Role of Endoproteolytic Processing in the Adhesive and Signaling Functions of αvβ5 Integrin*

Some integrin α subunits undergo a post-translational cleavage in their extracellular domain. However, the role of this cleavage in integrin function is unclear. Enzymes involved in this maturation belong to the subtilisin-like endoprotease family (convertases). To understand the role of the α subunit cleavage in integrin function, we have designed stable transfectants (PDX39P cells) expressing α1-PDX, a convertase inhibitor. Immunoprecipitation of cell surface proteins from PDX39P showed that α3, α6 and αvintegrins lack endoproteolytic cleavage. We have compared adhesion between PDX39P cells and mock-transfected cells on different extracellular matrix proteins. No difference in adhesion could be observed on laminin-1 and type I collagen, while attachment of PDX39P cells to vitronectin (ligand of the αvβ5integrin) was dramatically reduced. The reduced adhesion of PDX39P cells was not due to changes in integrin affinity as determined by solid-phase receptor assay in a cell-free environment. Intracellular signaling pathways activated by αv integrin ligation were also affected in PDX39P cells. It thus seems that the absence of endoproteolytic cleavage of αv integrins has important consequences on signal transduction pathways leading to alterations in integrin function such as cell adhesion.

Integrins are transmembrane glycoproteins, composed of noncovalently associated ␣ and ␤ subunits, that are involved in cell-extracellular matrix (ECM) 1 and cell-cell interactions (1). Many integrin ␣ chains undergo a post-translational endoproteolytic cleavage. The ␣ 3 , ␣ 5 , ␣ 6 , ␣ 7 , ␣ 8 , ␣ 9 , ␣ v , and ␣ IIb subunits are cleaved in the membrane-proximal extracellular region, resulting in a heavy chain that is disulfide-linked to a membrane spanning light chain (2). The ␣ 4 and ␣ E subunits can also be cleaved, but at unusual positions, near the middle and in the N-terminal region of the molecule, respectively (3,4). Endoproteolytic cleavage of integrin ␣ subunits occurs at specific sites comprising pairs of basic amino acids.
Post-translational proteolysis is a common mechanism required for the synthesis of biologically active proteins in bacteria, fungi, yeast, invertebrates, and mammals (5). However, the role of endoproteolytic cleavage of integrin ␣ subunits is not clear. The cleavage is conserved, not only in different ␣ chains but also across species, suggesting that it might be of functional importance. It has been established, by site-directed mutagenesis of cleavage sites, that uncleaved ␣ IIb ␤ 3 and ␣ 4 ␤ 1 are able to mediate cell adhesion to their ligands (3,6). However, it has been reported that a defect in ␣ 6 cleavage impairs the inside/ out signaling of uncleaved ␣ 6 ␤ 1 integrins induced by phorbol ester, indicating that cleavage is necessary for proper integrin function (7).
The search for mammalian analogues of kexin, a yeast endoprotease, has led to the discovery of the subtilisin/kexin-like family of proprotein convertases. These calcium-dependent serine proteases can be subdivided into four groups according to their cellular localization and tissue distribution (for reviews and updates, see Refs. 8 -11). Furin and PC7, belonging to the first group, are widely expressed in tissues and mainly localized to the trans-Golgi network. PC1 and PC2 are primarily expressed in neural and endocrine cells and are found essentially within secretory granules. The third group includes PC5A and PACE4, widely distributed in tissues and localized within the trans-Golgi network and secretory granules, and PC5B (also called PC6), mostly found within the trans-Golgi network. PC4, exclusively expressed in germ cells, constitutes the fourth group. The general consensus sequence cleaved by these enzymes contains the motif (K/R)-X n -(K/R)2, where n ϭ 0, 2, 4, or 6. In vitro and phage display data revealed that the X residue and those following the cleavage site define the fine specificity of each enzyme (12,13).
