Inhibition of Cell Migration by 24-kDa Fibroblast Growth Factor-2 Is Dependent upon the Estrogen Receptor*

The single-copy gene for fibroblast growth factor-2 (FGF-2) encodes for multiple forms of the protein with molecular masses of 24, 22.5, 22, and 18 kDa. We reported previously that the 24–22-kDa FGF-2 forms inhibit the migration of endothelial and MCF-7 cells by 50% and 70%, respectively. Here we show that this inhibition of migration is mediated by the estrogen receptor (ER). We have found that depletion of the receptor in either cell line abrogates the inhibitory activity of 24-kDa FGF-2 while re-introduction of the ER into deficient cells once again promotes the inhibitory response. To determine whether exposure to 24-kDa FGF-2 resulted in the activation of the estrogen receptor, 3T3 cells were cotransfected with estrogen receptor cDNA and an estrogen regulatory element-luciferase gene reporter construct and treated with 24- and 18-kDa FGF-2. The high molecular weight form stimulated luciferase activity 5-fold while 18-kDa FGF-2 at the same concentration had no effect. Treatment of ER-positive MCF-7 cells transfected with the reporter construct only showed the same results. Inclusion of the pure estrogen antagonist ICI 182,780 blocked the increase in luciferase activity by 24-kDa FGF-2, further indicating that the response was estrogen receptor dependent. Expression of dominant negative FGF receptor 1 inhibited ER activation, indicating that this was the cell surface receptor mediating the effect. Although growth factor-dependent

Multiple forms of FGF-2 1 are produced from a single-copy gene as a result of alternative translation initiation at an AUG codon or at three in-frame upstream CUG codons. This results in the synthesis of 4 FGF-2 isoforms of 24, 22.5, 22, and 18 kDa (1,2). The cellular localization and apparent functions of 18 kDa and the three higher molecular weight forms of FGF-2 (hmwFGF-2) differ. The 18-kDa FGF-2 is mostly cytoplasmic and is exported to the cell surface, where it is localized to the basement membrane or extracellular matrix in association with matrix heparins and heparans (3,4). In contrast, undetectable or extremely low levels of hmwFGF-2 are present in the media of cultured cells, suggesting that hmwFGF-2 is not released from cells. Instead, the majority of the cellular hmw-FGF-2 is directly translocated into the nucleus (5,6). The residues associated with nuclear translocation are RG repeats found at several sites within the amino-terminal region of hmwFGF-2 (7) that is absent from the 18-kDa FGF-2 form. Potential differences in function have also been reported for the 18-kDa versus the hmwFGF-2 forms. Transfection and expression of 18-kDa FGF-2 in 3T3 cells results in increased growth, motility, and levels of surface ␤ 1 integrins. In contrast, 3T3 cells transfected with the cDNA for hmwFGF-2 show enhanced growth but no increase in migration nor integrin expression (8,9). Additionally, PC12 cells overexpressing 18-kDa FGF-2 were found to differentiate toward the neuronal phenotype whereas overexpression of the higher molecular weight isoforms resulted in a stabilization of the endocrine phenotype (10). Thus, the family of FGF-2 growth factors demonstrate isoform-specific functions that differ depending on the cell type studied. Further control over the effect of the FGF-2 forms is provided by the extracelluar environment which regulates the ratio of their synthesis (11).
In a recently published article (12), we demonstrated that endothelial cells could be stimulated to secrete hmwFGF-2 in a regulated manner to levels capable of affecting cell behavior in an autocrine and paracrine fashion. The effects were 2-fold: stimulation of cell proliferation and inhibition of migration. The increase in proliferation was comparable to that promoted by an equal amount of 18-kDa FGF-2, indicating that the stimulation was independent of the additional amino-terminal peptide. On the other hand, the effect on migration was opposite to that of 18-kDa FGF-2. Although 18-kDa FGF-2 promotes cell motility, the hmwFGF-2 forms inhibited migration of endothelial cells by 50% and mammary carcinoma MCF-7 cells by greater than 70% even in the presence of unrelated mitogens that promote cell migration such as VEGF and IGF-1. Thus, we showed that, in addition to cellular effects promoted by hmw-FGF-2 transported into the nucleus, these forms could also dramatically alter cell function from outside the cells.
Among the various intracellular signaling pathways that are employed by peptide growth hormones to transmit signals into the nucleus is one that involves the activation of the ER. Although intuitively contrary to the characterization of these receptors as steroid hormone receptors, it has been shown repeatedly that they can be activated in a ligand independent * This work was supported by Grant R01 CA81209 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: La Jolla Inst. for Molecular Medicine, 4570 Executive Dr., San Diego, CA 92121. Tel.: 858-587-8788 (ext. 120); Fax: 858-587-6742; E-mail: glevin@ljimm.org. 1 The abbreviations used are: FGF-2, basic fibroblast growth factor; VEGF, vascular endothelial growth factor; hmwFGF-2, high molecular weight FGF-2; IGF-1, insulin-like growth factor-1; EGF, epidermal growth factor; ER, estrogen receptor; SDM, steroid-deficient medium; MAPK, mitogen-activated protein kinase; MAPKK, mitogen-activated protein kinase kinase, FGFR, fibroblast growth factor receptor; PAGE, polyarylamide gel electrophoresis; ERE, estrogen regulatory element. manner. Several peptide growth hormones including EGF and IGF-1 activate the receptor in a variety of cell types and in vivo (13)(14)(15)(16). In this study, we demonstrate that the ER mediates the inhibition of migration by 24-kDa FGF-2 in both endothelial cells and MCF-7 cells and that these forms of the growth factor can activate the ER leading to the transactivation of a reporter gene. In addition, we show that the mechanism by which 24-kDa FGF-2 activates the ER is different from that of other peptide growth hormones. This is the first report of an isoform-specific function of a growth factor being linked to an alternate intracellular signaling pathway.

