Interplay between αvβ3 Integrin and Nucleolin Regulates Human Endothelial and Glioma Cell Migration*

Background: Cell surface nucleolin (NCL) is a promising target for development of anticancer agents. Results: A novel pathway that includes ανβ3 integrin and leads to cell surface NCL localization and cell migration has been identified. Conclusion: ανβ3 can be used as a biomarker for the use of NCL antagonists. Significance: This pathway is active in endothelial and glioma cells, as well as in human glioblastomas. The multifunctional protein nucleolin (NCL) is overexpressed on the surface of activated endothelial and tumor cells and mediates the stimulatory actions of several angiogenic growth factors, such as pleiotrophin (PTN). Because αvβ3 integrin is also required for PTN-induced cell migration, the aim of the present work was to study the interplay between NCL and αvβ3 by using biochemical, immunofluorescence, and proximity ligation assays in cells with genetically altered expression of the studied molecules. Interestingly, cell surface NCL localization was detected only in cells expressing αvβ3 and depended on the phosphorylation of β3 at Tyr773 through receptor protein-tyrosine phosphatase β/ζ (RPTPβ/ζ) and c-Src activation. Downstream of αvβ3, PI3K activity mediated this phenomenon and cell surface NCL was found to interact with both αvβ3 and RPTPβ/ζ. Positive correlation of cell surface NCL and αvβ3 expression was also observed in human glioblastoma tissue arrays, and inhibition of cell migration by cell surface NCL antagonists was observed only in cells expressing αvβ3. Collectively, these data suggest that both expression and β3 integrin phosphorylation at Tyr773 determine the cell surface localization of NCL downstream of the RPTPβ/ζ/c-Src signaling cascade and can be used as a biomarker for the use of cell surface NCL antagonists as anticancer agents.

Nucleolin (NCL) 5 is a multifunctional protein with well characterized roles in the organization of nucleolar chromatin, packaging of pre-rRNA, rDNA transcription, and ribosome assembly (1). It also acts as a shuttling protein between the cytoplasm and nucleus (2) and is overexpressed on the plasma membrane of cancer (3)(4)(5), as well as activated endothelial (6 -8) cells. The importance of cell surface NCL has been suggested by studies showing that functional blockade or downregulation of cell surface NCL in endothelial cells inhibits migration and capillary-tubule formation (3,8) and causes endothelial cell apoptosis (7), whereas targeting NCL on the plasma membrane of cancer cells seems to be an effective way to inhibit cancer cell growth and angiogenesis in various in vitro and in vivo experimental models (3,9,10). The potential significance of targeting cell surface NCL is being proved by the fact that the guanosine-rich quadruplex-forming oligodeoxynucleotide AS1411 (11) and the NUCANT peptide (3,12) that interact with surface NCL, are currently being tested in Phase I/II clinical trials. The signals, however, that mobilize NCL from the nucleus to the cell surface still remain unclear and not all tumor cells express cell surface NCL.
Cell surface NCL interacts with receptors associated with malignancies, such as ErbB1, facilitating their activation and leading to enhanced cell growth (4). Moreover, it binds a variety of ligands that play critical roles in tumorigenesis and angiogenesis, such as hepatocyte growth factor (13), endostatin (14), tumor homing peptide F3 (6), laminin (15), P-selectin (16), and midkine (17). We have recently shown that NCL interacts with the heparin-binding growth factor pleiotrophin (PTN) in both chorioallantoic membrane vessels and human endothelial cells. This interaction is taking place on the cell surface and downregulation or blockade of cell surface NCL abolishes PTN-induced endothelial cell migration (18).
Integrins are cell surface heterodimeric receptors that mediate the physical and functional cell-cell or cell-matrix interactions. Among several integrins, ␣ v ␤ 3 is the most abundant and influential receptor regulating angiogenesis, and its tumor cell expression is correlated with disease progression in various tumor types (19). Growing evidence supports a central role for cooperative signaling between ␣ v ␤ 3 integrin and growth factor receptors, such as ErbB-2 (20), platelet-derived growth factor receptor ␤ (21,22), and vascular endothelial growth factor receptor 2 (21,23), mediating tumor cell adhesion, migration, invasion, and survival, as well as endothelial cell activation (19). In the same line, we have previously shown interaction of ␣ v ␤ 3 with the PTN receptor protein-tyrosine phosphatase ␤/ (RPTP␤/), which is required for PTN-induced endothelial cell migration (24).
