Epidermal Growth Factor Induces CD44 Gene Expression through a Novel Regulatory Element in Mouse Fibroblasts*

Growth factors coordinately regulate a variety of genes associated with pathological states including tumor invasion and metastasis. Overexpressed epidermal growth factor receptor (EGFR) on tumor cell surfaces is associated with enhanced cell attachment and migration into extracellular matrices, which promotes tumor aggressiveness. We have demonstrated that epidermal growth factor (EGF) up-regulates the cell surface adhesion molecule CD44 at both the mRNA and protein levels on mouse fibroblasts expressing full-length wild-type EGFR (NR6-WT) but not on EGFR-deficient cells (NR6-P). This increases cell attachment to hyaluronic acid. In this investigation, transcriptional regulation of CD44 by EGF was confirmed by defining an EGF-regulatory element. By employing human CD44 gene promoter-chloramphenicol acetyltransferase (CAT) constructs transfected into NR6-WT cells, EGF inducibility was observed within a 120-base pair (bp) DNA fragment located 450 bp upstream of the RNA initiation site. Differential EGF inducibility was found among different cell lines chosen, indicating a 3.2- and 1.8-fold enhancement in DU145 cells carrying exogenous wild-type EGFR and in MCF-7 cells, respectively, while minimal EGF induction was found in cervical cancer HeLa cells. Utilizing gel shift assays, a time-dependent increase of DNA-protein complex formation was found upon EGF stimulation in NR6-WT cells but not in NR6-P cells. Based upon these observations, a novel 22-bp EGF regulatory element (ERE) (5′-−604CCCTCTCTCCAGCTCCTCTCCC−583-3′) was isolated from the CD44 gene promoter. This ERE conferred DNA-protein binding ability in vitro, as well as the full functional recovery of EGF inducibility of CAT activity when linked to a homologous CD44 promoter or a SV40 promoter driving a CAT reporter gene. A two-base mutation of the ERE completely eliminated its binding activity as well as its EGF inducibility of CAT expression. Our studies indicate that EGF induces CD44 gene expression through an interaction between a specific ERE and putative novel transcriptional factor so as to regulate cell attachment to extracellular matrix.


Growth factors coordinately regulate a variety of genes associated with pathological states including tumor invasion and metastasis. Overexpressed epidermal growth factor receptor (EGFR) on tumor cell surfaces is associated with enhanced cell attachment and migration into extracellular matrices, which promotes tumor aggressiveness. We have demonstrated that epidermal growth factor (EGF) up-regulates the cell surface adhesion molecule CD44 at both the mRNA and protein levels on mouse fibroblasts expressing full-length wild-type EGFR (NR6-WT) but not on EGFR-deficient cells (NR6-P). This increases cell attachment to hyaluronic acid. In this investigation, transcriptional regulation of CD44 by EGF was confirmed by defining an EGF-regulatory element. By employing human CD44 gene promoter-chloramphenicol acetyltransferase (CAT) constructs transfected into NR6-WT cells, EGF inducibility was observed within a 120-base pair (bp) DNA fragment located 450 bp upstream of the RNA initiation site. Differential EGF inducibility was found among different cell lines chosen, indicating a 3.2-and 1.8-fold enhancement in DU145 cells carrying exogenous wild-type EGFR and in MCF-7 cells, respectively, while minimal EGF induction was found in cervical cancer HeLa cells. Utilizing gel shift assays, a time-dependent increase of DNA-protein complex formation was found upon EGF stimulation in NR6-WT cells but not in NR6-P cells. Based upon these observations, a novel 22-bp EGF regulatory element (ERE)
(5-؊604 CCCTCTCTCCAGCTCCTCTCCC ؊583 -3) was isolated from the CD44 gene promoter. This ERE conferred DNA-protein binding ability in vitro, as well as the full functional recovery of EGF inducibility of CAT activity when linked to a homologous CD44 promoter or a SV40 promoter driving a CAT reporter gene. A two-base mutation of the ERE completely eliminated its binding activity as well as its EGF inducibility of CAT expression. Our studies indicate that EGF induces CD44 gene expression through an interaction between a specific ERE and putative novel transcriptional factor so as to regulate cell attachment to extracellular matrix.
Growth factors in the extracellular milieu can convey signals to cells via transmembrane surface receptors to regulate a variety of downstream events. Activation of the epidermal growth factor receptor (EGFR) 1 upon binding of its ligands (EGF, transforming growth factor-␣, HB-EGF, and amphiregulin) under both normal and pathological conditions, initiates a cellular cascade which results in major pleiotropic changes in many cell types (1,2). Aberrant overexpression of EGFR on tumor cells is associated with increased proliferation, enhanced cell attachment, and migration into extracellular matrices, progression to a transformed phenotype and in vivo aggressiveness (3)(4)(5)(6)(7). Among the regulated genes are those encoding cell surface adhesion molecules (8 -10) as well as nuclear transcriptional factors that regulate gene expression. Cell surface glycoprotein CD44, an adhesion molecule that binds to the extracellular matrix components, hyaluronic acid (11) and osteopontin (12), has been postulated to modulate tumor invasion and metastasis (13)(14)(15). Our previous studies with mouse 3T3 fibroblasts demonstrated that EGF induced CD44 gene expression on both the protein and mRNA levels and that this was accompanied by enhanced cell attachment to the extracellular matrix component hyaluronic acid (16).
