Globular domains 4/5 of the laminin α3 chain mediate deposition of precursor laminin 5

Wounding of quiescent epidermis generates leading migratory keratinocytes that deposit precursor laminin 5 (pre-lam5; α3β3γ2) in the provisional basement membrane (PBM) of the wound (Aumailley, 2003; Borradori and Sonnenberg, 1999; Lampe et al., 1998; Nguyen et al., 2000b). Pre-lam5 contains a 200 kDa form of the laminin α3 chain (α3²⁰⁰), encoded by the LAMA3 gene (Ryan et al., 1994) with globular domains G4 and G5 (G4/5) (Aumailley, 2003; Nguyen et al., 2000b). Subsequent to deposition on exposed dermis, keratinocytes adhere and migrate on pre-lam5 via integrins α3β1 and/or α6β4 and on dermal collagen and fibronectin via integrins α2β1 and α5β1, respectively (Carter et al., 1990; Carter et al., 1991; Goldfinger et al., 1999; Mercurio et al., 2001; Nguyen et al., 2000b). Recently, we evaluated the coupling between deposition of pre-lam5 and subsequent adhesion and migration (Frank and Carter, 2003). We showed that deposition of pre-lam5 at the rear of migrating cells established linear polarized `processive migration' in the absence of chemotactic gradients. Here, we extended these studies to evaluate the function of G4/5 in deposition of pre-lam5 that precedes and is necessary for adhesion and processive migration.

Little is known about the role of G4/5 in the deposition of pre-lam5 onto dermis in wounds. Further, the order and interdependence of intracellular trafficking, deposition verses secretion, integrin ligation, and proteolytic processing of G4/5 domain have not been resolved. For deposition to occur, endogenous or exogenous pre-lam5 may interact with a cell surface receptor(s) followed by interaction with and retention in the dermis. For example, pre-lam5 is deposited into the PBM before other components of the mature BM including laminins 10/11 (α5β1γ1/α5β2γ1), collagen types IV and VII, nidogen and perlecan (Kainulainen et al., 1998; Lampe et al., 1998; Nguyen et al., 2000b). In the absence of other components, the deposited pre-lam5 may interact with uncharacterized components of the dermis. It is unclear how pre-lam5 makes the transition from the cytoplasm to the cell surface to deposits at the rear of leading keratinocytes (Frank, 2003). Binding of laminin 1 (α1β1γ1) onto the surface of a `competent' cell is dependent on the G4/5 domain of the laminin α1 chain (Li et al., 2003; Li et al., 2002; Smirnov et al., 2002). G4/5 in precursor α3²⁰⁰ and soluble G4/5 bind heparin (Nguyen et al., 2000a) and as a result candidate receptors for cell surface binding of the precursor α3²⁰⁰ chain of lam5 include sulfate-containing syndecans (Okamoto et al., 2003; Utani et al., 2001), dystroglycans (Hohenester et al., 1999; Winder, 2001) and glycolipids (Hall et al., 1997; Taraboletti et al., 1990). Integrin α3β1 was suggested to incorporate lam5 into its proper higher-order structure within the ECM of keratinocytes (deHart et al., 2003). Consistently, null defects in α3 (DiPersio et al., 1997) or β1 integrins (Grose et al., 2002; Grose and Werner, 2003; Raghavan et al., 2003) alter localization of deposited lam5 and adhesion. Significantly, lam5 is still deposited despite these adhesion defects. Null mutations in syndecans 1 (Stepp et al., 2002) and 4 (Echtermeyer et al., 2001), or integrin α6 (Georges-Labouesse et al., 1996) or β4 (Dowling et al., 1996; van der Neut et al., 1996) have no reported defects in lam5 deposition. Thus, receptors involved in adhesion to lam5 may be distinct from receptors involved in deposition of lam5. Significantly, inherited defects in the amino terminus of the γ2 chain of lam5 inhibit assembly in the BM (Gagnoux-Palacios et al., 2001). This may explain why laminin 6 (lam6; α3β1γ1) that contains the integrin-binding α3²⁰⁰ chain, but not γ2, fails to deposit and compensate for the adhesion defects in lam5 (Nakano et al., 2002).

Here, we evaluate the function of the G4/5 domain of the laminin α3²⁰⁰ chain in deposition of pre-lam5. As a definition for use here, endogenous cytoplasmic pre-lam5 can be either exported and deposited onto the substratum, termed deposition, or exported as a soluble protein, termed secretion. On the basis of three novel approaches, we conclude that the G4/5 domain contributes to efficient cell-mediated deposition of pre-lam5 onto a dermal equivalent.

Primary keratinocytes from normal human foreskins (HFKs) were grown as described previously (Boyce and Ham, 1985) in serum-free keratinocyte growth media (KGM; Clonetics, Corp., San Diego, CA). Keratinocytes from an individual with junctional epidermolysis bullosa gravis form (JEBG) (Nakano et al., 2002) were a gift from M. R. Pittelkow (Mayo Clinic College of Medicine, Rochester, MN) (Lim et al., 1996) and have homozygous null defect in the LAMB3 gene encoding the laminin β3 chain. Mouse keratinocytes (MKs) from wild-type mice (α3+/+ MKs) or mice with homozygous null mutations in the LAMA3 gene (α3–/– MKs) were immortalized with E6 and E7 oncogenes from human papilloma virus. α3–/– MKs on collagen-coated culture dishes (rat tail collagen I, Becton-Dickinson) and α3+/+ MKs, were maintained in KGM containing 60 μM calcium. Fibroblasts from human foreskin (HFs) or mouse skin (MFs) were grown in RPMI +10% fetal bovine serum.

