Osteopontin Is an Autocrine Mediator of Hepatocyte Growth Factor-induced Invasive Growth¹

HGF,³ also known as “scatter factor,” is a mesenchymal cytokine that acts on epithelial and endothelial cells by promoting a highly integrated biological program, hereafter referred to as “invasive growth.” This program involves coordinated control of basic cellular functions including dissociation and migration (“scattering”), invasion of extracellular matrix, proliferation, prevention of apoptosis, and polarization (1). As a consequence, complex developmental processes take place, such as epithelial branched morphogenesis and angiogenesis (2, 3). Aberrant activation of the invasive growth program in cancer cells confers invasive and metastatic ability (4). Interestingly, oncogenic activation by overexpression or point mutation of the gene encoding the tyrosine kinase receptor for HGF, c-MET (5), is involved in the progression of tumors toward the invasive-metastatic phenotype (6, 7). The invasive growth response to HGF requires days for fulfillment and is likely to involve transcriptional regulation of specific target genes. Indeed, on activation by HGF, the Met tyrosine kinase concomitantly regulates multiple signal transduction pathways including ras, phosphatidylinositol-3-kinase, phospholipase C-γ, and STAT (8, 9), which in turn have been shown to control the transcriptional status of the cell (10). Furthermore, HGF-stimulated scattering of epithelial cells is ablated by the inhibition of protein neosynthesis (11). Identification of HGF-regulated genes is, therefore, essential to clarifying the molecular basis of epithelial morphogenesis and tumor invasive growth. Toward this aim, we used cDNA microarrays to explore the transcriptional response to HGF of MLP-29 mouse embryo liver cells (12). The most highly induced transcript was identified as the product of the gene-secreted phosphoprotein 1 (spp1; Ref. 13), encoding an arginine-glycine-aspartate (RGD)-containing glycoprotein known as OPN (14), or early T-cell activation 1 (ETA-1; Ref. 15). Interestingly, OPN and its cell surface receptors, CD44 (16) and α_v-containing integrins (17, 18, 19), have been implicated in the progression of cancer toward the invasive-metastatic phenotype (20, 21, 22). It can, therefore, be hypothesized that OPN induction is a key event in the scattering and invasive growth responses elicited by HGF. To test this hypothesis, we evaluated: (a) the presence of OPN receptors at the surface of MLP-29 cells; (b) the effect of HGF or EGF stimulation on the ability to adhere to OPN; and (c) the consequences of constitutive OPN expression on the biological responses to HGF or EGF. EGF was included in the study because it does not induce scattering or invasive growth of MLP-29 cells (12). We found that HGF specifically promotes OPN transcription and its interplay with the CD44 receptor. OPN is thus a major player in the HGF-induced phenotype.

MLP-29 cells were cultured as described previously (12). For the scatter assay and expression analysis, cells were allowed to reach confluence in 10-cm dishes, and, subsequently, the serum concentration was lowered from 10 to 2% for 3 days before removing the serum for 24 h. From each 10-cm dish, the cells were plated in two 15-cm dishes in 2% bovine serum (Hyclone). Collagen cell cultures were performed as described previously (12). Stimulations were performed with recombinant HGF (23) or EGF (Sigma Chemical Co.) in the presence of 0.1 or 2% bovine serum. 293T cells were cultured in Iscove’s modified Dulbecco’s medium (IMDM, Sigma Chemical Co.) supplemented with 10% FBS in a humidified atmosphere of 5% CO₂. Recombinant mouse OPN and goat antimouse OPN antibody (AF808) were from R&D Systems. Swine antigoat IgG (H+L) R-phycoerithrin-conjugated antibody and rat monoclonal antibody to mouse CD44 were both from Caltag Laboratories. The rat antibody to mouse α_v (clone H9.2B8) was from PharMingen and the FITC-conjugated donkey antirat IgG (H+L) was from Jackson ImmunoResearch Laboratories, Inc. Human plasmatic FN was from Sigma Chemical Co.