Although it is known that EGF regulates gene transcription of the early responsive genes, c-myc, c-fos, and c-jun, through signaling mechanisms that are shared with the protein kinase C signaling pathway (17)(18)(19)(20)(21)(22), few EGF regulatory elements or their corresponding DNA-binding proteins have been identified. A serum-response element mediates EGF-stimulated c-fos transcription (23,24). Sp1 is required by EGF to stimulate the human gastrin gene upon binding to a 16-bp overlapping GCrich EGF response element (25), and elements in the prolactin and tyrosine hydroxylase promoter can serve as an EGF response element (ERE) (19,23,24,26). However, these sequences do not connote a specific ERE.
Structural analysis of the CD44 upstream regulatory region in human neuroblastoma cells (27) revealed that there were neither typical TATA nor CCAAT boxes in the promoter region, although approximately 150 bp adjacent to the 5Ј-end of the gene itself was sufficient to induce substantial basal transcription of the chloramphenicol acetyltransferase (CAT) gene. Despite the existence of three consensus hexanucleotide GGGCGG sequences, Sp1 binding to the CD44 gene upstream region has not been demonstrated. Rather, the cis-regulatory element, which was known to contribute to the down-regulation of CD44 in neuroblastoma cells, was identified as being a 120-bp DNA fragment located 450 bp upstream of the RNA initiation site. To gain further insight into the molecular mechanism by which EGF regulates CD44 gene expression, we investigated transcriptional regulation of CD44 gene by EGF using the bacterial CAT reporter system. Both in vitro DNA-protein binding ability and in vivo EGF inducibility were examined in both wild-type as well as mutational EREs. Herein, we report the existence of a transferable 22-nucleotide sequence which confers EGF responsiveness to transcriptional units.

EXPERIMENTAL PROCEDURES
Cell Culture and EGF Induction-Parental murine NR6 cells (P), which are 3T3 derivatives devoid of endogenous EGFR (28), were stably transduced via retrovirus-mediated infection with holo (wild-type) EGFR (29). These cell lines have been previously characterized, and are maintained at 37°C, in a 5% CO 2 environment, in minimal essential medium-␣ (Life Technologies, Inc.) supplemented with 7.5% fetal bovine serum, 100 g/ml penicillin and streptomycin, and 1 ϫ non-essential amino acid and sodium pyruvate with 350 g/ml G418 as a selective marker for the NR6-WT cells. Cells were carefully passaged before they reached 90% confluence to eliminate spontaneous morphological change. EGF induction was accomplished with 10 nM EGF (Life Technologies, Inc.) in complete medium containing 0.1% dialyzed fetal bovine serum for different time periods. Other cell lines used were: DU145 carrying exogenous wild-type EGFR (31) in Dulbecco's modified Eagle's medium/F-12 media supplemented with 100 g/ml penicillin and streptomycin, 7.5% fetal bovine serum, and 1 mg/ml G418 as a selective marker; MCF-7 and HeLa cells were maintained according to recommendations from the ATCC.
Analysis of Newly Synthesized CD44 mRNA in Nuclear Run-on Assays-Preparation of nuclei and RNA labeling followed standard protocols (30). Briefly, 1 ϫ 10 8 NR6-P or NR6-WT cells were treated with 10 nM EGF for 0, 1, 3, 6, or 12 h, washed with ice-cold phosphate-buffered saline, and preincubated in cold lysis buffer (10 mM Tris-Cl (pH 7.4), 3 mM CaCl 2 , and 2 mM MgCl 2 ). The cells were resuspended in 2 ml of lysis buffer followed by further lysis in 2 ml of Nonidet P-40 buffer (addition of 1% Nonidet P-40 to the lysis buffer) before disruption by 10 strokes in a cold Dounce homogenizer. Disruption was confirmed by light microscopic evaluation after trypan blue staining. Nuclei were pelleted by centrifugation at 500 ϫ g, at 4°C for 5 min, and then resuspended in ice-cold glycerol storage buffer (50 mM Tris-Cl (pH 8.3), 40% glycerol, 5 mM MgCl 2 , and 0.1 mM EDTA) followed by immediate return to dry ice before being stored at Ϫ70°C for later use.
Elongation of nascent RNA from 5 ϫ 10 7 nuclei was carried out for 30 min at 30°C in 200 l of 2 ϫ reaction buffer (10 mM Tris-Cl (pH 8.0), 5 mM MgCl 2 , 0.3 M KCl, 3 mM dithiothreitol, 35 units of RNasin, and 1 mM each of CTP, ATP, and GTP) in the presence of 10 l of [␣-32 P]UTP (760 Ci/mmol, 10 mCi/ml, Amersham). After treatment with 24 g of DNase I (Sigma) for 15 min at 30°C and 10 g of proteinase K for 30 min at 42°C, labeled RNA was extracted with phenol/chloroform and precipitated with 5% trichloroacetic acid. Precipitants were then applied to Whatman GF/A glass fibers which were further washed 3 times with 10 ml of 5% trichloroacetic acid, 30 mM sodium pyrophosphate and incubated in 1.5 ml of reaction buffer (20 mM HEPES (pH 7.5), 5 mM MgCl 2 , 1 mM CaCl 2 ) in the presence of 40 g of DNase I for 30 min at 37°C. RNA was eluted by heating the mixture to 65°C for 10 min after adding 45 l of 0.5 M EDTA and 68 l of 20% SDS and then treating with 10 g of proteinase K at 37°C for 30 min, before precipitating with 0.1 volume of 3 M sodium acetate and 2.5 volumes of ethanol. Aliquots of ϳ4 -5 ϫ 10 6 cpm of labeled RNA in 1 ml of TES solution (10 mM TES (pH 7.4), 10 mM EDTA, and 0.2% SDS) were hybridized to nitrocellulose membranes previously dot-blotted with 5 g of denatured and linealized plasmid DNA containing CD44 cDNA, ␣-actinin cDNA, or vector DNA in the presence of 1 ml of TES/NaCl solution (0.6 M NaCl ϩ TES solution) at 65°C for 36 h. After washing twice in 25 ml of 2 ϫ SSC for 1 h at 65°C followed by 8 g of RNase A treatment at 37°C for 30 min, nitrocellulose strips were then unraveled and exposed to x-ray film.
