The Role of CDP in the Negative Regulation of CXCL1 Gene Expression*

The CXC chemokine, melanoma growth stimulatory activity/growth-regulated protein, CXCL1 is an important modulator of inflammation, wound healing, angio-genesis, and tumorigenesis. Transcription of CXCL1 is regulated through several cis-acting elements including Sp1, NF- k B, and an element that lies immediately upstream of the NF- k B element, the immediate upstream region (IUR). A transcription element data base search indicated that the IUR element contains a binding site for the transcriptional repressor, human CUT homeodomain protein/CCAAT displacement protein (CDP). It is shown here that in electrophoretic mobility shift assays, complexes obtained with the IUR oligonucleotide probe are supershifted by anti-CDP antibodies and that a CDP polypeptide containing a high affinity DNA binding domain binds to the sequence GGGATCGATC in the IUR element. In Southwestern blot analyses, oligonucleotides containing the wild-type IUR sequence, but not a mutant oligonucleotide with substitutions in the GGGATCGATC sequence, bind a 170–180-kDa protein. Fur-thermore, overexpression of the CDP protein blocks CXCL1 promoter activity in reporter gene assays, whereas overexpression of an antisense CDP construct leads to a significant increase in CXCL1 promoter activity.

plays an important role in wound healing, inflammation, and tumorigenesis (1)(2)(3)(4)(5). The CXCL1 gene is not constitutively expressed in normal retinal pigment epithelial cells (RPE) but can be induced by cytokines such as interleukin-1 (IL-1) and tumor necrosis factor-␣ and bacterial products such as lipopolysaccharide. In contrast, Hs294T malignant melanoma cells exhibit high, constitutive levels of CXCL1 mRNA and protein.
IL-1 treatment of Hs294T cells does not significantly increase the already elevated transcription of the gene, although it does appear to stabilize CXCL1 mRNA (6).
Transcription of the CXCL1 gene is regulated largely through a 306-bp minimal promoter situated immediately upstream of the transcription start site. The four cis-acting elements comprising the minimal promoter include a TATA box (Ϫ25 to Ϫ30 bp), an NF-B-binding site (Ϫ67 to Ϫ77 bp), an AT-rich HMGI(Y)-binding element nested within the NF-B site, an immediate upstream region (IUR) (Ϫ78 to Ϫ96 bp), and a GC-rich SP1-binding site (Ϫ117 to Ϫ128 bp) (7,8). In RPE cells, IL-1 induction results in increased nuclear levels of NF-B p65 (RelA) and NF-B p50 subunits (9). This is a consequence of an IL-1-induced activation of the IB kinases (IKK1/IKK2) and subsequent phosphorylation, ubiquitination, and degradation of the IKK substrate, IB. In Hs294T cells, constitutively high nuclear levels of the p50 and p65/RelA proteins can be correlated with constitutive activity of IKK1/ IKK2 and enhanced degradation of the IB protein (10). The NF-B element therefore represents a crucial, inducible component of the putative CXCL1 enhanceosome.
The IUR is an ϳ20-bp sequence that is located immediately upstream of the NF-B element in the CXCL1 promoter. Previously, we demonstrated that in addition to the NF-B, the IUR element is essential for basal as well as cytokine-induced expression of the CXCL1 gene. In particular, point mutations within a putative TCGAT motif of the IUR element abolished basal and IL-1-induced transcription in reporter gene assays with RPE and Hs294T cells (8). Furthermore, in electrophoretic mobility shift assays (EMSA), these mutations blocked the ability of this element to compete with a constitutive, IURspecific complex in RPE and Hs294T nuclear extracts. UV cross-linking and Southwestern blot analyses revealed that at least one protein having a relative molecular size of 115 kDa bound the IUR element in a sequence-specific manner (5). Purification of the 115-kDa IUR-specific protein by oligonucleotide affinity chromatography revealed its identity as the poly-(A)DP-ribose polymerase (PARP) and demonstrated that it plays a role in the activation of the CXCL1 promoter (11).
In this study, the CXCL1 promoter was analyzed within a transcriptional element data base (Transfac) using a webbased search engine, the Transcription Element String Search (TESS). The search identified the CXCL1 IUR element as a putative binding site for the human CCAAT displacement protein (CDP) that is highly homologous to the Drosophila CUT protein. Genetic studies first identified the CUT locus as an important determinant of cell type specificity in Drosophila melanogaster (12)(13)(14). The human CDP is a homeodomain protein that is composed of an N-terminal, coiled-coiled domain, three highly homologous ϳ70 amino acid long CUT repeat domains, a C-terminal homeodomain, and two transcription repression domains (12,15). The three CUT repeats and the homeodomain bind DNA with differing affinities. However, high affinity sequence-specific DNA binding requires the cooperation of at least two domains, in particular CUT Repeat 3 and the homeodomain (16 -20). A consensus binding site for the CDP protein was identified by polymerase chain reaction-mediated site selection and was specified by the sequence (T/ G)ATCGAT(C/A) (18,19). CDP is an active repressor of cell cycle-dependent or differentiation-specific genes including gp91-phox, (21), p21 WAF1/CIP1/SDI1 (22), osteocalcin (23), thymidine kinase (24), cystic fibrosis related trans-conductance receptor (25), and c-myc (26). CDP also regulates the immunoglobulin heavy chain enhancer as a component of the factor NF-NR (27). CDP has been shown to interact with accessory factors such as cyclin A and p107 retinoblastoma-related protein to form proliferation-specific complexes (28). Similar complexes containing CDP, cyclin A, and polyoma virus large T antigen have been causally related to development of uterine leiomyomas (29).
Here, the IUR element of the CXCL1 promoter has been further characterized. The data indicate that the IUR can bind a protein having a molecular size of 170 kDa and that this protein is the 170-kDa CDP. In EMSAs, recombinant CDP polypeptides bound the IUR element in a sequence-selective fashion. In co-transfection experiments, overexpression of the CDP protein inhibited CXCL1 promoter activity, whereas overexpression of antisense CDP mRNA induced CXCL1 promoter activity 5-fold over the control. These results indicate that the transcription of the CXCL1 gene is negatively regulated by the CCAAT displacement protein.

