A Novel Protease-docking Function of Integrin at Invadopodia*

Invadopodia are membrane extensions of aggressive tumor cells that function in the activation of membrane-bound proteases occurring during tumor cell invasion. We explore a novel and provocative activity of integrins in docking proteases to sites of invasion, termed invadopodia. In the absence of collagen, α3β1 integrin and the gelatinolytic enzyme, seprase, exist as nonassociating membrane proteins. Type I collagen substratum induces the association of α3β1 integrin with seprase as a complex on invadopodia. The results show that α3β1integrin is a docking protein for seprase to form functional invadopodia. In addition, α5β1 integrin may participate in the adhesion process necessary for invadopodial formation. Thus, α3β1 and α5β1 integrins play major organizational roles in the adhesion and formation of invadopodia, promoting invasive cell behavior.


Invadopodia are membrane extensions of aggressive tumor cells that function in the activation of membranebound proteases occurring during tumor cell invasion.
We explore a novel and provocative activity of integrins in docking proteases to sites of invasion, termed invadopodia. In the absence of collagen, ␣ 3 ␤ 1 integrin and the gelatinolytic enzyme, seprase, exist as nonassociating membrane proteins. Type I collagen substratum induces the association of ␣ 3 ␤ 1 integrin with seprase as a complex on invadopodia. The results show that ␣ 3 ␤ 1 integrin is a docking protein for seprase to form functional invadopodia. In addition, ␣ 5 ␤ 1 integrin may participate in the adhesion process necessary for invadopodial formation. Thus, ␣ 3 ␤ 1 and ␣ 5 ␤ 1 integrins play major organizational roles in the adhesion and formation of invadopodia, promoting invasive cell behavior.
The integrin family of transmembrane adhesion proteins has been shown to exhibit multiple functions, including adhesion to extracellular matrix (ECM), 1 cytoskeleton organization, and signal transduction (1)(2)(3)(4). Because integrin and integrin-associated molecules are enriched at membrane protrusions called invadopodia (5)(6)(7)(8)(9)(10), we hypothesized that integrins may also be involved in recruiting proteases to these sites of cell invasion. In support of this hypothesis, the ␣ v ␤ 3 integrin has been shown to modulate ECM proteolytic activities by recruiting a major soluble protease, matrix metalloproteinase-2, to the cell surface (11). Moreover, both adhesive and signaling activities of integrins can be regulated by the interaction between integrins and the urokinase plasminogen activator/receptor (12). We have shown that in LOX melanoma cells, a 170-kDa membrane gelatinase, seprase, was localized to invadopodia and associated with the invasive phenotype (13)(14)(15)(16). Sequencing data on the 97-kDa protein subunit of seprase indicates only a short (six) amino acid sequence at the cytoplasmic amino terminus (14), suggesting that seprase localization at invadopodia may be dependent upon other membrane proteins such as integrins.
Here, we show immunoprecipitation, immunofluorescence, and cell surface cross-linking experiments demonstrating that seprase and ␣ 3 ␤ 1 integrin associate at invadopodia in a collagendependent manner.
Immunofluorescence Labeling of Seprase and Integrin-For direct immunofluorescence localization, purified rat mAb C27 was directly conjugated with fluorescein isothiocyanate (FITC hydrochloride, 10% on Celite, Research Organics Inc., Cleveland, OH) according to the manufacturer's instructions. Anti-seprase mAb D28 was similarly conjugated with rhodamine (tetramethylrhodamine, 10% on Celite, Research Organics Inc.). Cells were cultured on cross-linked gelatin films, fixed, and immunolabeled in a single step using these directly conjugated mAbs (8,18). Alternatively, LOX cells were indirectly labeled for integrin subunits following fixation in the presence (0.1% Triton X-100) or absence of detergent and secondary antibody detection of rat mAbs using Texas Red goat anti-rat antibodies (Jackson Laboratories, West Grove, PA) as described previously (8).
