS element and HeLa NE. EMSA was performed with 32P-labeled PS fragment 308 to 271 (underlined in Fig. 3) as described under "Experimental Procedures." The -fold molar excess of unlabeled competitors, i.e. PS, DM-PS, or the multicloning site fragment of pUC18, is indicated. Lane 7 contains 32P-labeled PS DNA alone. All reactions were analyzed on the same gel and exposed to film for the same length of time.

PS DNA binds to two closely migrating proteins with apparent molecular masses between 105 and 115 kDa in all three cell lines analyzed by Southwestern blot analysis (Fig. 5). It is interesting to note that the relative amount of these proteins is different in the various cell lines, with the heavier (115-kDa) protein prevailing in both HeLa and MDA cells. On the other hand, the ratio between the 105- and 115-kDa forms is reversed in MCF-7 cells. UV cross-linking of PS DNA with HeLa NE proteins reflects the mobility of the entire DNA-protein complex at approximately 145 kDa in SDS-PAGE (Fig. 5). Since this complex contains the ³²P-end-labeled PS DNA (49 bp), it migrates more slowly than the protein itself. Subtracting the contribution of the PS DNA fragment would result in a molecular weight that approximates that estimated by Southwestern blot analysis.

View larger version (23K): [in this window] [in a new window]   Fig. 5.   UV cross-linking of the PS element to HeLa NE protein(s) and Southwestern blot analysis of PS-binding proteins from HeLa, MDA, and MCF-7. Details of these assays are discussed under "Experimental Procedures." The numbers on both sides indicate the position of migration of various protein markers in kDa. The position of migration of unbound 32P-labeled PS DNA is noted.

Does a Known Protein Bind to the PS Element?-- Before embarking on purification of the silencer protein, we searched the literature for known repressors whose description matches the molecular weight and sequence specificity of the PS element. CTCF/NeP1 is an interesting zinc finger repressor that is able to recognize very divergent GC-rich sequences (45). We were able to eliminate CTCF/NeP1, because vimentin's PS element-binding protein did not compete with the CTCF/NeP1 binding site in EMSAs (data not shown). Similarly, we could also exclude the entire family of repressor proteins acting through the common co-repressor KAP-1 (39), since we did not observe any supershift in EMSA with anti-KAP-1 (Fig. 6A). Likewise, protein binding to the PS DNA was not due to the interaction of the Ku antigen, which has been shown to bind to the ends of DNA fragments (Fig. 6A) (46). However, a faint "supershifted" complex could be seen with anti-ZBP-89, a known Kruppel-like, zinc finger protein (Fig. 6A) (38).

View larger version (33K): [in this window] [in a new window]   Fig. 6.   EMSAs with various antibodies. A, the following antibodies were added to the EMSA reaction mixture together with NE: anti-ZBP raised against the first 521 amino acids of rat ZBP-89 (38), anti-KAP-1 raised to the co-repressor KAP (39), and anti-Ku raised to the small subunit of the Ku autoantigen. Antibodies were added to the reaction mixture prior to the addition of 32P-labeled PS DNA. The arrow indicates the position of the supershifted complex. B, Western blot analysis of His6-ZBP-89 and HeLa NE. HeLa NE (37 µg) and His6-ZBP-89 (0.5 µg) were separated on a 10% SDS-PAGE, transferred to nitrocellulose, and probed with anti-ZBP-89. Protein molecular mass markers are indicated in kDa. All reactions were run on the same gel and exposed to film for the same length of time.

Is ZBP-89 Vimentin's PS Element-binding Protein?-- We expressed ZBP-89 as a His₆-tagged protein (His₆-ZBP-89) using the Qiagen expression system. By Western blot analysis, we compared the mobility of recombinant His₆-ZBP-89 to the corresponding protein in HeLa NE (Fig. 6B). From both sources, a protein of identical mobility in SDS-PAGE was detected with anti-ZBP-89. Because ZBP-89 is a zinc finger protein, its binding to DNA can be abolished by the addition of the zinc chelator, o-phenanthroline (47). We also observed this result for vimentin's repressor from HeLa NE (Fig. 7). The addition of a similar aliquot of ethanol (the solute for o-phenanthroline) had no such effect. Moreover, the addition of zinc restored the repressor's ability to bind DNA. Therefore, it is clear that the PS element is bound by a zinc finger protein. Further confirmation came from EMSAs with recombinant His₆-ZBP-89 (Fig. 8) and with GST-ZBP-89 (data not shown), both of which bind ³²P-labeled PS DNA. Recombinant ZBP-89 binds the PS element specifically, since excess wild-type PS DNA, but not the nonfunctional double mutant (DM-PS), is able to totally abolish binding to the radioactive DNA probe.

