Deposition of Laminin 5 by Keratinocytes Regulates Integrin Adhesion and Signaling*

Deposition of laminin 5 over exposed dermal collagen in epidermal wounds is an early event in repair of the basement membrane. We report that deposition of laminin 5 onto collagen switches adhesion and signaling from collagen-dependent to laminin 5-dependent. Ligation of laminin 5 by integrin α6β4 activates phosphoinositide 3-OH-kinase (PI3K) signaling. This activation allows for adhesion and spreading via integrin α3β1 on laminin 5 independent of RhoGTPase, a regulator of actin stress fibers. In contrast, adhesion and spreading on collagen via α2β1 is Rho-dependent and is inhibited by toxin B, a Rho inhibitor. Deposition of laminin 5 and ligation of α6β4 increases PI3K-dependent production of phosphoinositide di- and triphosphates, PI3K activity, and phosphorylation of downstream target protein c-Jun NH2-terminal kinase. Conversely, blocking laminin 5-deposition with brefeldin A, an inhibitor of vesicle transport, or with anti-laminin 5 monoclonal antibodies abolishes the PI3K-dependent spreading mediated by α3β1 and phosphorylation of c-Jun NH2-terminal kinase. Studies with keratinocytes lacking α6β4 or laminin 5 confirm that deposition of laminin 5 and ligation by α6β4 are required for PI3K-dependent spreading via α3β1. We suggest that deposition of laminin 5 onto the collagen substratum, as in wound repair, enables human foreskin keratinocytes to interact via α6β4and to switch from a RhoGTPase-dependent adhesion on collagen to a PI3K-dependent adhesion and spreading mediated by integrin α3β1 on laminin 5.

Laminin 5 is a major adhesive component of many epithelial basement membranes (BMs) 1 and is secreted by epithelial cells in tissue and in culture. Injury of epidermis activates transcription and deposition of laminin 5 into the provisional BM by the leading keratinocytes of the epidermal outgrowth that migrates into the wound bed (1)(2)(3)(4). The activated expression of laminin 5 occurs within hours after injury and prior to expres-sion of laminin 10/11 or type VII collagen, attesting to the import of laminin 5 for tissue repair. Deposition of laminin 5 over the exposed collagen and fibronectin is required for the repair of the BM and for re-establishing anchorage of the epithelium to the BM via integrin ␣ 6 ␤ 4 in hemidesmosomes (5)(6)(7)(8)(9). Here, we have examined the function of laminin 5 deposited onto exposed dermal collagen in regulating adhesion and signaling in epidermal outgrowths.
However, in normal human keratinocytes, ␣ 6 ␤ 4 plays a primary anchorage role that is independent of actin or FAK phosphorylation (9,11,30). Conceivably, ␣ 6 ␤ 4 may have different adhesion and signaling functions in normal keratinocytes as compared with carcinoma cell lines.
The primary mediator of motility on laminin 5 in normal keratinocytes is integrin ␣ 3 ␤ 1 , not ␣ 6 ␤ 4 (9,11,18,31,32). Adhesion dependent signals mediated by ␤ 1 integrins are transmitted via tyrosine phosphorylation, changes in intracellular pH, activation of protein kinases and mitogen-activated protein kinase pathways, and recruitment of regulators involved in focal adhesion formation such as focal adhesion kinase (FAK). Focal adhesion formation and phosphorylation of FAK are downstream of RhoGTP activity (33,34). Curiously, integrin ␣ 3 ␤ 1 in keratinocytes is not readily detectable in focal adhesions on laminin 5 in contrast to the robust localization of ␣ 2 ␤ 1 in focal adhesions on collagen (35,36). This suggests that interactions of keratinocytes with laminin 5 differ from interactions with collagen and the differences may be attributable to signaling through Rho that affects assembly of focal adhesions. Recently, we showed that laminin 5 interactions with ␣ 3 ␤ 1 promote assembly of gap junctions and increase intercellular communication in basal keratinocytes (2). Collagen or fibronectin adhesion via ␣ 2 ␤ 1 and ␣ 5 ␤ 1 , respectively, does not promote gap junction intercellular communication. Ligation of ␣ 2 ␤ 1 by collagen, however, promotes expression of matrix metalloproteinase 1 (MMP1) that is necessary for migration on collagen in the wound bed (37)(38)(39)(40). Therefore, BM laminin 5 provides instructional signals that are distinct from those of dermal collagen or fibronectin. However, the signaling pathways resulting from ␣ 3 ␤ 1 -laminin 5 and ␣ 2 ␤ 1 -collagen ligations have yet to be characterized.
We wished to determine if deposition of laminin 5 onto collagen during wound repair changes signaling from collagen-dependent to laminin 5-dependent and if signaling pathways required for adhesion and spreading on laminin 5 were different than on collagen. We found that laminin 5 is deposited as a precursor form by a subpopulation of keratinocytes migrating at the leading edge of the wound outgrowths. Further, laminin 5 requires and stimulates signaling pathways for adhesion and spreading of keratinocytes that are different then collagen. Deposits of laminin 5 that interact with integrin ␣ 6 ␤ 4 are required for keratinocyte spreading on laminin 5 in the absence of RhoGTPases. We suggest that deposition of laminin 5 by keratinocytes at the leading edge of the wound outgrowth determines which signaling pathways are used by keratinocytes for motility. The advantage of dual signaling pathways for migration of keratinocytes on laminin 5 are discussed in relation to epithelial cell interactions and wound repair.