Preparation of Lysates, Immunoprecipitation, and Zymography-LOX cells were seeded onto cross-linked gelatin films (18) or hydrated collagen I films (rat tail type I collagen at 1 mg/ml according to manufacturers instructions, Collaborative Biomedical Products, Becton and Dickinson Labware, Bedford, MA) and cultured overnight until 80 -90% confluence. To harvest lysates, each 175-cm 2 plate was washed once with 25 ml of PBS, pH 7.4, at 25°C and then extracted with 25 ml of PBS containing 0.1% Triton X-100 and 0.02% NaN 3 by incubating for 2 h at 25°C on a rotary shaker (25 rpm, Bellco Orbital Shaker, Vineland, NJ). The cell layer and buffer (or gelatin plus cell layer and buffer) were transferred to a 50-ml conical tube and incubated a further 3 h at 4°C with end-over-end agitation. The extract was clarified by centrifugation at 10,000 ϫ g for 20 min at 4°C and the supernatants were used for immunoprecipitation reactions. Cell body (cb) and invadopodia membranes (in) were rapidly harvested by shearing the cell bodies in 25 ml PBS after a brief PBS wash. The invadopodia and cell bodies were extracted in 25 ml of extraction buffer as described above for lysates. Purified rat and mouse mAbs against membrane proteins (2.5 mg) were coupled to 1 ml of Sepharose 4 MB (50% slurry) and 0.25 ml used to immunoprecipitate complexes from 25 ml of cell extract overnight at 4°C with end-over-end agitation. After 3ϫ washes in 25 ml of extraction buffer, the beads with coupled antibody-antigen complexes were resuspended in extraction buffer (equal to the bead volume) and the sample subjected to 3 cycles of sonication on ice (setting 20, 10 s each using a KONTES Micro Ultrasonic Cell Disrupter). Immediately, the sample was transferred to an Amicon filter insert (0.45 m, 400-l capacity) and centrifuged 20 min at 10,000 rpm in an Eppendorf microfuge at 4°C. The bead filtrate was used either for Western blotting of integrin and seprase or for zymography to detect seprase gelatinase activity. To test for complete extraction of the seprase and integrin complexes from the beads, Laemmli sample buffer (equal to the bead volume) was added, and the samples were heated by microwaves (2 cycles on low setting, 30 s each, followed by 1 cycle on medium for 30 s). Then, the samples were immediately centrifuged at 25°C. The filtrates were subjected to immunoblotting and gelatin zymography as described (14,15). Antigens were essentially absent from post-sonication beads. We concluded that sonication detached antigens from antibodies, but the antibodies were only removed from the Sepharose beads after extraction with Laemmli sample buffer.
Biotinylation and Chemical Cross-linking-LOX cells were cultured on plastic or collagen overnight. After washing cells with PBS, pH 7.4, 1 mM Ca 2ϩ , 1 mM MgCl 2 at 4°C, cell surface proteins were cross-linked at 4°C for 30 min in the same buffer using BS-3 as described by the manufacturer (Pierce). Following additional washing in the above buffer, cells were surface biotinylated at 4°C for 120 min using SHbiotin (Pierce) as described by the manufacturer. Cross-linking and biotinylation of cell surface proteins were carried out at 4°C to prevent internalization of cell surface proteins. In addition, these reactions were carried out in the absence of detergent to ensure that intracellular pools of integrins and seprase would not be labeled or cross-linked. Cells were then solubilized in RIPA buffer and immunoprecipitated with C27 or D28. After multiple washes, immunoprecipitated complexes were solubilized in Laemmli sample buffer by boiling without reduction, and the cross-linked proteins were separated on a 7.5% SDS-polyacrylamide gel electrophoresis gel. Ferritin (440 kDa, Amersham Pharmacia Biotech) was used as the high molecular weight standard in addition to the routinely used 205-, 116-, 97.4-, 66-, 45-kDa standards (Sigma).

RESULTS
To determine invadopodial proteins that associate with seprase and participate in matrix degradation and invasion, the mAb C27 was generated using detergent soluble proteins derived from LOX melanoma cells that exhibited gelatinolytic activities. Immunoaffinity chromatography using mAb C27 identified two major bands in the LOX cell extract (Fig. 1A). The first band at 120 kDa, co-migrated with ␤ 1 integrin, and the second migrated at 150 kDa. To determine whether the 120-kDa C27 antigen was ␤ 1 integrin, the C27 antigen was isolated from LOX RIPA extracts by affinity chromatography using either mAb C27 or mAb 13 that recognizes ␤ 1 integrin (17). The eluates were immunoblotted with each of the mAbs and with polyclonal antibodies against ␤ 1 (3847) or ␤ 3 (antivitronectin receptor) integrin. The results in Fig. 1B indicate that the C27 antigen band at 120 kDa is ␤ 1 integrin. The C27 antigen was not ␤ 3 integrin, because ␤ 3 integrin was expressed at high levels in platelets but was not detected in LOX antigen preparations (Fig. 1B). The C27-120 and 150-kDa antigens were also affinity-purified from MDA-MB-231 breast carcinoma cells, and the 120-kDa band in C27 antigen preparations was identified as ␤ 1 integrin in these cells (Fig. 1C).
