Activated Epidermal Growth Factor Receptor Induces Integrin α2 Internalization via Caveolae/Raft-dependent Endocytic Pathway*

Elevated expression or activity of the epidermal growth factor (EGF) receptor is common in ovarian cancer and is associated with poor patient prognosis. Our previous studies demonstrated that expression of the constitutively active mutant form of the EGF receptor (EGFRvIII) in ovarian cancer cells led to reduction in integrin α2 surface expression, defects in cell spreading, and disruption of focal adhesions. Inhibition of EGFRvIII catalytic activity reversed the response, suggesting that EGF receptor activation regulates integrin α2. In this study we found that EGF treatment resulted in a transient loss of integrin α2 from the cell surface. Before EGF stimulation, integrin α2 and EGF receptors were associated based on biochemical and immuno-colocalization approaches. After EGF treatment, EGF receptor and integrin α2 were internalized and segregated into different compartments. Integrin α2, but not EGF receptor, was associated with caveolin-1 and GM1 (Gal_1,3GalNAc_1,4(Neu5Ac-_ 2,3)Gal_1,4Glc_1,1-ceramide) gangliosides, suggesting caveolae-mediated endocytosis. Moreover, integrin α2 was subsequently targeted to the Golgi apparatus and the endoplasmic reticulum. Together, these findings demonstrate that activated EGF receptor transiently modulates integrin α2 cell surface expression and stimulates integrin α2 trafficking via caveolae/raft-mediated endocytosis, representing a novel mechanism by which the EGF receptor may regulate integrin-mediated cell behavior.

Elevated expression or activity of the epidermal growth factor (EGF) receptor is common in ovarian cancer and is associated with poor patient prognosis. Our previous studies demonstrated that expression of the constitutively active mutant form of the EGF receptor (EGFRvIII) in ovarian cancer cells led to reduction in integrin ␣2 surface expression, defects in cell spreading, and disruption of focal adhesions. Inhibition of EGFRvIII catalytic activity reversed the response, suggesting that EGF receptor activation regulates integrin ␣2. In this study we found that EGF treatment resulted in a transient loss of integrin ␣2 from the cell surface. Before EGF stimulation, integrin ␣2 and EGF receptors were associated based on biochemical and immuno-colocalization approaches. After EGF treatment, EGF receptor and integrin ␣2 were internalized and segregated into different compartments. Integrin ␣2, but not EGF receptor, was associated with caveolin-1 and GM1 (Gal_1,3GalNAc_1,4(Neu5Ac-_ 2,3)Gal_1,4Glc_1,1-ceramide) gangliosides, suggesting caveolae-mediated endocytosis. Moreover, integrin ␣2 was subsequently targeted to the Golgi apparatus and the endoplasmic reticulum. Together, these findings demonstrate that activated EGF receptor transiently modulates integrin ␣2 cell surface expression and stimulates integrin ␣2 trafficking via caveolae/ raft-mediated endocytosis, representing a novel mechanism by which the EGF receptor may regulate integrin-mediated cell behavior.
Epithelial ovarian carcinoma accounts for 80 -90% of ovarian tumors and is the leading cause of death from gynecologic malignancy, resulting in 16,210 deaths in 2005 (1). Because of the current inability to detect disease confined to the ovary (stage I), ϳ75% of women are initially diagnosed with disseminated intra-abdominal disease (stage III-IV) and have a 5-year survival of Ͻ20%, whereas patients diagnosed with cancer localized to the ovary have a Ͼ90% 5-year survival. Clinically, tumors often involve the ovary and omentum, with diffuse intraperitoneal metastases and malignant ascites. Ovarian cancer metastasis results from numerous intraperitoneal adhesive events, suggesting that carcinoma cell integrins regulate subsequent invasive or metastatic behavior (2)(3)(4).
