Physical and Functional Interaction between Two Pluripotent Proteins, the Y-box DNA/RNA-binding Factor, YB-1, and the Multivalent Zinc Finger Factor, CTCF*

CTCF is a unique, highly conserved, and ubiquitously expressed 11 zinc finger (ZF) transcriptional factor with multiple DNA site specificities. It is able to bind to varying target sequences to perform different regulatory roles, including promoter activation or repression, creating hormone-responsive gene silencing elements, and functional block of enhancer-promoter interactions. Because different sets of ZFs are utilized to recognize different CTCF target DNA sites, each of the diverse DNA·CTCF complexes might engage different essential protein partners to define distinct functional readouts. To identify such proteins, we developed an affinity chromatography method based on matrix-immobilized purified recombinant CTCF. This approach resulted in isolation of several CTCF protein partners. One of these was identified as the multifunctional Y-box DNA/RNA-binding factor, YB-1, known to be involved in transcription, replication, and RNA processing. We examined CTCF/YB-1 interaction by reciprocal immunoprecipitation experiments with anti-CTCF and anti-YB-1 antibodies, and found that CTCF and YB-1 form complexes in vivo. We show that the bacterially expressed ZF domain of CTCF is fully sufficient to retain YB-1 in vitro. To assess possible functional significance of CTCF/YB-1 binding, we employed the very first identified by us, negatively regulated, target for CTCF (c-myc oncogene promoter) as a model in co-transfection assays with both CTCF and YB-1 expression vectors. Although expression of YB-1 alone had no effect, co-expression with CTCF resulted in a marked enhancement of CTCF-driven c-myc transcriptional repression. Thus our findings demonstrate, for the first time, the biological relevance of the CTCF/YB-1 interaction.

Our search for conserved transcription factors that could similarly regulate the highly divergent promoters of avian and mammalian c-myc oncogenes resulted in identification and cloning of the gene encoding CTCF, 1 which contains a DNAbinding domain comprising 11 ZFs (1-6) CTCF binds to remarkably different, ϳ50 bp long, DNA target sites in the promoters of chicken and human c-myc genes (5,6). Inspection of additional CTCF-binding sites characterized by us and others showed that the mere presence of the CCCTC motif is neither necessary nor sufficient for CTCF binding (4, 6 -12).
CTCF appears to act differently upon binding to functionally distinct regulatory elements, which include: 1) enhancer-blocking (insulator) elements in the two boundaries flanking the ␤-globin locus, in the T-cell receptor ␣/␥ locus site, and in a matrix attachment 5Ј-boundary of the chicken lysozyme gene (see Refs. 11,13, and 14 for details); 2) promoter-proximal sites in chicken, mouse, and human c-myc proto-oncogenes (5,6); 3) the S-2.4 lysozyme gene transcriptional silencer (8) and the 144 genomic element (15), both of which require CTCF for efficient negative regulation by thyroid nuclear receptors; and 4) the crucial regulatory site of the amyloid protein precursor gene promoter (9,16,17). Moreover, CTCF is recently found to be a parent of the origin-specific and methylation-sensitive functional component of the chromatin insulator(s) upstream of the H19 gene (12,18,19), thus suggesting a new role for CTCF in the control of genomic imprinting.
CTCF-dependent insulators do not act as general transcription repressors or activators, but, rather, block functional interaction between an enhancer and a promoter (11)(12)(13)(14). On the other hand, binding of CTCF to the chicken or human c-myc major promoter results in repression (4,6), while binding to the APP promoter activates transcription (9,17). This diverse variety of functions indicates that, following the interaction with different sequences, the unique CTCF⅐DNA complexes might acquire distinct conformations that behave as sequence-specific platforms for selective binding of other proteins, which, in turn, define specific functional role of CTCF in this setting. In support of this hypothesis, we previously demonstrated that interaction of CTCF with two different DNA sequences of identical length results in distinct patterns of protein protection against proteolytic attack (6).
Thus, identification of CTCF-interacting protein partners * This work was supported by a Royal Society research grant and by a research grant from The Research and Equipment Committee, University of Oxford. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence may be addressed: Section of Molecular Pathology, Laboratory of Immunopathology, NIAID, National Institutes of Health, Bldg. 7 appears to be necessary to elucidate biological consequences of CTCF binding to different targets. To search for proteins interacting with CTCF, we developed a reliable and simple affinity chromatography method for fractionation of nuclear extracts on a matrix coupled with purified CTCF protein expressed in a baculovirus system.
