Aberrant N-Glycosylation of β₁ Integrin Causes Reduced α₅β₁ Integrin Clustering and Stimulates Cell Migration

INTRODUCTION

Alterations of cell surface glycan structures are often associated with malignant transformation of cells. One of the most common is a significant increase in the levels of asparagine-linked glycans containing the β1,6 branch, β1,6-GlcNAc 3 linked to the α1,6-mannose of the trimannosyl core (1) . This GlcNAcβ1,6Man structure is synthesized by GnT-V (or Mgat5, EC 2.1.4.155), a key enzyme in the processing of multiantennary asparagine-linked glycans during the synthesis of glycoproteins (2 , 3) . Results from previous studies showed the expression of β1,6-GlcNAc branched N-linked oligosaccharides in human mammary, colon, hepatic, and glial tumors (4, 5, 6, 7, 8) and suggest that a relationship exists between N-linked β1,6 branching glycans on one hand and tumorigenesis and tumor invasion on the other.

Recent experiments have revealed the mechanisms by which oncogenesis causes changes of GnT-V activity. GnT-V transcription was stimulated by several oncogenes including src, her-2/neu, H-ras, and v-sis (9, 10, 11) , and this overexpression was regulated through the ras-raf-ets pathways. Stimulation of GnT-V activity has also been observed via the protein kinase C and phosphatidylinositol 3-kinase-protein kinase B signaling pathways (9, 10, 11, 12, 13) . GnT-V-deficient mice (GnT-V^−/− or Mgat5^−/−) have been used to study the effects of eliminating GnT-V activity on tumor progression (14) . When crossed with mice that express the polyoma middle T antigen under control of the mouse mammary tumor virus promoter, the progression of mammary tumors in the GnT-V^−/− mice was significantly reduced compared with that in the GnT-V^+/− mice. This study provides evidence that changes in N-linked β1,6 branching can affect carcinoma progression in vivo.

Since GnT-V cDNA was cloned by our and Taniguchi’s group (15 , 16) , several studies have transfected GnT-V sense or antisense cDNA into cultured cells, selected for clones with altered GnT-V expression, and investigated cell-adhesive properties (8 , 17, 18, 19) . Although these studies show that alterations in N-linked β1,6 branching may affect cell-adhesive behaviors, including cell adhesion and migration, the exact relationship were often inconsistent among the clones under study, and mechanisms involved in these events remain unclear. In one study, for example, two clones overexpressing GnT-V were selected and shown to have altered cell migration on plastic, high levels of apoptosis, and decreased adhesion to fibronectin and laminin, but the clone with intermediate GnT-V expression and β1,6 branching levels showed a greater alteration of adhesive properties than the clone with higher levels of GnT-V and β1,6 branches (17) .

Altered cell-cell and cell-ECM interactions are clearly critical during the acquisition of the invasive phenotype. Cell surface integrins, especially those binding to fibronectin and laminin, play essential roles in tumor cell invasion and metastasis. The fibronectin receptor α₅β₁ contains a total of 26 putative N-linked glycosylation sites (14 on α₅ and 12 on β₁), although it is not known which of these sites are used. Inhibitors of N-linked oligosaccharide processing have been shown to inhibit the ability of α₅β₁ to bind fibronectin but not to affect its cell surface levels or ability to dimerize (20 , 21) . The mechanisms by which changes in distal glycosylation, such as β1,6 branching, could affect cell adhesion and migration are, however, only now being understood.

To investigate in detail the relationship between increased N-linked β1,6 branching and α₅β₁-mediated adhesion to fibronectin and explore its possible mechanisms, we have used an inducible retroviral expression system. Our results show that increased expression of GnT-V activity and resulting increased levels of β1,6 branching inhibits attachment and spreading of cells on fibronectin but stimulates cell migration toward fibronectin and invasion through a reconstituted Matrigel basement membrane. Increased GnT-V expression cause increased levels of N-linked β1,6 branching on the β₁ subunit of the fibronectin receptor, but there is no evidence of these oligosaccharide structures on the α₅ subunit, even after increased GnT-V activity. Increased β1,6 oligosaccharide branching inhibits clustering of the α₅β₁ integrins and organization of F-actin into extended microfilaments when cells are plated on fibronectin-coated plates.

MATERIALS AND METHODS

Cell Lines and Materials.

HT1080 human fibrosarcoma cells (CCL-121) and Phoenix producer cells (SD3443) were obtained from the American Type Culture Collection; BSA, DMEM, human plasma fibronectin, swainsonine, 2-[bis(2-hydroxyethyl)amino]ethanesulfonic acid, SB3-10, and hexadimethrine bromide (Polybrene) were products of Sigma-Aldrich; retroviral plasmid (pTJ66) and the tet-off retroviral plasmid (pTJ68-IZG) were kindly provided by Dr. T. J. Murphy, Emory University. NHS-LC-biotin, Ultralink immobilization kit, and ECL assay kit were products of Pierce. Streptavidin-HRP was obtained from Rockland. Biotinylated l-PHA and ConA were products of Vector Laboratories. Rhodamine-phalloidin was from Molecular Probes. Protein A-agarose, protein-l-agarose, and Rhodamine-conjugated anti-rat IgG were from Santa Cruz Biotechnology. Gelatin gel (1%) was obtained from Bio-Rad. The 12-well chemotaxis Boyden chambers (Transwell) and the 24-well Matrigel-coated chemotaxis Boyden chamber (8.0 μm) were products of BD Falcon.

