Keratinocyte migration requires α2β1 integrin-mediated interaction with the laminin 5 γ2 chain

Interactions between cells and the extracellular matrix (ECM) are critical for many biological events and involve cell membrane receptors of the integrin family. Integrins are transmembrane receptors implicated in mediating cell-substratum attachment or cell-cell adhesion, and in transducing signals that regulate such diverse processes as growth, differentiation and migration.

In the epidermis of the skin, basal keratinocytes adhere to the basement membrane through integrins. Human keratinocytes express a number of integrins, including α2β, α3β and α6β restricted to the basal and suprabasal, proliferative cell layer, both in the epidermis and in stratified culture of keratinocytes (Carter et al., 1990a; Hertle et al., 1991; Marchisio, 1991). In the epidermis, α6β integrin is a component of hemidesmosomes (HD) (for a review, see Nievers et al., 1999). It mediates keratinocyte adhesion to the basement membrane by binding to laminin 5; this interaction is necessary for HD stability (Langhofer et al., 1993; Ryan et al., 1999). The α2β and α6β integrins are localized at the lateral plasma membranes and have been proposed to cooperate in cell-cell interactions (Kaufmann et al., 1989; Carter et al., 1990b; Symington et al., 1993).

According to tissue needs, the strength of attachment is subject to up- or downregulation during, for example, commitment to terminal differentiation or epidermal regeneration. In these situations basal cells detach from the basement membrane and migrate either into the suprabasal layers (Adams and Watt, 1990) or laterally (Larjava et al., 1993). These processes appear to involve changes both in the ligand-binding activities of certain integrins and in the expression of integrins on the cell surface. During wound healing, migration starts within hours after injury and continues until the epithelial surface is intact so as to finally regenerate the basement membrane (reviewed in Clark, 1996). Keratinocytes from the edges of the wound and within skin appendage structures migrate over a provisional matrix rich in newly synthesized fibronectin and laminin 5 (Larjava et al., 1993). Integrin polarization is lost and keratinocytes enlarge their integrin repertoire and modify their distribution. The appearance of the α6β integrin and a strong increase in the αvβ5 integrin are the major events occurring in the migratory region of the wound (Cavani et al., 1993; Juhasz et al., 1993; Larjava et al., 1993). Integrins α6β4 and α2β move from a lateral to a basal location in contact with ECM. The α6β4 integrin becomes more evenly distributed and appears along the entire cell surface of migrating keratinocytes in which HD are absent (Kurpakus et al., 1991).

Although the role of fibronectin in epidermal wound healing has been clearly demonstrated (for a review, see Humphries, 1991), it is certainly not the only ECM protein involved in this process. Dermal collagen I has been proposed to play a crucial role in keratinocyte migration by mediating induction of collagenase-1 through interaction with α2β integrins (Pilcher et al., 1997). The resulting cleavage of collagen I by collagenase-1 was proposed to be needed to initiate cell movement (Pilcher et al., 1997). Among basement membrane components involved in cellular interactions, laminin 5 is likely to play a role during cell migration. Laminin 5 is the major component of anchoring filaments of the skin and is a composed of the heterotrimeric association of the α3, β3 and γ2 laminin chains (Rousselle et al., 1991). It supports cell adhesion via interaction of its α? C-terminal extremity, called G-domain, with both the α6β4 and α6β4 integrins (Carter et al., 1991; Sonnenberg et al., 1993; Rousselle and Aumailley, 1994; Mizushima et al., 1997). Laminin 5 is synthesized as a 460 kDa precursor that undergoes specific processing to smaller forms of 440 and 400 kDa rapidly after secretion (Marinkovich et al., 1992). The size reduction is a result of a carboxy-terminal truncation of the 200 kDa α3 chain into a 165 kDa polypeptide (Goldfinger et al., 1998), and by shortening the 155 kDa γ2 chain at the amino terminus into a 105 kDa polypeptide (Vailly et al., 1994). A larger truncation has been reported for the γ2 chain synthesized by a rat tumour cell line (Giannelli et al., 1997).

Immunohistochemical and in situ hybridization studies have shown induction of laminin 5 expression in migrating keratinocytes during wound healing (Kurpakus et al., 1991; Larjava et al., 1993; Ryan et al., 1994; Lampe et al., 1998; Salo et al., 1999) or in malignant cells located at the invasion front of human carcinomas (Pyke et al., 1994). Several lines of evidence indicate that efficient migration of normal human keratinocytes (NHK) is dependent on endogenously deposited laminin 5 (Zhang and Kramer, 1996; Qin and Kurpakus, 1998; Goldfinger et al., 1999; Nguyen et al., 2000). Moreover, studies indicate that laminin 5 containing the 200 kDa α? chain is capable of inducing cell migration while impeding hemidesmosome assembly (Goldfinger et al., 1998; Goldfinger et al., 1999). Other studies provided evidence that processed-α3 containing-laminin 5 function in the nucleation of hemidesmosome assembly and as an adhesive factor that retards cell motility (Baker et al., 1996; Langhofer et al., 1993; O’Toole et al., 1997; Goldfinger et al., 1998).