EXPERIMENTAL PROCEDURES
Cells and Cell Culture-Primary keratinocytes from normal human foreskins (HFKs) and keratinocytes from individuals with junctional epidermolysis bullosa-pyloric atresia (JEB-PA; inherited defect in the ITGB4 gene) or from individuals with junctional epidermolysis bullosagravis (JEB-G; inherited defect in the LAMB3 gene) were prepared as described by Boyce and Ham (41). Primary and secondary keratinocyte cultures were maintained in serum-free keratinocyte growth media (KGM; Clonetics Corp., San Diego, CA). JEB-PA and JEB-G keratinocytes were infected with an LXSN retroviral vector (42) expressing E6 and E7 oncogenes from human papilloma virus to extend life in culture. When JEB-PA and JEB-G keratinocytes were compared with HFKs (Figs. 4 and 5), the HFKs were also infected with the same retroviral vector. Early passages (p1-15) were compared to avoid possible changes in adhesion or signaling due to later passages. JEB-PA individual JF/VS9-3-96 has premature termination mutations in both alleles of the ITGB4 gene encoding the ␤ 4 integrin subunit (JF/VS9-3-96 is proband from family two in Ref. 43). Three other individuals suspect for defects in ITGB4 were also examined. They are classified clinically as JEB-PA based on blistering phenotype, immature hemidesmosome structures, defective anchorage on laminin 5 in assays specific for ␣ 6 ␤ 4 , and decreased expression of ␤ 4 in tissue and culture. Causal mutations are currently under investigation. The individual with JEB-G does not deposit laminin 5 in culture and has homozygous premature termination mutations in both alleles of the LAMB3 gene encoding the ␤ 3 chain of laminin 5.
Explants of foreskin were grown on collagen layers (epibole cultures) as described previously (2) and used as a wound model. The collagen layers were prepared by drying rat tail collagen on Petri dishes (Collaborative Research, acid-soluble, diluted to 100 g of collagen/ml of water).
In Vitro Cell Adhesion Assay-Human placental collagen type I was prepared in the lab from human placenta as described (47) and coated onto 12-well culture plates overnight at 4°C at the concentration of 10 g/ml. Laminin 5-coated plates were prepared as described previously (11) and briefly as follows; HFKs were grown on TC plates, then detached by brief digestion with 0.5% trypsin (w/v)/EDTA (Sigma). Laminin 5 remaining on the TC surface was blocked with 0.5% w/v heatdenatured BSA/PBS for 30 min and used for adhesion studies. Cell adhesion to these laminin 5 plates was completely inhibited with antilaminin 5 mAb C2-9 (11). To generate laminin 5-conditioned collagen, HFKs were plated onto collagen-coated assay plates for 24 h. This allowed the cells to deposit laminin 5 onto the collagen substratum. The cells were trypsinized off, leaving the substrata for adhesion assay. Coated plates were blocked with 0.5% w/v heat-denatured BSA/PBS for 30 min. Cultured cells were either not treated for controls or pretreated with 50 ng/ml toxin B (gift from Laurie Neville, Tech Labs, Inc.), 50 nM wortmannin (Sigma), or 5 mM LY294002 (Sigma) for 3 h. Cells were then suspended with trypsin (w/v)/EDTA and labeled with 0.5 M calcein-AM (Molecular Probes, Eugene, OR). Labeled cells were then plated onto ligand-coated plates in the presence or absence of inhibitors (e.g. anti-␣ 6 antibody, G0H3) for 1 h as indicated. Total cell fluorescence (prewash) was read on CytoFluorII fluorescence plate reader (PerSeptive Biosystems, MA). Cells were washed three times with PBS, and remaining cell fluorescence (post-wash) was again read. An empty well containing just buffer medium where no cells were added was read to give base-line fluorescence that was subtracted from raw data. In every adhesion experiment, triplicates of each condition were done. Percent fluorescent cells adhered was calculated as: % fluorescent cell adhered ϭ (post-wash fluorescence reading Ϫ base-line fluorescence)/ (total (prewash) fluorescence reading Ϫ base-line fluorescence). Cells were then fixed with 2% formaldehyde in 0.1 M sodium cacodylate/0.1 M sucrose for 20 min and viewed under phase microscope for documentation of cell spreading.
In Vitro Tissue Adhesion Assay-The in vitro tissue adhesion assay was performed as a variation of the assay described (5)(6)(7)(8)(9). In brief, epidermis of neonatal foreskin was separated from the dermis by incubation with EDTA/NaCl to expose the underlying BM containing laminin 5. The dermis is abundant in interstitial collagen (types I and III) and fibronectin. Tissue sections (6-m thick) were cut from OCT-embedded frozen split skin and adhered onto lids of 35-mm Petri dishes, washed twice in PBS, and blocked in 0.5% heat-denatured BSA/PBS for 30 min. HFKs were untreated or pretreated with toxin B (50 ng/ml) for 3 h prior to assay. Calcein-labeled HFKs were plated onto the immobilized tissue sections and allowed to adhere for 1 h at 37°C. Nonadherent cells were washed twice with PBS. Adherent cells were fixed with 2% formaldehyde, mounted, and examined by phase and fluorescence microscopy.
