Regulation of Biological Activity and Matrix Assembly of Laminin-5 by COOH-terminal, LG4–5 Domain of α3 Chain*

The basement membrane protein laminin-5 (LN5; α3β3γ2) undergoes specific proteolytic processing of the 190-kDa α3 chain to the 160-kDa form after the secretion, releasing its COOH-terminal, LG4–5 domain. To clarify the biological significance of this processing, we tried to express a recombinant precursor LN5 with a 190-kDa α3 chain (pre-LN5), in which the cleavage sequence Gln-Asp was changed to Ala-Ala by point mutation. When the wild-type and mutated LN5 heterotrimers were expressed in HEK293 cells, the wild-type α3 chain was completely cleaved, whereas the mutated α3 chain was partially cleaved at the same cleavage site (Ala-Ala). pre-LN5 was preferentially deposited on the extracellular matrix, but this deposition was effectively blocked by exogenous heparin. This suggests that interaction between the LG4–5 domain and heparan sulfate proteoglycans on the cell surface and/or extracellular matrix is important in the matrix assembly of LN5. Next, we purified both pre-LN5 and the mature LN5 with the processed, 160-kDa α3 chain (mat-LN5) from the conditioned medium of the HEK293 cells and compared their biological activities. mat-LN5 showed higher activities to promote cell adhesion, cell scattering, cell migration, and neurite outgrowth than pre-LN5. These results indicate that the proteolytic removal of LG4–5 from the 190-kDa α3 chain converts the precursor LN5 from a less active form to a fully active form. Furthermore, the released LG4–5 fragment stimulated the neurite outgrowth in the presence of mat-LN5, suggesting that LG4–5 synergistically enhances integrin signaling as it is released from the precursor LN5.

The basement membrane proteins laminins plays essential roles in both tissue construction and regulation of cellular functions. The laminin molecules consist of the three subunits (or chains) of ␣, ␤, and ␥, linked by disulfide bonds to form the well known cross-shape structure. To date, more than 16 lami-nin isoforms with different combinations of five ␣, three ␤, and three ␥ chains have been identified (1)(2)(3). Each laminin chain contains many functional domains, which allow the laminins to interact with various molecules in the extracellular matrix (ECM). 1 For example, laminin ␣ chains (␣1 to ␣5) contain a large globular domain (laminin globular domain, LG domain) in their COOH termini, which is divided into five homologous subdomains (LG1-LG5). The LG domain is thought to be a major site to interact with specific receptors on cell surface, such as integrins, syndecans, and ␣-dystroglycan, whereas the NH 2 -terminal regions of the ␣, ␤, and ␥ chains contain functional domains that are mainly involved in the matrix assembly (1). Because of these complex structures, partial proteolysis of laminins alters their biological activities.
One of the laminin isoforms, laminin-5 (␣3␤3␥2; LN5), was originally found as an anchoring filament component of keratinocyte (4,5) and as a cell scatter factor secreted by gastric carcinoma cells (6). LN5 has strong activities to promote cellular adhesion, motility, and cell scattering in culture (7,8). These activities are mediated by integrins ␣3␤1, ␣6␤4, and ␣6␤1 (5,7,9). The LG domain of ␣3 chain contains several integrin-binding sites (10), and the LG3 domain plays a critical role in the expression of the unique activities of LN5 (9,11,12). The association of LN5 with integrin ␣6␤4 is essential for the stable hemidesmosome structure (13), and thereby it supports the stable anchorage of epidermal keratinocytes to the underlying connective tissues in vivo (14). On the other hand, the cell migration-promoting activity of LN5 is thought to contribute to wound healing (15).