One of the proteins retained by, and then eluted from, this matrix was found to be another multifunctional factor, YB-1. YB-1 is a member of the family of Y-box binding factors with each member containing some degree of homology to the cold shock domain (20). YB-1 is believed to participate in regulation of such diverse functions as transcription, replication, and RNA turnover (for reviews see Refs. 20 -24). In addition, YB-1 family members appear to be involved in mediating effects of UV irradiation (25), drug and interleukin-2 treatments (26,27), and DNA damage (28). Moreover, YB-1 and YB-1-like proteins participate in DNA repair. They can also promote DNA unwinding and DNA and RNA condensation (20,21,29,30). Given these properties, we predict that interactions between CTCF and YB-1 are likely to play multiple roles in regulation of major cellular processes.
In our initial attempt to assess whether there is a functional consequence of CTCF/YB-1 heterodimerization, we employed a transcriptional activity reporter system utilizing the very first CTCF-targeted promoter identified by us in the chicken c-myc oncogene locus (2)(3)(4)31).
We show here that, although expression of YB-1 alone had no effect on c-myc reporter activity, co-expression of YB-1 with CTCF resulted in a marked enhancement of CTCF-dependent transcriptional repression. These findings provide the first example of what may be multiple functional roles for CTCF/YB-1 interactions.

EXPERIMENTAL PROCEDURES
CTCF Expression in the Baculovirus System-The baculovirus expression system was utilized to obtain pure CTCF protein. To generate baculovirus that would expresses recombinant His-tagged CTCF we engineered the pVLH-CTCF vector based on the pVLH 6 plasmid kindly provided by R. M. Marais (CRC Center for Cell and Molecular Biology, London). The pVLH-CTCF was co-transfected with the Baculogold baculovirus DNA into SF9 cells, and recombinants were selected, characterized, and amplified following the manufacturer's instructions (BacVector System Manual, Novagen). The CTCF protein (termed "baculoCTCF") was purified to 80 -90% purity from infected SF9 cells using Ni-affinity chromatography with a linear gradient of imidazole for elution and subsequent gel filtration on an S-200 column (see Fig. 1A).
BaculoCTCF Affinity Chromatography-BaculoCTCF protein (see Fig. 1A) served as a "molecular hook" to pull out CTCF-interacting proteins from cell extracts. For this purpose, the baculoCTCF was first covalently coupled to the Sepharose 4B matrix using N-hydroxysuccinimidyl chloroformate-activated Sepharose 4B (both reagents from Sigma) to attach the protein. The protein solution was dialyzed against 1 M NaCl/0.1 M Na 2 HPO 4 , pH 8.2, and mixed with the activated resin in a protein/matrix ratio of 2 mg/1 ml. The binding reaction was performed overnight at ϩ4°C. The suspension was filtered to remove non-incorporated material and then blocked with 1 M ethanolamine, pH 8.2. Next, cell extracts were obtained from the chicken erythroid precursor cells HD3 (32) in buffer containing 50 mM Tris/HEPES, pH 8.0, 1% Nonidet P-40, 2 mM EDTA, 0.5 M NaCl, and 1 mM phenylmethylsulfonyl fluoride and incubated at 4°C for 30 min. An equal amount of the same buffer without NaCl or Nonidet P-40 was added, and the lysates were clarified by centrifugation at 15,000 ϫ g for 10 min. The cleared extracts in the 0.25 M NaCl-containing buffer were passed through the bacu-loCTCF affinity column. Weakly bound proteins were washed off with the low-salt buffer while strongly interacting proteins were subsequently eluted with the same buffer containing 1 M NaCl.
Identification of the Proteins Obtained by CTCF Affinity Chromatography-The eluted baculoCTCF-interacting partners were resolved by SDS-PAGE and visualized by Coomassie Blue staining. Well resolved protein bands were excised from gels and subjected to the "in-gel" digestion according to the method previously described by Shevchenko and colleagues (33). The MALDI-MS system on the LaserMat 2000 (ThermoBioAnalysis, Middlesex, UK) was utilized to map the mass of the tryptic mixture derived from each band (34). The mass values obtained by the analysis of the tryptic digests were used to inspect the SwissProt protein data base using the MS-Fit program available on the WEB from the UCSF Mass Spectrometry Facility. One of the proteins was tentatively identified as the Y box binding protein-1 (YB-1).