Immunofluorescence Staining-Cells were plated onto surfaces coated with immobilized anti-␣ 3 antibodies or laminin 5 for 30 min. Cells were then treated Ϯ toxin B (50 ng/ml) or Ϯ wortmannin (50 nM) for 3 h. Cells were fixed for 20 min with 2% formaldehyde in 0.1 M sucrose, 0.1 M sodium cacodylate (pH 7.2), permeabilized with 0.5% Triton X-100 for 10 min, and then blocked in 0.5% heat-denatured BSA in PBS for 30 min. P1F2 (anti-␣ 3 ), VIN-11-5 (anti-vinculin; Sigma), and anti-FAK (ICN) were incubated with the cells for 1 h at room temperature. The cells were washed, and bound antibody was reacted with fluorescein isothiocyanate-conjugated, goat anti-mouse antibody (Cappel). Rhodamine-conjugated phalloidin (Molecular Probes) was incubated for 10 min and washed. For laminin 5 staining, attached cells were extracted with Triton, then the matrix was stained with biotinylated anti-laminin 5 mAb (P1E1). Bound biotinylated mAb was detected with rhodamine-avidin. Immunofluorescence was visualized using a Zeiss microscope equipped with epifluorescence. Images were photographed with TMAX 400 ASA film.

Leading Keratinocytes in Epidermal Outgrowths Express
Laminin 5 and Are Sensitive to Toxin B-We used skin explants placed in organ culture to study keratinocyte migration over collagen and the effects of deposited laminin 5. Biopsies of skin were placed onto collagen-coated surfaces, allowing keratinocytes to migrate out as a continuous epithelial outgrowth.
Keratinocytes at the leading edge of the outgrowth are in direct contact with collagen, and express elevated cytoplasmic levels of a precursor form of laminin 5 recognized by mAb D2-1 (45,46). This distinguishes the leading keratinocytes from the following cells that express little cytoplasmic laminin 5 ( Fig. 1a; arrowheads indicate leading cells, arrows identify following cells). Precursor laminin 5 is deposited by the leading cells onto the collagen substratum over which the following cells migrate. Extraction of the epithelial outgrowth with Triton X-100 detergent allows antibody access to the substratum and exposes the deposited laminin 5 under the following cells (Fig. 1b, arrow indicates direction of migration). Note that the extraction also removes the cytoplasmic precursor laminin 5 from the leading cells (dotted line demarks leading edge). mAb D2-1 was used to characterize the precursor form of laminin 5 (␣ 3 ␤ 3 ␥ 2 ) synthesized and deposited by leading cells. Results in Fig yields ␣ 3 -165 and ␣ 3 -35/37. (iii) Peptides ␣ 3 -35/37 bound heparin and were affinity purified on immobilized heparin (Fig. 2B). Partial sequencing of purified ␣ 3 -35/37 identified the amino acids sequence identical to residues 1484 -1497 near the amino-terminal end of the LG4 subdomain of ␣ 3 -200 (45,46). Based on the relative molecular mass of ␣ 3 -35/37 and the presence of amino acid sequences from both LG4 and LG5, we suggest that cleavage of ␣ 3 -200 occurs extracellularly at two sites between LG3 and LG4 to release the heparin-binding ␣ 3 -35/37. Based on the staining with mAb D2-1 (Fig. 1, a and b), we conclude that precursor laminin 5 is synthesized and deposited primarily by the leading keratinocytes in epidermal outgrowths and the precursor and or processed laminin 5 deposited on the collagen substratum interacts with the following keratinocytes.
We examined signaling differences in leading and following keratinocytes that interact with collagen and deposited laminin 5, respectively. We compared the sensitivities of leading and following keratinocytes to toxin B, an inhibitor of RhoGT-Pase that regulates actin stress fibers. Toxin B glucosylates Thr 37 of Rho proteins, thereby rendering them inactive (45,46). Toxin B (50 ng/ml) was added to cultures of skin explants grown on collagen, and the epidermal outgrowth was photographed at the zero time point (Fig. 1c, black arrow indicates leading edge) and also at 10 min (Fig. 1, d and e; panel e is a 2-fold magnification of d to better show cell rounding at the leading edge). Toxin B selectively rounded the leading cells, whereas following cells remained adherent and spread. Thus, adhesion and spreading of leading keratinocytes on collagen is toxin B-sensitive, whereas adhesion and spreading of following cells on deposits of laminin 5 is toxin B-insensitive.
Integrin ␣ 2 ␤ 1 Interaction with Collagen Is Toxin B-sensitive Compared with ␣ 3 ␤ 1 -Laminin 5-We compared toxin B sensitivity of HFKs adherent on exogenous collagen via ␣ 2 ␤ 1 to cells adherent on laminin 5 via ␣ 6 ␤ 4 and ␣ 3 ␤ 1 . Cultured HFKs were plated onto exogenous laminin 5 (Fig. 3, a and c) or collagen (b and d) for 30 min. Toxin B (50 ng/ml) was added, and cells were photographed at time 0 (a and b) and at 30 min (c and d). Cells were stained with rhodamine-conjugated phalloidin for actin. Consistent with results in Fig. 1, HFKs in the presence of toxin B on collagen rounded up while cells on laminin 5 retained their spread morphology.