This regulation, as well as the up- and downregulation of the strength of attachment occurring during cell migration, is likely to be modulated by external mediators such as growth factors and cytokines. A cytokine with a broad spectrum of effects in cell-matrix interactions is the transforming growth factor-β (TGF-β) peptide family (for a review, see Massague, 1990). In addition to controlling cell growth, these molecules play an important role in the remodelling of tissue by regulating the transcription of a wide spectrum of ECM proteins, increasing their production while decreasing their proteolysis and modulating their interactions with cellular integrin receptors. Moreover, TGF-β promotes the migration of the majority of cell types that participate in repair processes. Indeed, TGF-β1 has the ability to induce keratinocyte migration in vitro by regulating integrin expression (Zambruno et al., 1995). Moreover, it was shown to induce upregulation of the laminin α3 mRNA (Korang et al., 1995; Virolle et al., 1998).

In this study, we attempt to understand the function of laminin 5 during NHK migration. Since NHK have no tendency to migrate during normal culture conditions, we have constrained them to migrate by exposing them to TGF-β1. In order to exclude any artifact caused by the use of TGF-β1, we have verified our findings with untreated cells. We show that migration is accompanied by a significant decrease in integrin-mediated adhesion to laminin 5. Moreover, we provide evidence that the α6β4 integrin does not promote migration while the α2β integrin is absolutely necessary. Finally, we demonstrate that the α2β integrin binds to the short arm of the laminin 5 γ2 subunit.

NHK cultures were established from foreskin as previously described (Rousselle et al., 1991) and subcultured in Keratinocyte Serum Free Medium K-SFM (Gibco BRL, Cergy Pontoise, France). Cells were harvested for subculturing or for subsequent experiments using 0.05% trypsin and 0.02% EDTA in phosphate-buffered saline, pH 7.4 (PBS). NHK were used between passages 1-3. Cell culture reagents were from Seromed-Biochrom (Polylabo, Strasbourg, France) and plasticware was from Falcon (Dutscher, Brumath, France). NHK were harvested at a cell density of 1×10⁵ cells/cm², and after reaching subconfluency they were incubated in SFM for 24 hours in the presence or absence of 10 ng/ml of human platelet-derived purified TGF-β1 (R&D Systems, Abingdon, UK).

After the cells were grown in the presence or absence of TGF-β1, the medium was removed and the cells were washed in sterile PBS. The cells were removed by treating them for 5 minutes in sterile 20 mM NH₄OH (Gospodarowicz, 1984). The matrix was washed three times in sterile water and was then removed from the substratum by solubilization in 0.325 M Tris-HCl, pH 6.9, 25% glycerol, 10% SDS with or without 5% β-mercaptoethanol, for electrophoretic analysis.

Human plasma fibronectin and mouse EHS collagen IV were purchased from Becton Dickinson (Le Pont de Claix, France). EHS laminin 1 was graciously provided by Prof. Aumailley and human laminin 5 was purified from the culture medium of human SCC25 cells as previously described (Rousselle and Aumailley, 1994). Characterization of laminin 5 monoclonal antibody (mAb) BM165 (anti-α3chain) and polyclonal antibody (pAb) 4101 was described elsewhere (Rousselle et al., 1991; Marinkovich et al., 1992) and pAb L132 was raised against purified native laminin 5. mAb D4B5 against domain III of the γ2 chain was purchased from Chemicon (Euromedex, Souffelweyersheim, France). pAbs to thrombospondin (Ab-8) and collagen XVII (L 766P) were provided by Dr Clezardin (Inserm U403, Lyon, France) and Dr Meneguzzi (Inserm U385, Nice, France), respectively. mAb (T2H5) to tenascin was purchased from Chemicon and mAb NP32 to the NC1 domain of type VII collagen was provided by Prof. Burgeson (CBRC, Charlestown, USA). The following function-blocking mAbs against integrin subunits were used: P4C10 against β1, P1E6 against α2, P1B5 against α3, P1D6 against α5, all from Telios (Gibco BRL); CS-1 against α3 from Chemicon and GoH3 against α6 (Genosys, Becton Dickinson). Nonfunction-blocking mAb A3×8 to the α3 integrin subunit was graciously provided by Dr Hemler (Dana Farber Cancer Institute, Boston, USA).

Multiwell tissue culture plates (96-wells, Costar, Dutscher) were coated with the indicated concentrations of laminin 5, fibronectin and collagen IV substrates by overnight adsorption at 4°C. After saturation of the wells with 1% BSA (fraction V, Sigma Chimie, France), the plates were immediately used for cell adhesion assays (3×?? ⁵ cells/well) in serum-free medium as detailed previously (Rousselle and Aumailley, 1994). The extent of adhesion was determined after fixation of the adherent cells with 1% glutaraldehyde in PBS, by staining with 0.1% Crystal Violet, and color reading at 570 nm with an ELISA reader (MR5000 Dynatech, Guernsey, Channel Islands). A blank value corresponding to BSA-coated wells was automatically subtracted. For all experiments, each assay point was determined in triplicate.