Immunoprecipitation-For immunoprecipitation, HeLa cells (ϳ2 ϫ 10 6 ) were collected, washed twice with the ice-cold phosphate-buffered saline, and lysed in 500 l of high salt RIPA buffer containing 0.5 M NaCl, 50 mM Tris-HEPES (pH 7.5), 1% Nonidet P-40, 2 mM EDTA, supplemented with 1 mM of phenylmethylsulfonyl fluoride, as described previously (35). After incubation for 30 min at 4°C, 500 l of RIPA buffer without NaCl was added to cell lysates, which were then cleared by centrifugation at 15,000 ϫ g for 10 min. CTCF and YB-1 were immunoprecipitated, respectively, by polyclonal affinity-purified rabbit antibodies to CTCF (Upstate Biotechnology, Lake Placid, NY) and YB-1 (kindly provided by H.-D. Royer and A. Lee (36)). The anti-CTCF antibodies were raised against the bacterially expressed human CTCF carboxyl-terminal domain and recognize CTCF in all vertebrates due to the exceptionally high evolutionarily conservation of CTCF protein sequence (5,6). Other antibodies for co-immunoprecipitation experiments were used against several nuclear proteins, including p21, ubiquitous nuclear receptor UR, thyroid receptor TR␣, hTAF II 130, and MYC (Santa Cruz Biotechnology, Santa Cruz, CA).
The immunoprecipitates obtained with the rabbit polyclonal antibodies were then incubated with 20 l of protein A-Sepharose-4B-Fast (Sigma), while the precipitates obtained with the mouse monoclonal antibodies were incubated with 20 l of protein G-Sepharose-CL4B (Sigma) for 1 h at 4°C. The Sepharose beads were then washed five times with 1 ml of RIPA buffer containing 0.25 M NaCl. Immobilized proteins were resolved by 10% SDS-PAGE, transferred onto Immobilon P membranes (Millipore), and probed by Western blot analysis procedure as described below.
Expression of the Amino-terminal, Zinc Finger, and Carboxyl-terminal Domains of CTCF in Bacterial System-We initially prepared the pCITE4b-hCTCF-C plasmid containing the 0.8-kbp NcoI to EaeI DNA fragment from the p7.1 human CTCF cDNA plasmid (6) cloned into the NcoI and NotI sites of pCITE-4b (Novagen, Madison, WI) downstream of the viral cap independent translational enhancer (CITE). This vector encodes the carboxyl-terminal domain of CTCF extending from the Met residue in the middle of the 11th zinc finger to the stop codon (amino acids 589 -727, see Fig. 4A). Next, we constructed a bacterial expression vector pET16b-CTCF-C for purification of the His-tagged CTCF-C protein. The NdeI-XhoI DNA fragment from the pCITE4b-hCTCF-C plasmid was re-ligated to the NdeI and XhoI ends of the pET16b (Novagen) resulting in pET/hCTCF-C. This vector contains, under an inducible promoter, a coding region for the in-frame fusion between the His-Tag polypeptide, Met-Gly-(His 10 -tag)-Ser-Ser-Gly-His-Ile-Glu-Arg-His, the S-tag polypeptide derived from the pCITE4b, and the hCTCF-C region. To improve the coupling of the bacterially expressed proteins to the matrix, we modified the parental plasmid pET16b-CTCF-C to pET16b-CTCF-Cϩcys by adding a tag composed of four cysteine residues. To this end, the pET16b-CTCF-C plasmid was cut with AgeI and XhoI enzymes, and the backbone was then re-ligated with the double stranded annealed DNA fragment 5Ј-CCGG-TGT-TGC-TGT-TGC-TAG-CTA-AC-3Ј encoding the tag created by the four-cysteine amino acids (underlined) in-frame with the end of the CTCF protein, followed by the stop codon and flanked by the AgeI and XhoI (shown in bold italic) sites for cloning into pET16b-CTCF-C digested with AgeI plus XhoI. The resulting plasmid was named pET16b-CTCF-Cϩcys, and its vector backbone obtained after digestion with NdeI and AgeI enzymes was utilized to produce the pET16b-CTCF-Nϩcys and p16b-CTCF-Znϩcys vectors for expression of the CTCF amino-terminal and zinc finger domains, respectively (see Fig. 4A). The corresponding domains were polymerase chain reaction-amplified from the pCI7.1 vector (6) with primers containing a NdeI restriction site in-frame with the pET16b translation start codon and AgeI at the stop site. The primers were (NdeI and AgeI sites are underlined): 1) for the CTCF aminoterminal domain (amino acids 1-278) (see Fig. 4A), the forward primer (NdeI, nucleotides 279 -300, numbering according to ref. 6), 5Ј-AGAG-GCAGGGCATATGGAAGGT-3Ј, and the reverse primer (AgeI, nucleotides 1125-1105), 5Ј-ACGCCGTGGACACCGGTAACT-3Ј; or for the CTCF zinc finger domain (amino acids 264 -584) (see Fig. 4A), the forward primer (NdeI) (nucleotides 1081-1104), 5Ј-AAGACATTCCAG-CATATGCTTTGC-3Ј with the reverse primer (AgeI) (nucleotides 2043-2023), 5Ј-CTCTACGCCATACCGGTCAGC-3Ј. The polymerase chain reaction-produced DNA fragments, corresponding to the amino-and Zn-domains of CTCF, were then cut with NdeI and AgeI and ligated into the pET16b-CTCF-Cϩcys vector digested with NdeI and AgeI so that the four Cys residues remained in the backbone of the vector. The resulting plasmids, pCTCF-ZFϩcys and pCTCF-Nϩcys, were verified by sequencing and transformed into Escherichia coli strain BL21 (DE3) purchased from Invitrogen (Carlsbad, CA).