Initial adhesion and spreading of HFKs on collagen was inhibited by toxin B in a dose-dependent manner (Fig. 4A, open circles). In contrast, adhesion and spreading of HFKs on laminin 5 was unaffected by treatment with toxin B (Fig. 4, A (black squares) and B (panel b)). Although both ␣ 6 ␤ 4 and ␣ 3 ␤ 1 contribute to adhesion on laminin 5, the spreading on laminin 5 in the presence of toxin B was blocked by anti-␣ 3 antibodies (P1B5; Fig. 4B, panel c) and not by anti-␣ 6 antibodies (data not shown). This confirmed that ␣ 3 ␤ 1 mediates spreading on laminin 5 in the presence of toxin B.
We used a tissue adhesion assay to determine whether native collagen in tissue could support adhesion of HFKs in the presence of toxin B (45,46). HFKs were treated with toxin B (50 ng/ml for 3 h), fluorescently labeled with calcein, and then assayed for adhesion on cryostat sections of skin (Fig. 4C). For these studies, the epidermis was separated from the skin by incubation with EDTA/NaCl in order to expose the underlying BM containing laminin 5. The dermis is rich in interstitial collagens, including types I and III, and fibronectin. Untreated control cells adhered to both laminin 5 in the BM and to the dermis that contains collagen (Fig. 4C, panels a and c). In contrast, toxin B-treated HFKs only adhered to the BM and not to the dermis (Fig. 4C, panels b and d). Thus, ␣ 2 ␤ 1 failed to mediate adhesion in the presence of toxin B even on native collagen in dermis. In contrast, toxin B had no observable inhibitory effects on adhesion to laminin 5 in the BM. Figs. 1, 3, and 4, HFKs interacting with laminin 5 can spread via ␣ 3 ␤ 1 in the presence of toxin B. We evaluated the import of laminin 5-deposition on this toxin B-insensitive spreading. HFKs were plated onto immobilized anti-␣ 3 antibody surfaces. The cells adhered, spread, and deposited laminin 5 onto the anti-␣ 3 substratum in the absence (Fig. 5, A (Control, panel a) and B (panel a)) or in the presence of toxin B (Fig. 5, A (ToxB, panel d) and B (panel b)). Blocking laminin 5 deposition with brefeldin A (Fig. 5B, panel d) inhibited HFK spreading in the presence of toxin B (Fig. 5A, panel  g). Blocking HFK interactions with laminin 5 using inhibitory antibody C29 also inhibited cell spreading in the presence of toxin B (Fig. 5A, compare panels b and e). Keratinocytes from an individual with JEB-G with inherited defects in laminin 5 expression (50) were unable to spread on anti-␣ 3 immobilized surface in the presence of toxin B (Fig. 5A, panel f compared  with untreated control panel c). These JEB-G keratinocytes do not express nor deposit laminin 5 in culture (data not shown). However, addition of exogenous laminin 5 allowed JEB-G keratinocytes to spread in the presence of toxin B (Fig. 5A, panel  h). Therefore, either endogenous or exogenous laminin 5 on the substrate is required for the toxin B-insensitive cell spreading via ␣ 3 ␤ 1 .

Laminin 5 Is Required for Spreading via ␣ 3 ␤ 1 in the Presence of Toxin B-As seen in
Toxin B-insensitive Spreading via ␣ 3 ␤ 1 on Laminin 5 Requires ␣ 6 ␤ 4 and PI3K-Since ␣ 6 ␤ 4 also contributes to adhesion to laminin 5, we determined if interaction of ␣ 6 ␤ 4 with laminin 5 participates in the toxin B-insensitive spreading via ␣ 3 ␤ 1 . Individuals with JEB-PA bear mutations in the ITGB4 gene that encodes the integrin ␤ 4 subunit (51, 52) and display severe to lethal defects in anchorage function of ␤ 4 (13). Keratinocytes cultured from a JEB-PA individual (Fig. 6A, striped bars) attached and spread on exogenous laminin 5 via ␣ 3 ␤ 1 in the absence of toxin B comparable to normal HFK controls (Fig. 6A,  solid bars). However, JEB-PA keratinocytes failed to adhere and spread on exogenous laminin 5 in the presence of toxin B (Fig. 6A, ϩ Toxin B, striped bars). Keratinocytes from three additional individuals suspect for JEB-PA were also examined for sensitivity to toxin B. Toxin B inhibited adhesion and spreading on laminin 5 for all three of the suspect JEB-PA keratinocytes (data not shown). Therefore, ligation of ␣ 6 ␤ 4 by laminin 5 is required for keratinocyte spreading via ␣ 3 ␤ 1 in the presence of toxin B.