The lower face of the nucleopore filters (pore size: 8 μm) present in Transwell chambers (6 wells; Costar, through Dutscher) was coated or not with fibronectin, collagen IV or laminin 5 (10 μg/ml) and post-coated with BSA. Cells in monolayers were radioactively labeled by incubation with ³⁵S-methionine (NEN, Les Ulis, France) for 1 hour in methionine-free SFM. Labelled cells were seeded in the upper part of the chambers (3×10⁵ cells/well) in SFM with 0.2% BSA and allowed to migrate for 6 hours. For cell migration perturbation experiments with cycloheximide (Sigma) or various antibodies, suspended cells were mixed with cycloheximide (25 μg/ml) for 30 minutes at room temperature or dilutions of antibody prior to being seeded in the transwell. At the end of the assay, cells on the upper face of the nucleopore filters were manually removed except for a set of filters used for the determination of total cell number. Radioactivity associated with the filters was measured in a scintillation counter. The number of cells that had migrated was expressed as a percentage of radioactivity associated with the lower face versus both faces of the filters. Radioactivity associated with BSA-coated filters was subtracted from the values.

Sterile round glass coverslips deposited on the bottom of 24-well plates (Costar, Dutscher) were seeded with 1×10⁵ keratinocytes. After the cells reached subconfluency, they were incubated in SFM for 24 hours in the presence or absence of 10 ng/ml of TGF-β1. After washing the wells with PBS, the cells were fixed with 2% paraformaldehyde in PBS for 15 minutes, permeabilized with 0.2% Triton X-100 for 1 minute, rinsed several times with PBS, and incubated for 45 minutes with the first antibody. Cy3-conjugated second antibodies against mouse or rabbit immunoglobulins (Jackson, distributed through Immunotech) were applied for another 45 minutes. In the case of double immunolabelling, the procedure was followed with an additional 45 minutes incubation with pAb L132, followed by the appropriate FITC-conjugated second antibody (Biosys, Compiègne, France). Glass coverslips were mounted onto slides in FluoPrep mounting medium (BioMérieux, Marcy l’Etoile, France) and cells were observed by epifluorescence with an Axiophot Zeiss microscope.

Cell lysates containing α2β integrin were prepared from detergent extracts of either cultured human keratinocytes or fibroblasts. The cells were washed in PBS and solubilized in 50 mM Tris-HCl buffer, pH 7.4, containing 50 mM n-octyl-1-thio-β-D-glucopyranoside (Roche Diagnostics, Meylan, France) and 2 mM phenylmethylsulfonyl fluoride (PMSF). After 10 minutes, the extract was centrifuged at 14,000 g for 15 minutes and the supernantant was dialysed against 50 mM Tris-HCl buffer, pH 7.4, 0.2 M NaCl (TBS) containing 2 mM MgCl₂ and 1 mM MnCl₂. α2β integrin binding assays on immobilized laminins were performed as followed. Multiwell plates (96 wells, Greiner, Dutcher) were incubated with a concentration of laminin 1 or 5 varying from 20 to 100 μg/ml dissolved in 20 mM Na₂Co₃, pH 9.2, for 18 hours at 4°C. The wells were then blocked with 5 mg/ml BSA in TBS for 1 hour at room temperature. BSA was removed and the soluble cell lysate added for 2 hours at room temperature. After the wells were washed three times with TBS 0.02% Tween 20 containing 2 mM MgCl₂, 1 mM MnCl₂, they were exposed to a constant amount of diluted pAb against human α2β integrin (Chemicon International, Euromedex, Mundolsheim, France) followed by a typical enzyme-immunoassay reaction with peroxidase-conjugate of donkey specific antirabbit immunoglobulin (Jackson Immuno Research Laboratories, through Interchim, Montlucon, France) as second antibody and 2,2′-azino-bis(3-ethylbenthiazoline-6-sulfonic acid) as the chromogenic substrate. Color yields were determined at 405 nm in an ELISA reader. Negative control values, from wells in which the soluble ligands were omitted, were subtracted from the binding data. For inhibition assays, soluble cell lysates and immobilized proteins were incubated with inhibitors for 1 hour at room temperature before being exposed.

The 5-day-old-wounding skin biopsy was provided by the Centre des Grands Brûlés, Pavillon I-5, Hôpital Edouard Hérriot, Lyon (France). It was sampled from the back of a 2^nd degree burn patient subject to spontaneous wounding. The 5 mm² biopsy was taken during a dressing change. After a rapid wash in PBS, it was immediately frozen in Tissue-Tek OCT Compound (Sakura Finetek, Poly Labo) and stored at –70°C. Consecutive frozen sections of 8 μm thickness were prepared using a Reichert-Jung Leica cryostat at –20°C and placed on glass slides. Sections were then processed for indirect immunofluorescence as described above and finally incubated with Trypan Blue prior mounting with FluoPrep mounting medium.

Protein preparations were separated by SDS-PAGE using 5% or 3%-5% acrylamide gels. Proteins were either stained with Coomassie Brilliant Blue or transferred to nitrocellulose filters for western blotting according to standard procedures. Molecular mass markers were from Bio-Rad Laboratories (Ivry, France).

In agreement with previous reports (Boland et al., 1996; Zicha et al., 1999), we have found that TGF-β1 increased the cell motility in vitro. Cell migration assays, using Transwell chambers, demonstrated that NHK ability to migrate was greatly enhanced. We tested the ability of cells to migrate towards substrates such as laminin 5, fibronectin and collagen IV).Fibronectin was the best migrating substrate for NHK whilst there was no migration toward purified laminin 5 (Fig. 1). Although TGF-β1 treatment did not enhance cell migration towards purified laminin 5, it considerably increased cell migration towards fibronectin and collagen IV.

