Co-operative Effect of c-Src Tyrosine Kinase and Stat3 in Activation of Hepatocyte Growth Factor Expression in Mammary Carcinoma Cells*

We have previously shown coexpression of hepatocyte growth factor (HGF) and its receptor Met in the invasive tumor front of human breast carcinomas. We have also demonstrated secretion of HGF, constitutive activation of Met, and increased invasion in a murine breast carcinoma cell line, SP1. These observations suggest the presence of an HGF autocrine loop in some breast carcinoma cells, which confers increased survival, growth, and invasiveness during tumor progression and metastasis. c-Src tyrosine kinase, which is critical in regulating the expression of many genes, is activated in SP1 carcinoma cells, as well as in most human breast cancers. We therefore examined the role of c-Src kinase in HGF expression in breast carcinoma cells. Expression of activated c-Src in SP1 cells increased transcription from the HGF promoter and expression of HGF mRNA and protein, while dominant negative c-Src had the opposite effect. Using deletion analysis, we showed that the region between −254 and −70 base pairs was required for c-Src responsiveness of the HGF promoter. This region contains two putative consensus sequences (at −110 and −149 base pairs) for the Stat3 transcription factor, which bind protein complexes containing Stat3 (but not Stat1, -5A, or -5B). Coexpression of activated c-Src and Stat3 synergistically induced strong HGF promoter activity in SP1 cells, as well as in a nonmalignant epithelial cell line, HC11 (HGF negative). c-Src kinase activity correspondingly increased the tyrosine 705 phosphorylation and DNA binding affinity of Stat3 (but not Stat1, -5A, or -5B). Collectively, our data indicate a cooperative effect of c-Src kinase and Stat3 in the activation of HGFtranscription and protein expression in breast carcinoma cells. This process may be important in overriding the strong repression ofHGF expression in nonmalignant epithelium, and thereby promote tumorigenesis.

Scatter factor, also known as hepatocyte growth factor (HGF), 1 is a multifunctional cytokine. Through binding to its receptor (Met), HGF can induce cell survival (1), growth (2), differentiation (3), and motility (4). It has been shown that both HGF and Met are essential for embryo development. Disruption of HGF expression in mice results in lethality in early development (5), while deletion of Met causes underdevelopment of limb buds (6). During development of the mammary gland, HGF is expressed by stromal cells, whereas epithelial cells express Met, but not HGF (7). Paracrine stimulation of normal breast epithelium with HGF, in cooperation with other growth factors (e.g. neuregulin), promotes branching morphogenesis (8). The tissue-specific suppression of HGF expression in normal epithelial cells provides a tightly controlled regulation of mammary ductal morphogenesis (9).
In contrast to normal breast epithelium, HGF and Met are frequently overexpressed in breast carcinomas (10 -12) as well as many other cancer types (10,11,13,14). This high level of HGF and Met expression has been identified as a possible independent predictor of poor survival in breast cancer patients (11). Our laboratory has previously shown that invasive human carcinoma cells coexpress HGF and Met, particularly at the migrating tumor front (12). We have also found that breast carcinoma cell lines frequently express HGF and Met, whereas most nonmalignant epithelial cell lines express Met, but not HGF. 2 Furthermore, overexpression of HGF or a constitutively active mutant form of Met (Tpr-Met) in transgenic mice (15,16) or in transformed cell lines (17,18) promotes tumorigenesis and metastasis. Together, these results suggest that establishment of an autocrine HGF loop and sustained activation of the Met signal transduction pathway in carcinoma cells may promote tumor progression. However, the mechanisms leading to aberrant expression of HGF in carcinoma cells are not known.
A number of signaling molecules, such as c-Src (19), Grb2/ Ras (17), and phosphatidylinositol 3-kinase (1), have been shown to be part of the HGF/Met signaling pathway. Activation of Met through binding of HGF causes autophosphorylation of two specific tyrosine residues in the cytoplasmic tail of the receptor tyrosine kinase (20). These phosphorylated tyrosine residues act as multifunctional docking sites that bind the SH2 domain of specific cytoplasmic signaling molecules and causes their activation. The c-Src nonreceptor tyrosine kinase is expressed in many cell types, and its activity is increased in response to HGF and binding to Met (19). Increased activation of the tyrosine kinase c-Src occurs in many human cancer cells, and c-Src plays a critical role in breast cancer. Overexpression of an activated form of c-Src in transgenic mice induces mammary hyperplasia (21). Furthermore, c-Src kinase is required in polyoma middle T-induced mammary tumorigenesis in transgenic mice (22). We have shown previously that c-Src kinase is constitutively activated in a mouse breast carcinoma cell line, SP1, which expresses both HGF and tyrosine-phosphorylated Met and which exhibits spontaneous invasion through matrigel (19,23,24). Furthermore, c-Src kinase activity is required for HGF-dependent cell motility and anchorage-independent growth of SP1 cells (19). Collectively, these findings indicate that c-Src kinase is an important requirement, but is not sufficient, for mammary tumorigenesis.