MATERIALS AND METHODS
Cell Lines and Treatments-Hs294T melanoma cells are a continuous cell line established from a human melanoma metastatic to the lymph node. These cells were obtained from American Type Culture Collection (Manassas, VA). RPE cells are normal retinal pigment epithelial cells that were cultured by Dr. Glenn Jaffe from the North Carolina Organ Donor and Eye Bank within 24 h of death (30). RPE and Hs294T cells were cultured as described previously (9).
Reporter and Expression Vectors-The CXCL1 minimal promoter region (Ϫ306 to ϩ45 bp), which was originally constructed in a CAT reporter system (7), was inserted in the pGL2 Basic vector (Promega, Madison, WI). The mutant reporter mIUR.Luc was generated by using the GeneEditor in vitro site-directed mutagenesis system (Promega, Madison, WI). The mutagenic oligonucleotide employed for this purpose had the sequence: 5Ј-cgggaATGaCctggaactccgggaatttccctggcc-3Ј. Uppercase letters represent nucleotide replacements in the GGGATC-GATC motif of the IUR element, which is underscored. The reporter vectors, p2RT.Luc, p2mRT.Luc, and p2dRT.Luc, were generated by inserting XhoI/HindIII-flanked oligonucleotides in the XhoI and Hind-III sites of pGL2. The oligonucleotide sequence for the wild-type reporter construct, 2 RT.Luc, was 5Ј-tcgaggggatcgatctggaactccgggatcgatctggaactcctggcccgggggctccgggctttccagccccaaccatgcataaaaggaagctta-3Ј. The underscored letters represent the GGGATCGATC motif of the IUR element. The oligonucleotide sequence for the mutant reporter 2dRT.Luc was tcgagcctcgatctggaactcctcgatctggaactccggatgctggcccgggggctccgggctttccagccccaaccatgcataaaaggaagctta-3Ј. The underscored letters represent the truncated IUR motif in which the 5Ј-GGGA sequence has been deleted. An expression vector containing the full-length CDP cDNA in the sense orientation (pMX.HSCDP) and another with the CDP cDNA in the antisense orientation (pMXASCDP) were previously described (19,20). All reporter and expression constructs were confirmed by sequencing.
Reagents-Monoclonal and polyclonal antibodies raised against the CDP protein as well as recombinant CDP polypeptides have been characterized previously, and the specificity of these antibodies to the various domains of the CDP protein has been demonstrated (31).
Electrophoretic Mobility Shift Assays-Nuclear extracts were prepared as described previously. All extracts contained a 1ϫ concentration of Complete protease inhibitor mixture (Roche Molecular Biochemicals). The 2R oligonucleotide, 5Ј-gggatcgatctggaactccgggatcgatctggaactcc-3Ј, was radiolabeled by extension of an annealed primer, 5Ј-ggagttccagatc-3Ј. The probe 2mR had the oligonucleotide sequence 5Ј-gggaagtacctggaactccgggaagtacctggaactcc-3Ј and was primed by an oligonucleotide primer, 5Ј-ggagttccaggta-3Ј. (Underscored letters represent the nucleotide replacements in the GGGATCGATC motif of the IUR element.) The probe 2dR, 5Ј-cctcgatctggaactcctcgatctggaactccggatgc-3Ј, was primed with 5Ј-gcatccggagttcca-3Ј. The primed oligonucleotide probes were then radiolabeled with the Klenow fragment of Escherichia coli DNA polymerase, dNTPs, and [␣-32 P]dCTP. The probe containing the NF-B-binding site 5Ј-agttgaggggactttcccaggc-3Ј was from Promega (Madison, WI) and was radiolabeled using T4 polynucleotide kinase and [␥-32 P]ATP. A typical binding reaction involved a 15-min preincubation with 10 g of nuclear extract, 2 g of the nonspecific competitor poly(dI⅐dC), 200 ng of single-stranded oligonucleotide, 20 mM Hepes-NaOH (pH 7.6), 100 mM NaCl, 1 mM dithiothreitol, 2% glycerol, followed by a 20-min incubation with 50,000 cpm (40 fmol) of radiolabeled probe. In oligonucleotide competitions, 1000-fold molar excess of cold, double-stranded oligonucleotide was added to the preincubation mix. In some oligonucleotide competition experiments, the CDP-binding site oligonucleotide (wCDP) used had the upper-strand sequence, 5Ј-aaaagaagcttatcgataccgt-3Ј. Mutations in the CDP-binding site, previously shown to affect binding to the CDP DNA binding domains (18), were included in the oligonucleotide (mCDP) that had the following sequence, 5Ј-aaaagaagcttatTCataccgt-3Ј. Underscored letters represent the core CDP-binding motif, and uppercase letters indicate the nucleotide replacements in this sequence. In supershift analyses, antibodies raised against CDP domains other than the DNA binding domain (W, A, N) were added after the binding reaction and incubated for an additional 10 min at room termperature prior to electrophoresis. In cases where the antibody was against the CDP DNA binding domain (AK), the antibody was included in the preincubation reaction. The concentration of antibody in each EMSA reaction was 2 g/10 g nuclear extract. Complexes were resolved by electrophoresis for 2 h at 170 V on a 4% native, polyacrylamide gel, which was later dried and processed for autoradiography. In cases where purified CDP polypeptides were used, the reaction mixture contained no cold competitor, and the complexes were resolved on a 6% native acrylamide gel.
Southwestern and Western Blot Analysis-Nuclear extracts (25 g) were heated at 90°C for 3 min in 50 mM Tris⅐Cl (pH 6.8), 100 mM dithiothreitol, 2% SDS, and 10% glycerol, resolved on 4% stacking, 8% resolving SDS-polyacrylamide gel, and electrophoretically transferred to nitrocellulose membranes (Bio-Rad). For Southwestern analysis, membranes were rocked for 15 min in PBS, blocked for 4 h at room temperature with buffer A (20 mM Hepes-NaOH, 50 mM NaCl, 12.5 mg/ml skim milk powder, 2.5 mg/ml bovine serum albumin, 100 g/ml native salmon sperm DNA), and incubated overnight at room temperature in up to 2 ml of buffer A ϩ 10 7 cpm (8 -10 pmol) of radiolabeled probe. Probes used were either 2R, 2mR, or 2dR which are described under the "Electrophoretic Mobility Shift Assays." Membranes were subjected to three washes of 15 min each at room termperature in a buffer containing 20 mM Hepes-NaOH, 50 mM NaCl, 1 g/liter skim milk powder, and 0.025% Nonidet P-40, prior to drying and autoradiography.