Procedures for skin wounds in mice were approved by the Fred Hutchinson Cancer Research Center Animal Care Committee and were executed as follows: mice (C57BL/Ks) were anesthetized and shaved. Full thickness punch biopsies were created on the dorsal surface of the mice. Mice were euthanized with IP injections of sodium pentobarbital at 7 and 14 days after wounding and wounds were harvested, embedded in OCT and cyrostat sectioned (6 μm).

Mouse monoclonal antibodies (Mabs) against human laminin β3 chain (Mab A2′-2) and α3 chain (Mab P5H10), or rat Mabs against mouse lam5 (Mab P3C8) and against mouse and human β4 integrin (Mab P4G11) were isolated in this lab. The anti-laminin α3 chain Mab (P5H10) immunoblots denatured or nondenatured precleaved or noncleavable α3 chain, and immunoprecipitates denatured or nondenatured precleaved or noncleavable α3 chain in laminin 5 and laminin 6. An antibody against laminin β1 and γ1 chains (rabbit polyclonal antibody R5922) was prepared against denatured β1 and γ1 chains and is able to immunoblot denatured β1 or γ1 chains and immunoprecipitate denatured or nondenatured β1 or γ1 chains in laminins 1 and 6. Mouse Mabs P1H5 (inhibitory anti-α2 integrin), P1B5 (inhibitory anti-α3 integrin), P4C10 (inhibitory anti-β1 integrin), D3-4 (anti-human and mouse laminin α3), C2-5 (anti-human laminin α3), B4-6 (anti-human laminin γ2), C2-9 (inhibitory anti-human laminin α3) and D2-1 (anti-human laminin α3 G4/5 domain) have been described (Gil et al., 1994; Gil et al., 2002; Wayner and Carter, 1987). R5808-f1, a rabbit Mab directed against a peptide sequence (n-EQVYLGLSPSRKSKSLP-c) from the G5 domains of mouse laminin α3 chain was developed in this lab in conjunction with Elizabeth Wayner in the FHCRC hybridoma facility. The Mab is specific for the 200 kDa mouse precursor laminin α3 chain and the 37 and 39 dDa G4/5 products released by proteolytic processing of the α3 chain. Goat anti-laminin β3 chain antibodies are from Santa Cruz Inc. (Santa Cruz, CA). Rabbit polyclonal Ab against phosphorylated tyrosine residue 397 in focal adhesion kinase (FAKp-Y397) is from Bio-source (Hopkinton, MA).

α3–/– MKs were attached to glass coverslips for 3 hours in KGM then blocked for 15 minutes with 0.5% heat-denatured BSA in PBS generating attached round cells. The cells were then cultured in a mixture of 50% fresh KGM, 50% KGM conditioned by HFKs or JEBG keratinocytes, and 0.025% BSA, for 24 hours. Alternatively, precursor laminin 5 in HFK-conditioned media was separated from cleaved laminin 5 by heparin chromatography (Nguyen et al., 2000a). Thrombin digestion was performed at 37°C for 1 hour with 1 NIH unit of thrombin/ml, after which 2 units/ml of hirudin were added to block further thrombin activity. Fixed and permeabilized cultures were examined for spreading by phase contrast, and for deposition of human laminin 5 by immunostaining with antibodies. The area covered by deposits was measured, using Metamorph software, by counting pixels with a signal intensity above a threshold value and dividing by the number of cells (nuclei) in the same field. Data in Fig. 4B is expressed as the deposit area/cell in the presence or absence of lam5.

HFs or MFs were grown on glass coverslips (11 mm) at confluence (about 5 days) to establish ECM used as a dermal equivalent. The coverslips were transferred to KGM and 10⁴ HFKs were added to each coverslip, and incubations were continued at 37°C for the indicated times. Thrombin (1 unit/ml) with or without hirudin (2 units/ml) was added during or after co-culturing. Treatments after the appointed culture period were for 1 hour. Lam5 deposits were visualized with human specific Mabs described above, and fibronectin was stained with a rabbit polyclonal Ab against fibronectin (R790). Digital images, 6 μM thick, were captured in 0.2 μM optical sections and deconvoluted using Deltavision software. ELISA data (Fig. 5B) is the average and standard deviation of six wells for each time point.

Oligonucleotide primers for the isolation of a full-length cDNA clone of the human LAMA3 gene (GenBank accession #L34155) were designed based on the sequence of partial cDNA clones described previously (Ryan et al., 1994); 5′ primer (5′ CTGGTTCCGGCGAGATGCCTCCAGCAGTGAGG 3′), 3′ primer (5′ CTCGCTCCGGCGACTTGGGTTACTGGTCAGG 3′). Total RNA was isolated from log-phase HFKs. cDNA was reverse-transcribed using superscipt II (Gibco-BRL) and the 3′ primer. A PCR product consistent with a full-length cDNA of the α3 chain was obtained using taq DNA polymerase (Gibco-BRL) and cloned into an engineered bluescript vector (Stratagene) using ligation independent cloning (Haun and Moss, 1992). Point mutations in this 5 Kb clone were repaired by resection with restriction fragments from partial cDNAs. The corrected clone was inserted into the pcDNAIIzeo expression vector (Invitrogen) to yield pZeoα3, expressing human LAMA3 under control of the CMV promoter, and conferring zeocin resistance to transfected cells.