LN5 is synthesized and secreted as a precursor form consisting of a 190-kDa ␣3 chain, a 135-kDa ␤3 chain, and a 150-kDa ␥2 chain. After secretion, the ␣3 and ␥2 chains undergo specific extracellular proteolytic processing to convert to the mature form containing a 160-kDa ␣3 chain and a 105-kDa ␥2 chain (16,17). Much attention has been focused on the physiological consequence of this processing of LN5. Laminin ␥2 chain is cleaved at its NH 2 -terminal region by BMP-1/mTLD proteinases (18,19). In rat LN5, the cleavage of the 150-kDa ␥2 chain to a 80-kDa form by gelatinase A (matrix metalloproteinase-2) or membrane type-1-MMP elevates its cell motility activity (20,21). Similarly, the cleavage of the 150-kDa ␥2 chain to a 105-kDa form in human LN5 increases its cell motility activity but decreases its cell adhesion activity (22). The proteolytic proc-essing of the ␣3 chain occurs much more efficiently than that of the ␥2 chain. Therefore, it is difficult to detect the precursor LN5 with the unprocessed ␣3 chain in the conditioned medium and ECM of LN5-producing cells (17). We recently showed that the 190-kDa ␣3 chain of LN5 is cleaved between the LG3 and LG4 domains, releasing an LG4 -5 fragment (11,23). However, the physiological meaning of the ␣3 chain cleavage is controversial. It was reported that the cleavage of the 190-kDa ␣3 chain to a 160-kDa form decreases the cell motility activity of LN5 and promotes stable cell anchorage (24). The mature LN5 with the 160-kDa ␣3 chain induces the nucleation of hemidesmosome assembly through interaction with integrin ␣6␤4 (13). When the skin is injured, the precursor LN5 with the unprocessed ␣3 chain is deposited in the provisional matrix by leading keratinocytes, which appears to support migration of following cells toward the wound edge (15,24,25). To the contrary, other studies have shown that the LN5 with the processed ␣3 chain supports rapid migration of various types of cells (6,7,9,26). In a case of laminin-6 (LN6; ␣3␤1␥1), the same cleavage of the ␣3 chain activates this laminin, increasing both the cell adhesion and cell migration activities (27). Thus, the biological significance of the ␣3 chain cleavage still remains to be unclear. No past studies have directly compared the functional difference between the processed and unprocessed LN5 forms, using the purified proteins. In this study, therefore, we prepared a recombinant LN5 with an unprocessed ␣3 chain and compared the biological activities of the two LN5 forms with the processed or unprocessed ␣3 chain.

MATERIALS AND METHODS
Cell Culture-LN5-HEK and ␤3␥2-HEK cell lines were previously established by transfecting HEK293 cells with the cDNAs of human laminin ␣3, ␤3, and ␥2 chains and with those of the ␤3 and ␥2 chains, respectively (28). LN5-HEK cells secrete the human recombinant LN5 heterotrimer at a high level. The non-tumorigenic epithelial cell line (BRL) derived from Buffalo rat liver has been used in previous studies (6,23). Human fibrosarcoma cell line HT1080 was obtained from the Japanese Cancer Resource Bank (Tokyo, Japan). These cell lines were maintained in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium, DMEM/F-12 (Invitrogen), supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin G, and 0.1 mg/ml streptomycin sulfate. A spontaneously immortalized human keratinocyte cell line, Ha-CaT (29), was a generous gift from Dr. N. E. Fusenig (Deutsches Krebsforschungszentrum, Heidelberg, Germany), and maintained in DMEM (Nissui, Tokyo, Japan) supplemented with 10% FCS. PC12/HS, a clone of PC12 rat pheochromocytoma cells that responds to nerve growth factor (NGF) with a high sensitivity, was kindly given by Dr. Hiroshi Iizuka (30) and cultured in RPMI 1640 medium (Nissui) supplemented with 10% horse serum and 5% FCS on plastic dishes precoated with 30 g/ml of type-I collagen (Koken, Tokyo, Japan).
cDNA Construction and Transfection-Human laminin ␣3 chain (␣3A) cDNA has been cloned from a cDNA library of gastric cancer cells (31). To convert two amino acid residues at the processing site of laminin ␣3 chain, Gln 1337 and Asp 1338 , to Ala, the site-directed mutagenesis was performed by the inverted PCR method reported by Imai et al. (32). The EcoRI/SmaI fragment of the ␣3 cDNA (nucleotides 3293-4406) was cloned into the EcoRI/HincII sites of pGEM-3Zf (ϩ) and used as a template. The whole sequence of the template was amplified with a mutagenesis primer pair designed in the inverted tail-to-tail direction. The sequence of primers used were as follows: the primer ␣3/AA-3Ј (sense, 4010 -4031), 5Ј-ccGcCACACCAGTGGCCTC-CCC-3Ј; and the primer ␣3/AA-5Ј (antisense, 3983-4009), 5Ј-ctAG-CAGCTGGTTGATACGAAAAGTCT-3Ј, in which the small letters indicate the base substitution to convert both Gln 1337 and Asp 1338 to Ala, and the underlined letters indicate additional base substitution to introduce a unique NheI site for rapid screening of mutated clones. The amplified linear DNA was self-ligated to transform Escherichia coli. The mutated clones thus created were screened by NheI digestion, and these sequences were verified with DNA sequence analysis. The ClaI/ ApaI fragment of laminin ␣3 cDNA (3730 -4403) in pGEM-LS/CX (11) was replaced with the corresponding region of the mutated fragment to construct the full-length cDNA of mutant ␣3 chain (␣3AA) (pGEM-LS/ CX/AA). The full-length cDNA of ␣3/AA chain was released with XbaI from pGEM-LS/CX/AA and cloned into the XbaI site of the pcDNA3.1-Hygro(ϩ) mammalian expression vector (Invitrogen) to make ␣3/AA-pcDNA3.1-Hygro. The ␣3/AA-pcDNA3.1-Hygro was transfected into ␤3␥2-HEK cells as described previously (28). The ␣3/AAtransfected ␤3␥2-HEK cells were cloned and expanded in 10% FCS-DMEM/F-12 containing 500 g/ml Geneticin (Sigma), 300 g/ml Zeocin (Invitrogen), 100 g/ml hygromycin (Wako, Osaka, Japan). One of the clones highly expressing the introduced ␣3/AA chain was used in this study (hereafter referred as LN5AA-HEK).