MATERIALS AND METHODS
Synthesis of 24-kDa FGF-2-To generate pure recombinant 24-kDa FGF-2, full-length 24-kDa FGF-2 cDNA (obtained from Dr. M. Stachowiak, State University of New York, Buffalo, NY) was inserted in frame into a pPIC9K yeast expression vector (Invitrogen) between the SnaBI and AvrII restriction sites of the pPIC9K vector directly downstream of the DNA encoding the X-factor secretion signal region. The His4 mutant of P. pastoris GS 115, the methanol utilization-positive phonetype (Mutϩ), was transformed by electroporation with the pPIC9K construct vector linearized with SacI. The Hisϩ yeast transformants were grown on MD plates, and those carrying the kanT gene further selected for multiple integrated copies by replating on plates containing 4.0 mg/ml G418 antibiotic. The multicopy transformant was grown in 50 ml of BMGY medium with glycerol as the sole carbon source until the cultures reached an A 600 ϭ 2-6 (ϳ16 -18 h). The cells were collected by centrifugation at 1500 -3000 ϫ g for 5 min at room temperature and the cell pellet resuspended in BMMY medium to induce expression. The cells were grown for 4 days, and 100% methanol was added every 24 h to a final concentration of 2% to maintain protein expression. The medium was cleared of the yeast by filtration and mixed with 5 ml of heparin-Sepharose for 2 h at 4°C. The gel was washed with 50 ml of buffer A (20 mM Tris-HC1, pH 7.4, 5 mM EDTA, 2 mM EGTA) and then with buffer A containing 0.5 M NaCl. The recombinant 24-kDa basic fibroblast growth factor was eluted with buffer A containing 3 M NaCl. The eluate was dialyzed against 4 liters of 20 mM Tris, pH 7.4, 145 mM NaCl. Protein purity was assessed by 10% SDS-PAGE and Coomassie Blue staining and Western blot analysis using antibodies generated against a 12-amino acid peptide found within the amino-terminal domain of 24-kDa FGF-2 (17).
Cross-linking and Immunoprecipitation-125 I-18-kDa FGF-2 (10 Ci/ml; Amersham Pharmacia Biotech) was added to the cells cultured in 60-mm dishes for 2 h at 4°C. Cells were washed twice with phosphate-buffered saline and disuccinimidyl suberate (0.37 mg/ml; Pierce) added for 15 min at room temperature. The reaction was stopped with 0.2 M ethanolamine; after rinsing in phosphate-buffered saline, the cells were solubilized with 1% Triton X-100 in 50 mM HEPES, pH 7.5, 50 mM NaCl, 5 mM EDTA. Insoluble material was removed by centrifugation and the supernatant used for immunoprecipitation. For immunoprecipitation of FGFR1-4, 200 g of each cell lysate was incubated with 2 g of the polyclonal antibodies against FGFR1-4 (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C, and immune complexes were collected on protein A-Sepharose (Amersham Pharmacia Biotech). Complexes were washed twice in a buffer containing 20 mM HEPES, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, and 10% glycerol, then once in phosphate-buffered saline. Bound proteins were eluted by boiling in 2ϫ Laemmli sample buffer, separated by SDS-polyacrylamide gel electrophoresis (PAGE), and placed on film (Kodak BioMax).
Immunoblot Analysis-Cells were washed twice in Dulbeco's phosphate-buffered saline and then lysed with 0.5% Triton X-100 in 10 mM imidazole, pH 7.15, 40 mM KCl, 10 mM EGTA, 1 mM phenylymethylsulfonyl fluoride, 10 mM benzamidine, and 100 g/ml leupeptin. The lysates were spun for 2 min at 14,000 ϫ g to pellet cell debris and then subjected to the bicinchoninic acid protein assay (Pierce). The Triton X-100-insoluble material remaining on the flask, containing the cell nuclei, was washed twice with the lysis buffer and then solubilized in SDS-PAGE reducing sample buffer (50 mM Tris, pH 6.8, 5% ␤-mercaptoethanol, 2% SDS, and 10% glycerol). Samples representing equivalent amounts of each fraction were subjected to electrophoresis on 12% polyacrylamide gels. Upon completion of electrophoresis, proteins were transferred to nitrocellulose membranes, blocked for 1 h at room temperature in 5% nonfat milk in 20 mM Tris, pH 7.4, 145 mM NaCl (Tris-buffered saline) with 0.05% Tween 20 and then incubated with dilutions of primary antibodies in Tris-buffered saline with Tween 20 containing 0.1% bovine serum albumin. Antibodies used in this study were anti-18-kDa FGF-2 monoclonal antibody (Sigma), anti-human ER-␣ monoclonal antibody (Neomarkers), anti-smooth muscle cell actin monoclonal antibody (Sigma), and anti-FGFR1 monoclonal antibody (Santa Cruz Biotechnology). Bound antibody was detected using donkey anti-mouse or rabbit IgG conjugated to horseradish peroxidase (Jackson Laboratories) and the ECL chemiluminescence reagent. The stained immunoblots were placed against x-ray film (BioMax MR, Kodak) to generate chemiluminographs.