Plasmid Constructs, DNA Transfection, and CAT Assays-Plasmids containing different fragments of the human CD44 gene regulatory region placed upstream of the bacterial CAT gene (pCATbasic, Stratagene) (27) were a kind gift from Dr. Emma Shtivelman (SYSTEMIX, CA). Specifically, these plasmids were: pRBCAT with 1.9 kilobases of genomic CD44 sequences extending upstream from the BamHI site in the first exon; pSacBCAT and pRVBCAT were 5Ј truncated versions of pRBCAT, with 0.65-and 0.53-kilobase inserts upstream of the CAT gene, respectively; and pRB-SRCAT which was a deletion version of the pRBCAT sequence in which a 120-bp fragment between the SacI and EcoRV sites were excluded (illustrated in Fig. 2).
To further define the putative ERE, three repeats (22 bp, 5Ј-CCCTCTCTCCAGCTCCTCTCCC-3Ј) were amplified by polymerase chain reaction methods from a synthesized oligonucleotide template (5Ј-ACCCAAGCTTCGCCCTCTCTCCAGCTCCTCTCCCGCATGCGTC-GACCGCCCTCTCTCCAGCTCCTCTCCCGCTAGCCGCCTTCTCTCC-AGCTCCTCTCCCGTCGACGAT-3Ј) using the 5Ј primer (5Ј-ACCCAA-GCTTCGCCCT-3Ј) and the 3Ј primer (5Ј-ATCGTCGACGGGAGAG-3Ј), and placed upstream of the pRVBCAT plasmid as p3ERE-RVBCAT. p3ERE-RVBCAT then underwent internal restriction enzyme digestion with SphI and SalI followed by religation to establish a p2ERE-RVBCAT and p1ERE-RVBCAT which contained 2 or 1 ERE repeats, respectively. A heterologous SV40-driven CAT reporter system was also utilized to verify the ERE function. Synthesized oligonucleotides of the ERE as 5Ј-pGATCTAAGCTTGGGAGAGGAGCTGGAGAGAGGGCGA-TATCA-3Ј and 5Ј-p-GATCTGATATCGCCCTCTCTCCAGCTCCTCTC-CCAAGCTTA-3Ј were annealed and ligated into a BglII site in a pPromoterCAT vector (Promega) to obtain both a sense oriented (pERE(s)SV 40 CAT) and antisense oriented (pERE(as)SV 40 CAT) CAT construct. Two-base mutations of the ERE (from T to C) synthesized from oligonucleotides as 5Ј-CCCTCTCCCCAGCTCCCCTCCC-3Ј and its compliment, were placed in front of an SV 40 promoter of the pPromot-erCAT vector at a Klenow-blunted BglII site to establish a heterologous ERE mutation construct (pERE(Mu)SV 40 CAT), after polymerase chain reaction screening and DNA sequencing confirmation.
DNA transfection was performed on 1 ϫ 10 7 of 80% confluent NR6-P and NR6-WT cells using equimolar amounts of pCAT constructs corresponding to 20 g of pCAT-basic DNA (Promega) by an electroporation method (Bio-Rad) under constant conditions (400 V, 500 microfarads capacity, and 0 resistance). Plasmids containing a ␤-galactosidase gene driven by an SV40 promoter were used to normalize the transfection efficiency for basal levels of CAT expression. In the EGF induction CAT assays, transfected cells were equally divided, allowed to recover in complete medium for 24 h, and then treated with or without 10 nM EGF for an additional 24 h in 0.1% dialyzed fetal bovine serum containing medium. CAT activity was measured from cell extracts by the CAT Enzyme Assay System (Promega) based on liquid scintillation counting using 14 C-labeled chloramphenicol (Amersham) and n-butynyl coenzyme A. The level of CAT activity from pRVBCAT transfectants without EGF induction was set at 100% and the percent (fold) increase in the other transfectants determined and compared with the standard. Five independent experiments, each in duplicate, were performed for each plasmid transfection. Statistical significance was determined by use of the Student's t test.
Nuclear Protein Extraction-5 ϫ 10 7 cells were lysed, and nuclei collected as above, following 10 nM EGF induction of both NR6-P and NR6-WT cells for 0, 1, 3, 6, or 12 h. Disruption of the nuclear membranes was accomplished by sequentially adding 1/2 packed nuclear volume of a low salt buffer (20 mM HEPES (pH 7.9), 25% glycerol, 1.5 mM MgCl 2 , 60 mM KCl, 0.2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol) and then an equal volume of a (1/2 packed nuclear volume) high salt buffer (1.2 M KCl as 60 mM in low salt buffer). The solution was gently mixed for 30 min before centrifugation at 14,500 rpm for 30 min. Supernatants were tube dialyzed (6,000 -8,000 M r cut-off) against a buffer containing 20 mM HEPES (pH 7.9), 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol at 4°C for 2 h followed by centrifugation to remove precipitated protein and nucleic acids. Supernatants containing nuclear extracts were further purified by heparin-affinity chromatography (Bio-Rad) and eluted from a concentration gradient of 0.1, 0.2, 0.4, 0.6, 0.8, 1.0, and 1.5 M KCl in TM buffer (25 mM Tris-Cl (pH 8.0), 6 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, and 15% glycerol). Aliquots that contained the most significant binding to the DNA probe were stored at Ϫ70°C and utilized in the subsequent gel shift and DNA footprinting assays. The final concentration of extracts was estimated by both a Bio-Rad protein assay and quantitative gel shift assay.