Preparation of Conditioned Media and ECMs-HEK cells expressing recombinant LN5 were grown to confluence in 90-mm culture dishes with 10% FCS-DMEM/F-12. Confluent cultures were washed twice with Ca 2ϩ -Mg 2ϩ -free phosphate-buffered saline (PBS) and further incubated in serum-free DMEM/F-12 for 2 days. The resultant serum-free conditioned media were collected, dialyzed against pure water, and concentrated by lyophilization. After collecting the serum-free conditioned media, the cells in the dishes were lysed with 20 mM NH 4 OH (24), and the remaining ECM components on the dishes were washed with PBS and then solubilized in a sample buffer containing 3% SDS in 100 mM Tris-HCl (pH 6.8). To examine the effect of heparin on LN5 deposition, LN5AA-HEK cells were seeded at a density of 2 ϫ 10 5 cells per 35-mm culture dish in 10% FCS-DMEM/F-12. After incubation overnight, the media were changed to fresh serum-free DMEM/F-12 with or without heparin (Wako) at various concentrations, and further incubated for 2 days. The resultant conditioned media and ECMs were prepared as described above.
SDS-PAGE and Immunoblotting Analyses-SDS-PAGE was performed on 6% gels under reducing conditions or 4 -7.5% gradient gel under non-reducing conditions. In the case of purified proteins, separated proteins were stained with a Wako silver staining kit II (Wako). For immunoblotting, proteins resolved by SDS-PAGE were transferred onto polyvinylidene difluoride membranes. The three subunits of LN5 were detected with specific antibodies using ECL-kit (Amersham Biosciences). The monoclonal antibodies used are a mouse monoclonal antibody against the NH 2 -terminal region of human laminin ␣3 chain (LS␣c1), a mouse monoclonal antibody against human laminin ␤3 chain (Anti-Kalinin B1) (BD Transduction Laboratory), and a mouse monoclonal antibody against domain III of human laminin ␥2 chain (D4B5). LS␣c1 and D4B5 were prepared in our laboratory.
Purification of Two LN5 Forms-Serum-free conditioned medium of LN5AA-HEK cells was prepared and fractionated by molecular-sieve chromatography on a Sepharose 4B column (2.6 ϫ 98 cm; Amersham Biosciences) according to the previously reported method (28) with some modifications. The collected conditioned medium was previously incubated with 2 mM EDTA and 2 mM phenylmethylsulfonyl fluoride to inactivate proteinase activities, and the column was previously equilibrated with 50 mM Tris-HCl (pH 7.5) buffer containing 0.1% CHAPS, 0.01% Brij 35, and 1 mM EDTA (TCBE buffer) supplemented with 500 mM NaCl and 0.5 mM phenylmethylsulfonyl fluoride. The fractions containing recombinant LN5 were pooled, concentrated by ultrafiltration, diluted with TCBE buffer to decrease the NaCl concentration below 0.25 M, and then applied to a heparin-Sepharose CL-6B column (2.6 ϫ 10 cm; Amersham Biosciences) pre-equilibrated with TCBE buffer supplemented with 0.25 M NaCl and 0.5 mM phenylmethylsulfonyl fluoride. Bound proteins were sequentially eluted with 0.4, 0.5, and 1.0 M NaCl. The recombinant precursor LN5 with the 190-kDa ␣3 chain (pre-LN5) was purified from the 1.0 M NaCl fraction by immunoaffinity chromatography with the anti-laminin-␣3 monoclonal antibody (LS␣3c4) as described previously (28). Similarly, the mature LN5 with the 160-kDa, processed ␣3 chain (mat-LN5) was purified from the 0.4 M NaCl fraction.
Cell Adhesion Assay-The cell adhesion assay was performed as described previously (11) with some modifications. Briefly, 96-well microtiter plates (Corning Costar, Acton, MA) were coated with test samples in PBS at 4°C overnight and then blocked with 1.2% (w/v) bovine serum albumin in PBS at 37°C for 1.5 h. Cells were trypsinized and recovered in 10% FCS-containing medium at 37°C for 30 min. The cells were then washed three times with and suspended in serum-free DMEM/F-12. The cell suspension (3 ϫ 10 4 HaCaT cells or 4 ϫ 10 4 BRL cells per 0.1 ml) was inoculated into each well and incubated at 37°C for 1 h. Adherent cells were fixed with 2.5% glutaraldehyde, stained with 0.5% (w/v) crystal violet in 20% (v/v) methanol for 10 min, and washed with tap water. The dye was extracted with 0.1 M citrate in 50% methanol (v/v) for 30 min and measured for the absorbance at 590 nm using a microplate reader.