For Western analysis, following transfer, membranes were blocked in Tris-buffered saline containing 0.05% Tween 20 and 5% skim milk (TBST) and probed with rabbit polyclonal anti-CDP antibody at a dilution of 1:2500 (0.8 g/ml) using standard procedures. The specificity of this antibody has been demonstrated previously (31). The signal was visualized by enhanced chemiluminescence (ECL) assay (Amersham Pharmacia Biotech) according to the manufacturer's recommendations.
Transient Transfection and Reporter Activity Assay-Hs294T or RPE cells were plated in 60-mm dishes at a density of 2 ϫ 10 5 . The following day, cells at nearly 70% confluency were transfected using the Lipo-fectAMINE/Plus method (Life Technologies, Inc.) with 1 g of the reporter construct pGL2.CXCL1, pGL2mIUR, pGL2.2mRT, or pGL2.2RT, and whenever necessary, 1 g of the CDP expression vector either pMXHSCDP or the CDP antisense construct pMXASCDP. In addition, all samples received 1 g of pCMV-␤-gal. The total amount of DNA was kept constant by supplementing with either pGL2 or pUC19 vector DNA. Cells were harvested 48 h after transfection, and luciferase activity was measured using the luciferase assay system (Promega, Madison, WI) and Monolight 2010 Luminometer (Analytical Luminescence Laboratory, San Diego, CA). All values were normalized to ␤-galactosidase expression to correct for transfection efficiency. Each experiment was performed 3-6 times, with 3-4 different plasmid preparations. Each time the samples were in duplicate. The transfection efficiency of the sense and antisense CDP constructs was about 45-60%. Mean values and standard deviation was determined from 3 to 6 separate experiments. Student's paired t test p values were determined using Microsoft Excel.

Sequence Analysis of the CXCL1 Promoter Reveals a Putative
Binding Site for the CCAAT Displacement Protein-The immediate upstream region (IUR) has been described previously as essential for CXCL1 transcription. It has been shown earlier that this element contains a TCGAT motif, which is essential for binding to a 115-kDa protein. The factor binding to this element has been purified by oligonucleotide affinity chromatography and has been identified by mass spectroscopy/matrixassisted laser desorption ionization spectroscopy as PARP (11). In this report we show that the IUR binds a second transcription factor that has specificity to the GGGATCGATC motif of the IUR element and negatively regulates CXCL1 gene transcription.
In order to identify putative transcription factor-binding sites in the IUR element (Ϫ96 to Ϫ78 nt) of the CXCL1 promoter, we employed a web-based search engine, TESS (www.cbil.upenn.edu/tess) (32), and analyzed the sequence corresponding to Ϫ130 bp to ϩ20 of the minimal CXCL1 promoter (Fig. 1). In particular, our objective was to identify putative transcription factor-binding sites within the sequence extending from Ϫ96 to Ϫ78 nt (underscored letters in Fig. 1) which we designate here as sequence R. The search was restricted to a maximum of 20% mismatch within an element length of 6 nucleotides or greater. As a negative control, a mutant promoter sequence with replacements in the TCGAT motif of the IUR element was also used. Fig. 1 describes the sequences identified in the Ϫ130 to ϩ20 nt region by the TESS search. In addition to the previously characterized Sp1 and NF-B sites, an 8-nt sequence (Ϫ94 to Ϫ87 nt) located within the IUR element exhibited a 100% match with the putative binding site for the transcriptional repressor, CDP. The CDP-binding site was not detected by a similar search of the mutant promoter sequence in which the TCGAT motif was altered. The consensus sequence for the CDP-binding element is (T/G)ATCGAT(C/ A). This consensus sequence perfectly matches the GATCGATC motif in the IUR element. Other putative sites identified by the search included binding sites for the transcription factors, Sp1, C/EBP␤, opaque-1, and dorsal. In particular, putative Sp1, opaque, and C/EBP ␤ sites were detected within the Ϫ100 to Ϫ77-nt region of the CXCL1 promoter.
The IUR Binds a 170-kDa Protein-We first examined sequence-specific binding to the IUR element. Fig. 2A describes the nucleotide sequence Ϫ96 to Ϫ67 nt of the CXCL1 promoter. Uppercase letters represent the 20-bp IUR element, which we designate here as sequence R. Lowercase letters represent the adjacent NF-B element. The TCGAT motif is underscored. Oligonucleotide probes containing a single copy of the sequence R showed weak binding in electrophoretic mobility shift assays. However, probes containing two tandem copies of sequence R showed significantly stronger binding (data not shown). Fig. 2A shows the sequence of the oligonucleotide probes 2R and 2mR used in binding experiments in this study. The oligonucleotide probe 2R contains two tandem copies of the wild-type IUR sequence. The oligonucleotide probe 2mR contains multiple nucleotide replacements in the TCGAT motif each indicated by an asterisk. The underscored letters indicate the TCGAT motif.
In Fig. 2B, sequence-specific binding to the 2R and 2mR oligonucleotides is examined. Nuclear extract from RPE cells was analyzed by electrophoretic mobility shift assay on 4% native acrylamide gels ( Fig. 2B) for binding to 32 P-labeled oligonucleotides corresponding to the 2R (lanes 1-3) or the 2mR sequence (lanes 4 -6). Two specific complexes were generated by the 2R probe but not the 2mR probe (Fig. 2B, compare lanes 1 and 4). The 2R-specific complexes were then competed with 1000-fold molar excess of cold oligonucleotide corresponding to either the wild-type (2R) or the mutant (2mR) sequences. The complexes with the 2R probe were resistant to competitions with 1000-fold molar excess of the oligonucleotide 2mR (Fig.  2B, lane 2) but were completely eliminated by a similar excess of the 2R oligonucleotide. Since the only difference between the two oligonucleotide probes is the sequence of the TCGAT motif, the results of this experiment indicate that these complexes are specific to the TCGAT motif of the IUR element.