The noncleavable form of the α3 chain (Fig. 6) was obtained by deleting the sequence for 46 amino acids (S1308 to C1354) in the spacer region separating the G3 and G4 domains. The precleaved form was constructed by inserting five His codons and a stop codon after His 1364. The modifications were generated via oligonucleotide directed mutagenesis of appropriate restriction fragments using a proofreading polymerase (Deep Vent, New England Biolabs) for PCR. The modified restriction fragments were gel purified and spliced back into the pZeoα3 expression vector using standard techniques. All constructs were verified by DNA sequencing.

The mutated human laminin α3 expression vectors described above were transfected into the α3–/– MK cell line by electroporation. Zeocin resistant colonies (5 μg/ml) were only present in cultures receiving plasmid DNA during the electroporation. Zeocin-selected cells were further selected for their ability to adhere and survive on Petri dishes and finally on BSA-coated Petri dishes. Only those cells synthesizing and depositing a minimal level of lam5 were able to adhere and survive on these surfaces.

For preparation of ECM, keratinocytes were grown to confluence in multiwell Petri dishes (Falcon) and extracted with three changes of 20 mM ammonium hydroxide (Gospodarowicz, 1984). Collagen-coated control surfaces were prepared by incubating 10 μg/ml rat tail collagen (Becton-Dickinson) in 0.05 N HCl on Petri dish for 2 hours. All surfaces were blocked with heat-denatured BSA (5 mg/ml in PBS). Lam5 and lam6 were immobilized onto plastic surfaces using antibody trapping as described (Gil et al., 2002; Xia et al., 1996). Adhesion assays were performed as previously described (Carter et al., 1991) in 24-well Petri plates on ECM prepared by NH₄OH extraction or on collagen-coated surfaces.

Cells were plated at 10⁴ cells per well in 24-well Petri dishes (Fig. 8A) and counted every 24 hours thereafter, for 7 days using a Metamorph software package.

Duplicate plates were analyzed by ELISA for deposition of the human laminin α3 chain into the ECM (Fig. 8B). Cell-cycle analysis (Fig. 8D) was by flow cytometry, and this data was deconvoluted using Multicycle software.

Assembly of β4 integrin into Triton X-100 insoluble SACs was quantitated by ELISA and by immunofluorescent staining (Fig. 10). SACs were examined by immuno-fluorescence as described previously (Carter et al., 1990b) using anti-integrin β4 (P4G11), anti-human laminin α3 (P5H10) or rat Mab against mouse lam5 (P3C8). For ELISA, the Triton insoluble SACS were incubated with the same Mabs used for microscopy then incubated with peroxidase conjugated anti-mouse or anti-rat secondary antibodies. Optical densities were measured at 405 nm in a microplate reader. Results are presented as the average and standard deviation from triplicate wells.

Keratinocytes were grown to confluence (4 days) on Petri dishes. The cell layers were removed with 20 mM NH₄OH, and the remaining ECM was collected by scraping into 0.1% SDS and equal volumes were analyzed as shown in Fig. 4.

Adherent cells were labeled with ³⁵S-met and ³⁵S-cys (2 hours in 3.5 ml met-free KGM containing 100 μCi/ml). For pulse-chase studies, the labeling media was collected at the 0 hour time point, and the incubation was continued for 3, 8 or 24 hours in KGM containing 100 μg/ml BSA as a carrier protein. For studies with labeled ECM and/or conditioned medium labeled cells were incubated with 3.5 ml KGM containing 100 μg/ml BSA, and grown for 3 days. The conditioned media containing labeled secreted proteins was collected and protease inhibitors added (NEM, PMSF, 5 mM EDTA). The cell layer was extracted with 1% Triton X-100 in PBS containing protease inhibitors, then with 2 M urea, 1 M NaCl. The remaining ECM was collected into 8 M urea, 0.1% Triton X-100, and protease inhibitors in 0.05 N HCl, by scraping. Samples were neutralized with NaOH before immunoprecipitation. Labeled proteins were precipitated as previously described (Carter et al., 1991). Antibodies used (P5H10 and R5922, see Antibodies, above) were able to immunoprecipitate denatured antigens.

Pre-lam5 containing G4/5 in the α3²⁰⁰ chain was deposited into the PBM of epidermal wounds by leading keratinocytes in mice (Fig. 1). Precursor mouse lam5 was detected using rabbit monoclonal antibody R5808-f1 that recognizes a peptide epitope in G5 of mouse α3²⁰⁰ (Fig. 1a). Pre-lam5 was deposited into the PBM within 6-8 hours of injury and at 7 days (Fig. 1a). Detectable α3²⁰⁰ was lost by 14 days post injury (Fig. 1b), while lam5 remained detectable (Fig. 1d). This suggests that α3²⁰⁰ is proteolytically converted to α3¹⁶⁵ with removal of G4/5.

We wished to determine whether G4/5 in α3²⁰⁰ contributes to deposition and retention of pre-lam5 on the substratum. Keratinocytes secrete several soluble proteins (Katz and Taichman, 1999) including lam5, lam6, fibronectin, thrombospondin and an uncharacterized laminin, probably laminin 1 (Carter et al., 1991). However, lam5 is more efficiently deposited than the other secreted proteins and is therefore dominant in adhesion (Carter et al., 1991; Frank and Carter, 2003; Nguyen et al., 2000a). Therefore, we speculated that the G4/5 domain of precursor α3²⁰⁰, in addition to the γ2 chain, facilitates deposition and/or retention of pre-lam5 in the ECM.