Cell-scattering Assay-Assay of cell scattering activity was performed using BRL cells as reported before (33). For analysis of soluble proteins, BRL cells (7000 cells) were seeded per well of 24-well culture plates (Sumibe Medical, Tokyo, Japan) containing 0.5 ml/well of DMEM/F-12 plus 1% FCS with a test sample. For analysis of insoluble proteins, test samples were previously coated on 24-well plates as described above. After 2 days in incubation, cell scattering was judged by microscopic observation. For the quantification of cell scattering, total cells and scattered, single cells were counted in three randomly selected microscope fields. At least 500 cells were counted in each field. The degree of cell scattering was expressed by the percentage of single cells in each field.
In Vitro Wound-healing Assay-HaCaT cells in 10% FCS-DMEM medium were densely plated onto a 96-well plate (Sumibe) and allowed to attach to the wells. After 2-or 3-h incubation, each culture formed a monolayer sheet of cells. The confluent cell monolayer was scratched with a plastic pipette tip to create a cell-free zone in each well. The wells were washed with serum-free DMEM extensively to remove cellular debris. The wells were then incubated with test samples in serum-free DMEM at 37°C for 30 min to establish a fresh substrate surface on the cell-free zones, followed by washing with serum-free DMEM. The Ha-CaT cells were allowed to migrate from the cut edge of the scratch in fresh serum-free DMEM. Photographs were taken at 0 and 13 h.
Assay of Neurite Outgrowth-Each well of 24-well plastic plates was coated with test samples at indicated concentrations in 500 l of PBS at 4°C overnight, and then washed with PBS. PC12/HS cells were plated at a density of 1 ϫ 10 5 or 40,000 cells per well containing 1 ml of serum-free RPMI1640 medium supplemented with 100 ng/ml NGF (Mouse 2.5S NGF, Wako) and incubated at 37°C for 6 h or 24 h. The degree of neurite outgrowth was expressed as the total length of extended neurites per total cells in each field.

Expression of LN5 with Unprocessed Laminin ␣3
Chain in HEK293 Cells-The 190-kDa, precursor ␣3 chain of LN5 is almost completely cleaved to the 160-kDa mature form immediately after secretion in many LN5-producing cells. To prepare the LN5 with the 190-kDa ␣3 chain, we expressed a processingresistant, mutated ␣3 chain in ␤3/␥2-HEK cells, which had previously been transfected with the cDNAs of human laminin ␤3 and ␥2 chains (28). The cleavage of the ␣3 chain by an endogenous proteinase occurs at the Gln 1337 -Asp 1338 bond in the spacer region between the LG3 and LG4 domains, releasing the LG4 -5 domain (Fig. 1A) (23). The Gln-Asp sequence is commonly found at the corresponding site of LN5 in human, mouse, and rat. To prevent the proteolytic cleavage, we converted these two amino acid residues to alanine by the site-directed mutagenesis of the ␣3 chain cDNA (Fig. 1B). The mutated ␣3 cDNA (␣3/AA) was transfected into ␤3/␥2-HEK cells. An HEK transfectant clone, which expressed all the three chains at high levels as the LN5 heterotrimer, named LN5AA-HEK, was selected.
Processing of ␣3 Chain in Two HEK Transfectants-The recombinant LN5 produced by LN5AA-HEK cells was compared with that produced by LN5-HEK cells, which had been transfected with the cDNAs of the wild-type ␣3, ␤3, and ␥2 chains (28). When analyzed by immunoblotting, the conditioned medium of LN5-HEK cells contained the 160-kDa, processed ␣3 chain with a trace of 140-kDa ␣3 chain but not the 190-kDa form at all (Fig. 2A). The 140-kDa ␣3 chain is produced by a proteolytic cleavage within the NH 2 -terminal, domain IIIa of the native chain (3). Unexpectedly, the conditioned medium of LN5AA-HEK cells contained the 160-kDa ␣3 chain as a major band with a smaller amount of the 190-kDa ␣3 chain. This showed that the amino acid substitution could not sufficiently prevent the proteolytic processing.
We also analyzed LN5 deposited on ECM by the two HEK transfectants. The ECM of LN5-HEK cells, like their conditioned medium, did not contain the 190-kDa ␣3 chain (Fig. 2B). In contrast, the ECM of LN5AA-HEK cells contained the 190-kDa, precursor ␣3 chain. The relative amount of the 190-kDa ␣3 chain was higher than that of the 160-kDa form. These results indicated that LN5AA-HEK cells expressed the precursor LN5 with the 190-kDa ␣3 chain, which was efficiently deposited on the ECM.