View larger version (47K): [in this window] [in a new window]   Fig. 7.   The effect of zinc on EMSAs with the PS element and HeLa NE. EMSA was performed as described under "Experimental Procedures." However, before the reaction components were added, 1 µl of o-phenanthroline (800 nM final concentration, lanes 2 and 4) or 1 µl of ethanol alone (lane 3), was placed at the bottom of the tube. The ethanol was allowed to evaporate at room temperature for 10-15 min. Reagents were added as described under "Experimental Procedures." ZnSO4 (50 nM final concentration) was added to lane 4. All reactions were run on the same gel and exposed to film for the same length of time.

View larger version (28K): [in this window] [in a new window]   Fig. 8.   EMSA with the PS element and His6-ZBP-89. EMSA was performed as described under "Experimental Procedures." Lanes 3 and 4 contain a 250-fold molar excess of the appropriate competitor, PS or DM-PS, as indicated. Lane 1 contains 32P-labeled PS DNA only. All reactions were run on the same gel and exposed to film for the same length of time.

ZBP-89 Heterodimerizes with Sp1-- We observed that the complex of DNA with His₆-ZBP-89 migrates in EMSAs faster than the complex from HeLa NE (Fig. 9A, compare lane 8 with lane 1). However, when Sp1 is added together with increasing amounts of His₆-ZBP-89, the association of these two proteins into a larger complex with reduced mobility is observed (Fig. 9A, lanes 3-7). Moreover, the mobility of this assembled complex matches the mobility of the complex present in HeLa NE. Surprisingly, purified transcription factor Sp1, another GC-box-binding protein, by itself does not bind the PS element (Fig. 9A, lane 2). This result suggests that Sp1 can associate with His₆-ZBP-89 bound to the PS element but has little or no affinity for the PS DNA itself. Moreover, the addition of anti-Sp1 (-Sp1) either before (Fig. 9B, lanes 2 and 3) or after (Fig. 9B, lane 4) the PS DNA yielded a supershift of the PS DNA protein complex in HeLa NE. This confirms that Sp1 is one of the components of the PS DNA-protein complex in HeLa NE.

View larger version (21K): [in this window] [in a new window]   Fig. 9.   EMSA with the PS element and recombinant His6-ZBP-89, purified Sp1, and -Sp1. A, EMSA with recombinant His6-ZBP-89 and purified Sp1. The details of the EMSA are discussed under "Experimental Procedures." Lane 1 contains 32P-labeled PS DNA incubated with HeLa NE. Lanes 2-7 contain 32P-labeled PS DNA and increasing amounts (from approximately 20 to 180 ng) of recombinant His6-ZBP-89 (lanes 3-7) with a constant amount (1 footprinting unit) of purified recombinant Sp1 (lanes 2-7). Lane 8, which contains 32P-labeled PS DNA and recombinant His6-ZBP-89, was imported from a duplicate gel. B, supershift of the PS element-binding protein from HeLa NE with anti-Sp1 (note arrow). The EMSA was performed as described under "Experimental Procedures." Antibody in the amount indicated above each lane was added to the reaction mixture either prior to (lanes 2 and 3) or following (lane 4) the addition of 32P-labeled PS DNA.

