Occurrence of Oligosialic Acids on Integrin (cid:1) 5 Subunit and Their Involvement in Cell Adhesion to Fibronectin*

Integrin (cid:1) 5 (cid:2) 1 , a major fibronectin receptor, functions in a wide variety of biological phenomena. We have found that (cid:1) 2–8-linked oligosialic acids with 5 < degree of polymerization (DP) < 7 occur on integrin (cid:1) 5 subunit of the human melanoma cell line G361. The integrin (cid:1) 5 subunit immunoprecipitated with anti-integrin (cid:1) 5 antibody reacted with the monoclonal antibody 12E3, which recognizes oligo/polysialic acid with DP > 5 but not with the polyclonal antibody H.46 recognizing oligo/polysi-alic acid with DP > 8. The occurrence of oligosialic acids was further demonstrated by fluorometric C 7 /C 9 analy- sis on the immunopurified integrin (cid:1) 5 subunit. Oligosialic acids were also found in the (cid:1) 5 subunit of several other human cells such as foreskin fibroblast and chronic erythroleukemia K562 cells. These results suggest the ubiquitous modification with unique oligosialic acids occurs on the (cid:1) 5 subunit of integrin (cid:1) 5 (cid:2) 1 . The adhesion of human melanoma G361 cells to fibronectin was mainly mediated by integrin (cid:1) 5 (cid:2) 1 . Treatment of cells with sialidase from Arthrobacter ureafaciens cleaving (cid:1) 2–3-, (cid:1) 2–6-, and (cid:1) 2–8-linked sialic acids inhibited adhesion to fibronectin. On the other hand, N -acetyl-neuraminidase II, which cleaves (cid:1) 2–3 and (cid:1) 2–6 but not (cid:1) 2–8 linkages, showed no inhibitory activity. After the loss of oligosialic acids, integrin (cid:1) 5 (cid:2) 1 failed to bind to fibronectin-conjugated Sepharose, indicating that the oligosialic acid on the (cid:1) 5 subunit of integrin (cid:1) 5 (cid:2) 1 plays important roles in cell adhesion to fibronectin. (Juntendo Y. Inoue (Aca-demia respectively. Horseradish peroxidase- conjugated goat affinity-purified antibody to mouse IgM, horseradish peroxidase-conjugated goat IgG fraction to rabbit IgG, and mAb P4C10 against the human integrin (cid:2) 1 subunit were purchased from Phar-maceuticals, OH). Synthetic Gly-Arg-Gly-Asp- Ser-Pro and Gly-Arg-Gly-Glu-Ser-Pro, were obtained from Inc. (San G-Sepharose, CNBr-activated Sepharose 4B, and purchased from Amersham Pharmacia Biotech. Sigma, N (cid:2) prestained molecular weight markers Bio-Rad. kindly (cid:2) 1 to fibronectin dependent on the presence of oligosialic acids. Cell lysates obtained from G361 cells were treated with or without 20 milli-international units/ml NAN- ase-II, A. ureafaciens sialidase, or Endo-N at 37 °C for 3 h and then incubated with 5 mg/ml fibronectin-Sepharose at 4 °C for 2 h. Sepha- rose beads with associated proteins were then precipitated by centrifugation and washed four times in lysis buffer. Bound proteins were analyzed by Western blotting using the anti-integrin (cid:1) 5 subunit polyclonal antibody.

Integrins comprise a structurally and functionally complex family of integral membrane glycoproteins that mediate cell adhesion to extracellular matrices, intracellular adhesion, and binding of plasma proteins. Such interactions also seem crucial for many normal and disease processes including those of wound healing (1), embryonic development (2), the maintenance of tissue integrity (3), and metastasis of cancer cells (4,5). The interaction of cells expressing integrin ␣ 5 ␤ 1 with fibronectin results in augmented cell adhesion, migration, and differentiation and has been implicated in important events in early embryogenesis such as gastrulation (6). Although the molecular basis of the integrin-ligand interaction is not understood completely, a line of experimental evidence suggests that post-translational modifications of integrin ␣ 5 ␤ 1 with carbohydrates are essential for cells to attach to fibronectin. The ␣ and ␤ subunits of integrin ␣ 5 ␤ 1 contained 14 and 13 potential asparagine-linked glycosylation sites in their extracellular region, respectively (7). Previous studies indicated that each of the respective chains contained asparagine-linked carbohydrates accounting for a molecular mass of 20,000 -30,000 daltons (8), and their carbohydrate modifications were involved in the binding to fibronectin (9 -11). In the present study, we demonstrated the occurrence of ␣2-8 oligosialic acids on integrin ␣ 5 ␤ 1 by applying highly sensitive immunochemical and chemical methods (12,13).