Expression in E. coli and Purification of the Amino-terminal, Zinc Finger, and Carboxyl-terminal Domain of CTCF -Transformants carrying the plasmids expressing the three CTCF distinct regions, were grown in LB media supplemented with ampicillin (50 g/ml) for 3 h at 37°C. Protein expression was induced by the addition of 0.4 mM isopropyl ␤-D-thiogalactoside with further incubation for 3 h at 37°C. For purification of each of the desired protein, the bacterial cells were pelleted, washed twice with 0.1 volume of cold phosphate-buffered saline followed by lysis in 0.1 volume of the original culture in the cold fresh lysis buffer containing 8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0. The lysates were subjected to immobilized metal ion affinity chromatography for further purification. Total bacterial lysates containing His-tag expressed proteins were 1) supplemented with 20 mM imidazole; 2) loaded onto the nickel-charged His-Bind resin (R&D Systems, Europe Ltd.); 3) washed with one bed volume of the washing buffer containing 8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0, and 20 mM imidazole; 4) finally, the His-tagged proteins were eluted by10 ml of the buffer with 8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-HCl, pH 8.0, and 0.5 M imidazole. Purity and identity of the isolated proteins were verified by 10% SDS-PAGE followed by Coomassie Blue staining and Western blot assays with anti-His-tag (H-3) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA).
Immobilization of the Bacterially Expressed Amino-terminal, Zinc Finger, and Carboxyl-terminal Domains of CTCF on the Matrix-Cystamine-coupled Sepharose 4B was utilized to bind the bacterially produced distinct CTCF domains shown below in Fig. 4A. Cystamine was first converted into aminoethylthiol by a reducing reaction with 50 mM dithiothreitol in TE buffer (50 mM Tris-EDTA, pH 8.3) for 30 min at room temperature, then treated with 5 mM 2,2-dipyridyldisulfide for 2 h. The activated matrix was washed with the TE buffer. Each protein was reduced by incubation with 5 mM dithiothreitol for 1 h at room temperature, passed through a G50 column equilibrated with TE to desalt the proteins, and then incubated with the activated matrix overnight at ϩ4°C at a protein to Sepharose v/w ratio of 5 mg/1 ml. The amounts of the proteins retained on the matrix were monitored by protein assay (Bio-Rad) according to the manufacturer's instructions. The protein-Sepharose conjugates were finally washed with the TE buffer to remove non-incorporated materials and stored in the buffer containing 20% glycerol, 50 mM KH 2 PO 4 , pH 7.0, and 0.2% Na 3 N.
Interaction Assay-On a rotating platform, we incubated ϳ50 l of the Sepharose suspension carrying each of different of CTCF protein domains (see Fig. 4A) mixed with 1 ml of HeLa lysate in 0.25 M RIPA buffer for at least 6 h. Each suspension was then washed six times with 0.25 M RIPA buffer, boiled in SDS sample-loading solution for 5 min, run on 10% SDS-PAGE, and Western blotted. The presence of the CTCF-interacting partner, YB-1, was detected with anti-YB1 antibodies.