The secreted heparin-binding growth factor PTN has established roles in cancer development, either directly by acting on cancer cells, or indirectly by affecting tumor angiogenesis (25). PTN binding to its receptor RPTP␤/ in endothelial cells leads to dephosphorylation and thus activation of c-Src (26), leading to Tyr 773 phosphorylation of ␤ 3 integrin and PTN-induced endothelial cell migration (24). Because both ␣ v ␤ 3 (24) and NCL (18) are required for PTN-induced cell migration, the aim of the present study was to investigate interplay between NCL and ␣ v ␤ 3 that regulates human endothelial and glioma cell migration. Our data show that ␣ v ␤ 3 expression and ␤ 3 phosphorylation at Tyr 773 , as a result of the RPTP␤//c-Src signaling cascade, induces cell surface localization of NCL through phosphoinositide 3-kinase (PI3K), shedding light into the mechanisms that underlie NCL overexpression on the surface of activated endothelial and cancer cells. Moreover, evidence is presented that ␣ v ␤ 3 integrin may be a useful biomarker to be used for identification of patients that could benefit from strategies targeting cell surface NCL.
RNA Interference-Cells were grown to 50% confluence in medium without antibiotics. Transfection was performed in serum-free medium for 4 h using annealed RNA for RPTP␤/ (26), NCL (18), ␣ v and ␤ 3 (24). JetSI-ENDO (Polyplus Transfection, Illkirch, France) was used as transfection reagent. Cells were incubated for another 48 h in serum-containing medium and lysed, serum starved, or fixed before further experiments. Double-stranded negative control siRNA (Ambion, Austin, TX) was also used in all experiments.
Transient Transfection-M059K cells were transfected with pCDNA3.1 vector, wild-type ␤ 3 , or wild-type ␣ v constructs using jetPEI HUVEC (Polyplus Transfection). At 50% confluence, cells were incubated with DNA and jetPEI-HUVEC in an N/p ϭ 5 ratio for 4 h at 37°C, as previously described (24). The transfection medium was replaced by fresh serum-containing medium, and 24 h later, the cells were examined for NCL expression in different cell compartments by Western blot or immunofluorescence analyses.
Western Blot Analysis-Proteins were analyzed by SDS-PAGE and transferred to Immobilon P membranes. Blocking was performed by incubating the membranes with Tris-buffered saline (TBS), pH 7.4, with 0.05% Tween (TBS-T), containing 5% nonfat dry milk in all cases. Membranes were incubated with primary antibodies for 16 h at 4°C under continuous agitation, washed 3 times with TBS-T, and incubated with secondary antibodies for 1 h at room temperature. Detection of immunoreactive bands was performed using the enhanced chemiluminescence (ECL) detection kit (Pierce Biotechnology, Rockford, IL). The protein levels that corresponded to the immunoreactive bands were quantified using the ImagePC image analysis software (Scion Corp., Frederick, MD).
Immunoprecipitation Assay-Cells were lysed with RIPA buffer, as previously described (24). Three mg of total protein were incubated with primary antibody for 16 h at 4°C under continuous agitation. Protein A-and protein G-agarose beads (Merck Millipore) were added, samples were further incubated for 2 h at 4°C, and beads with bound proteins were collected by centrifugation and washed twice with ice-cold PBS. Immuno-precipitated proteins were resuspended in 50 l of SDS loading buffer and analyzed by Western blot.
Immunohistochemistry-Human Brain Cancer Tissue Array (USBiomax Inc., Rockville, MD) sections were deparaffinized with xylene (3 ϫ 5-min incubations) followed by treatment with serial dilutions of ethanol (100, 100, 95, and 95%, 10 min each) and by two changes of ddH 2 O. Antigen unmasking was achieved by boiling the slides (95-99°C) for 10 min in 10 mM sodium citrate, pH 6.0. Sections were rinsed three times with ddH 2 O, immersed in 3% H 2 O 2 for 20 min, washed twice with ddH 2 O and once with TBST (TBS, 0.1% Tween 20), and blocked for 30 min with Image-iT FX signal enhancer (InVitrogen, Carlsbad, CA). Mouse anti-CD51/CD61 and rabbit anti-NCL antibodies (Abcam, Cambridge, MA) were diluted 1:50 and 1:100, respectively, in blocking solution (TBST, 5% normal goat serum, Cell Signaling Technology, Inc., Beverly, MA) and incubated with the sections overnight at 4°C. Following incubation with the antibodies, sections were washed three times with TBST and incubated for 1 h at room temperature with the corresponding Alexa fluorescent secondary antibodies diluted in blocking solution (1 g/ml). Sections were finally washed three times, 5 min each, with TBST and mounted with DAPIcontaining Vectashield mounting medium (Vector Labs, Burlingame, CA). Images were captured with a Nikon 80i Upright Microscope equipped with a Nikon Digital Sight DS-Fi1 color camera, using the Metamorph image acquisition software. All images were captured and processed using identical settings. Co-localization analysis was performed using Intensity Correlation Analysis plug-in within WCIF ImageJ software (30). Results are expressed with the Mander's overlap coefficient (ranges between 1 and 0, with 1 being high and 0 being low co-localization) and Pearson's correlation coefficient (value close to 1 indicates reliable co-localization).