Integrins are the major family of cell surface receptors that mediate attachment to the extracellular matrix, and these integrin-mediated adhesive interactions are intimately involved in the regulation of many cellular functions, including tumor cell growth, apoptosis, and metastasis (5)(6)(7). After disseminated primary ovarian tumor cells attach to the peritoneal mesothelial monolayer via CD44 (8 -10), integrin-mediated cell-matrix interaction potentiates intraperitoneal invasion. Ovarian carcinoma cells extend cytoplasmic processes through the junctional margins of neighboring mesothelial cells, inducing cellular retraction and exposure of the submesothelial extracellular matrix, followed by integrin-mediated adhesion to the newly exposed matrix (11,12). Analysis of adhesive preferences and integrin expression profiles of established and primary cultures of ovarian carcinoma cells demonstrates high level expression of ␣2, ␣3, and ␤1 subunits and preferential adhesion to interstitial collagen types I and III (13)(14)(15)(16) as well as laminins (4) mediated by the ␣2␤1 and ␣3␤1 integrins.
Numerous studies have demonstrated cooperation between integrin and epidermal growth factor (EGF) 2 -mediated signaling pathways in the control of mitogenic, motogenic, and cell survival pathways (17)(18)(19)(20). The EGF/ErbB family of receptor tyrosine kinases has been shown to play a key role in normal ovarian follicle development and cell growth regulation of the ovarian surface epithelium (21). Moreover, EGF receptor activation can modulate integrin function by regulating the expression and/or activity of numerous integrins, leading to altered adhesion, motility, and invasive capacity (22)(23)(24). Integrin ␣2 expression is selectively modulated by EGF receptor activation but not ␤1 integrin in several cell types (22,25,26), and the ␣2 integrin cytoplasmic domain is required for EGF-stimulated migration in NMuMG-3 cells (27,28). Furthermore, co-localization and direct interaction between integrin ␣2␤1 and the EGF receptor at sites of cell:cell contact has been reported in human epithelial A431 cells (18). Collectively, these observations suggest that aberrant regulation of EGF receptor activity may alter integrin ␣2 expression and/or function.
Our previous studies demonstrated that expression of a constitutively active mutant form of the EGFR (EGFRvIII) in ovarian cancer cells led to reduction in integrin ␣2 surface expression, defects in cell spreading, and disruption of focal adhesions. These responses were reversed upon inhibition of EGFRvIII catalytic activity (29). In the course of those studies we observed that integrin ␣2 and the wild type (wt) EGF receptor were co-localized when EGFRvIII activity was inhibited; therefore we decided to investigate potential regulation of integrin ␣2 by ligand-activated EGF receptor in ovarian tumor cells.
Here we report that EGF induces transient internalization of integrin ␣2, but not integrin ␤1. In the absence of ligand, integrin ␣2 interacts with the EGF receptor based on co-localization, co-immunoprecipitation, and chemical cross-linking studies. Following EGF stimulation, integrin ␣2 and EGF receptor interaction was disrupted and they were internalized by distinct pathways. EGF receptor activation promoted integrin ␣2 internalization via a caveolae/lipid raft-mediated, rather than the clathrin-dependent, endocytic pathway. Furthermore, internalized integrin ␣2 was localized to Golgi and endoplasmic reticulum (ER). Based on these findings, we propose that activated EGF receptor down-regulates surface integrin ␣2 by a caveolae/raft-mediated endocytic pathway and presents a novel mechanism for EGF-dependent regulation of integrins.

MATERIALS AND METHODS
Cell Culture and Treatment-Ovarian carcinoma cell line OVCA 433 was generously provided by Dr. Robert Bast Jr., MD Anderson Cancer Center, Houston, TX, and grown as described previously (29). For experiments involving EGF (Biomedical Technologies, Stoughton, MA), OVCA 433 cell lines were placed into minimal essential medium containing 0.1% (w/v) bovine serum albumin for 24 h prior to growth factor addition as described previously (29).