We wished to identify the signaling pathway involved in the toxin B-insensitive spreading on laminin 5. A panel of known inhibitors of adhesion was examined for effects on HFK adhesion and spreading on laminin 5 in the presence of toxin B. In preliminary studies, wortmannin or LY294002, inhibitors of phosphoinositide 3-OH-kinase (PI3K; Ref 53), selectively inhibited ␣ 3 ␤ 1 -mediated spreading on laminin 5 when toxin B was also present. Quantitative experiments were performed (Fig.  6B). HFKs were pretreated with toxin B alone, 50 nM wortmannin alone, or the combination of both, then assayed for adhesion to exogenous laminin 5 (solid bars) or collagen (striped bars). Untreated HFKs adhered and spread on both laminin 5 and collagen. Toxin B (TxB) selectively inhibited collagen adhesion whereas on laminin 5 the cells still adhered and spread. Wortmannin (Wt) by itself did not inhibit adhesion or spreading on either laminin 5 or collagen. However, the combination of toxin B and wortmannin (TxB ϩ Wt) inhibited adhesion and spreading on both laminin 5 and collagen. Note that LY294002 generated results similar to wortmannin (results not shown) and had no inhibitory effects on the deposition of laminin 5 (Fig. 5B, panel c). Note that the assay was done in the presence or absence (data not shown) of anti-␣ 6 mAb (GoH3) to block ␣ 6 ␤ 4 contribution to adhesion. In the absence of anti-␣ 6 mAb spreading was inhibited by toxin B and wortmannin but adhesion occurred. Thus, ␣ 3 ␤ 1 mediates adhesion and spreading on laminin 5 via two pathways; one is PI3K-dependent, and the other is toxin B-sensitive.
Similar results were obtained when we assayed cell spreading on immobilized antibodies (Fig. 6C). We compared normal HFKs to JEB-PA keratinocytes adherent on immobilized anti-␣ 3 antibodies under all assay conditions. JEB-PA keratinocytes spread in the absence but not the presence of toxin B (Fig.  6C, compare panels c and f) and was unaffected by wortmannin (panel i). HFKs spread on anti-␣ 3 in the absence (panel a) or presence of toxin B (panel d) or wortmannin (panel g). HFK spreading was blocked only in the presence of a combination of inhibitors (panel j). Therefore, integrin ␣ 6 ␤ 4 plays a critical role in regulating ␣ 3 ␤ 1 spreading via the PI3K-dependent pathway.
Adhesion of HFKs on immobilized anti-␣ 2 antibody was apparent under all assay conditions but spreading was inhibited in the presence of toxin B or toxin B ϩ wortmannin but not by wortmannin alone (Fig. 6C, panels e, k, and h, respectively). The behavior of HFKs on immobilized anti-␣ 2 is similar to the behavior of JEB-PA keratinocytes on anti-␣ 3 .
Inhibition of PI3K Increases Focal Adhesions and FAK Phosphorylation-Laminin 5 interaction with ␣ 6 ␤ 4 is required for cell spreading on laminin 5 via ␣ 3 ␤ 1 in the presence of toxin B (Figs. 5 and 6). Further, PI3K is required for spreading on laminin 5 via ␣ 3 ␤ 1 in the presence of toxin B. We hypothesized that inhibition of PI3K would increase Rho dependence and increase Rho activity. Rho activity is essential for focal adhesion formation when cells interact with ECM ligands via integrin receptors (33,34). We examined focal adhesions in HFKs plated on laminin 5 after treatment with toxin B or wortmannin. Untreated control HFKs did not efficiently localize ␣ 3 ␤ 1 , vinculin (VIN), or FAK into focal adhesion structures (Fig. 7,  panels a, d, and g). Consistent with inactivating Rho, toxin B treatment decreased localization of all three components in focal adhesions (panels b, e, and h). Note that toxin B caused disappearance of actin stress fibers; however, the cells retained  6. Toxin B-insensitive adhesion and spreading mediated by ␣ 3 ␤ 1 requires ␣ 6 ␤ 4 and PI3K. A, HFKs (solid bars) and keratinocytes from an individual (JF/VS 9-3-96) with JEB-PA (striped bars) lacking integrin ␣ 6 ␤ 4 were treated with (ϩ) or without (Ϫ) 50 ng/ml toxin B for 3 h, trypsin-suspended, labeled with calcein, and assayed for adhesion to exogenous laminin 5. B, HFKs were untreated (Control), or treated with toxin B (TxB; 50 ng/ml), wortmannin (Wt; 50 nM), or toxin B ϩ wortmannin combination for 3 h, trypsin-suspended, labeled with calcein, and added to surfaces coated with exogenous laminin 5 (solid bars) or collagen (striped bars). The assay was done in the presence or absence (data not shown) of anti-␣ 6 mAb (G0H3). In either case no spreading occurred in the presence of toxin B ϩ wortmannin. In the absence of anti-␣ 6 , adhesion was unaffected but spreading was inhibited with toxin B ϩ wortmannin. C, phase images of HFKs that were untreated (Control, panels a and b), or treated with toxin B (TxB, d and e), wortmannin (Wort., g and h), or the combination of toxin B and wortmannin (j and k) for 3 h, and plated onto their spread morphology (Fig. 7, compare panels j and k). Inhibiting PI3K with wortmannin caused a major increase in localization of ␣ 3 ␤ 1 , vinculin, FAK, and actin into focal adhesion sites (Fig. 7, panels c, f, i, and l). Note that LY294002 shows the same effects (data not shown).
