β1-Integrin-mediated Glioma Cell Adhesion and Free Radical-induced Apoptosis Are Regulated by Binding to a C-terminal Domain of PG-M/Versican*

Integrins are cell-surface glycoproteins that mediate cell activities, including tissue morphogenesis, development, immune response, and cancer, through interaction with extracellular proteins. Here we report a novel means by which integrin signaling and functions are regulated. In pull-down assays and immunoprecipitation, β1-integrin bound to the C-terminal domain of PG-M/versican, an extracellular chondroitin sulfate proteoglycan. This was confirmed by cell-surface binding assays. Binding was calcium- and manganese-dependent. Upon native gel electrophoresis, β1-integrin comigrated with the C-terminal domain of PG-M/versican. The interaction of β1-integrin with the C-terminal domain of PG-M/versican activated focal adhesion kinase, enhanced integrin expression, and promoted cell adhesion. As a result, cells expressing the C-terminal domain of PG-M/versican were resistant to free radical-induced apoptosis. As the PG-M/versican peptide used in this study does not contain the RGD consensus-binding motif for integrins, the mechanism of the observed binding represents an entirely new function.

Immunoprecipitations-The cell lysate was prepared in lysis buffer for protein interaction (20 mM Tris-Cl (pH 7.4), 150 mM NaCl, 1 mM CaCl 2 , 2 mM MnCl 2 , 2 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 2 mM EDTA, 5 g/ml BSA, and 1.5% Triton X-100), followed by centrifugation to recover proteins in the supernatant. The protein sample was incubated overnight at 4°C with protein G that had been saturated with monoclonal antibody 4B6 (36), followed by blocking with 5% BSA. After washing, the beads were boiled in 1ϫ protein loading dye for 5 min, analyzed by Western blotting, and probed with anti-␤ 1integrin antibody JB1a. In addition, the cell lysate was incubated with antibody JB1a-saturated protein G under the same conditions as described above to co-immunoprecipitate G3⌬EGF. Bound products were analyzed by Western blotting and probed with antibody 4B6. Cells were subjected to native gel electrophoresis, harvested, homogenized, and then centrifuged. Cell membrane fractions were resuspended in 1% Triton X-100 and subjected to electrophoresis on 3-20% gradient gels. The separated proteins were analyzed by Western blotting and probed with antibody JB1a or 4B6.
Western Blotting-Protein samples were subjected to SDS-PAGE on separating gel containing 10 -12% acrylamide. The buffer system was 1ϫ Tris/glycine buffer (Amresco) containing 1% SDS. Separated proteins were transblotted onto a nitrocellulose membrane in 1ϫ Tris/ glycine buffer containing 20% methanol at 60 V for 2 h in a cold room. The membrane was blocked in TBST (10 mM Tris-Cl (pH 8.0), 150 mM NaCl, and 0.05% Tween 20) containing 10% nonfat dry milk powder (TBSTM) for 30 min at room temperature to saturate nonspecific binding sites on the membrane. The membrane was then incubated overnight at 4°C with monoclonal antibody 4B6 (which recognizes an epitope at the leading peptide of the link protein, which was engineered in all recombinant constructs) or with the anti-His tag monoclonal antibody prepared in TBSTM. The membranes were washed with TBST (3 ϫ 30-min washes) and then incubated for 2 h with horseradish peroxidase-conjugated goat anti-mouse IgG antibodies (1:50,000 dilution) in TBSTM. After washing as described above, the bound antibodies were visualized by chemiluminescence (ECL kit).
Protein Purification-G3 recombinant proteins containing an N-terminal His tag were purified on Ni-NTA affinity columns (catalog no. 30230, QIAGEN Inc., Chatsworth, CA) as previously described (31). Briefly, the G3 domain was amplified by PCR using two primers, 5Ј-aaaggatccggacaggatccatgcaaa and 5Ј-aaagcatgcgcgccttgagtcctgccacgt. The product was subcloned into the bacterial expression vector pQE30 (catalog no. 32149, QIAGEN Inc.), and the resulting construct contained an N-terminal MRGS His tag. The G3 construct was expressed in Escherichia coli strain M15 using pQE30. Protein expressed in bacteria was prepared, and peptides were purified on an Ni 2ϩ -NTA affinity column under denaturing conditions according to the manufacturer's instructions. The protein was dialyzed and renatured in a buffer containing 10 mM HEPES, 10 mM MgCl 2 , 50-150 mM NaCl (serial increase NaCl), 0.1 mM EDTA, 2-10% glycerol (serial increase), and 0.2-1 mM dithiothreitol (serial increase).