Chemical Attachment of Fluorescent Tags to GM1 Gangliosides-Alexa Fluor 488 hydrazide-conjugated gangliosides were prepared via a modified adaptation of reaction methods previously described (30,31). Bovine brain gangliosides (1 mg/ml) were suspended in 100 mM sodium acetate buffer, pH 5.5, with 1 mM sodium meta-periodate. The oxidation reaction was allowed to proceed for 30 min on ice. The vesicle suspension was then purified and concentrated by ultrafiltration in the same buffer using YM-30 Microcon centrifugal filter devices (Millipore Corp., Bedford, MA), repeating five times to remove the sodium meta-periodate. 10 mM Alexa Fluor 488 hydrazide was added to the oxidized GM1 and allowed to react with agitation for 2 h at room temperature. The fluorescent GM1 conjugates were freed of unreacted dye by repeated pelleting and resuspension in PBS buffer as described above until the supernatant was optically clear. The labeled GM1 was dried under vacuum and stored as a powder under nitrogen at Ϫ20°C.
Immunofluorescence and Confocal Microscopy-OVCA 433 cells were treated with 25 nM EGF and fixed with freshly prepared 3.7% (w/v) formaldehyde in PBS (137 mmol/liter NaCl, 2.7 mmol/liter KCl, 8.1 mmol/liter Na 2 HPO 4 , and 1.5 mmol/ liter KH 2 PO 4 , pH 7.4) containing 0.8 mmol/liter MgCl 2 and 0.18 mmol/liter CaCl 2 for 10 min at room temperature, permeabilized with 0.1% Triton X-100 in PBS for 5 min at room temperature, and blocked with 3% bovine serum albumin/PBS for 1 h at 37°C. For dual staining, fixed cells were incubated with mouse anti-integrin ␣2, or rabbit anti-integrin ␣2 and sheep anti-EGFR, or mouse anti-EGFR and chicken anti-Rab 7, or mouse anti-TfR overnight at 4°C. After washing three times with PBS, samples were incubated with fluorescein isothiocyanate-conjugated anti-mouse IgG, anti-sheep IgG or anti-rabbit IgG, and the Cy3-conjugated anti-mouse IgG, anti-sheep IgG or anti-chicken IgG. For triple staining, fixed cells were incubated with mouse anti-integrin ␣2, sheep anti-wtEGFR, and rabbit anti-caveolin-1 or goat anti-clathrin and then incubated with fluorescein isothiocyanate or Cy3-conjugated anti-mouse IgG or fluorescein isothiocyanate or Cy3-conjugated antisheep IgG and Cy5-conjugated anti-rabbit or goat IgG antibodies. For GM1 staining, fixed cells were incubated with mouse anti-integrin ␣2 or mouse anti-EGFR, and then samples were incubated with Cy3-conjugated anti-mouse IgG. After washing three times with PBS, 150 nM Alexa 488-GM1 was added for 30 min at 37°C. For Golgi and ER marker staining, OVCA 433 cells were fixed with 70% methanol/30% acetone (Ϫ20°C) for 10 min at room temperature and incubated with rabbit anti-58K or anti-GRP94 and mouse anti-integrin ␣2 overnight at 4°C; fluorescein isothiocyanate-conjugated anti-mouse and Cy3-conjugated anti-rabbit IgG were added to sample. Confocal images were acquired at room temperature using a Zeiss LSM510 system equipped with argon and HeNe lasers for excitation at 488 nm (green), 543 nm (red), and 633 nm (blue). Samples were viewed with the 63 ϫ 1.4 oil immersion objective lens.
Cross-linking Assay-Cells were treated as described in the figure legends, washed three times with PBS, and incubated with 2 mM 3,3Ј dithiobis (sulfosuccinimidylpropionate) (Pierce) in PBS at 23°C for 30 min. The cross-linking reaction was quenched with buffer containing 10 mM Tris-HCl, pH 7.5, 0.9% (w/vol) NaCl, and 0.1 M glycine. Cells were then washed three times with PBS and collected for further analysis.