Construction of Retrovirus Plasmids Containing GnT-V cDNA.

A 2.5-kb fragment containing the full-length mouse GnT-V cDNA was cut from the pMMVGnT-V plasmid with SmaI. This fragment was then ligated with SfiI adapters (formed using 3′-GATCCGGA and 5′-CTAGGCCTACA) and cloned into the noncohesive universal cloning SfiI site of retroviral plasmid pTJ66 to produce pTJ66GnT-V. The orientation of the insert was confirmed by BstXI restriction digestion mapping. This vector uses a 5′-long terminal repeat promoter, includes an internal ribosome entry signal, and produces a chimera mRNA that simultaneously expresses the protein of interest and the Zeo:eGFP reporter/selection marker. pTJ68-IZG denotes the retroviral plasmid with inducible cytomegalovirus/tet-operon (tetO-CMV) promoter, which expresses the tetracycline trans-activator (tTA) protein (tet-off) and uses Zeo:eGFP as the reporter marker. For the inducible expression of GnT-V, pTJ68-IZG was linearized by digestion with BstXI, followed by dephosphorylation, and then ligated, after its 3′-overhang was removed using T₄ polymerase, with the 2.5-kb fragment of GnT-V to produce pTJ68-IZGGnT-V. BstXI and SapI digestions were used to identify the orientation of the insert.

Cell Culture, Retroviral Production, and Infection.

HT1080 and Phoenix producer cells were maintained and passaged routinely at 37°C in 5% CO₂ in DMEM growth medium containing 10% heat-inactivated fetal bovine serum (Atlanta Biologicals) supplemented with 0.1 mm nonessential amino acids, 2 mm l-glutamine, 1 mm sodium pyruvate, 100 units/ml penicillin, and 100 μg of streptomycin (Sigma-Aldrich).

Retroviral production and infection were performed as described previously (22) . Briefly, infectious retroviral supernatants were produced by a transient, helper virus-free protocol with pTJ66GnT-V and pTJ68-IZGGnT-V plasmids. Phoenix producer cells were grown to 50–80% confluence in DMEM in 100-mm culture dishes and transfected with retroviral plasmids (pTJ66GnT-V and pTJ68-IZGGnT-V, respectively) using 2 m CaPO₄ and 25 mm chloroquine in 50 mm 2-[bis(2-hydroxyethyl)amino]ethanesulfonic acid buffer, pH 7.0, for 6–8 h, and then the dishes were refed with 25 ml of growth medium. Twenty-four hours after transfection, the growth medium was aspirated, replaced with 9 ml of growth medium, and incubated in 5% CO₂ at 32°C to increase the retroviral titer. Supernatants containing infectious retrovirus were collected every 12 h up to 106 h after transfection and then filtered using a 0.45-μm pore size filter, aliquoted, snap frozen in liquid nitrogen, and stored at −80°C. At the same time, pTJ66 and pTJ68-IZG plasmids without GnT-V cDNA insert were used to generate control retroviral supernatants.

For the infection, HT1080 cells were grown in six-well plates and infected by adding retroviral supernatant and Polybrene (8 μg/ml). The plates were then incubated at 32°C for 15 min, followed by centrifuging at 2500 rpm for 30 min at 32°C. The infectious supernatant was aspirated; the cells were refed with growth medium and then placed back to the incubator with 37°C. The infection was repeated 8–12 h later. Cells were then cultured for 48 h after the second infection before selection in medium containing 800 μg/ml Zeocin for 3–5 days. The surviving cells (expressing GFP) were transferred into a culture flask and used for experiments. After two rounds of retroviral infections, >95% of the cells showed GFP fluorescence in each infection experiment when viewed with an inverted fluorescence microscope.

Transfected cells were maintained in DMEM plus 10% FBS. The cells with tet-inducible promoter were maintained in growth medium with 0.1 μg/ml tetracycline to suppress exogenous GnT-V expression (noninduced). Before each experiment, the cells containing the inducible promoter were induced to express GnT-V by the withdrawal of tetracycline from the culture medium (induced), and levels of expression were determined by altering the duration of time that cells were maintained in the medium without tetracycline. Unless specified, induction time was at least 36 h. For noninduced cells, parallel cultures were maintained with growth medium containing 0.1 μg/ml tetracycline.

GnT-V Activity Assay.