We sought to identify the 150-kDa band by subjecting it to N-terminal peptide sequencing. The following sequence was obtained: F N L D T R F L. The data base search program "FindPatterns" was used to search for the identical protein sequence allowing zero mismatches in the protein data banks (PIR-Protein and SwissProt). This sequence is 100% identical to the N-terminal residues 1-8 of the human integrin ␣ 3 chain/ galactoprotein ␣ 3 /very late antigen-␣ 3 chain; residues 38 -45 of human integrin VLA-3 ␣ 3 chain precursor; and residues 38 -45 of golden hamster cell surface glycoprotein ␣ 3 precursor (20 -22). Also, the MDA-MB-231 150-kDa band was sequenced and found to be ␣ 3 integrin. We conclude that mAb C27 recognizes ␤ 1 integrin on Western blots, and appears to preferentially immuno-isolate ␣ 3 ␤ 1 heterodimers from these two cell lines.
To probe further the possible interaction between ␣ 3 ␤ 1 integrin and seprase, immunoprecipitation was performed on lysates of LOX cells cultured on plastic ( Fig. 2A) or collagen or gelatin (Fig. 2, B and C) with mAbs D8 or D28 to detect seprase (14,15,18) and C27 to detect ␤ 1 integrin. In three independent experiments, a stable association of integrin and seprase was not detected when cells were cultured on plastic ( Fig. 2A) but was reproducibly detected in lysates prepared from cells that were cultured on collagen or gelatin (Fig. 2B). This association was specific because anti-␣ v integrin or rat mAb E19 (control) did not co-immunoprecipitate either ␤ 1 integrin or seprase (Fig. 2B).  1. Characterization of C27 antigen isolated by affinity chromatography as ␤ 1 integrin. A, silver-stained gel of C27 antigen affinity isolated from LOX melanoma cell extracts using RIPA buffer. B, immunoblotting of affinity isolated antigens from LOX. mAbs C27 and 13 (␤ 1 specific) were used to isolate antigen. Platelet lysates (10 g) were loaded as positive control for ␤ 3 integrin. Anti-␤ 1 polyclonal (3847) and mAb 13 antibodies detected ␤ 1 integrins, and anti-vitronectin receptor antibodies detected the ␤ 3 subunit of the vitronectin receptor. C, C27 antigen affinity-isolated from MDA-MB-231 breast carcinoma cell lysates. Four lanes are immunoblots of C27 isolated proteins using mAbs 13 and C27.
To determine the localization of the seprase-␣ 3 ␤ 1 complex at invadopodia, we fractionated LOX cells into an invadopodiaenriched fraction (in) and the cell body fraction (cb) as described previously (8,18). Association of seprase and integrin occurred specifically in the invadopodia fraction rather than in the cell body fraction (Fig. 2B, IP: seprase and ␤1, cb versus in) despite the predominant localization of ␤ 1 integrin in the cell body fraction (Fig. 2B, IP and BLOT: ␤ 1 , cb versus in). As previously demonstrated (18), we found that seprase was concentrated in the invadopodia-enriched membrane fraction (in) with very little detected in the remaining cell body (Fig. 2B, IP and BLOT: seprase, cb versus in). These data suggest the existence of a stable invadopodial complex consisting of seprase and ␤ 1 integrin. Furthermore, gelatin zymography detected a 170-kDa gelatinase activity in immunoprecipitates of anti-seprase mAb D28 (Fig. 2C, IP: seprase) or anti-␤ 1 integrin (Fig.  2C, IP: ␤ 1 ). Lysates (ly) from cells cultured on cross-linked gelatin films or on collagen I layers contained equal amounts of seprase gelatinase activity co-immunoprecipitating with ␤ 1 integrin. This demonstrated that both native or denatured collagen matrices were equally effective in eliciting co-immunoprecipitation of seprase-and ␤ 1 -integrin (Fig. 2C, IP: ␤ 1 ). Similar to what was observed by immunoblot detection of seprase, the association of seprase gelatinase activity with integrin occurred predominantly in the invadopodia fraction (Fig. 2C). Control immunoprecipitations using mAb E19 (control) and anti-␣ v mAb did not result in immunoprecipitation of any detectable seprase gelatinase activity (Fig. 2C).