Due to their potential pharmacological interest, several inhibitors of proprotein convertases have been designed (14). In humans, a naturally occurring mutation of ␣ 1 -antitrypsin (␣ 1 -AT), known as ␣ 1 -AT Pittsburgh, changes the specificity of this serpin from an inhibitor of elastase into a potent inhibitor of thrombin (15). Another variant of ␣ 1 -AT, called ␣ 1 -AT Portland (␣ 1 -PDX), has been engineered by Anderson et al. (16) and described as a potent inhibitor of convertases. Recent findings shows that ␣ 1 -PDX is a selective inhibitor for furin and, to a lesser extent, for PC5B (14). This inhibitor blocks the convertases dependent processing of various precursors (17)(18)(19)(20)(21).
In a previous report (22), we showed that ␣ 3 , ␣ 6 , and ␣ v integrin subunits are not processed in the furin-deficient LoVo-C5 cell. Moreover, pro-forms of ␣ integrin subunits are cleaved both by recombinant convertases in vitro and ex vivo after overexpression in LoVo cells (22,23). In the present study, we investigated the functional importance of the ␣ subunits cleavage during the interaction of integrins with ECM proteins. We have designed stable transfectants of ␣ 1 -PDX in HT29-D4 adenocarcinoma cells. These transfected cells, bearing uncleaved integrin ␣ subunits, displayed a reduced attachment to vitronectin. This alteration was correlated with defects in the intracellular signaling pathways activated by ␣ v integrin ligation.
Stable ␣ 1 -PDX-expressing Transfectants-HT29-D4 cells, derived from a human colon adenocarcinoma, were transfected with a pBK-CMV expression vector containing the full-length ␣ 1 -PDX cDNA (gift of G. Thomas, Portland, OR). Clones of stable transfected cells (controlled by Northern blot) were isolated and tested for their sensitivity to Pseudomonas exotoxin A (PEA), a toxin activated after cleavage by convertases, which has proved to be an indicative assay for furin activity (22,27). Among the clones tested, the clone PDX39 presented a pronounced increase in PEA resistance compared with HT29-D4 cells or to the clone PDX0 transfected with the empty vector. To generate cells expressing higher levels of ␣ 1 -PDX inhibitor, the clone PDX39 was cultured in the presence of 1 g/ml PEA. This cell subpopulation, resistant to PEA and expressing high levels of ␣ 1 -PDX (controlled by reverse transcription polymerase chain reaction), was referred as PDX39P.
Flow Cytometric Analysis and Cell Surface Labeling-The cell surface expression of integrin subunits was determined by flow cytometry and by immunoprecipitation of biotinylated cell surface proteins as described elsewhere (28).
Cell Adhesion Assay-Adhesion assays were performed as described previously (28,29). Briefly, cells were harvested in single cell suspension by treatment with 0.53 mM EDTA in PBS, added to wells coated with purified ECM proteins, and allowed to adhere to the substrata for 2 h at 37°C. After washing, attached cells were stained by 0.1% crystal violet and lysed with 1% SDS. Absorbance was then measured at 600 nm.
Solid-phase Receptor Assay-Serum-starved cells in suspension were washed twice with 10 mM Tris, pH 7.4, and resuspended in the same buffer containing a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 500 units/ml aprotinin, 1 g/ml leupeptin, 1 M pepstatin, 1 mM iodoacetamide, and 1 mM o-phenanthroline). Cells were disrupted with a Potter homogenizer and homogenate was centrifuged at 600 ϫ g for 10 min at 4°C. Plasma membranes were recovered by centrifuging the supernatant at 50,000 ϫ g for 45 min and then solubilized for 45 min at 4°C with 50 mM Tris, pH 7.4, 150 mM NaCl (TBS) containing 1% n-octyl-␤-D-glucopyranoside, 1 mM MgCl 2 , 1 mM CaCl 2 , and the mixture of protease inhibitors. The extract was clarified by centrifugation at 15,000 ϫ g for 10 min at 4°C, and the protein concentration was determined by the Bio-Rad DC protein assay.