DNA-Protein Binding in a Gel Shift Assay-HindIII-digested pSacB-CAT plasmid was Klenow end-labeled with [␣-32 P]dCTP before a second digestion with EcoRV. The released 120-bp labeled DNA fragment was fractionated and purified on agarose gel. Complementary synthetic oligonucleotides of putative ERE sequence ( Fig. 6) were annealed before T4 kinase 5Ј-end-labeling with [␥-32 P]ATP. 2 ϫ 10 4 cpm of each probe was applied to 20 l of premixed gel shift reaction containing 25 mM Tris-Cl (pH 8.0), 6.25 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 10% glycerol, and 40 ng of poly(dI⅐dC) in the presence of 4 g of nuclear extract at 25°C for 1 h. Poly(dI⅐dC) at 0.2 or 1 g, unlabeled 120-bp DNA probe at 1 or 100 ng, and unlabeled pRVBCAT insert DNA at 1 or 100 ng were preincubated with the nuclear extracts at 25°C for 10 min before being placed in reaction with the labeled probe which served as the competition control. Annealed synthetic oligonucleotides were 5Јend-labeled with ␥-32 P by T4 kinase before being used in gel shift reactions. DNA-protein complexes were resolved on a native 4% non-denatured polyacrylamide gel containing 45 mM Tris base, 45 mM boric acid, and 2.5% glycerol under nondenaturing conditions. After exposure to x-ray film, gel shift bands were quantified by densitometry (Image 1.47 software).
DNase I Protection Footprinting Assays-DNase I protection assays were performed in 50 l of reaction with 0.5, 2, or 4 g of nuclear extract from NR6-WT cells treated with or without 10 nM EGF for 6 h and 2 ϫ 10 4 cpm of the 120-bp [␣-32 P]dCTP-labeled DNA probe as described in the gel shift assay. The nuclear extract was preincubated without probe for 10 min on ice in the same buffer as used in the gel shift assay, and incubated for another 30 min on ice after the addition of probe. Digestion was initiated with the addition of 1 unit of DNase I (Promega) in 25 mM CaCl 2 at room temperature and terminated after 1 min with 90 l of stop solution (200 mM NaCl, 30 mM EDTA, 1% SDS, and 100 g/ml yeast RNA) prior to phenol-chloroform extraction and ethanol precipitation. Digested DNA and sequencing reactions of the probe using sequencing primer, 5Ј-AGCTTCTCCCTCTTTCCAC-3Ј carried out by the Sequenase version 2.0 (U. S. Biochemical Corp.), was resolved on a 5% polyacrylamide, 7 M urea gel followed by autoradiography.

RESULTS
EGF Up-regulates CD44 Gene Transcription-Our previous data indicated that EGF up-regulated cell attachment of NR6 mouse fibroblastic cells to hyaluronic acid by increasing the expression of the cell surface adhesion molecule CD44 at both the protein and mRNA levels (16). EGF treatment of NR6 cells expressing wild-type EGFR resulted in a 3-4-fold up-regulation of CD44 protein expression with a maximum at 24 h, as well as a 3-fold increase of mRNA levels after 6 -24 h of EGF treatment (16). The CD44 mRNA in NR6 cells were determined by nuclear run-on analysis to confirm the transcriptional regulatory mechanism (Fig. 1). Both NR6 parental cells without endogenous EGFR (ϪP) and NR6 cells with transfected EGFR (ϪWT) were preincubated with 10 nM EGF for various times before collecting and labeling new transcripts. In agreement with earlier Northern blot analysis, there was no perceptible change in the rate of CD44 transcription with NR6-P cells, while a 3-fold induction was observed maximally at 6 h in NR6-WT cells in the presence of EGF. This was seen to decline by a third at 12 h, identical to the transient wave of mRNA seen on Northern blot analysis (data not shown). These results support the contention that EGF regulates CD44 mRNA levels by increasing transcription in a time-dependent manner at least during short-term EGF exposure.
Functional Characterization of the EGF Regulatory Element (ERE) in the CD44 Gene-To identify the sequence that mediates EGF responsiveness in the CD44 gene, a series of DNA fragments containing the sequence of the human CD44 gene upstream regulatory region (pRBCAT) and restriction enzy-matic internal truncations (pSacBCAT and pRVBCAT) and deletions of the human CD44 gene upstream region (pRB-SRCAT) were obtained and placed into a bacterial CAT reporter gene ( Fig. 2A). CAT activities were measured on electroporated transfected NR6-WT cells in the presence or absence of EGF. Consistent with the previous finding in human neuroblastoma cells (26), all constructs that contained the 150-bp sequence upstream of the transcription initiation site conferred substantial basal CAT activity on the constructs (Fig.  2B) when compared with the promoterless CAT plasmid.
The observation that not only did pSacBCAT elicit a 2-fold higher basal CAT activity than pRVBCAT or pRB-SRCAT, both of which were lacking a 120-bp fragment between SacI and EcoRV sites (Sac-RV fragment), but that the same order of magnitude change was seen in comparison to pRBCAT, which contains this fragment, suggested that a negative cis-element upstream of the SacI site and a stimulatory cis-element on the Sac-RV fragment may contribute to the basal promoter activity. Moreover, as there was an increase in EGF-mediated induction of CAT activity in pRBCAT and pSacBCAT (2.5-and 2.3-fold, respectively) (Fig. 2C), the only construct with a Sac-RV fragment sequence, this was strong presumptive evidence that an ERE was present in this region.