Immunoprecipitations of individual ␣ subunits of integrin were used to determine the specificity of seprase interactions with integrins. Lysates from LOX cells cultured on plastic or collagen were immunoprecipitated using anti-␣ 2 , ␣ 3 , or ␣ 6 integrin mAbs or anti-seprase mAb D28 (Fig. 3). Western blotting of immunoprecipitates revealed that anti-␣ 3 mAbs, but not mAbs against ␣ 2 or ␣ 6 , were able to precipitate seprase from cells cultured on collagen (Fig. 3A, BLOT seprase, IP ␣ 3 versus IP ␣ 2 or ␣ 6 obtained from lysates of LOX cells cultured on collagen, lane 7 versus lanes 6 and 8). In the complementary immunoprecipitation, anti-seprase mAb D28 only co-precipitated ␤ 1 integrin from cells cultured on collagen (Fig. 3A, BLOT ␤ 1 , IP seprase obtained from lysates of LOX cells cultured on collagen, lane 5). Western blotting using secondary antibody only (control) revealed background bands that were present particularly in the ␣ 6 lane (see Fig. 3A, brackets, rat IgG). These bands, however, were not related to seprase as demonstrated by zymography (Fig. 3B). The increased background bands might be due to impurities that were present in the ␣ 6 antibody preparation or the fact that this particular antibody results in higher background binding to lysate proteins. Zymography was used to detect seprase gelatinolytic activity. Activity was only immunoprecipitated from cells cultured on collagen using anti-␣ 3 or seprase mAbs (Fig. 3B). Thus, we conclude that collagenous matrix can induce the seprase-␣ 3 ␤ 1 association in invadopodia, which results in the localization of the 170-kDa gelatinase activity at sites of matrix degradation.
In vitro experiments to determine the interaction between ␣ 3 ␤ 1 and seprase are not very feasible, because the association of these dimeric membrane molecules requires cell attachment to matrix and occurs only in membranes isolated from the invadopodia-enriched fraction. Therefore, cross-linking and immunoprecipitation analyses were used to explore the cell surface association of seprase and integrin. Chemical cross-linking experiments were used to determine whether seprase interacts directly with ␤ 1 integrin complexes on the cell surface. Because immunoprecipitation experiments demonstrated that this interaction occurred only in cells cultured on collagen, but not on plastic, we expected that cells cultured on plastic would not form oligomers of seprase and ␣ 3 ␤ 1 integrin. And, conversely, cells cultured on collagen would be expected to contain complexes of a molecular weight corresponding to seprase dimer (170 kDa) plus ␣ 3 ␤ 1 dimer (ϳ270 kDa), thus a complex of about FIG. 2. Co-immunoprecipitation of ␤ 1 integrin and seprase from cells cultured on plastic, collagen, or gelatin. A, seprase and ␤ 1 are not co-immunoprecipitated from extracts of LOX cells cultured on plastic. Equal amounts of lysates were incubated with beads directly conjugated with anti-seprase mAbs D8 (seprase, on left) and D28 (seprase, on right), control rat mAb F4 (control), anti-␤ 1 integrin mAb C27 (␤ 1 ), anti-␣ v mouse mAbs (␣ v ) or no primary antibody (no 1 o ). Immunoprecipitates were analyzed by Western blotting under reducing conditions using anti-␤ 1 integrin mAb 13, anti-seprase mAb D28, and a negative control, rat mAb E26. B, seprase and integrin co-immunoprecipitated from total extract or invadopodia fractions of LOX cells cultured on cross-linked gelatin or collagen. Amounts of lysates (ly) or subcellular fractions containing invadopodia membranes (in) or the cell bodies (cb) representing equal numbers of cells were immunoprecipitated using control rat mAb E19 (control), mAb D28 (seprase), mAb C27 (␤ 1 ) or anti-␣ v integrin mouse mAb (␣ v ). One half of each immunoprecipitate was analyzed under nonreducing conditions by immunoblotting with D28, C27, or control rat mAb E19 (15). C, gelatin zymography of the other half of each of the samples from B .   FIG. 3. Specific co-immunoprecipitation of seprase with ␣ 3 integrin from total lysates. A, immunoprecipitation of cells cultured on plastic or collagen using mAb D28 directed against seprase or anti-␣ subunit mAbs. Anti-␣ 2 integrin (mouse mAb clone P1E6), anti-␣ 3 integrin (mouse mAb clone P1B5), and rat anti-␣ 6 (clone GoH3) were used to perform immunoprecipitation. For blotting, mAb D8 against seprase, mAb 13 against ␤ 1 integrin and control rat IgG were used. Only anti-␣ 3 mAb co-immunoprecipitated seprase from cells cultured on collagen but not on plastic surfaces (seprase). Anti-␣ 2 mAb immunoprecipitated only low levels of ␤ 1 integrin with no associated seprase. Background bands are revealed in the rat IgG control blot. B, gelatin zymography (ZYM) of seprase-␣ 3 ␤ 1 complex. Anti-␣ 3 but not ␣ 2 or ␣ 6 integrin mAbs immunoprecipitated seprase gelatinase activity that co-migrated with authentic seprase immunoprecipitated using mAb D28 (seprase). 440 kDa. Comparison of D28 and C27 immunoprecipitates from lysates of cells cultured on plastic versus collagen, confirmed our prediction that seprase and ␣ 3 ␤ 1 integrin were associating in a direct manner at the cell surface (Fig. 4, Specifically, a complex of ϳ430 kDa was precipitated both by C27 and D28 demonstrating that the complex contains both ␣ 3 ␤ 1 integrin and seprase dimers (Fig. 4, coll1 lanes). This high molecular weight band was only observed in the lanes derived from cells cultured on collagen (Fig. 4, ␣ 3 ␤ 1 ϩ seprase, IP D28 and C27, compare plastic versus coll1).