First, we established a source of lam5 (α3²⁰⁰β3γ2) and lam6 (α3²⁰⁰β1γ1). HFKs and JEBG keratinocytes were metabolically labeled with ³⁵S-met. Labeled lam5 and lam6 were immunoprecipitated from conditioned media (Fig. 2). From HFKs, lam5 containing primarily cleaved α3¹⁶⁵ chain and small amounts of uncleaved α3²⁰⁰ were precipitated with Mab against α3 (C2-5) or γ2 (B4-6) (Fig. 2, HFK, –thrombin, lanes 2 and 3). Lam6 was also precipitated from HFK media with anti-α3 (C2-5), but not anti-γ2 (B4-6), and appeared to contain only uncleaved α3²⁰⁰ (Fig. 2, HFK, –thrombin, lane 2). The lam6 was confirmed to be uncleaved because it also precipitates with Mab D2-1 against G4/5 (Fig. 2, HFK, –thrombin, lane 1). From JEBG media, only lam6 containing α3²⁰⁰ was precipitated with anti-α3, or anti-G4/5 but not anti-γ2 (Fig. 2, JEBG, –thrombin). Therefore, the conditioned media of HFKs contains soluble lam6 with uncleaved α3²⁰⁰ chain and soluble lam5 containing primarily cleaved α3¹⁶⁵ chain with only small amounts of uncleaved α3²⁰⁰. By contrast, the conditioned media of JEBG cells contains lam6 with uncleaved α3²⁰⁰ and no lam5.

Next, a method was developed to selectively remove G4/5 from pre-lam5 and pre-lam6 containing α3²⁰⁰. In preliminary studies, we determined that thrombin, a serum protease (Backes et al., 2000), selectively cleaves α3²⁰⁰ at two candidate PRS cleavage sites. The potential cleavage sites at R1343 and R1389 localize to the linker sequence between G3 and G4 [see sequence in Ryan et al. (Ryan et al., 1999)]. After thrombin cleavage, lam5 and lam6 were immuno-precipitated with a Mab against the α3 chain (C2-5; Fig. 2, + thrombin, lane 2). The α3²⁰⁰ chain in both laminins 5 and 6 was cleaved to α3¹⁶⁵. Thrombin digestion released G4/5 as two bands of 37 and 35 kDa (+thrombin, lane 1, marked g4/g5) that immunoprecipitated with D2-1 and were indistinguishable from cleavage products made by the endogenous protease (–thrombin, lane 1). Significantly, there was no detectable cleavage in β3 or γ2 chains (Fig. 2) testifying to the specificity of the thrombin cleavage even at 10× the concentration of thrombin used in Fig. 2 (i.e. 10 NIH units/ml for 1 hour, results not shown). Hirudin, a thrombin inhibitor, blocked digestion (results not shown) confirming that thrombin was required for cleavage. Consistently, Mabs against thrombospondin (TSP) did not co-precipitate with any of the laminin chains and the thrombospondin was unaffected by the thrombin digestion (Fig. 2, lane 4). To the best of our knowledge this is the first report that thrombin selectively cleaves α3²⁰⁰ releasing G4/5 and α3¹⁶⁵.

Two assays were developed to compare deposition of lam5 and lam6 and how this affects adhesion. Soluble lam5 and 6 from HFK-conditioned media, or lam6 from JEBG-conditioned media (characterized in Fig. 2) were trapped on surfaces using immobilized Mabs and assayed for adhesion of the α3–/– MKs (Fig. 3A). Immobilized Mab against the α3 chain (P5H10) trapped both lam5 and lam6. Mab against the γ2 chain (B4-6) trapped only lam5 on the substratum. Control Mab did not trap adhesive activity. Both immobilized lam5 and 6 induced adhesion and spreading of the α3–/– MKs. This indicates that soluble lam6 from JEBG keratinocytes is inherently adhesive like lam5, if trapped on the substratum, and probably would be adhesive if deposited.

Next, we determined if α3–/– MKs could utilize either exogenous soluble lam5 or 6 to deposit an adhesive ECM (Fig. 3B). The α3–/– MKs were adhered to glass coverslips then the surface was blocked with BSA. The adherent but round α3–/– MKs were incubated with conditioned media from HFKs or JEBG keratinocytes. The soluble lam5 in HFK-conditioned culture media caused cell spreading of the α3–/– MKs within 1 hour (Fig. 3Ba). By contrast, α3–/– MKs given JEBG-conditioned media remained round (Fig. 3Bb). Therefore, exogenous soluble lam5, but not lam6, promoted spreading of α3–/– MKs. Significantly, fibronectin, laminin 1 and thrombospondin in the JEBG-conditioned media also failed to induce cell spreading. In the absence of soluble lam5, the α3–/– MKs remained round (Fig. 3Bb,d). The round cells were overexposed so they could be seen on the substratum (Fig. 3Bd). The fluorescence of the round cells in Fig. 3Bd was also useful in Fig. 4Aa-c to detect the round cells that did not spread or deposit soluble lam5 (see next section).

Significantly, the spread α3–/– MKs had deposited the human lam5 onto the substratum in the characteristic flower pattern for lam5 (Carter et al., 1991). This deposition is consistent with results reported by Baker et al. (Baker et al., 1996). The deposited lam5 reacted with anti-α3 Mab (P5H10; Fig. 3Bc) and anti-G4/5 specific Mab (D2-1; see Fig. 4d-f, next section). This suggested that exogenous soluble pre-lam5, not cleaved lam5, was selectively bound and deposited by the α3–/– MKs. In controls, the deposited pre-lam5 also reacted with anti-α3 Mabs (Fig. 3Bc) and anti-human γ2 chain Mabs (see Fig. 4). As expected, no lam6 deposition could be detected in the presence of JEBG-conditioned media (Fig. 3Bd). This indicates that soluble lam6 is not deposited even though it is inherently adhesive when immobilized (Fig. 3A). In contrast to the α3–/– MKs, adherent NIH3T3, CHO or HT-1080 cells did not deposit the soluble human lam5 (results not shown). This suggested that cells that can adhere and spread on exogenous lam5 may lack cell binding required for deposition.