Immunoprecipitation and Immunoblot Analysis-Cells were lysed in ice-cold lysis buffer (1% Nonidet P-40, 150 mM NaCl, 50 mM Tris, pH 7.2, 5 g/ml leupeptin, 5 g/ml aprotinin, 1 mM NaVO 4 , and 1 mM phenylmethylsulfonyl fluoride) for 30 min at 4°C and clarified at 13,000 rpm at 4°C for 10 min. The supernatants of cell lysates were precleared by incubation with 100 l of a 50/50 protein A agarose bead suspension overnight at 4°C. After preclearing, sample supernatant was collected following centrifugation and incubated overnight at 4°C with 20 l of protein A beads and 5 l/500 g sample of either rabbit anti-wtEGFR, rabbit anti caveolin-1, or rabbit anti-integrin ␣2 overnight at 4°C. Nonspecific rabbit IgG was used as a control. The beads were washed once with lysis buffer and four times with buffer without detergent and then suspended in 50 l of 2ϫ loading buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 25% glycerol, and 0.1% Bromphenol blue). For immunoblot analysis, total cellular proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes, blocked with 5% nonfat milk in Tris-buffered saline (TBS) (10 mmol/liter Tris, 150 mmol/liter NaCl, pH 8.0) and 0.05% Tween 20 for 1 h at room temperature, probed with rabbit anti-EGFR, rabbit anti-integrin ␣2, rabbit anti-caveolin-1, or ␤-tubulin, washed three times with TBS and 0.05% Tween 20, and incubated with horseradish peroxidase-conjugated anti-rabbit IgG secondary antibody.
Flow Cytometry-Cells were washed twice with Dulbecco's phosphate-buffered saline (DPBS) and harvested with trypsin-EDTA. 10 6 cells were blocked with 3% bovine serum albumin/DPBS for 10 min at 4°C, incubated with primary antibody mouse PE-conjugated anti-integrin ␣2 or mouse PE-conjugated anti-wtEGFR for 50 min at 4°C, and washed twice with DPBS. Additional control samples included cells without antibody, mouse antiintegrin ␤1 and primary antibodyblocked control. Cells were incubated with non-PE-conjugated primary antibody, either mouse anti-integrin ␣2, anti-integrin ␤1, or mouse anti-wtEGFR for 30 min at 4°C, washed and then mouse PE-conjugated anti-integrin ␣2, ␤1, or mouse PE-conjugated anti-wtEGFR were added and samples incubated for 50 min at 4°C. Cells were pelleted and then resuspended in 0.5 ml of DPBS. Flow cytometric analysis was performed on a BD Biosciences FACScan flow cytometer (Immunocytometry Systems, San Jose, CA). Mean fluorescence intensity for three independent experiments is shown, and error bars represent Ϯ S.D.

Activated EGF Receptor Initiates Transient Internalization of
Integrin ␣2-We reported previously that integrin ␣2 was dynamically modulated by activity of the constitutively active EGF receptor mutant EGFRvIII in an ovarian tumor cell line, OVCA 433 (29). To establish whether ligand-stimulated EGF receptor also modulates integrin ␣2, we examined integrin ␣2 localization and protein expression after EGF treatment. As shown in Fig. 1A, integrin ␣2 displayed cell surface localization before EGF treatment, and after addition of ligand, integrin ␣2 surface staining intensity became more diffuse and of diminished intensity. The cell surface localization of integrin ␣2 was largely restored within 4 h of EGF treatment (Fig. 1A). This result is supported by flow cytometry to detect cell surface integrin ␣2 and integrin ␤1 levels (Fig. 1B) and immunoblot analysis of whole cell lysates (Fig. 1C). Cell surface integrin ␣2 expression was decreased by 32% within 30 min and returned to control levels by 24 h (Fig. 1B). This response differed from that of the EGF receptor, which was persistently down-regulated in the presence of ligand (Fig. 1B). A similar decrease and recovery of total integrin ␣2 protein levels was detected by immunoblot analysis (Fig. 1C). The recovery of cell surface integrin ␣2 did not require new protein synthesis. There were no significant differences in integrin ␣2 surface levels in cells treated with or without cycloheximide, and integrin ␣2 surface expression in cycloheximide-treated cells was 90% of untreated control after 6 h of EGF treatment. In contrast to integrin ␣2, cell surface integrin ␤1 expression was not significantly altered as a consequence of EGF receptor activation by mutation (29) or ligand (Fig. 1B).