Biotinylation of Cell Membrane Proteins-HUVEC were washed twice with PBS and incubated with 0.5 mg/ml of NHS-D-biotin for 40 min at 4°C, as previously described (24). Three mg of total proteins from the cell lysate were used for immunoprecipitation with primary antibody or 0.5 mg of streptavidin magnetic particles (Roche Applied Science) for 24 h at 4°C. The immune complexes or the particles were subjected to SDS-PAGE and Western blot analysis for NCL.
Migration Assays-Migration assays were performed as described previously (24) in 24-well microchemotaxis chambers (Corning, Inc., Lowell, MA) using uncoated polycarbonate membranes with 8-m pores. Serum-starved CHO cells were harvested and resuspended at a concentration of 10 5 cells/0.1 ml in serum-free medium containing 0.25% BSA. HUVEC, U87MG, M059K, and C6 cells grown in serum-containing medium were harvested and resuspended at a concentration of 5 ϫ 10 4 cells/0.1 ml (except from HUVEC that were resuspended at a concentration of 10 5 cells/0.1 ml) in serum-containing medium. In all cases, the bottom chamber was filled with 0.6 ml of the corresponding medium and the tested substances. The upper chamber was loaded with 0.1 ml of medium containing the cells and incubated for 4 h at 37°C. After completion of the incubation, the filters were fixed and stained with 0.33% toluidine blue solution. The cells that migrated through the filter were quantified by counting the entire area of each filter, using a grid and an Optech microscope at ϫ20 objective (Optech Microscope Services Ltd., Thame, UK).
PI3K p85 ELISA-The levels of total and phosphorylated PI3K p85 were quantified using Fast Activated Cell-based ELISA (Active Motif, Carlsbad, CA). Briefly, cells were cultured in 96-well plates 1 day prior to manipulation. Serumstarved CHO stably transfected cells were treated with 100 ng/ml of PTN for 10 min, fixed, and incubated with anti-phospho and anti-total p85 antibodies, according to the manufacturer's instructions.
Subcellular Fractionation-Subcellular fractions of cells comprising cytosolic, nuclear, and cell membrane extracts were prepared as previously described (18). Cell monolayers in 100-mm plates were washed extensively with PBS before being scraped and pelleted. Washed cells (30 ϫ 10 9 ) were then disrupted in a hypotonic solution (10 mM Hepes, pH 6.9, 10 mM KCl, 2 mM MgCl 2 , 0.1 mM PMSF) on ice. Nuclei were pelleted at 400 ϫ g for 5 min and washed twice in PBS before extraction in the lysis buffer (10 mM Tris-HCl, pH 7.6, 400 mM NaCl, 1 mM EDTA, 0.1 mM PMSF, and 1% Triton X-100). This suspension was centrifuged at 12,000 ϫ g for 10 min and the pellet was referred to as the nuclear fraction. The supernatant obtained after pelleting intact nuclei was further centrifuged at 14,000 ϫ g for 30 min and the supernatant corresponding to the cytoplasmic fraction was recovered, whereas the pellet was resuspended in lysis buffer containing 150 instead of 400 mM NaCl. This latter suspension was recentrifuged at 14,000 ϫ g for 30 min to separate the cytoskeletal (the pellet) and membrane (supernatant) fractions. Equivalent amounts of total protein from all fractions were immunoprecipitated for RPTP␤/ or ␣ v ␤ 3 and then analyzed by Western blot analysis for NCL.
Mass Spectrometry-Three mg of total protein from cell lysates were immunoprecipitated for ␣ v ␤ 3 and subjected to reduction by pre-treatment with dithiothreitol. After SDS-PAGE, proteins were visualized by silver staining. Relevant bands were excised and treated for in-gel digestion as described (31). Briefly, the silver was destained using Farmer's reagent and trypsin (porcine, modified, sequence grade, Promega, Madison, WI) was introduced to the dried gel pieces. After overnight tryptic digestion, the peptides were bound to a C18 ZipTip and after washing, they were eluted with acetonitrile containing the matrix (alfa-cyano 4-hydroxycinnamic acid) directly onto the target plate. The mass list was generated by MALDI-TOF mass spectrometry on an Ultraflex TOF/TOF from Bruker Daltonics. The search for identity was performed using the search engine ProFound by scanning the current version of the NCBI nr sequence database. The spectrum was internally calibrated using autolytic tryptic peptides, and the error was set at Ϯ0.02 Da. One missed cleavage was allowed, and methionine could be oxidized. The significance of the identity was judged from the search engines scoring system. The occurrence of the few missed cuts was either on a terminal basic residue or surrounded by acidic amino acid residues.