Total RNA was extracted by the guanidinium thiocyanate protocol according to Chomczynski and Sacchi (24). Poly(A)⁺ RNA was isolated from total RNA by two subsequent rounds of purification on Oligotex resin (Quiagen), according to the manufacturer’s protocol. The comparative analysis of cells stimulated with HGF or with control medium was performed using Gene Expression Microarrays from Incyte Genomics containing 8735 murine genes, as described previously (25, 26). Raw expression data were entered into Excel (Microsoft) for subsequent normalization, averaging, and comparative analysis. For Northern blot analysis, 5 μg of total RNA were separated by electrophoresis on 0.8% denaturing agarose gels and transferred to Hybond-N membranes (Amersham). Probes were labeled with [α-³²P]dCTP using the Megaprime Labeling kit (Amersham), according to the manufacturer’s protocol. Hybridization was carried out overnight at 42°C in ULTRAhyb hybridization solution (Ambion). The membranes were washed twice in 1× SSC-0.1% SDS at room temperature for 15 min, and twice in 0.5× SSC-0.1% SDS at 65°C for 15 min. Bound radioactivity was detected and quantified using a STORM 840 PhosphorImager (Molecular Dynamics). The filter was subsequently rehybridized with mouse glyceraldehyde-3-phosphate dehydrogenase cDNA as a probe, to normalize for the amount of loaded RNA.

MLP-29 cells, incubated for different times at 37°C in the absence or in the presence of 40 units/ml HGF or of 10 ng/ml EGF, were washed twice with cold PBS and lysed in boiling Laemmli buffer (27). Cell lysates were incubated for 5 min at 94°C, sonicated, clarified by centrifugation at 13,000 rpm at 4°C for 15 min and quantified by the BCA assay (Pierce). Proteins were resolved on SDS-PAGE and subsequently transferred onto Hybond-C filters (Amersham). Filters were blocked with 10% BSA for 1 h at 45°C and were probed with OPN antibodies diluted in TBS (Tris-buffered saline)-5% BSA, for 16 h at 4°C. After extensive washing, immunocomplexes were detected with horseradish peroxidase-conjugated secondary antiserum followed by enhanced chemiluminescence reaction (ECL; Amersham).

Immunostaining was performed according to Lin et al. (28), with minor modifications. Briefly, cells were harvested with 10 mm EDTA in PBS, washed in PBS and incubated on ice for 40 min with antibodies to either CD44 or α_v in PBS-1% bovine serum. Cells were washed twice with the same buffer and incubated for 40 min on ice with fluorescently labeled secondary antibodies. After two washes in PBS-1%FBS, cells were analyzed by flow cytometry with a Becton Dickinson FACS Calibur.

Microtiter plates (96-well; Nunc, Naperville, IL) were coated with FN or OPN (2.5–20 μg/ml) in PBS. Proteins were allowed to bind overnight at 4°C before washing the wells with PBS and blocking for 2 h at 37°C with 2% heat-denatured BSA in PBS. Starved cells were harvested and added to the wells (10⁴ cells/0.1 ml) in the presence or absence of growth factors and antibodies. After incubation for 45 min at 37°C, the wells were gently washed with PBS. Adherent cells were fixed in 11% glutaraldehyde in PBS and stained with 0.1% crystal violet in 20% methanol. Cell numbers were obtained by counting all of the cells in the well using a phase-contrast light microscope at ×25. Data presented are the means ± SD of triplicate wells from two experiments (three experiments in the case of CD44 antibody treatments).