Adhesion to Laminin 5 Increases Synthesis of Phosphoinositides and PI3K Activity-We compared steady-state levels of phosphotidylinositol di-phosphate (PIP 2 ) synthesized by PI3K in HFKs that are adherent on either laminin 5 or collagen (Fig.  9A). Cells were labeled for 1 h with 32 P i in original adherent (A) or in suspension (S) or during re-adhesion to laminin 5 (L5) or collagen (Col) with (ϩ) or without (Ϫ) wortmannin. Extracts from the labeled cells were fractionated by thin layer chromatography and levels of PIP 2 were quantitated. HFKs adhering to laminin 5 synthesized over 2 times higher levels of PIP 2 than cells adhering on collagen (L5Ϫ: 39.35;ColϪ: 19.15) and this increased level could be inhibited by wortmannin (L5ϩ: 18.32; Colϩ: 13.74). These results suggest that HFKs adhering to laminin 5 had higher levels of PI3K activity compared with cells on collagen.
Samples were taken at 15, 30, 60, and 120 min of adhesion. Adhesion and spreading of HFKs on anti-␣ 3 ( Fig. 9B; quantitated in Fig. 9C, open circles) generated a more rapid and higher levels of PIP 2 and PIP 3 than cells adhering on anti-␣ 2 (Fig. 9, B and C, black squares). This confirms that adhesion to laminin 5 via ␣ 3 ␤ 1 generates a higher level of PI3K activity than adhesion to collagen via ␣ 2 ␤ 1 .
Deposition of Laminin 5 Regulates PI3K-dependent JNK Phosphorylation-To confirm the results from the functional adhesion studies on laminin 5 and collagen and the PI3K immobilized anti-integrin ␣ 3 (P1B5; a, d, g, and j) or anti-integrin ␣ 2 (P1H5; b, e, h, and k) 9. Interaction with laminin 5 activates PI3K activity compared with collagen. A, steady-state levels of phosphotidylinositol diphosphate (PIP2) synthesized by PI3K were compared in HFKs that are adherent on either laminin 5 or collagen. Cells were labeled for 1 h with 32 P i in original adherent (A) or labeled in suspension (S) or during adhesion to laminin 5 (L5) or collagen (Col) with (ϩ) or without (Ϫ) wortmannin. B, time course of PIP 2 and PIP 3 synthesis resulting from PI3K activation upon cell adhesion via ␣ 3 ␤ 1 and ␣ 2 ␤ 1 . HFKs were labeled for 1 h with activation assays, we examined downstream signaling responses induced by adhesion to laminin 5 and collagen. We found that adhesion of HFKs to laminin 5 (Fig. 10, A (lane 2) and B) induces phosphorylation of Jun kinase (JNK) 2-fold over adhesion to collagen (Fig. 10, A (lane 3) and B) when assayed by quantitative immunoblotting. In contrast, HFK adhesion to collagen induced higher levels of phosphorylated ERK2 seen as a slower migrating band (Fig. 10A, lane 3 compared with lane  2). Phosphorylation of ERK2 is inhibited by toxin B (results not shown). This confirmed that adhesion of HFKs on laminin 5 stimulates different signals then collagen. We determined which integrin, ␣ 3 ␤ 1 or ␣ 6 ␤ 4 , was responsible for induction of JNK phosphorylation as a result of adhesion to laminin 5.
Deposition of Laminin 5 on Collagen Switches HFK Spreading from Toxin B-sensitive to -insensitive-We have shown that laminin 5 is deposited onto substratum during epithelial migration on collagen (Fig. 1), and that this deposition of laminin 5 is required for toxin B-resistant spreading on immobilized anti-␣ 3 mAbs (Fig. 5) and induction of PI3K-dependent phosphorylation of JNK (Fig. 10F). We sought to determine if deposition of laminin 5 is sufficient to convert cell spreading from toxin B-sensitive to toxin B-resistant. We compared HFK adhesion and spreading on exogenous laminin 5, collagen, and laminin 5-conditioned collagen (see "Experimental Procedures"). HFKs were untreated (Fig. 11, Control) or treated with toxin B (50 ng/ml; TxB), LY294002 (5 nM; LY), or the combination of toxin B and LY294002 (TxBϩLY) for 3 h. Cells were then trypsinized and in the presence of inhibitory G0H3 were plated on exogenous laminin 5 (Lam5; black bars), collagen (Col I; striped bars), or laminin 5-conditioned collagen (Lam5 cond; gray bars). Toxin B was unable to inhibit HFK adhesion and spreading on the laminin 5-conditioned collagen, whereas it inhibits on collagen alone. Inhibiting PI3K with LY294002 in conjunction with toxin B blocked cell adhesion and spreading on all surfaces. Thus, laminin 5 deposited over collagen substratum, as in the case of cell migration in epidermal wound, is sufficient to convert HFKs from a toxin B-sensitive to a toxin B-resistant adhesion and spreading.