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assay-Nuclear extracts were prepared according to procedures described in detail previously (37). A 32 P-labeled DNA probe (ϳ1 ng), the chicken c-myc fragment FPV harboring the binding site V for CTCF, was prepared by labeling with T4 polynucleotide kinase and [␥-32 P]ATP as described previously (5,38). The probe was incubated in 1ϫ EMSA binding buffer (the 5ϫ EMSA binding buffer stock contained 20% Ficoll, 30 mM NaCl, 100 mM HEPES, pH 7.85, and 1 mM MgCl 2 ) with purified baculoCTCF protein, nuclear extracts prepared from COS6 cells, or with no protein as a control, in the presence of nonspecific doublestranded competitors poly(dI-dC) (200 g) and poly(dG-dC) (40 g) in a total volume of 20 l. This EMSA reaction mixture was kept on ice for 30 min and loaded on the 10% native polyacrylamide gel. The samples were run in 0.33ϫ TBE at 250 V (ϳ25 mA) for 2 h. The gel was fixed in 10% glacial acetic acid with 20% methanol, dried, and exposed to the XAR-5 film (Kodak).
Western Blot Analysis-Proteins resolved on the 10% SDS-PAGE gels were transferred to Immobilon P polyvinylidene difluoride filters (Millipore) by electroblotting. Membranes were incubated with the primary antibody (at 1:200 dilution for the anti-CTCF and 1:5000 dilution for the anti-YB-1 antibodies) for 1 h at room temperature. After washing, the membranes were incubated with anti-rabbit-peroxidase-conjugated antibodies (1:10,000 dilution) for 1 h at room temperature and then processed following the manufacturer's instructions for the enhanced chemiluminescence detection (ECL kit, Amersham Pharmacia Biotech).
Cell Culture, Transient Transfections, and CAT Assay-On the day prior to transfection, COS6 cells were plated at a density of 10 5 cells/ well using12-well plates (Corning), and maintained overnight in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. On the day of transfection the cells were ϳ80% confluent. A standard calcium phosphate DNA transfection method was utilized. Measurements of chloramphenicol acetyl transferase (CAT) activity produced by the reporters in equal numbers of the transfected cells were carried out as described previously earlier (5,38).

RESULTS
We wished to determine whether CTCF is involved in protein-protein interactions and to identify different CTCF-binding factors. For this purpose, we developed an affinity chromatography approach that utilizes the matrix coupled with the CTCF protein produced in the baculovirus expression system. In addition to high specificity, another advantage of this approach is that the same CTCF matrix can be reused for isolation of CTCF partners from different cell extracts.
Expression and Purification of the Full-length CTCF Protein-The baculovirus expression system was employed to generate the CTCF protein in quantities sufficient for immobilization of CTCF on the matrix. Approximately 5 mg of ϳ90% pure full-length His-tagged baculoCTCF protein (schematically depicted in Fig. 4A) was obtained (Fig. 1). In contrast to bacterially produced full-length CTCF that forms aggregates difficult to solubilize into properly refolded individual CTCF molecules, 2 baculoCTCF apparently has the native conformation. In comparisons with the endogenous CTCF, the 6xHis-bacu-loCTCF has the same major biochemical activities (Fig. 1). First, baculoCTCF co-migrates as the endogenous CTCF in SDS-PAGE. Second, anti-CTCF antibodies recognize it equally well. Third, it has DNA-binding properties identical to that of the endogenous CTCF as demonstrated by EMSA (Fig. 1C). Thus, the isolated full-length His-tagged baculoCTCF protein appears to be in the native conformation. This property is of obvious importance for use as the molecular bait in a search for interacting partners.
Affinity Chromatography to Isolate Proteins Which Strongly Interact with the BaculoCTCF-BaculoCTCF covalently coupled to matrix was used to prepare a column as described under "Experimental Procedures," and nuclear extracts prepared from the chicken erythroid precursor cells HD3 were passed through the column. Proteins binding to CTCF with very low affinity or nonspecifically, were eluted by washing with lowsalts buffers. The proteins strongly interacting with CTCF were subsequently eluted with 1 M NaCl. The latter protein fraction was resolved on 10% SDS-PAGE mini gels. Silver staining revealed several major and multiple minor protein bands (Fig. 2). To identify the major proteins, larger amounts of high-salt-eluted proteins were separated by SDS-PAGE in larger gels, stained with the Coomassie Blue, and the visible bands were excised. These were subjected to the "in-gel " digestion and tryptic mixture mass mapping. The protein with a molecular mass of 36 kDa (Fig. 2, lane E2) appeared to be the Y box binding protein 1, or YB-1 (SwissProt protein data base accession number Q06066).