In Situ Proximity Ligation Assay-For detection of proteinprotein interactions, in situ proximity ligation assay (PLA) was performed. The components used (Olink Bioscience Uppsala, Sweden) were as follows: anti-mouse PLA plus probe, anti-rabbit PLA minus probe, and Detection Reagents Orange.
Statistical Analysis-All experiments were performed at least 3 independent times. Where applicable, the significance of variability between the results from each group and the corresponding control was determined by unpaired t test or analysis of variance. All results are expressed as mean Ϯ S.E. from at least 3 independent experiments.

RESULTS
␣ v ␤ 3 Expression and ␤ 3 Tyr 773 Phosphorylation Are Required for Cell Surface NCL Localization-To examine the cell surface localization of NCL in different types of cells that differentially express ␣ v ␤ 3 , fluorescence immunostaining of endothelial and glioma cells for NCL was performed. As shown in Fig. 1A, NCL was not detected in the extranuclear compartments of M059K and C6 cells that do not express ␣ ␤ 3 (24), whereas it was localized on the plasma membrane of HUVEC (also supplemental Fig. S1) and U87MG cells that express ␣ v ␤ 3 (24). To investigate FIGURE 1. Expression of ␣ v ␤ 3 correlates with cell surface NCL localization. A, representative immunofluorescence images of M059K, C6, HUVEC, and U87MG cells cultured in serum-containing medium and stained for NCL (green) and nucleus (blue). B, representative immunofluorescence images for NCL (green) and ␣ v ␤ 3 (red) of U87MG cells treated with a negative control siRNA or siRNAs for ␤ 3 or ␣ v ; mock-transfected M059K glioma cells, M059K cells transiently transfected to overexpress wild-type ␤ 3 or ␣ v ; mock-transfected CHO cells, CHO cells stably transfected to overexpress wild-type ␤ 3 or mutant ␤ 3 Y773F, ␤ 3 Y785F, or ␤ 3 Y773F/Y785F. C, cell lysates after treatment of U87MG cells with a negative control siRNA or siRNAs for ␤ 3 or ␣ v , or after overexpression of wild-type ␤ 3 or ␣ v in M059K cells, or different forms of ␤ 3 in CHO cells, were analyzed by Western blot for ␣ v , ␤ 3 , or actin (used as a loading control). D, for detection of cell surface NCL, intact cells were incubated with biotin, cell lysates were immunoprecipitated for magnetic streptavidin beads, and immunoprecipitates were analyzed by Western blot for NCL (membrane NCL). Whole cell protein extracts were analyzed by Western blot for NCL (total NCL) or actin, used as a loading control. siNeg, cells transfected with a negative control siRNA; si␤ 3 , cells transfected with siRNA for ␤ 3 ; si␣ v , cells transfected with siRNA for ␣ v ; vector, cells negative for ␤ 3 , transfected with the plasmid vector; wt␤ 3 , cells overexpressing wild-type ␤ 3 ; ␤ 3 Y773F/Y785F, cells overexpressing double mutant ␤ 3 Y773F/Y785F; ␤ 3 Y773F, cells overexpressing single mutant ␤ 3 Y773F; ␤ 3 Y785F, cells overexpressing single mutant ␤ 3 Y785F; wt␣ v , cells overexpressing wild-type ␣ v . Scale bars in all cases correspond to 10 m.
whether expression of ␣ v ␤ 3 is playing a role in cell surface localization of NCL, we down-regulated ␤ 3 expression in U87MG cells by using siRNA, or overexpressed ␤ 3 in M059K, as well as ␤ 3 in CHO cells that do not normally express it (Fig. 1, B and C), as previously described (24). Down-regulation of ␤ 3 but not ␣ v in U87MG cells resulted in decreased extranuclear NCL, as evidenced by immunofluorescence microscopy (Fig. 1B) and biotin labeling of cell membrane proteins (Fig. 1D). In the same line, overexpression of wild-type ␤ 3 in M059K and CHO cells resulted in extranuclear localization of NCL, as also evidenced by immunofluorescence microscopy (Fig. 1B) and biotin labeling of cell membrane proteins (Fig. 1D).
To determine whether ␤ 3 Tyr 773 or/and Tyr 785 is responsible for the cell surface localization of NCL, we used CHO cells overexpressing ␤ 3 Y773F, ␤ 3 Y785F, or ␤ 3 Y773F/Y785F instead of wild-type ␤ 3 and performed immunofluorescence assays. As shown in Fig. 1, B and D, NCL was localized in the extranuclear fraction of CHO cells overexpressing ␤ 3 Y785F, similarly to CHO cells overexpressing wild-type ␤ 3 , whereas in cells overexpressing ␤ 3 Y773F or ␤ 3 Y773F/Y785F it was restricted in the nucleus. In all cases, the total cellular amounts of NCL were not affected by the presence or absence of ␣ v ␤ 3 integrin (Fig. 1D).