We determined whether the removal of G4/5 from soluble pre-lam5 would inhibit cell binding and/or deposition by the α3–/– MKs (Fig. 4). Soluble pre-lam5 was affinity purified from HFK-conditioned media by chromatography on immobilized heparin (Nguyen et al., 2000a). Pre-lam5 was digested with or without thrombin to remove G4/5, as seen in Fig. 4C. The cleaved and noncleaved lam5 preparations were inclubated with the α3–/– MKs adhered to glass coverslips (Fig. 4A). α3–/– MKs bound exogenous soluble human pre-lam5, deposited it onto the substratum and spread on the deposits, (Fig. 4Ad-f). Deposition and spreading was eliminated by the removal of G4/5 with thrombin (Fig. 4Ag-i). The deposits of pre-lam5 were detected with anti-γ2 (B4-6), anti-α3 (C2-5) and anti-G4/5 Mab (D2-1). Visual data was confirmed by quantitation of the deposits by morphometric analysis (Fig. 4B). In controls, inclusion of hirudin, a thrombin inhibitor, blocked removal of G4/5 and prevented the effects of thrombin (Fig. 5, next section). Treatment of the deposits with thrombin subsequent to deposition removes G4/5 staining but does not remove lam5 in deposits detected by staining with anti-α3 or -γ2 Mabs (Fig. 5, next section). Thus, G4/5 is required for binding and deposition, but not retention, of soluble precursor human lam5 onto a culture dish by α3–/– MKs.

We determined whether pre-lam5 could deposit onto a dermal equivalent and if the deposition required G4/5 (Fig. 5). For this purpose we established an in vitro model for dermis. Primary mouse fibroblasts (MFs) or human fibroblasts (HFs) were grown to confluence to assemble a dense ECM containing at least fibronectin, tenascin and collagen types I, III and VI assembled in fibrils (Carter, 1982) but lacking lam5 (Carter et al., 1991). HFKs were seeded at low density onto the apical surface of the dermal equivalent. HFKs attached in colonies on the apical surface of the MFs ECM and were distinguished from fibroblasts by staining with Mabs against either epithelial specific antigens (β4 integrin or lam5) or human-specific antigens (CD44) or by the appearance of DAPI-stained nuclei. Keratinocyte nuclei (arrowheads, Fig. 5A) display uniform blue staining with DAPI while fibroblast nuclei also display blue staining of nucleoli in spots. The HFKs deposited a prominent PBM at the interphase with the MF ECM that stained with anti-pre-lam5 Mab (D2-1; Fig. 5Ad,e,f) or α3 chain (P5H10; Fig. 5Aa,b,c) or γ2 chain (B4-6, not shown). At the light microscope level, the deposited pre-lam5 (Fig. 5Aa, green) was co-linear with fibronectin fibrils (Fig. 5Aa, red, see inset) or tenascin fibrils (not shown). This suggested that lam5 interacts directly or indirectly with one or more components of the fibrils. In controls (results not shown), deposits of pre-lam5 co-aligned with MF ECM that had been de-cellularized by extraction with Triton X-100 detergent. Significantly, pre-lam5 deposits co-aligned with integrin α6β4 in hemidesmosme-like SACs of the HFKs (Fig. 5Ba-c). This indicates that pre-lam5 deposits were restricted to the points of contact of the HFKs with the MF ECM. The deposits of lam5 formed a `nest' surrounding the colony of HFKs and was not deposited over the MFs or ECM distant to the HFKs.

Significantly, thrombin digestion during co-culture (Fig. 5Ac,f), but not after deposition (Fig. 5Ab,e), reduced deposition. ELISA quantitation of deposited lam5 using anti-γ2 chain Mab (B4-6) (Fig. 5C) confirmed that thrombin decreased deposition. Hirudin, a specific inhibitor of thrombin, partially blocked the effects of thrombin (Fig. 5C). Thrombin efficiently inhibited deposition of lam5 when detected by fluorescence microscopy (Fig. 5A) but only partially blocked the lam5 deposition when detected by ELISA (Fig. 5C). This is because the ELISA assay does not distinguish pre-lam5 in the cytoplasm of HFKs that is resistant to thrombin, from lam5 in the PBM. We conclude that HFKs deposit a PBM of pre-lam5 at the junction with a dermal equivalent. Removal of G4/5 prevents deposition, but not retention, of lam5.

We used recombinant forms of lam5 to further evaluate the role of G4/5 in deposition. A modified form of human α3 chain was expressed in mouse α3–/– MKs (Fig. 6). `Precleaved α3' (α3<sup>165</sup>*) was designed to mimic the α3 chain after proteolytic removal of G4/5. The sequence for residues H1364 to the stop codon was removed and replaced with the sequence for a six-histidine tag. `Noncleavable' α3 (α3²⁰⁰*) contains an internal deletion, with residues S1308 to C1354 removed. This deletion produces a G-domain resembling the laminin α1 chain, which is not cleaved (Timpl et al., 2000). The deletion removes one of the thrombin cleavage sites at R1343, and removes a cleavage site at Q1337 for endogenous protease (Tsubota et al., 2000).

Before using the recombinant laminin 5 forms for evaluating the role of G4/5 in deposition, it was necessary to evaluate their function in assembly, cell proliferation, adhesion and signaling.