To confirm that Sp1 and ZBP-89 interact with each other, we performed co-immunoprecipitation experiments using -Sp1 antibodies bound to protein G-Sepharose. We hoped to "pull down," from the HeLa NE, ZBP-89 bound to Sp1 (Fig. 10A). However, only a negligible amount (if any) of ZBP-89 co-precipitated with Sp1 as detected by Western blotting analysis (Fig. 10A, lane 2). In fact, this amount was equivalent to that obtained by ZBP-89 nonspecifically binding to the resin alone (Fig. 10A, lane 1). It should be noted that the anti-Sp1 resin (data not shown) precipitated an immense amount of Sp1 itself. However, when PS DNA was added to the NE and -Sp1-Sepharose incubation mixture, a stronger ZBP-89 band could be observed by Western blotting (Fig. 10A, lane 3), suggesting that only ZBP bound to PS DNA is able to interact with Sp1. On the other hand, the addition of DM-PS to the reaction mixture did not result in increased binding of ZBP-89 above background levels (data not shown). In this experiment, we have precisely chosen low stringency conditions for resin washing in order to favor the formation of the specific quarternary complex, -Sp1-Sp1-ZBP-89-PS. However, these conditions resulted in a high background of protein binding to protein G-Sepharose alone (Fig. 10A, lane 1). Any effort to reduce this background by increasing wash stringency and/or changing salt conditions during antigen binding resulted in diminished binding of PS-ZBP-89 to Sp1. One explanation for this result is that during washing DNA dissociates from the protein component of the complex. Therefore, we decided to reverse our strategy and pull down from NE Sp1 molecules associated with ZBP-89 using DNA affinity chromatography (65). Indeed, we were able to co-purify Sp1 with ZBP-89 using PS-Sepharose (Fig. 10B, PS). At the same time, only residual binding of ZBP-89 to the DM-PS-Sepharose was observed (Fig. 10B, DM-PS). This observation agrees with our mutagenesis results, where DM-PS is a poor repressor of gene activity (Fig. 3). Concomitantly, DM-PS-Sepharose purification resulted in a reduced amount of Sp1 as detected by Western blotting (Fig. 10B, DM-PS).

View larger version (19K): [in this window] [in a new window]   Fig. 10.   Co-immunoprecipitation and affinity chromatography of Sp1 and ZBP-89. A, co-immunoprecipitation of Sp1 and ZBP-89. protein G-Sepharose (30 µl) was incubated with HeLa NE only (lane 1) or preincubated with rabbit anti-Sp1 (lane 2) and then HeLa NE (lane 3). Incubation was carried out for 3 h at 4 °C in the absence (lanes 1 and 2) or presence (lane 3) of wild type PS DNA fragment (1 mM final concentration). The resin was than washed with PBS and analyzed by 10% SDS-PAGE and Western blotting using rabbit anti-ZBP-89 antibodies. The details are described under "Experimental Procedures." The band, marked as Sp1 IgGH, is the heavy chain of anti-Sp1, which is detected by the secondary antibody (anti-rabbit IgG) used in Western blotting. B, DNA affinity chromatography with PS DNA and HeLa NE. Affinity resins, PS-Sepharose (PS) and DM-PS-Sepharose (DM-PS), were incubated with 700 µg of HeLa NE for 4 h at 4 °C. Following incubation, the resins were washed with buffer containing 0.1% Nonidet P-40 and then analyzed by SDS-PAGE and Western blotting using rabbit polyclonal anti-Sp1 (PEP2X; Santa Cruz Biotechnology) and rabbit polyclonal anti-ZBP-89 antibodies. Two different exposures are shown for anti-ZBP-89, with the longer exposure using anti-ZBP-89 alone on the left.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

All in vitro studies on the interaction between regulatory factors and their target genes are hampered by their questionable application to genes as positioned in chromatin. The success of newer in vivo footprinting methods using LMPCR has proven invaluable in assessing the relevance of in vitro generated results.

Interestingly, no DMS hypersensitivity sites were detected in vivo on any of vimentin's regulatory elements, be they positive or negative acting, in the MCF-7 cell line, where vimentin is not synthesized. This observation has led us to speculate that the vimentin gene is "off" in this cell line due to the contextual effect of chromatin. In support of this hypothesis, we and others have found various vimentin 5'-end CAT constructs to be expressed when transfected into MCF-7 cells (13, 16). Thus, despite the fact that some of vimentin's regulatory factors are present in MCF-7 cells, the endogenous gene is not expressed. Altogether these results suggest that chromatin may repress vimentin gene transcription in such cell types. This might be as expected for an epithelially derived cell line not known to express vimentin at an earlier stage of development. Probably, the gene remains buried in inactive areas of chromatin, inaccessible to binding by regulatory and transcription factors.