In summary, ␣ 3 ␤ 1 mediating adhesion and spreading on laminin 5 utilizes either Rho or PI3K. The PI3K pathway in HFKs requires expression of ␣ 6 ␤ 4 and ligation of laminin 5 and 32 P i in suspension (0) and then adhered onto immobilized anti-␣ 3 (P1B5) or anti-␣ 2 (P1H5) mAbs. Samples were taken at 0, 15, 30, 60, and 120 min of adhesion. C, densitometry reading of PIP 2 and PIP 3 bands from B, comparing time course between cells on immobilized anti-␣ 3 (open circles) or anti-␣ 2 (black squares). D, PI3K immunoprecipitated from extracts of HFKs that were suspended, or plated onto immobilized anti-␣ 3 (P1B5), anti-␣ 2 (P1H5), or anti-␣ 6 (G0H3) and PI(3,4)P 2 was used as an exogenous substrate (see "Experimental Procedures"). E, densitometry reading of relative ratios of PIP 3 on TLC from D. functions to convert cell signaling pathways from toxin Bsensitive to toxin B-insensitive. Finally, this PI3K pathway enables wound keratinocytes to adhere, spread on laminin 5 via ␣ 3 ␤ 1 in the absence of RhoGTPase activity. DISCUSSION We report that the deposition of laminin 5 by keratinocytes onto the substrate is a key factor in regulating mechanisms of both cell spreading and signaling. The findings suggest that primary human keratinocytes utilize at least two different signaling pathways when adhering and spreading on ECM. Which pathway is utilized by the keratinocytes is established by the substrate ligand. One pathway is dependent on PI3K signaling and the other on RhoGTPases. Adhesion and spreading via ␣ 3 ␤ 1 on laminin 5 is blocked only when both pathways are inhibited using toxin B and wortmannin. In contrast, adhesion and spreading on collagen via ␣ 2 ␤ 1 is blocked with toxin B alone, suggesting a primary dependence on RhoGTPases. Deposition of laminin 5 onto collagen substratum switches signaling from a Rho-dependent to a PI3K-dependent pathway.
Laminin 5-␣ 6 ␤ 4 Interactions Regulate Spreading and Signaling Mediated by ␣ 3 ␤ 1 -Toxin B-insensitive spreading via ␣ 3 ␤ 1 requires ligation of both ␣ 6 ␤ 4 and ␣ 3 ␤ 1 to laminin 5 deposited by keratinocytes. PI3K signaling is also required. Consistently, PI3K is activated through the ligation of ␣ 6 ␤ 4 to laminin 5. Recent findings indicate that ␣ 6 ␤ 4 signals through PI3K to RacGTPases to mediate migration and invasion of carcinoma cell lines (29). In contrast to carcinoma cell lines, primary HFKs utilize integrin ␣ 3 ␤ 1 for cell spreading, not ␣ 6 ␤ 4 , yet at the same time is dependent upon ␣ 6 ␤ 4 -laminin 5 signaling. Further, we observed the PI3K-dependent spreading on laminin 5 but not on collagen, suggesting that ␣ 2 ␤ 1 does not respond to signals from PI3K or that ligation of collagen inhibits PI3Kdependent spreading. We have also examined toxin B sensitivity in cell adhesion and spreading in primary human foreskin fibroblasts (HFF, data not shown). HFFs express ␣ 3 ␤ 1 and can interact with exogenous laminin 5 (29). However, HFFs in the presence of toxin B fail to adhere and spread on laminin 5. This characteristic of HFFs is similar to keratinocytes isolated from JEB-PA individuals lacking ␣ 6 ␤ 4 function. This suggests that the toxin B-insensitive spreading on laminin 5 is unique to cells possessing ␣ 6 ␤ 4 signaling. It remains to be established if the expressed ␣ 6 ␤ 4 must also be able to function in hemidesmo-somes. In addition, we have also tested keratinocyte cell lines immortalized with E6/E7 oncogenes from human papilloma virus. These cells promiscuously adhered and spread in the presence of toxin B to all tested matrix surfaces (data not shown). Hence, the specific regulation of signaling pathways that differentiate between laminin 5 and collagen is lost in transformed cells. Therefore, collagen invasion by carcinomas or transformed keratinocyte cell lines may originate as a subversion of a normal wound repair pathway that is selective for ␣ 3 ␤ 1 on laminin 5 in normal epithelium.
Links between the Rho-dependent and PI3K-dependent Pathways-Our results establish that integrin ␣ 3 ␤ 1 utilizes two different signaling pathways for adhesion and spreading: Rhodependent and PI3K-dependent. The PI3K-dependent pathway requires laminin 5 ligation of ␣ 6 ␤ 4 . Inhibition of PI3K by wortmannin increased focal adhesions containing ␣ 3 ␤ 1 , vinculin, FAK, and actin stress fibers as well as increased phosphorylation of FAK. In this condition, adhesion and spreading mediated by ␣ 3 ␤ 1 becomes dependent upon Rho and can be inhibited by toxin B. Interestingly, keratinocytes from JEB-PA (␤ 4 -deficient) individuals also had increased focal adhesions and stress fiber assembly on laminin 5 (data not shown). Consistently, toxin B blocked JEB-PA keratinocyte adhesion to laminin 5, suggesting that ␣ 6 ␤ 4 and PI3K signaling were required for Rho independence of ␣ 3 ␤ 1 . Thus, the active PI3K pathway induced by laminin 5 in primary HFKs may serve to suppress the Rho-dependent pathway on laminin 5.