Differential EGF Induction of CAT Gene Transcription between Different Cells-To test whether the observed relatively low EGF induction in NR6 cells (2.3-fold) was due to low promoter activity or is the result of a cell-specific effect, three transformed human cell lines, which express EGFR and CD44 and are derived from prostate, DU145-WT, breast, MCF-7, or cervix HeLa, were employed in our transfection/CAT system (Fig. 3). All cells were transfected by pCAT, pRVBCAT, and pSacBCAT constructs via standard electroporation methods. Basal level of CAT activity was determined by cotransfection of ␤-galactosidase and was shown to be similar among these cells (data not shown). After EGF stimulation for 24 h following transfection, changes in CAT activity was determined by comparing EGF-treated cells with their non-EGF-treated counterparts. In agreement with the data from NR6 cells, no significant CAT induction was found following mock (pCAT basic vector) and pRVBCAT transfections in all cell lines tested; however, a differentiated CAT induction by EGF stimulation was demonstrated in these cells, being 2. tions of the assay. Moreover, additional experiments demonstrated that CAT induction of NR6-WT cells began at a low dose of 5 nM EGF and was sustained at an EGF concentration of up to 20 nM (data not shown), indicating no EGF dose-dependent bidirectional effect.
A EGF Inducible Nuclear Protein Binds to the Sac-RV Fragment of the CD44 Promoter-The presence of an EGF inducible DNA binding activity was sought by an electrophoretic mobility shift assay using the 32 P-labeled Sac-RV DNA fragment (which contained the putative ERE described in the functional analysis), and nuclear extracts from NR6-P and -WT cells. Pilot experiments indicated that nuclear extracts eluted from heparin-affinity columns with 0.4 M KCl buffer contained the DNAprotein complexes of interest. Binding specificity was assessed by examining the effect of excess amounts of poly(dI⅐dC) as nonspecific DNA, unlabeled Sac-RV fragment as specific competitor, and probe containing the CD44 promoter sequence from the pRVBCAT plasmid as random nonspecific DNA (Fig.  4). It was noted that three bands (labeled 2, 3, and 4) were eliminated by preincubating the labeled probe with 0.5 g of poly(dI⅐dC), while the major band (labeled 1) retained significant binding, even in the presence of excess poly(dI⅐dC) (Fig.  4A). Furthermore, the major band could be competed only with increasing amounts of specific unlabeled Sac-RV probe (10 -100 ng, equivalent to 20 -200-fold excess of the labeled probe) (Fig.  4B) but not to excessive nonspecific probe derived from pRVB-CAT (Fig. 4C).
The time course of EGF induction of this ERE response was assessed using nuclear extracts from both NR6-P and -WT cells, treated by EGF for 1, 3, 6, or 12 h (Fig. 5). Extracts were normalized to cell number with the loading amount ranging from 3.8 to 4.0 g; 0.5 g of poly(dI⅐dC) (500-fold excess) was also included in the reaction mixture to minimize nonspecific FIG. 4. Specific DNA-protein complex formation on a gel shift assay. A 120-bp DNA fragment between the S and RV sites of the CD44 regulatory region (Sac-RV fragment) was 5Ј-end-labeled with [␣-32 P]dCTP and incubated with a nuclear extract from NR6-WT cells. The specificity of DNA-protein binding was tested by using different competitors, as indicated, prior to the binding of the nuclear extract with labeled probe: A, unlabeled poly(dI⅐dC) as a nonspecific competitor; B, the unlabeled Sac-RV fragment as a specific competitor; C, an unlabeled RV fragment (DNA fragment of the CD44 regulatory region from the pRVB CAT construct) as a random nonspecific competitor. DNAprotein complexes were resolved on a native 4% nondenatured polyacrylamide gel and exposed to x-ray film.

FIG. 2. EGF induction of CD44 promoter using a transient transfection (CAT) assay.
Truncation and deletion versions of the CD44 upstream regulatory region was linked to a CAT reporter gene as shown (A): pRBCAT, CD44 gene regulatory region from the first exon to the RI (EcoR I) site; pSacBCAT, pRBCAT truncation to the S (SacI) site; pRVB, pSacBCAT truncation to the RV (EcoRV) site; pRB-SR deletion version of pRBCAT between the S and RV sites. pCAT, without the exogenous CD44 regulatory region. The RNA initiation site is indicated by the arrow. Corresponding basal levels of CAT activity were measured as the percent of CAT activity from the pRVB construct and normalized utilizing cotransfected ␤-galactosidase gene activity (B). EGF inducibility was expressed as the "fold" increase over the basal CAT activity for each construct (C). Data was from five separate experiments performed in duplicate and statistical significance was validated by use of Student's t test (*, p Ͻ 0.01).

FIG. 3. Differential EGF induction of CAT among different cells.
20 g of CAT construct DNA for pCATbasic, pRVBCAT, and pSacBCAT were individually transfected into 1 ϫ 10 7 cells: NR6-WT, DU145-WT, MCF-7, and HeLa cells by electroporation method. Cells were divided into two and incubated in complete medium for 24 h before cells were treated with or without 10 nM EGF for 24 h. Fold increase of CAT activities were measured from EGF-treated cells compared with untreated cells. Statistic analysis was from five independent experiments of duplicates (*, p Ͻ 0.01).
binding. No change was observed in protein-DNA complex formation upon EGF stimulation in the NR6-P cells. However, in NR6-WT cells, the induction became evident (2-fold increase, n ϭ 3), 3-6 h after exposure to EGF, and persisted for at least 12 h (Fig. 5B). It was noted also that this time-dependent induction of protein-DNA complex formation paralleled the mRNA wave demonstrated in both Northern blot and nuclear run-on assays, suggesting that the induction response required functional EGFR expressed on cell surfaces and that the Sac-RV fragment contained an ERE.