In order to determine the relative molecular sizes of the factors interacting with the 2R probe, nuclear extract from RPE cells was examined for binding to the 2R oligonucleotide probe using Southwestern blot analysis. RPE nuclear extracts were first resolved on an 8% SDS-polyacrylamide gel and transferred to nitrocellulose membranes, and the blots were incubated with 32 P-labeled 2R probe. Similar blots probed with the 2mR oligonucleotide probe served as a negative control. Results of such an experiment are represented in Fig. 2C. The 2R oligonucleotide bound to two distinct proteins having relative molecular sizes of 170 and 80 kDa (lane 2). However, when probed with the 2mR oligonucleotide, no binding was detected FIG. 1. Transcription element string search of the CXCL1 promoter. The sequence contained within nucleotides Ϫ130 and ϩ20 nt of the CXCL1 promoter was analyzed using a web-based search engine (TESS). The sequence extending from Ϫ96 to Ϫ78 nt has been underscored and is designated here as sequence R. Bold arrows indicate transcription factor-binding sites identified by TESS that have a high likelihood of occurrence with minimum number of mismatched bases. Broken arrows represent sites that had a greater degree of mismatches with the CXCL1 promoter sequence and are therefore predicted to be less likely to occur. The direction of the arrow indicates the sense or the antisense strand of the CXCL1 promoter that scored a hit. The putative CDP-binding site as identified by TESS is located between the nucleotides Ϫ94 and Ϫ87 and is indicated by 7 . The lightface arrow indicates the initiation site for CXCL1 transcription.
in these extracts (lane 4). The results are consistent with data from the EMSA experiment in which the 2R oligonucleotide was bound by two specific complexes, whereas the 2mR oligonucleotide probe failed to bind any complex. The data together indicate that binding of both polypeptides appears to be specific to the TCGAT motif of the IUR element because mutations in this motif appear to block binding to the IUR element.
Results from a transcription element search indicated that the IUR element contained a binding site for the CDP, which has a molecular size of 170 -180 kDa. One of the proteins that bound to the 2R oligonucleotide probe also migrated with a relative molecular size of 170 kDa. We next examined the expression of CDP in RPE nuclear extracts using anti-CDP antibody in order to verify the presence of CDP in RPE nuclear extracts and determine the relationship of the two proteins p170 and p80 that bound to the 2R probe in Southwestern blot analysis. We reasoned that if the antibody detected only the 170-kDa polypeptide then the 80-kDa polypeptide was unrelated to CDP and distinct from it. However, if it recognized both polypeptides it would mean that the 80 kDa polypeptide was either a cleavage product of the 170-kDa CDP or antigenically similar to it. We examined RPE nuclear extracts by Western blot analysis using anti-CDP antibodies. In Fig. 2D, RPE ex-tracts were first electrophoresed on an 8% SDS-acrylamide gel and transferred to nitrocellulose membrane, and the blots were incubated with antibodies raised against the C-terminal domain of the CDP protein. The data indicate that anti-CDP antibody detected two polypeptides with molecular sizes 170 and 80 kDa in RPE nuclear extracts, a result that closely resembled the binding data from the Southwestern blot analysis using the 2R oligonucleotide probe in Fig. 2C. Taken together, results from the TESS analysis and the Southwestern and the Western blot assays suggest the possibility that the 80-kDa polypeptide may be a specific cleavage product of the 170-kDa CDP protein or an alternatively spliced product of the CDP gene, analogous to the 76-kDa CDP spliced product, CASP (33).
The Consensus Sequence for the CCAAT Displacement Protein (CDP)-binding Site Competes with Binding to the 2R Probe-If CDP indeed binds to the IUR element, we reasoned that it should be present in the complexes associated with the 2R oligonucleotide probe in EMSA. Furthermore, recombinant CDP should be able to bind specifically the 2R oligonucleotide probe in EMSA. We employed three approaches to test whether CDP is a constituent of the 2R complexes in EMSA. First, the effect of competition with wild-type and mutant CDP-binding FIG. 2. The IUR element binds two proteins, p170 and p80. A, sequences of the Ϫ96 to Ϫ67 nt of the CXCL1 promoter and the oligonucleotide probes, 2R and 2mR, are shown. Uppercase letters in the CXCL1 promoter sequence represent the IUR element, and lowercase letters show the adjacent NF-B-binding site. The sequence underscored in probes 2R and 2mR represents the TCGAT motif between nucleotides Ϫ93 and Ϫ89 in the CXCL1 promoter. Asterisks below the 2mR sequence indicate nucleotide replacements in the TCGAT motif. B, EMSA. RPE nuclear extracts were incubated with the 2R probe (lanes 1-3) or the 2mR probe (lanes 4 -6). Two distinct complexes migrating as a doublet are obtained with the 2R probe. Complexes were competed with 1000-fold molar excess of cold 2R or 2mR oligonucleotides as indicated above the panel. site oligonucleotides on complexes generated with the probe 2R was examined. Second, we tested whether recombinant polypeptides corresponding to specific domains of the CDP protein possessed the ability to bind selectively the IUR element. Third, we tested the effect of antibodies raised against different domains of the CDP protein on complexes generated with the 2R probe. Based on the work of Harada et al. (19) a high affinity CDP-binding site was selected as a potential competitor. Sequences for this oligonucleotide and its corresponding mutant are described under "Materials and Methods." Fig.  3A shows a non-scaled schematic representation of the human CDP and the recombinant polypeptides corresponding to the two DNA binding domains of the protein. Fig. 3B shows the sequences of the three oligonucleotide probes used in this experiment. In addition to the 2mR mutant, we used a third oligonucleotide probe, 2dR, which lacked the Ϫ96to Ϫ93-nt sequence 5Ј-GGGA-3Ј. The underscored letters represent the TCGAT motif of the IUR element. In Fig. 3C, complexes generated with 2R (lanes 1-3) were competed with a 1000-fold molar excess of oligonucleotides that contained either the consensus CDP-binding site (lane 3) or a specific mutant of that site (lane 2). The wild-type CDP-binding sequence, but not the mutant CDP sequence, diminished the complexes obtained with this probe but had no significant effect on complexes generated with an oligonucleotide probe containing an NF-Bbinding site (lanes 5 and 6), suggesting that both complexes associated with probe 2R contain the CCAAT displacement protein.