Cells were trypsinized, pelleted, and lysed with 50 mm 4-morpholinepropanesulfonic acid (pH 6.5), 150 mm NaCl, and 1% Triton X-100. Insoluble debris was pelleted by microcentrifugation (10 min at 4°C), and the supernatant was used for the GnT-V activity assay with UDP-[ 3 H]GlcNAc as donor substrate. Sep-Pak columns (Waters) were used to separate the substrates and products as described (23) . The column methanol eluant was added to scintillation fluid and radioactivity measured by scintillation counting. Assays were performed in duplicate, the data were averaged, and results were expressed as specific activity (nmol/h/mg). Protein concentration in the cell lysates was determined using the bicinchoninic acid assay (Pierce).

Lectin Blotting.

Cells were harvested, rinsed with PBS, and lysed with 1% Triton X-100 in PBS. Cell lysates containing 30 μg of protein were boiled in SDS sample buffer with or without β-mercaptoethanol, loaded on 4–20% SDS-PAGE gels, and then transferred onto a PVDF membrane. After being blocked with 1% BSA, the membrane was incubated with 1 μg/ml biotinylated l-PHA in blocking solution for 1 h followed by the incubation with a 1:5000 dilution of streptavidin-HRP for another 30 min at room temperature. The blots were washed and developed with the ECL detection system using X-ray film. Quantification was obtained using a Fluor-S scanner (Bio-Rad).

Assays of Cell-Fibronectin Adhesion and Spreading.

Cell attachment was assayed as described (24) . Briefly, 96-well microtiter plates were coated with 0.1 ml of human plasma fibronectin in PBS, incubated at 37°C for 1 h, and blocked by 1% BSA at 37°C for 30 min after washing. Cells (3 × 10⁴) suspended in 100 μl of serum-free DMEM were added to each coated well and incubated at 37°C for 30 min. Wells were gently washed three times with 100 μl of ice cold PBS to remove unbound cells, followed by fixation of adherent cells using 3.5% formaldehyde for 15 min. Cells were then stained with a 0.5% crystal violet solution. After the wells were washed twice with PBS, the absorbance of each well at 595 nm was measured using an automated microtiter plate spectrophotometer. For the antibody inhibition experiments (25) , function-blocking mAb13 (50 μg/ml) against β₁ integrin or mAb16 (2 mg/ml) against α₅ integrin was added to the cell suspension before addition of the cell suspension to the fibronectin-coated wells. The data were expressed as the mean of triplicate wells.

For assessment of cell spreading on fibronectin, the same procedure as above was used. After the adherent cells were fixed and stained with crystal violet, the number of spread cells was quantitated by counting the percentage of spread cells versus the total number of cells. Cells were considered as spread if they became flattened and lost nuclear refractivity. Each experimental point was the result of counting cells in 3–5 fields/well in at least 2 wells viewed at ×100 magnification. The values from 6–10 fields were used to calculate the mean and SE. Typically, 50–150 cells were present per field.

Wound Healing Assay.

Cells (3 × 10⁵) were seeded into fibronectin (10 μg/ml)-coated six-well plates in 2 ml of serum-free DMEM overnight. A clear area was then scraped in the monolayer with a 200-μl yellow plastic tip. After being washed with serum-free DMEM, the plate was incubated for 6 h at 37°C in serum-free DMEM. In some experiments, 1 μg/ml swainsonine was added to the cells for 24 h as GnT-V was induced to express. For inhibitory experiments, mAb13 (50 μg/ml), or mAb16 (2 mg/ml) was added into the well after wounding. Migration of cells into wounded areas was evaluated with an inverted microscope and photographed.

Transwell Cell Migration and Invasion Assays.

Migration assays were performed using 12-well Transwell units with 8-μm pore size polycarbonate inserts. Briefly, the undersides of insert membranes were coated with fibronectin (10 μg/ml) overnight at 4°C and blocked with 1% BSA for 1 h at 37°C. Cells (3 × 10⁴) suspended in 400 μl DMEM plus 0.1% BSA were added to the upper compartment of the Transwell unit, and 1.5 ml DMEM plus 0.1% BSA were added to the lower chamber. Cells were allowed to migrate for 6 h at 37°C in a humidified atmosphere containing 5% CO₂. The cells on the upper side of the membrane were removed using a cotton swab, whereas the cells that migrated to the underside were fixed and stained with crystal violet. The number of cells on the underside of the membrane was counted in five different fields with a light microscope at ×100, and the mean and SD were calculated. Typically, each field contained ∼300 cells. For some experiments, mAb13 (50 μg/ml) or mAb16 (2 mg/ml) was added into the cells before they were added to the Transwell apparatus. For swainsonine-treated cells, 1 μg/ml swainsonine was added to the culture medium for 24 h before cells were removed for addition to the upper chamber of the Transwell apparatus. Using Matrigel-coated 24-well Boyden chambers, invasion assays were performed using the same procedure as the migration assays, except that the incubation time of the experiment was prolonged to 21 h. Cells that migrated through the membrane were stained and counted as described above.

Gelatin Gel Zymography.