Association of Integrin ␣2 and EGF Receptor in Unstimulated Cells-Physical association between integrin ␣2␤1 and the EGF receptor at intercellular adhesion sites has been reported in A431 cells (18). Confocal microscopy of OVCA 433 cells immunostained for EGF receptor and integrin ␣2 reveals co-localization prior to EGF treatment ( Fig. 2A). After EGF treatment, both EGF receptor and integrin ␣2 were rapidly internalized, but no longer co-localized ( Fig. 2A). Co-immunoprecipitation and chemical cross-linking approaches were used to determine whether the apparent co-localization represented biochemical interaction between the EGF receptor and integrin ␣2. In OVCA 433 cell lysates, anti-integrin ␣2 antibodies immunoprecipitated the EGF receptor, and conversely anti-EGF receptor antibodies immunoprecipitated integrin ␣2 (Fig. 2B) before EGF treatment. This interaction was greatly diminished within 10 min of EGF receptor activation (Fig. 2B), which was consistent with results obtained by confocal microscopy (Fig. 2A). These findings support the conclusion that the EGF receptor and integrin ␣2 were associated before EGF treatment.
EGF-stimulated Internalization of EGF Receptor and Integrin ␣2 through Distinct Pathways-Several mechanisms of integrin endocytosis have been reported (32)(33)(34). The trafficking of integrin ␣V␤6 and ␤1 has been reported to take place through an endosomal-mediated endocytic recycling pathway, but integrin ␣L␤2 was internalized and rapidly recycled upon chemoattractant stimulation via a clathrin-independent, cholesterolsensitive pathway. Because integrin ␣2 is initially associated with the EGF receptor (Fig. 2), which is endocytosed primarily via clathrin-dependent mechanisms (35), and ␣v␤6, ␤1 integrins have been reported to be internalized via clathrindependent mechanisms (33,34), we investigated whether integrin ␣2 is internalized by the clathrin endocytic pathway. Several endocytic markers (clathrin, Rab7, TfR) of clathrin-dependent internalization were used to examine integrin ␣2 trafficking in response to EGF. Untreated and treated OVCA 433 cells were triple-labeled using antibodies recognizing the EGF receptor, integrin ␣2, and clathrin. As shown in Fig. 3A, EGF receptor co-localized with clathrin; however, there was no detectable co-localization between clathrin and integrin ␣2. Rab7 protein belongs to a superfamily of small molecular weight GTPases associated with late endosomes. Rab7 regulates the later stages of the endocytic pathway for a number of proteins, including the TfR, a marker of recycling endosomes that undergoes multiple rounds of clathrin-mediated endocytosis and reemergence at the cell surface (36,37). Dual staining confocal microscopy was used to detect localization of EGF receptor or integrin ␣2 with Rab 7 (Fig. 3B) or TfR (Fig. 3C). As expected, there was punctate staining of EGF receptor with Rab 7 (Fig. 3B, arrows) or TfR, but no co-localization of Rab 7 (Fig.  3B) or TfR (Fig. 3C) with integrin ␣2 was detected. These findings suggest that EGF-stimulated internalization of integrin ␣2 does not occur through a clathrin-dependent pathway and prompted investigations of alternative mechanisms.