These observations were followed by exploring the ability of various sub-domains of the CDP protein to bind the 2R probe. A sub-domain of CDP comprising the CUT repeat 3 and home- . Polypeptides corresponding to two different DNA binding domains, one comprising CR1 and CR2 (C1C2) and the other comprising CR3 and HD-DBD (C3HD), were used in this study. B, sequences of oligonucleotide probes. The probe 2R contains two tandem copies of the 20-bp IUR element. Mutant 2mR contains multiple nucleotide replacements in the TCGAT motif, which are indicated by an asterisk below. The mutant 2dR contains a truncated IUR sequence in 2 tandem repeats. The mutation involves deletion of the 5Ј-GGGA-3Ј sequence corresponding to the region between nucleotides Ϫ97 and Ϫ94 of the CXCL1 IUR element. The deleted nucleotides are indicated by dashes. Underscored letters in all three oligonucleotide sequences represent the TCGAT motif of the IUR element. C, oligonucleotide competition. Complexes obtained in EMSA with the 2R oligonucleotide probe were competed with a 1000-fold molar excess of oligonucleotides containing either the mutant (lane 2) or wild-type-binding site sequence (lane 3) and were resolved on a 4% native polyacrylamide gel. The sequences for the wild-type CDP or the mutant CDP oligonucleotides are described under "Materials and Methods." Arrows indicate the two complexes that are observed with the 2R probe. D, EMSA. Radiolabeled probes 2R (lanes 1 and 4), 2mR (lanes 2 and 5), or 2dR (lanes 3 and 6) were incubated with polypeptides comprising CUT repeat-3 homeodomain (C3HD, lanes 1-3) or CUT repeat 1-Cut repeat 2 (C1C2, lanes 4 -6), and the complexes were analyzed by EMSA on a 4% native polyacrylamide gel. Arrows indicate the sequence-selective complexes formed by the 2R probe with the C3HD polypeptide. Data shown here are representative of Ͼ3 independent experiments.  1-7) or a probe (NF-B) representing the consensus NF-B-binding site (lanes 8 -12) contained monoclonal antibodies (mAB) against various domains of the CDP binding as indicated above each panel or a nonspecific rabbit IgG (lanes 2 and 9) as a control. The arrows indicate the mobility of the supershifted complexes. odomain (C 3 HD) (Fig. 4A) was demonstrated to have high affinity, sequence selective binding to the (T/G)ATCGAT(C/A) motif, whereas a sub-domain spanning CUT repeats 1 and 2 (C 1 C 2 ) was unable to bind this sequence (19,20). Our next objective was to examine whether the C 3 HD polypeptide exhibited sequence-selective binding to the 2R probe. We selected the C 1 C 2 polypeptide as a negative control for sequence-specific binding to the 2R probe as this domain was reported to exhibit only low affinity and sequence non-selective DNA-binding properties (19). EMSA in Fig. 3D shows that the C 3 HD subdomain produced two distinct complexes with the 2R but not with the 2mR or the 2dR probes. The data indicate that the high affinity, sequence-specific binding exhibited by the C 3 HD polypeptide requires not only the TCGAT motif of the IUR element but also the GGG sequence at the 5Ј end of the motif 5Ј-GGGATCGATC-3Ј. In contrast to the sequence-selective binding exhibited by the C 3 HD subdomain, the polypeptide, C 1 C 2 , failed to bind any of the three probes. These results are consistent with the reported binding properties of the two subdomains. The two band pattern generated by C 3 HD is probably a consequence of the protein binding one or both the copies of the IUR element in the 2R probe.
Anti-CDP Antibodies Recognize Complexes Generated by the 2R Sequence-Next, the effect of antibodies against various domains of the CDP protein (shown in Fig. 4A) was examined in supershift analyses with the 2R probe. In experiments represented in Fig. 4B, the 2R-specific complexes were generated in nuclear extracts from RPE cells by EMSA. EMSA reactions were incubated with specific antibodies raised against various domains of the CDP protein. Almost all the anti-CDP antibodies either eliminated the retarded complexes or produced a supershifted complex. In contrast a nonspecific antibody did not affect the migration of the complexes (Fig. 4B, lane 2). In particular, monoclonal antibodies against the CUT Repeat 3 alone , virtually eliminated the two complexes associated with the 2R probe. The antibodies had no effect on complexes associated with the consensus NF-B element (lanes 8 -12). In addition, antibodies against the other transcription factors such as Sp1, Sp3, c-Jun, c-Fos, and C/EBP␤ had no significant effect on these complexes (data not shown). Similar results were obtained with nuclear extracts from Hs294T melanoma cells, HEK293, and HeLa cells (data not shown). These data clearly demonstrate that CDP is a constituent of the specific complex bound to the IUR element.
CDP Transcriptionally Represses CXCL1 Promoter Activity-To determine whether CDP transcriptionally regulates the CXCL1 promoter, the effects of CDP overexpression in transient transfection and reporter gene assays (Fig. 5) were examined. RPE cells (Fig. 5A) or Hs294T cells (Fig. 5B) were cotransfected with expression constructs of the CDP protein in either the sense or the antisense direction along with luciferase reporter constructs, driven by either the 306-bp wild-type CXCL1 promoter or a mutant CXCL1 promoter that had replacements in the TCGAT motif of the IUR sequence. Forty eight h after transfection, cells were harvested. Western blot analysis of RPE cell lysates (Fig. 5A) with an anti-CDP antibody (top panel) showed that cells transfected with the CDP construct in the sense orientation (Fig. 5A, lane 2) had at least 2-3-fold higher levels of the 170-kDa CDP protein as compared with lysates from untransfected cells (compare lanes 1 and 2). In contrast, in cells transfected with the antisense CDP construct, almost no CDP protein was detected, indicating that the antisense CDP RNA efficiently ablated CDP levels in these cells ( compare lanes 1 and 3). Of the two polypeptides, p170 and p80, detected by the anti-CDP antibody (Fig. 2D), only the 170-kDa protein is shown in this experiment. An antibody against a 115 constitutive nuclear protein, PARP, showed no differences in the levels of PARP with overexpression of either construct (Fig. 5A, bottom panel). Cell lysates from the above transfectants were assayed for luciferase activity. The results indicate that overexpression of CDP in RPE cells did not significantly affect the low basal activity of the CXCL1 promoter (Fig. 5B). However, an almost 50% inhibition of the constitutively high basal CXCL1 promoter activity in Hs294T cells was observed. Overexpression of the CDP sense construct clearly inhibited the CXCL1 promoter in Hs294T cells (Fig. 5C). Expression of the CDP gene in the antisense direction, however, led to a striking 4 -5-fold increase in transcription from the CXCL1 promoter in both cell lines. The results indicate that a decrease in CDP levels particularly in RPE cells relieves CDPmediated repression of the CXCL1 promoter in these cells.