Cells were cultured to subconfluence in six-well plates coated with fibronectin (10 μg/ml). Then, cells were washed and incubated with 2 ml of serum-free medium for 24 h. The media were collected, centrifuged, and frozen until experiment. To detect gelatinase, 10 μl of conditioned medium were loaded to 10% SDS-PAGE gel containing 1% gelatin at 4°C under nonreducing conditions. After electrophoresis, the gels were rinsed in renaturing buffer (2.5% Triton X-100) for 1 h and incubated at 37°C overnight in buffer containing 50 mm Tris-HCl (pH 7.7), 5 mm CaCl₂, and 0.02% NaN_3. MMP-9 and MMP-2 were indicated by clear bands at 92 and 72 kDa that appeared after staining with Coomassie Brilliant Blue. Gels were then treated with resin and dried.

Cell Surface Labeling and Immunoprecipitation.

Cell labeling with biotin and α₅β₁ immunoprecipitation were performed as described by Rigot et al. (26) with minor modifications. Subconfluent cells were washed and detached using 2 mm EDTA. Cells were then washed twice with ice-cold PBS and incubated with 1 mg/ml NHS-LC-biotin in PBS for 20 min at 4°C on a rocking platform. After a washing with PBS, cells were lysed by incubation with buffer I (50 mm Tris-HCl, pH 7.5; 150 mm NaCl; 1% NP40; 0.5% sodium deoxycholate; Complete mini protein inhibitor cocktail; 0.7 μg/ml pepstatin; and 1% Triton X-100). Lysates were cleared by centrifugation and incubated with protein-l-agarose (50 μl of beads/ml of lysate) at 4°C under agitation for 3–5 h to remove nonspecific adsorption to the agarose beads. After the determination of total protein using a bicinchoninic acid assay, cell lysates were incubated with 5 μg/ml mAb13 or mAb16 overnight at 4°C under agitation, followed by incubation with 50 μl/ml protein-l-agarose at 4°C for 3–5 h under agitation. The pellets were washed twice in buffer I, followed by two washes in buffer II (50 mm Tris-HCl, pH 7.5; 500 mm NaCl; 0.1% NP40; 0.05% sodium deoxycholate), and once in buffer III (50 mm Tris-HCl, pH 7.5; 0.1% NP40; 0.05% sodium deoxycholate) and subjected to SDS-PAGE under reducing and nonreducing conditions. The gels were then transferred onto a polyvinylidine difluoride membrane and probed with streptavidin-HRP (1:5000).

Fluorescence Staining.

Cells were plated onto fibronectin-coated chamber slides (10 μg/ml) at 37°C in serum-free medium for either 30 or 120 min. Cells were then fixed with 3.5% formaldehyde in PBS for 15 min followed by three washes with PBS, permeabilized with 0.5% Triton X-100 (for actin staining only) for 15 min, and blocked with 1% BSA. For the localization of α₅β₁ integrins, chamber slides were incubated with 10 μg/ml mAb13 for β₁ or mAb11 for α₅ integrin at 37°C for 30 min, followed by incubation with Rhodamine-conjugated antirat IgG (1:250) at 37°C for another 30 min. For actin filament staining, chamber slides were incubated with Rhodamine-phalloidin (1:40). After a washing with PBS, the chamber slides were mounted and imaged using an immunofluorescence confocal microscope (Bio-Rad).

α₅β₁ Integrin Purification.

Purification of α₅β₁ integrin was performed according to the procedure of Pytela et al. (27) with minor modifications. Briefly, 2 × 10⁸ cells were harvested and lysed in PBS buffer containing mini protease inhibitor cocktail (1 tablet/10 ml), 0.7 μg/ml pepstatin, and 100 mm N-octylglucoside plus 0.5% SB3-10 (Sigma-Aldrich). Lysed cells were loaded onto an immunoaffinity column consisting of mAb16 (against α₅ integrin) coupled to agarose beads using the Ultralink immobilization kit. Briefly, 2 mg of mAb16 in 200 μl of PBS were mixed with 0.25 g of dry beads in 4 ml of 4-morpholinepropanesulfonic acid coupling buffer and allowed to swell and couple for 1 h. The reaction was terminated by incubating beads in 5 ml of 3 m ethanolamine for 3 h. Noncoupled protein was removed by washing the beads with 8 m urea in PBS, followed by rinsing with at least 10 ml of equilibration buffer (PBS containing 0.5% SB3-10). Beads were incubated on ice for 1 h before use. After cell lysate was applied to the 1-ml column of beads at 4°C, the column was washed with 10 column volumes of equilibration buffer and eluted with low pH elution buffer (0.1 m glycine adjusted to pH 2.9 with acetic acid) at 4°C. Fractions of 1 ml each were collected and subjected to SDS-PAGE under nonreducing conditions. The identity of the two distinct bands corresponding to 140 and 120 kDa under nonreducing conditions was confirmed to be the α₅ and β₁ subunits, respectively, by Western blotting with a combination of 20 μg/ml mAb11 and 20 μg/ml mAb13.