In order to elucidate cell behavior during the migration process, the cell adhesion properties of keratinocytes, treated for 24 hours with TGF-β1, to purified laminin 5, fibronectin and collagen IV were characterized. The adhesion of untreated cells to each of the three substrates was in accordance with that obtained previously and served as a control (Fig. 2A). Adhesion of TGF-β1-treated cells to purified laminin 5 was significantly decreased (Fig. 2B), while it was dramatically increased on fibronectin and collagen IV.

As shown by fluorescence analysis in Fig. 3A and as previously documented, laminin 5 is deposited into the matrix of cultured NHK and is present upon the substrate underneath and around the cells. As expected and shown in Fig. 3B, the TGF-β1-induced migration was accompanied by laminin 5 deposition, which when stained allowed visualization of the cell path. Therefore, the molecular form of laminin 5 present in the matrix of migrating TGF-β1 treated keratinocytes was analyzed by SDS-PAGE, and compared to that of untreated keratinocytes and purified laminin 5 (Fig. 4). After disulfide reduction, three major species were visualized by staining with Coomassie Blue in the ECM of confluent cultured keratinocytes (Fig. 4A, lane 2). These bands corresponded to those seen in the purified laminin 5 preparation (lane 1), confirming that the major component of cultured keratinocyte ECM is laminin 5. Analysis of the α? subunit by western blotting (Fig. 4B, lanes 2) revealed that most laminin 5 molecules present in normally cultured keratinocytes ECM contained the processed α3 chain (165 kDa), while faint staining of the 200 kDa form was seen. Staining of the γ2 subunit with D4B5 mAb revealed the presence of two bands, a minor band running in the position of 155 kDa and a major one in the position of 105 kDa corresponding to unprocessed and processed γ2 subunits, respectively (not shown). Indeed, nonreducing conditions (Fig. 4C) confirmed that the 400 kDa fully processed laminin 5 (α3,165; β3,140; γ2,105 kDa) was the major component of the ECM of untreated cells (lane 2). Although this form was predominant, the 460 and 440 kDa bands were also seen, confirming the presence of unprocessed laminin 5 in this ECM. As shown in lane 1 and as previously documented (Rousselle et al., 1991), purified laminin 5 contains both 440 and 400 kDa forms, corresponding to an equal mixture of the fully processed laminin 5 and the unprocessed γ2 subunit containing laminin 5. Analysis of the ECM of TGF-β1 treated keratinocytes under reducing conditions revealed the presence of three major bands running at approximately 200, 155 and 140 kDa (Fig. 4A, lane 3). Western blotting analysis of these bands revealed that the three Coomassie Blue stained bands were recognized by pAb 4101 (Fig. 4B, left panel, lane 3), suggesting that laminin 5 was expressed under different molecular forms. Blotting the reduced sample with mAb BM165 (Fig. 4B) identified the major 200 kDa band as the unprocessed α3 subunit while the 165 kDa form was minor (lane 2). Blotting reduced samples with the D4B5 mAb confirmed Coomassie Blue staining and revealed a single band running at 155 kDa, corresponding to the unprocessed γ2 chain in the matrix of migrating cells (not shown). Analysis of the material under nonreducing conditions (Fig. 4C, lane 3) revealed a major band running at the position of 460 kDa, showing that mainly unprocessed laminin 5 (α3, 200; β3, 140; γ2, 155 kDa) was found in the ECM of TGF-β1 induced migratory keratinocytes.

To search for the potential receptors involved in the interactions with newly synthesized laminin 5 during cell locomotion, cell migration experiments were performed in the presence of function blocking-antibodies (Fig. 5). Fibronectin or collagen IV were used as adhesion substrates on the lower face of the Transwell chambers. First of all, inhibition of protein synthesis was performed by using cycloheximide (Fig. 5A) and induced significant reduction of migration on both fibronectin (down to 30%) and collagen IV (down to 3%). Therefore, antibodies directed against laminin 5 (Fig. 5A), as well as antibodies against the various integrin subunits, were used as described in Materials and Methods. Migration experiments conducted on fibronectin or collagen IV (Fig. 5A) revealed that pAb L132, directed against laminin 5, significantly abrogated keratinocyte migration, indicating that laminin 5 is necessary for migration. Moreover, mAb D4B5 raised against domain III of the γ2 subunit (Mizushima et al., 1998) almost totally blocked migration (down to 25%). Interestingly, mAb BM165, directed against the major laminin 5 cell binding site, enhanced cell migration (up to 150%). Control experiments conducted with function-blocking mAbs to integrin subunits (Fig. 5B,C) revealed that, as expected, mAb P1D6 against the α? integrin subunit inhibited migration on fibronectin and mAb P1E6 against the α2 integrin subunit prevented cell migration on collagen IV. Interestingly, mAb P1B5 against the α3 integrin subunit enhanced cell migration on both substrates, in a manner comparable to that obtained with mAb BM165, while mAbs directed against the α6 integrin subunit (GoH3) only weakly affected cell migration. Of particular interest was the significant inhibition of migration caused by mAb P1E6 on fibronectin (down to 25%), suggesting involvement of the α2β integrin in keratinocyte migration. In order to verify that this observation was not an artifact of the TGF-β1 treatment, the experiment was also performed with untreated NHK on fibronectin (Fig. 5C). Despite their low ability to migrate, these cells produced similar results. In view of the results obtained with P1B5 mAb, a number of control experiments were performed with other mAbs to the α3 integrin subunit (Fig. 6). The mAb ASC-1, known to block the α6β4 integrin-mediated adhesion to laminin 5 (Lichtner et al., 1998), also enhanced migration, while the nonfunction-blocking mAb A3×8 (Weitzman et al., 1993) had no effect on migration. Morover, incubating cycloheximide-pretreated cells with the P1B5 mAb prevented the cell migration enhancement that was previously obtained.