PTN Induces Cell Surface NCL Localization through RPTP␤/ -mediated ␣ v ␤ 3 Activation-Because cell surface localization of NCL is induced by angiogenic growth factors, such as VEGF (8), we questioned whether PTN alters the intracellular trafficking of NCL. HUVEC displayed increased levels of cell surface NCL upon PTN stimulation, similarly to the known effect of serum (2) or VEGF ( Fig. 2A). This increase was not due to increased protein levels of NCL ( Fig. 2A) and was confirmed by immunofluorescence studies (Fig. 2B). Notably, PTN-induced cell surface NCL localization was not observed in CHO cells overexpressing ␤ 3 Y773F (Fig. 2C). Accordingly, PTN induced migration of CHO cells overexpressing wild-type ␤ 3 or ␤ 3 Y785F, but had no effect on CHO cells overexpressing ␤ 3 Y773F (Fig. 2D), suggesting that cell surface localization of NCL through ␤ 3 Tyr 773 phosphorylation relates to PTN-induced cell migration.
c-Src and PI3K That Lie Up-and Downstream of ␣ v ␤ 3 , Respectively, Are Required for PTN-induced Cell Surface NCL Localization-To further elucidate signaling molecules that affect PTN-induced cell surface NCL localization, we used inhibitors of several signaling molecules known to affect cell migration. As shown in Fig. 3A, inhibition of c-Src (also supplemental Fig. S2) and PI3K abolished PTN-induced cell surface localization of NCL, whereas inhibition of p38, JNK, and ERK 1 ⁄ 2 had no effect. PTN is known to activate c-Src, PI3K, and ERK 1 ⁄ 2

FIGURE 2. PTN induces cell surface NCL localization through RPTP␤/-mediated ␣ v ␤ 3 activation.
A, serum-starved HUVEC were stimulated with PTN (100 ng/ml) or VEGF (10 ng/ml) or FBS (15%) for 5 h at 37°C and then incubated with biotin, to label cell surface proteins, as described under "Experimental Procedures." In the upper panel, cell lysates were immunoprecipitated with magnetic streptavidin beads and immunoprecipitates were analyzed by Western blot for NCL. In the lower panel, total cell lysates were analyzed by Western blot for NCL. Actin was used in both cases as a loading control. B, representative immunofluorescence images stained for NCL (green) and nucleus (blue) in serum-starved HUVEC treated with PTN or VEGF for 5 h at 37°C. C, representative immunofluorescence images stained for NCL (green) and nucleus (blue) in serum-starved CHO cells expressing wild-type ␤ 3 (wt␤ 3   To investigate whether PI3K is up-or downstream of ␣ v ␤ 3, the effect of PI3K inhibition in PTN-induced ␤ 3 Tyr 773 phosphorylation was studied. The p38 inhibitor was used as a negative control and the c-Src inhibitor as a positive control, because c-Src is known to lay upstream of ␤ 3 Tyr 773 phosphorylation (24). As shown in Fig. 3B, PI3K inhibition did not affect PTN-induced ␤ 3 Tyr 773 phosphorylation, suggesting that PI3K lies downstream of ␣ v ␤ 3 . This was confirmed by ELISA for activated PI3K, using CHO cells overexpressing wild-type ␤ 3 , ␤ 3 Y773F, or ␤ 3 Y785F subunit. PTN significantly induced PI3K activation in CHO cells overexpressing wild-type ␤ 3 or ␤ 3 Y785F, whereas it had no effect in cells overexpressing ␤ 3 Y773F (Fig. 3C).
Cell Surface NCL Directly Interacts with ␣ v ␤ 3 -To study whether NCL directly associates with ␣ v ␤ 3 , immunoprecipitation assays were performed in HUVEC lysates. NCL was found to interact with ␣ v ␤ 3 by mass spectrometry (Table 1) and Western blot (Fig. 4A) analysis. In the former, among the proteins co-immunoprecipitated with ␣ v ␤ 3 , NCL (accession number AAA59954) was found to be a binding partner with a minimum sequence coverage of 19%. This relatively low sequence coverage can be explained by the particular structure of NCL, which is very rich in acidic (24.9%) and basic (16.3%) residues in highly basic regions, e.g. the C terminus (32), producing tryptic fragments that are too short for analysis by MALDI-TOF. Co-localization of ␣ v ␤ 3 with NCL on the cell surface was observed by immunofluorescence studies (Fig. 4B), whereas co-immunoprecipitation experiments using different cellular fractions (Fig.  4C) or cell lysates of biotinylated cells (Fig. 4D), as well as PLA assays (Fig. 4E), demonstrated direct formation of NCL-␣ v ␤ 3 complexes in cells expressing ␣ v ␤ 3 .