Activation of c-Src kinase can lead to increased expression of many genes, including growth factors such as vascular endothelial growth factor (25,26) and parathyroid hormone-related peptide (27). We therefore hypothesized that elevated c-Src activity can promote increased HGF expression and the establishment of an HGF autocrine loop in SP1 cells. We observed that the c-Src tyrosine kinase inhibitor PP2 causes a 2-fold reduction in HGF transcription in SP1 cells. In addition, expression of a dominant negative mutant of c-Src (SRC-RF) in SP1 cells leads to similar levels of reduction in HGF mRNA and functional protein. Using deletion mutants of the HGF promoter, we have located a region (between Ϫ254 and Ϫ70) of the HGF promoter responsive to increased c-Src kinase activity in SP1 cells. This region contains two putative consensus binding sites for Stat3. Stat3 is a transcription factor originally described as the target of interferon receptors (28), but recent reports have indicated that Stat3 can be activated by c-Src kinase via platelet-derived growth factor (29) and HGF receptors (30), and is important in mammary differentiation (30). We therefore examined the role of Stat3 in c-Src-dependent regulation of HGF transcription. The results indicate that while expression of Stat3 alone increased HGF promoter activity, simultaneous expression of Stat3 and activated c-Src led to strong cooperative activation of HGF transcription in both nonmalignant epithelial and carcinoma cells. Expression of mutant c-Src kinases in breast carcinoma cells altered both the tyrosine phosphorylation status and DNA binding activity of Stat3. While activated c-Src induced Stat3 tyrosine phosphorylation and DNA binding activity, a dominant negative mutant of c-Src reduced tyrosine phosphorylation and DNA binding. Together these data suggest that c-Src kinase and Stat3 act cooperatively in the activation of HGF expression in breast carcinoma cells, and may be important in overriding the strong repression of HGF expression in nonmalignant epithelial cells.

MATERIALS AND METHODS
Antibodies and Reagents-Rabbit anti-c-Src IgG was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal antibody EC10 against chicken c-Src was a gift from Dr. S. Parsons. Rabbit anti-sheep IgG conjugated with horseradish peroxidase was from Jackson Immu-noResearch Laboratories (West Grove, PA). Sheep anti-HGF IgG was a gift from Genentech (San Francisco, CA). Rabbit anti-HGF antibody was generated against recombinant glutathione S-transferase-HGF-(1-120) protein in our laboratory at Queen's University, this antibody recognizes only the N-terminal portion of HGF (data not shown). Anti-Stat1, -Stat3, -Stat5A, and -Stat5B and anti-phospho-Stat3 (Y705) antibodies were obtained from Upstate Biotechnology (Lake Placid, NY). c-Src family kinase inhibitor PP2 was obtained from Calbiochem (La Jolla, CA).
Plasmid Construction-c-Src expression plasmids were constructed by subcloning activated (Y527F) and dominant negative (K295R,Y527F) chicken c-src cDNAs (gift from Drs. J. Brugge and D. Shalloway) into the EcoRI site of DNA polymerase I (Klenow fragment)treated pBabePuro plasmid to generate pBabe Y527F and pBabe Src-RF. A reporter construct containing the full-length HGF promoter region fused to luciferase (2.7 HGF-luc) was constructed by ligating the HindIII/XbaI fragment (treated with DNA polymerase I (Klenow fragment)) of 2.8 HGF-CAT (gift from Dr. R. Zarnegar) into the HindIII site of pGL2-Basic (Promega), also treated with DNA polymerase I (Klenow fragment). Further deletions were constructed by cutting 2.7 HGF-luc with SmaI, SacI, and BglII, followed by re-ligation to generate 0.5 HGF-luc, 0.3 HGF-luc, and 0.1 HGF-luc, respectively. The 1.2 HGF-luc was constructed by ligating the 1.4-kb SalI fragment from 2.7 HGF-luc into the XhoI site of pGL2-Basic. An internal deletion mutant 0.5⌬ HGF-luc was constructed by digestion of 0.5 HGF-luc with PvuII/BglII and treatment with DNA polymerase I (Klenow fragment) before religation. The ⌬1 HGF-luc was constructed by ligating the SmaI fragment of 2.7 HGF-luc into the same site of 0.5⌬ HGF-luc. The ⌬2 HGF-luc was constructed by ligating the SmaI fragment of 2.7 HGF-luc into 0.8 HGF-luc. The ⌬⌬ HGF-luc was made by ligating the SmaI fragment of ⌬2 HGF-luc into the same site of 0.5⌬ HGF-luc. For normalization of transfection efficiency of each sample, pSG5␤gal (a gift from Dr. M. Petkovich) or pCHC␤gal (a gift from Dr. F. Kern) (31), which expresses ␤-galactosidase under the control of SV40 and cytomegalovirus promoters, respectively, was used.
Tissue Culture and Cell Lines-The SP1 tumor cell line is derived from a spontaneous poorly metastatic murine mammary intraductal adenocarcinoma, and expresses both HGF and Met. The characterization of the SP1 cell line has been described previously (19,23,24). Maintenance medium for SP1 cells was RPMI 1640 supplemented with 7% fetal bovine serum. HC11 is a mammary epithelial cell line (32) and was maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, insulin (5 g/ml), and epidermal growth factor (10 ng/ml).
Cell Transfection-All transfections were carried out with Lipo-fectAMINE Plus reagent (Canadian Life Technology, Burlington, ON, Canada) according to manufacturer's instructions. Cells (15,000) were seeded in a 24-well plate and transfected with 0.4 g of reporter plasmid, 0.1 g of pSG5␤gal, and up to 0.4 g of expression plasmids (such as c-Src) as indicated. After 48 h, transfected cells were harvested and lysed. One-fifth of the cell lysate was used to assay for ␤-galactosidase activity, an equal amount of lysate was used for a luciferase assay using PharMingen Luciferase Substrates (BD PharMingen, Mississauga, ON). Luciferase activity was measured using a luminometer with wavelength at 562 nm. Luciferase activity of each sample was normalized to the corresponding ␤-galactosidase activity. For immunoprecipitation and in vitro c-Src kinase assays, 2.5 ϫ 10 5 cells were seeded in a 100-mm tissue culture plate and transfected with 4 g of reporter plasmid, 1 g of pSG5-␤-galactosidase, and up to 4 g of expression plasmids as indicated. One-tenth of the cells was used for a luciferase assay, and the remaining cells were lysed and used for immunoprecipitation.