In view of the results obtained in the cell migration assay, analysis of the keratinocyte α2β integrin expression was performed by indirect immunofluorescence before and after migration was induced by TGF-β1. Indeed, the staining of the α2 integrin subunit under TGF-β1 treatment (Fig. 7), present in focal contacts localized at the leading edge of migrating cells, confirmed a possible involvement of this integrin in the migration process. The α2β integrins were organized into thick and well-separated aligned patches located across lamellipodia (Fig. 6Aa,b). In the case of untreated keratinocytes and in addition to strong staining at cell-cell borders, these α2 integrin-containing patches were also found in lamellipodia of cells located at the periphery of the clones (Fig. 6Ac,d). When the focus of the microscope was adjusted to the matrix located in the immediate cell environment, α2 integrin subunit staining could then be detected as light fluorescent patches attached to the substrate left behind by the migrating cells (Fig. 7B). As double immunostaining with D4B5 mAb was not feasible, study of laminin 5 expression was performed with pAb 4101 (Fig. 7C) in the same experiment. It revealed an expression pattern similar to that of α2β integrin, corresponding to the ‘left behind’ track. Analysis of the matrix beneath the cells, after removal of the cells, revealed numerous and intense footprints positively stained with the α2 integrin subunit antibody (Fig. 7G,I), which appeared to colocalize in many places with the laminin 5 staining (Fig. 7F,H). This unusual colocalization pattern with the α2 integrin subunit could not be duplicated with antibodies to other components likely to be involved in this process such as tenascin, thrombospondin, collagen XVII, NC1 domain of collagen VII (see Materials and Methods, not shown). In the case of untreated clones, the leftover α2 integrin subunit staining remained at cell-cell borders (Fig. 7E). Concomitant analysis of α6 and α3 integrin subunits under TGF-β1 treatment also revealed, for both integrin subunits, a staining similar to that of laminin 5 (not shown).

As keratinocyte migration appeared to involve an α2β integrin interaction with laminin 5, and as the antilaminin γ2 subunit mAb D4B5 interfered with cell migration in a manner comparable to that of mAb P1E6, further experiments were performed in order to verify that these two phenomena were related. To test the potential involvement of the α2β integrin binding to the short arm of the laminin γ2 chain, solid-phase assay binding studies were performed using a soluble integrin preparation obtained from a keratinocyte lysate interacting with immobilized laminin 5 (Fig. 8). Since purified laminin 5 preparation is composed of both processed and unprocessed γ2-containing laminin 5 molecules (Fig. 4C), such substrate was used in this experiment. To confirm previous studies showing the direct interaction between α2β integrin and laminin 5 (Orian-Rousseau et al., 1998), the binding of α2β integrin to immobilized laminin 5 was measured in comparison to laminin 1 (Fig. 8A). The specificity of both molecular interactions was verified by the complete inhibition obtained with mAb P1E6 (Fig. 8B). Incubation of the immobilized ligand with mAb D4B5, against domain III of the laminin γ2 chain prior to the interaction, totally prevented binding of the α2β integrin to laminin 5, while mAb 6F12 against the laminin β3 chain did not. None of these mAbs affected the binding to laminin 1 (Fig. 8B).

To gain insight into the possible involvement of the α2β integrin in cell adhesion to laminin 5 during cell migration, cell adhesion experiments using normal keratinocytes and TGF-β1 induced migrating keratinocytes were carried out with anti-integrin-specific antibodies including mAb P1E6 and D4B5 (Fig. 9). In a manner comparable to that obtained previously with NHK (Rousselle and Aumailley, 1994), adhesion of TGF-β1 treated keratinocytes to laminin 5 was totally abrogated by mAbs BM165, P4C10 and P1B5 while GoH3 had no effect (Fig. 9A). Preincubation of the substrate with mAb D4B5 prior to the cell adhesion experiment was not inhibitory and use of mAb P1E6 produced a substantial disturbance of cell adhesion as compared to the control. Analysis of cell spreading by phase-contrast microscopy and measurement of cell diameters (Fig. 9B) revealed that mAb P1E6, and D4B5 to a lesser extent, was inhibitory of cell spreading rather than cell adhesion itself. The effect of both mAbs on cell spreading on laminin 5 was minimal in the case of untreated keratinocytes. Indeed, mAb P1E6 impeded cell spreading either on laminin 5 or on the ECM of TGF-β1-treated keratinocytes (not shown) as compared to the controls without antibodies.