Interestingly, by using the same experimental approaches, NCL was also found to co-localize and co-immunoprecipitate with RPTP␤/, not only on the cell membrane, but also in the

TABLE 1 Identification of NCL (accession number AAA59954) by peptide mass fingerprint analysis (see text for details)
Monoisotopic uncharged masses are shown.  (Fig. 4, A-E). In M059K cells, NCL-RPTP␤/ complexes were limited and restricted in the cell nucleus (Fig.  4E).

Measured mass
Correlation between Cell Surface NCL Localization and ␣ v ␤ 3 Expression in Glioblastomas-To further confirm that ␣ v ␤ 3 expression is a prerequisite for cell surface NCL localization, we performed immunohistochemistry in tissue microarrays, containing samples from normal brains, grade 2 astrocytomas and grade 4 glioblastomas. As shown in Fig. 5, the localization pattern of both NCL and ␣ v ␤ 3 seems to correlate in all cases, although obvious diversity between samples can be observed. Intensity correlation analysis revealed increased co-localization in astrocytomas and the highest co-localization in glioblastomas compared with normal brains.

Expression of ␣ v ␤ 3 Determines the Biological Effect of Cell Surface NCL Targeting Agents on Endothelial and Cancer Cell
Migration-Based on the observation that NCL is preferentially localized on the surface of cells expressing ␣ v ␤ 3 , we wanted to evaluate the efficacy of anti-cell surface NCL strategies in relationship to ␣ v ␤ 3 expression. Anti-C23 monoclonal antibody raised against NCL, which has been shown to dramatically inhibit angiogenic functions of VEGF stimulated HUVEC (3), as well as HB-19 and Nucant 6L pseudopeptides, which are known to target and down-regulate only surface NCL (3, 10), were tested for their effect on migration of cells differentially expressing ␣ v ␤ 3 . As shown in Fig. 6A, the anti-C23 antibody caused a concentration-dependent inhibition of U87MG cell migration, which was maximal at 10 g/ml. Anti-C23 antibody significantly inhibited migration of all tested cells expressing FIGURE 4. Cell surface NCL directly interacts with ␣ v ␤ 3 and RPTP␤/. A, HUVEC were lysed with RIPA buffer and immunoprecipitated for NCL, RPTP␤/, ␣ v , or ␤ 3 integrin subunit. Immunoprecipitates (IP) were analyzed by Western blot (WB) for the presence of RPTP␤/, ␣ v , or ␤ 3 integrin subunit as shown. Immunoprecipitation for IgG was used as a negative control in every case. B, representative immunofluorescence images showing colocalization (yellow) of NCL (green) and RPTP␤/ (red) or ␣ v ␤ 3 (red) on the surface of HUVEC. Nuclei are shown blue. C, total proteins from nuclear, cytosolic, and membrane fractions of U87MG cells were immunoprecipitated for RPTP␤/ or ␣ v ␤ 3 and analyzed by Western blot for NCL. D, intact U87MG cells were incubated with biotin and cell lysates were immunoprecipitated for PTN, RPTP␤/, ␣ v , ␤ 3, NCL, magnetic streptavidin beads (s.b.) or IgG. Immunoprecipitates were analyzed by Western blot for NCL (left). The same blot was stripped and biotinylated proteins were detected by incubation with peroxidase-conjugated streptavidin (right). The arrow shows the band that corresponds to NCL. E, in situ PLA signals were detected as red dots, indicating the direct formation of NCL-RPTP␤/, NCL-␣ v , and NCL-␤ 3 complexes in different cell types. The interaction of PTN with its receptor RPTP␤/ was used as a positive control. Omission of the primary antibodies was used as negative control. Scale bars in all cases correspond to 10 m. activated ␣ v ␤ 3 , such as U87MG, HUVEC, CHO wt␤ 3 , and CHO ␤ 3 Y785F cells, whereas it did not affect or affected to a limited degree migration of M059K, C6, and CHO ␤ 3 Y773F cells (Fig.  6B). Similarly, HB-19 and Nucant 6L pseudopeptides significantly inhibited migration of U87MG cells and HUVEC, whereas they had a minor effect on the migration of M059K and C6 cells (Fig. 6, C and D).

DISCUSSION
Extranuclear distribution of NCL has been observed in both activated endothelial (6) and cancer (3)(4)(5) cells. However, not all tumor cells express cell surface NCL and the mechanisms involved are still unclear. Molecules known to affect tumor growth or/and angiogenesis that have been reported to increase cell surface NCL localization are serum (2), VEGF (8), laminin-1 (33), and low density lipoprotein receptor-related protein 1 (10). In the present study it is shown that PTN also increases cell surface NCL localization, and that expression of ␣ v ␤ 3 and ␤ 3 Tyr 773 phosphorylation are required for the redistribution of NCL from the nucleus to the cell membrane. VEGF has been shown to induce expression (34) and c-Src-mediated phosphorylation of ␤ 3 at Tyr 773 (35), which based on our data, provide an explanation for the VEGF-stimulated NCL cell surface localization. Laminin-1-induced increase of NCL cell surface localization relies on an integrin-dependent cascade (33), supporting our notion that integrins regulate NCL cell surface localization. In the same line are also data showing that many of the molecules that have been described to bind and act through cell surface NCL are also described to act through integrins, and especially ␣ v ␤ 3 . This is true for VEGF (34,35), laminins (15,36,37), hepatocyte growth factor (13,38), endostatin (14,39), and PTN (18,24).