Recent studies have suggested that metalloproteinases of the BMP-1/mTLD family, which specifically recognize Asp at the P1Ј site, are responsible for the processing of the laminin ␣3 and ␥2 chains (18). To determine the cleavage site in the mutated ␣3 chain ␣3/AA, we purified the LG4 -5 fragments from the serum-free conditioned media of LN5AA-HEK and LN5-HEK cells, and analyzed their NH 2 -terminal amino acid sequences. The yield of the purified LG4 -5 from LN5AA-HEK cells was about one-third that of LN5-HEK cells:1.5 g/liter for LN5AA-HEK versus 4.5 g/liter for LN5-HEK. The NH 2 -terminal sequence of the LG4 -5 derived from LN5-HEK cells started at residue 1338 (Asp), in agreement with our previous result (23) (Table I). Unexpectedly, the NH 2 -terminal sequence of the mutated LG4 -5 derived from LN5AA-HEK also started at residue 1338 (Ala) with the same following sequence as that of the wild-type LG4 -5. We could not identify any additional sequence. This indicated that the mutated ␣3 chain was cleaved at the same position (between residues 1337 and 1338) as the wild-type ␣3 chain, despite the substitution of the cleavage site from Gln-Asp to Ala-Ala.
Cellular morphology was also compared between LN5-HEK and LN5AA-HEK cells. Both cell lines exhibited flattened morphology with prominent lamellipodia as compared with parent HEK cells or ␤3␥2-HEK cells (data not shown). There was no apparent difference in their cell growth rates in serum-containing medium (data not shown).

Deposition of Processed and Unprocessed LN5 Forms onto ECM-The
LG4 -5 fragment of the ␣3 chain is known to have strong affinity to heparin (23). To test whether or not the heparin affinity of the LG4 -5 domain contributes to the efficient deposition of the precursor LN5 on the ECM, LN5AA-HEK cells were incubated in the presence or absence of various concentrations of heparin. The analysis of the resultant conditioned medium and ECM by immunoblotting showed that the amount of the 190-kDa ␣3 chain decreased in the ECM but increased in the conditioned medium as the heparin concentration was increased (Fig. 3). This suggested that the ability of LG4 -5 to bind heparan sulfates contributed to the efficient deposition of the precursor LN5 on the matrix. The deposition of the 160-kDa ␣3 chain was also inhibited by heparin, suggesting that other heparin-binding sites also contribute to the deposition of the processed LN5.
Purification of LN5 Containing Unprocessed or Processed ␣3 Chain-To show the difference of the biological activity between the LN5 forms with the unprocessed and processed ␣3 chains, we purified both LN5 forms from the serum-free conditioned medium of LN5AA-HEK cells. First, the conditioned medium of LN5AA-HEK cells was subjected to molecular-sieve chromatography on a Sepharose 4B column. The fractions rich in the 190-kDa ␣3 chain and those rich in the 160-kDa ␣3 chain were separately pooled and applied to a heparin-Sepharose column. The 190-kDa, unprocessed ␣3 chain was eluted at 0.6 -0.75 M NaCl, whereas the 160-kDa ␣3 chain was completely eluted below 0.5 M NaCl (data not shown). Finally, the LN5 containing the 190-kDa ␣3 chain and one containing the 160-kDa ␣3 chain were purified by affinity column chromatography with the anti-laminin-␣3 monoclonal antibody.
When analyzed by SDS-PAGE under reducing conditions (Fig. 4A) and immunoblotting with the antibodies to the three subunits (Fig. 4B), the LN5 eluted at 0.6 -0.75 M NaCl was resolved into three major bands of the 190-kDa ␣3 chain, the 135-kDa ␤3 chain and the 105-kDa ␥2 chain. In addition, this unprocessed LN5, designated as pre-LN5, contained a trace of the 160-kDa ␣3 chain. On the other hand, the LN5 eluted below 0.4 M NaCl, designated as mat-LN5, was resolved into three major bands of the 160-kDa ␣3 chain, the 135-kDa ␤3 chain, and the 105-kDa ␥2 chain.
Cell Attachment and Spreading Activities of pre-LN5 and mat-LN5-The LN5 with the unprocessed ␣3 chain (pre-LN5) and the mature one with the processed ␣3 chain (mat-LN5) were compared for their cell attachment activities using the following three cell lines: the rat liver-derived epithelial cell line BRL, the human immortalized keratinocyte line HaCaT, and the human fibrosarcoma cell line HT1080.