To understand better the possible involvement of the α2β?integrin in keratinocyte migration in vivo, human cutaneous wound biopsies from a full-thickness burn patient were assessed by immunolabeling of laminin 5 and various integrin subunits. Our findings were in accordance with earlier studies demonstrating an increase of laminin 5 expression and deposition by cells located at the regenerating epidermal margin of skin wounds (Larjava et al., 1993; Ryan et al., 1994; Kainulainen et al., 1998) (not shown). Analysis of integrin subunits revealed that migrating cells present at the leading edge stained brightly with mAb P1E6 (Fig. 10B), suggesting that these cells were actively expressing the α2β integrin possibly involved in cell-matrix interactions. Staining domain III of the laminin γ2 subunit with D4B5 mAb on serial section revealed a similar localization (Fig. 10C). Observation at higher magnification shows strong intracellular and extracellular staining at the migrating front, for both antigens (Fig. 10E,F).

In this study, we have analyzed the role of laminin 5, newly synthesized and secreted by migrating keratinocytes. Since NHK do not tend to move in normal culture conditions, but on the contrary, can form stable hemidesmosomal-like structures, we have induced them to migrate by exposure to TGF-β1. Moreover, since TGF-β1 prevents NHK proliferation (reviewed in Moses et al., 1990), this important feature permitted the study to be focused on cell migration without interference from cell proliferation. In agreement with studies conducted with other epithelial cells (Miettinen et al., 1994; Boland et al., 1996), the data presented here demonstrate that TGF-β1 induced significant increase in NHK migration when cells were placed on either fibronectin or collagen IV. Absence of migration on purified laminin 5, whether cells were TGF-β1-treated or not, confirms previous studies showing that processed-α? containing laminin 5 does not support migration of keratinocytes or other epithelial cell lines such as SCC12 and MCF10A (O’Toole et al., 1997; Giannelli et al., 1997; Goldfinger et al., 1998). In contrast, this substrate has been shown to support migration of a number of cell lines including glioma or HT1080 fibrosarcoma cells (Fukushima et al., 1998; Hirosaki et al., 2000; Koshikawa et al., 2000). Recent data indicates that the ability for these particular cells to constitutively migrate on laminin 5 is linked to their capability to specifically express the plasma-membrane-bound MT1-MMP metalloprotease and cleave the γ2 chain to an 80 kDa peptide (Koshikawa et al., 2000). This mechanism is likely to be similar to that observed in scattering of these cells by the addition of soluble laminin 5 (Miyasaki et al., 1993; Kikkawa et al., 1994; Koshikawa et al., 2000). Adding soluble laminin 5 to NHK does not have a scattering effect (O’Toole et al., 1997). On the contrary, it was shown to inhibit collagen I-driven motility migration (O’Toole et al., 1997). It is likely that this process is strictly dependent on the cell type and that keratinocyte migration may require a different type of interaction with laminin 5.

Our cell adhesion studies, showing dramatic enhancement of cell adhesion to fibronectin and collagen IV with TGF-β1 treatment, reveal a change in integrin expression similar to that previously documented (Sheppard et al., 1992; Gailit et al., 1994). Of novel interest are our cell adhesion data reporting a significant decrease in cell adhesion to purified laminin 5 (down to 40%) when cells were forced to migrate, suggesting an absolute necessity for the cells to decrease their adhesion to processed laminin 5 in order to migrate. In addition, our results corroborate previous observations describing a TGF-β1 induced decrease in α6β4 integrin expression in various human cell lines (Heino and Massague, 1989; Kumar et al., 1995) and in keratinocytes (Tennenbaum et al., 1996; Zambruno et al., 1995). Despite its decreased adhesion-promoting activity for TGF-β1-induced migrating keratinocytes, laminin 5 synthesis has, however, been shown to be upregulated by TGF-β1 (Korang et al., 1995), and more generally so during the in vivo epidermal cell migration processes (Larjava et al., 1993; Ryan et al., 1994; Kainulainen et al., 1998; Salo et al., 1999). Moreover, endogenous laminin 5 has been proposed to play a crucial role in keratinocyte migration in vitro (Zhang and Kramer, 1996). Our immunohistochemical and electrophoretic analysis of the ECM beneath TGF-β1 induced migrating cells revealed that its main component is laminin 5. Under these conditions, we demonstrate that laminin 5 is mainly present in its unprocessed form of 460 kDa, suggesting that processing of both α? and γ2 chain is prevented. In contrast, the predominant molecular form of laminin 5 present in the matrix of confluent keratinocytes is the fully processed form, confirming that, in normal conditions, laminin 5 is rapidly processed after secretion. In the case of NHK, various amounts of the unprocessed form were found, depending on the level of cell confluency (P. Rousselle, unpublished observation). Our observations are in agreement with a recent report showing a migratory promoting activity of unprocessed α?-containing laminin 5 (Goldfinger et al., 1998). In addition, our results suggest that the lack of cleavage of the γ2 subunit may also be associated with keratinocyte migration.