Mouse OPN coding sequence was obtained by reverse transcription-PCR from MLP-29 cells using the sense primer 5′-AGCTAGCCAAGGACTAACTACGACCATGAGA-3′ and antisense primer 5′-ACTCGAGTTCGTTGACCTCAGAAGATGAACTCTC-3′, designed to amplify the OPN coding sequence excluding the 3′ termination codon and including restriction sites for NheI and XhoI at the 5′ and 3′ ends, respectively. Using NheI and XhoI, the OPN PCR product was incorporated in the pEGFP-N1 plasmid (Clontech), upstream of the enhanced GFP coding sequence and in the same translational frame. The resultant plasmid, named pOPN-GFP, was sequence-verified. The OPN-GFP fusion cDNA was excised from pOPN-GFP with NheI and NotI, and subsequently bluntized and ligated into the lentiviral vector pRRLsin.PPT.hCMV-GFP.pre (29) previously digested with BamHI and SalI to excise GFP, bluntized, and dephosphorylated. The resulting transfer vector was named pLC-OG, which stands for Lentiviral (vector) CMV (promoter) OPN-GFP (transgene). The virus was produced by transient transfection of 293T cells with pLC-OG together with the VSV-G and pCMVΔR8.93 plasmids, as described previously (29). Viral p24 concentration was determined by HIV-1 p24 Core profile ELISA (NEN Life Science Products). Transduction experiments were performed by adding a 1:2 dilution of the virus-containing 293T supernatant onto 10⁵ cells in six-well plates (Costar) in the presence of polybrene (8 μg/ml). Transduced cells were checked by flow cytometry and Western blot for expression of the recombinant protein.

Cells were plated onto 24-well plates (Costar) containing 1.4-cm² glass coverslips. After incubation in the presence or absence of HGF (40 units/ml), cells were fixed in a freshly prepared solution of 3% formaldehyde (from paraformaldehyde) and 2% sucrose in PBS (pH 7.6) for 5 min at room temperature. After saturation with PBS-2% BSA for 15 min, CD44 antibodies were layered onto cells and incubated in a moist chamber for 30 min. After rinsing in PBS-0.2% BSA, coverslips were incubated with rhodamine-tagged secondary antibody for 30 min. Coverslips were mounted in Mowiol 4–88 (Hoechst AG) and observed with a DM-IRB inverted photomicroscope (Leica) equipped with mercury short arc epifluorescence lamp. Images were captured using a cooled digital CCD Hamamatsu ORCA camera (Hamamatsu Photonics), digitally recorded and elaborated with ImageProPlus 4.0 imaging software (Media Cybernetics).

To detect genes that are induced on HGF stimulation, we used the mouse embryo liver cell line MLP-29, that specifically responds to HGF by scattering and invasive growth (12). We optimized the conditions for a reproducible scatter response in large-scale cultures so that RNA could be extracted in sufficient amounts for microarray analysis. We subsequently performed time course stimulation experiments to determine the optimal stimulation time and growth conditions. As shown in Fig. 1, a specific response to HGF was already detectable after 6 h of stimulation: the colony boundaries became irregular, and single cells protruded and initiated detachment and migration. At this time, neither EGF nor serum induced a similar effect. After 24 h, the HGF-induced scattering was complete, and EGF and serum induced minor morphological modifications of the colonies. We, therefore, considered 6 h as the best time point to detect transcriptional regulation of effectors that initiated cell scattering and would sustain it over the following hours. We also observed that optimal scattering of MLP-29 required, besides HGF, the presence of a minimal amount (1–2%) of serum in the medium (data not shown), and, therefore, performed HGF stimulation and subsequent gene expression profiling in both 0.1 and 2% serum. mRNA was prepared from cells that had been stimulated with HGF or with control medium for 6 h, and comparative analysis was performed using cDNA microarrays covering around 8500 murine genes. After averaging the results of the experiments produced in 0.1and 2% serum, the most highly induced gene was identified as OPN, with an induction level of 4.2 ± 0.7 (Fig. 2). The total number of genes regulated above 3-fold or 2-fold was, respectively, 6 and 22. As shown in Table 1, several other transcripts encoding matrix proteins were explored by the microarray, including FN, seven classes of laminins, nine types of procollagen chains, but none of them turned out to be significantly regulated. Moreover, no gene involved in matrix remodeling was found to respond to HGF above the 2-fold threshold of significance (not shown). To validate and further characterize OPN gene regulation, we performed time course Northern blot analysis on cells stimulated with HGF and EGF (Fig. 3,A). OPN mRNA induction by HGF was already detectable after 1 h, peaked ∼9-fold at 6 h, and decreased almost to baseline at 24 h. Intriguingly, EGF was also found to induce OPN, although with a different, long-term kinetics. Induction of OPN by HGF and EGF was confirmed at the protein level by Western blotting (Fig. 3 B). In this case, high levels of OPN protein remain evident in HGF-stimulated cells at 24 h, probably because of its accumulation in the extracellular matrix, whereas EGF-stimulated cells display a lower OPN accumulation. Real-time PCR analysis showed that OPN induction by HGF is not limited to MLP-29 cells, but also occurs in A549 human lung cancer cells (2-fold at 24 h) that express the HGF receptor and scatter in response to HGF (not shown). Both Northern and Western blot analyses indicate that the microarray underestimated OPN induction. It is, therefore, conceivable that many of the genes that did not reach the 2-fold threshold of significant regulation (52 above 1.5-fold) are indeed regulated by HGF. To address this point, a more extensive microarray analysis is currently in progress.