To obtain stably transfected cells, SP1 cells were plated at 70% confluence in 60-mm plates and transfected with 2 g of plasmids expressing various mutants of c-Src. Puromycin (2 g/ml, Sigma, Oakville, ON) was added to cells 24 h following transfection, and was maintained until all cells in the mock transfection were killed. Puromycin-resistant cells were then collected and used as pooled cell lines. Expression and activity of c-Src mutants in transfected cells were checked using Western blotting analysis and a c-Src kinase assay. Total c-Src protein was immunoprecipitated with an excess amount of antic-Src (pan) antibody to maximize the amount of antibody-protein complex formed. We have previously found that these c-Src mutants are quite effective, and that relatively small levels of expression can result in significant phenotypes (19).
RNA Isolation and RT-PCR-Cells grown to 80% confluence on a 100-mm dish were washed and lysed with TriZol reagent (Canadian Life Technology). Phase separation was achieved by addition of chloroform and centrifugation at top speed in a microcentrifuge for 10 min. Aqueous phase containing total RNA was removed to a new tube and precipitated with an equal volume of isopropyl alcohol for 10 min at room temperature. The RNA pellet was recovered by centrifugation and washed with 70% ethanol. After brief drying, the RNA pellet was resuspended in diethyl pyrocarbonate-treated water. RNA concentration was determined by spectrophotometry. An aliquot (1 g) of total RNA was used for reverse transcription with avian myeloblastosis reverse transcriptase at 42°C for 15 min. One-tenth of the reaction was used in PCR analysis with end-labeled oligonucleotides specific for HGF (5Ј-TGTCGCCATCCCCTATGCAG-3Ј and 5Ј-GGAGTCACAAGTCT-TCAACT-3Ј) and ␤-glucuronidase (GUSB) sequences, as previously described (33). The PCR reaction conditions were 2 min at 95°C, followed by 25 cycles of 1 min at 95°C, 1 min at 55°C, 1 min at 72°C, and a final cycle of 10 min at 72°C. The reaction was then analyzed on a 2% agarose gel by electrophoresis. The bands corresponding to the HGF and GUSB products were excised and the amount of radioactivity was determined by scintillation counting.
Copper Affinity Column Chromatography-Conditioned media were collected and HGF was partially purified using copper (II) affinity column chromatography, as described previously (34). Cells were grown to 80% confluence. The cell monolayer was washed with fresh Dulbecco's modified Eagle's medium and incubated in serum-free Dulbecco's modified Eagle's medium for 24 h. Conditioned media were collected, and cell debris was removed by centrifugation. Conditioned medium (10 ml) from each cell line was then loaded onto a copper (II) affinity column. The copper (II) affinity column was prepared by chelating Cu 2ϩ ions on a 1-ml HiTrap Chelating column (Amersham Pharmacia Biotech, Baie d'Urfe, PQ), and equilibrated with equilibration buffer (20 mM sodium phosphate, pH 7.2, 1 M NaCl, 1 mM imidazole). The conditioned medium was recycled through the column 5 times to ensure binding of all HGF proteins, and the column was washed thoroughly with 15 volumes of equilibration buffer. HGF protein was eluted from the column with equilibration buffer containing 80 mM imidazole at a flow rate of 1 ml/min. Fractions of 1 ml each were collected; previous experiments have determined that essentially all HGF was eluted in fraction 2 (Ref. 34 and data not shown). The fraction containing HGF was concentrated by centrifugation with Microcon centrifugal filter devices (Millipore Corp., Bedford, MA) with a 10-kDa molecular mass cut off. The samples were analyzed on a denaturing 10% SDS-PAGE gel, followed by Western blotting with anti-HGF antibody.
Immunoprecipitation, Western Blotting Analysis, and c-Src Kinase Assay-Cells were grown to confluence and treated as indicated. After three washes with cold phosphate-buffered saline, cells were lysed in lysis buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 1 mM Na 3 VO 4 , 50 mM NaF, 2 mM EGTA, 2 g/ml aprotinin, 2 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride. Cell debris was removed by centrifugation and protein concentrations were determined by a bicinchoninic acid protein assay (Pierce, Rockford, IL). For immunoprecipitation, equal amounts of lysate were incubated with the indicated antibodies at 4°C for 2 h or overnight. Immunoprecipitates were collected on protein A-Sepharose (Amersham Pharmacia Biotech), washed three times with lysis buffer, separated by SDS-PAGE gel, and transferred to a nitrocellulose membrane. Western blotting analysis was performed as described previously (19).