Cell-surface Binding Assays-G3⌬EGF-and vector-transfected cells harvested by trypsinization were washed three times with TBS (20 mM Tris-HCl and 150 mM NaCl (pH 7.4)) containing 2 mM CaCl 2 and probed with antibody 4B6 for 1 h on ice in the presence of 2 mM CaCl 2 , followed by staining with FITC-conjugated secondary antibody for 1 h on ice in the dark. Cells were washed after each treatment. Binding of G3⌬EGF to the cell surface was analyzed by flow cytometry. Vector-transfected U87 cells were also incubated with the culture medium from G3⌬EGFor vector-transfected cells at 37°C for 1 h or with purified His 6 -G3 (expressed by bacteria) at room temperature for 1 h to assess binding of G3⌬EGF or His 6 -G3 products to cells. CaCl 2 and MnCl 2 were included in the system at a concentration of 2 mM each. Bacterially expressed His 6 -G3 was purified on Ni-NTA affinity columns and dialyzed overnight against TBS containing CaCl 2 at room temperature. Equal numbers of cells were incubated with 0.1% BSA in Hanks' balanced salt solution as a negative control. The cells were washed three times with cold TBS containing CaCl 2 and MnCl 2 (2 mM each), followed by antibody 4B6 staining as described above.
Activation of FAK-Tissue culture plates were precoated with the culture medium from vector-or G3⌬EGF-transfected cells or with purified His 6 -G3. Plates were then seeded with cells transfected with G3⌬EGF or the vector control and incubated at 37°C for 1 h. The cell lysate was prepared in lysis buffer for phosphorylation assay (50 mM HEPES (pH 7.5), 0.5% Triton X-100, 100 mM NaF, 10 mM sodium phosphate, 4 mM EDTA, 2 mM sodium vanadate, 2 mM sodium molybdate, 2 g/ml aprotinin, 2 g/ml leupeptin, and 2 g/ml phenylmethyl-sulfonyl fluoride), followed by incubation with protein G that had been presaturated overnight with anti-FAK monoclonal antibody (Transduction Laboratories) at 4°C. After washing, the beads were boiled in 1ϫ protein loading dye for 5 min, followed by Western blot analysis and probing with anti-FAK or anti-phosphotyrosine (clone PY-20, Transduction Laboratories) antibody to detect FAK phosphorylation.
Adhesion Assay-Vector-transfected cells (2 ϫ 10 5 cells) were seeded on six-well plates precoated with the culture medium from vector-or G3⌬EGF-transfected cells and incubated on the plates at 37°C for 1-1.5 h as described in the figure legends. Unattached cells were removed by washing, and the remaining cells were photographed. Ni-NTA-purified His 6 -G3 was also used to coat six-well plates. The coated wells were washed with DMEM, followed by incubation with vectortransfected cells (2 ϫ 10 5 cells) in DMEM at 37°C for 7 h. The cultures were photographed, and attached cells were counted.
To test whether cell adhesion is proportional to a decreasing concentration of G3⌬EGF product, the culture media collected from vectorand G3⌬EGF-transfected U87 cells were diluted with phosphate-buffered saline to 100, 50, 25, 12.5, and 6.25% and used to coat six-well tissue culture plates at 4°C overnight. U87 cells (2 ϫ 10 5 cells/well) were seeded on the wells in DMEM containing 1% FBS and incubated at 37°C for 1 h. The medium was removed, and the cells were washed with phosphate-buffered saline. The cells remaining on the plate were photographed and counted.
The effect of the products of mini-versican and G3⌬EGF on the adhesion of different cells was also examined. The culture media from vector-, mini-versican-, and G3⌬EGF-transfected U87 cells were used to coat six-well tissue culture dishes at 4°C for 4 h. Different types of cells (U87 cells, NIH3T3 fibroblasts, Jurkat cells, and breast cancer cell line MDA-MB-231; 2 ϫ 10 5 cells/well) were inoculated on the wells in DMEM containing 1% FBS and incubated at 37°C for 1 h. The culture media were removed, and the cells were washed three times with phosphate-buffered saline. The cells remaining on plates were counted.