Identification of an ERE in the CD44 Gene Promoter-This work strengthened the hypothesis that EGF up-regulates CD44 gene expression through induction of transcriptional factor(s) binding to an ERE in the gene promoter region. To determine the specific sequence to which the transcriptional factor bound, we performed DNase I footprinting by using a nuclear extract from NR6-WT cells induced by EGF for 6 h and [␣-32 P]dCTP 5Ј-end (Klenow) labeled Sac-RV as the DNA probe. Two "footprints" were observed when 0.5 g of poly(dI⅐dC), as nonspecific DNA competitor, was included in the binding reaction: footprint I over the sequence Ϫ5Ј-CTC-CTCTCCC Ϫ583 and footprint II over the sequence -Ϫ604 CCCTCTCTCC Ϫ595 (Fig. 6). Both sites were better protected from DNase I digestion when an increasing amount of nuclear extract (0.5, 2.0, and 4.0 g in lanes 2, 3, and 4, respectively, from non-EGF treated cells; lanes 5, 6, and 7, respectively, from EGF-treated cells) was added. The specificity of the protein binding was also demonstrated by competition with excess poly(dI⅐dC), as nonspecific DNA, which abolished a putative third footprint and with unlabeled Sac-RV fragment, as specific probe, which competed away all footprints while the RVB fragment, acting as an unlabeled random nonspecific competitor, did not affect the DNase I protection (data not shown). It should be noted that both footprints are partially palindromic in themselves and are inverse repeats of each other, and that the 2-base nucleotides between them were partially protected. A sequence search in GenBank against known transcriptional factors did not reveal any significant homology to the two separate footprints, either alone or together. Therefore, this sequence from Ϫ583 to Ϫ604 on the CD44 gene upstream regulatory region appears to be a novel EGF-regulatory element in mouse fibroblast NR6 cells.
EGF-induced Nuclear Protein Specifically Binds to the ERE-The sequence-specific binding of the NR6 cell nuclear extract was further studied in a gel shift assay by using the 32 P-5Ј-end-labeled Sac-RV fragment and an unlabeled 22-bp double-strand synthetic oligonucleotide as competitor: 5Ј-CCCTCTCTCCAGCTCCTCTCCC-. The protein-DNA complex formation recognized as the EGF-inducible nuclear protein was noted to decrease as an increasing amount of unlabeled syn-thetic oligonucleotides (1-10 pmol, equivalent to 10-and 100fold excess of labeled probe, respectively) was applied to the gel shift reaction mixture (Fig. 7B). This suggested that this synthetic sequence could competitively bind the same EGF-induced protein.
To confirm the binding specificity of this ERE, the Sac-RV fragment (probe 1), the 22-bp oligonucleotide sequence (probe 2), and two oligonucleotides containing either the 5Ј or 3Ј half of the palindromic structure (5Ј-CCCTCTCTCCAG-and 5Ј-CAGCTCCTCTCCC-) (probes 3 and 4, respectively) were 5Јend-labeled with [␥-32 P]ATP followed by incubation with the NR6-WT nuclear extract and a gel shift assay was performed. As shown in Fig. 7A, only the 22-bp ERE (probe 2), but not the half-sequences (probe 3 and 4), was capable of binding the band FIG. 6. Specific EGF response element binding to nuclear protein in a DNase I footprint assay. Nuclear extracts from NR6-WT cells treated with or without 10 nM EGF for 6 h were incubated with 20,000 cpm of ␣-32 P-labeled SacB fragment DNA probe before DNase I digestion for 1 min. After phenol-chloroform extraction, the reaction mixture and the comparable sequencing reaction were fractionated on a 5% sequencing gel. A, increasing amounts, 0, 0.5, 2, and 4 g, of nuclear extract from NR6-WT cells without EGF treatment were loaded onto lanes 1-4, respectively. Increasing amounts of nuclear extract from NR6-WT cells with EGF treatment, 0.5, 2, and 4 g, in lanes 5-7, respectively, were similarly loaded. DNase I protection sequences (I and II region) were read from parallel sequencing reactions shown in B. B, the DNase I footprint assay in the left panel. Lanes 1 and 2, DNA probe only. Lanes 3 and 4, DNA probe with 4 g of nuclear extract from non-EGF treated and EGF treated NR6-WT cells, respectively. The specific DNA binding sequencing was indicated in the highlighted box as seen in the right panel.

FIG. 5. EGF induced DNA-protein complex formation on a gel shift assay.
Nuclear extracts from NR6 parental and WT cells treated with 10 nM EGF for 0, 1, 3, 6, or 12 h were incubated with a [␣-32 P]dCTP-end-labeled Sac-RV fragment in the presence of 0.5 g of poly(dI⅐dC). DNA-protein complex formation was resolved on a native 4% nondenatured polyacrylamide gel (A). Quantitation was performed by densitometry scanning of the specific bands (B).
compatible with the Sac-RV probe, indicating that EGF-induced protein-DNA binding requires a full-length sequence.
The time-dependent EGF responsiveness was further demonstrated by using nuclear extracts from EGF-treated NR6-WT cells and labeled ERE (Fig. 7C). EGF inducibility was evident from 3 to 6 h after EGF stimulation, similar to the findings using the labeled Sac-RV probe. Together, these results further strengthen the conclusion that the 22-bp sequence is an EGFregulatory element which regulates CD44 gene expression.