Interestingly, ␤ 3 but not ␣ v , appears to be the subunit responsible for NCL cell surface localization. Cell surface NCL is only observed in ␤ 3 -overexpressing CHO and M059K cells, but not ␣ v overexpressing cells. In the same line, down-regulation of ␤ 3 abolishes cell surface localization of NCL in U87MG cells. However, down-regulation of ␣ v does not cause such an effect. The latter could be attributed to formation of a functional complex of ␤ 3 with its other partner subunit ␣ IIb , which can be ectopically expressed in several tumor cell lines (40).
Cell surface NCL localization depends on the Tyr 773 phosphorylation of ␤ 3 integrin subunit. PTN induces ␤ 3 Tyr 773 phosphorylation through RPTP␤/ and c-Src activation (24). Down-regulation of RPTP␤/ or c-Src inhibition completely abolishes PTN-induced ␤ 3 Tyr 773 phosphorylation (24) and cell surface localization of NCL (present study), suggesting that both RPTP␤/ and c-Src lie upstream of ␣ ␤ 3 . c-Src is also known to lie upstream of and be responsible for VEGF-induced ␣ v ␤ 3 phosphorylation (23), although possible involvement of RPTP␤/ is currently being investigated. Downstream of ␣ v ␤ 3 phosphorylation, NCL cell surface localization is mediated by PI3K (Fig. 7). The latter has important roles in cell migration (41) and the remodeling of actin filaments induced by growth factors (42) and integrins (43), and is involved in PTN-induced endothelial cell migration (26). Both the regulation of nucleocytoplasmic shuttling of NCL by phosphorylation (44) and the reported physical interaction between NCL and PI3K (16,45), indicates a possible direct regulation of NCL by PI3K upon ␤ 3 Tyr 773 phosphorylation. Alternatively, PI3K may drive NCL membrane shuttling through regulation of other targets that interact with NCL. For example, it has been shown that proteins containing pleckstrin homology domains are recruited in the membrane following PI3K activation in response to different growth factors (46). Although NCL does not possess related sequences, it has been proposed as a potential ligand for such proteins through its acidic motifs (47).
It is well known that ERK 1 ⁄ 2 is activated by ␣ v ␤ 3 and participates in integrin-mediated cell migration (48). The present study shows that ERK 1 ⁄ 2 is not involved in PTN-induced cell surface NCL localization, although it is required for PTN-induced endothelial cell migration (26). One possible explanation is that ERK 1 ⁄ 2 may lay downstream of ␣ v ␤ 3 -mediated cell surface NCL localization. Although few data exist on the participation of cell surface NCL to activation of cell signaling pathway(s), it has been shown that cell surface NCL enhanced signaling through the mitogen-activated protein kinase (MAPK) pathway (49), in agreement with our data.
Interestingly, cell surface NCL directly interacts with ␣ v ␤ 3 and RPTP␤/ on the cell surface. This is in line with the emerg-ing notion that cell surface NCL may participate in functional complexes with other receptors, such as urokinase plasminogen activator receptor (50) and ErbB receptors (4). Interaction of NCL with ␣ v ␤ 3 has been previously observed in MDA-MB-231 and HeLa cancer cells by bioconjugated semiconductor quantum dots; however, it was then discussed as a methodology FIGURE 6. Targeting of cell surface NCL is effective only in cells expressing ␣ v ␤ 3 . A, U87MG cells were treated with various concentrations of the monoclonal anti-C23 antibody against NCL and with a high concentration (20 g/ml) of human IgG. Different cell types that expressed or not ␣ v ␤ 3 were treated with anti-C23 antibody (10 g/ml) (B), HB-19 pseudopeptide (10 g/ml) (C), and Nucant 6L pseudopeptide (10 g/ml) (D). Cell migration assays were performed in FBS-containing medium using microchemotaxis chambers as described under "Experimental Procedures." Data are the mean Ϯ S.E. percentage change in number of migrating cells versus the corresponding untreated cells (default ϭ 100). Asterisks in A denote statistical significant differences from untreated cells. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. Binding of exogenous PTN to its receptor RPTP␤/ on the surface of endothelial or cancer cells, leads to c-Src activation, Tyr 773 phosphorylation of ␤ 3 integrin subunit, and PI3K phosphorylation. Activated PI3K induces, either directly or indirectly, the translocation of NCL from the nucleus to the cell membrane. Cell surface NCL interacts with both RPTP␤/ and ␣ v ␤ 3 and as previously shown (18), is required for PTN-induced cell migration.