MaxiSorb TM microtiter plate wells (Nunc) were coated for 2 h at 37°C with 10 g/ml vitronectin in 50 l of TBS and blocked with 1% BSA for 1 h at room temperature. Vitronectin-coated wells were incu-bated with 100 l of solubilized membranes (300 g of protein) overnight at 4°C. Unbound proteins were removed by four washes with TBS containing 0.2% Tween 20 and 1% BSA. ␣ v binding was detected by sequential incubations for 1 h at 37°C with 100 l of 10 g/ml mouse mAb AMF-7 and at room temperature with rabbit anti-mouse IgG (1/1000) and anti-rabbit IgG-horseradish peroxidase-conjugated antibodies. Finally, 100 l of peroxidase substrate was added for 10 min at room temperature, the reaction was stopped by adding 100 l of 0.5 mol/liter H 2 SO 4 , and absorbance was measured at 450 nm.
Detection of Activated MAPK and Tyrosine Phosphorylation of FAK and Paxillin-Confluent cells were washed twice with serum-free medium and incubated in Dulbecco's modified Eagle's medium containing 0.1% BSA for 24 h at 37°C. Serum-starved cells were harvested in single cell suspension, added (5 ϫ 10 6 cells in 1 ml) to 9.6-cm 2 wells coated with 10 g/ml poly-L-lysine or vitronectin, and allowed to adhere to the substrata for the indicated periods of time at 37°C. After three washes with PBS, attached cells were lysed with 20 mM Tris-HCl, pH 8, 200 mM NaCl, 1 mM EDTA, and 1% Triton X-100 (RIPA buffer) containing 10 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 10 mM NaF, and the mixture of protease inhibitors. Lysates were clarified by centrifugation and analyzed by immunoblot, after immunoprecipitation, or directly with an anti-active MAPK polyclonal antibody as described previously (30). For immunoprecipitation, 600 -900 g of proteins were incubated with 1 g of anti-FAK or 2 g of anti-paxillin antibodies overnight at 4°C and then with protein G-agarose for 45 min. Pellets were washed three times with RIPA buffer, three times with RIPA buffer, 500 mM NaCl, and once with PBS. Immunoprecipitated proteins were resolved by SDS-PAGE and blotted onto a nitrocellulose sheet. Membranes were blocked in PBS, 4% BSA, 0.2% Tween 20 and probed overnight at 4°C with PY20 antibody in PBS, 0.8% BSA, 0.2% Tween 20. Blots were then visualized with horseradish peroxidase-conjugated secondary antibodies.

RESULTS
Cell Surface ␣ 3 , ␣ 6 , and ␣ v Integrin Subunits Are Uncleaved in PDX39P Cells-␣ 1 -PDX has been described as a potent inhibitor of convertases (16), which are responsible for the cleavage of ␣ integrin subunits (22,23). We therefore expressed the ␣ 1 -PDX inhibitor in HT29-D4 cells, and we selected stable transfectants. The clone PDX39 was further selected on the basis of its resistance to PEA, a toxin activated by convertases (27), to obtain a population (PDX39P cells) expressing high levels of the ␣ 1 -PDX inhibitor (data not shown).
Among the cleavable integrins, HT29-D4 cells express ␣ 3 ␤ 1 , ␣ 6 ␤ 4 , ␣ v ␤ 5 , and ␣ v ␤ 6 (26, 28). The ␣ 6 subunit is synthesized as a 140-kDa precursor and then converted to a 120-kDa species by endoproteolytic processing (31). To determine whether ␣ 1 -PDX expression impairs the endoproteolytic processing of ␣ 6 , cells were surface-biotinylated, and integrins were immunoprecipitated with specific antibodies against ␣ 6 subunit. Immunopurified proteins were then analyzed by SDS-PAGE under nonreducing (NR) or reducing (R) conditions (Fig. 1). As illustrated in Fig. 1A, nonreduced ␣ 6 immunoprecipitates, composed of ␣ 6 chains (140 kDa) associated with ␤ 4 subunits (190 kDa), displayed a similar pattern in the different cell populations. However, when the disulfide bridges were broken, the electrophoretic profiles of ␣ 6 immunoprecipitates were obviously different depending on the cell population. Indeed, in HT29-D4 cells and in cells transfected with the empty vector (PDX0 cells), the apparent molecular mass of ␣ 6 was reduced from 140 to 120 kDa upon reduction, while a major band of 150 kDa corresponding to noncleaved subunit (␣ 6 NC ) was observed in PDX39P cells. Only a limited amount of cleaved ␣ 6 subunit (␣ 6 C ) was seen on the gel for PDX39P. These data, similar to those obtained with the cleavage-deficient LoVo cells (22), indicate that in PDX39P the large majority of the ␣ 6 integrin subunit was not endoproteolytically processed. In the case of PDX39 cells, about half of the ␣ 6 chains were cleaved, confirming that the selection of PDX39 cells in the presence of a high concentration of PEA increased ␣ 1 -PDX expression and led to a strong impairment of ␣ 6 cleavage.