CTCF and YB-1 Proteins Co-immunoprecipitate from Cell Extracts-To determine if CTCF and YB-1 interact in vivo we carried out a series of co-immunoprecipitation assays with the extracts from HeLa cells. Proteins bound by anti-CTCF or anti-YB-1 antibodies were incubated with protein A-Sepharose-4B-Fast, washed, separated on SDS-PAGE, and analyzed by ECL immunoblotting. Fig. 3A (lane 8) shows that YB-1 immunoprecipitates obtained from HeLa (lane 8) or HD3 (lane 9) contain the CTCF band co-migrating with the CTCF protein from the whole HeLa cell lysate (lane 6). CTCF could not be detected in precipitates obtained with the pre-immune serum (lane 7), or with antibodies against several nuclear proteins, including p21, ubiquitous nuclear receptor UR, thyroid receptor TR␣, hTAF II 130, and MYC (lanes 1-5, respectively). CTCF was specifically co-immunoprecipitated with anti-YB-1 antibodies from extracts obtained from a large number of cell lines (data not shown), thus suggesting both ubiquitous expression and association of these two proteins.
In reciprocal experiments, we performed co-immunoprecipitation assays with anti-CTCF antibodies combined with immunoblotting for YB-1. As shown in Fig. 3B, the YB-1 protein band was detected in the CTCF-immunoprecipitated material, and the position of this band was identical to that of the YB-1 protein in the unfractionated HeLa cell lysate (Fig. 3B, lane 1). Taken together, these results indicate that at least a fraction of both YB-1 and CTCF form specific complexes in vivo.
YB-1 Interacts with CTCF through the CTCF Zinc Finger Domain-Although our findings demonstrate that CTCF complexes in vivo with YB-1, further investigation was required to determine whether this interaction is direct or mediated by other proteins. To this end, we produced the three consecutive distinct domains of CTCF, schematically shown in Fig. 4A, in a bacterial expression system. These three proteins, CTCF-Nϩcys, CTCF-ZFϩcys, and CTCF-Cϩcys, contained in-frame additions of a 10xHis-tag at the amino terminus and a 6xCys residue tag at the carboxyl terminus (Fig. 4A). All three proteins, equally well expressed in E. coli, were obtained at about 90% purity, as shown by the Coomassie Blue staining and immunoblotting with anti-His-tag antibodies (Fig. 4, B and C).
To identify the region(s) of CTCF necessary for interaction with YB-1, HeLa cell lysates were incubated with immobilized CTCF-Nϩcys, CTCF-ZFϩcys, and CTCF-Cϩcys proteins. As shown in Fig. 4D (lane 2), the zinc finger domain of CTCF retained a protein of about 45 kDa from HeLa lysates, recognized by specific antibodies for YB-1. YB-1 protein of the same size was also detected in the HeLa cell lysate (Fig. 4D, lane 5). The apparent molecular mass (45 kDa) of this protein is higher than previously described for YB-1 (Fig. 2 and Refs. 26, 29, and 36). This difference may be due to specific post-translational modifications of YB-1 in HeLa cells (30). No interaction of the HeLa 45-kDa YB-1 with the matrix-coupled CTCF-Nϩcys, CTCF-Cϩcys, or bovine serum albumin proteins was detected  1-5, respectively). B, the Western assay with the anti-YB-1 antibodies was performed after immunoprecipitation of the CTCF protein from HeLa lysates with the CTCF antibodies. The immunocomplexes were resolved by SDS-PAGE and blotted, and the membrane was probed with the anti-YB1 antibodies. The arrow indicates the position of YB-1 in the HeLa cell lysate (lane 1) and immunoprecipitates with the anti-CTCF antibodies (lane 2). (Fig. 4D, lanes 1, 3, and 4). When the same membrane was reprobed with the preimmune serum, only nonspecific bands outside of the 45-kDa region were noted (not shown) thus confirming that the 45-kDa band is the YB-1 of HeLa cells. Further experiments are in progress to identify more precisely the parts of the CTCF zinc finger domain interacting with YB-1, and also to delineate the smallest region(s) in YB-1 required to interact with CTCF. Preliminary results of in vitro pull-down assays with [ 35 S]methionine-labeled in vitro translated YB-1 and the CTCF-ZFϩcys polypeptide on beads showed retention of the near-full-length YB-1 (not shown). Taken together, our protein interaction experiments suggest that CTCF and YB-1 are likely to bind one another without requiring additional partner(s).