RESULTS

To develop an experimental system in which GnT-V activity could be overexpressed without the need for selection of single-cell clones, we developed two retroviral expression plasmids to be used to infect HT1080 human fibrosarcoma cells. A cDNA-encoding mouse GnT-V was inserted into two retroviral expression vectors with GFP expression to monitor the infection efficiency, one of which used tet-off inducible expression. In all experiments to be described, >95% of each population of cells expressed GFP fluorescence 24 h after infection with retroviral supernatants.

After infection of HT1080 cells with a plasmid with no GnT-V sequence and a noninducible promoter (mock transfected cells), cells showed a GnT-V specific activity of 0.18 ± 0.02 nmol/h/mg, whereas cells infected with a similar plasmid that encoded GnT-V (GnT-V-transfected cells) showed a specific activity of 10.7 ± 1.3 nmol/h/mg. As shown in Fig. 1A ⇓ ,cells infected with the plasmid containing GnT-V insert and the tet-off inducible promoter (GnT-V-transfected cells) showed an increase in GnT-V activity as early as 3 h after the removal of tetracycline and reached maximum induction (about 10–12-fold) 24 h after withdrawal of tetracycline from culture medium. These cells also showed a dose-dependent increase in GnT-V activity as the concentration of tetracycline in the culture medium was reduced (Fig. 1B) ⇓ . When maintained in tetracycline concentrations of >0.1 μg/ml, these cells showed ∼1.5-fold higher GnT-V activity than did mock transfected cells, showing that the promoter activity of the GnT-V plasmid could not be completely suppressed. In these cells, mRNA that encoded both GnT-V and GFP showed a tetracycline dose-dependent induction as the concentration of tetracycline in the culture medium decreased (data not shown).

The level of expression of cell surface N-linked oligosaccharides that contained β1,6 oligosaccharide branches increased after transfection of many cells types by several oncogenes, including, src, her-2/neu, ras, and sis (9, 10, 11) , caused by the up-regulation of GnT-V transcription. To determine whether the overexpression of GnT-V showed the expected increase in β1,6 oligosaccharide branching on the cell surface of infected HT1080 cells, the levels of β1,6 branched oligosaccharides present on the glycoproteins of GnT-V-transfected cells were compared before and after induction by using the l-PHA, which specifically binds to N-linked oligosaccharides containing Gal(β1,4)GlcNAcβ1,6Manα1,6. Cell lysates were subjected to SDS-PAGE, followed by lectin blotting and ECL detection of l-PHA binding. The results (Fig. 1C) ⇓ confirmed that there were substantial increases in the levels of β1,6 branching of glycoproteins after induction of GnT-V activity.

We first studied the growth properties of HT1080 cells overexpressing GnT-V. The growth rate of GnT-V-transfected cells cultured on plastic in serum-containing medium was similar to that of mock transfected cells, and no significant difference in the number of colonies was observed when both types of transfected cells were grown in soft agar (data not shown), showing that high levels of GnT-V activity did not influence cell proliferation. Then, we measured the effects of GnT-V-increased expression on cell spreading and adhesion to fibronectin. A population of HT1080 cells overexpressing GnT-V activity showed significantly reduced (>50%) spreading on fibronectin-coated plates after 30 min (Fig. 2A) ⇓ . Spreading on fibronectin was inhibited by the addition of either of two function-blocking mAbs, mAb13, specific for the human β₁ integrin subunit, and mAb16, specific for the α₅ subunit. These results demonstrate that the spreading of the cells on fibronectin resulted from adhesion through the α₅β₁ integrin receptors and that increased expression of GnT-V caused significantly reduced spreading. When HT1080 cells transfected with GnT-V under control of the tet-off promoter were induced to overexpress GnT-V (Fig. 2B) ⇓ , inhibition was also observed on spreading on fibronectin. In fact, some inhibition of spreading could be observed when noninduced cells with GnT-V insert were compared with mock transfected cells, suggesting that even relatively small increases in GnT-V activity, caused by the leaky tet-off promoter, could affect spreading on fibronectin.

The effects of GnT-V overexpression were next tested on simple cell attachment to fibronectin. The results of these experiments show that inducing the expression of GnT-V in HT1080 cells caused decreased cell attachment to fibronectin (Fig. 3) ⇓ . The attachment of noninduced cells was slightly but consistently inhibited compared with that of control vector-infected cells. Adhesion of the HT1080 cells was inhibited by mAb13 and mAb16, showing that the adhesion was α₅β₁ dependent (data not shown). These results show that the adhesion of HT1080 cells to fibronectin was significantly inhibited by increased GnT-V expression.

Experiments were then designed to determine whether GnT-V overexpression could affect cell migration, assayed by both wound healing on fibronectin substrates and haptotaxis toward fibronectin. The results in Fig. 4 ⇓ show that induction of GnT-V activity caused a striking increase of cell migration and that this effect was strongly inhibited by antibodies against the integrin β₁ subunit and to a lesser extent by antibodies against the integrin α₅ subunit. Swainsonine inhibits the Golgi α-mannosidase II, ultimately causing the inhibition of N-linked β1,6 oligosaccharide expression upstream of the action of GnT-V. Cells preincubated in swainsonine before wounding but after induction of GnT-V showed little or no increase in migration.