Blocking Assay-Vector-transfected cells (1.5 ϫ 10 4 cells) were incubated in 100 l of DMEM in the presence or absence of antibody JB1a at a concentration of 30 g/ml at room temperature for 30 min. The cells were then seeded on 96-well culture plates precoated with purified His 6 -G3 and maintained at 37°C for 7 h. Unattached cells were removed by washing. The remaining cells were photographed, and cell attachment was determined by cell counting.
Competition Assay-Nylon membranes were incubated with the U87 cell lysate (membrane preparation) or heat-denatured bovine serum at room temperature for 1 h, followed by blocking with heat-denatured bovine serum at room temperature for 1 h. After washing, the nylon membranes were incubated at room temperature with the culture medium from G3⌬EGF-transfected U87 cells for 1 h to absorb G3⌬EGF. The treated medium, untreated G3⌬EGF-containing medium, or medium containing BSA was incubated with vector-transfected U87 cells for attachment assays as described above.
U87 cells (5 ϫ 10 4 cells) were also incubated with the culture medium containing purified His 6 -G3 or eluate from the vector control in the presence of 2 mM each CaCl 2 and MnCl 2 at room temperature for 30 min. The cells were then inoculated on 12-well tissue culture plates precoated with purified His 6 -G3 and maintained at 37°C for 7 h. Unattached cells were removed, and attached cells were photographed and counted.
Apoptosis Assay-U87 cells (2 ϫ 10 5 cells/well) in monolayer culture on six-well plates or in suspension culture were treated with H 2 O 2 (at 1.5 mM for 6 h or at 0.1 mM for 48 h, respectively). Trypan blue staining was used to determine viability. Cells were also subjected to annexin V-FITC staining (15 min in the dark) using an annexin V apoptosis detection kit (Santa Cruz Biotechnology, Inc.) to monitor apoptosis as described (37,38). Fig. 1A) containing an engineered His 6 tag was stably expressed in human astrocytoma cell line U87. The cell lysate was prepared under native conditions and purified with Ni-NTA resin. The purified product was analyzed by Western blotting and probed with antibody JB1a. The untreated lysate was used as a positive control to detect integrin, and the cell lysate from vector-transfected cells that had been subjected to the same purification procedure was used as a negative control. The results indicated that ␤ 1 -integrin was pulled down by the His-tagged mini-versican (Fig. 1B). Expres-sion of the mini-versican was confirmed by Western blotting (Fig. 1C). Our previous studies suggested that U87 cells interact with a versican mutant construct containing only the CRD and CBP motifs (G3⌬EGF) expressed by COS-7 cells and with the versican G3 domain expressed by bacteria (His 6 -G3) (33,34). To examine the molecular sites of interaction at the subdomain level, the cell lysate prepared from G3⌬EGF-transfected cells was subjected to the pull-down assay using Ni-NTA resin as described above. Probing with antibody JB1a indicated that ␤ 1 -integrin coprecipitated with G3⌬EGF ( Fig. 1, B and C). His 6 -G3 expressed in bacteria was purified and incubated with intact U87 cells for 2 h, followed by pull-down assay. The results indicated that His 6 -G3 bound to ␤ 1 -integrin (Fig. 1D). To confirm this binding, co-immunoprecipitation was performed on the lysate from G3⌬EGF-transfected cells using monoclonal antibodies BJ1a and 4B6 (which recognizes an epitope at the leading peptide of G3⌬EGF). Consistent with the above results, ␤ 1 -integrin was co-immunoprecipitated with G3⌬EGF ( Fig. 1E), and G3⌬EGF was co-immunoprecipitated with antibody BJ1a (Fig. 1F). As a final experiment, the lysate prepared from G3⌬EGF-or vector-transfected cells was subjected to native gel electrophoresis and Western blotting and then probed with antibody JB1a or 4B6. Both antibodies detected bands with similar patterns, migrating at sizes ranging from 250 to 500 kDa (Fig. 1G). This indicated that the majority of the cell-associated G3⌬EGF interacts with ␤ 1 -integrin under native conditions and suggested that versican is an important partner in regulating the integrin signaling pathway, although G3⌬EGF does not contain the RGD consensus-binding sequence for integrins (14).