CTCF and YB-1 Cooperate in the Regulation of the c-myc Promoter-As outlined previously, both CTCF and YB-1 are involved in several, perhaps intertwined, but mechanistically different molecular processes related to regulation of gene expression. One model for evaluating functional role(s) for interaction of these proteins was designed to test if YB-1 affects activity of CTCF-dependent promoters. For this purpose, we performed transient co-transfection studies with vectors expressing full-length CTCF (pCTCF) and YB-1 (pYB-1, or pREP9) and the CAT-based reporter construct pK232-S1CAT containing ϳ1-kb fragment of the chicken c-myc 5Ј-noncoding region (Fig. 5A) bearing the FPV binding site for CTCF (1,3,4).
As shown in Fig. 5, and in line with our previous observations, ectopic expression of CTCF led to repression of the c-myc promoter. In contrast, ectopic expression of YB-1 had no effect on expression from this reporter. However, a marked synergism of CTCF and YB-1 was detected in repressing the c-myc promoter when both plasmids were overexpressed together, especially at the higher input of YB-1 (Fig. 5B). No CTCF-dependent repression, and no synergy with YB-1, was noted with a mutated PFV site not capable of CTCF binding (5) (not shown). These results indicate that a functional interaction between CTCF and YB-1 is important for regulation of the c-myc promoter. DISCUSSION The initial aim of this study was to develop a reliable system for identifying proteins that interact with CTCF. To identify such factors, we employed the full-length CTCF expressed in the baculovirus system (Fig. 1). Because the purified bacu-loCTCF possessed major features characteristic for the native protein ( Fig. 1), it was assumed to be the appropriate bait to utilize for making a reusable affinity matrix carrying the covalently bound CTCF. The simple procedure for isolating potential CTCF-interacting partners from cell extracts has been developed as described under "Results." Employing this procedure, several proteins, retained on the CTCF matrix, have been identified. One of them turned out to be another transcription multifunctional factor, the Y box binding protein 1, YB-1 (Fig. 3).
YB-1 is implicated in regulation of multiple cellular functions. It has unusual transcription factor activities such as site-specific DNA unwinding, interaction with RNA, and damaged DNA (30, 39 -45). A multitude of functionally different genes appears to be regulated by YB-1 (26, 40, 43, 46 -50). Among them are germ-line specific genes (20) and genes associated with cell growth control (24). Moreover, YB-1 is involved in multiple protein-protein interactions. For instance, it forms functional heterodimers with transcription factors Pur-␣, 65 P2 (51), YY-1 (52), PCNA (28), cardiac ankyrin repeat protein CARP (53), RelA (54), T-antigen (55), and Tat (42). Taken together these properties of YB-1 indicate diverse and multipurpose utilization of this factor in cells (for reviews see (20 -22, 45, 56 -59). We showed here that CTCF and YB-1 form specific complexes both in vivo and in vitro, and that interaction with YB-1 requires the zinc finger domain of CTCF (Figs. 3 and 4). Furthermore, CTCF/YB-1 complexes were observed in various cell types (Fig. 3A) possibly implying "universal" functions for the association of these two ubiquitously expressed factors. It is tempting to speculate that CTCF may, therefore, be involved in modulation of at least some of the multiple functions mediated by YB-1, including those that are not mediated through direct DNA recognition. It also seems reasonable to suggest in a reciprocal manner that interaction with YB-1 may equip CTCF to perform functions beyond transcriptional regulation mediated by binding to promoters (5,6,9,16,17), hormone-responsive silencers (8,15), or enhancer-blocking elements in the globin genes loci or the differentially methylated imprinting control region upstream of the H19 gene (11-14, 18, 19). Precedent for this comes from studies on the Wilms' tumor suppressor gene WT1 with a DNA-binding domain of four zinc fingers capable of mediating binding to at least two different sequences. WT1 is suggested to be "more then just a transcription factor" due to its apparent role in RNA processing and other functions not related to transcription (for review see Ref. 60).