In the Boyden chamber (Transwell) migration assays, HT1080 cells induced to overexpress GnT-V showed a ∼2-fold increase in rate of migration compared with mock transfected cells (Fig. 5A) ⇓ . These increased rates were inhibited by antibodies to α₅ and β₁ and by preincubation with swainsonine. Taken together, these results are consistent with the conclusion that increased β1,6 oligosaccharide branching caused altered adhesion to fibronectin mediated by the α₅β₁ integrin that results in increased rates of migration.

We next examined whether GnT-V-increased expression could alter in vitro invasive behavior through a reconstituted Matrigel basement membrane. Fig. 5B ⇓ shows that HT1080 cells with induced GnT-V activity showed almost a 3-fold increase in in vitro invasiveness after 21 h of incubation and that this effect was inhibited by antibodies to α₅ and β₁ or by pretreatment of cells with swainsonine. Noninduced cells, which showed a 1.5-fold increase in GnT-V activity, migrated only slightly faster than control cells, suggesting that the ability of cells to penetrate and to migrate through Matrigel was very sensitive to relatively small increases in GnT-V activity.

We also detected the secretion of MMP-9 and MMP-2 in the culture medium of HT1080 cells grown on the fibronectin-coated plates, but no significant difference in the MMP-9 or MMP-2 levels was found after induction of GnT-V overexpression (Fig. 5C) ⇓ . Taken together, these results show that the increased ability of HT1080 cells overexpressing GnT-V to migrate through Matrigel resulted from increased expression of β1,6 branching.

To investigate the mechanisms underlying the effects of GnT-V expression on cell-adhesive behavior, F-actin filaments were first visualized in GnT-V-overexpressing cells. At early times (30 min) after plating cells on fibronectin-coated wells, phalloidin staining of permeabilized cells revealed increased F-actin staining in membrane protrusions in cells overexpressing GnT-V, suggesting increased motility of these cells (Fig. 6) ⇓ . After longer times on fibronectin (120 min), the network of F-actin stress fibers appeared to be less organized in the GnT-V-induced cells, as compared with mock transfected cells, suggesting that increased β1,6 branching had a destabilizing effect on F-actin filaments.

One possible explanation of the effects of GnT-V levels on fibronectin-mediated cell adhesion and migration could be that changes in β1,6 oligosaccharide branching affect levels of α₅β₁ integrin expression on the cell surface. To test this possibility, cell surface α₅β₁ integrin receptors were quantitated by biotinylation, immunoprecipitation, and blotting using streptavidin-peroxidase. As shown in Fig. 7A ⇓ ,levels of α₅ or β₁ on HT1080 cells were similar after induction of GnT-V expression. In addition, the total mRNA levels of β₁ integrin determined by RT-PCR were also unchanged in GnT-V-expressing cells compared with control cells (data not shown). These results suggest that the effects observed on adhesion/migration after induction were not the result of altered levels of α₅β₁ on the cell surface.

Fluorescent staining patterns of cell surface α₅ and β₁ using mAbs, however, showed clear differences after induction of GnT-V activity (Fig. 7B) ⇓ . Cells with induced GnT-V grown on fibronectin plates for 2 h demonstrated a more diffuse surface staining of both α₅ and β₁ subunits compared with mock transfected cells, suggesting decreased α₅β₁ integrin clustering. This observation is consistent with a less adhesive and more motile cellular phenotype (28) .

To determine whether levels of β1,6 branching on the α₅β₁ subunits were specifically altered after induction of GnT-V, the α₅β₁ receptor was purified from HT1080 cells before and after induction of GnT-V using an antibody affinity column, followed by chromatography on immobilized wheat germ agglutinin. Equal amounts of the purified receptor were subjected to SDS-PAGE under nonreducing conditions, blotted, and probed with various antibodies and lectins. The results of these experiments (Fig. 7C) ⇓ showed that the subunits could be distinguished and that each bound the lectin ConA. Surprisingly, however, only the β₁ subunit and not the α₅ subunit was bound by l-PHA, demonstrating that β1,6 branching could be detected only on the β₁ subunit. When GnT-V activity was induced, significant increases in l-PHA staining (>3-fold) were observed on the β₁ subunit. Even after induction, however, no l-PHA binding to the α₅ subunit was observed, suggesting a selective glycosylation of the β₁ subunit by GnT-V. These results demonstrate that induction of GnT-V activity caused increased β1,6 branching on the β₁ subunit and that the effects of increased GnT-V expression on fibronectin-mediated adhesive phenotypes likely result from this altered glycosylation.