In vitro c-Src kinase assays were performed as described previously (19). Briefly, each lysate was immunoprecipitated with anti-c-Src IgG (Santa Cruz Biotechnology) as described above. One-half of each immunoprecipitate was subject to SDS-PAGE under nondenaturing conditions and Western blot analysis to confirm the amount of c-Src protein present. The other half of each immunoprecipitate was assayed for c-Src kinase activity by incubating with 10 l of reaction buffer (20 mM PIPES, pH 7.0, 10 mM MnCl 2 , 10 M Na 3 VO 4 ), 1.4 g of freshly prepared acid-denatured enolase (Sigma), and 10 Ci of [␥-32 P]ATP. After a 10-min incubation at 30°C, reactions were terminated by the addition of 2 ϫ SDS sample buffer, and samples were subjected to 8% SDS-PAGE. Serine and threonine phosphorylations were hydrolyzed by incubating the acrylamide gel in 1 M KOH at 45°C for 30 min, followed by fixing in 45% MeOH and 10% acetic acid for 30 min at room temperature. The gel was dried under vacuum. Autoradiograms were produced and analyzed with a Storm PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assays-Nuclear extracts were prepared as described previously (35). Briefly, 10 7 cells were washed once with phosphate-buffered saline before resuspension in cold buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 mM sodium orthovanadate). Cells were allowed to swell on ice for 10 min before lysis by brief vortexing. Nuclei were pelleted and resuspended in buffer C (20 mM HEPES, pH 7.9, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 25% glycerol, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 mM sodium orthovanadate). High salt extraction was performed by incubation on ice for 30 min in buffer C and centrifugation at 4°C. The protein content of the supernatant (nuclear extract) was determined using a Bradford protein assay (Bio-Rad, Mississauga, ON).
Electrophoretic mobility shift assays were performed as described by Mohan et al. (36). Briefly the binding reaction was performed by incubating 5 g of nuclear extracts with 0.1 pmol of 32 P-labeled oligonucleotide probe in the presence of binding buffer (10 mM HEPES, pH 7.9, 60 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol), 9% glycerol, and 4 g of poly(dI-dC) (Amersham Pharmacia Biotech). Binding was allowed to proceed at room temperature for 10 min before analysis on 5% nondenaturing PAGE gel in Tris glycine buffer (40 mM Tris-HCl, pH 8.4, 266 mM glycine). When unlabeled oligonucleotides were added, 10-fold molar excess was included in the binding reaction. For supershifting experiments, nuclear extracts were incubated with 2 g of the indicated antibody at room temperature for 20 min prior to the binding reaction. After electrophoresis, the gel was fixed in 7% acetic acid, 40% methanol for 30 min, and dried under vacuum. The gel was then exposed to a PhosphorImager screen, and analyzed using a Storm PhosphorImager.

RESULTS
Inhibition of Activity of c-Src Family Kinases Impairs HGF mRNA Expression-To study the regulation of HGF expression in breast carcinoma cells, we used the mouse mammary carcinoma cell line SP1, which coexpresses HGF and tyrosine-phosphorylated Met (23). Semi-quantitative RT-PCR was performed to determine the levels of HGF mRNA in SP1 cells. We first examined the dose-dependent effect of an inhibitor of c-Src family kinases, PP2 (37). Total RNA was isolated from SP1 cells treated with different concentrations of PP2 and used for cDNA synthesis by reverse transcription. Relative HGF mRNA levels were determined by RT-PCR with HGF-specific primers, and each sample was normalized to the expression of a housekeeping gene ␤-glucuronidase (GUSB) (33). The results showed that the PP2 inhibitor reduced HGF mRNA expression in a dose-dependent manner up to 40% of untreated cells (Fig. 1A). In addition, we examined the level of transcription of the HGF gene using a reporter plasmid. A plasmid containing a 2.7-kb fragment 5Ј of the HGF transcriptional start site ligated to the firefly luciferase gene was transiently transfected into SP1 cells. Bell et al. (38) have previously shown that this 2.7-kb fragment of the HGF promoter contains all the necessary sequence to direct HGF expression and mimics the expression pattern of the endogenous HGF gene in transgenic mice. Following transfection, these cells were treated with different concentrations of the PP2 inhibitor under conditions used in Fig. 1A. After a 24-h incubation, the cells were lysed and luciferase activity in each sample was determined and compared with control cells. The results show a similar dose-dependent reduction of HGF transcription following PP2 treatment (Fig. 1B). These findings suggest that the activity of c-Src kinase family members is important in the regulation of HGF transcription and mRNA expression.
c-Src Kinase Activity Regulates HGF Expression at Both mRNA and Protein Levels-We further investigated the role of c-Src tyrosine kinase in HGF expression by transfecting chicken c-Src mutants (SRC-Y527F and SRC-RF) with altered kinase activity into SP1 cells. The SRC-Y527F mutant contains a phenylalanine substitution at tyrosine 527 which results in constitutive kinase activity (39,40). The SRC-RF mutant contains a double substitution at tyrosine 527 to phenylalanine and at lysine 295 to arginine, which produces a dominant negative phenotype (26). We have previously shown that expression of a similar dominant negative form of murine c-Src in SP1 cells reduces endogenous c-Src kinase activity and also impairs anchorage-independent growth in soft agar (19). As predicted, expression of the dominant negative form of chicken c-Src (SRC-RF) also decreased total c-Src kinase activity in SP1 cells, when compared with untransfected cells (Fig. 2, top panel). In addition, expression of the activated form of c-Src (SRC-Y527F) dramatically increased total c-Src kinase activity in SP1 cells. Expression of the chicken c-Src mutants was detected by an antibody (EC10) specific for avian c-Src (Fig. 2,  bottom panel).