The results with the mutant promoter having replacements in the TCGAT motif are intriguing. Oligonucleotide probes with replacements in the TCGAT motif fail to generate 2Rspecific CDP containing complexes in EMSA and do not bind the 170-and 80-kDa proteins in Southwestern blot assays (Fig.  2, B and C). Overexpression of CDP represses CXCL1 promoter activity in luciferase assays in Hs294T cells, and overexpression of the antisense CDP construct leads to a 4 -5-fold increase in transcription from the CXCL1 promoter in RPE cells (Fig.  5B) and 2-2.5-fold in Hs294T cells (Fig. 5C). If CDP-mediated events at the IUR element are indeed eliminated by this mutation, then one could expect that the TCGAT mutation would relieve CDP-mediated repression of the CXCL1 promoter and allow an increase in promoter activity compared with the wildtype promoter.
We considered several possibilities to address this question. First, the consensus sequence for CDP binding is considerably degenerate. It is possible that while the mR mutation is effective in eliminating CDP binding in EMSAs and Southwestern assays, it may only partially affect CDP binding in an in vivo situation. A second possibility is that reduction in the levels of the CDP protein by antisense RNA overexpression could affect expression of other trans-acting factors involved in the regulation of the CXCL1 promoter, notably the Rel family of transcription factors. A TESS of the relA promoter reveals a putative binding site for the CDP repressor, although there is no evidence in the literature linking the two proteins. A third possibility is that the IUR element may contain a binding site for a second protein, which could be an activator of CXCL1 transcription, and the TCGAT mutant may have blocked binding to both the 170-kDa protein as well as the activator protein.
In support of the last assumption, previous studies from our laboratory demonstrated that the TCGAT motif of the IUR element is crucial to the binding of the 115-kDa protein poly-(A)DP-ribose polymerase that has a distinct positive role in the transcriptional activation of the CXCL1 promoter (11).
The IUR element of the CXCL1 promoter extends from Ϫ96 to Ϫ78 nt. The sequence between positions Ϫ96 and Ϫ87 nt, 5Ј-GGGATCGATC-3Ј, contains a binding site for the transcriptional repressor CDP. Our next objective was to generate a mutant sequence of the IUR, which more effectively eliminated IUR-CDP interactions in vivo. We considered a mutant oligonucleotide in which the first four nucleotides 5Ј-GGGA-3Ј of the IUR motif 5Ј-GGGATCGATC-3Ј are deleted. We designate this sequence 2dR. The 2dR sequence is a significant departure from the consensus CDP-binding site. The 2dR sequence failed to bind the high affinity DNA binding (C 3 HD) domain of the CDP protein (Fig. 3C, lane 2). In an EMSA experiment with RPE extracts, this sequence did not generate a complex or compete with complexes generated by the 2R sequence (Fig.  6A). In Southwestern blot analysis (Fig. 6B), the 2R oligonucleotide probe, containing the wild-type IUR sequence, bound a 170-and an 80-kDa polypeptide in RPE nuclear extracts. The mutant 2dR oligonucleotide failed to bind either of the two polypeptides detected by the 2R sequence.
By having confirmed that the 2dR sequence did not bind the 170-and 80-kDa polypeptides interacting with the wild-type (2R) sequence, our next objective was to evaluate the effects of the dR mutation on reporter gene activity. In order to rule out the possibility that effects of the antisense CDP overexpression on CXCL1 activity may involve other transcriptional regulators of the CXCL1 gene, we generated luciferase reporter constructs driven by a basic transcriptional module containing only the CXCL1 TATA box and two copies of either the R sequence (2RT.Luc) or the dR sequence (2dRT.Luc). In Fig. 6C, RPE cells were transfected with pCMV-␤GAL and either 2RT.Luc or 2dRT.Luc, and luciferase activity was measured after 48 h. Relative light unit values obtained were normalized for ␤-galactosidase expression. The 2RT.Luc construct was expressed at levels 3-4-fold higher than the 2dRT.Luc construct (p Ͻ 0.01, Student paired t test). Taken together with the data from binding experiments in Fig. 6, A and B, the results from the reporter assay using the 2RT and 2dRT constructs strongly suggest that in RPE cells the wild-type IUR sequence represses reporter gene activity, and this repression can be relieved by mutations, such as the dR mutation, which effectively block the binding of CDP to this element.
If the dR mutation indeed abolishes CDP-IUR interactions, then overexpression of the CDP cDNA in the sense or the antisense direction should have absolutely no effect on transcription directed by the 2dR oligonucleotide. To verify whether this is indeed the case, we examined the effect of CDP overexpression on reporter activity driven by either the 2RT.Luc or the 2dRT.Luc promoters. RPE cells were co-transfected with either the 2RT.Luc or 2dRT.Luc constructs along with expression constructs of the CDP cDNA in either the sense or the antisense orientation. Luciferase activity was measured 48 h after transfection. Results in Fig. 6D are representative of three separate experiments, each performed in duplicate. The results indicate promoter activity of the 2dRT.Luc construct was found to be at least 5-fold higher than that of the 2RT.Luc construct (p Ͻ 0.01, Student paired t test). Overexpression of CDP sense or antisense CDP constructs had no effect on this high level of expression (p Ͻ 0.01, Student paired t test). In striking contrast to effects on the 2dRT.Luc construct, co-transfection with the antisense CDP construct led to a nearly 4-fold stimulation of the 2RT.Luc promoter activity, (p Ͻ 0.01, Student paired t test), clearly demonstrating that CDP-mediated repression of the 2RT construct occurred outside the context of the CXCL1 promoter and was not due to the neighboring NF-B or Sp1 elements in the CXCL1 promoter. The high transcriptional activity of the 2dRT.Luc construct relative to in the 2mR oligonucleotide probe, shown in Figs. 2 and 3. Cells also received expression constructs containing the CDP cDNA in either the sense or the antisense orientation. Data shown here are from a representative of six independent experiments. Error bars indicate S.D. values. Luciferase activity values obtained with the antisense construct alone were found to be significantly different than those with the empty vector as determined by Student's paired t test, and p value was found to be Ͻ0.01, indicated by an asterisk. were co-transfected with pCMV-␤-gal and a reporter construct containing either the wild-type CXCL1 promoter (CXCL1.Luc) or a mutant promoter (mIUR.Luc) that contained the multiple replacements in the TCGAT motif of the IUR element. The replacements were the same as the 2RT.Luc construct also points to the possibility that this sequence could harbor a positive cis-acting element. This concept is consistent with our earlier observations that mutations in the TCGAT motif significantly inhibited basal as well as cytokine-induced transcription (8) and our observation that this motif has binding site for PARP which is associated with transcriptional activation of the CXCL1 promoter (11).