Culture of MCF-7, Endothelial, and NIH3T3 Cells-All cells were maintained at 37°C under 5% CO 2 in humidified incubators and subcultured to cell densities of one third to one sixth that of confluent cultures. MCF-7 cells were maintained in phenol red-free minimal essential media supplemented with 1 mM sodium pyruvate, 10% fetal calf serum, and 10 g/ml insulin. For the generation of estrogen receptor-negative populations, the cells were subcultured in steroid-deficient medium (SDM) prepared with phenol red-free minimal essential medium and 10% dextran/charcoal-stripped fetal calf serum. These MCF-7 cells were cultured in SDM for at least 6 months prior to use. The NIH3T3 cells were grown in Dulbecco's modified Eagle's medium containing 1 mM each penicillin and streptomycin and 10% calf serum. Bovine aortic endothelial cells were grown in Dulbecco's modified Eagle's medium containing 25 mM HEPES and 4.0 g/liter glucose supplemented with 1 mM sodium pyruvate, 1 mM nonessential amino acids, 1 mM penicillin/streptomycin, and 10% fetal calf sera.
Migration and Growth Assays-Endothelial or MCF-7 cells were harvested with trypsin, counted, centrifuged, and resuspended at 1 ϫ 10 5 cells in 0.5 ml of Dulbecco's modified Eagle's medium, 0.5% bovine serum albumin. Cells were added to the upper well of a Boyden chamber containing an 8.0-m pore size polycarbonate membrane separating the two chambers of a 6.5-mm Transwell (Costar). The upper wells were placed into lower wells containing 0.75 ml of Dulbecco's modified Eagle's medium/0.5% bovine serum albumin to which 10 ng/ml IGF-1(Sigma) or 10 ng/ml VEGF (Sigma) was added as a chemoattractant. Both chambers contained the 24-kDa FGF-2 at the appropriate concentrations. After 4 -6 h of incubation at 37°C in 5% CO 2 , nonmigratory cells on the upper membrane surface were removed with a cotton swab, and the cells that traversed and spread on the lower surface of the filter were fixed and stained with Diff-Quik (Dade-Behring). The filter was mounted on a glass slide, and four phase-contrast photomicrographs/ membrane were taken (magnification, ϫ100). The number of cells per field was counted from contact sheets and the results compared with control chambers that had no 24-kDa FGF-2 added. To measure growth rates, MCF-7 cells (6 ϫ 10 3 ) were plated in growth medium for 48 h, the medium changed to assay medium containing phenol red-free modified Eagle's medium supplemented with 1 mM sodium pyruvate and 0.3% lactalbumin hydrolysate plus or minus growth factors, and the cultures were allowed to incubate another 24 h. Two hours prior to the termination of the experiment, [ 3 H]thymidine was added. The cultures were washed with phosphate-buffered saline and then ice cold methanol (2ϫ), 5% trichloroacetic acid was added two times for 10 min each and discarded, and the DNA extracted with 0.3 N NaOH. The number of counts/min were determined on a liquid scintillation counter.
Plasmids and Transfection-MCF-7 cells or NIH3T3 cells were plated for transfection at a density of 1 ϫ 10 5 cells/well in a six-well tissue culture dish for 24 h. ER-positive MCF-7 cells were transfected with 2 g of pVitEREluc, a reporter construct containing three copies of the Xenopus laevis vitellogenin estrogen regulatory element proximal to the thymidine kinase promoter driving expression of the luciferase cDNA (provided by Dr. C. Glass, University of California, San Diego, CA). NIH 3T3 cells were cotransfected with 2 g of the reporter plasmid and 3 g of pSG5 plasmid containing the full-length wild type cDNA for the human ER (pSG5hER, HEGO) or one of two truncated forms of the ER (HE15 containing amino acids 1-282 of the ER or HEG19 containing amino acids 178 -595) provided by Dr. P. Chambon (Unite 84 de Biologie Moleculaire et de Genetique de I'NSERM, Strasbourg Cedex, France) (18). A mutant of the estrogen receptor in which serine 118 was changed to alanine was generated by the method of Kunkel (19), and the mutation was confirmed by sequencing. The FGFR1 expression plasmid was constructed by cloning the entire coding sequence of the three Ig-like domain form of human FGFR1 into pcDNA3.1 (Invitrogen). The dominant negative mutant lacking the tyrosine kinase domain was constructed by truncating the full-length receptor at the KpnI site (between Gly-402 and Thr-403) in the juxtamembrane domain. The DNA was mixed with LipofectAMINE (Life Technologies, Inc.) in 1.8 ml of Opti-MEM I (Life Technologies, Inc.), allowed to sit for 45 min at room temperature, and the mixture was added to the cells that had been washed twice with Opti-MEM I. Transfection was allowed to occur for 5.5 h at 37°C. The transfected cells were cultured for 24 h in their respective growth medium and the vehicle or agonists added at concentrations described under "Results" for another 24 h. When applicable, ICI 182,780 (Zeneca Pharmaceuticals) or the MAPKK inhibitor PD98059 was added 30 min prior to the addition of agonist. The cells were lysed with 0.1 M potassium phosphate, pH 7.8, 0.2% Triton X-100, 1 mM dithiothreitol, and the insoluble material was removed by centrifugation at 10,000 ϫ g. ER transactivation ability was determined by measuring luciferase activity (20).