Recent studies have reported that protein kinase C-dependent integrin ␣2 internalization and C8-LacCer-stimulated ␤1 integrin internalization occur via caveolar endocytosis (38,39). Therefore, we investigated the possibility of a caveolae-dependent mechanism for EGF-mediated integrin ␣2 internalization. Caveolin-1 is a hallmark protein for caveolae and caveosomes, so interactions between caveolin-1 and integrin ␣2 were investigated using confocal microscopy and immunoprecipitation approaches. In the absence of EGF, caveolin-1 co-localized with integrin ␣2 and EGF receptor (Fig. 4A, upper panels). Thirty minutes after EGF addition, caveolin-1 co-localized with integrin ␣2, but not the EGF receptor (Fig. 4B, lower panels). We confirmed FIGURE 2. Interaction between integrin ␣2 and EGF receptor. A, co-localization of integrin ␣2 with EGF receptor. OVCA 433 cells were treated with EGF (25 nM) for 30 min or 2 h, fixed in 3.7% formaldehyde for 10 min, permeabilized with 0.1% Triton X-100 for 5 min, labeled with sheep anti-EGF receptor or mouse anti-integrin ␣2 antibody, and imaged by confocal immunofluorescence microscopy. The merged images present EGF receptor in green and integrin ␣2 in red. Co-localization of EGF receptor and integrin ␣2 is shown by yellow before EGF treatment, and after EGF exposure integrin ␣2 and EGF receptor stain separately. Scale bars, 5 m. B, co-immunoprecipitation of integrin ␣2 with EGF receptor. OVCA433 cells were treated with EGF (25 nM) for 10 or 30 min as indicated in the Fig. 1. After treatment, cells were lysed with Nonidet P-40 buffer and the lysate was immunoprecipitated with anti-EGFR (EGFR) or anti-integrin ␣2. Immunoprecipitates were run on 7.5% SDS-PAGE and immunoblotted with EGF receptor polyclonal antibody (top, left panel) or anti-integrin ␣2 polyclonal antibody (bottom, left panel). After treatment, OVCA 433 cells were incubated with 3,3Ј dithiobis (sulfosuccinimidylpropionate) (DTSSP) for 30 min at room temperature and quenched with buffer containing glycine. Cells were lysed and immunoprecipitations conducted as described above (right panel).
the apparent co-localization of integrin ␣2 with caveolin-1 through co-immunoprecipitation and chemical cross-linking techniques. Before EGF treatment, caveolin-1 was associated with the EGF receptor and integrin ␣2 (Fig. 4B). After 10 min of EGF stimulation, caveolin-1 and EGF receptor were no longer associated but interactions between caveo-lin-1 and integrin ␣2 were largely retained (Fig. 4, B and C). Interaction between caveolin-1 and integrin ␣2 remained apparent 30 min after EGF exposure (data not shown).
Caveolae are associated with lipid raft fractions that concentrate the ganglioside GM1, and GM1 serves as a marker for caveolae-like rafts (40,41). As further confirmation that integrin ␣2 is present in caveolar rafts, we examined localization of the EGF receptor or integrin ␣2 with fluorescently labeled Alexa 488-GM1. The results were comparable with those obtained for caveolin-1 (Fig. 4A) in which both EGF receptor and integrin ␣2 colocalized with GM1 before EGF treatment (Fig. 5) but only integrin ␣2 was associated with GM1 after EGF treatment (Fig. 5). Based on these findings, we conclude that integrin ␣2 is concentrated in a subset of GM1-rich caveolae-like rafts and, in response to EGF, internalized by a caveolae-mediated endocytic pathway.