It is not apparent why HFKs in the presence of toxin B can adhere and spread on laminin 5, immobilized anti-␣ 3 but not on collagen or immobilized anti-␣ 2 mAbs. HFKs deposit precursor laminin 5 onto the collagen and presumably interact with the deposits via ␣ 3 ␤ 1 and ␣ 6 ␤ 4 . Further, we established that deposited laminin 5-␣ 6 ␤ 4 interactions signal through PI3K to control spreading through ␣ 3 ␤ 1 . On laminin 5, the PI3K signals suppressed FAK phosphorylation and focal adhesion formation that require Rho (Figs. 7 and 8). Conceivably, interaction of collagen with ␣ 2 ␤ 1 and signaling through Rho may be dominant to the PI3K signals resulting from the laminin 5 deposits. However, this dominance must be short-lived; as seen in Fig.  11, extended deposition of laminin 5 on the collagen surface eventually overrides the toxin B-sensitive interactions with collagen making them toxin B-resistant. How HFKs establish these adhesion and signaling preferences remains to be established.
In relation to the above point, laminin 5 is deposited onto the collagen substrate by leading keratinocytes in a precursor form containing the heparin-binding LG4/5 subdomain at the carboxyl terminus of the ␣ 3 -200 chain. The LG4/5 domain is released by proteolytic processing of the precursor laminin 5 as a post-secretory event generating the mature laminin 5. Hypothetically, the heparin binding activity and/or the proteolytic processing of precursor laminin 5 at either the ␣ 3 or ␥ 2 chains may participate in regulating the activity of PI3K and/or Rho. The addition of exogenous processed laminin 5 that lacks the LG4/5 to JEB-G keratinocytes rescued the PI3K signaling pathway (Fig. 5, panel h). Thus, LG4/5 does not appear to be required for toxin B-resistant spreading. Future studies will determine if the precursor form of laminin 5 promotes toxin B-sensitive spreading of HFKs on substratum.
Advantage of the Dual Signaling Pathways during Wound Repair-There are multiple possible advantages for a dual signaling pathway mediating adhesion and spreading on laminin 5 deposits. Stimulation of PI3K activity by laminin 5 may promote survival signals during stress-activated events such as tissue injury. Indeed, in mammary epithelial cells, laminin acts as a survival ligand through ␤ 1 integrins and cooperates with FIG. 11. Laminin 5 deposited over collagen substratum converts HFKs to toxin B-insensitive spreading. HFKs were untreated (Control) or treated with toxin B (50 ng/ml; TxB), LY294002 (5 nM; LY), or the combination of toxin B and LY294002 (TxBϩLY) for 3 h. Cells were then trypsinized, treated with G0H3, and plated on exogenous laminin 5 (Lam5; black bars), collagen (ColI; striped bars), or laminin 5-conditioned collagen (Lam5 cond; gray bars). Cells were also noted whether they were spread (ϩ) or not spread (Ϫ). growth factor signals to suppress apoptosis (54). Several recent reports have shown that FAK signaling to PI3K and Cas is important for matrix survival signals (55,56).
Laminin 5 dual signals could also regulate communication between cell-substrate and cell-cell adhesions. We recently reported that laminin 5, but not collagen or fibronectin, promoted gap junctional intercellular communication (GJIC; Ref. 54). Our findings suggest that Rho was required for laminin 5 induction of GJIC. RhoGTP and RacGTP are both required for E-cadherin-based cell-cell adhesion in epithelial cells (57)(58)(59)(60). During epidermal wound repair, keratinocytes move as an integrated epithelial sheet. Because Rho is required for cell-cell adhesion and GJIC on laminin 5, PI3K may be necessary to substitute for Rho in cell-substrate adhesion. Conversely, leading edge keratinocytes migrating over collagen do not communicate via gap junctions. Thus, leading keratinocytes on collagen depend on RhoGTPases for substrate interactions (Fig. 1). Active RhoGTPases may be a limiting factor in keratinocytes, and regulation may be controlled by cellular localization that is dependent on laminin 5. Collard's group (60) has shown that collagen and laminin induced different levels of intercellular adhesion; adhesion to substrate collagen decreased intercellular adhesion, Tiam1/RacGTPases localization at cell-cell contacts, and induced cell invasion. Adhesion to laminin increased Tiam1/RacGTPases localization to cell-cell junctions and inhibited cell invasion. Hence, regulation of the subcellular localization of RhoGTPases by cell-substrate ligands may affect intercellular adhesion, GJIC, and invasion.
In conclusion, we suggest that deposition of laminin 5 by keratinocytes in epidermal outgrowths plays a critical role in regulating adhesion, spreading, and signaling. Adhesion of HFKs to collagen is sufficient to promote Rho-dependent spreading and deposition of laminin 5 as part of a BM repair process. Interaction of following keratinocytes with this deposited laminin 5 activates the PI3K-dependent pathway for adhesion, spreading, and downstream JNK signaling as well as promotes gap junctional intercellular communication. Thus, instructions from collagen and laminin 5 allow leading and following keratinocytes, respectively, to perform distinct functions necessary for repair of the BM while maintaining the barrier function of a continuous impermeable epidermis.