In cells cultured on plastic, D28 seprase dimer was the major species observed, whereas C27 precipitated ␣ 3 ␤ 1 dimer as well as a prominent band at the position of ␤ 1 monomer (Fig. 4, ␣ 3 ␤ 1 and ␤ 1 ). In cells cultured on collagen, seprase dimers were detected in D28 immunoprecipitates (Fig. 4, seprase dimer, IP D28, coll1 lane). Bands co-migrating with the expected molecular weight of seprase monomer also appear in the D28 and C27 lanes (Fig. 4, seprase monomer, IP D28 and C27, coll1 lanes). We conclude that 1) the immunoprecipitates observed in Fig. 4 are derived from the cell surface; 2) the high molecular weight complex at 430 kDa is formed only when cells are cultured on collagen 1; and 3) this high molecular weight complex corresponds to seprase dimer plus ␣ 3 ␤ 1 dimer, because it is immunoprecipitated by anti-seprase or anti-integrin antibodies.
Cell surface expression of seprase and ␣ 3 ␤ 1 integrin distribution was also examined using fluorescence-activated cell sorter and immunofluorescence microscopy of cells in suspension or cultured on collagenous substrata. Comparison of the mean values from fluorescence-activated cell sorter analysis of 10,000 suspended cells for each antibody, relative to secondary antibody alone controls, revealed no difference in the levels of either of these proteins on either substratum (data not shown). Immunofluorescence studies on cells cultured in the absence of collagen demonstrated seprase and ␤ 1 protein at membrane ruffles and diffusely distributed on the remainder of the plasma membrane. To confirm that seprase and integrin are associated in invadopodia during localized ECM degradation, double label immunofluorescence experiments were performed. We found that mAb C27 directed against ␤ 1 integrin did colocalize with seprase in the same invadopodia (Fig. 5A). Furthermore, anti-␣ 3 mAb P1B5 labeled invadopodia directly overlying the sites if localized matrix degradation (Fig. 5B, closed arrows) as well as membrane extensions at the cellular margin (Fig. 5B, open arrow). However, the ␣ 5 ␤ 1 integrin, a fibronectin receptor, localized in focal adhesions throughout the cell (anti-␣ 5 mAb 11; Fig. 5C, open arrow) as well as in fine adhesion structures surrounding the base of invadopodia (Fig. 5C, solid  arrow). Double label experiments using combinations of anti-␣ 3 /mAb 13, anti-␣ 5 /mAb 13, and C27/mAb 13 demonstrated the selective staining of invadopodia and membrane protrusions by anti-␣ 3 ␤ 1 and C27, and the recognition of focal adhesions by anti-␣ 5 ␤ 1 mAbs (data not shown). Anti-␤ 1 mAb 13 staining was a combination of the two, both invadopodia and focal adhesion staining. Detergent extraction (0.1% Triton X-100) during fixation of cells effectively reduced diffuse membrane staining by anti-integrin mAbs while retaining invadopodia staining in the case of anti-␣ 3 ␤ 1 and C27 or focal adhesion staining in the case of anti-␣ 5 ␤ 1 staining (data not shown). This suggests that a subpopulation of integrin is anchored to the cytoskeleton at invadopodia as well as focal adhesions. The presence of ␣ 5 in punctate sites at the base of invadopodia and in streaks at the cellular periphery suggests that these structures are both related to focal adhesions, sites of actin-membrane anchorage to the membrane. Finally, we compared the immunolocalization of ␣ 3 and the related ␣ 6 integrin, because we had found that ␣ 3 but not ␣ 6 co-immunoprecipitated with seprase. The ␣ 6 subunit localized in a pattern that was indistinguishable with that of ␣ 3 in LOX cells cultured on cross-linked gelatin. Anti-␣ 6 mAbs labeled filopodia and invadopodia associated with sites of degradation (data not shown). Thus, ␣ 3 integrin associates with seprase even though both (␣ 3 and ␣ 6 ) integrins localize to the same filopodia and invadopodia structures. We therefore suggest that ␣ 3 ␤ 1 integrin is a protease-docking protein for directing seprase to invadopodia, and ␣ 5 ␤ 1 may participate in the adhesion process necessary for formation and extension of invadopodia (Fig. 5D).