On the other hand, LMPCR studies on the human vimentin gene promoter have detected two regions of altered sensitivity to DMS cleavage in cell lines, which do express vimentin. Here, we have analyzed the upstream region. Our transfection analyses using both wild type and selected mutant sequences demonstrate that this region can be further divided into negative and positive acting subregions. However, that subregion originally reported to contain a negative element actually exerts a positive effect on gene expression in HeLa cells (15). The true negative element is located 3' to this region.

An interesting question is, "Why is there a difference between the previous report and our findings?" We can offer some possible explanations, which illustrate how presumably minor differences in experimental design could lead to conflicting conclusions. The previous study reported the location of a negative element within a 187-bp fragment fused to either a synthetic vimentin NF-B site or a heterologous desmin enhancer, inserted into HSV-tkCAT, and transfected into HeLa cells (15). Deletion of 19 bp within this region (construct NE19NFB-tkCAT) abolished negative activity and led to the conclusion that it was required for repression. These studies were conducted on considerably rearranged DNA fragments or deletions thereof, fused to heterologous promoters. In some of these experiments, DNA fragments contained muscle-specific elements (myoD1, myogenin) not pertinent to vimentin expression. Experiments were not carried out on homologous vimentin promoter constructs; nor were base-specific mutations constructed to confirm DNA deletion results. Moreover, these constructs lacked elements, e.g. vimentin's GC-box 1, which we have shown is required both for basal and regulated vimentin gene activity (17). Thus, substitution of a promoter not containing an equivalent GC-box binding the appropriate protein could lead to a different conclusion. As has been recently shown, changing the specific DNA context of a regulatory site may result in misunderstanding the role of a DNA-binding transcription factor in the regulation of gene expression (66). In our experiments, we have introduced specific nucleotide changes in relevant regulatory elements within the "native" context of vimentin's 5'-end fused to the CAT gene. Furthermore, we have avoided large DNA deletions, which could alter spatial requirements and generate artificial results. In conclusion, our study is quite different from the previous report, and we see a direct correlation between results obtained in vivo and in vitro.

The presence of additional hypersensitive sites 3' to the aforementioned region led to the delineation of the true silencer region. From the mutational studies, we were able to precisely choose the appropriate DNA sequence for further investigation. The final identification of vimentin's PS element-binding protein to be the Kruppel-like, zinc finger repressor protein ZBP-89 is substantiated by several facts. First, the core portion (in boldface capital letters) of vimentin's PS element (ggaCCcCCcCC) is identical to the consensus DNA binding site for ZBP-89 (gccCCtCCxCC) as found in the promoters of other ZBP-regulated genes. Second, both, His₆-ZBP-89 and protein(s) from NE are recognized by anti-ZBP-89 antibody, and their mobilities are identical as detected by SDS-PAGE and Western blot analysis. It is interesting that anti-ZBP-89 recognizes a closely migrating doublet in HeLa NE (Fig. 10B) and on Southwestern blots (Fig. 5B). Moreover, the relative amount of these two protein species is similar in cell lines which express vimentin (HeLa and MDA) with the slower migrating protein species being more abundant. Conversely, in a cell line that does not synthesize vimentin (MCF-7), the faster migrating protein prevails. Since both of these protein species are recognized by anti-ZBP-89, we assume that they represent two different, post-translationally modified forms of ZBP-89. It is tempting to speculate that one of these forms represents the transcriptionally active form and the other an inactive factor. Such a hypothesis would be in agreement with our results obtained by in vivo footprinting. Third, antibodies to ZBP-89 were capable of supershifting the PS DNA-protein complex as found in crude NE. Unrelated antibodies gave no supershift. Fourth, the PS element-binding protein requires zinc for binding to DNA, which is required for ZBP-89 activity. Fifth, both purified recombinant His₆-ZBP-89 and GST-ZBP-89 fusion proteins bind to PS DNA, whereas no binding occurs to the nonfunctional DM-PS DNA. All of these results are compatible with vimentin's PS element-binding protein being ZBP-89.