The mechanism by which OPN may contribute to the responses elicited by HGF is by binding cells through two known receptors: α_V-containing integrins and CD44. Flow-cytometric analysis showed that whereas MLP-29 cells express high levels of CD44, they do not express detectable α_V at their surface (Fig. 4,A). We have previously shown that HGF activates integrins and renders epithelial cells capable of adhering to integrin-binding substrates, such as FN (30). To verify the ability of HGF to trigger a similar response in MLP-29 cells, we evaluated the ability of these cells to adhere to OPN- or FN-coated wells. Interestingly, HGF-stimulated cells displayed increased binding to OPN and FN, with respect to unstimulated and EGF-stimulated cells (Fig. 4,B). Specificity of adhesion to OPN was confirmed by the use of blocking anti-OPN antibodies that completely inhibited cell adhesion to OPN without interfering with adhesion to FN. Increased adhesion to OPN in response to HGF was also significantly impaired in the presence of anti-CD44 antibodies (Fig. 4 C). HGF stimulation for the duration of the adhesion assay (1 h) was not found to modify CD44 or α_V expression (not shown). These data indicate that HGF specifically induces adhesion of MLP-29 cells to OPN through the CD44 receptor.

To generate MLP-29 cells constitutively expressing OPN and directly detect the recombinant protein, we generated a fusion between OPN and GFP. We incorporated an OPN coding sequence lacking the stop codon in an enhanced GFP expression plasmid vector (pEGFP-N1; Clontech), thereby generating the OPN-GFP fusion, as illustrated in Fig. 5,A (see also “Materials and Methods”). The construct was first tested by transient transfection of 293T cells. Western blot analysis with anti-OPN antibodies confirmed that the transiently transfected cells expressed a novel band migrating at ∼95 kDa, corresponding to the sum of the molecular masses of OPN and GFP (Fig. 5B). We subsequently introduced OPN-GFP in a lentiviral vector (29) for stable, CMV-driven expression (Fig. 5,A). MLP-29 cells were subjected to four subsequent rounds of infection with high titers of OPN-GFP vector (multiplicity of infection, >1), so that the entire cell population expressed the recombinant protein without requiring clonal selection procedures. OPN-GFP expression was confirmed by Western blot analysis with anti-OPN antibodies (Fig. 5 B). The transduced cell population was named 29-OG, where OG stands for OPN-GFP. A control population (29-G) was generated by infection of MLP-29 cells with a lentiviral vector encoding GFP alone.