DISCUSSION
Considering that integrins are linked both to ECM components and to cytoskeleton, and that seprase possesses a putative cytoplasmic domain of only 6 amino acids, it seems likely that ␣ 3 ␤ 1 integrin actively forms a docking site for seprase. Additionally, ␣ 3 ␤ 1 integrin may immobilize seprase to invadopodia following ␣ 3 ␤ 1 attachment to the cytoskeleton and collagenous matrices. This argument is not without precedent, because ligation or aggregation of integrins by fibronectin or by anti-integrin antibodies recruits cytoskeletal and signaling molecules to the membrane (6,7,9,23,24). Our immunofluorescence, cross-linking, and immunoprecipitation studies demonstrate the association between ␣ 3 ␤ 1 integrin and seprase that occurs only when cells are cultured on matrix. This scenario is also consistent with the state of tyrosine phosphorylation of these cytoskeletal/signaling molecules and their localization at invadopodia (8,9). In addition, genistein, a tyrosine kinase inhibitor, inhibits tyrosine phosphorylation of proteins at invadopodia as well as the degradative and motile activities of invadopodia (8,9). Immunoprecipitation and immunofluorescence data shown in this paper demonstrate that ␣ 3 ␤ 1 may participate in the formation of invadopodia by docking seprase, and ␣ 5 ␤ 1 integrin appears to function in adhesion structures throughout the cell, particularly at the base of invadopodia. ␣ 3 ␤ 1 is primarily a receptor for the basement membrane-associated molecules epiligrin and laminin/merosin, but also for fibronectin and collagen types I and IV (25)(26)(27). A related laminin binding integrin, ␣ 6 ␤ 1 , does not associate with seprase, even though we found that it was localized at the membrane in filopodia and invadopodia in a pattern very similar to ␣ 3 ␤ 1 (data not shown). However, ␣ 6 ␤ 1 integrin previously was demonstrated to play a role in signal transduction that promotes invadopodial activities (28) suggesting that these integrins may coordinately regulate the activities of invadopodia. Taken   FIG. 4. Chemical cross-linking of seprase and ␣ 3 ␤ 1 integrin on the surface of LOX cells. Cross-linked and biotinylated cell surface proteins were immunoprecipitated using anti-seprase (D28) or anti-␤ 1 integrin (C27) antibodies. When LOX cells were cultured on plastic (plastic), D28 precipitated predominantly the seprase dimer that migrates at 170 kDa (seprase dimer) from cell lysates, whereas C27 immunoprecipitated primarily ␤ 1 integrin monomer (120 kDa, ␤ 1 ) and ␣ 3 ␤ 1 dimer (ϳ270 kDa, ␣ 3 ␤ 1 ). In contrast, following cell culture on collagen (coll1), D28 immunoprecipitated a ϳ430-kDa complex of ␣ 3 ␤ 1 plus seprase (␣ 3 ␤ 1 (ϳ270 kDa) ϩ seprase (170 kDa)), seprase dimer (170 kDa, seprase dimer), ␤ 1 monomer (120 kDa, ␤ 1 ), and seprase monomer (95 kDa, seprase monomer). C27 immunoprecipitated a high molecular weight complex of the same size as that immunoprecipitated by D28 (␣ 3 ␤ 1 ϩ seprase).
together, these results suggest that proteolytic activity at the tip of the invadopodia degrades the matrix to weaken resistance to invasion and that collagen-induced ␣ 3 ␤ 1 association with seprase participates in this process. We speculate that ␣ 5 ␤ 1 supports the localized membrane attachment for extension of invadopodia into the matrix.