Lam5 deposited by transfected α3–/– MKs was characterized by western blotting of ECM (Fig. 7). α3–/– MKs transfected with the noncleavable human α3²⁰⁰* chain expressed a 195 kDa human α3 chain detected with a Mab (P5H10) specific for human α3 chain (Fig. 7A, lane 5). For comparison, the wild-type α3²⁰⁰ was detected in CCM of HFKs (lane 2). Endogenous protease failed to remove G4/5 from the recombinant α3²⁰⁰*. A human α3 chain at 165 kDa was detected in the ECM from α3–/– MKs expressing the precleaved human α3¹⁶⁵* chain (lane 6). In controls, no human α3 chain was detected in the ECM from either nontransfected α3–/– MKs (lane 3) or wild-type α3+/+ MKs (lane 4). A polyclonal antibody against both mouse and human laminin β3 chain detected the β3 chain in all samples except the α3–/– ECM (Fig. 7B, lane 3). In Fig. 7C, uncleaved mouse α3²⁰⁰ chain was detected only in ECM from α3+/+ MKs by an antibody that recognizes only G4/5 domains (R5808-f1) in the mouse laminin α3²⁰⁰ chain (lane 4). We conclude that expression of recombinant human α3¹⁶⁵* or α3²⁰⁰* α3–/– MKs is sufficient to restore assembly and deposition of endogenous chimeric human/mouse lam5.

The α3–/– MKs do not deposit lam5 and as a result require an exogenous ligand, such as lam5 or collagen I, to adhere and proliferate on a polystyrene Petri dish (Ryan et al., 1999). When the α3–/– MKs express a human α3 chain, they are able to adhere and proliferate on a Petri dish in the absence of exogenous ligand (Fig. 8). The α3–/– MKs expressing human α3¹⁶⁵* or α3²⁰⁰* or the α3+/+ MK adhered and proliferated. The nontransfected α3–/– MKs were unable to increase cell numbers. The two α3 transfected cell lines grew with nearly identical kinetics; however, significantly greater amounts of noncleavable lam5 were deposited compared with precleaved (Fig. 8B). We will return to this difference later.

Phase contrast microscopy of the MKs remaining attached to Petri dishes after 7 days of culturing (Fig. 8C) showed that the α3+/+ cells and the human α3 transfected α3–/– cells were confluent and spread. The α3–/– MKs were sparse, round and weakly attached. The α3–/– cells were still viable, however, as they could re-plate onto collagen coated dishes (results not shown). Cell cycle profiles indicated that α3–/– MKs have a low proportion of cells in S phase compared with α3+/+ MKs, and α3–/– MKs transfected with human α3 chains (Fig. 8D). Thus, expression of human α3 in α3–/– MKs followed by assembly and deposition of lam5 partially restores adhesion, cell cycle progression and cell replication.

Adhesion assays were performed to determine which integrins ligate the precleaved and noncleavable lam5 deposited by transfected α3–/– MKs (Fig. 9). Lam5 deposited by the transfected α3–/– MKs, but not α3–/– MKs, supported adhesion of HFKs in 30 minute adhesion assays. Adhesion to the deposited chimeric lam5 and to lam5 deposited by HFKs is inhibited by Mab against human laminin α3 chain (C2-9; Fig. 9) but not mouse lam5 deposited by α3+/+ MKs. Adhesion to the deposited recombinant ECM was inhibited by Mab against the integrin α3 subunit (P1B5), and integrin β1 subunit (P4C10). Significantly, adhesion and inhibitor results were not affected by the presence of the G4/5 domains in the recombinant lam5. Adhesion to deposited lam5 was only partially inhibited by antibodies to α3 and β1 integrins because integrin α6β4 also contributes to adhesion (Xia et al., 1996) (see next section). Adhesion to collagen I, but not any of the lam5 substrates, was blocked by Mab against integrin α2 (P1H5). We conclude that adhesion via integrins α3β1 is indistinguishable on noncleavable or precleaved lam5.

The α3–/– MKs transfected with noncleavable (Fig. 10Ae-f and i-j) or precleaved (g-h and k-l) human α3 chain or α3+/+ MKs (panel b) each deposited lam5 that codistributed with integrin β4 in Triton-insoluble hemidesmosome-like SACs. Lam5 is not detected (panel c) and Triton-resistant β4 integrin (panel d) is absent in α3–/– MKs on collagen. We conclude that precleaved and noncleavable lam5 both promote the assembly of α6β4 in SACs.

We determined whether deposits of noncleavable and precleaved lam5 that rescued adhesion also rescued transmembrane cell signals. α3–/– MKs or HFKs were plated onto surfaces of deposited recombinant lam5 for 30 minutes. Extracts of the cells were blotted with Abs against focal adhesion kinase that is phosphorylated on tyrosine 397 (FAK PY-397; Fig. 11). FAK PY-397 was present in α3–/– MKs adhering to recombinant precleaved or noncleavable lam5, but not in suspended cells or cells plated onto a nonadhesive BSA blocked surface or cells plated onto ECM from α3–/– MKs (Fig. 11). Similar results were obtained with HFKs adhering to the same surfaces. In controls, FAK was also phosphorylated in α3–/– MKs or HFKs plated on collagen I, on ECM from HFKs or ECM from α3+/+ MKs.