Using the GFP moiety incorporated in the fusion, we analyzed OPN-GFP localization in unstimulated and stimulated cells by fluorescence microscopy (Fig. 6,A). In the absence of HGF, OPN-GFP was accumulated in the inner portions of the colonies, forming matrix aggregates in the intercellular spaces. During HGF-induced scattering, the matrix aggregates were actively remodeled and kept attached to the moving cells. Double staining with CD44 antibody indicated that only a small fraction of CD44 is engaged with OPN even in HGF-stimulated cells. In particular, although the surface of lamellipodia exploring the surrounding environment was essentially devoid of OPN and rich in CD44, colocalization of the two molecules appeared more evident near the sites of emergence of long and thin cellular processes or at sites of cell-cell contact. It is, therefore, likely that OPN-CD44 interaction plays a role in the organization of the distinctive “cell-net” that is observed during scattering, in which motile, isolated cells are connected to each other by thin cytoplasmic bridges to form a net-like structure (see Fig. 1).⁴ Indeed, 29-OG cells were clearly more prone to scattering than control, 29-G cells. As illustrated in Fig. 6 B, 29-OG cells had a higher basal motility that was further increased by HGF stimulation, which indicated that OPN overexpression facilitates HGF-induced cell scattering. Conversely, simultaneous treatment with antibodies to OPN and CD44 significantly impaired the response of 29-G cells to HGF. Single antibody treatments, as well as administration of exogenous recombinant OPN, did not exert detectable effects (not shown). The partial inhibition observed may be a consequence of poor accessibility of the sites of autocrine OPN-CD44 interaction to the blocking antibodies. Alternatively, redundant signals may render OPN and/or CD44 partially dispensable for the HGF response (see “Discussion”).

The assay that best demonstrates how the multiple effects of HGF are indeed part of a single complex phenomenon is the induction of branching morphogenesis in MLP-29 cells grown in type I collagen gels (12). We, therefore, cultured 29-G and 29-OG cells in tridimensional type I collagen gels as described previously (Fig. 7; Ref. 12). The behavior of 29-G cells was similar to that of the parental line: on stimulation with HGF, they underwent an ordered morphogenetic process, forming unbroken and sometimes branching sprouts that retained contact with the colony of origin. Unstimulated 29-OG cells formed irregularly shaped colonies with protruding cells. HGF stimulation strongly increased their motility and invasiveness, inducing single cells to spread in all directions through the collagen matrix, so that the colonies of origin were no longer distinguishable. Specificity of the response to HGF was confirmed by stimulating 29-G and 29-OG cells with EGF, which induced only minor morphological changes. Also in this case, exogenous administration of recombinant OPN could not mimic the effect of endogenously overexpressed OPN (not shown), which indicates that autocrine interactions between OPN and CD44 have some specific advantage in promoting invasiveness.

This work defines OPN as a major transcriptional target of HGF in normal liver cells and assigns a relevant role for OPN in the biology of cell scattering and invasive growth. OPN is a highly phosphorylated sialoprotein of the extracellular matrix, with an unusually wide spectrum of cell surface receptors and biological activities. Interestingly, many of the basic cellular functions affected by OPN were known to be controlled also by HGF and its receptor, Met. By interacting with α_vβ₁, α_vβ₃, and α_vβ₅ integrins, OPN promotes cell attachment, spreading, nitric oxide production, and migration, and more complex events like vascular remodeling, bone mineralization, and tumor metastasis (21). OPN also binds the cell surface hyaluronate receptor CD44 (16), a protein that has been implicated, among the others, in cell-cell and cell-extracellular matrix interactions (31), lymphocyte extravasation and homing (32, 33, 34), tumor cell metastasis (20, 22), and regulation of hematopoiesis and apoptosis (35, 36, 37, 38, 39). OPN and Met are frequently overexpressed in a large variety of malignant cells (6, 40), and both have been implicated in cell invasive behavior and tumor progression. OPN expression increases the invasiveness of neoplastic cells (41), and a similar effect has been observed for Met (4, 42). During metastatic spread, neoplastic cells have to overcome anoikis, an apoptotic response to cell-cell and cell-matrix detachment (43, 44). In this regard, both OPN (28, 45, 46) and HGF (47) protect cells from apoptosis, which further explains how they promote invasive migration.