When BRL cells were seeded onto plastic plates pre-coated with various concentrations of LN5, they attached to the mat-LN5 substrate at slightly lower concentrations than the pre-LN5 substrate. The effective concentration for the half-maximal cell attachment (ED 50 ) was 0.07 g/ml for pre-LN5 and 0.045 g/ml for mat-LN5 (Fig. 5A, upper panel). When HaCaT cells were seeded on the plates, they attached to the mat-LN5 substrate much more efficiently than the pre-LN5 substrate: the ED 50 of pre-LN5 (0.19 g/ml) was about 5-times higher than that of mat-LN5 (0.04 g/ml) (Fig. 5A, lower panel). HT1080 cells showed essentially the same dose-response to both LN5 forms as HaCaT cells (data not shown). We verified  . 3. Effect of heparin on deposition of LN5 onto ECM in LN5AA-HEK cell culture. LN5AA-HEK cells were incubated in serum-containing medium for 24 h and then in serum-free medium with or without the indicated concentrations of heparin (ϩHep) for 2 days. The resultant extracellular matrix (ECM) and conditioned medium (CM) were analyzed by immunoblotting with the anti-␣3-chain antibody. Arrowheads, unprocessed (190) or processed (160) forms of the ␣3 chain.
that there was no significant difference in the coating efficiency between the two LN5 forms.
Next, we examined morphological differences on the plates pre-coated with either form of LN5 at the minimal concentration required for the maximal cell attachment, i.e. 0.25 g/ml mat-LN5 and 0.50 g/ml pre-LN5. The most significant difference was obtained with HaCaT cells (Fig. 5B). HaCaT cells spread well on the mat-LN5 substrate, whereas they spread poorly on the pre-LN5 substrate. Similar but less prominent differences were obtained with HT1080 cells and BRL cells (data not shown). All these results showed that the mature LN5 (mat-LN5) had significantly higher cell adhesion activity than the unprocessed LN5 (pre-LN5).
Cell-scattering and Cell Migration Activities-LN5 supports not only cell adhesion but also cell migration and scattering. The cell-scattering activity of LN5 is thought to reflect its cell migration activity. Both soluble and insoluble (or coated) forms of LN5 are able to stimulate cell migration and cell scattering (33). First, the cell-scattering activity toward BRL cells was assayed for pre-LN5 and mat-LN5. For the assay, each LN5 was directly added into the culture medium of BRL cells in the presence of 1% FCS and incubated for 2 days. As shown in Fig. 6A, both LN5 forms at 40 ng/ml showed the typical cell scattering, but at 10 ng/ml the scattering was less evident in pre-LN5 than mat-LN5. In the experiment with varied concentrations of LN5, the ED 50 for cell scattering was determined to be 14 ng/ml for pre-LN5 and 3.5 ng/ml for mat-LN5 (Fig. 6B). Thus, mat-LN5 showed about 4-times higher cell-scattering activity than pre-LN5. The differential cell-scattering activity was reproduced when the cell scattering was assayed with the plates pre-coated with various concentrations of the LN5 forms (Fig. 6C).
The cell migration activity of the two LN5 forms was assayed by an in vitro wound healing assay with HaCaT keratinocytes. The "wounded" zone introduced by scratching the monolayer sheet of HaCaT cells was treated with or without pre-LN5 or mat-LN5 in serum-free medium. The cultures were then incubated to allow the cell migration onto the scraped (or wound) zones. The treatment with either LN5 form caused rapid repair of the scraped area compared with the control (Fig. 7A). When compared between the two LN5 forms, the wound closure was faster on mat-LN5 than pre-LN5. The relative repair rates of control, mat-LN5, and pre-LN5 were determined to be ϳ35: 100:53. These data indicate that the loss of the LG4 -5 domain from the precursor ␣3 chain increases the cell migration activity of LN5.
Neurite Outgrowth Activities of pre-LN5 and mat-LN5-LN5 is know to promote neurite outgrowth (34 -36). Therefore, we also assessed the neurite outgrowth activity of the two LN5 forms toward PC12/HS cells. pre-LN5, mat-LN5, and type-I collagen were individually coated on plastic plates, and then PC12/HS cells were incubated on the substrate-coated plates in serum-free culture medium supplemented with NGF for 24 h. As shown in Fig. 8A, both pre-LN5 and mat-LN5 promoted the neurite outgrowth of PC12/HS cells much more efficiently than type-I collagen. When compared between the two LN5 forms, the neurite outgrowth was more prominent on the mat-LN5 substrate than on the pre-LN5 substrate. Quantitative experiments showed that the ED 50 of neurite outgrowth was 0.02 g/ml for mat-LN5 and 0.1 g/ml for pre-LN5 (Fig. 8B).