These observations were extended to the ␣ v and ␣ 3 subunits using the same experimental approach. As seen on Fig. 1B, most ␣ v chain, that associated with ␤ 5 and ␤ 6 subunits, failed to be cleaved in PDX39P cells, as observed with ␣ 6 subunit. The amount of normally processed ␣ 3 is more difficult to evaluate because cleaved form (␣ 3 C ) comigrated with reduced ␤ 1 (Fig.  1C). However, the bands corresponding to noncleaved (␣ 3 NC ) and nonreduced ␣ 3 chains from PDX39P have the same intensity, suggesting that virtually all ␣ 3 subunits are in an uncleaved form. Thus, the expression of high levels of ␣ 1 -PDX inhibitor blocked almost totally the endoproteolytic maturation of all the cleavable ␣ integrin subunits expressed in PDX39P cells. Moreover, the ratio cleaved/noncleaved form correlated with the expression level of ␣ 1 -PDX (data not shown).
␣ 1 -PDX Does Not Alter Cell Surface Expression of Integrins-As observed above after cell surface biotinylation and immunoprecipitation, the expression of ␣ 1 -PDX does not seem to affect the labeling intensity of integrin subunits. To confirm these results by a more quantitative approach, cells were examined for cell surface expression of integrin subunits by indirect immunofluorescence using specific mAbs. Flow cytometry analysis indicated that the extent of integrin expression at the cell surface is quite similar whatever the cell population studied (Fig. 2). This demonstrates that the defect in ␣ chains processing did not impair the exportation of the heterodimers to the plasma membrane.
Expression of ␣ 1 -PDX Decreases Adhesion of PDX39P Cells on Vitronectin-To examine the importance of the ␣ chain cleavage for integrin function, we first measured the adhesion of transfected cells to purified ECM proteins and to poly-Llysine as an integrin-independent substrate. Adhesion of PDX39P cells to laminin-1 and collagen type I was quite similar to control cells, as was the attachment to poly-L-lysine (Fig. 3A). On the contrary, adhesion of PDX39P cells to vitronectin was significantly reduced when compared with HT29-D4 or PDX0 cells.
The integrin ␣ v ␤ 5 , the unique receptor for vitronectin in HT29-D4 (26) and in PDX0 and PDX39P cells (not shown), does not require activation to mediate ligand binding. However, the absence of cleavage of ␣ subunits could lead to the suppression of the constitutive integrin activation. We, therefore, assessed whether PMA or the divalent cation Mn 2ϩ could restore a normal attachment of ␣ 1 -PDX-expressing cells on vitronectin. As illustrated in Fig. 3B, stimulation of PDX39P cells by PMA failed to restore cell adhesion to vitronectin. Moreover, treatment of cells by 1 mM (Fig. 3B) or higher concentrations (not shown) of Mn 2ϩ does stimulate cell adhesion, but in the same proportion for all cell types. These data indicate that uncleaved ␣ v ␤ 5 integrin present on PDX39P cells seems to be constitutively active, but it can be further activated by divalent ions as does cleaved integrin.