Depletion of the ER and Loss of Inhibition-A requirement
for the presence of ER in 24-kDa FGF-2-mediated inhibition of migration was tested in MCF-7 and endothelial cells by depleting the receptor content and testing the effect on 24-kDa FGF-2-mediated migration. To reduce the level of the ER in endothelial cells, cultures were maintained in the presence of 500 nM ICI 182,780 which promotes changes in ER conformation which target the receptor for degradation (21). Immunoblot analysis for ER content demonstrates the reduction in the level of both the cytosolic (cyt) and nuclear (nuc) ER of the ICItreated cells (Fig. 1A). ICI-treatment reduced the amount of the 67-kDa ER by greater than 70%, in the cytoplasm and to undetectable levels in the nucleus as compared with parallel cultures of the same cells cultured in the absence of the antiestrogen. Equivalent loading of cellular material was demonstrated by reprobing the blot with an anti-␤ actin peptide antibody. The loss of the ER had no significant effect on the base rate of endothelial cell migration (1.0 versus 0.86 Ϯ 0.16, ER-positive versus ER-negative) nor VEGF-induced enhanced migration (1.8 Ϯ 0.2 versus 2.0 Ϯ 0.2, ER-positive versus ERnegative), indicating that basal motility and chemokinetic stimulation through the VEGF receptor were not impeded by the loss of the ER (Fig. 1B). However, the reduction in ER levels resulted in the loss of the inhibitory response to 24-kDa FGF-2. The migration rate of receptor positive endothelial cells was reduced to 48 Ϯ 8% of control values but declined only 10% to 90 Ϯ 16% with ER-negative cells. The same was true in the presence of VEGF. The migration rate was reduced dramatically by 24-kDa FGF-2 in VEGF-treated ER-positive cells (to 40 Ϯ 12% of ER-positive control cells), while the ER-negative cells showed no such decline. Further evidence for the involvement of the ER in the inhibition of migration in endothelial cells was obtained with short term exposure of the cells to 10 nM ICI 182,780, which results in the inactivation of ER function without the loss of ER antigen (21). Simultaneous exposure of endothelial cells to ICI and 24-kDa FGF-2 resulted in a loss of the inhibitory potential of the 24-kDa FGF-2; motility was inhibited only 25% in cultures treated with ICI versus a 60% decline of migration in its absence (data not shown). Thus, ER function as well as ER protein is required for the inhibition of migration by the 24-kDa FGF-2.
Depletion of the ER had the same effect on 24-kDa FGF-2mediated inhibition of MCF-7 migration (Fig. 1, C and D). MCF-7 cells were cultured and maintained in SDM, which results in a nonclonal population of cells that no longer exhibit a mitogenic response to ␤-estradiol (12). This lack of estrogen responsiveness is accompanied by a decrease in ER levels as assessed by immunoblot analysis (Fig. 1C). Cytoplasmic ER was undetectable, while nuclear ER levels were substantially reduced. 24-kDa FGF-2 inhibited the migration of these ERpositive cells by 77 Ϯ 18%, while in ER-negative cells the migration rates actually increased to 1.3 Ϯ 0.22 times those of control, ER-positive cells (Fig. 1D).
Reestablishing the 24-kDa FGF-2 inhibition of migration in ER-negative MCF-7 cells could be achieved by re-introducing the estrogen receptor to its normal levels. MCF-7 cells propagated in SDM (ER-negative; Fig. 2A, lane 2) were placed in serum-free Opti-MEM I for 4 h and then returned to SDM for 24 h. This short term exposure to Opti-MEMI resulted in the appearance of endogenous ER at levels approximating those of the MCF-7 cells retained in normal medium ( Fig. 2A, compare  lanes 1 and 3). Fig. 2B shows that the migration rate of the cells remaining in SDM (corresponding to lane 2, panel A) is not affected by 24-kDa FGF-2 but replicate cultures stimulated to produce the ER (lane 3, panel A) only migrated 37.1 Ϯ 10.7% as fast as the control cells. Although 24-kDa FGF-2 inhibits the migration of MCF-7 cells, it acts as a stimulus for cell proliferation equally as well as 18-kDa FGF-2 (17). Therefore, the effect of ER-depletion on the growth promoting activities of 24-kDa FGF-2 was examined using MCF-7 cells cultured in SDM or regular medium (Fig. 2C). Cells were incubated with 24-kDa FGF-2 or 18-kDa FGF-2 and the rate of proliferation determined by thymidine incorporation studies. No change in the rate of cell proliferation was observed in either case regardless of the presence or absence of the ER (3.7 Ϯ 0.3 versus 3.3 Ϯ FIG. 1. A, prolonged incubation in ICI 182,780 depletes endothelial cells of ER antigen. Endothelial cells were incubated in complete medium containing 500 nM ICI 182,780 for 10 days and the cultures extracted with 0.5% Triton X-100. After removing the soluble material, the nuclei were extracted with SDS. Both samples were fractionated by SDS-PAGE and the ER detected by immunoblotting with anti-ER antibodies. The blots were reprobed for actin to confirm equal loading. B, depletion of the ER abrogates the inhibitory effect of 24-kDa FGF-2. ICI-treated cells were used to test the effect of 5 ng/ml 24-kDa FGF-2 on the migration rates of the estrogen-negative versus estrogen-positive endothelial cells. Cell motility was tested in Boyden Chamber assays in the presence and absence of 10 ng/ml VEGF. The number of migrating cells were counted and compared with the number migrating in the absence of 24-kDa FGF-2. ER deficiency had no effect on the basal migration of endothelial cells with or without VEGF but abrogated the inhibitory effect of 24-kDa FGF-2. C, reduction of ER antigen in MCF-7 cells by prolonged culture in steroid-deficient medium. Replicate cultures of MCF-7 cells were maintained continuously in either complete (RM) or SDM. With SDM the level of ER was undetectable in the cytoplasm by Western blot analysis and significantly reduced in the nucleus. D, the effect of 5 ng/ml 24-kDa FGF-2 on ER-negative and positive MCF-7 cell mobility was tested in Boyden Chamber assays with 10 ng/ml IGF-1 as a chemoattractant. The data show that the inhibition of migration is dependent on the ER but basal motility is not. The results are presented as a ratio of the number of 24-kDa FGF-2-treated cells migrating onto the bottom of the membrane to that occurring in the absence of 24-kDa FGF-2. The data are presented as the mean Ϯ S.D. cyt, cyto, cytosolic; nuc, nuclear; C, control. Thus, the involvement of the ER in 24-kDa FGF-2 function is specific to the inhibition of migration and is not relevant to its effect on cell division.

24-kDa FGF-2 Activates the Estrogen Receptor
Activation of the ER by 24-kDa FGF-2-To assess the effect of 24-kDa FGF-2 on ER activity, we measured the ability of the growth factor to modulate the transcriptional activity of ER in transient transfection assays. 3T3 cells were cotransfected with a plasmid containing a triple estrogen regulatory element (ERE) proximal to the thymidine kinase promoter driving luciferase reporter gene expression and a plasmid containing the cDNA for the full-length human ER-␣ (22). Fig. 3A shows the low level of endogenous ER in the 3T3 cells and the increase in ER levels 24 h after transfection of the ER cDNA. Treatment with estradiol for 24 h had no effect on the intracellular content of the recombinant ER (Fig. 3B). Treatment of these cells with 10 nM estradiol increased the luciferase activity in the cell extracts 3.6 Ϯ 0.1-fold. Dose-dependent increases in luciferase activity also were observed with increasing doses of 24-kDa FGF-2 from 5 to 20 ng/ml. As the concentration of the growth factor was raised, there was a corresponding increase in luciferase activity up to 2.5 times control levels. Higher doses (30 ng/ml) had no additional effect. When the cells were transfected with the reporter plasmid only (no ER expression), no increase in luciferase activity was observed, even at 20 ng/ml, demonstrating that the ER was necessary for the stimulation of luciferase gene transcription. In contrast to the higher molecular weight FGF-2, 18-kDa FGF-2 at 20 ng/ml concentration had little effect on luciferase activity, establishing the specificity of ER activation for the higher molecular weight forms of

FIG. 3. 24-kDa FGF-2 activates ER transcriptional activity. A,
Western blot analysis of extracts of 3T3 cells transfected with the ER gene. NIH 3T3 at 50% confluence in six-well tissue culture dishes were transiently transfected with 3 g of plasmid pSG5hER containing the human ER cDNA and 24 h later the cell extracts analyzed for ER by Western blot analysis. Replicate cultures were treated with 10 nM estradiol (E 2 ) for 24 h to determine whether the steroid had any effect on ER levels. B, transactivation of the ER by 24-kDa FGF-2. NIH 3T3 cells were cotransfected with plasmid pSG5hER containing the cDNA for human ER and 2 g of 3xERE -luciferase reporter construct and 24 h later treated with 10 nM estradiol, 5-30 ng/ml 24-kDa FGF-2, or 20 ng/ml 18-kDa FGF-2. After an additional 24 h, the cultures were extracted and the samples assayed for luciferase activity. Replicate cultures were transfected with the reporter construct only and treated with 20 ng/ml 24-kDa FGF-2 (ERE only). The failure of the FGF-2 alone to promote luciferase activity demonstrates that the activity observed in the cotransfected cultures was ER-dependent. Data are presented relative to those from untreated (control) cultures (mean Ϯ S.D.). C, activation of endogenous ER by 24-kDa FGF-2 in MCF-7 cells. MCF-7 cells maintained in normal media (ER-positive) were transfected with the luciferase reporter construct for 24 h and 10 nM estradiol or 2-10 ng/ml 24-kDa FGF-2 added for an additional 24 h. Cells were extracted and the samples assayed for luciferase activity. Data are presented relative to those from untreated cultures (mean Ϯ S.D.). D, inhibition of 24-kDa FGF-2 transactivation of ER by the pure estrogen antagonist ICI 182,780. MCF-7 cells transfected with the ERE-luciferase reporter plasmid and treated with 10 ng/ml 24-kDa FGF-2 and 0.1, 1, and 10 nM ICI for 24 h before luciferase activity was measured. A dose-dependent reduction in ER transactivation is observed.

24-kDa FGF-2 Activates the Estrogen Receptor
FGF-2. Mammalian cell-derived recombinant hmwFGF-2 produced in COS cells (17) gave identical results (data not shown). The ability of 24-kDa FGF-2 to activate the ER also was tested in MCF-7 cells using the natural endogenous ER and the luciferase reporter construct. Cells were transfected with the reporter construct only and treated with estradiol or 24-kDa FGF-2. Treatment with 10 nM estradiol induced an increase in luciferase activity to 2.6 Ϯ 0.4 times control (Fig. 3C). Addition of 24-kDa FGF-2 created a dose-dependent increase in luciferase activity to 2.3 Ϯ 0.3 times control at 10 ng/ml. Further evidence for ER-dependent 24-kDa FGF-2 activation of luciferase expression was acquired with the pure anti-estrogen ICI 182,780 (Fig. 3D). This anti-estrogen inhibited the level of luciferase activity achieved with 10 ng/ml 24-kDa FGF-2 in a dose-dependent manner. Thus, consistent results were observed in 3T3 cells expressing an exogenous source of ER and in MCF-7 cells using the endogenous receptor.
18-kDa FGF-2 binds to and activates four different high affinity tyrosine kinase receptors, FGFR1/flg to FGFR4, although not all are found in a single cell type (23)(24)(25)(26). To determine which of the four receptors were present in the MCF-7 cells, aortic endothelial cells, and 3T3 cells employed in these studies, cells extracts were analyzed by cross-linking 125 I-18-kDa FGF-2 to the cells and performing immunoprecipitation with antibodies against each of the four receptors (Fig.  4A). FGFR1 was the only receptor detected in each cell type and was the only one present in the 3T3 and endothelial cells. The doublet observed in 3T3 cells represents the two and three IG-like domain isoforms of FGFR1 (data not shown). The possibility that FGFR1 mediates the activation of the ER by 24-kDa FGF-2 was tested in 3T3 cells by expressing a dominant negative FGFR1 in which the tyrosine kinase domain had been deleted. Expression of this receptor and the wild type receptor following transient transfection is shown in Fig. 4B (right panel). In cells expressing the mutant FGFR1, no increase in ER activation occurred with 10 ng/ml 24-kDa FGF-2 treatment while the activation by estradiol was not affected. When the experiments were repeated with cells transfected with a plasmid containing wild type FGFR1 cDNA, no difference in ER activation was found when compared with control cells.
Binding of FGF-2 to the extracellular domain of FGFR results in activation of the intrinsic kinase activity causing tyrosine phosphorylation of a number of signaling intermediates including MAPKs (27)(28)(29)(30)(31)(32)(33). In COS and HeLa cells, ER activation by EGF depends on the phosphorylation of the ER (34). The intracellular pathway responsible for this phosphorylation event is the MAPK signal transduction cascade and results in the phosphorylation of Ser 118 of the ER, which is also the target of a phosphorylation event induced by estradiol (35)(36)(37). To determine whether 24-kDa FGF-2 can activate the MAPK signal transduction cascade, 3T3 cells were treated with 18-and 24-kDa FGF-2 and the activation of Erks 1 and 2 examined by Western blot analysis with antibodies against the phosphoform of the enzymes (Fig. 5). Both growth factors were added to the cells at 10 ng/ml for increasing amounts of time, and the levels of phospho and total protein analyzed. MAPK phosphorylation was elevated by both 18-and 24-kDa FGF-2 within 10 min and remain elevated for at least 1 h after treatment. The constant total levels of Erk1/2 show that this increase reflected the phosphorylation of the protein and not differences in loading. Addition of antibodies to the 18-kDa COOH-terminal domain or the amino-terminal end of 24-kDa FGF-2 neutralized the effect on MAPK phosphorylation (lower panel). Direct evidence for the involvement of the MAPK pathway in the activation of the ER was sought with the MAPKK inhibitor PD98059. Dose titration analysis of the effect of this inhibitor on the phosphorylation of ERK1/2 in 3T3 cells showed that 1 M or greater was capable of inhibiting the phosphorylation of MAPK induced by 24-kDa FGF-2 (Fig. 6). However, instead of reducing ER activation in parallel with the loss of MAPK phosphorylation, 1 M PD98059 treatment increased the level of luciferase activity in response to 10 ng/ml 24-kDa FGF-2 by 1.5

FIG. 4. Activation of the ER by 24-kDa FGF-2 is mediated by FGFR1.
A, analysis of 3T3, MCF-7, and endothelial cells for the presence of FGF receptors. Cross-linking experiments were performed with 125 I-18-kDa FGF-2 and cell lysates immunoprecipitated with antibodies to each of the FGFRs. Immunoprecipitates were fractionated by SDS-PAGE and the gels exposed to x-ray film. T47D cell lysates were analyzed concurrently to demonstrate positive staining for each receptor type. The doublet in the 3T3 cells represents the two and three Ig-like domain isoforms of FGFR1. The numbers underneath the panel identify the receptor specifically identified by the antibody used. N, nonimmune control. B, loss of activation with the expression of a FGFR1 tyrosine kinase-negative receptor. 3T3 cells were transfected with plasmids containing wild type FGFR1 or the tyrosine kinase mutant and the expression analyzed by immunoblot analysis with antibodies to the receptor (right panel). To test the effect of the mutant receptor, cells were cotransfected with the plasmids containing the tyrosine kinase negative FGFR1 cDNA, estrogen receptor, and the reporter gene construct and treated with either 10 ng/ml 24-kDa FGF-2 or 10 nM estradiol (filled bars). Control cultures were transfected in the same manner except the mutant plasmid was replaced with pcDNA3 vector at the same concentration (open bars). FGFR1, The experiments were repeated with cells transfected with plasmid containing wild type FGFR1 cDNA (hatched bar). Results are presented relative to those from replicate cultures containing the appropriate plasmid but left untreated. wt, wild type; mut, mutant.

FIG. 5. Induction of MAPK phosphorylation by 24-kDa FGF-2.
3T3 cells were treated with 10 ng/ml 24-kDa FGF-2 or 18-kDa FGF-2 for increasing amounts of time, the cells were extracted, and the protein lysates were immunoblotted with antibodies to either phospho-MAPK (P-MAPK) or total MAPK. In some experiments, antibodies to the 18-kDa domain or the amino-terminal end of 24-kDa FGF-2 (␣18kD and ␣24kD, respectively) were incubated for 30 min with the protein and the mixture added to the cells for 10 min. Cells were then analyzed for phospho-MAPK. C, control.
Specific activation factor domains of the ER have been related to the activation of the receptor by estradiol or peptide growth factors. In COS and HeLa cells, it is the AF-1 domain that mediates the growth factor-induced transactivation while the AF-2 domain is required for estradiol-induced activation (38 -40). To determine whether 24-kDa FGF-2 activation of the ER can be localized to a specific domain of the receptor, 3T3 cells were transfected with truncated forms of the ER cDNA containing either the AF1 or AF2 domains plus the DNA binding domain (Fig. 7). In contrast to what has been found with other growth factors, increases of the reporter luciferase activity in response to 24-kDa FGF-2 occurred with the truncated ER containing the AF2 domain (HEG19; 5.4 Ϯ 2.0-fold) but not the AF1 domain (HE15; 0.9 Ϯ 0.1-fold). The increase with HEG19 was comparable to that observed with estradiol, which also occurred through the AF2 domain (7.4 Ϯ 2.5-fold). Little difference was observed between 24-kDa FGF-2 activation of the truncated HEG19 ER and the wild type (HEGO; 4.4 Ϯ 2.0-fold). In addition to the domain specificity of EGF-induced ER activation, phosphorylation of the receptor at Ser 118 is required in COS1 and HeLa cells (39). Conversion of this serine to alanine eliminated activation of the receptor. To determine whether Ser 118 phosphorylation was involved with 24-kDa FGF-2 activation of the ER, this residue was substituted with alanine by site directed mutagenesis. Comparison of the effect of 24-kDa FGF-2 on cells containing the mutant ER versus the wild type receptor showed a 5.6 Ϯ 0.22-fold increase in luciferase activity in cells containing the mutant ER and 6.0 Ϯ 0.29fold with the wild type receptor. DISCUSSION The synthesis of various molecular weight isoforms of FGF depends upon the mechanism of alternative translation initiation (1,2). This process employs three CUG initiation sites upstream from a traditional AUG start site. This results in three higher molecular weight forms containing an aminoterminal peptide of increasing length coupled to the 18-kDa FGF-2. The presence of this peptide alters the cellular localization of FGF-2 from the cytoplasm (18 kDa) to the nucleus due to the presence of nuclear localizing RG repeats within the amino-terminal extension (3)(4)(5)(6). In addition to differences in localization, the hmwFGF-2 forms also have been found to modify cell behavior differently from the 18-kDa form. This difference was first suggested by experiments in which 18-kDa or hmwFGF-2 forms were overexpressed in 3T3 cells (8,9). All forms stimulated proliferation, but, whereas the smaller form stimulated migration, the hmwFGF-2 had no effect. Overexpression of the different forms also had distinct effects on PC12 cell phenotype with the higher molecular weight FGF-2-expressing cells displaying neuronal properties and 18-kDa FGF-2-expressing cells being more endocrine-like (10). Thus, not only is localization of the proteins altered by the amino-termi-

24-kDa FGF-2 Activates the Estrogen Receptor
nal extension, so is their function. Further control of cell behavior by this mechanism can occur since the different isoforms are expressed in response to changing environmental conditions (41), e.g. untransformed endothelial, epithelial, and fibroblastic cells express primarily the 18-kDa FGF-2 while stress conditions stimulate the synthesis of the CUG forms (11).
The difference in the function of 18-kDa and hmwFGF-2 forms was further demonstrated by studies in which recombinant hmwFGF-2 was added to endothelial cells and MCF-7 cells and their effects on cell migration evaluated (12). Cell migration was inhibited to greater than 50% by hmwFGF-2. This occurred even in the presence of saturating amounts of mitogens that stimulate cell motility (VEGF, IGF-1, 18-kDa FGF-2), suggesting that the effect of hmwFGF-2 can override the signals generated by these other factors. In this study we provide evidence that this effect on migration results from a change in the intracellular signaling pathway that the cell employs to transmit FGF-2-activated signals. With the synthesis of the amino-terminal end, FGF-2 can activate pathways that lead to the transactivation of the ER. An association between inhibition of migration and the activation of the ER is demonstrated by the parallel changes between the two with reduction in ER accompanied by loss of the inhibitory effect. This link was found in two different cell types (endothelial and MCF-7), which had been depleted of their ER by two different methods (maintenance in SDM and ICI 182, 780 treatment, respectively). In addition, short term treatment with ICI 182,780 blocked the inhibition of migration indicating that the ER must not only be available but also needs to be functional. The latter premise is consistent with our observations that 24-kDa FGF-2 has the capacity to transactivate the ER as demonstrated by the expression of an estrogen regulatory element controlled gene. Again, this occurred in two different cell types regardless of whether the ER was expressed from a transfected plasmid (3T3 cells) or was the natural product of the endogenous gene (MCF-7 cells).
The activation of the ER by a peptide growth hormone is not unusual; direct evidence for peptide growth factor activation of the ER has been obtained in multiple cell types (13)(14)(15)(16) and in ER knockout animals (42). Transforming growth factor-␣ or EGF treatment of HeLa, Ishikawa (endometrial), and BG-1 (ovarian) cells transiently transfected with ER-responsive target genes results in the dose-dependent activation of theER in an estrogen-independent manner (38). In primary rat uterine cell cultures (43), GH3 pituitary cells (16), and the neuroblastoma cell line, SK-ER3 (44), ER-dependent gene expression is stimulated by IGF-1. Growth factors also mimic estrogen in their ability to increase the expression of ER target genes, such as the progesterone receptor and the iron-binding glycoprotein, lactoferrin (45). EGF and IGF-1 also promote cell growth through the ER both in vitro and in vivo (38,43). What distinguishes hmwFGF-2 from these other growth factors is the fact that it affects motility, which has not yet been associated with peptide activation of ER. The mechanism by which EGF activates the ER has been shown to occur through the EGF plasma membrane receptor (38) and activation of specific MAPKs resulting in the phosphorylation of serine 118 of the ER. This occurs, however, in only certain cell types (e.g. COS and HeLa cells). Studies employing vascular smooth muscle cells failed to show involvement of either MAPKs or Ser 118 phosphorylation, indicating that multiple pathways are available for ER activation by these peptide hormones (20). In fact, overexpression of the dominant positive MAPKK in endothelial cells inhibited basal ER activity, suggesting that in these cells there is a reverse relationship between MAPKK activity and ER activity.
Our results are more consistent with those reported for the smooth muscle cells than the studies performed with COS or HeLa cells. Although 24-kDa FGF-2 clearly stimulates the phosphorylation of MAPK, inhibition of that phosphorylation with the MAPKK inhibitor PD98059 did not reduce the level of ER activation. In fact, we observed an increase in the basal level of ER activity and an increase in the 24-kDa FGF-2 induced activation of the ERwith the addition of the MAPKK inhibitor. Despite the independence from the MAPK pathway, it is apparent that a fully functional FGFR1 receptor must be present for 24-kDa FGF-2 to activate the ER. It is also likely that the tyrosine kinase activity of FGFR1is responsible for the transduction of the 24-kDa FGF-2 generated signals for ER activation. The diversity of the different transducer proteins associating with FGFR1 and the number of signaling intermediates that are activated may reflect the capacity of the different forms of FGF-2 to exert the complex array of biological responses in the same cell, a characteristic shared by other tyrosine kinase receptor-binding growth factors (46 -48).
In addition, the mechanism of ER activation by FGFR1 appears to involve distinct domains of the ER. Previously, the AF-1 domain was shown to be required for peptide growth factor ER activation while the AF-2 domain mediated the estradiol effect. The use of the same truncated forms of the ER transfected into 3T3 cells demonstrated the opposite effect with 24-kDa FGF-2. Thus, activation of the ER by 24-kDa FGF-2 requires the AF-2 and not the AF-1 domain. Consistent with these results is our observation that the elimination of the putative MAPK phosphorylation site Ser 118 by conversion to alanine has no effect on the activation of the receptor, a result consistent with the effect of growth factors on the ER in smooth muscle cells but not in COS or HeLa cells.
Alternative translation site usage was first characterized as occurring during viral replication, particularly among the picornaviruses. It is a mechanism that allows for the expansion of gene diversity within a highly restricted genome by changing the localization and the functions of a single gene. Although much less common, alternative translation initiation using noncanonical codons also exists for mammalian proteins other than FGF-2. Among these are several whose function changes with the synthesis of their alternate forms. For example, there are two forms of GATA-1 that differ in their transactivation potential and developmental expression (49), the cAMP-responsive element modulator is converted from a transcriptional activator to a repressor (50), and Egr3 transcription factors differ in their ability to activate transcription (51). As transcription factors, the changes in these three proteins most likely involve how they interact directly with DNA or other DNA-binding proteins. In contrast, changes in 24-kDa FGF-2 function involve a more complex process involving multiple steps between cell surface receptor activation, ER activation, and the inhibition of migration. This represents the novelty of our observations on FGF-2 that, in addition to the functional change that occurs with the additional amino-terminal peptide, there is also an activation of a separate signaling pathway that can be directly linked to the change in function.