DISCUSSION

To test directly the hypothesis that changes in GnT-V activity and β1,6 branching can cause altered cell-matrix adhesion and migration and allow study of the underlying mechanisms, we transfected mouse GnT-V cDNA into human HT1080 cells using retroviral vectors, one of which contained a tet-off promoter. Compared with standard plasmid transfection used in previous studies (8 , 17, 18, 19) , retroviral infection is a highly efficient gene transfer method applicable to many cell types (29) , that yields high expression efficiency and the ability to modulate expression (29 , 30) . In the present study, both retroviral and tet-off retroviral systems were used to express GnT-V, at efficiencies of >95%, as judged by GFP fluorescence of infected cells. Because of the high efficiency, we were able to examine total populations of transfected cells, ensuring that any observed alterations in cell behaviors could be attributed to GnT-V expression and not to phenotypic variation of selected clones. Stable populations of HT1080 cells were produced with expressions of GnT-V that could be regulated by changing levels of tetracycline in the growth medium. In both infected cell types, increased expression of β1,6 branching followed increased activity of GnT-V.

We first investigated the effect of GnT-V expression on cell proliferation and found that GnT-V levels had minimal effects on the rate of cell growth on plastic or anchorage-independent growth in soft agar. In addition, no effect of GnT-V overexpression was observed on the apoptotic behavior of cell populations (data not shown), in contrast to the results of an earlier report (17) . Altering GnT-V expression did, however, cause significant changes in cell-matrix interaction. We found that both cell adhesion to and cell spreading on fibronectin was markedly decreased after the induction of GnT-V expression in HT1080 cells, and these events were most likely mediated by α₅β₁ integrin confirmed by antibodies that blocked α₅β₁ adhesion. After exogenous expression of GnT-V, reduced cell-matrix adhesion was observed in mink lung epithelial cells (17) and hepatocarcinoma cells (31) , but not in mouse mammary cancer cells (19) . This difference in cell adhesion might result from the variation between individual clones used in the experiment. Our results together with previous studies suggest strongly, therefore, that changes in N-linked glycosylation caused by increased β1,6 branching had direct effects on α₅β₁-mediated cell spreading and adhesion.

We used different types of assays to test whether GnT-V overexpression could affect rates of migration. We observed with both the wound healing assay and the haptotaxis assay that cell motility was significantly increased after induced expression of GnT-V. The increased migration of HT1080 cells induced by GnT-V expression could be suppressed by adding function-blocking antibodies against α₅ or β₁, indicating that this receptor was involved in regulating this migratory behavior. The increased migration of HT1080 cells overexpressing GnT-V was inhibited by swainsonine treatment, consistent with the conclusion that β1,6 oligosaccharide branching may play a direct role in α₅β₁ integrin-mediated cell motility. Cell migration is a complex behavior regulated by multiple mechanisms (32) . A critical factor that regulates the rate of cell migration is the degree of adhesion of cells to their ECM substrata. Migration is maximal under conditions of intermediate levels of cell adhesion (33) . Thus, the increased rate of migration on fibronectin is consistent with decreased cell adhesion and spreading observed for GnT-V overexpressing cells.

Increased β1,6 branching also promoted the ability of cells to invade through Matrigel. Antibodies against both the β₁ and α₅ integrin subunits and swainsonine treatment blocked increased HT1080 invasion caused by the induced expression of GnT-V, indicating the involvement of α₅β₁ integrin and β1,6 branching in the invasion process. HT1080 cells produce MMPs, including both MMP-2 and MMP-9, which can degrade the components of basement membranes and ECM and increase cell invasion (34) . To determine whether GnT-V overexpression enhanced the secretion of MMP-2 and MMP-9 in HT1080 cells, levels of MMP-2 and MMP-9 in culture medium were analyzed by zymography using gelatin as substrate. No significant differences were observed in levels of either MMP-2 or MMP-9 after induction expression of GnT-V in HT1080 cells. This result supports the conclusion that increased cell invasion because of increased β1,6 branching is the result of increased cell migration rather than an increased secretion of MMPs.

Focal adhesions are formed when cells contact the ECM, linking the cytoplasmic tails of integrins to the actin cytoskeleton, and play a pivotal role in cell spreading and migration (35 , 36) . In our study, we found increased F-actin staining in membrane protrusion at early times and decreased stress fiber formation in the cells with overexpressed GnT-V. These observations are consistent with a less adhesive, more motile phenotype, as suggested by Duband et al. (28) . It was shown that after cell spreading on fibronectin, activation of FAK, and transient inhibition of a small GTPase, RhoA, proved to be essential for promoting membrane protrusion and polarity (37 , 38) ; Moreover, FAK-dependent transient inhibition of RhoA is required to promote focal adhesion turnover and actin filament reorganization (39 , 40) . It is reasonable, therefore, to postulate that the signaling pathway mediated by FAK and/or RhoA activity might be affected in GnT-V overexpressing cells.