Reduction in levels of the CDP function by antisense overexpression clearly seems to relieve trans-repression at the CXCL1 wild-type promoter. In fact, this effect closely resembles the effect of the mutation dR, which also relieves repression of reporter gene expression. Results from these experiments suggest negative regulatory roles for the CDP protein.
The data indicate that the IUR element of the CXCL1 promoter contains a negative regulatory sequence, which binds the 170-kDa transrepressor CDP.
Findings in this study raise the possibility that CDP may play a role in repression of basal CXCL1 transcription in normal cells. In cases where the CXCL1 gene is overexpressed, such as in malignant melanoma, CDP-mediated trans-repres- FIG. 7. CDP binding activity in normal and melanoma cells. A, nuclear extracts from 1 normal, NHEM, and four melanoma cell lines, Hs294T, SKMEL5, WM115, and A375, were incubated with the 2R probe, and complexes were resolved on a 4% acrylamide gel. Arrows indicate 2R-specific complexes. B, nuclear extracts from RPE T476 and SKMEL5 cells were probed for binding to the 2R probe in an EMSA experiment. Complexes were resolved on a 4% acrylamide gel. Arrows indicate 2R-specific complexes.
above the gel. Complexes were resolved on a 4% native polyacrylamide gel. In this particular case, the free probe was run out to obtain a better resolution of the complexes. B, Southwestern blot assay. 30 g of RPE nuclear extract was resolved on an 8% SDS-polyacrylamide gel, transferred to nitrocellulose membranes, and probed with radiolabeled oligonucleotide probes 2R (left panel) or 2dR (right panel). Relative mobility of molecular size standards is indicated on the left of each panel. Nuclear extracts from RPE cells were incubated with either the radiolabeled oligonucleotide probe 2R (lanes 1-6) or the radiolabeled probe 2dR. The reaction mixes contained a 1000-fold molar excess of the cold competing oligonucleotides 2R, 2mR, or 2dR as indicated in the panel sion could be a potential target for de-regulation of the CXCL1 gene. We explored the possibility that binding of the CDP protein to the IUR element is reduced in Hs294T melanoma cells in comparison to normal RPE cells. Nuclear extracts from normal, NHEM, and malignant melanoma, Hs294T, SKMEL5, WM115, and A375 cells were bound to the 2R oligonucleotide probe in an EMSA experiment (Fig. 7A). Similarly, IUR-specific binding was compared between nuclear extracts from normal RPE and melanoma SKMEL5 cells (Fig. 7B). Both normal and melanoma nuclear extracts produced similar binding patterns with the 2R probe. There were no consistent differences in CDP binding between the normal and melanoma nuclear extracts. When examined by Western blot analysis using an anti-CDP antibody, nuclear extracts from two RPE and five melanoma cell lines showed no striking differences in CDP levels between the cell types (data not shown). It is likely that the trans-repression activity of this protein may be altered in melanoma cells presumably by post-translational modification of the protein.

DISCUSSION
The 170-kDa CCAAT Displacement Protein Binds Specifically to the IUR Element and Negatively Regulates Expression of the CXCL1 Gene-In experiments described here, the role of the IUR element in the regulation of the CXCL1 gene is examined. In this report, we describe specific binding to the IUR sequence, sequence R (Ϫ96 to Ϫ78 nt). The data indicate that the GGGATCGATC motif in the IUR element contains the binding site for two polypeptides, 170 and 80 kDa, interacting with the IUR element of the CXCL1 promoter.
Sequence analysis and transcription element data base search revealed that the CXCL1 IUR element could be a potential binding site for the human CDP. We present several lines of evidence to support this conclusion. Supershift analysis with anti-CDP antibodies, EMSA with recombinant CDP polypeptides, together with reporter gene assays using sense and antisense CDP expression constructs demonstrate that CDP binds the IUR element and represses the CXCL1 promoter in transient transfection assays.
CDP has a relative molecular size of 170 kDa. This is comparable to the size of the IUR-specific 170-kDa protein, which requires an intact GATCGATC motif for binding. In this context, the nature of the 80-kDa polypeptide is somewhat unclear. Both the 170-kDa protein as well as the 80-kDa IURspecific proteins appear to have identical specificities to the GGGATCGATC sequence of the IUR element. In Southwestern assays, binding to both proteins is lost when the TCGAT motif is altered (mR) or the 5Ј-GGG sequence is deleted (dR). In Western blot assays, an anti-CDP antibody detects two polypeptides (170 and 80 kDa) in RPE nuclear extracts. This pattern resembles the Southwestern blot profiles observed with the 2R oligonucleotide. It is possible that the 80-kDa polypeptide may be cleavage product of the 170-kDa protein. A second possibility is that the 80-kDa polypeptide may represent the 76-kDa alternatively spliced product of CDP, CASP (33). The binding of the 170-kDa protein to the 2R probe or the anti-CDP antibody is consistently reproduced across several different cell lines and employing different batches of nuclear extracts. In contrast, the 80-kDa polypeptide is not consistently observed in Southwestern assays with the 2R probe or in Western blots probed with anti-CDP antibodies. We also observed no correlation between the presence and absence of the 80-kDa IURbinding polypeptide and CXCL1 expression. The physiological relevance of the 80-kDa protein in terms of its activity as a DNA binding transcriptional repressor remains to be determined.