We also examined the neurite outgrowth activity of the LG4 -5 fragment (␣3LG4 -5), which was released from the 190-kDa ␣3 chain of the precursor LN5 by an endogenous proteinase (Fig. 9). When LG4 -5 was coated alone on a culture plate, it scarcely induced the neurite outgrowth of PC12/HS cells. However, when LG4 -5 was co-coated with a low concentration (0.01 g/ml) of mat-LN5 on a culture plate, it significantly promoted the neurite outgrowth as compared with the plates coated with either LG4 -5 or mat-LN5 alone. These data suggest that the LG4 -5 domain is able to enhance the neurite outgrowth activity of mat-LN5 when it is released from the precursor ␣3 chain. FIG. 7. In vitro wound healing assay of mat-LN5 and pre-LN5 using HaCaT keratinocytes. Monolayer cultures of HaCaT cells on a 96-well plate were scratched with a yellow tip to introduce a wounded zone, followed by washing with PBS. The denuded areas were coated with serum-free medium (None), 1.0 g/ml mat-LN5, or 1.0 g/ml pre-LN5. HaCaT cells were allowed to migrate from the cutting edge of the scratch in fresh serum-free medium. Phase-contrast photomicrographs were taken at 0 and 13 h.

DISCUSSION
To clarify the biological significance of the ␣3 chain cleavage in human LN5, we tried to prepare the precursor LN5 with a non-cleavable, 190-kDa ␣3 chain (pre-LN5). This attempt was partially successful, because the point mutation of the cleavage sequence Gln-Asp to Ala-Ala in the ␣3 chain did not completely block the proteolytic cleavage. It has been reported that the BMP-1/mTLD families are most likely responsible for the processing of the laminin ␣3 and ␥2 chains (18,19). Interestingly, the analysis of the released LG4 -5 fragment indicated that the mutated ␣3 chain was cleaved at the same position as the wild-type ␣3 chain. Because these proteinases specifically recognize Asp at the P1Ј site, other proteinases may be responsible for the cleavage of the Ala-Ala sequence. Serine proteinases such as plasmin (24) and thrombin (37) have been reported to cleave the ␣3 chain between LG3 and LG4 domains. Based on the cleavage bond, these serine proteinases seem to not be candidates for the processing enzyme of the LN5 ␣3 chain at least in the culture systems examined so far (23).
LN5-HEK cells, which had been introduced with the wildtype cDNAs of the ␣3, ␤3, and ␥2 chains, did not show the precursor LN5 with the 190-kDa ␣3 chain in either the conditioned medium or ECM fraction. Only the processed LN5 with the 160-kDa ␣3 chain was detected in both the conditioned medium and ECM. However, LN5AA-HEK cells, which expressed the mutated ␣3 chain, produced the LN5 with the 190-kDa ␣3 chain (pre-LN5) in the ECM at a much higher level than in the conditioned medium. This clearly indicated that pre-LN5 is deposited on the ECM much more efficiently than the LN5 with the 160-kDa ␣3 chain (mat-LN5). Very recently, Sigle et al. (37) reported that the LG4 -5 domain plays an important role in the deposition or assembly of the precursor LN5 into the ECM of keratinocytes. They introduced cDNA either for a non-cleavable ␣3 chain lacking the spacer sequence between LG3 and LG4 or for a pre-cleaved ␣3 chain lacking LG4 -5 into mouse keratinocytes deficient in the endogenous laminin ␣3 chain. The unprocessed LN5 was deposited on the matrix more efficiently than the pre-cleaved LN5. They have also suggested that exogenous precursor LN5, but not the processed LN5, is assembled into the matrix through interaction with an unidentified cell surface receptor (37). Another group, using integrin ␣3␤1-deficient keratinocytes, has reported that this integrin plays an important role in the organized incorporation of LN5 into the matrix (38). The LG4 and LG5 domains of the ␣3 chain contain heparan sulfate-binding sites (10), and a recombinant LG4 domain interacts with syndecans-1, -2, and -4 (27,39). Furthermore, it has been reported that the precursor LN5 interacts with syndecan-1 through its LG4 -5 domain (25). In the present study, the deposition of pre-LN5 on the matrix was effectively blocked by exogenous heparin in a dosedependent manner. These results suggest that pre-LN5 may be assembled into the matrix through the interaction between its LG4 -5 domain and heparan sulfate proteoglycans on the cell surface and/or ECM. Because the matrix deposition of the processed LN5 without LG4 -5 was also inhibited by heparin, heparin-binding sites other than the LG4 -5 domain may also contribute to the deposition of LN5 on the ECM. In this regard, it should be noted that the short arm of laminin ␥2 chain plays a role in the deposition of LN5 (40).