ERE Is Necessary for EGF-induced Transcription-We demonstrated that the ERE sequence within the CD44 gene promoter is critical for both EGF-induced DNA-protein complex formation and EGF-induced transcription of the CAT gene. The identity of the ERE in the context of EGF induction of CD44 gene transcription was further assessed in a homologous and a heterologous CAT reporter system. Multiple EREs were placed in front of the CD44 gene promoter in the pRVBCAT construct, which lacked EGF inducibility of CAT gene transcription as described above. They were p1ERE.RVBCAT, p2ER-E.RVBCAT, and p3ERE.RVBCAT containing 1, 2, and 3 ERE, respectively. pRVBCAT and pSacBCAT served as the negative and positive controls. After transfection of NR6-WT cells with the various constructs, half the cell groups were stimulated with 10 nM EGF for 24 h and compared in the CAT assay to the untreated matched control for increase of CAT activity secondary to EGF induction. All ERE-containing constructs elicit similar significant induction of CAT activity by EGF (2.8-, 2.4-, and 2.5-fold for p1ERE.RVBCAT, p2ERE.RVBCAT, and p3ERE.RVBCAT, respectively), similar to the positive control (2.3) fold (Fig. 8A). The fact that p1ERE.RVBCAT yielded at least equal EGF induction may imply that accessibility of the ERE because of its relative shorter sequence may be important in this process, although other mechanisms may also account for the observed changes.
To rule out the possibility that the CAT induction by EGF was the result of transcription initiated from the transcription start site of the transfected CD44-CAT reporter gene, the ERE sequence was introduced into a heterologous SV 40 promoterdriven CAT reporter system in both a sense and antisense orientation (pERE(s)SV 40 CAT and pERE(as)SV 40 CAT, respectively). Both constructs and the pSV 40 CAT vector control conferred high constitutive basal level of CAT activity (data not shown). However, only the sense ERE sequence was able to fully restore the EGF induction of CAT transcription in NR6-WT cells (2.3-fold increase), while minimal (1.2-fold) to no induction was observed in pERE(as)SV 40 CAT and pSV 40 CAT vector, respectively.
The Integrity of the 22-bp ERE of CD44 Promoter Is Necessary for Full Transcriptional Induction-EGF-induced CD44 transcription is initiated by a specific nuclear protein binding to the ERE sequence of the CD44 gene promoter. We performed an experiment to test the specificity of the ERE sequence in the EGF induction event by point mutating two "T" nucleotides to two "C" (Fig. 9C), to disturb the potential binding motif (palindromic structure). The binding activity to the nuclear protein was measured in a gel shift assay. EGF inducibility of CAT transcription was also verified in a pSV 40 CAT reporter system by placing the mutated ERE in front of the SV 40 promoter. In this analysis, the specific DNA-complex formation was observed to be enhanced when an unmutated ERE (WT) and an increasing amount of nuclear extract from EGF-treated NR6-WT were placed in the gel shift reaction mixture. Minor bands were, however, noted above the specific bands and were thought to be the result of nonspecific binding since no increasing binding effect was found when increasing amounts of nuclear extract was applied. Poly(dI⅐dC), a nonspecific competitor, was able to eliminate these bands without affecting the major specific DNA-protein complex formation. In contrast, no significant DNA-protein complex formation was detected using the mutated ERE even in the presence of an excess amount of FIG. 7. EGF-induced protein-ERE complex formation in a gel shift assay. The DNA-protein complex formation was measured by using a 32 P-labeled double SacB strand fragment (probe 1), 22-bp ERE (probe 2), 12-bp first half of ERE (probe 3), and 12-bp second half of ERE (probe 4), and by using nuclear extracts from NR6-WT cells as indicated. A, DNA-protein complex formation of each probe, with or without nuclear extracts, from the NR6-WT cells as indicated. B, ERE (probe 2) competition in protein-DNA complex formation. 1 and 10 pmol of annealed double-strand ERE were incubated with a nuclear extract from NR6-WT cells before adding labeled SacB fragment. C, EGF induced protein-ERE binding in the gel shift assay. Nuclear extracts from NR6-WT cells, with or without EGF treatment, for different time periods as indicated were incubated with labeled double probe 1 or probe 2 in a gel shift reaction. Lanes 1 and 2, probe 1 only, probe 1 with nuclear extract, respectively. Lane 3, probe 2 only. Lanes 4 -8, probe 1 was incubated with nuclear extracts from NR6-WT cells treated with 10 mM EGF for 0, 1, 3, 6, and 12 h, respectively.
FIG. 8. Functional analysis of ERE in homologous and heterologous CAT system. NR6-WT cells were transfected with various CAT constructs containing ERE as described before. Fold increase was measured as EGF induction from EGF-treated cells compared with untreated cells. A, homologous ERE constructs in EGF-irresponsive pRVBCAT vector containing 1, 2, and 3 ERE as p1ERE.RVBCAT, p2ERE.RVBCAT, and 3ERE.RVBCAT, respectively, as well as pCAT, pRVBCAT, pSacBCAT controls, were transfected into NR6-WT cells for EGF induction of CAT activity. B, ERE constructs in heterologous SV 40 -driven CAT vector (pSV 40 CAT) with both sense orientation (pERE(s)SV40CAT) and antisense orientation (pERE(as)SV 40 CAT), were transfected into NR6-WT cells for EGF induction of CAT activity. Data from four independent experiments of duplicates (*, p Ͻ 0.01).
nuclear extracts (10 g) (Fig. 9A). More importantly, transfection of the wild-type ERE (WT)-SV 40 CAT reporter resulted in a full restoration of EGF induction of CAT activity (2.3-fold increase) in NR6-WT cells, while the same reporter with a mutated ERE (Mu) devoid of DNA-protein binding capacity exhibited no induction (1.0-fold) (Fig. 9B). Although inducible activity was relatively low in this experiment, the results clearly establish the necessity for the integrity of this ERE sequence for EGF induction. DISCUSSION EGF is known to regulate many different gene transcripts and thereby modulate cell adhesion, attachment, migration, and differentiation, under both physiologic and pathologic conditions (31). We have earlier shown that EGF up-regulates the cell surface adhesion molecule CD44 protein and its mRNA level to enhance mouse fibroblast NR6 cell attachment to the extracellular matrices (16). In the present study, using different techniques, we have confirmed that EGF induces CD44 gene transcription and that this induction is mediated by a novel EGF regulatory element found in the upstream regulatory region of the CD44 gene.