Our cell migration inhibition experiment after either cycloheximide treatment or use of the pAb L132 against laminin 5, showed that laminin 5 deposition appears to be critical for cell migration. Of particular interest, our results showed enhancement of migration when cells were incubated with the mAb P1B5 directed against the α3 integrin subunit. The use of another mAb known to block α6β4 integrin adhesion to laminin 5 (ASC-1, Lichtner et al., 1998) also induced migration enhancement, indicating that the α6β4 integrin does not promote migration but, on the contrary, it impedes this process and renders the cells more adherent. Our results showed absence of migration enhancement with the noninhibitory mAb A3×8 (Weitzman et al., 1993), as well as the absence of migration enhancement by the P1B5 mAb when cells were pretreated with cyclohemide. These data strongly suggest that enhanced migration is a result of an inhibition of the α6β4 integrin adhesion to laminin 5 rather than an activation caused by the P1B5 antibody. Similar observations were reported in previous studies conducted with lung carcinoma cells (Barr et al., 1998) and NHK (Kim et al., 1992; O’Toole et al., 1997), showing that a decrease in α6β4 integrin expression or a blocking of its function was associated with an increased cell migration ability. In vivo findings support this hypothesis by showing that a reduction of α6β4 is associated with the increased virulence of epithelial cancers (Adachi et al., 1998; Bartolazzi et al., 1995; Koukoulis et al., 1991). In accordance with O’Toole et al., 1998, our results show that preincubation of the cells with mAb BM165 resulted in an enhancement of their migratory capacity in a manner comparable to that obtained with the mAb P1B5. These data strengthen the hypothesis that the mAb BM165 epitope on laminin 5 is involved in stable cell adhesion and maintenance of HDs. As interactions with both α6β4 and α6β4 integrins are thought to occur with sites on laminin 5 corresponding to the mAb BM165 epitope (Rousselle and Aumailley, 1994), it was not clear which integrin was blocked when mAb BM165 was used. Since function-blocking antibodies to the α6 integrin subunits produced weak perturbation of cell migration, we propose that the α6β4 interaction with newly synthesized laminin 5 strongly impedes keratinocyte migration. The latter suggestion is reinforced by recent work from Sterk and coworkers, suggesting that α6β4 integrin-CD151 clusters may precede and facilitate the formation of HDs (Sterk et al., 2000). However, our findings contradict data presented by others, suggesting a crucial role for the interaction between the α6β4 integrin and the α? chain of laminin 5 during migration (Zhang and Kramer, 1997; Goldfinger et al., 1999). It is conceivable that, in the vitro wound-closure models presented in both studies, antibodies to either laminin 5 or α6β4 integrins inhibited cell migration by interfering with cell adhesion. Since laminin 5 is the main component synthesized and deposited by migrating cells, absence of a provisional matrix or another substrate for the cells to attach to may be the reason for the observed inhibition. This may also be the case in the experiments in which authors performed their cell migration assays on immobilized laminin 5 (Fukushima et al., 1998).

In accordance with our results, in vitro studies conducted with α6β4 knock-out keratinocytes indicate that the α6β4 integrin may trans-dominantly inhibit fibronectin and collagen IV receptor functions, confirming that α6β4 integrin does not facilitate migration processes on these substrates (Hodivala-Dilke et al., 1998). Moreover, data obtained using breast carcinoma cell lines propose a mechanism involving a possible α6β4 integrin-negative trans-regulation of the α2β integrin (Lichter et al., 1998). Furthermore, there is growing evidence suggesting that the α6β4 integrin may have a role to play in ECM assembly by establishing and/or maintaining integrity of the basement membrane (DiPersio et al., 1997; Wu et al., 1995). However, and as our model is restricted to the study of migrating cells in the Transwell chambers system only, we cannot exclude the possibility that the α6β4 integrin may actively participate in the re-epithelialization process by mediating different and cooperative signaling pathways.

Interestingly, our results demonstrate a role for the α2β integrin during keratinocyte migration whether the cells are TGF-β1 treated or not. Our experiments showing inhibition of keratinocyte migration by mAb D4B5 directed against domain III of the γ2 subunit (Mizushima et al., 1998), in a manner comparable to that obtained with mAb P1E6 against the α2 integrin subunit, were a primary indication of a possible relationship between these two events. A crucial role for the γ2 subunit in cell migration was also proposed in a study showing that a pAb to domain III was able to block cell migration of mouse keratinocytes (Salo et al., 1999). Other studies reported a mechanism by which a specific and unique cleavage of the γ2 subunit by matrix metalloprotease-2 (MMP2), or Membrane Type1-MMP, induced migration of various epithelial cell types on purified laminin 5 (Giannelli et al., 1997; Koshikawa et al., 2000). The mechanism occurring in our experiments is likely to be different since we never detected the migration-related 80 kDa cleaved γ2 fragment in the ECM of TGFβ1-treated migrating keratinocytes. Furthermore, our immunolabelling experiments showing perfect colocalization of the α2β integrin with laminin 5 in tracks left behind by migrating cells, strongly suggested that the α2β integrin was likely to bind the unprocessed form of laminin 5. We also detected α2β integrin-containing-focal contacts in peripheral lamellipodia of untreated cells located at the edges of growing clones. Laminin 5 was shown to be actively synthesized by growing keratinocytes (Rousselle et al., 1991) and recent data reported identification of the unprocessed α? chain at the periphery of growing epithelial clones (Goldfinger et al., 1999). Since processing of the γ2 chain was shown to occur at a slower rate than that of the α? chain (Marinkovich et al., 1992), it is conceivable that the γ2 chain may also be unprocessed in this precise location. This would suggest that the α2β integrin may transiently interact with unprocessed laminin 5 after its secretion. Previous studies (Orian-Rousseau et al., 1998) reported an interaction between laminin 5 and purified α2β integrin. Our solid-phase interaction studies demonstrating that solubilized α2β integrin binding to purified laminin 5 is inhibited by mAb D4B5, show for the first time that the α2β integrin binds to the short arm of the γ2 subunit in a specific manner.