Trans-repression by CDP Can Occur Outside the Context of the CXCL1 Promoter-In reporter gene assays, constructs containing two copies of sequence R show a striking decrease in promoter activity when compared with similar constructs containing two copies of the dR sequence. This result has three important implications. First, the IUR element contains a negative regulatory sequence. Transcriptional activity of the 2dRT.Luc promoter, which lacks the 5Ј-GGGA sequence, is 3-4-fold higher than that of the 2RT.Luc promoter. This increase in transcriptional activity could be due to a loss of repressor-binding site. Moreover, a similar increase in transcriptional activity is observed with the wild-type 2RT.Luc construct when levels of the CDP protein are ablated by the antisense CDP overexpression. The data indicate that the GG-GATCGATC motif of the IUR element is a binding site for the 170-kDa, trans-repressor, CDP. Second, CDP-mediated transrepression of the CXCL1 promoter is independent of the Sp1, NF-B, or HMGI(Y)-binding sites in the CXCL1 promoter. This point is supported by the observation that a transcriptional module that lacks these sites, 2RT.Luc, is responsive to CDP ablation by antisense CDP overexpression and exhibits an activity that is at least 3-4-fold higher than that observed in the absence of the antisense CDP construct. Third, the increased transcriptional activity of the 2dR construct may involve a positive cis-acting element contained within the Ϫ93to Ϫ88-nt region of the IUR element. Support for the involvement of a trans-activator interacting with the IUR element comes from a recent report in which we identified sequence-specific binding of a 115-kDa transcriptional co-activator, PARP, to the IUR element, and interactions of this protein require the TCGAT motif of the element (11). Since interactions with the 170-kDa CDP protein also require the TCGAT motif, this sequence may be a potential region of overlap between the binding sites for the two proteins. Negative Regulation of the CXCL1 Promoter by CDP-The CXCL1 promoter, under the control of five cis-acting elements including the Sp1, IUR, NF-B, HMGI(Y), and TATA boxbinding sites, is a template for a multiprotein complex, the CXCL1 enhanceosome. Trans-repression by CDP is thought to involve either promoter occlusion, by preventing binding to other adjacent or overlapping cis-acting elements, or by active repression (25,35). Both these mechanisms may be involved in the regulation of the CXCL1 promoter. Results from transient transfection and reporter gene assays suggest that CDP actively represses the CXCL1 promoter within the context of the minimal CXCL1 promoter, or outside it. It is possible that CDP-mediated repression may also involve displacement of other trans-activating factors that bind to the CXCL1 promoter, such as NF-B, Sp1, HMGI(Y), PARP, or factors contributing to the stability of the CXCL1 enhanceosome. However, we have no direct evidence at this time to support this hypothesis.
Interestingly, Nourbakhsh et al. (34) have identified a novel, negative regulatory cis-acting element downstream to the NF-B-binding site in the promoter of the IL-8 gene, which is referred to as CXCL8, based on a recently adopted system of nomenclature (1). This element controls the basal as well IL-1-induced expression of the CXCL8 gene in cooperation with the NF-B element contiguous with it (34). The IUR element in the CXCL1 promoter may have an analogous function in basal and IL-I-induced expression of the gene. Negative regulation appears to be an important control mechanism for promoters of cytokine genes such as IL-2 (36), IL-3 (37), IL-4 (38,39), IFN-␣ (40), IFN-␤, IFN-␥ (41), tumor necrosis factor-␣ (42), and CXCL8 (34). Enhanceosome models for cytokine gene expression, analogous to the CXCL1 paradigm, have been proposed for the regulation of IL-6 and CXCL8 (IL-8) promoters (43).
Both promoters have binding sequences for NF-B, C/EBP, and the TATA box. The strongest promoter activation relies on the p65 NF-B subunit, which specifically recruits CREB-binding protein (CBP/p300) to the site. Engagement of CBP/p300 in the enhanceosome and its histone acetylase activity have been proposed to stabilize the enhanceosome and stimulate transcription from these promoters (43). In an independent study, CDP has been shown to interact physically with CBP/p300 and is a target for acetylation at specific residues near the homeodomain (44). These models strongly implicate antagonistic roles for CBP/p300 and CDP in the regulation of IL-6 and CXCL8 transcription, although direct involvement of CDP in regulation of either promoter has not been established. Transcription repression by CDP may also involve its ability to recruit a histone deacetylase activity, HDAC1, leading to deacetylation of histones, a phenomenon consistent with transcriptionally inert chromatin (25). Similar interactions between NF-B, CBP/p300, and CDP may be involved in the regulation of CXCL1 gene regulation.
The Potential Role of CDP in Chronic Inflammation and Melanoma-The relevance of the IUR-binding factor, CDP, is of potential interest in disorder states such as chronic inflammatory conditions and malignant melanoma, where constitutive overexpression of the CXCL1 gene contributes to disease etiology. Interactions of CDP with the IUR cis-acting element may allow for tight repression of the CXCL1 gene. The loss or displacement of CDP may be an important phenomenon in the short term induction of the CXCL1 gene, usually associated with acute inflammatory responses, or in the constitutive, high level expression of CXCL1 observed in chronic inflammatory processes, tumorigenesis, and malignant melanoma.
We have examined levels and binding profiles of CDP in two normal and five melanoma cell lines by Western blot, EMSA, and Southwestern blot analyses. We have not observed any consistent differences in the levels of the CDP protein or IURspecific binding of CDP between the cell types. We postulate that trans-repression of the CXCL1 promoter by CDP may be regulated at the level of post-translational modification of the CDP protein. CDP-mediated trans-repression is regulated by de-phosphorylation of the CDP protein, and this occurs in a cell stage-specific manner. To determine whether or not activity of this protein is altered in malignant melanoma cells, a detailed analysis of CDP-phosphorylation, IUR-specific binding, and CXCL1 trans-repression by CDP in cell-stage synchronized cells will be necessary. Insights from these investigations may uncover a relationship between CDP and progression of malignant melanoma.