Little attention has been paid to the occurrence and functions of such short sialyl oligomers on glycoproteins. Polysialic acids on glycoproteins were first found in a glycoprotein isolated from unfertilized eggs of rainbow trout (Oncorhynchus mykiss) (14). After this discovery, ␣2-8-linked oligo/polysialic acid 1 has been shown to occur in various animal specimens obtained from insect to human more frequently than hitherto recognized (12,(15)(16)(17)(18). Although the functions of oligo/polysialic acids are not fully understood, it has been shown that ␣2-8-linked polysialic chains are attached to the embryonic form of neural cell adhesion molecule (N-CAM) 2 in mammals (16,19,20), and their chain length dramatically decreases to oligosialic acids in the adult brain (19). Polysialic acids are known to negatively modulate N-CAM-associated cell-cell interactions during neurite outgrowth and synaptogenesis because of weak adhesive properties as compared with N-CAM lacking polyanionic polysialic acids (21,22). Using highly sensitive chemical and immunochemical methods to detect oligosialic acids recently established by Sato et al. (12,16), oligosialic acids have been found in embryonic and adult mammalian brains, and expression of oligosialic acids was shown to vary during development and cell differentiation in a stage-specific manner (17). Furthermore, oligosialic acids were expressed in a number of cultured mammalian cell lines such as human myeloid cell line HL-60, tetracarcinoma cell line PA-1, mouse neuroblastoma cell line Neuro2A, myoblastoma cell line C2C12, and preadipose cell line 3T3-L1, and their expression was reported to change drastically by the induction of differentiation (23). Oligosialic acids are thus strongly suggested to play important roles in embryonic development and differentiation. Our present study demonstrates that ␣2-8linked oligosialic acids on the ␣ 5 subunit of integrin ␣ 5 ␤ 1 are integral for the integrin to bind to fibronectin.
Cell Culture-Human melanoma G361 cells and human chronic erythroleukemia K562 cells were obtained from the American Type Culture Collection (Manassas, VA). Fibroblasts were prepared from human fetal foreskin. Cells were cultured in Dulbecco's modified Eagle's medium (ICN Biochemicals, Inc., Aurora, OH) supplemented with 10% (v/v) fetal calf serum at 37°C under a humidified 95% air/5% CO 2 atmosphere.
Immunoprecipitation-After being washed, the cells were solubilized by lysis buffer (20 mM Tris-HCl, pH 8.0, 0.2% Triton X-100, 1.5 mM MgCl 2 , 0.15 M NaCl 2 , 1 mM phenylmethanesulfonyl fluoride, and 2 mM N-ethylmaleimide) for 30 min at 4°C. The extract was clarified by centrifugation for 5 min at 1,000 ϫ g and then for 10 min at 13,000 ϫ g. Protein G-Sepharose was preincubated with 50 g/ml anti-integrin ␣ 5 antibody (mAb KH/33) or normal mouse IgG for 2 h at 4°C. After incubation with protein G-Sepharose, lysates were incubated with protein G-Sepharose pretreated with mAb KH/33 or normal mouse IgG. After incubation for 2 h at 4°C, the beads were collected by centrifugation and washed with lysis buffer, and the immunoprecipitates were eluted by boiling at 60°C for 20 min with Laemmli buffer. In other experiments, cell lysate was concentrated and then dialyzed with PBS using Ultrafree®-C3TK (Millipore). Immunoprecipitated proteins or cell lysates were subjected to electrophoresis on 6% SDS-polyacrylamide gel electrophoresis (PAGE).
Immunoblotting-Proteins were electrophoretically transferred to a nitrocellulose filter or a PVDF membrane for immunoblotting as described (25). The blots were incubated for at least 2 h in 1% bovine serum albumin (BSA) in PBS (blocking buffer) followed by the incubation with polyclonal rabbit antibodies against integrin ␣ 5 subunits (diluted 1:500 in the blocking buffer) at room temperature for 2 h, or mAb 12E3 and polyclonal antibody H.46 against oligosialic acid and polysialic acid, respectively (diluted 1:500), at 4°C overnight. Secondary horseradish peroxidase-conjugated goat IgG fraction to rabbit IgG (diluted 1:5,000) and horseradish peroxidase-conjugated goat affinity-purified antibody to mouse IgM (diluted 1:1,000) were used at room temperature for 1 h. An ECL system was used for the detection of the immunoblotting analysis. To confirm whether the bands detected on the membrane were specially recognized by mAb 12E3, transblotted membrane was pretreated with A. ureafaciens exosialidase (250 milli-international units) or Endo-N (20 milli-international units) in 2.0 ml of 1% BSA/PBS containing 0.05% Tween 20 at 37°C for 20 h after blocking. The treated membranes were washed three times with PBS containing 0.05% Tween 20 for 10 min, blocked, and immunostained with mAb 12E3 or polyclonal antibody H.46.
Fibronectin Binding Assay-5 mg/ml fibronectin (bovine plasma, Seikagaku Co.) was coupled to CNBr-activated Sepharose 4B according to manufacturer instructions. Cell lysates were treated with or without 20 milli-international units/ml sialidases at 37°C for 3 h in 20 mM Tris-HCl, pH 7.5, containing 0.15 M NaCl and then incubated with fibronectin-Sepharose 4B in 20 mM Tris-HCl, pH 8.0, containing 1% n-octyl-␤-D-glucopyranoside, 1.5 mM MgCl 2 , 0.15 M NaCl 2 , 1 mM phenylmethanesulfonyl fluoride, and 2 mM N-ethylmaleimide for 2 h at 4°C. Sepharose beads and associated proteins were then precipitated by centrifugation, washed four times in the lysis buffer, resuspended in SDS-PAGE sample buffer, and boiled for 2 min. The released proteins were resolved on 6% polyacrylamide gels under nonreducing conditions.