To assess the effect of c-Src kinase activity on HGF mRNA expression, RT-PCR analysis was carried out on RNA extracted from SP1 cells expressing the different c-Src mutants, or treated with the PP2 inhibitor (Fig. 3A). Expression of the dominant negative SRC-RF mutant or treatment with PP2 reduced the HGF mRNA level in SP1 cells by ϳ60%. Conversely, expression of the constitutively active SRC-Y527F mutant increased HGF mRNA expression by about 2-fold. In a parallel approach, the level of secreted HGF protein was compared in conditioned media collected from the same cells and under the same conditions described in Fig. 3A. Our laboratory has previously shown that HGF is a Cu(II)-binding protein, which can be purified with high recovery from conditioned media with copper (II) affinity chromatography (34) and analyzed on a denaturing SDS-PAGE gel (Fig. 3B). Using this method, we showed that expression of the dominant negative SRC-RF mutant or treatment with PP2 significantly decreased the amount of HGF protein secreted by SP1 cells. In contrast, expression of activated c-Src (SRC-Y527F) increased the amount of secreted HGF protein. Together these data suggest that HGF expression (both at the mRNA and protein levels) is regulated by c-Src kinase activity.
c-Src Kinase Activity Induces HGF Expression through a Specific cis-Acting Region on the HGF Promoter-To determine the effect of c-Src kinase mutants on HGF promoter activity, we constructed a series of reporter plasmids with the luciferase gene linked to different fragments of the 2.7-kb region 5Ј of the HGF transcriptional start site (Fig. 4B). These reporter constructs were co-transfected into SP1 cells with a control vector, or vectors expressing the SRC-Y527F or SRC-RF mutants of chicken c-Src kinase, and luciferase activity of the transfected cells was compared (Fig. 4A). The results show that expression of activated c-Src increased up to 2-fold the activity of the 2.7-kb HGF promoter, whereas dominant negative c-Src had the opposite effect. Deletions of up to Ϫ538 bp (0.5⌬ HGF) had no significant effect on the c-Src dependent activity of the HGF promoter, although some fluctuations in basal activity of the promoter were apparent. A further deletion of Ϫ273 bp (0.3 HGF-luc) significantly reduced the basal HGF promoter activity, while some c-Src dependent activity remained. The remaining c-Src kinase responsiveness was eliminated when all but 72 bp (0.1 HGF-luc) of the HGF promoter was removed. This suggests that a cis-acting element responsive to c-Src kinase activity is located within Ϫ273 and Ϫ70 bp of the HGF promoter. An internal deletion construct lacking the Ϫ70 to Ϫ254-bp region (named 0.5⌬ HGF-luc) was used to confirm the c-Src responsiveness of this region. As predicted, the 0.5⌬ HGF-luc reporter did not respond to expression of SRC-Y527F, although basal activity remained. A similar pattern of repression of the luciferase activity among all the HGF promoter deletion mutants used was seen when dominant negative c-Src (SRC-RF) was coexpressed with the HGF-luc constructs.
To confirm the importance of the regions of the promoter responsive to activated c-Src, several internal deletion mutants were constructed. Full-length reporter constructs missing Ϫ273 to Ϫ70 bp (⌬1), Ϫ1231 to Ϫ755 bp (⌬2), or both regions (⌬⌬) of the HGF promoter were transfected into SP1 cells in the presence or absence of the SRC-Y527F and SRC-RF mutants (Fig. 4A). As predicted, ⌬1 and ⌬⌬ deletion mutants exhibited neither induction nor repression of HGF promoter activity when activated c-Src or dominant negative c-Src was expressed, respectively. In contrast, the ⌬2 mutant showed strong induction of HGF promoter activity corresponding to expression of the activated SRC-Y527F mutant, and strong repression of HGF promoter activity when the SRC-RF mutant was expressed. This finding shows that only the region between Ϫ254 and Ϫ70 bp of the HGF promoter is important for c-Src responsiveness of HGF expression in SP1 cells. We will refer to this region as the c-Src responsive region.
Stat3 Activates HGF Transcription in Cooperation with Activated c-Src-Examination of the c-Src responsive region of the HGF promoter revealed several Stat3-binding sites. This consensus sequence is highly conserved among mouse, rat, and human (100% identity), while this conservation is lost in re-gions upstream of Ϫ500 bp of the HGF promoter (41). Since Stat3 activation by Src induces specific gene expression and is required for cell transformation (42,43), we examined whether expression of Stat3 in the presence or absence of the activated c-Src mutant (SRC-Y527F) has any effect on HGF promoter activity. A reporter plasmid containing the Ϫ2.7-kb full-length HGF promoter was co-transfected with a constant amount of the SRC-Y527F, and varying amounts of Stat3, expression plasmids. Expression of activated c-Src (SRC-Y527F) alone increased HGF transcription by about 2-fold (Fig. 5A). Likewise, expression of Stat3 alone increased HGF transcription by about 2-fold, and maintained a plateau value with even 0.05 g of plasmid DNA. However, in cells coexpressing both the activated c-Src mutant and increasing amounts of Stat3, HGF transcription increased up to 5-fold. This result indicates that there is a cooperative effect between c-Src kinase activity and

FIG. 3. c-Src kinase activity modulates HGF mRNA and protein levels in SP1 cells. Panel A, SP1 cells transfected with dominant
negative Src (SRC-RF) or activated Src (SRC-Y527F) or empty vector (control) were prestarved overnight. PP2 (40 M) was added to one plate of SP1 cells and incubated for an additional 24 h. A nonmalignant breast epithelial cell line HC11 was used as a negative control. Total RNA was isolated, and the amount of HGF mRNA in each sample was quantitated using RT-PCR and normalized to GUSB mRNA as described in the legend to Fig. 1. The level of HGF mRNA expression in each group was expressed as a percentage of that in untreated (control) cells. Values represent the mean of two experiments Ϯ range. Panel B, serum-free conditioned media were collected for 24 h from HC11 cells, PP2-treated SP1 cells, and SP1 cells transfected as in Panel A. HGF protein from the conditioned media was purified using copper (II) affinity chromatography (34). The fraction containing HGF protein was concentrated in Microcon concentrators and subjected to denaturing SDS-PAGE. Recombinant HGF (100 ng) was included in one lane as a control. After electrophoresis, the proteins were transferred onto nitrocellulose and the blot was probed with anti-HGF antibody. Immunoreactive bands were revealed using Enhanced Chemiluminescence kit. The nonmalignant mammary epithelial cell line, HC11, shows at least a 15-fold lower level of HGF transcription and no detectable HGF protein, compared with SP1 carcinoma cells (data not shown). We therefore determined whether coexpression of c-Src and Stat3 can activate HGF transcription in HC11 cells. Expression of activated c-Src induced expression by about 4-fold (Fig. 5B). In contrast to SP1 cells, expression of Stat3 alone in HC11 cells did not significantly induce HGF transcription. However, when activated c-Src and Stat3 were coexpressed, HGF transcription was synergistically induced up to 17-fold. Similarly, two clones of HC11 cells, stably transfected with Stat3 followed by transient expression of activated c-Src, showed up to a 20-fold increase in HGF promoter activity (data not shown). Together, these results suggest that increased c-Src kinase activity and Stat3 expression can override the repression of HGF transcription in nonmalignant mammary epithelial cells.