Integrin receptors regulate cell adhesion, spreading, migration, and proliferation, as well as gene transcription (25 , 32 , 41 , 42) . The α₅β₁ integrin, a primary receptor of fibronectin, contains N-linked glycans. Previous studies demonstrated that α₅β₁ function can be modulated by N-linked oligosaccharide expression (20 , 43) . To investigate further glycosylation differences after GnT-V induction, the α₅β₁ integrin was purified from induced and noninduced cells and subjected to lectin blotting with l-PHA. Increased β1,6 branching was observed on the β₁ subunit after induction of GnT-V, but no β1,6 branching could be detected on the α subunit either before or after induction. ConA bound to both α₅ and β₁ subunits, by contrast, and little change in binding could be observed before and after the induction of GnT-V expression. More information on the N-linked oligosaccharide expression on the integrin subunits will await further detailed structural characterization. The effect of increased β1,6 oligosaccharide branching on α₅β₁ function, however, appears to result solely from changes in β₁ glycosylation. Kobata’s laboratory suggested that changes in β1,6 branching of the β₁ subunit of transformed 3T3 cell lines correlated well with increased tumorigenicity and metastatic potential (44) . More recently, Skinner et al. (45) reported that overexpression of the 16-kDa membrane subunit of vacuolar ATPase could inhibit β1,6 branching of β₁ integrin subunit, resulting in decreased tumor cell invasiveness. These studies are consistent with our direct demonstration of increased β1,6 branching on the β₁ subunit after induction of GnT-V expression.

In our study, the levels of both α₅ and β₁ subunits on HT1080 cell surfaces remained relatively unchanged when GnT-V was induced (Fig. 7) ⇓ . We did observe changes in cell surface α₅β₁ integrin clustering on fibronectin, however, after induction of GnT-V overexpression, using antibodies to visualize separately the α₅ and β₁ subunits. Formation of focal adhesions is a critical event as cells attach to matrix components, initiated by the binding of integrins to specific ECM ligands and subsequent clustering of these receptors (46) . We found more diffuse staining of α₅β₁ integrins on cells spreading on fibronectin after induced GnT-V expression, indicating that the clustering of α₅β₁ integrins was reduced in GnT-V overexpressing cells. This conclusion was further supported by our recent experiments with 12G10, an activating anti-β₁integrin mAb which was reported to induce cell-cell adhesion and β₁ integrin clustering in HT1080 cells (47) . We found that overexpression of GnT-V in HT1080 cells could significantly inhibit β₁ integrin clustering induced by 12G10 (unpublished data), suggesting that N-linked glycosylation of the β₁ integrin may be a critical factor for regulating integrin clustering. In another study (8) , human glioma cells transfected with GnT-V resulted in different localization pattern of α₃β₁ integrin, where the immunostaining of α₃β₁ integrin showed a strong localization of the integrin heterodimer to leading lamellipodia of GnT-V-transfected cells, but not of parental or mock transfected cells. It is therefore clear that glycosylation of integrins by GnT-V causes functional alteration in β₁-containing integrin heterodimers.

The function of integrin clustering in the regulation of cell adhesion strength has been studied by several groups. Yauch et al. (48) reported that the deletion of the α₄ integrin tail region significantly impaired static cell adhesion by severely restricting β₁ integrin clustering. Without affecting ligand-binding affinity, integrin-mediated adhesion appeared to be regulated by the rate of receptor diffusion/clustering. Another study showed that leukocyte function-associated antigen-1 or integrin α_Lβ₂ ligand binding strength to intercellular adhesion molecule-1 was regulated by lateral diffusion of the integrins and their interaction with the cytoskeleton (49) . Kassner et al. (50) showed that an α₄ integrin chimera expressing α₂ or α₅ tails greatly enhanced localization of the chimeric very late antigen-4 into focal adhesions and concluded that relatively more mobile α₄ integrin, less engaged with cytoskeletal interaction, is likely to augment cell migration and diminish cell spreading and adhesion strengthening. Consistent with these reports, our data suggest that altered integrin clustering is one of the major consequences of increased β1,6 branching of β₁ integrin and results in a less adhesive and more motile phenotype through regulating cell spreading, adhesion, as well as actin-filament reorganization.

There are now clear examples of the inhibition of signal transduction pathways by mutations in specific N-acetylglucosaminyltransferases. Altered glycosylation of the Notch cell surface receptor during Drosophila development can clearly affect the signaling function of this receptor (51) and is caused by mutations in the Fringe N-acetylglucosaminyltransferase in the O-linked pathway. In addition, a recent study has documented that inactivating mutations in a human N-acetylglucosaminyltransferase, which also functions in the biosynthesis of specific O-linked glycans, causes an inhibition of a particular cell surface signaling pathway, which, in turn, results in a form of muscular dystrophy known as muscle-eye-brain disease (52) . Glycosylation of the CD8αβ coreceptor stalks on native thymocytes has recently been shown to regulate interactions between the CD8 “head” domain and major histocompatibility complex class I (MHCI) tetramers, thereby modulating thymic selections (53) . In light of these results, it is reasonable to suggest that increased β1,6 branching of the β₁ integrin subunit by GnT-V that is up-regulated after oncogenesis results in increased cell migration and invasion by modulating integrin clustering and subsequent signal transduction pathways.