The ERE sequence differs from previously reported EGF response elements found in genes encoding c-myc, c-fos, c-jun, and other genes whose protein products include gastrin (31), prolactin, tyrosine hydroxylase (19,23,25,26), transin (20), and pS2 (21). These studies have shown that EGF stimulates transcription of genes through different cis-regulatory sequences. The ERE of c-fos is mediated by a serum response element (32) and the Sp1 site (GGGCGG) modulates the gastrin gene, whereas EGF induction of tyrosine hydroxylase and prolactin is mediated by the nearly identical sequence GGAA-GAGGATGCC (19,23). Although an ERE sequence containing the sequence GGGCGG has been identified as a high affinity Sp1 site (33), Sp1 is unlikely to mediate CD44 transcriptional activation by EGF in our model system because the pRVBCAT plasmid, which contains this sequence, did not show EGF induction. Moreover, gel shift assays using labeled probe containing the Sp1 sequence, released from the pRVBCAT construct, did not demonstrate significant binding to nuclear extracts of NR6 cells (data not shown). As noted above, to exclude other possible transcriptional factors which may bind to this ERE, we searched data bank files and confirmed that the CD44 gene's ERE is not similar to other described regulatory elements.
EGF induction of CAT activity in our ERE sequence is relatively low (ϳ2.3-fold). This most likely is due to low intrinsic promoter activity, similar to that found in other gene promoters, i.e. the ␤-fibroblast growth factor regulatory element of the skeletal ␣-actin promoter (34). This also suggests that EGF may modulate cell attachment to the extracellular matrices via the coordination of multiple cell surface adhesion molecules (8 -10) rather than exclusively by CD44. However, the possibility of cell-specific differential induction could not be completely ruled out because in our transfection system a 2.3-, 3.2-, and 1.8-fold induction of CAT activity was detected in mouse fibroblasts, human prostate cancer, and human breast cancer cells, respectively.
Unlike other cis-acting EGF regulatory elements that may also mediate transcriptional activation by phorbol esters (35)(36)(37), we have preliminary data that other growth factors including transforming growth factor-␤ and basic fibroblast growth factor, but not phorbol 12-myristate 13-acetate, induced CD44 protein levels in NR6 cells. This would suggest that EGF stimulates CD44 transcription through a distinct receptor signaling event other than the protein kinase C signaling pathway. Whether these growth factors activate the same transcription element or different factors which bind to unrelated cis-regulatory sequences needs to be addressed. Like the serum response element-mediated c-fos transcription, our ERE binds to nuclear proteins from EGFR-devoid NR6-P cells and unstimulated NR6-WT cells so as to activate basal transcription in the context of the CD44 promoter when linked to CAT reporter gene. The specific DNA-protein complex formation derived from NR6-WT cells increased after EGF stimulation in a timedependent manner. This implies that stimulation of CD44 gene transcription by EGF results from either an increased expression or increased activation of this cis-acting DNA-binding protein.
We noted that the 22-bp ERE sequence contains the palindrome of 5Ј-CCCTCTC-and -CTCTCCC-3Ј. Although the significance of this is currently unknown, it might serve as a DNA recognition site for either multimeric proteins composed of identical subunits or is needed to form hairpin loops to generate structures that might be differentially recognized by specific DNA-binding proteins. As demonstrated in the gel shift assay, each half of the ERE sequence was unable to form a stable DNA-protein complex. It is most likely that binding of nuclear protein to each CCCTCTC sequence of the CD44 ERE is required for protein-DNA recognition. Cloning of the gene encoding the ERE-binding protein will permit further characterization of its structure and the process by which EGF exerts a stimulatory effect on CD44 gene expression.
In summary, we conclude that EGF transcriptionally upregulates CD44 gene expression. The 22-bp ERE from the CD44 gene promoter region is orientation-specific, and is necessary to enable EGF-induced CD44 gene transcription through a specific interaction between the ERE motif and a putative novel transcriptional factor. Acknowledgments-We thank Dr. Emma Shtivelman (SYSTEMIX, CA) for the kind gift of the cloned CD44 promoter constructs. Further-FIG. 9. Binding and EGF inducibility of mutated ERE. Wildtype ERE sequence (ERE(WT)) was mutated by two base pair changes from "T" to "C" (ERE(Mu)) (C), and both were inserted into a heterologous CAT construct pSV 40 CAT in sense orientation as pERE(s)SV 40 CAT and pERE(Mu)SV 40 CAT, respectively. A, protein-DNA complex formation from ERE(WT) and ERE(Mu) in gel shift assay. 20,000 cpm of labeled oligos were incubated with increasing amounts of nuclear extracts (0, 1, 2, and 10 g) from EGF-treated NR6 cells and fractionated on a native 4% nondenatured polyacrylamide gel as described. B, NR6-WT cells were transfected with ϳ20 g of constructs pSV 40 CAT, pERE(s)SV 40 CAT, and pERE(Mu)SV 40 CAT and were treated with or without 10 nM EGF for 24 h. Fold increase of CAT activity was measured as described before as EGF inducibility. Data was from four independent experiments (*, p Ͻ 0.01). more, we thank Victor Krasnykh and Dr. David Curiel, Gene Therapy Program, University of Alabama at Birmingham, for assistance in the oligonucleotide synthesis.