Until now, four genes have been shown to be regulated by the repressor ZBP-89: gastrin (38), ornithine decarboxylase (47), and two type I collagen genes (48). In all of these cases, ZBP-89 appears to compete with another protein, like Sp1, Sp2, Sp3, Sp4, or WT1, for binding to an overlapping site in DNA (47). For example, ZBP-89 is involved in regulation of the ornithine decarboxylase gene together with the transcriptional activator Sp1, and its DNA binding site overlaps that of Sp1. Therefore, we questioned if other proteins might interact with ZBP-89 and be involved in vimentin gene repression. There were several indications that this might be the case. For example, we noticed that His₆-ZBP-89 bound to PS DNA moves faster in EMSAs than the corresponding DNA complex with NE protein(s), suggesting that more than one protein might be present in the PS DNA-protein complex in vivo. As a first candidate, we chose to analyze the complex for the presence of the transcription activator Sp1. Our rationale was that GC-box 1 and its binding factor was already shown to be required for interaction between sequence-specific regulatory factor(s) and the basal transcription machinery (17). Similarly, Sp1 binds to a GC-box sequence, stimulates the expression of many different genes, and interacts with a variety of proteins, some of which are known components of the transcriptional machinery, i.e. YY1, TBP, NF-B, RB, E2F, STAT-1, TAF_II55, and dTAF_II110 (13, 49-58). Thus, it has been suggested that Sp1 is a mediator between sequence-specific and general transcription factors, as we have suggested for vimentin's GC-box 1-binding protein (17). Moreover, some of the above mentioned factors, i.e. NF-B and STAT-1, have already been implicated in vimentin gene regulation (12, 59). In addition, Sp1 is essential in early stages of embryonic development, which correlates with the time frame for vimentin synthesis (60).

Therefore, we decided to verify whether Sp1 could compete or interact with ZBP-89. Indeed, EMSAs showed that the addition of Sp1 to ZBP-89 and PS DNA resulted in the generation of a tertiary complex, containing DNA and both proteins. Several explanations suggested how such a complex could be assembled. The first possibility was that only ZBP-89 can bind to the PS element and that Sp1 is held in this complex via protein-protein interaction with ZBP-89. The second possibility was that, since the PS element is a GC-rich sequence, a binding site for Sp1 is located close to or overlapping ZBP's binding site as has been suggested for other ZBP-regulated genes (47). Therefore, Sp1 would be able to bind to the PS element along with, but independent from, ZBP-89. Finally, it was possible that after binding to DNA, the conformation of ZBP-89 is changed such that Sp1 could bind either by protein-protein interaction or by enhanced recognition to a portion of the PS element or both.

Several lines of evidence suggest that only ZBP-89 can bind to the PS element and that Sp1 is held in this complex via protein-protein interactions. The experiment with a fixed amount of Sp1 and an increasing amount of ZBP-89 (Fig. 9) suggests that by itself either Sp1 does not bind or binds very weakly to the PS element. Increasing 3-fold the amount of Sp1 did not result in binding to PS DNA (data not shown). In addition, a co-immunoprecipitation experiment (Fig. 10A) using anti-Sp1 or a co-purification experiment using PS as well as DM-PS elements immobilized on protein-G-Sepharose (Fig. 10B) proves that ZBP-89 and Sp1 interact with each other. The presence of PS DNA but not DM-PS (which contains a mutated ZBP-89 core element but unchanged sequences outside of this binding site) is indispensable for this interaction. In all of these experiments, we never observed competition between Sp1 and ZBP-89 binding to the PS DNA. A portion (in boldface type) of the PS element (GgaCCcCCCCC) exhibits substantial homology (8/11 match) to a GC-box consensus sequence ((A/G)(T/C)(T/C)CCGCCCC(A/C)), but it lacks the crucial G of the core sequence. Therefore, we postulate that the Sp1 transcription factor contributes to the regulation of vimentin expression via ZBP-89 bound to the PS element.