To determine whether the c-Src responsive region of the HGF promoter is involved in the observed cooperative effect between c-Src and Stat3, the transcriptional activity of a mutant HGF reporter lacking the c-Src responsive region (⌬1 HGF-luc) was compared with that of the full-length (2.7 HGFluc) HGF reporter. Each reporter construct was transfected into SP1 cells alone, or in combination with Stat3, and the activated c-Src (SRC-Y527F) mutant, expression plasmids. Expression of the activated c-Src mutant induced activation of the full-length HGF promoter, but not of the deletion mutant (⌬1 HGF-luc) (Fig. 6). Similarly, Stat3 expression increased the activity of the full-length HGF promoter, and only marginally affected that of the deletion mutant (⌬1 HGF-luc), this result suggests that Stat3 activates the HGF promoter. The level of induction due to Stat3 expression is even higher than that due to activated c-Src alone. This effect is probably due to a limiting amount of endogenous Stat3 in SP1 cells. When both Stat3 and activated c-Src were coexpressed, HGF promoter activity in the full-length construct was strongly induced, this effect was not seen in the deletion mutant (⌬1 HGF-luc). These results show a cooperative effect between Stat3 and activated c-Src in the induction of HGF transcription, and imply the presence of specific Stat3-binding sites on the HGF promoter.
c-Src Kinase Regulates Tyrosine 705 Phosphorylation and DNA Binding Activity of Stat3-Previous reports have found that c-Src activates Stat3 by inducing tyrosine phosphorylation of Stat3 and increasing its DNA binding affinity (42,44). We therefore examined the effect of c-Src kinase activity on Stat3 tyrosine 705 phosphorylation in SP1 cells. We found that expression of activated c-Src induced Stat3-specific tyrosine 705 phosphorylation, while expression of dominant negative c-Src had the opposite effect (Fig. 7). c-Src kinase activity similarly affected the nuclear protein binding affinity of the Stat3 consensus sites on the HGF promoter (Fig. 8). We used electrophoretic mobility shift assays to examine the Stat3 consensus DNA binding affinity of nuclear protein extracts from cells expressing different mutants of c-Src. Radiolabeled oligonucleotide probes with DNA sequences corresponding to the two Stat3 consensus binding sites in the region between Ϫ254 to Ϫ70 of the HGF promoter were used to detect putative Stat3 binding (Fig. 8). Binding of probes corresponding to each Stat3 consensus site (Ϫ110 or Ϫ149) was detected in nuclear protein extracts of SP1 cells (lane 1 in Fig. 8, A and B, respectively). These DNA binding activities were specific since the presence of the corresponding unlabeled probes abolished the binding (second lane), while a probe with an unrelated DNA sequence had no effect (third lane). In addition, when comparing first, fourth, and seventh lanes (Fig. 8), it is apparent that there was less specific DNA binding in nuclear extracts from SP1 cells expressing dominant negative c-Src than in control cells expressing no exogenous c-Src. Moreover, nuclear extracts from SP1 cells expressing activated c-Src had higher binding activity than that from untransfected cells. This finding indicates that the expression of dominant negative c-Src reduces nuclear protein binding to the Stat3 consensus sites, whereas activated c-Src has the opposite effect. Thus specific binding of nuclear protein to the Stat3 consensus sites correlates with phosphorylation at tyrosine 705 of Stat3 in these cells.
Although there is a strong indication of Stat3 being the transcription factor binding to the c-Src responsive region of the HGF promoter, other Stat proteins (such as Stat1, Stat5A, and Stat5B) can also bind to a Stat3 consensus site, albeit at lower levels (28,45,46). Therefore, antibodies against specific Stat proteins were used in supershift experiments to determine the composition of the DNA binding complex (Fig. 9). Nuclear extracts from SP1 cells were preincubated with antibodies against Stat1, Stat3, Stat5A, or Stat5B prior to the addition of the radiolabeled probe. Both Ϫ110 (Fig. 9A) and Ϫ149 (Fig. 9B) probes formed DNA-protein complexes when nuclear extracts were added. However, only anti-Stat3 antibody could efficiently bind to these complexes to form a supershift band. Antibodies to Stat1 (data not shown), Stat5A, or Stat5B did not retard the DNA-protein complex further, despite the fact that these tran- scription factors were present in SP1 cells (data not shown). This observation suggests that Stat3 is preferentially involved in the DNA-protein complexes which bind to the c-Src responsive elements in the HGF promoter.

DISCUSSION
During normal breast development, HGF is expressed primarily by mesenchymal cells, while its receptor Met is expressed by epithelial cells (7). However, HGF is expressed in regions of human invasive breast carcinoma, and in various breast carcinoma cell lines (11)(12)(13). During tumorigenesis HGF stimulates angiogenesis, invasion, and metastasis (47,48). Our laboratory (1) and others (49) have shown that HGF can stimulate survival of carcinoma cells. Therefore, acquired HGF expression leading to an HGF autocrine loop in breast carcinoma cells may be an important step during mammary tumorigenesis. However, the regulation of HGF expression in breast carcinoma cells is not very well understood, although some studies have been done in fibroblasts (50 -54). In the present study, we examined the role of c-Src kinase, which shows increased activity in human breast cancer (55), in controlling HGF expression in breast carcinoma cells.
We previously described a mammary breast carcinoma cell line, SP1, which expresses both HGF and activated Met (23). In SP1 cells, several downstream signaling molecules, such as phosphatidylinositol 3-kinase, phospholipase C␥, and focal adhesion kinase, are constitutively phosphorylated on tyrosine residues in SP1 cells, consistent with the presence of an autocrine loop (1,23). We have also found that c-Src tyrosine kinase in SP1 cells is constitutively active and is required for several HGF-dependent processes, such as cell motility and anchorageindependent growth (19).
In this report, we showed that inhibition of c-Src kinase activity in SP1 cells, through either the presence of c-Src kinase inhibitors or the expression of a dominant negative mutant of c-Src, caused a decrease in HGF mRNA and protein levels. Expression of an activated c-Src kinase had the reverse effect. This finding suggests that c-Src is important in regulating the basal level of HGF transcription in epithelial and carcinoma cells, and can induce elevated expression of HGF. However, since inhibition of c-Src kinase activity cannot completely eliminate HGF basal expression, other transcription factors may play roles in maintaining HGF basal expression. Indeed, in our system, the Sp1 transcription factor is essential in maintaining HGF basal level transcription, but has no effect on c-Src-induced HGF expression (data not shown). Furthermore, aggregates of SP1 cells expressing the activated form of c-Src, in which HGF protein level was high, showed spontaneous scattering when plated on plastic, compared with the parent cell line which required addition of exogenous HGF. 3 The higher level of endogenous HGF expression in SP1 cells expressing the activated form of c-Src may be sufficient to induce spontaneous scattering of these cells. Together, these findings suggest that c-Src kinase activity is important in regulating HGF expression.
By using deletion mutants of the HGF promoter, we mapped the c-Src responsive element to Ϫ254 to Ϫ70 bp. Since there is significant homology among the mouse, rat, and human HGF promoter sequences between Ϫ500 and ϩ1 (41), the regulation of HGF expression by c-Src kinase through this element is probably conserved among these species. Previous studies in fibroblast cells have demonstrated several transcription factors which regulate HGF expression: C/EBP (Ϫ4 bp) (50), an epithelial cellspecific repressor (Ϫ16 bp) (9), Sp1/Sp3 (Ϫ318 bp) (52), estrogen receptor (Ϫ872 bp) (51), and chicken ovalbumin upstream promoter-transcription factor (Ϫ860 bp) (51). Transgenic mouse studies showed that 0.7 kb of the HGF promoter exhibited the same expression pattern as the full-length (2.7 kb) promoter (38). Although in our system we observed that Sp1/Sp3 maintain the basal level expression of HGF in breast carcinoma cells, these sites are not responsible for c-Src induced expression of HGF (data not shown). The C/EBP site appeared to have no transcriptional activity in vivo (38). Binding sites for estrogen receptor and chicken ovalbumin upstream promoter-transcription factor are likely to be involved in estrogen-induced expression of HGF since the upstream sequence between Ϫ2.7 and Ϫ0.7 kb has been shown to be necessary for maximal inducibility of the HGF promoter (such as after partial hepatectomy) (38). However, the c-Src responsive region (Ϫ254 to Ϫ70 bp) described here has not been previously reported.
In the c-Src responsive region of the HGF promoter there are two consensus binding sites for Stat3 (at Ϫ110 and Ϫ149), both of which are completely conserved among human, mouse, and rat. Our results showed that Stat3, in cooperation with c-Src kinase, can activate HGF promoter, this activation is completely dependent on the presence of these Stat3-binding sites and implies a role of Stat3 as a downstream effector of c-Src kinase. We therefore examined the mechanism by which c-Src regulates Stat3 activity in SP1 carcinoma cells. Stat3 has been shown to be regulated by both tyrosine and serine phosphorylations (56 -58). Although there is no direct evidence that Stat3 is phosphorylated directly by c-Src, some reports suggest that c-Src and Stat3 interact physically (30,59). Therefore, it is possible that c-Src regulates Stat3 through tyrosine phosphorylation. Our results showed that expression of a dominant negative form of c-Src reduced tyrosine phosphorylation of Stat3 and the expression of constitutively active c-Src mutant had the opposite effect. In addition, we found that the formation of a DNA-protein complex with the two Stat3binding sites in the c-Src responsive elements was dependent on 3 B. Elliott, unpublished results. Stat2, -4, and -6 are not normally expressed in mammary tissues (60 -63), and are therefore unlikely to be involved in the formation of DNA-protein complexes in SP1 cells. Both Stat1 and Stat3 have been shown to be activated by c-Src in fibroblast cells when they are stimulated with various growth factors (30,45,64), while Stat5 is expressed and activated during mammary development (60). Moreover, both Stat3 and Stat5 have been found to be constitutively active in cells transformed by v-Src, v-Abl, and other oncoproteins (42-44, 59, 65-67). Therefore, other Stat proteins cannot be ignored as part of the complex. Supershift studies with antibodies against specific Stat proteins allowed us to identify Stat3, and exclude Stat1, -5A, or -5B, as a component of the DNA-protein complex. Furthermore, since there is only one DNA complex formed with each probe and each probe can effectively abolish DNA-protein complex formation with the other (data not shown), the same DNA-binding protein(s) must be involved in binding to each of these regions. Since Stat3 protein binds as dimers to its binding sites, it is reasonable to assume that Stat3 dimers are binding to both sites in the c-Src responsive region. Together, these observations suggest that c-Src kinase may regulate Stat3-dependent transcriptional activation through direct or indirect tyrosine phosphorylation of Stat3, resulting in increased DNA binding ability. In contrast to SP1 carcinoma cells, the nonmalignant mammary epithelial cell line, HC11, showed a very low level of HGF transcription with no detectable HGF protein. Furthermore, expression of activated c-Src (Y527F) had very little effect on HGF transcription in HC11 cells, possibly due to the presence of the epithelial cell type-specific repressor (9). However, coexpression of Stat3 and activated c-Src caused a strong synergistic induction of HGF transcription in HC11 cells, implying that the lack of c-Src kinase activity and the low level of activated Stat3 may be limiting for HGF transcription in HC11 cells. Increased activities of these proteins can possibly override the repression by the cell type-specific repressor and allow expression of HGF in epithelial cells. Interestingly, we found that fibroblast cells, which normally express HGF, also require c-Src kinase activity to regulate HGF expression, and that this FIG. 7. c-Src kinase activity regulates phosphorylation of tyrosine residue 705 of Stat3. SP1 cells transfected with SRC-RF or SRC Y527F, or untransfected SP1 cells, were lysed. Equal amounts of proteins from each cell lysate were subjected to denaturing SDS-PAGE. The proteins were then transferred onto nitrocellulose and the blot was probed with antibody specific for phosphotyrosine 705 of Stat3 (Panel A). The blot was subsequently reprobed with anti-Stat3 (pan) antibody (Panel B).
FIG. 8. c-Src kinase activity regulates nuclear protein binding to the Stat3 consensus sites (at positions ؊110 and ؊149) of the HGF promoter. Nuclear extracts were prepared from SP1 cells transfected with SRC-RF, SRC-Y527F, or untransfected cells. Equal amounts of each nuclear extract were used in binding studies with radiolabeled probes containing either the Ϫ110 (Panel A) or the Ϫ149 region (Panel B) of the HGF promoter. 10-fold molar excess of an unlabeled probe containing the Ϫ110, Ϫ149 or a nonspecific sequence (NS), respectively, was included in the binding reaction where indicated. The gel was fixed, dried, and analyzed using a Storm Phospho-rImager as described under "Experimental Procedures." The arrow indicates the position of the protein-DNA complex.
FIG. 9. Stat3 forms part of the DNA-protein complex at both the ؊110 and ؊149 consensus sites. Nuclear extracts were prepared from SP1 cells as described under "Experimental Procedures." For supershift assays, nuclear extracts were incubated with anti-Stat3, Stat5A, or Stat5B antibody on ice for 30 min prior to electrophoretic mobility shift assay analysis. After incubation with labeled Ϫ110 (Panel A) or Ϫ149 (Panel B) probes, the reaction was subjected to nondenaturing PAGE. The asterisk indicates the position of supershift band. regulation of HGF is dependent on the same region of the HGF promoter as our breast carcinoma cell model (data not shown). These results suggest a similar regulation pattern between fibroblast cells that express HGF endogenously, and carcinoma cells, which acquire the ability to express HGF. It is possible that during epithelial-mesenchymal transition, epithelial cells acquire different genetic mutations leading to the activation of c-Src kinase. For example, increased expression and activity of HER2/Neu, an epidermal growth factor-like receptor tyrosine kinase, in breast carcinoma cells has been shown to activate c-Src kinase (64,68). Activation of c-Src, in turn, may lead to a de-repression of HGF expression, giving these cells a growth advantage compared with nontransformed epithelial cells. This step may be an important initial step in tumorigenesis.
Here, we have reported that c-Src kinase and Stat3 act cooperatively in stimulating HGF gene expression in breast carcinoma cells, most likely via regulation of Stat3-dependent transcriptional activation of the HGF promoter. Although many reports have indicated that increased Src kinase activity (particularly through the expression of v-Src) can activate gene expression via Stat3, in this study we identify a target region (Ϫ254 to Ϫ70 bp) of the HGF promoter responsive to elevated activity of c-Src kinase in breast carcinoma cells. There is recent evidence suggesting that an HGF autocrine loop can provide selective survival and growth advantage to carcinoma cells and that overexpression of HGF can be a reliable indicator of poor survival of breast cancer patients (11). Our findings therefore provide an important link between breast cancer progression and HGF expression, and suggest that the c-Src/Stat3 pathway regulating HGF expression can be a potential target for therapy in breast cancer treatment.