To test whether interaction with YB-1 might modulate any of many CTCF functions, we measured the effects of ectopically expressed CTCF on regulation of promoters bearing CTCFbinding sites (4,6,9,17). The very first identified regulatory targets for CTCF were the vertebrate c-myc gene promoters (4 -6). We demonstrated here that YB-1 acts synergistically with CTCF to repress the c-myc promoter (Fig. 5). Expression of YB-1 alone had no effect on promoter activity (Fig. 5), suggesting that the putative target sequences for YB-1 within the c-myc promoter are not transcriptionally active. There are several possible reasons explaining this finding: 1) an optimal YB-1-binding consensus sequence CTGATTGG(C/T)(C/T)AA (61) is absent; 2) the only CCAAT motif found in our c-myc reporter constructs was positioned intronically downstream of the first noncoding exon within the region not previously identified as important for the promoter regulation (62)(63)(64); and 3) the CCAAT core motif may be masked from YB-1 by other proteins recognizing the same core sequence, including NF-Y, NF1/CBP, C/EBP, and the CAAT-binding displacement factor. Therefore, our results suggest that YB-1 does not have to bind the c-myc promoter to cooperate with CTCF, but that interaction with YB-1 is functionally important in regulation of CTCFdependent promoters. Whether YB-1/CTCF binding may have any effect on the CTCF-driven enhancer-blocking (insulator) activity (11-14, 18, 19) remains to be investigated.
Because both CTCF and YB-1 demonstrate multiple interactions with other proteins, an obviously important question for future studies is whether the association between CTCF and YB-1 is mutually exclusive with respect to other YB-1 and CTCF interacting proteins. The quantitative in vitro interaction assay or EMSA with CTCF, YB-1, and a third interacting protein would perhaps allow us to assess whether three or more proteins could form competitive heterodimers or cooperative heterotrimers.
Our pull-down assays revealed that the zinc finger domain of CTCF is utilized for interaction with YB-1 (Fig. 4). The concept of alternative involvement of DNA-binding zinc fingers in protein-protein interactions is increasingly appreciated (65). Besides DNA binding, the C 2 H 2 and C 2 HC classes of ZFs provide functionally important interactions with other proteins. For example, transcriptional co-factors of the friend of GATA (FOG) family bind to GATA proteins by means of multiple ZFs (66). Interestingly, the unusual single carboxyl-terminal C 2 HC-type ZF that is necessary and sufficient for binding of FOG-1 to GATA-1 is a structural homologue of the carboxylterminal 11th ZF of CTCF (not shown).
Because the multiple functions of CTCF are quite different, there must be specific molecular mechanisms that enable the same protein to perform distinct functions. Based on the results of DNA binding experiments with "missing finger" CTCF ZF domains, the multiple sequence specificity of CTCF was interpreted as reflecting the ability of CTCF to employ different groups of individual ZFs to recognize highly divergent sequence (6,8,9,12,15). One can imagine that the multiplicity of CTCF effects that derive from occupation of different DNA sites can be determined by the allosteric effects that result from assignment of particular group of ZFs for making contacts with DNA leaving the others for interaction with other proteins. Therefore, conformational changes in CTCF following interaction with different DNA sequences are likely to specify its interactions with cofactors that ultimately determine the biological readout(s).
A growing body of evidence implicates both CTCF and YB-1 in tumorigenesis. CTCF is located at the cancer-associated "hot spot" on 16q22 and is found to have a rearranged ZF domain in some breast cancer samples (67). It is also a negative regulator of the c-myc oncogene for which dysregulation without genomic alterations is frequently found in different cancers (68 -70). On FIG. 5. CTCF and YB-1 cooperate in repressing the c-myc promoter. Approximately 10 5 COS6 cells per well were plated in 12-well plates and transfected, and CAT assays were performed as described under "Experimental Procedures." A, the reporter construct p232-S1-CAT schematic map. It was co-transfected with the vectors expressing CTCF and/or YB-1. An "empty" vector, pSG5, was used as a control. B, CAT activities are presented in absolute units. Transfections were performed in triplicates in five independent experiments, and statistically significant (by t test) results are shown. the other hand, a correlation has been established between YB-1 expression and the development of a malignant phenotype in several tumors including breast cancer (26), colorectal carcinoma (71), ovarian serous adenocarcinoma (72), and osteosarcoma (73). Normally localized in the cytoplasm, YB-1 is translocated into the nucleus following environmental stress (25), and cellular redistribution is reported to be associated with certain cancer phenotypes characterized by the P-glycoprotein overexpression (26,73). Taken together, these findings suggest that deregulated and/or abnormal CTCF/YB-1 interactions play a role in tumor development, a subject of our ongoing studies.