Cleavage Does Not Affect Binding of Soluble ␣ v ␤ 5 Integrins-Integrin-mediated cell adhesion may be altered by changes either in affinity of individual receptors for ligand or in integrin avidity (32). It has been proposed that integrin interaction with cytoskeletal proteins that serve to anchor and cluster integrins regulates receptor avidity. We therefore used a solid-phase binding assay to test the interaction of the ␣ v ␤ 5 integrins with vitronectin in a cell-free environment. In these conditions, we observed no significant difference of binding between the solubilized ␣ v ␤ 5 integrins from PDX0 and those from PDX39P cells (Fig. 4). The interaction integrin/vitronectin was RGD-and divalent cation-dependent and was enhanced to the same extent in the presence of Mn 2ϩ . These results suggest that the reduced adhesion of PDX39P cells to vitronectin was not due to changes in the interaction of the integrins with their ligand.
Integrin Signaling Is Impaired in PDX39P Cells-Previous studies on the role of cleavage on integrin function have focused on ligand binding (3,6,7). However, cell adhesion to ECM results in the integrin-dependent activation of a number of signaling pathways that can be affected by cleavage. We therefore examined whether ␣ 1 -PDX-expressing cells were able to elicit intracellular signal transduction events upon ␣ v ␤ 5 integrin ligation. Protein tyrosine phosphorylation is one of the earliest events detected in response to cell attachment to an ECM protein-coated surface. Adhesion of PDX0 cells to vitronectin led to a time-dependent increase in tyrosine phosphorylation of proteins in the molecular mass range of 110 -125 kDa and around 70 kDa, whereas no enhanced phosphorylation could be observed with PDX39P cells (data not shown).
In most cell types, the FAK accounts for a large proportion of the tyrosine phosphorylation in the 110 -125-kDa region. Therefore, PDX0 and PDX39P cells were plated on vitronectincoated dishes, and the phosphorylation status of FAK was assessed after immunoprecipitation. We observed a time-dependent increase in tyrosine phosphorylation of FAK from PDX0 cells after adhesion on vitronectin (Fig. 5A). In PDX39P cells in suspension, the FAK phosphorylation level was very low compared with control cells and did not enhance in vitronectin-adherent cells. We repeated the experiment with an antibody against paxillin, a FAK-associated phosphoprotein. As observed with FAK, paxillin from PDX0 cells was tyrosine-phosphorylated in response to integrin ligation, while only a very slight phosphorylation could be detected in PDX39P cells (Fig. 5B). The absence of tyrosine phosphorylation of FAK in PDX39P cells could be due to a defect in the FAK signaling pathway. However, as shown in Fig. 5C, FAK could still be tyrosine phosphorylated after stimulation of PDX39P cells by Proteins from plasma membranes were solubilized with a buffer containing 1% n-octyl-␤-D-glucopyranoside and incubated overnight at 4°C in plates coated with 10 g/ml vitronectin in the absence (control) or the presence of the synthetic peptides RGDS or RGES (1 mM), EDTA (10 mM) or MnCl 2 (1 mM). Binding of the ␣ v ␤ 5 integrin to vitronectin was quantified by spectrophotometry using a mAb against ␣ v subunit as described under "Materials and Methods." Specific absorbance (total absorbance minus absorbance in the absence of anti-␣ v antibody) was calculated for each condition. Data represent the mean Ϯ S.D., from one experiment representative of six performed in triplicate, and are expressed as ␣ v binding relative to that obtained with PDX0 cells without addition.
FIG. 5. ␣ v ␤ 5 integrin-dependent FAK signaling in PDX39P cells. Serum-starved cells were harvested in single cell suspension, added to wells coated with 10 g/ml vitronectin, and allowed to adhere for the indicated times (A and B). In C, adherent cells were incubated at 37°C in the absence or in the presence of 10 nM neurotensin (Nt) or 1 M nocodazole (Nz) for 30 min and 2 h, respectively. Proteins from attach ed cells were submitted to immunoprecipitation by anti-FAK (A and C) or anti-paxillin (B) antibodies, and immunoprecipitated proteins were resolved by SDS-PAGE. The phosphorylation status of FAK or paxillin was assessed by incubating the nitrocellulose membrane with the anti-Tyr(P) antibody PY20 (upper panels). After stripping, the membrane was reprobed with anti-FAK or anti-paxillin antibodies to determine the amount of immunoprecipitated protein (lower panels).
neuropeptides, such as neurotensin, or by treatment with the microtubule-disrupting agent nocodazole.