Internalized Integrin ␣2 Is Targeted to the Golgi Apparatus and the ER-The caveosome is an endocytic compartment that is distinguished from the early endosome by neutral pH and by the presence of caveolin-1 (42). SV40 and cholera toxin are internalized by caveosome-mediated mechanisms and then sorted to Golgi and ER, respectively (43,44). To investigate the fate of integrin ␣2 after internalization, we used the Golgi marker 58K and ER marker GRP94. OVCA 433 cells were treated with EGF for 30 min and labeled with 58K or GPR94 and integrin ␣2. Co-localization of integrin ␣2 and 58K or GPR94 was observed in response to incubation with EGF (Fig. 6). These data suggest that internalized integrin ␣2 has been delivered to Golgi apparatus and ER. DISCUSSION Functional interactions between integrins and receptor tyrosine kinases through reciprocal activation and cooperation in signal transduction have been described (45). Often these interactions occur in an integrin-, receptor-, and cell type-dependent manner. Based on co-immunoprecipitation studies, the EGF receptor forms complexes with integrin ␣6␤4 (24) and integrin ␣2 (18), suggesting a close linkage between EGF receptor and the functions of these integrins. The integrin ␣2 cytoplasmic domain is required for EGF-stimulated migration (27,28), and integrin ␣2 is reportedly required for serum-independent activation of the EGF receptor at sites of cell:cell contact in A431 cells (18). We reported previously that a constitutively active form of the EGF receptor EGFRvIII down-regulated integrin ␣2 protein, and the loss of integrin ␣2 was accompanied by aberrant spreading and focal adhesion formation on type I collagen (29). Inhibition of EGFRvIII catalytic activity restored integrin ␣2 expression, cell spreading, and assembly of focal adhesions, suggesting that EGF receptor kinase activity regulated integrin ␣2 functions. In this study we find that ligand-activated EGF receptor causes transient internalization of integrin ␣2 through a caveolae/raftmediated mechanism.
There are many examples of integrin trafficking from the cell membrane with evidence for selective recycling of specific integrins and internalization by different mechanisms (46). Clathrin-dependent mechanisms have been described and integrins are involved in viral entry through clathrin-mediated pathways (33,46,47). ␣v␤5 is reportedly recruited to clathrincoated pits, and integrin ␤1 and L1 adhesion molecule complexes are internalized by clathrin-dependent endocytosis (46,48). Despite the observed association between integrin ␣2 and the EGF receptor before ligand addition, integrin ␣2 did not gain entry through the clathrin-dependent pathway with the EGF receptor in ovarian tumor cells (Fig. 2).
Clathrin-independent and caveolae-mediated internalization of integrins has been described for certain integrins, including integrin ␣2␤1 (38,39). Caveolae are cholesterol-and sphingolipid-rich smooth invaginations of the plasma membrane that partition into raft fractions and whose expression is associated with caveolin-1 (49 -52). A number of integrins, including ␣v␤3 and ␣5␤1, associate with caveolin-1, and ␣2␤1 redistributes to caveolae after integrin clustering (38,46). In studies using human osteosarcoma cells transfected with ␣2 integrin, integrin ␣2␤1 was internalized into caveosome-like structures but direct interaction with caveolin-1 was not determined (38). We found co-localization and biochemical interaction between caveolin-1 and integrin ␣2 in resting and EGFstimulated cells (Fig. 4) that persisted during integrin ␣2 internalization. Furthermore, integrin ␣2 internalized with GM1, which has been extensively used as a marker for glyco- and treated (lower panel) with EGF (25 nM) for 30 min, were fixed in 3.7% formaldehyde, permeabilized with 0.1% Triton X-100, labeled with sheep anti-EGFR, mouse anti-integrin ␣2, and rabbit anti-caveolin-1 antibody, and imaged by confocal immunofluorescence microscopy. The merged images represent EGF receptor in green, integrin ␣2 in red, and caveolin-1 in blue. Co-localization of EGF receptor, integrin ␣2, and caveolin-1 are shown in white, and co-localization between integrin ␣2 and caveolin is shown in purple. Scale bars, 5 m. B, EGF receptor, integrin ␣2, or caveolin-1 were immunoprecipitated from OVCA 433 cell lysates treated without or with EGF for 10 min. Aliquots of the immunoprecipitates were immunoblotted with anti-caveolin-1 (left panels) or anti-EGF receptor (top right panel) or anti-integrin ␣2 (lower right panel) antibody. Immunoprecipitates were run on 15 or 7.5% SDS-PAGE and immunoblotted with caveolin-1 or EGF receptor, anti-integrin ␣2 polyclonal antibodies. Nonspecific rabbit IgG was used as a control. C, OVCA 433 cells were incubated with 3,3Ј dithiobis (sulfosuccinimidylpropionate) (DTSSP) and quenched with glycine. Then EGF receptor, integrin ␣2 (top panel), or caveolin-1 (lower panel) were immunoprecipitated from OVCA 433 cell lysates treated without or with EGF for 10 min as described in panel B. lipid raft domains (40,41) and is associated with caveolae (53,54). These findings suggest that integrin ␣2 is concentrated in caveolar rafts and internalized by a caveolae-mediated endocytic pathway in response to EGF receptor activation.
Caveolae do not appear to be involved in constitutive integrin endocytosis, but this internalization pathway can be activated by a variety of mechanisms (46). Trafficking by caveolae is less well understood than other endocytic pathways, but it is stimulated by a number of viruses binding to certain integrins, including ␣2␤1 or glycosphingolipids (42). Other mechanisms promote caveolar endocytosis, such as C8-LacCer or cholesterol (39), leading to rapid internalization of ␤1 integrins and protein kinase C-dependent internalization of ␣2␤1 (38). Our studies indicate that EGF receptor activation is an additional mechanism to stimulate caveolar trafficking pathways.
Recent studies (55,56) have shown that a variety of cell surface receptors, including the EGF receptor, are present in caveolae and lipid rafts. A caveolin binding motif within the kinase domain of the EGF receptor mediates the interaction of EGF receptor with caveolin-1 (56,57). In human glioblastoma cell lines, EGF receptor rapidly moves out of caveolae domains in response to EGF and then internalizes and degrades via a clathrin-dependent endocytic pathway (56,58,59). In our studies we detected co-localization and biochemical association between the EGF receptor, integrin ␣2, and caveolin-1 without EGF treatment. After EGF binding, interaction between the EGF receptor and integrin ␣2 or caveolin-1 was no longer detected and EGF receptor was located in clathrin-associated vesicles (Figs. 3 and 4). Although the EGF receptor was degraded (Fig. 2), integrin ␣2 was trafficked to the Golgi apparatus and ER and surface levels of integrin ␣2 were restored to near control levels within 4 -6 h after EGF treatment. This targeting of integrin ␣2 is consistent with findings that cholera toxin and SV40 are delivered to the Golgi and endoplasmic reticulum, respectively, following caveolae-mediated internalization (42,49,60).
Trafficking is a well recognized mechanism to modulate signaling of receptors such as the EGF receptor (61); similarly, integrin endocytosis is understood to regulate cell function (42,46). Inhibition of integrin recycling interferes with cell spreading and migration, and integrin internaliza-  A and B, bottom panels) with EGF (25 nM) for 30 min, were fixed in 3.7% formaldehyde for 10 min at room temperature, labeled with (A) mouse anti-EGFR and (B) anti-integrin ␣2 and Cy3-conjugated anti-mouse antibody, incubated with 150 nM Alexa 488-GM1 for 30 min at 37°C, and then imaged by confocal immunofluorescence microscopy. The merged images present integrin ␣2 in red and GM1 in green with co-localization between integrin ␣2 and GM1 in yellow. Scale bars, 5 m. After fixation with dry cold 70% methanol/30% acetone, the Golgi apparatus was detected using rabbit anti-58K and ER using rabbit anti-GPR94. The merged confocal images represent integrin ␣2 in green and 58K or GPR94 in red with co-localization between integrin ␣2 and 58K or GPR94 in yellow. Scale bars, 5 m. tion is proposed to play a role in speed and directionality of migrating cells (46). Our findings that ligand-activated EGF receptor results in transient integrin ␣2 internalization and that constitutively activated EGFRvIII leads to persistent integrin ␣2 down-regulation (29) suggest a novel mechanism by which EGF receptor activation may regulate cell behavior and ovarian cancer metastasis.