Here we used the recombinant lam5 characterized above to evaluate the role of G4/5 in secretion and deposition of lam5. α3–/– MKs expressing human α3²⁰⁰* or α3¹⁶⁵* chains were metabolically labeled with ³⁵S-Met. The adherent cells were removed with ammonium hydroxide and adherent ECM solubilized in urea and immunoprecipitated with the indicated antibodies (Fig. 12A). Noncleavable lam5 containing α3²⁰⁰* was readily detected in the ECM deposits (Fig. 12A lane 3). By contrast, relatively little precleaved lam5 containing α3¹⁶⁵* was detected in the ECM (Fig. 12A lane 5). Clearly, lam5 containing α3¹⁶⁵* was deposited sufficiently to rescue adhesion and survival but less than lam5 containing α3²⁰⁰* (Fig. 8B). Assays in Figs 8 and 12 are quite different. In Fig. 12, it was necessary to solubilize the ECM deposits with urea in a form that could still be immunoprecipitated. In controls, immunoprecipitation of the ECM extracts (Fig. 12A) with anti-β1γ1 antibody (R5922) failed to detect deposited lam6 or presumptive laminin 1. However, presumptive laminin 1 was immunoprecipitated with the anti-β1γ1 antibody (R5922) and was secreted into the conditioned culture medium by α3–/– MKs expressing α3²⁰⁰* (Fig. 12B lane 2) and α3¹⁶⁵* (Fig. 12B, lane 3) or untransfected α3–/– (results not shown). This indicated that transfected cells were expressing similar levels of other secreted laminins and proteins, as untransfected cells, even though they were not being deposited.

To corroborate the findings in Fig. 12A, we determined whether the differences in deposition of noncleavable and precleaved lam5 were reflected in levels of secreted lam5. MKs expressing precleaved α3¹⁶⁵* or noncleavable α3²⁰⁰* were pulse-labeled with ³⁵S-Met and chased for the indicated times (Fig. 12C). MKs expressing α3²⁰⁰* chain secreted almost no lam5 but relatively large amounts of lam6 (Fig. 12C, noncleavable). By contrast, MKs expressing α3¹⁶⁵* secreted primarily lam5 and no lam6 (Fig. 12C, precleaved). It is also apparent from Fig. 12C, that noncleavable α3²⁰⁰*, but not α3¹⁶⁵*, assembles into secreted lam6. This point will be the subject of a separate study (R.O.S., unpublished). In Fig. 12D, quantitation of deposited verses secreted human α3 chain in noncleavable and precleaved lam5 confirmed the qualitative results from Fig. 12A. The data shows that noncleavable lam5 is efficiently deposited while precleaved lam5 is poorly deposited but efficiently secreted. It should be pointed out that most of the noncleavable human α3 chain is secreted as lam6, as shown in Fig. 12B. As a control, secreted laminin β1 and γ1 chains were quantitated to assure that the α3–/– MKs transfected with noncleavable or precleaved human α3 chain were secreting comparable levels of lam1 while depositing different levels of lam5. For the study here, we conclude that noncleavable lam5 is efficiently deposited but poorly secreted, whereas precleaved lam5 is efficiently secreted but poorly deposited.

The results establish, for the first time, that the G4/5 domain of the α3²⁰⁰ chain facilitates deposition of pre-lam5 in two culture models for assembly of the PBM. This conclusion is based on the following. (1) Pre-lam5 is deposited into the PBM of epidermal wounds before other BM components with which lam5 interacts. After deposition, α3²⁰⁰ is cleaved to the α3¹⁶⁵ form releasing g4/5 (Fig. 1). (2) Exogenous soluble pre-lam5 was bound and deposited by α3–/– MKs (Fig. 3). The deposition was prevented by removal of G4/5 with thrombin before binding. However, thrombin did not detach lam5 after deposition, suggesting that G4/5 is not required for retention (Figs 4 and 5). (3) HFKs cultured on MF ECM deposited endogenous pre-lam5 along fibronectin fibrils analogous to deposition on dermis in wounds. Deposition of pre-lam5 into the PBM was blocked by thrombin digestion during, but not after, deposition (Figs 5 and 6). (4) Expression of recombinant precleaved or noncleavable human α3 chain in α3–/– MKs rescued assembly and deposition of lam5 sufficient for adhesion via both integrins α3β1 and α6β4. Adhesion rescued phosphorylation of FAK and survival. However, noncleavable lam5 was efficiently deposited but not secreted, whereas precleaved lam5 was efficiently secreted but not deposited. We conclude that the G4/5 domain of α3²⁰⁰ facilitates cell binding and deposition of pre-lam5. We suggest that G4/5 facilitates deposition of pre-lam5 into the PBM of epidermal wounds and that the deposition is required for subsequent cell adhesion, signaling and survival.

Recently we have shown that pre-lam5 deposited onto collagen, localized into `deposition contacts' at the rear of migrating HFKs and is necessary to generate polarized and linear migration of keratinocytes, termed processive migration (Frank and Carter, 2003). These deposition contacts localize to sites of collagen remodeling/removal. After deposition, the pre-lam5 ligates integrin α3β1 to mediate adhesion and processive migration. Processive migration via α3β1 is followed by a switch to integrin α6β4 in SACs correlating with the end to migration. Deposition and subsequent adhesion do not occur if HFKs are treated with microtubule poisons (Frank and Carter, 2003). The results presented here indicate that G4/5 facilitates cell binding to an unidentified receptor(s) and deposition of pre-lam5 before ligation of α3β1 and processive migration.