The α2β integrins were originally identified as major receptors for several collagen types (Elices and Hemler, 1989; Languino et al., 1989; Kirchhofer et al., 1990), but they have also been shown to bind to laminin 1, although with lower efficiency (Pfaff et al., 1994). Integrin α2β binding to laminin 1 was originally mapped to fragment E1XNd, corresponding to the N-terminal portion of the molecule (Pfaff et al., 1994). Recent studies proposed an interaction with the N-terminal globular domain VI of the laminin α? chain (Ettner et al., 1998) as well as of the laminin α2 chain (Colognato et al., 1997). Our results suggest that binding of the α2β integrin occurs within the γ2 chain of laminin 5. In that regard, recent work reported identification of an α2β integrin binding sequence within domain IV of the γ1 subunit in a region with high homology with the γ2 subunit (Nomizu et al., 1997). We can speculate that the α2β integrin binding site on the γ2 chain may be localized within the N-terminal portion, which is removed after proteolytic cleavage during maturation of the molecule (Vailly et al., 1994; Amano et al., 2000). This hypothesis is strengthened by our results showing loss of the α2β integrin colocalization with processed laminin 5 in the ECM confluent keratinocytes (Fig. 7D,E). However, we cannot confirm this suggestion since the mAb D4B5 epitope is localized within the portion of domain III that remains on the γ2 subunit after cleavage (Mizushima et al., 1998). Inhibition of migration could then be explained by steric hindrance caused by the D4B5 mAb. Further studies are currently being conducted to resolve this question.

Other studies have proposed a role for the α2β integrin in the migration process, with absence of an obvious adhesive function (Lochter et al., 1999), suggesting a role for these integrin-ECM interactions in the regulation of matrix metalloproteinase expression. Our adhesion experiments propose a role for the α2β integrin in keratinocytes spreading on purified laminin 5. Indeed we have shown that incubation of TGF-β1 treated keratinocytes with mAb P1E6 prior to cell adhesion to purified laminin 5 or unprocessed laminin 5-containing matrices did not affect adhesion but strongly impeded spreading. Why the function of the α2β integrin in keratinocytes spreading on laminin 5 is seen only when the cells have been exposed to TGF-β1 remains to be elucidated. As previously suggested (Hodivala-Dilke et al., 1998; Lichtner et al., 1998), the α6β4 integrin may trans-dominantly restrict α2β integrin function, and it is conceivable that the downregulated expression of the α6β4 integrin by TGF-β1 may allow the α2β integrin to interact with the laminin γ2 chain. We can speculate that, when cells become stationary, the involvement of the α2β integrin in migration may be abolished upon re-expression of the α6β4 integrin.

The crucial participation of the α2β integrin was shown for initiation of keratinocyte migration over dermal type I collagen (Pilcher et al., 1997). A recent study using a similar model has provided evidence that keratinocytes deposit laminin 5 while migrating over collagen I, and that two different cell populations can be distinguished at the wound margin, depending on their α2β integrin-mediated binding to collagen I or their α6β4 integrin-mediated binding to deposits of laminin 5 (Nguyen et al., 2000). While our approach does not allow us to assign our findings to a particular cell population, we consider the possibility that α2β integrin binding to laminin 5 may represent a secondary event occurring after migration has been induced.

The strong pericellular expression of the α2β integrin in migrating keratinocytes during the process of wound healing, correlated to the staining we observed with mAb D4B5, indicates that the transient interaction of α2β integrin with the laminin γ2 chain may be of physiological relevance. In summary, the present study indicates that, among integrins involved in laminin 5 interactions, α2β seems to have a role to play during migration of keratinocytes while α6β4 retards it. We show that the α2β integrin binds to the short arm of the γ2 chain. In addition to controlling cell growth, TGFβ certainly has an important role to play in epidermal repair by regulating the transcription of laminin 5 and by controlling its interactions with keratinocytes by modulating the integrin expression at the cell surface.

We are indebted to Prof. Burgeson for the generous gift of 6F12 and NP32 mAbs, Dr Hemler for A3×8 mAb, Dr Meneguzi for pAb L 799P, Dr Clezardin for pAb Ab-8, Prof. Aumailley for the kind gift of laminin 1 and to Dr Braye for providing us with the wound biopsy. We thank M. Gonzalez for excellent technical assistance and A. Bosch for artwork. This work was supported by grants from the CNRS, the Association pour la Recherche sur le Cancer, the Ligue Nationale contre le Cancer (Rhône-Alpes, Loire), a CNRS-INSERM grant (Adhésion Cellules/Biomatériaux) and support as JSID’s International Fellowship Shiseido Award to P. Rousselle. F. Décline was a recipient of a studentship from the French government and the Fondation pour la Recherche Médicale.