Because MAPK cascade is an important signaling pathway activated by integrins, we analyzed MAPK activation in response to cell adhesion using an antibody raised against the dually phosphorylated region within the catalytic core of the active form of ERK1 and ERK2. As illustrated in Fig. 6A, adhesion of PDX0 cells to vitronectin resulted in the activation of ERK2, while ERK1 was barely activated. In the case of PDX39P, ERKs are already activated in suspended cells and adhesion to vitronectin resulted in an additional increase in activation of ERK2. On the contrary, activation of ERK1 decreases with time of adhesion.
A recent report demonstrated that activation of MAPK might suppress the activation of integrins and thus negatively regulate cell adhesion and motility (33). Thus, it is possible that the reduced attachment of PDX39P cells to vitronectin could be due to the higher activity of MAPK in these cells. We therefore measured cell adhesion when MAPK is activated by epidermal growth factor or, on the contrary, when its activation is impaired by PD98059. As illustrated in Fig. 6B, we observed no difference in cell adhesion (upper panel), whatever the activation status of MAPK (lower panels). Thus, neither impediment of MAPK activation in PDX39P cells, nor activation of MAPK in PDX0 cells had any effect on adhesion to vitronectin (Fig.  6B). DISCUSSION In the present work we have examined the importance of ␣ subunit cleavage for integrin function. We have generated stable transfectants (PDX39P cells) expressing high levels of ␣ 1 -PDX, a potent and selective inhibitor of convertases and especially of furin and PC5B (14). The ␣ 1 -PDX-expressing cells displayed plasma membrane integrins mainly under an uncleaved form, although a minor cleavage could be observed. This result confirms that the convertase family of serine proteases is involved in the post-translational processing of integrins ␣ chains, as we reported previously (22,23).
Endoproteolytic cleavage is often required to generate active proteins from inactive precursors. Here we report that PDX39P cells, which display uncleaved integrins on cell surface, showed a reduced adhesion to vitronectin through the ␣ v ␤ 5 integrin. However, we found no evidence that cleavage of the ␣ subunits has any major effect on cell adhesion to other ECM proteins, such as laminin-1 and collagen type I. This absence of effect is likely due to the involvement of noncleavable ␤ 1 integrins, such as ␣ 1 ␤ 1 or ␣ 2 ␤ 1 , which support HT29-D4 cell adhesion to these ECM proteins (26,28). We neither observed any difference between cells when using the integrin-independent substrate poly-L-lysine.
The reduced adhesion of PDX39P cells to vitronectin was not due to changes in ␣ v ␤ 5 expression, as neither transfection nor PEA selection altered the amount of integrin subunits at the cell surface. For several reasons, abolishing ␣ subunit cleavage neither seems to cause important alterations in integrin structure. (i) ␣/␤ association still occurred because heterodimers containing the uncleaved ␣ 3 , ␣ 6 , or ␣ v subunits were expressed on the cell surface, as also reported for the furin-deficient cell line LoVo (22) and mutated ␣ IIb and ␣ 4 subunits (3, 6). (ii) PDX39P cells attached with the same efficiency than parental cells or empty vector transfected cells to two mAbs (69.6.5 and AMF-7) recognizing distinct ␣ v epitopes (not shown). (iii) The function-blocking mAb 69.6.5 inhibited adhesion to vitronectin with similar dose-effect profiles whatever the cell type (not shown).
The integrin adhesive function can be regulated by intracellular signals, a phenomenon known as inside-out signaling (reviewed in Refs. 34 and 35). This can be done either by altering the affinity of the individual integrins or by integrin clustering, which allows more efficient interaction and increased binding between cells and ECM. The reduced adhesion of PDX39P cells to vitronectin is most likely not due to changes in integrin affinity, as we noticed no significant difference of interaction between cleaved and uncleaved integrins in a solidphase receptor assay. This result is in agreement with previous reports showing that the absence of cleavage of the ␣ 4 and ␣ IIb subunits did not affect ligand binding of the ␣ 4 ␤ 1 and ␣ IIb ␤ 3 integrins (3,6). However, these studies have no examined the role of integrin cleavage on intracellular signaling pathways. In the present study we show that the phosphorylation level of FAK and its associated phosphoprotein paxillin, in response to ligation of ␣ v ␤ 5 integrin, was very low in ␣ 1 -PDX-expressing cells.
Various signaling proteins have been involved in the insideout modulation of integrin function. Thus, FAK phosphorylation induced by integrin ligation appears to contribute to stabilization of cell adhesion (36). Such a positive feedback loop is also thought to occur in the context of protein kinase C activation during adhesion and spreading (37). It thus can be hypothesized that the absence of cleavage of the ␣ v ␤ 5 integrin leads to the impairment of signal transduction by FAK (or another molecule) upon cell adhesion, that in turn might result in a reduced efficiency of PDX39P cells attachment to vitronectin. In support of this hypothesis, Delwel et al. (7) have shown that the cleavage of the ␣ 6 A subunit is essential for activation of the integrin by the phorbol ester PMA, a protein kinase C activator.
MAPKs (ERK1 and ERK2) are activated in response to a variety of extracellular signals and thus represent a conver-FIG. 6. Activation of MAPK pathway during cell adhesion to vitronectin. A, serum-starved cells were harvested in single cell suspension and allowed to adhere for the indicated periods of time to wells coated with 10 g/ml vitronectin. Attached cells were lysed and 10 g of proteins were resolved by SDS-PAGE. The nitrocellulose membrane was probed with an anti-active MAPK antibody (upper panel). After stripping, the membrane was reprobed with anti-MAPK antibodies (lower panel). B, serum-starved cells were pretreated or not for 90 min with 10 M PD98059 before performing the cell adhesion assay in wells coated with 10 g/ml vitronectin in the absence or the presence of 10 ng/ml epidermal growth factor. At the end of the adhesion assay (2 h at 37°C), the activation status of MAPK was checked as described above. Results are representative experiments of three performed. gence point for many signaling pathways. It is now clear that integrin-mediated cell adhesion can also lead to the activation of the MAPK cascade (for review, see Ref. 1). In this work we show that ERK2 and, to a lesser extent, ERK1 are activated upon interaction of the ␣ v ␤ 5 integrin with vitronectin. We also reported that ERK2 was already activated in PDX39P cells kept in suspension and that the enzyme was further activated during adhesion. There is increasing evidence that suppression of integrin activation may be a physiological mechanism to control integrin-dependent cell adhesion and migration. The existence of signaling pathways acting as inhibitors of integrin activation has been proposed, and recently Ras/Raf-1-initiated activation of MAPK pathway has been shown to suppress the activation of ␤ 1 and ␤ 3 integrins (33). Thus, one possibility is that the reduced adhesion of PDX39P cells to vitronectin was the result of the high activity level of MAPK in these cells. However, this is likely not the case, because neither MAPK activation nor impediment of MAPK activation had any effect on adhesion to vitronectin. An alternative is the suppression of the constitutive integrin activation due to the absence of ␣ subunits cleavage. Nevertheless, the integrin ␣ v ␤ 5 was already present as an active form in ␣ 1 -PDX-expressing cells, as neither PMA nor the divalent cation Mn 2ϩ , two integrin activators, could restore a normal attachment of ␣ 1 -PDX-expressing cells on vitronectin.
It thus appears that the absence of endoproteolytic cleavage of the ␣ v subunit affects integrin function by altering ␣ v ␤ 5 -dependent signaling pathways. The molecular mechanism responsible of this dysfunction remains to be determined.