Noncleavable α3²⁰⁰* chain assembled into recombinant lam5 and lam6. However, lam5 was deposited, and lam6 was secreted (Fig. 12). This suggests an important role for the β3 or γ2 chains in deposition as reported (Gagnoux-Palacios et al., 2001). However, when lam5 was assembled with precleaved α3¹⁶⁵*, it was deposited inefficiently but efficiently secreted (Fig. 12C). This suggested that the γ2 chain was insufficient for deposition of lam5 and that G4/5 was also necessary. Consistently, α3–/– MKs bound and deposited exogenous soluble pre-lam5, but not soluble pre-lam6 or cleaved lam5. In the absence of G4/5, precleaved lam5, or thrombin-digested lam5, were soluble and not efficiently deposited. However, precleaved is deposited at least on plastic dishes in culture sufficiently to rescue adhesion and survival. Together, the results suggest that the mechanism of deposition and retention of lam5 is complex and may involve multiple receptors or ligand-ligand interactions. For example, the binding and deposition of exogenous soluble pre-lam5 was accomplished by the α3–/– MKs, but not HT1080, CHO, NIH 3T3, HF or MF cells. HT1080 cells can adhere to exogenous lam5 via α3β1 (Carter et al., 1991) or syndecan 1 (Okamoto et al., 2003) but are not competent for binding or deposition of exogenous pre-lam5. This suggests that neither α3β1 nor syndecan 1 are sufficient for binding and deposition of pre-lam5. Consistently, neither null defects in integrins α3 (DiPersio et al., 1997), β1 in epidermis (Grose et al., 2002; Grose and Werner, 2003; Raghavan et al., 2003), nor syndecan 1 (Stepp et al., 2002) nor 4 (Echtermeyer et al., 2001) have reported defects in lam5 deposition, only in organization of deposits. By contrast, the G4/5 domains of the laminin α2 (Li et al., 2002) and α3 chains appear to play similar roles in cell binding or deposition: The G4/5 domains may participate in laminin deposition by interacting with a nonintegrin receptor(s). Only after deposition does interaction with integrins facilitate adhesion (Frank and Carter, 2003). Our studies did not evaluate the interactions that retain lam5 along the fibronectin fibrils (Fig. 5) nor in the provisional BM of the wound (Fig. 1). Conceivably, the γ2 chain of lam5 may play a role in retention by interacting with dermal ligands while G4/5 interacts with a cell receptor involved in binding or deposition. Significantly, thrombin digestion that removed G4/5 and inhibited deposition, did not cleave the γ2 chain, an event reported to inhibit deposition of laminin 5 (Gagnoux-Palacios et al., 2001).

Lam5 interacts with integrin α6β4 in hemidesmosome-like stable anchoring contacts (SACs) (Carter et al., 1990b) and hemidesmosomes (Ryan et al., 1999) and with α3β1 in adhesion and motility (Carter et al., 1991; Frank and Carter, 2003). It has been suggested that cleaved lam5 is necessary for the nucleation of hemidesmosome-like structures in epithelial cell lines (Goldfinger et al., 1998). The G3 domain is reported as the major binding site for integrins but mutation studies failed to distinguish separate binding sites for α3β1, α6β1 or α6β4 integrins (Hirosaki et al., 2000; Hirosaki et al., 2002). Here, noncleavable and precleaved lam5 both promoted the assembly of α6β4 in SACs and cell spreading and signaling via α3β1. The results (Figs 10 and 11) do not disprove the proposed role of G4/5 as a regulator of α3β1 verses α6β4 interactions; however, the current studies found no evidence for such regulation. In an alternative view, studies have raised the possibility that changes in cell signaling, not the proteolytic removal of G4/5, may regulate functional switching between α3β1 and α6β4 (Frank and Carter, 2003; Mercurio et al., 2001; Nguyen et al., 2000a; Nguyen et al., 2001).

Proteolytic cleavage of the G4/5 by endogenous protease was prevented by removal of a 46 amino acid sequence in the spacer region separating the G3 and G4 domains. Modeling of this region, based on the crystal structure of the laminin α2 G4/5 domains, predicts an exposed, flexible loop, easily accessible to proteases (Timpl et al., 2000). A variety of proteases can cleave the α3²⁰⁰ to α3¹⁶⁵ chain in this region, including mammalian tolloid (Veitch et al., 2003), plasmin (Goldfinger et al., 1998) and thrombin (Fig. 2). Thrombin selectively converts α3²⁰⁰ to α3¹⁶⁵ with release of G4/5 from both pre-lam5 and pre-lam6 (Fig. 2). Whatever the identity of the protease(s) responsible for G4/5 cleavage in cultured keratinocytes, deletion of the spacer region eliminates G domain cleavage without inactivating deposition or adhesive function. Conceivably, removal of the 46 amino acid spacer may generate conformational changes that disrupt interactions of G4/5 with G1-3 that mimic proteolytic removal of G4/5. For example, deletion of the spacer between G3 and G4 may expose binding sites in G1-3 for integrin α6β4. By analogy, recombinant lam6 reportedly requires cleavage of the G domains to function as an adhesive ligand (Hirosaki et al., 2002). However, our results indicate that unprocessed pre-lam6 will support cell adhesion and spreading but only when immobilized on a surface (Fig. 3).

Reasonably, the α3–/– MKs, and the recombinant precleaved and noncleavable lam5 provide novel tools to evaluate the coupling between deposition of pre-lam5 and processive migration via α3β1 (Frank and Carter, 2003). Further, we identified a novel cell surface receptor, p80/gp140/CDCP1, whose tyrosine phosphorylation is responsive to the deposition and adhesion to lam5 (Brown, 2004; Xia et al., 1996). The use of these tools may facilitate identification of a receptor that mediates binding and deposition of pre-lam5 onto a dermal equivalent.

This work was supported by: National Institutes of Health Grants CA49259 (WGC) and DK59221 (JO/WGC); University of Washington Engineered Biomaterials (UWEB) NSF EEC 9529161UWEB (WGC); and BioStratum Inc. Durham NC (WGC). Medical Student Research Training Grant from the NIDDK (MB).