The interaction between Sp1 and ZBP-89 is an interesting observation because it documents the fact that proteins of opposite activity (repressor versus activator) can directly contact each other. Blocking Sp1 activation is the suggested mode of action for the von Hippel-Lindau repressor, a known tumor suppressor (61). In this case, von Hippel-Lindau repressor and Sp1 must directly interact in order to repress gene activity. The overexpression of ZBP-89 has been shown to affect cell proliferation by slowing the growth of a rat pituitary, adenoma cell line, GH₄ (62). Vimentin expression is similarly cell cycle-regulated (14, 44). Moreover, it is a marker for tumor progression and invasiveness, since the high expression of vimentin correlates with malignancy (7, 40). Thus, it is possible that the inactivation of vimentin's repressor protein could contribute to enhanced cell growth and a tumorigenic phenotype, which would make ZBP-89 a candidate for a tumor suppressor as has been postulated for the von Hippel-Lindau repressor (63).

It is tempting to speculate that ZBP-89's binding of Sp1 may abolish activation of the vimentin gene by prohibiting its positive interaction with GC-box 1. However, a number of Sp family members exist (Sp1, Sp2, Sp3), any one of which could be the required GC-box 1-binding protein. Our preliminary data suggest that Sp1 can recognize the sequence of GC-box 1 in vitro. Interestingly, vimentin expression depends strongly on GC-box 1 at position 64, although there are eight such GC-box elements present throughout the vimentin promoter. Of these, GC-box 2 is the closest at position 147. A possible explanation for the preference of the first GC-box is that the distance between GC-box 1 and the PS element equals 220 bp, the approximate length of DNA wrapped around a single nucleosome. Thus, these two regulatory elements, GC-box 1 and the PS element, could reside on the same face of a single nucleosome or across two adjacent nucleosomes. In this context, the correct spatial location of their respective binding proteins could exist and favor protein-protein interaction. Once we have unequivocally identified the required GC-box 1-binding protein in vivo, we can initiate more detailed studies into the mechanism by which ZBP-89 binding to the PS element represses vimentin gene expression. For now, it appears to be more complex than the straightforward competition models suggested for other ZBP-regulated genes.

	ACKNOWLEDGEMENT

We express our thanks to Dr. S. Snyder for helpful comments, Dr. S. Rittling for the gift of the human vimentin DNA, Drs. L. Burke and R. Renkawitz for the NeP1/CTCF expression plasmid and its DNA binding site, and Dr. F. J. Rauscher III for antibodies to the co-repressor, KAP.

	FOOTNOTES

^* This work was supported by NHLBI, National Institutes of Health (NIH), Grant HL-45422 (to Z. E. Z.) and by NIDDK, NIH, Grant DK-45729 (to J. L. M.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

^§ Present address: Biochemistry Group, Inst. of Organic Chemistry, Biochemistry, and Biotechnology, Wroclaw University of Technology, 50-370 Wroclaw, Poland.

^** Investigator of the Howard Hughes Medical Institute.

To whom correspondence should be addressed: Dept. of Biochemistry and Molecular Biophysics, P.O. Box 980614, Virginia Commonwealth University, Richmond, VA 23298. Tel.: 804-828-8753; Fax: 804-828-1473; E-mail: Zehner@hsc.vcu.edu.

	ABBREVIATIONS

The abbreviations used are: IF, intermediate filament; anti-ZBP, antibodies raised against the protein ZBP-89; CAT, chloramphenicol acetyltransferase; CTCF, CCCTC-binding factor; DM-PS, double mutant of proximal silencer; DM-19, double mutant of 19; DMS, dimethylsulfate; DS, distal silencer; EMSA, electrophoretic mobility shift assay; NeP1, negative protein 1; PCR, polymerase chain reaction; LMPCR, ligation-mediated PCR; NE, nuclear extract(s); PS, proximal silencer; PAGE, polyacrylamide gel electrophoresis; His₆-ZBP-89, hexahistidine-tagged ZBP-89 fusion protein; STAT, signal transducers and activators of transcription; PBS, phosphate-buffered saline; bp, base pair(s).

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES