Interaction between the Adhesion Receptor, CD44, and the Oncogene Product, p185 HER2 , Promotes Human Ovarian Tumor Cell Activation*

In this study we have examined the interaction between CD44s (the standard form) and the p185 HER2 proto-oncogene in the ovarian carcinoma cell line. Surface biotinylation followed by wheat germ agglutinin column chromatography and anti-CD44-mediated immunoprecipitation indicate that both CD44s and p185 HER2 are expressed on the cell surface and most importantly, that these two molecules are physically linked to each other via interchain disulfide bonds. We have also determined that hyaluronic acid stimulates CD44s-associated p185 HER2 tyrosine kinase activity, leading to an increase in the ovarian carcinoma cell growth. After transfection of the ovarian carcinoma cell line with the adenovirus 5 E1A gene, which is known to repress p185 HER2 expression, we observed that both surface CD44s expression and CD44s-mediated cell adhesion to hyaluronic acid are significantly reduced in the transfectant cells compared with the control cells. These data suggest that down-regulation of p185 HER2 blocks CD44s expression and subsequent adhesion function. Our findings also indicate that the CD44s-p185 HER2 interaction is both functionally coupled and biosynthetically regulated. We believe that direct “cross-talk” between these two surface molecules (i.e. CD44s and the p185 HER2 ) may be one of the most important signaling events in human ovarian carcinoma development.

CD44 is a transmembrane glycoprotein that is widely distributed in different cell types and tissues (for reviews see Refs. 1 and 2). The extracellular domain of CD44 is extensively modified by N,O-glycosylation and glycosaminoglycan addition (2,3) and is responsible for the binding of extracellular matrix materials such as hyaluronate and collagen (2)(3)(4) as well as fibronectin (5). The cytoplasmic domain of CD44 contains at least one ankyrin binding site (6). Post-translational modification of the CD44 cytoplasmic domain by either acylation (7), protein kinase C (8), or GTP binding (9) enhances the binding between CD44 and ankyrin. The transmembrane interaction between CD44 and the cytoskeleton appears to play an important role in the early signal transduction events leading to cell activation (10).
The gene encoding the CD44 protein (located on human chromosome 11) contains 20 exons, of which only 10 are expressed in the standard form (CD44s). 1 The remaining 10 exons are expressed in different combinations in the extracellular domain of variant isoforms of the protein (CD44v), generated by alternative splicing of the mRNA (11,12). Some of these CD44v isoforms are selectively expressed on certain tumor cell surfaces during metastasis (13)(14)(15). In addition, high levels of CD44s/CD44v expression have been correlated with the progression of various carcinomas (13)(14)(15). At the present time, information concerning possible factors involved in triggering CD44v formation and CD44s overexpression during tumorigenesis or metastasis is very limited.
Ovarian carcinoma is the most lethal tumor of the female genital tract and continues to be the major cause of mortality in female cancer patients. Ovarian cancer has a unique pattern of spreading, which is primarily by intraperitoneal seeding (16). Recently, it has been shown that ovarian cancer cells express CD44s predominantly, which causes very strong adhesion to peritoneal mesothelium (17,18). It has also been reported that a significant reduction in tumor implants occurred in nude mice 5 weeks after intraperitoneal injection of ovarian cancer cells incubated with anti-CD44 antibody compared with injected cells pretreated with antibodies to other cell surface proteins (17,18). These findings suggest that CD44 plays an important role in the implantation of ovarian cancer metastasis.
The HER2 oncogene (also called c-erbB-2 or neu) encodes a 185-kDa (p185 HER2 ) membrane protein that contains a single transmembrane spanning region, two cystine-rich extracellular domains and a tyrosine kinase-associated cytoplasmic domain (19). This protein belongs to the epidermal growth factor receptor subgroup of the receptor-linked tyrosine kinase superfamily (20). Overexpression and amplification of HER2 oncogenes have been found to correlate with poor survival of many known cancers including ovarian cancer (21,22). Hung and co-worker (23) have shown that the HER2 oncogene is also overexpressed in certain ovarian carcinoma cell lines, such as SKOV3.ipl, shown to display high tumorigenic and metastatic potential. Therefore, coexpression of both CD44s and p185 HER2 appears to be closely associated with ovarian cancer progression.
In this study we have addressed the question of whether there is any interaction between CD44s and p185 HER2 in ovarian tumor cells using various human ovarian carcinoma cell lines including SKOV3.ipl (established from ascites of a nu/nu mouse given an intraperitoneal injection of SKOV-3 human ovarian carcinoma cell line) and derivatives (e.g. two stable transfectant cell lines designated as SKOV3.ipl.E1A (expressing E1A genes) and SKOV3.ipl.Efs (E1A frameshift mutants)) as model systems. Our results clearly indicate that these two surface molecules (i.e. CD44s and p185 HER2 ) are physically linked, functionally coupled, and biosynthetically regulated together. We proposed that the interaction between CD44s and p185 HER2 may be critically important for the onset of tumorigenesis and spreading during ovarian tumor development.

MATERIALS AND METHODS
Cell Lines and Culture-The SKOV3.ipl cell line was established from ascites that developed in a nu/nu mouse given an intraperitoneal injection of SKOV-3 human ovarian carcinoma cell line (obtained from the American Type Culture Collection) as described previously (23). Cells were grown in Dulbecco's modified Eagle's medium, F12 medium supplement (Life Technologies, Inc.) supplemented with 10% fetal bovine serum.
DNA Transfection-Two stable transfectant cell lines, SKOV3. ipl.E1A transfectants (expressing E1A genes) and E1A frameshift mutants (SKOV3.ipl.Efs) were prepared as a control cell line (to make sure the changes in transformation phenotypes (if any) in ipl.E1A transfectants were not due to the selection process or to transfection of the plasmids and the pSV2-neo gene) (23).
Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) and Southern Blot Analysis-Total RNA was extracted from Ϸ1 g of human ovarian cells (e.g. SKOV3.ipl) and processed for RT-PCR (using specific primer pairs (e.g. exon 5 and 15 primer pairs)) and Southern blot analyses as described previously (15).
Northern Blot Analysis-Twenty g of total RNA isolated from SKOV3.ipl cells was separated on a 1.2% agarose-formaldehyde gel, transferred to nylon membrane, and hybridized to the CD44s cDNA probe at 42°C for 16 h.
Cell Surface Labeling Procedures-Human ovarian carcinoma cells (SKOV3.ipl) and derivatives (e.g. SKOV3.ipl.E1A2 and SKOV3.ipl.Efs) suspended in PBS were surface-labeled using the following biotinylation procedure. Briefly, cells (10 7 cells/ml) were incubated with sulfosuccinimidobiotin (Pierce) (0.1 mg/ml) in labeling buffer (150 M NaCl, 0.1 M HEPES (pH 8.0) for 30 min at room temperature. Cells were then washed with PBS to remove free biotin. Subsequently, the biotinylated cells were used for anti-CD44-mediated immunoprecipitation as described below.
Immunoprecipitation and Immunobloting Techniques-Surface-biotinylated cells (5 ϫ 10 5 cells) were washed in 0.1 M PBS (pH 7.2) and solubilized in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100 buffer followed by sequential wheat germ agglutinin-Sepharose chromatography and immunoprecipitation by rat anti-CD44 antibody plus goat anti-rat IgG. The immunoprecipitated material was solubilized in either nonreducing or reducing SDS sample buffer and analyzed by SDS-PAGE (with 6 or 7.5% gel). Separated polypeptides were then transferred onto nitrocellulose filters. After blocking nonspecific sites with 3% bovine serum albumin, the nitrocellulose filters were incubated with ExtrAvidin-peroxidase (Sigma) or with anti-p185 HER2 antibody (5 g/ml) plus peroxidase-conjugated goat anti-mouse IgG (1:10,000 dilution) followed by an addition of peroxidase substrate (Pierce). The blots were developed using Renaissance chemiluminescence reagent (Amersham Life Science, Inc., UK) according to the manufacturer's instructions.
In some cases, surface-biotinylated SKOV3.ipl cells (treated with hyaluronic acid (HA) (50 g/ml) for 1 h, or pretreated with anti-CD44 followed by HA treatment (50 g/ml) for 1 h, or untreated) were processed for anti-CD44-mediated immunoprecipitation to obtain the CD44s⅐p185 HER2 complex. This material was then analyzed by SDS-PAGE under reducing conditions, transferred to the nitrocellulose fil-ters, and incubated with ExtrAvidin peroxidase (Sigma) to detect surface-biotinylated CD44 and/or anti-phosphotyrosine antibody plus peroxidase-conjugated goat anti-mouse IgG (1:10,000 dilution) to detect p185 HER2 tyrosine kinase activity. In some experiments, cells (e.g. SKOV3.ipl.E1A2 and SKOV3.ipl.Efs) were immunoprecipitated with anti-CD44 or immunoblotted with various immunoreagents such as mouse anti-p185 HER2 antibody, anti-E1A (M73) antibody, and anti-␣actin followed by incubation with peroxidase-conjugated goat antimouse IgG (1:10,000 dilution) at room temperature for 1 h. After an addition of peroxidase substrate, the blots were developed using Renaissance chemiluminescence reagent according to the manufacturer's instructions.
Protein Biotinylation-After SDS-PAGE, the 280-kDa gel band that was detected under nonreducing conditions was excised, eluted from the gel, and further labeled using the following biotinylation procedure. Briefly, eluted 280-kDa protein was incubated with sulfosuccinimidobiotin (0.1 mg/ml) in labeling buffer (150 M NaCl, 0.1 M HEPES (pH 8.0) for 30 min at room temperature followed by extensive dialysis against PBS buffer (0.1 M phosphate buffer (pH 7.5) and 150 mM NaCl). This biotinylated material was analyzed by SDS-PAGE under reducing conditions, transferred to the nitrocellulose filters, and incubated with ExtrAvidin peroxidase. After an addition of peroxidase substrate, the blots were developed using Renaissance chemiluminescence reagent according to the manufacturer's instructions. In some cases, the biotinylated material was analyzed by SDS-PAGE under reducing conditions followed by immunoblotting/immunoprecipitation with anti-p185 HER2 and/or anti-CD44 antibody according to the procedures described above.
Double Immunofluorescence Staining-Human ovarian carcinoma cells such as SKOV3.ipl were washed with PBS buffer and fixed by 2% paraformaldehyde. Fixed cells were then incubated with rat anti-CD44 antibody (10 g/ml) and fluorescein-labeled goat anti-rat antibody (10 g/ml) to detect surface CD44 expression. Subsequently, these fluorescein-labeled cells were rendered permeable by ethanol treatment and stained with monoclonal mouse anti-p185 HER2 followed by rhodamineconjugated goat anti-mouse IgG. To detect nonspecific antibody binding, fluorescein-labeled cells were incubated with normal mouse IgG followed by rhodamine-conjugated goat anti-mouse IgG. No labeling was observed in such control samples.
The fluorescein-and rhodamine-labeled samples were examined with a confocal laser scanning microscope (MultiProbe 2001 inverted confocal laser scanning microscope system, Molecular Dynamics, Sunnyvale, CA).
Cell Adhesion Assay-Human ovarian carcinoma cell lines including SKOV3.ipl and derivatives (e.g.SKOV3.ipl.E1A2 and SKOV3.ipl.Efs) were metabolically labeled with Tran 35 S label (20 Ci/ml) as described previously (10). After labeling, the cells were washed in PBS and incubated in PBS containing 5 mM EDTA at 37°C to obtain a nonadherent single-cell suspension. Labeled cells (9.1 ϫ 10 5 cpm/10 5 cells) were incubated on HA-coated plates (prepared as described previously) (10) at 4°C for 30 min either alone or in the presence of a unique monoclonal rat anti-CD44 antibody (50 g/ml). After incubation, the wells were washed three times in PBS, the adherent cells were solubilized in PBS containing 1% SDS, and the well-bound radioactivity was determined by liquid scintillation counting.
In Vitro Cell Growth Assays-SKOV3.ipl cells (5 ϫ 10 3 cells/well) (treated with HA (50 g/ml), pretreated with anti-CD44 followed by HA treatment (50 g/ml), or untreated) were plated in 96-well culture plates in 0.2 ml of Dulbecco's modified Eagle's medium, F12 medium supplement containing either 0.5% fetal bovine serum or no serum for 24 h at 37°C in 5%CO 2 , 95% air. In each experiment, a total of five plates (10 wells/treatment (e.g. HA treatment or anti-CD44 plus HA treatment or no treatment)/plate) was used. Experiments were repeated three times. The in vitro growth of these cells was analyzed by measuring increases in cell number using the 3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyl tetrazolium bromide assays (CellTiter 96 R nonradioactive cell proliferation assay according to the procedures provided by Promega). Subsequently, viable cell-mediated reaction products were recorded by a Molecular Devices (Spectra Max 250) enzyme-linked immunosorbent assay reader at a wavelength of 450 nm.

CD44 Expression in Human Ovarian
Tumor Cells-The expression of CD44 is known to be closely correlated with the metastatic and proliferative behavior of a variety of tumor cells including human ovarian tumor cells (13-15, 17, 18). To examine the expression of CD44 transcripts at the mRNA level, total RNA from ovarian tumor cells (SKOV3.ipl) was isolated and analyzed by RT-PCR with exon 5 and 15 primer pairs (designed to amplify the open reading frame region of CD44 involved in the alternative splicing of several exons) (15). By employing ethidium bromide staining (Fig. 1A, lane 1) of agarose gels followed by Southern blot hybridization (Fig. 1A, lane 2), we have identified a major PCR product of 156 base pairs that comigrates with an amplimer of CD44s (the standard form) in human ovarian tumor cells (e.g. SKOV3.ipl). Northern blot analysis also reveals a single 2.8-kilobase CD44s transcript that is expressed in these cells (Fig. 1B). Nucleotide sequence data confirm that this transcript represents the CD44 standard form, CD44s (data not shown). CD44 expression at the mRNA level does not always correlate with cellular protein expression. Therefore, it is important to further characterize CD44 protein expression in the human ovarian tumor cells. To examine CD44 expression on the surface of ovarian tumor cells, we have utilized surface biotinylation techniques and a specific monoclonal anti-CD44 antibody that recognizes the standard form of CD44 (CD44s) as well as other variant isoforms (15). Our results indicate that a single surface-biotinylated polypeptide (molecular mass Ϸ 85 kDa) displaying immunological cross-reactivity with CD44s is preferentially expressed on the cell surface of human ovarian cells (Fig. 1C, lane 1). No CD44s-containing material is observed in control samples when normal rat IgG is used (Fig. 1C, lane 2). Therefore, we believe that the major CD44 isoform expressed on the surface of the human ovarian tumor cells (SKOV3.ipl) is CD44s. These results are consistent with previous findings indicating ovarian cancer cells express predominantly CD44s (17,18).
Analysis of a Complex Formed between CD44s and p185 HER2 -Both CD44s and the p185 HER2 oncogene are often overexpressed (as much as 100-fold) in human tumor cells as a consequence of gene amplification and/or transcriptional regulation (15,24,25). In this study we addressed the question regarding whether there is an interaction between CD44s and p185 HER2 in human ovarian tumor cells. Using the nonionic detergent Triton X-100 to extract surface-biotinylated cells followed by sequential wheat germ agglutinin-Sepharose column chromatography and anti-CD44-mediated immunoprecipitation, we found that only one surface-biotinylated 85-kDa CD44s band is revealed under reducing conditions by SDS-PAGE (Fig. 2, lane 1). However, under nonreducing conditions, we have detected a large polypeptide (Ϸ280 kDa) that is coisolated with 85-kDa CD44s (Fig. 2, lane 2). To further characterize this large polypeptide, the 280-kDa protein was first eluted from the nonreducing SDS-PAGE gel and processed for biotinylation labeling. Under reducing conditions, we found that this biotinylated 280-kDa polypeptide is readily reduced into two proteins (Fig. 2, lane 3). One corresponds to p185 HER2 (185 kDa) and the other is CD44s (85 kDa) as verified by anti-p185 HER2 (Fig. 2, lane 4) and anti-CD44s immunoblot (Fig.  2, lane 5), respectively. Preliminary data indicate that the p185 HER2 associated with the plasma membrane of SKOV3.ipl cells is not surface-biotinylated. However, if p185 HER2 is denatured by SDS treatments, the sites for biotinylation of this molecule become available. This is the reason why we eluted the 280-kDa complex (Fig. 2, lane 2) and biotinylated again to detect both p185 HER2 and CD44s (Fig. 2, lane 3). In addition, we carried out anti-p185 HER2 or anti-CD44s-mediated precipitation followed by anti-CD44s immunoblot or anti-p185 HER2 immunoblot, respectively, using reducing SDS-PAGE analyses.
Our results indicate that the CD44s band is revealed in anti-p185 HER2 -immunoprecipitated materials (Fig. 2, lane 6). The p185 HER2 band can also be detected in the anti-CD44s-immunoprecipitated materials (Fig. 2, lane 7). These findings clearly indicate that CD44s and p185 HER2 are closely associated with each other in a complex involving interchain disulfide bonds in Based on the predicted sequence obtained from cDNA cloning of the human CD44s gene, several cysteine residues have been identified, including 6 -7 cysteine residues at the external domain, one Cys (amino acid 286) in the transmembrane region, and one Cys (amino acid 295) in the cytoplasmic domain (11). The functional significance of the cysteine residues at the external domain of CD44s is not clear at the present time. However, point mutation of Cys 286 (but not Cys 295 ) of human CD44s has been reported to cause a reduction (or loss) of HA binding (26). Certain cysteine residues (e.g. Cys 286 (in the transmembrane domain) and/or Cys 295 (in the cytoplasmic domain)) in mouse CD44s have been shown to play an important role in signal transduction and cell adhesion. For example, these residues appear to be involved in CD44s fatty acylation required for CD44s and ankyrin interaction (7). Therefore, it is possible that some of these cysteine residues of CD44s may also be involved in the disulfide linkage with one or more cysteine residues in p185 HER2 . The question of which cysteine residues in human CD44s are directly responsible for forming disulfide linkages with p185 HER2 is currently under investigation in our laboratory. The association of p185 HER2 with other surface molecules via disulfide linkages has been reported previously (27). Further analyses using double immunofluorescence labeling and confocal microscopy also support the notion that CD44s (Fig. 3A) and p185 HER2 (Fig. 3B) are co-localized at the periphery of ovarian tumor cells (Fig. 3C). Therefore, we believe that CD44s and p185 HER2 are physically linked in a complex in vivo in ovarian tumor cells such as SKOV3.ipl.
Effects of HA on CD44s-associated Function-HA has been shown to play an important role in several important physiological functions such as maintaining cartilage integrity, balancing homeostasis of water and plasma proteins in the intercellular matrix, and promoting mitosis and cell migration (28). In addition, degradation products of HA containing 3-25 di-saccharide units are known to promote angiogenesis, which involves cell proliferation, migration, and tubule formation (29 -31). CD44 is the major hyaluronan cell surface receptor (32) and cellular adhesion molecule in many different cell types (1,2). Specific HA binding motifs have been identified and localized in the extracellular domain of a number of CD44 isoforms (33).
In this study we examined the ability of ovarian tumor cells to adhere to HA-coated plates. Our data indicate that ovarian tumor cells display a high level of CD44-specific cell adhesion to HA-coated plates (Table I). After treatment with specific rat anti-CD44 antibody, the cells were largely inhibited from displaying HA-mediated cell adhesion (Table I). These results confirm that CD44s is a hyaluronan receptor required for HAmediated cell adhesion in these ovarian cells.
Adenovirus 5 E1A (E1A) products have been shown to be involved in regulating the expression of a number of important regulatory molecules (e.g. p185 HER2 and nm23) in fibroblast model systems (34,35). For example, the E1A products act as effective suppressors of invasion and metastasis in p185 HER2overexpressed 3T3 mouse fibroblasts (34). In addition, E1A is involved in elevating the expression of nm23 (a metastatic tumor suppressor gene) in ras-transfected rat embryo fibroblasts (35). These results suggest that E1A genes play an important role in suppressing certain cellular events leading to the metastatic cascade. Previously, it has been shown that introduction of the E1A gene into SKOV3.ipl ovarian cancer cells (e.g. SKOV3.ipl.E1A2) results in the suppression of p185 HER2 -induced metastatic properties, including tumor cell adhesion and invasion (23,35).
In this study we determined the effects of E1A genes on regulating CD44s and p185 HER2 interaction in ovarian tumor cells. Our results indicate that control cells (SKOV3.ipl.Efs) without E1A (Fig. 4, A-D, lane 1) express very high levels of p185 HER2 (Fig. 4B, lane 1). However, in transfectants (e.g. SKOV3.ipl.E1A2) (Fig. 4, A-D, lane 2) containing E1A (Fig. 4A,  lane 2)), p185 HER2 oncogene expression is significantly suppressed (Fig. 4B, lane 2). These results are consistent with previous findings reported by Hung and co-workers (22,23). Most importantly, CD44s expression in those transfectants containing E1A is also concomitantly decreased (Fig. 4C, lane 2) as compared with that in the control cells (Fig. 4C, lane 1). Our recent results indicate that up-regulation of p185 HER2 and CD44s promotes CD44s-mediated cell adhesion (36). The fact that CD44s-mediated cell adhesion to HA-coated plates dis- FIG. 3. Double immunofluorescence staining and confocal microscopic analysis of CD44 and p185 HER2 in human ovarian tumor cells (SKOV3.ipl). SKOV3.ipl cells were first fixed with 2% paraformaldehyde and processed for double immunofluorescence staining using anti-CD44 and anti-p185 HER2 , respectively. The labeled samples were then examined by confocal laser microscope according to the procedures described under "Materials and Methods." A, localization of CD44 in SKOV3.ipl cells using rat anti-CD44 antibody followed by fluorescein-labeled goat anti-rat IgG. B, localization of p185 HER2 in SKOV3.ipl cells using mouse anti-p185 HER2 antibody followed by rhodamine-conjugated goat anti-mouse IgG. C, co-localization of surface CD44 and p185 HER2 using rat anti-CD44 antibody and mouse anti-p185 HER2 antibody followed by fluorescence-labeled goat anti-rat IgG and rhodamine-labeled goat anti-mouse IgG, respectively.  (Table I) suggests that down-regulation of these two molecules (e.g. CD44s and p185 HER2 ) blocks CD44s-mediated cell adhesion function. Other cellular proteins, such as actin, do not show any detectable changes in the E1A transfectants (Fig. 4D, lane 2) as compared with the untransfected cells (Fig. 4D, lane 1). This tightly coupled interaction between CD44s and p185 HER2 may be very important for the early onset of the metastatic cascade during ovarian cancer. HA is also known to be involved in a number of pathophysiological processes. For example, high levels of HA in solid tumors (e.g. breast, lung, hepatic, and bladder tumors) appear to be closely associated with tumor progression and metastasis (37,38). In this study, we found that CD44s-associated p185 HER2 tyrosine kinase activity (Fig. 5A, lower panel, lane 1 and subsequent ovarian tumor cell growth are significantly stimulated by HA treatment, as detected by anti-phosphotyrosine-mediated immunoblotting techniques (Fig. 5A, upper  panel, lane 1) and an in vitro growth assay (Fig. 5B, a). We believe that HA-triggered p185 HER2 kinase activation (Fig. 5A  upper panel, lane 1) and cell growth up-regulation (Fig. 5B, a) are CD44s-specific since control samples (either without HA treatment (Fig. 5, A, upper panel, lane 2 and B, b) or pretreated with anti-CD44 to inhibit HA binding effects (10) (Fig. 5, A,  upper panel, lane 3 and B, c) followed by HA treatments) display very low levels of CD44s (Fig. 5, A, lower panel, lanes 2  and 3)-associated p185 HER2 tyrosine kinase activity (Fig. 5A,  upper panel, lanes 2 and 3) and tumor cell growth (Fig. 5, B, b  and c). These findings clearly demonstrate that both CD44sassociated p185 HER2 kinase activity and ovarian tumor cell growth are stimulated by HA. Recently, we found that the formation of a CD44⅐p185 HER2 complex plus HA-mediated p185 HER2 kinase activation and cell growth also occur in other ovarian tumor cells (e.g. SKOV-3 and CAOV, etc.) as well as breast tumor cells (e.g. Met-1 and MDA231, etc.). 2 A number of other molecules (including epidermal growth factor, neuregulins, neu-activating factor, ascites sialglycoprotein-2, a factor from bovine kidney (NEL-GF), and unknown factors in serum) have been implicated as potential agonists for the activation of p185 HER2 -linked tyrosine kinase activities (20). Therefore, we believe that the results of this study provide new evidence that the physiological ligand for CD44s, HA, may also play an important role in activating CD44s-associated p185 HER2 kinase activity required for the onset of tumor cell growth and spreading during ovarian tumor development.