Interestingly, we recently found that HGF is involved in monocyte-macrophage activation (48), a process in which OPN plays a key role, as revealed by targeting experiments (49). Nevertheless, whereas HGF null mice die in utero, OPN null mice are viable (50), indicating that functional compensation may actually take place. This is a paradigm for cell adhesion molecules: for example, single mutations in the genes encoding ligands for the a_Vβ₃ integrin (OPN, thrombospondin-1, tenascin-C, and vitronectin) do not lead to developmental defects, whereas inactivation of the a_V receptor component is lethal (51).

The functional link between Met activation and OPN expression described herein provides a molecular basis for the strong biological correlations described above. Inspection of the OPN promoter sequence provides clues for identifying the signaling pathways mediating OPN regulation by HGF. The OPN promoter contains elements that are activated by HGF stimulation, like the AP-1 and AP-2 elements and multiple ETS-family binding sites (52, 53). Notably, the observed peak of OPN induction at 6 h concords with the genetic program triggered by HGF, which involves an early dissociation phase followed by the organization of branched and polarized structures. Tight regulation of OPN expression in this later phase was found to be crucial, because cells that constitutively overexpress OPN remained highly motile and invasive even at later times of HGF stimulation and were not capable of organizing tubular structures.

In partial agreement with data obtained from human mammary epithelial cells (41, 54), constitutive OPN expression is not sufficient for MLP-29 cells to fully exploit their motile-invasive potential: on HGF stimulation, a second signal from the activated Met tyrosine kinase renders them capable of adhering and responding to OPN with further enhancement of scatter and invasive growth. Interestingly, only this second signal turned out to be specific, because EGF is also capable of inducing OPN expression but does not promote the OPN-CD44 interplay.

Lack of α_v integrin expression indicates that the biological activity of OPN on MLP-29 cells is mediated by the CD44 receptor. Involvement of integrins in this cell type is not completely excluded by our data, because CD44 can cooperate with β₁-containing integrins in promoting cell adhesion to OPN (55). Further analysis will be required to clarify this point and to determine which CD44 isoform is responsible for the observed OPN effects.

An intriguing issue is how activation of the Met receptor by HGF renders MLP-29 cells competent for binding OPN through the CD44 receptor. Indeed, the following lines of existing evidence correlate HGF and CD44, and suggest that integrins, CD44, and the Met receptor may coexist at least transiently in a supramolecular membrane complex: (a) HGF amplifies CD44-induced integrin-mediated adhesion of colon cancer cell lines to endothelial cells (56); (b) membrane-localized constitutively activated Tpr-Met oncoprotein induces expression of hyaluronic acid and of a particular isoform of CD44 (57); (c) a CD44 splice variant efficiently binds HGF via its heparin sulfate side chain and facilitates HGF-induced Met activation and phosphorylation (58); and (d) both Met and CD44 may associate members of the Ezrin-Radixin-Moesin family, and this association affects cell polarity and spatial organization (59, 60). It is, therefore, conceivable that a supramolecular complex containing Met and CD44, together with other proteins, is stabilized and functionally activated by the concerted action of HGF and OPN. Because OPN, CD44, and Met are overexpressed in several types of tumors, the observed functional collaboration may be instrumental in promoting tumor growth and metastasis.

We thank Alberto Bardelli and Michela Riba for helpful discussion. The technical assistance of Lorenza D’Alessandro, Sabrina Arena, Raffaella Albano, Laura Palmas, and Giovanna Petruccelli is gratefully acknowledged. We thank Antonella Cignetto for secretarial help and Elaine Wright for editing the manuscript.