Cell Adhesion Assay-Cell adhesion assays were performed as described previously with minor modifications (27). Briefly, 100 l of bovine plasma fibronectin (10 g/ml in PBS) were dispensed into each well of 96-well polystyrene microtiter plates (Costar Corp., Cambridge, MA) and adsorbed overnight at 4°C. The coated wells were washed three times with RPMI 1640 medium containing 10 mM HEPES, pH 7.4, and 0.03% BSA, blocked with 100 l of RPMI 1640 medium containing 10 mM HEPES, pH 7.4, and 3% BSA at 37°C for 2 h, and washed with RPMI 1640 medium containing 10 mM HEPES, pH 7.4, and 0.03% BSA.
Cells were harvested by washing four times with PBS (without Ca 2ϩ and Mg 2ϩ cations) and incubating in 2 ml of PBS (without Ca 2ϩ and Mg 2ϩ cations) containing 0.1% trypsin (Difco) and 0.01% EDTA for 1 min. The cells were dislodged by shaking, suspended in 10 ml of standard growth medium, allowed to recover from trypsin treatment for 15 min, and counted using a hemocytometer. The cells were then centrifuged and resuspended in RPMI 1640 medium containing 10 mM HEPES, pH 7.4, and 0.3% BSA at a concentration of 1 ϫ 10 6 cells/ml. Cells were labeled with 2Ј,7Ј-bis-(2-carboxyethyl)-5 (and -6)carboxyfluorescein (Molecular Probes) and suspended with RPMI 1640 containing 10 mM HEPES, pH 7.4, and 0.3% BSA. The cells were transferred into coated wells at 3 ϫ 10 4 cells/well and then incubated at 37°C for 30 min. Nonadherent cells were removed with four 21-gauge needle aspirations. Bound cells were quantified in the 96-well plate using a Cyto Fluor® multiwell plate reader series 4000 (Perseptive Biosystems, Foster City, CA). Specific attachment was calculated by subtracting the number of cells attached to BSA from the number of cells attached to fibronectin. The percentage of adherence was calculated by the following formula: percentage adherence ϭ (fluorescence intensity (FI) of experimental wells Ϫ FI of BSA-coated wells)/(FI of unwashed wells Ϫ FI of BSA-coated wells) ϫ 100.
Inhibition of Cell Adhesion-Labeled cells were preincubated with mAbs at 37°C for 20 min. The cells were then added to precoated wells and incubated for 30 min at 37°C. The wells were then washed, and adherent cells were quantified as described above for the cell adhesion assay.
Treatment of Cells with Sialidases-The treatment of cells with sialidases was carried out before the cell adhesion assays. Cells (1 ϫ 10 5 cells) were treated with 0.5 international units/ml exosialidase from A. ureafaciens, 0.5 international units/ml NANase-II from C. perfringens, or 0.1 international units/ml Endo-N from bacteriophage K1F (24) in 200 l of buffered medium (RPMI 1640 medium/PBS (1:1), pH 6.8) for 1 h at 37°C.
Flow Cytometry Analysis-Cells were treated with or without sialidase or mAb 12E3, suspended at 10 6 cells/ml and incubated with mAb JBS5 in PBS containing 0.3% BSA for 60 min for 4°C. The cells were washed with PBS containing 0.3% BSA, incubated with fluorescein isothiocyanate-conjugated goat anti-mouse IgG for 30 min at 4°C, washed, and analyzed by EPICS XL (Coulter Corp., Miami, FL). Nonimmune mouse IgG was used as a negative control.

Integrin ␣ 5 Subunit of Human Melanoma G361 Cells Susceptible to Exosialidase Capable of Cleaving ␣2-8-linked Sialic
Acids-To probe the occurrence of ␣2-8-linked oligosialic acids, the specimen was digested with several different sialidases. NANase-II can cleave ␣2-3 and ␣2-6 but not ␣2-8 linkages of Neu5Ac and N-glycolylneuraminic acid, whereas exosialidase from A. ureafaciens can split ␣2-3, ␣2-6, and ␣2-8 linkages of Neu5Ac and N-glycolylneuraminic acid. Treatment of the cell lysate of human melanoma G361 cells with NANase-II hardly affected the apparent M r of integrin ␣ 5 subunit even if the incubation time was prolonged ( Fig. 1). Digestion with exosialidase from A. ureafaciens yielded a sharp band with an M r of 145,000 at the faster moving position than that of intact ␣ 5 subunit with an M r of 150,000 ( Fig. 1). Although optimal pH values for sialidases from A. ureafaciens and NANase-II were reported to be 5.0 and 6.0, respectively, enzyme digestion was performed at pH 7.5 in 20 mM Tris-HCl containing 0.15 M NaCl, because oligosialic acids are sensitive to acidic conditions. The NANase-II-insensitive sialic acid linkages could have been cleaved by exosialidase from A. ureafaciens. Taking all these into consideration, the occurrence of ␣2-8-linked oligosialic acids of integrin ␣ 5 subunit of human melanoma G361 cells has been strongly suggested.
Immunochemical Evidence for the Occurrence of Oligosialic acids on Integrin ␣ 5 Subunit from Human Melanoma G361 Cell Line-The immunoprecipitate of human melanoma G361 cells with the anti-integrin ␣ 5 antibody mAb KH/33 was subjected to SDS-PAGE and subsequently transferred to a PVDF membrane. When the membrane was visualized with the anti-integrin ␣ 5 polyclonal antibody, a broad band with an apparent M r value of ϳ150,000 was detected (Fig. 2). The band, corresponding to integrin ␣ 5 subunit, was also stained with the anti-oligo/ polysialic acid antibody mAb 12E3, which recognizes ␣2-8linked Neu5Ac oligomers with a degree of polymerization (DP) Ն 5. However, the integrin ␣ 5 subunit hardly reacted to equine polyclonal antibody H.46, which recognizes ␣2-8-linked Neu5Ac polymers with a DP Ն 8. These results suggest that the integrin ␣ 5 subunit possesses oligosialic acids with 5 Յ DP Յ 7. The electroblotted membranes were pretreated with exosialidase from A. ureafaciens, Endo-N, or peptide-N4-(N-acetyl-␤glucosaminyl)asparagine amidase and then stained with mAb 12E3. Endo-N is known as an endosialidase that hydrolyzes ␣2-8-linked oligo/polysialic acids (24). The staining of the integrin ␣ 5 subunit with mAb 12E3 completely disappeared after the treatment with exosialidase from A. ureafaciens, Endo-N, and peptide-N4-(N-acetyl-␤-glucosaminyl)asparagine amidase. However, pretreatment of the membrane with NANase-II did not diminish the staining with mAb 12E3. Thus, the results indicate that the integrin ␣ 5 subunit contained N-linked glycan chains with unique oligosialic acids with an average degree of polymerization of 5-7. Integrin ␣ 5 ␤ 1 can be immunoprecipitated as a heterodimer in the presence of a divalent cation. Although the integrin ␤ 1 subunit migrated to the position at an M r of 120,000 on SDS-PAGE (data not shown), the immunoreactivity at the position of M r 120,000 with 12E3 was not observed, indicating that oligosialic acid occurs in the integrin ␣ 5 but not ␤ 1 subunit. Although G361 cells express the 140-kDa N-CAM isoform with oligosialic acid (16,19,20), the immunoprecipitate with mAb KH/33 did not stain with the anti-N-CAM antibody (data not shown), showing that staining with mAb 12E3 was attributable solely to oligosialic acid of ␣ 5 subunit but not to N-CAM.
Fluorometric Detection of Internal Sialic Acid Residues-For chemical detection of ␣2-8-linked oligosialic acid, the periodate oxidation/fluorometric HPLC method (12) was applied to immunopurified integrin ␣ 5 subunit. The internal sialic acid residues in ␣2-8-linked oligosialyl chains were determined after sequential periodate oxidation, borohydride reduction, and mild acid hydrolysis according to a modified procedure of van Lenten and Ashwell (28). Briefly, when an ␣2-8-linked oligosialyl chain is subjected to periodate oxidation, the nonreducing terminal residue is oxidized to give rise to the C 7 analogue of the Neu5Ac or N-glycolylneuraminic acid (5-acetoamido-3,5dideoxy-L-arabino-2-hepturosonic acid or 5-hydroxyacetoamide-3,5-dideoxy-L-arabino-2-hepturosonic acid), whereas internal residues linked by ␣2-8 linkages remain intact as the C 9 analogue of Neu5Ac or N-glycolylneuraminic acid (2-keto-5acetoamido-3,5-dideoxy-D-glycero-D-galacto-nonoic acid or 2-keto-5-hydroxyacetoamido-3,5-dideoxy-D-glycero-D-galactononoic acid) (29). Thus, the detection of C 9 analogues in the periodate oxidation products would further strongly support the presence of ␣2-8-linked oligosialic acids. Immunoprecipitated integrin ␣ 5 was electrophoresed and transferred to a PVDF membrane. When the membrane was stained with Coomassie Brilliant Blue, the band with an M r value of 150,000 corresponding to integrin ␣ 5 was detected (Fig. 3). The membrane pieces containing integrin ␣ 5 were excised and subjected to fluorometric C 7 /C 9 analysis. C 7 and C 9 derivatives of Neu5Ac were detected in the molar ratio of 143:5 after complete hydrolysis of the oxidized and reduced integrin ␣ 5 subunit. Detection of the C 9 derivative was a strong ground for the occurrence of oligomeric sialic acids in the integrin ␣ 5 subunit. Assuming that an oligo-sialic acid of the integrin ␣ 5 subunit consists of an average of five internal sialic acid residues (refer to the previous section), an oligosialic acidcontaining chain occurs in 1 of 143 termini of N-linked oligosaccharides. According to the structural analysis of N-linked oligosaccharides of integrin ␣ 5 ␤ 1 (30), one N-linked sugar chain is estimated to bear an average of 1.70 termini capped by ␣2-3 or ␣2-6 sialic acid residue. Because integrin ␣ 5 contains 14 potential N-linked glycosylation sites (7), integrin ␣ 5 contains 23.8 termini capped by ␣2-3 or ␣2-6 sialic acid residue. Thus, an oligosialic acid-containing chain appears at the frequency of one of six molecules of integrin ␣ 5 ␤ 1 . In the control experiment in which the immunoprecipitate was made with normal IgG in place of the mAb KH/33, the C 9 derivative was not detected by fluorometric C 7 /C 9 analysis.
Ubiquitous Modification of Integrin ␣ 5 Subunit with Oligosialic Acids-Whether integrin ␣ 5 subunits derived from other cells of different origins were likewise modified with oligosialic acid was determined using human foreskin fibroblast and chronic erythroleukemia K562 cells. The immunoprecipitate of each cell with the mAb KH/33 was analyzed using anti-integrin ␣ 5 polyclonal antibodies, mAb 12E3 and polyclonal antibody H.46. The integrin ␣ 5 subunit of both cells reacted to mAb 12E3 but not to polyclonal antibody H.46 (Fig. 4A). Cell lysates of fibroblast and K562 cells were treated with NANase-II or sialidase from A. ureafaciens, separated by SDS-PAGE, and stained using an anti-integrin ␣ 5 antibody and mAb 12E3. Integrin ␣ 5 subunits of fibroblast and K562 cells were hardly affected by treatment with NANase-II, although these were sensitive to exosialidase from A. ureafaciens (Fig. 4B). It then seems that ␣2-8 oligosialic acids occur on integrin ␣ 5 subunits of human foreskin fibroblast and K562 cells. When immunopurified integrin ␣ 5 subunit from human foreskin fibroblast and K562 cells (Fig. 3) were subjected to fluorometric C 7 /C 9 analysis, C 7 and C 9 derivatives of Neu5Ac in fibroblast and K562 cells were detected in the molar ratios of 50:5 and 45.6:5, respectively. Detection of the C 9 derivative supports the occurrence of ␣2-8-linked oligosialic acids, and it is calculated that an oligosialic acid-bearing chain occurs every other molecule of integrin ␣ 5 in both samples. The results indicate that integrin ␣ 5 subunits of human foreskin fibroblast and K562 cells contain oligosialic acids and that further modification of the integrin ␣ 5 subunit with oligosialic acid is ubiquitous, implying its biological significance.

Adhesion of Human Melanoma G361 Cells to Fibronectin
Mediated Mainly by Integrin ␣ 5 ␤ 1 -To establish the functional role of oligosialic acid of integrin ␣ 5 subunit in cell adhesion, the mediation of integrin ␣ 5 ␤ 1 in cell adhesion to fibronectin was determined in human melanoma G361 cells. Integrin ␣ 5 ␤ 1 is known as a major fibronectin receptor, recognizing the Arg-Gly-Asp (RGD) sequence in fibronectin (31). The adhesion of G361 cells to a 10 g/ml fibronectin-immobilized plate was dose-dependently inhibited by the synthetic Gly-Arg-Gly-Asp-Ser-Pro peptide (Fig. 5A). Treatment of G361 cells with function-blocking anti-integrin ␣ 5 or ␤ 1 mAb resulted in the inhibition of adhesion of G361 cells to fibronectin. The anti-integrin ␤ 1 antibody mAb P4C10 (32) and anti-integrin ␣ 5 antibody mAb KH/33 inhibited cell adhesion by 70 and 90%, respectively (Fig. 5B). Thus, adhesion of human melanoma G361 cells to fibronectin was shown to be mainly mediated by integrin ␣ 5 ␤ 1 .
Abolishment of the Fibronectin Binding Capacity of ␣ 5 ␤ 1 after the Loss of Oligosialic Acids-We investigated the binding ability of integrin ␣ 5 ␤ 1 to fibronectin after treatment with various sialidases to evaluate the roles of oligosialic acids of the integrin ␣ 5 subunit in fibronectin binding. G361 cell lysates were treated with sialidases and incubated with 5 mg/ml fibronectin-Sepharose at 4°C for 2 h, and bound proteins were analyzed by Western blotting using an anti-integrin ␣ 5 polyclonal antibody. Binding of integrin ␣ 5 ␤ 1 to fibronectin was not affected by NANase-II treatment, whereas integrin ␣ 5 ␤ 1 treated with sialidase from A. ureafaciens failed to bind to fibronectin (Fig. 6, lanes 2 and 4, respectively). Because integrin ␣ 5 ␤ 1 mixed with A. ureafaciens sialidase without incubation bound to fibronectin, it is suggested that A. ureafaciens sialidase effect on immobilized fibronectin was ruled out (Fig.   FIG. 3. Purity of immunoprecipitated integrin ␣ 5 subunit. Immunoprecipitates from cell lysates with the mAb KH/33 raised against integrin ␣ 5 subunit were electrophoresed, transferred to a PVDF membrane, and stained with anti-integrin ␣ 5 polyclonal antibody (lane 1) or Coomassie brilliant blue (lanes 2-4). The bands corresponding to integrin ␣ 5 , indicated by the bar, were excised and then subjected to C 7 /C 9 chemical analysis. Lanes 1 and 2, melanoma G361 cells; lane 3, K562 cells; lane 4, fibroblast.

FIG. 4. Oligosialic acids occurring on integrin ␣ 5 subunit of human foreskin fibroblast and human chronic erythroleukemia K562 cells.
A, the integrin ␣ 5 subunit was immunopurified (IP) from fibroblast and K562 cell lysates using an anti-integrin ␣ 5 subunit antibody (mAb KH/33). The immunopurified integrin ␣ 5 subunit was analyzed by Western blotting using the anti-integrin ␣ 5 subunit polyclonal antibody mAb 12E3 and polyclonal antibody H.46. B, cell lysates from fibroblast and K562 cells were pretreated with or without NANase-II or sialidase from A. ureafaciens and then analyzed by Western blotting using an anti-integrin ␣ 5 polyclonal antibody and mAb 12E3. 6, lane 3). The effects of possible proteases in cell lysate on the integrin ␣ 5 were excluded, because the integrity of integrin ␣ 5 was not affected when cell lysate was incubated without sialidases under the same conditions (Fig. 6, lane 1). These results clearly indicate that ␣2-8-linked oligosialic acids of integrin ␣ 5 ␤ 1 are essential for the binding of integrin ␣ 5 ␤ 1 to fibronectin.
Involvement of Oligosialic Acids of Integrin ␣ 5 Subunit in Cell Adhesion to Fibronectin-Adhesion of G361 cells to a fibronectin-coated dish was not affected by pretreatment of cells with NANase-II (Fig. 7A). However, cell adhesion of cells pretreated with A. ureafaciens sialidase was decreased to 70% of the control. These results suggest that integrin ␣ 5 ␤ 1 -mediated cell adhesion to fibronectin was suppressed by the loss of oligosialic acids. Contrary to the experiment shown in Fig. 2, Endo-N showed no inhibitory activity in the cell adhesion experiment. Endo-N is reported to require a minimal DP of 5 to work on oligo/polysialic acids. Because the average DP of oligosialic acid of ␣ 5 subunit has been shown to be 5-7, Endo-N could trim the oligosialic acid of ␣ 5 subunit enough to lose its reactivity to mAb 12E3 but not enough to affect its binding capacity to fibronectin. Fig. 7A, inset, supports our notion by showing that Endo-N-treated ␣ 5 subunit moved to the same position as that of the control and the ␣ 5 subunit pretreated with NANase-II.
Whether the conformational change of integrin ␣ 5 was induced by the treatment with A. ureafaciens sialidase was in-vestigated. The reactivity of mAb JBS5 toward cells with or without A. ureafaciens sialidase treatment was investigated by flow cytometry. It has been proposed that the epitope of mAb JBS5 lies close to the RGD recognition site of ␣ 5 integrin (33). The pretreatment of cells with A. ureafaciens sialidase decreased reactivity to mAb JBS5 (Fig. 7B), whereas NANase-II showed no effect (data not shown). These results suggest that the conformation of integrin ␣ 5 was changed by the loss of oligosialic acids from integrin ␣ 5 subunit, possibly resulting in the suppression of binding ability to fibronectin.
Inhibition of Cell Adhesion to Fibronectin by the Treatment with Colominic Acids or mAb 12E3-To further demonstrate that oligosialic acids of integrin ␣ 5 subunit function in cell adhesion, the effects of colominic acids and mAb 12E3 on cell adhesion to fibronectin were investigated. Adhesion of G361 cells to fibronectin was not inhibited by colominic acid, a polymer of ␣2-8-linked sialic acid obtained from E. coli (Fig.  8A), suggesting that oligosialic acids of integrin ␣ 5 do not reside in the interface between integrin and fibronectin binding. To obtain direct evidence that oligosialic acids are involved in binding between integrin ␣ 5 ␤ 1 and fibronectin, cells were treated with the anti-oligosialic acid antibody mAb 12E3. Interestingly, treatment of cells with mAb 12E3 activated integrin ␣ 5 ␤ 1 -mediated cell adhesiveness to fibronectin by 40% (Fig. 8B), and the treatment with mAbs against integrin ␣ 5 and ␤ 1 inhibited adhesion to fibronectin as discussed in the preceding section (Fig. 5B). Because mAb 12E3 is composed of IgM, nonspecific interaction caused by IgM should be excluded. Therefore, a control experiment was performed using mAb LY111 (subclass IgM) recognizing chondroitin sulfate proteoglycan expressed on the cell surface of G361 cells. Although heparan sulfate proteoglycan is reported to be involved in cell adhesion to fibronectin mediated by integrin ␣ 5 ␤ 1 (34,35), there has been no report yet on the involvement of chondroitin sulfate proteoglycan. The G361 cell adhesion to fibronectin was not affected by the treatment with mAb LY111, suggesting that activation of cell adhesion  6. Binding of integrin ␣ 5 ␤ 1 to fibronectin dependent on the presence of oligosialic acids. Cell lysates obtained from G361 cells were treated with or without 20 milli-international units/ml NANase-II, A. ureafaciens sialidase, or Endo-N at 37°C for 3 h and then incubated with 5 mg/ml fibronectin-Sepharose at 4°C for 2 h. Sepharose beads with associated proteins were then precipitated by centrifugation and washed four times in lysis buffer. Bound proteins were analyzed by Western blotting using the anti-integrin ␣ 5 subunit polyclonal antibody. by mAb 12E3 should be ascribed to specific binding of mAb to oligosialic acid. Whether the stimulative effects of mAb 12E3 could be attributed to the exposure of the fibronectin recognition site of integrin ␣ 5 or the enhanced clustering of integrin ␣ 5 ␤ 1 was investigated. The reactivity of mAb JBS5 toward cells with or without mAb 12E3 treatment was determined by flow cytometry. Pretreatment of cells with mAb 12E3 enhanced reactivity to mAb JBS5 (Fig. 8C). Therefore, it is highly likely that the treatment of cells with mAb 12E3 caused a conformational change in integrin that unmasked sites involved in ligand recognition. Furthermore, immunofluorescent microscopic observation indicated that clustering of integrin ␣ 5 ␤ 1 was not induced by the treatment of cells with mAb 12E3 (data not shown). Taken together, it is suggested that the modification of integrin ␣ 5 with oligosialic acids enhances the binding capability of integrin ␣ 5 ␤ 1 to fibronectin, possibly by holding the conformation in the high affinity form to the ligand. DISCUSSION Glycoproteins with oligosialic acids have not yet been identified except for the adult form of N-CAM (16,19,20) and sodium channels (20). A previous structural study on N-linked oligosaccharides obtained from human placenta integrin ␣ 5 ␤ 1 did not mention oligosialic acids (30). By applying immunochemical and chemical analyses (12,16), we have succeeded in finding oligosialic acids of the integrin ␣ 5 subunit of human melanoma G361 cells (Fig. 2). In addition, modification of the integrin ␣ 5 subunit with oligosialic acids seems ubiquitous because occurrence of oligosialic acids was observed in the integrin ␣ 5 subunit of cells tested thus far including human melanoma G361, foreskin fibroblast, and chronic erythroleukemia K562 cells (Figs. 2 and 3). Extensive work on other cells may possibly prove the wide distribution of oligosialic acid on the ␣ 5 integrin subunit of other cell types. Integrin ␣ 5 ␤ 1 consists of ␣ and ␤ subunits, but oligosialic acids did not occur in the ␤ 1 subunit. Although the present study focused on oligosialic acids of the integrin ␣ 5 subunit, we do not rule out the possibility that oligosialic acids can also be present on proteins other than integrin ␣ 5 in human melanoma G361 cells.
␣2-8-Linked oligosialic acids in the ␣ 5 subunit were stained with mAb 12E3, which reacts with oligosialic acids with a DP Ն 5, but not with polyclonal antibody H.46, which recognizes polymers with a DP Ն 8. Treatment of the ␣ 5 subunit with A. ureafaciens sialidase and Endo-N completely abolished the reactivity to mAb 12E3, but treatment with NANase-II did not decrease the reactivity. Endo-N requires a minimal DP of 5 to work (24). From these studies, the degree of polymerization of ␣2-8-linked oligosialic acids in the ␣ 5 subunit was estimated to be 5-7. The treatment of G361 cells with A. ureafaciens sialidase brought about the loss of binding capability to fibronectin but failed to do so with Endo-N. Endo-N may attack the sialyl linkage near the nonreducing terminal of oligosialic acid chains, leading to a decrease in the degree of polymerization, which still allowed cells to interact with fibronectin, but did not support mAb 12E3 binding to the residual chains of oligosialic acids. The occurrence of internal sialic acids was confirmed further by the C 7 /C 9 analysis. On the assumption that the average degree of polymerization is 6, it was calculated from the ratio that integrin ␣ 5 contains an oligosialic acid-bearing chain every six integrin ␣ 5 molecules in G361 cells and every two molecules in fibroblast and K562 cells.
In addition to the N-CAM and sodium channels being intensively studied, oligosialic acids were implied to play a role in embryonic development because their amount changed in a stage-specific manner (17). However, direct evidence on the functional roles of oligosialic acids is yet to be determined. We have investigated the function of oligosialic acids of integrin ␣ 5 ␤ 1 , known as a fibronectin receptor, and its interaction with fibronectin is considered to be involved in various biological phenomena. Previous studies showed that cell adhesion mediated by integrin was regulated by glycosylation. Akiyama et al. (9) reported that blockage of N-glycan processing by 1-deoxymannojirimycin, an inhibitor of Golgi mannosidases, resulted in inhibition of the fibronectin-binding capability of integrin ␣ 5 ␤ 1 of mouse 3T3 cells. Zheng et al. (10) showed that Nglycosylation of both the ␣ and ␤ subunits of integrin ␣ 5 ␤ 1 is essential for association of these subunits and for the receptor to function normally. Koyama and Hughes (36) described that cell adhesion to fibronectin was impaired in the N-acetylglucosaminyl transferase I-deficient cells, which were unable to synthesize any hybrid or complex carbohydrates including those carrying oligosialic acids. In other studies, phorbol ester treatment of human multipotential hematopoietic cell line K562 and promonocytic cell lines U937 and THP-1 induced the decrease in sialylation of integrin ␣ 5 ␤ 1 , leading to the suppression of cell adherence to fibronectin (11,37). In our study, we have demonstrated that unique ␣2-8-linked oligosialic acids located on N-linked outer chains of the integrin ␣ 5 subunit are involved in cell adhesion to fibronectin. Adhesion of human melanoma G361 cells to fibronectin mediated mainly by ␣ 5 ␤ 1 was inhibited when cells were treated with sialidase from A. ureafaciens but not with NANase-II (Figs. 6 and 7). The decrease in binding to fibronectin was not caused by dissociation of integrin ␣ 5 and ␤ 1 subunits, because the ␤ 1 subunit was co-immunoprecipitated using integrin ␣ 5 mAb whether cell lysate had been treated with sialidases or not (data not shown).
Previous studies provided evidence that oligosaccharides of integrin ␣ 5 are indispensable for the binding to fibronectin. However, explanations have not been given on how the effect of the sugar moiety is achieved. In the present study, we showed that the treatment of cells with A. ureafaciens sialidase decreased reactivity of the ␣ 5 subunit to mAb JBS5, the epitope of which is closely located at the RGD recognition site of ␣ 5 (Fig.  7). Thus, it is suggested that the conformational change has been induced after the loss of oligosialic acids, leading to masking of the ligand-binding domain of ␣ 5 . As shown in Fig. 8, the mAb 12E3 binding to ␣ 5 enhanced its reactivity to JBS5, suggesting that conformational change has also been induced. Treatment with mAb 12E3 may have exaggerated the conformational change introduced by the occurrence of oligosialic acid. Oligosialic acids of integrin ␣ 5 thus contribute to the binding capability of integrin ␣ 5 ␤ 1 to fibronectin, possibly by bringing the conformation into the high affinity form to the ligand.
␣2-8-Linked oligosialic acids are known to be common structural units of gangliosides implicated in various biological processes such as cell adhesion, cell differentiation, signal transduction, and surface expression of stage-specific antigen. Cell adhesion regulated by integrins is modulated by co-existing gangliosides. It was reported that based on immunohistochemical and biochemical analyses (38), GD2 and GD3 gangliosides were closely associated with the vitronectin receptor (integrin ␣ v ␤ 3 ) in human melanoma. Wang et al. (39) reported that GT1b regulated integrin ␣ 5 ␤ 1 -mediated adhesion of epithelial cells to fibronectin through carbohydrate-carbohydrate interactions between GT1b and the integrin ␣ 5 subunit. However, in human melanoma G361 cells, the occurrence of gangliosides containing ␣2-8 oligosialic acids such as GD2, GD3, or GT1b was not confirmed using thin layer chromatography analysis (data not shown). Thus, the loss of cell adhesion to fibronectin by sialidase treatment was suggested not to involve ␣2-8-oligosialylated gangliosides. Koyama and Hughes (36) reported that adhesion of BHK cells to fibronectin was not affected by treatment of 1-deoxymannojirimycin. Their finding is in contrast to our and previous studies (9) that indicate that hybrid or complex carbohydrates are important for the normal function of integrin ␣ 5 ␤ 1 . Such a conflict may possibly be caused by gangliosides on the cell surface. The adhesion of G361 cells to fibronectin were remarkably inhibited by 1-deoxymannojirimycin treatment (data not shown), probably because G361 cells scarcely express gangliosides containing ␣2-8 sialic acids. In the study by Koyama and Hughes, it is conceivable that gangliosides expressed on BHK cells compensated the loss of function of integrin by the treatment with 1-deoxymannojirimycin.
Of the known ␣2-8 sialyltransferases (ST8Sia), ST8Sia III is most likely to be the enzyme responsible for the synthesis of oligosialic acids on glycoproteins. Recently, Angata et al. (40) reported that oligosialic acids could be synthesized by ST8Sia III. However, although integrin ␣ 5 ␤ 1 is widely distributed, no expression of ST8Sia III was reported in tissues other than brain or testis. Thus, we may propose the presence of other types of ST8Sia for the oligosialylation of the integrin ␣ 5 subunit. It will be of interest to search for the biosynthetic enzyme of oligosialic acids by which the function of integrin is regulated.
The interaction of integrin ␣ 5 ␤ 1 with fibronectin plays an essential role in development. The integrin ␣ 5 null and fibronectin null embryos have pronounced defects in mesodermal structures, suggesting the role of both interacting molecules in mesoderm formation (41). However, it remains to be determined whether oligosialic acids of integrin ␣ 5 function in the interaction between integrin ␣ 5 ␤ 1 and fibronectin while mesoderm formation takes place. Based on the present study, the hypothesis that oligosialic acids are important for the interaction between integrin ␣ 5 ␤ 1 and fibronectin in the developmental stage seems quite plausible, particularly taking into account the occurrence of and dynamic change in oligosialic acids and the regulation of N-CAM function by oligosialylation during embryonic development.