We expressed and purified the two forms of recombinant LN5 with the unprocessed ␣3 chain (pre-LN5) and with the processed ␣3 chain (mat-LN5) and compared their biological activities. When they were coated on plastic plates or directly added into culture medium, pre-LN5 supported cellular attachment, spreading, scattering, and migration less efficiently than mat-LN5. LN5 is also known to stimulate the neurite out-growth of primary neural cells (34) as well as some neural cell lines through integrin ␣3␤1 (35,36). The expression of LN5 in the floor plate of the developing neural tube (41) suggests its potential role in neural development. In the present study, mat-LN5 also promoted neurite outgrowth more efficiently than pre-LN5. All these results suggest that the processed LN5 induces integrin signaling more efficiently than the precursor LN5. It can be concluded that the proteolytic processing of the ␣3 chain between the LG3 and LG4 domains converts LN5 from a less active form to an active form. These results contrast with the finding by Goldfinger et al. (24) that the proteolytic processing of the ␣3 chain suppresses the cell motility activity of LN5 but enhances its activity to support stable adhesion. We recently prepared a series of recombinant LN6 (␣3␤1␥1) forms with a partially deleted LG domain and showed that the deletion of LG4 -5 or LG5 domain of the ␣3 chain activates LN6 with respect to its capacity to promote cell adhesion and migration (27). HaCaT keratinocytes and HT1080 fibrosarcoma cells did not spread on the LN6 with the full LG domain, whereas they efficiently spread on LN6 lacking the LG4 -5 or LG5 domain. It was speculated that the integrin-binding site present in the LG1-3 domain is partially masked by the LG4 -5 domain, especially by LG5, hence the deletion of LG5 or LG4 -5 enhances the integrin binding efficiency of LG1-3. These previous findings about laminin-6 are consistent with the present data. However, the precursor LN5 with LG4 -5 (pre-LN5) appears to be more active than the precursor LN6, because pre-LN5 supported cell spreading as the concentration was increased. Recently, Kü nneken et al. (42) prepared recombinant LN5 fragments comprising a heterotrimeric, COOH-terminal part of the coiled-coil domain and the full or partially deleted G domain and showed that the removal of the LG4 -5 domain increases the binding affinity of the LN5 fragment to integrin ␣3␤1. Their results also support our present findings.
The process of the epidermal wound repair has been extensively studied by many groups. Although the precise mechanism remains unclear, there is a general concept that LN5 plays a major role in the regulation of the adhesion and migration of keratinocytes in the wound repair process (15,43,44). When the skin is injured, the precursor LN5 with the 190-kDa ␣3 chain is actively synthesized by keratinocytes at the wound edge and deposited onto the wound area, resulting in the elevated migration of following keratinocytes (15,37,44). After the cell migration onto the wound area, the precursor LN5 with the 190-kDa ␣3 chain is proteolytically converted to the mature LN5 with the 160-kDa ␣3 chain, which supports the stable adhesion rather than the migration of the following keratinocytes (24). These findings give rise to a hypothesis that the change of keratinocyte behavior from the active migration to the stable adhesion is regulated by the proteolytic conversion of the precursor LN5 to the processed one. However, our present study showed that the processed LN5 had higher activity in both adhesion and migration than the precursor LN5.
Based on these results, we propose the following model. The LG4 -5 domain plays an important role in the deposition of LN5 on the matrix, presumably through its interaction with heparan sulfate proteoglycans on the cell surface and ECM. This is also supported by a recent report (37). After secretion and deposition, the precursor LN5 is converted to the processed form, releasing the LG4 -5 fragment. The processed LN5 supports active migration and proliferation of keratinocytes at the wound edge. We have recently shown that the processed LN5 in a soluble form is able to bind to integrin ␣3␤1 on the cell surface at the wound edge to stimulate cell migration (33). Therefore, the soluble, processed LN5 is also likely to contribute to the keratinocyte migration. When the wound area is covered with the migrated keratinocytes, the processed LN5, which is assembled into the matrix, nucleates the assembly of hemidesmosomes to produce the stable cell adhesion via integrin ␣6␤4. It has been suggested that the organizational state of LN5 has an influence on its function (38). Therefore, we speculate that the switch from the active migration on the LN5 substrate via integrin ␣3␤1 to the stable adhesion via integrin ␣6␤4 may be achieved by the receptor-mediated, organized assembly of the mature LN5 into the matrix, rather than the proteolytic processing. In addition, cell-cell interaction through cadherin and gap junction may also regulate the LN5-integrin interaction (45).
Our present and previous studies imply that the LG4 -5 domain in the ␣3 chain has a suppressive role in the interaction of LN5 with integrins. However, when released from LN5 by proteolysis, LG4 -5 supports weak cell adhesion through heparan sulfate proteoglycans on the cell surface (25,27). LG4 -5 also cooperates with the processed LN5 to promote cell migration (23) and cell adhesion (27). In the present study, LG4 -5 also stimulated neurite outgrowth in the presence of the processed LN5. The recombinant LG4 of the ␣3 chain has been reported to stimulate neurite outgrowth (46). These results suggest that the LG4 -5 fragment enhances the LN5/integrin signaling, presumably through its interaction with syndecans. Many studies have shown that syndecans modulate integrin functions (47). It seems likely that the LG4 -5 fragment cooperates with the processed LN5 and other integrin ligands to regulate cellular functions in some physiological conditions.