␤ 1 -Integrin Binds to the C-terminal Domain of PG-M/Versican-A mini-versican (
To examine binding of G3⌬EGF to the cell surface, G3⌬EGFand vector-transfected cells were analyzed by flow cytometry. Labeling with antibody 4B6 demonstrated that cells transfected with G3⌬EGF expressed and interacted with this protein strongly (77% binding) ( Fig. 2A). When untransfected U87 cells were incubated with the culture medium from G3⌬EGFor vector-transfected COS-7 cells, large amounts of bound G3⌬EGF were detected (51% binding) (Fig. 2B). Incubation with His 6 -G3 purified from a bacterial expression system produced similar results (Fig. 2C). In the presence of either calcium (Fig. 2D) or manganese (Fig. 2E), binding of His 6 -G3 to the cell surface was concentration-dependent. However, manganese had the greater effect on binding, and 5 mM manganese had the same effect on the binding activity as a combination of calcium and manganese (2 mM each) (data not shown). The binding activity disappeared when EDTA was added (Fig. 2F), confirming the requirement for calcium or manganese. Binding of G3⌬EGF to the cell surface was also confirmed by confocal microscopy. Labeling with antibody 4B6 showed that G3⌬EGF bound to the surface of G3⌬EGF-transfected U87 cells (Fig.  2G), but no staining was detected in vector-transfected cells (Fig. 2H).
The G3 Domain of PG-M/Versican Activates FAK and Enhances Cell Adhesion-To investigate the biological effect of versican-␤ 1 -integrin interaction, we examined its effect on FAK activity, the immediate intracellular step in the integrin-medi-  G3⌬EGF. B, the cell lysate from mini-versican-or G3⌬EGF-transfected U87 cells was purified on Ni 2ϩ -NTA affinity columns. The products were analyzed by Western blotting and probed with antibody JB1a. control, the cell lysate without Ni 2ϩ -NTA purification. C, expression of mini-versican and G3⌬EGF was confirmed by Western blotting and probing with antibody 4B6. D, purified His 6 -G3 (or eluate from the vector control) was incubated with U87 cells, followed by pull-down assay as described under "Experimental Procedures" and analysis by Western blotting and probing with antibody JB1a. E, in the co-immunoprecipitation (co-IP) assay, protein G beads complexed with antibody 4B6 were incubated with the lysate from G3⌬EGF-transfected cells. The precipitated proteins were analyzed by Western blotting and probed with antibody JB1a. control, the cell lysate without antibody 4B6 precipitation. F, protein G beads complexed with anti-␤ 1 -integrin antibody JB1a and anti-␣ 5 integrin antibody (BD PharMingen) or protein G beads alone were precipitated with the lysate from G3⌬EGF-transfected cells, followed by Western blot analysis and probing with antibody 4B6. control, the cell lysate without treatment. G, the cell lysate prepared from G3⌬EGF-transfected cells was subjected to native gel electrophoresis, followed by Western blotting and probing with antibody 4B6 or JB1a. Antibodies detected bands with similar patterns. ated pathway (1-7). The cell lysate prepared from G3⌬EGF-or vector-transfected U87 cells was immunoprecipitated with anti-FAK antibody, followed by Western blot assay and probing with anti-phosphotyrosine antibody. G3⌬EGF-transfected cells had a higher level of phosphotyrosine activity compared with vector-transfected cells, although probing with anti-FAK antibody showed that the levels of FAK protein expression were similar (Fig. 3). Treatment of U87 cells with purified His 6 -G3 also resulted in enhanced phosphotyrosine activity of FAK compared with BSA treatment.
It is known that integrin-mediated cell adhesion results in activation of FAK activity (6,7). To investigate whether the observed G3⌬EGF-induced FAK activation is accompanied by changes in cell adhesion, we inoculated G3⌬EGF-and vectortransfected cells on tissue culture plates for attachment assays. As predicted, G3⌬EGF expression induced cell attachment (Fig. 4A) compared with the vector control (Fig. 4B). On bacte-rial Petri dishes, G3⌬EGF-transfected cells were able to attach and grow as monolayer cultures (Fig. 4C), whereas vectortransfected cells aggregated and remained in suspension (Fig.  4D). These results suggested that G3⌬EGF expression had enhanced cell adhesion. To confirm this effect, plates were coated with the medium containing G3⌬EGF or purified His 6 -G3 and inoculated with U87 cells. G3⌬EGF in the culture medium (Fig. 4E) or purified His 6 -G3 (Fig. 4F) enhanced cell attachment compared with plates coated with the culture medium from vector-transfected cells (Fig. 4G) or BSA alone (Fig. 4H).