artifact (51). In the present study, by using several different techniques, this interaction has been clearly demonstrated. One possibility for the functional significance of these interactions is activation of cell signaling pathways. Besides activation of MAPKs (49), plasma membrane NCL has been shown to increase cytoplasmic Ca 2ϩ levels (52), or activate PI3K (53). On the other hand, accumulating evidence suggests that signaling events related to cell migration require ligand/receptor endocytosis (54). VEGF and VEGF receptor 2 internalization are required for endothelial recovery in a wound healing assay (55), whereas low platelet-derived growth factor (PDGF) concentrations induce cell migration by promoting clathrin-mediated endocytosis of PDGF receptor (56). Moreover, functions of integrins and their signaling leading to cell migration are strictly controlled by their endocytic trafficking (57). The well demonstrated role of NCL in the internalization of specific ligands and transport to the cell nucleus (6,14,17,18,58) prompt the speculation that NCL mediates the translocation of PTN or/and RPTP␤/ inside the cell, thus triggering or amplifying signaling events. This hypothesis is strengthened by the fact that both PTN (18) and RPTP␤/ (present study) co-localize with NCL in the cell nucleus, whereas down-regulation of NCL expression caused significant reduction in nuclear localization of PTN (18) and RPTP␤/ (supplemental Fig. S4). The functional significance of this internalization is currently under investigation.
The relevance of ␣ v ␤ 3 -regulated cell surface NCL localization to human cancer was addressed by analyzing brain cancer tissue microarrays. Increased ␣ v ␤ 3 expression was observed in grade 4 glioblastomas compared with grade 2 astrocytomas and normal brain, in line with previous data showing that ␣ v ␤ 3 expression was significantly increased in high-grade glioblastomas compared with low-grade gliomas (59). Extranuclear NCL expression was also increased in grade 4 glioblastomas, and in all cases correlated with ␣ v ␤ 3 , further strengthening our notion that ␣ v ␤ 3 is playing a role for NCL cell surface targeting.
It was recently shown that the phosphorylation of NCL by protein kinase Cand casein kinase 2 mediates the surface translocation of NCL via regulating its interaction with heat sock cognate 70 (hsc70) (60). Although NCL phosphorylation (1,53,61) and glycosylation (52) have been mentioned to play a role in NCL cell surface localization, the mechanisms remain mostly unexplored. In the present study we did not look for NCL phosphorylation; however, we showed that ␤ 3 Tyr 773 phosphorylation is required for NCL cell surface localization. ␣ v ␤ 3 integrin is found to form a complex with hsc70 in lipid raft microdomains of the cytoplasmatic membrane of susceptible intestinal epithelial cells, and both act as receptors for rotavirus entry (62). Therefore, it is tempting to speculate that ␣ v ␤ 3 and hsc70 may cooperate for cell surface NCL localization and in that case direct or indirect NCL phosphorylation by protein kinase C-and casein kinase 2 may have a role.
Based on the observation that NCL is preferentially localized on the surface of cells that express ␣ v ␤ 3 , we evaluated the efficacy of an anti-NCL antibody, as well as HB-19 and Nucant 6L pseudopeptides on cell migration, in relationship to ␣ v ␤ 3 expression, in the presence of serum that is known to increase levels of cell surface NCL (Ref. 2 and Fig. 2A). Several cell lines were tested and significant reduction of cell migration was observed only in cells that express ␣ v ␤ 3 . This is in line with data showing that a pseudopeptide that is known to bind the C-terminal RGG domain of the cell surface-expressed NCL and block its function, similarly to HB-19 and Nucant 6L pseudopeptides, completely abolished PTN-induced HUVEC migration (18) and data showing that the surface NCL antagonists, HB-19 and related NUCANT pseudopeptides, exert different inhibitory effects in different tumor cell types. Although the reason for the latter observation has not been studied, this has been attributed to different levels of surface NCL or/and different NCL partners in tumor cells (10). Our data clearly show that levels of cell surface NCL are different among different types of cells and correlate with ␣ v ␤ 3 expression, suggesting that ␣ v ␤ 3 could be a biomarker for prediction of the biological outcome of cell surface NCL antagonists.
In summary, the current study unravels at least one of the pathways that participate in cell surface localization of NCL in endothelial and glioma cells, and provides evidence that ␣ v ␤ 3 expression could be a useful biomarker for targeting cell surface NCL as a therapeutic strategy in treatment of gliomas or/and other types of angiogenesis-dependent cancers.