We then performed independent assays similar to those described in the legend to Fig. 4 to quantify the cell attachment effected by G3⌬EGF. We demonstrated that G3⌬EGF-expressing cells had much greater activity to attach to tissue culture plates (Fig. 5A) and Petri dishes (Fig. 5B). U87 cells had greater activity to attach to tissue culture plates coated with G3⌬EGF (Fig. 5C) and purified His 6 -G3 (Fig. 5D) products.

FIG. 2.
Interactions of G3⌬EGF with the cell surface. Flow cytometry analysis indicated that the transfected cells expressed high levels of G3⌬EGF (red; 77% staining) compared with untransfected control cells (black; no staining) and vector-transfected cells (green; probed with secondary antibody alone) (A). The culture medium from G3⌬EGF-transfected COS-7 cells also interacted with the cell surface as indicated by the antibody 4B6 labeling (red; 51%) (B). The G3 domain expressed by bacteria (His 6 -G3) also interacted with the U87 cell surface (red; 33% staining; green; background staining) (C). This binding was calcium-dependent (D) and manganese-dependent (E), and the interaction was abolished when EDTA (5 mM) was added (F). G3⌬EGFtransfected (G) and vector-transfected (H) U87 cells were immunostained with antibody 4B6 and examined with a confocal microscope. G3⌬EGF labeled the cell surface. FIG. 3. G3⌬EGF induces FAK activity. The cell lysate prepared from U87 cells incubated with the culture medium from G3⌬EGF-or vector-transfected U87 cells was incubated with anti-FAK antibodysaturated protein G beads. The precipitates were analyzed by Western blotting and probed with PY-20, an anti-phosphotyrosine antibody. G3⌬EGF transfection enhanced FAK phosphorylation. The cell lysate prepared from U87 cells treated with or without His 6 -G3 was also subjected to anti-FAK immunoprecipitation (IP). Treatment with His 6 -G3 enhanced FAK phosphorylation, but FAK expression remained similar in treated and untreated cells. Dilution of the G3⌬EGF product reduced its ability to enhance cell attachment (Fig. 5E).
To test the function of mini-versican and G3⌬EGF in the attachment of different cell types, we compared their effects on the attachment of U87 cells, NIH3T3 fibroblasts, the MDA-MB-231 breast cancer cell line, and the Jurkat tumor cell line. The products of mini-versican and G3⌬EGF enhanced the attachment of all types of cells (Fig. 6A). Both mini-versican and G3⌬EGF had the highest activity to enhance the attachment of U87 cells, but the lowest activity to enhance the attachment of Jurkat cells. The enhancement by these products of the attachment of NIH3T3 fibroblasts is shown in Fig. 6 (B-D).
In blocking assays, U87 cells pretreated with or without anti-integrin antibody were seeded on plates precoated with purified His 6 -G3. Antibody treatment reduced cell attachment (Fig. 7A). In competition assays, U87 cells were also incubated with the G3⌬EGF-containing medium or the G3⌬EGF-containing medium pre-absorbed by integrin. Integrin treatment par-tially abolished the enhanced adhesion effects of the G3⌬EGFcontaining medium (Fig. 7B). Purified His 6 -G3 or eluate from the vector control was incubated with U87 cells before they were seeded on His 6 -G3-coated plates (24), and the incubation of free His 6 -G3 inhibited cell attachment (Fig. 7C). U87 cells were also incubated with the G3⌬EGF-containing medium in the presence (Fig. 7D) or absence (Fig. 7E) of free His 6 -G3, followed by immunostaining with antibody 4B6. Treatment with free His 6 -G3 reduced G3⌬EGF binding to the cells. Taken together, our results provide strong evidence that G3⌬EGF or purified His 6 -G3 plays an important role in enhanced cell adhesion through binding to integrin and activation of FAK.
The G3 Domain of PG-M/Versican Enhances ␤ 1 -Integrin Expression and Protects Cells from Free Radical-induced Apoptosis-As integrin expression is always associated with cell adhesion, we examined whether G3⌬EGF-induced cell adhesion affects integrin expression. G3⌬EGF-and vector-transfected U87 cells were seeded on tissue culture plates for 1, 2, 3, and 24 h. Analysis of integrin expression by Western blotting indicated that G3⌬EGF-transfected cells expressed higher levels of integrin during cellular attachment compared with vector-transfected cells (Fig. 8). However, no difference could be detected after 24 h of cell inoculation, when both types of cells were well spread on the plates.
As integrin-mediated cell adhesion is essential for cell survival (1-7), we investigated whether G3⌬EGF-induced cell adhesion and integrin expression can promote cell viability. We used the free radical H 2 O 2 to induce cell death and examined the effects of G3⌬EGF. H 2 O 2 was added to monolayer and suspension cultures of G3⌬EGF-and vector-transfected cells maintained in DMEM containing 10% FBS. H 2 O 2 -induced apoptosis was reduced in G3⌬EGF-transfected cells as measured by cell survival (Fig. 9, A and B) and by labeling with annexin FIG. 5. G3⌬EGF-enhanced cell adhesion is concentration-dependent. In cell attachment assays, G3⌬EGF-and vector-transfected cells were cultured in DMEM containing 1% FBS on tissue culture plates for 1 h. Adherent cells were counted, and cell adhesion was determined (A). G3⌬EGF-and vector-transfected cells were also cultured on bacterial Petri dishes for 4 days to test cell adhesion (B). Tissue culture plates coated with the G3⌬EGF-containing medium enhanced cell attachment (1-h incubation) compared with plates coated with the medium from vector-transfected cells (C). Tissue culture plates coated with purified His 6 -G3 enhanced cell attachment (1-h incubation) compared with plates coated with BSA alone (D). When the medium containing G3⌬EGF was diluted, cell adhesion decreased (E). V (Fig. 9, C and D). We then analyzed the effect of G3⌬EGF on integrin expression. H 2 O 2 was added to G3⌬EGF-and vectortransfected cells before (Fig. 9E) and after (Fig. 9F) cell spread-ing, and the cell lysate was analyzed by Western blotting. Integrin expression decreased in vector-transfected cells, but not in G3⌬EGF-transfected cells. Coating the plates with purified His 6 -G3 reduced H 2 O 2 -induced apoptosis (Fig. 9G). DISCUSSION Our study has shown for the first time that the aggregating chondroitin sulfate proteoglycan PG-M/versican binds to ␤ 1integrin. Initially, we observed that the mini-versican product, the G3 fragment, and the G3 fragment lacking the two EGFlike motifs (G3⌬EGF) all bound to the U87 cell surface. Further investigation involving pull-down and co-immunoprecipitation assays indicated that ␤ 1 -integrin was involved in binding. When His 6 -tagged versions of mini-versican and G3⌬EGF (both expressed in COS-7 cells) and the G3 domain (expressed in E. coli) were bound to Ni-NTA resin, we were able to copurify ␤ 1 -integrin. Binding was further confirmed by co-immunoprecipitation. Antibody against an epitope located at the leading peptide of G3⌬EGF was able to coprecipitate G3⌬EGF and ␤ 1 -integrin. Similarly, antibody against ␤ 1 -integrin was able to coprecipitate the G3⌬EGF product.
Integrins have been well known to interact with ECM molecules such as fibronectin, collagen, and vitronectin and with soluble ligands such as fibrinogen (39). Integrins also interact with other cell-surface receptors such as ICAM (intercellular adhesion molecule) and tetraspan (40,41). The chondroitin sulfate proteoglycan PG-M/versican has been shown to interact with a number of ECM and cell-surface glycoproteins, including hyaluronan, fibronectin, tenascin, fibulin-1, fibulin-2, CD44, and L-selectin (20, 24 -28). However, the binding activity reported here represents the first observation of interaction between PG-M/versican and integrin. Furthermore, a well known consensus-binding sequence for integrins is the RGD peptide (8,9). The PG-M/versican G3 domain does not contain an RGD peptide; thus, the binding observed in this study may represent a novel mechanism of integrin binding.
The diverse binding activities of PG-M/versican are achieved via its multiple heterogeneous domains: hyaluronan binds to the N-terminal fragment of PG-M/versican; the CD44, L-selectin, and P-selectin interactions involve the glycosaminoglycan chains of PG-M/versican; and tenascin and fibulin-1 bind to the CRD motif of PG-M/versican. The minimal sequence present in all products studied here is G3⌬EGF, and our aim was to dissect the motif in G3⌬EGF (e.g. CRD or CBP) that interacts with ␤ 1 -integrin. However, we did not obtain any binding activity with the CRD or CBP construct. Only in the presence of both CRD and CBP (the G3⌬EGF construct) was the interaction with ␤ 1 -integrin observed. It seems that the binding requires the cooperation of CRD and CBP to generate the optimal conformation for binding to ␤ 1 -integrin.
To test whether G3⌬EGF can directly interact with ␤ 1 -integrin, we used yeast two-hybrid assays. We generated constructs containing different fragments of ␤ 1 -integrin for the assays, but no positive result was obtained. It could be that the binding requires a certain conformational structure and/or glycosylation modification of ␤ 1 -integrin that can be obtained only in mammalian cells. The binding might also require the presence of the partner of ␤ 1 -integrin, the ␣ chain. We have tried (␣ 2 and ␣ 5 ) to identify the ␣ chain involved without success. From the binding results, we could not exclude the possibility that the interaction of PG-M/versican with integrin requires other components in forming complexes for their binding. Nevertheless, our competition results that anti-␤ 1 -integrin antibody blocked cell attachment to G3 domain-coated plates suggest a direct interaction between ␤ 1 -integrin and G3.
When the cell lysate prepared from cells stably transfected with the G3⌬EGF construct was subjected to native gel elec-

FIG. 7. Involvement of integrin and G3⌬EGF in cell adhesion.
In a blocking assay, treatment with antibody JB1a reduced cell attachment to His 6 -G3-coated plates (A). In competition assays, U87 cells were incubated with one of the following: G3⌬EGF-containing medium (control), G3⌬EGF-containing medium pre-absorbed by integrin (G3⌬EGF Ϫ ), or medium containing BSA. Treatment with integrin (G3⌬EGF Ϫ ) reduced cell attachment (B). Purified His 6 -G3 (6xHis-G3 ϩ ) or eluate from the vector control was incubated with U87 cells, and the cells were seeded on His 6 -G3-coated plates. Preincubation with His 6 -G3 reduced cell attachment (C). U87 cells were incubated with the G3⌬EGF-containing medium in the presence (D) or absence (E) of free His 6 -G3, followed by probing with antibody 4B6 and FITC-conjugated secondary antibody. Treatment with free His 6 -G3 reduced the binding activity.
FIG. 8. G3⌬EGF enhances integrin expression. G3⌬EGF-and vector-transfected U87 cells were seeded on tissue culture plates in DMEM containing 10% FBS for 1, 2, 3, and 24 h at equal cell density. The cell lysate was prepared, analyzed by Western blotting, and probed with antibody JB1a. G3⌬EGF-transfected cells expressed higher levels of integrin within 3 h after cell inoculation, but no difference was detected after 24 h of cell inoculation. trophoresis and Western blotting, antibody against ␤ 1 -integrin or G3⌬EGF detected bands with similar patterns. This implies that these two molecules are important partners; they interacted with each other and comigrated upon gel electrophoresis. This result did not exclude the possibility that ␤ 1 -integrin and G3⌬EGF could also interact with other molecules. For example, integrins may have interacted with other ECM molecules, resulting in large complexes that could not enter the gel. G3⌬EGF might have interacted with other (smaller) molecules, forming smaller complexes that migrated out of the gel. However, our immunocytometry results indicated that G3⌬EGF strongly interacted with the cell surface, suggesting a physiological significance of G3⌬EGF binding to ␤ 1 -integrin.
Although all products produced by U87, COS-7, and E. coli cells interacted with U87 cells, the binding affinities varied. The stable U87 cell line transfected with G3⌬EGF had the highest affinity for G3⌬EGF binding as assayed by flow cytometry. The COS-7 cells expressing G3⌬EGF had moderate affinity for binding U87 cells, whereas the bacteria-produced G3 domain had the lowest affinity. G3⌬EGF secreted by U87 cells was perhaps the closest to the physiological state required for ␤ 1 -integrin binding. Exogenously added G3⌬EGF produced by COS-7 cells might have already interacted with other components in the culture medium or was less accessible to integrins on the cell surface. The G3 domain produced by E. coli may not have exhibited the proper confirmation (compared with that expressed by mammalian cells) and thus had lower binding activity for the cell surface. Notably, these binding data are in agreement with the experimental results of the cell attachment and apoptosis assays. Plates coated with the E. coli-expressed G3 domain showed reduced cell attachment compared with those coated with the COS-7 cell-expressed product (Fig. 4), and the E. coli-expressed G3 product only partly protected U87 cells from free radical-induced cell apoptosis (Fig. 9). Nevertheless, these results indicated that the purified G3 product inter-acted with U87 cells, enhanced cell attachment, and reduced apoptosis of cells exposed to free radicals.
Although the products of mini-versican and G3⌬EGF enhanced the attachment of different cell types, including U87 cells, NIH3T3 fibroblasts, MDA-MB-231 breast cancer cells, and Jurkat tumor cells, the degree of enhancement varied. This might be due to the difference in the expression levels of ␤ 1 -integrin, which interacts with the G3 domain of versican. At the moment, it is not known which ␣ chain of the integrins is required to combine with the ␤ 1 chain and binds versican. It is possible that only one ␣ chain is involved in this binding. It is also not clear if other surface proteins of U87 cells play roles in binding to the versican G3 domain. Although CD44 has been reported to bind versican (29), it does not bind to the G3 domain.
It was interesting to observe that the expression of ␤ 1 -integrin decreased greatly in vector-transfected U87 cells after free radical treatment. These data support the hypothesis that ␤ 1integrin plays an important role in protecting the cells from free radical-induced apoptosis. However, the results shown did not exclude the possibility that the decreased detection levels upon Western blotting were due to an increase in protein degradation rather than a decrease in protein synthesis. To examine this, we harvested proteins in the culture medium by trichloroacetic acid precipitation and analyzed them by Western blotting. We did not detect a difference in the degraded fragment of ␤ 1 -integrin (data not shown). It seems that in the presence of free radical, integrin expression was reduced, although degradation of the protein was not excluded. In all of these cases, the presence of G3⌬EGF elevated the levels of cell-associated ␤ 1 -integrin in addition to increasing signaling, and this protected the cells from free radical-induced apoptosis. This was further demonstrated by coating the culture plate with the purified G3 domain, which decreased cell apoptosis in response to free radical treatment. and apoptosis (C and D; annexin V-FITC labeling) indicated that G3⌬EGF-transfected cells had a much higher rate of cell survival and a lower rate of apoptosis. G3⌬EGF-and vector-transfected U87 cells in DMEM containing 10% FBS and 0.1 mM H 2 O 2 were seeded on tissue culture plates and incubated for 0 or 10 min or 4 h (E). Probing with antibody JB1a showed that integrin expression increased in G3⌬EGF-transfected cells, but decreased in vector-transfected cells. G3⌬EGF-and vector-transfected U87 cells were cultured on tissue culture plates in DMEM containing 10% FBS (F). After 24 h, H 2 O 2 (0.1 mM) was added, and the treatment lasted for 30 min or 1 or 4 h. Integrin expression decreased in vector-transfected cells, but not in G3⌬EGFtransfected cells. U87 cells seeded on tissue culture plates precoated with purified His 6 -G3 (blue) or BSA (red) were treated with 1.5 mM H 2 O 2 at 37°C for 4 h, followed by labeling with annexin V-FITC (G). The presence of His 6 -G3 protected the cells from H 2 O 2 -induced apoptosis.
In summary, our study has demonstrated that G3⌬EGF interacts with ␤ 1 -integrin and binds to the cell surface. These interactions are correlated with FAK activation, enhanced cell adhesion, and protection of cells from free radical-induced apoptosis. The exact molecular mechanism of this binding and its physiological significance are prospects for future study. It has been established that PG-M/versican also binds to hyaluronan through the G1 domain and decreases cell adhesion (20,30,34), a role similar to that of the aggrecan G1 domain (42). This suggests an intriguing new model for regulation of cell adhesion in vivo: the PG-M/versican-hyaluronan interaction (via the G1 domain) may counteract the effects of the G3 domainintegrin interaction. When the G1 domain is not present, the G3-integrin interaction increases cell adhesion and therefore promotes cell survival. These counteractive effects of the G1 and G3 domains have also been extensively studied in the product processing of versican and aggrecan (43)(44)(45)(46)(47)(48)(49). It would be interesting to understand how the G1 and G3 domains modulate the functions of these chondroitin sulfate proteoglycans. Also of great importance is the fact that the G3⌬EGFintegrin binding reported here represents an entirely new mechanism for integrin binding because G3⌬EGF does not contain an RGD sequence, the consensus-binding sequence for integrins. Determination of the sequence involved in the reported interaction will require further studies.