Structural Requirement of Carboxyl-terminal Globular Domains of Laminin α3 Chain for Promotion of Rapid Cell Adhesion and Migration by Laminin-5*

The basement membrane protein laminin-5, a heterotrimer of laminin α3, β3, and γ2 chains, potently promotes cellular adhesion and motility. It has been supposed that the carboxyl-terminal globular region of the α3 chain consisting of five distinct domains (G1 to G5) is important for its interaction with integrins. To clarify the function of each G domain, we transfected cDNAs for the full-length (wild type (WT)) and five deletion derivatives (ΔGs) of the α3 chain into human fibrosarcoma cell line HT1080, which expressed and secreted the laminin β3 and γ2 chains but not the α3 chain. The transfectants with the α3 chain cDNAs lacking G5 (ΔG5), G4–5 (ΔG4–5), G3–5 (ΔG3–5), and G2–5 (ΔG2–5) secreted laminin-5 variants at levels comparable to that with WT cDNA. However, the transfectant with the cDNA without any G domains (ΔG1–5) secreted little laminin-5, suggesting that the G domains are essential for the efficient assembly and secretion of the heterotrimer α3β3γ2. The transfectants with WT, ΔG5, and ΔG4–5 cDNAs survived in serum-free medium longer than those with ΔG3–5, ΔG2–5, and ΔG1–5 cDNAs. The transfectants with WT, ΔG5, and ΔG4–5cDNAs secreted apparently the same size of laminin-5, which lacked G4 and G5 due to proteolytic cleavage between G3 and G4, and these laminin-5 forms potently promoted integrin α3β1-dependent cell adhesion and migration. However, the laminin-5 forms of ΔG3–5 and ΔG2–5 hardly promoted the cell adhesion and motility. These findings demonstrate that the G3 domain, but not the G4 and G5 domains, of the α3 chain is essential for the potent promotion of cell adhesion and motility by laminin-5.

Laminins are a family of extracellular matrix proteins that are localized mainly in basement membranes and regulate various cellular functions such as adhesion, motility, growth, differentiation, and apoptosis through interaction with specific integrins on the cell surface (1,2). The three subunits of laminins, designated ␣, ␤, and ␥ chains, form the well known cross-shaped structure linked together by disulfide bonds. Five ␣, three ␤, and three ␥ chains and at least 12 structural iso-forms of laminin (laminin-1 to -12) having distinct chain combinations have been identified in human thus far (3,4). Among these laminin isoforms, laminin-5, which consists of the laminin ␣3, ␤3, and ␥2 chains, is unique in the structure and biological activity. Laminin-5 was originally found as a keratinocyte-derived matrix protein named epiligrin, kalinin, or nicein (5-7) and a laminin-like cell scattering factor, ladsin, secreted by human gastric carcinoma cells (8). Laminin-5 has a feature lacking some domains found in the amino-terminal regions (or the short arms) of the three subunits of other laminin isoforms (2). The laminin ␣3 chain is found in laminin-6 (␣3␤1␥1) and laminin-7 (␣3␤2␥1) besides laminin-5, but the laminin ␤3 and ␥2 chains are found only in laminin-5. More interestingly, laminin-5 has much higher activity to promote adhesion, migration, and scattering of various types of cells than laminin-1, fibronectin, and vitronectin (8 -10).
Most cultured cell lines utilize integrin ␣ 3 ␤ 1 as a major receptor to adhere and migrate on the laminin-5 substrate, but integrins ␣ 6 ␤ 4 and ␣ 6 ␤ 1 also act as the additional receptors of laminin-5 depending on cell types (5,9,10). Laminin-5 is an important component of basement membranes of the skin and many other epithelial tissues (5,6,11). The interaction of laminin-5 with integrin ␣ 6 ␤ 4 in the hemidesmosome structures is essential for the stable anchorage of basal epithelial cells to the underlying connective tissues. Defects of laminin-5 genes cause Herlitz-type junctional epidermolysis bullosa (H-JEB), which is characterized by splitting of epidermal/dermal junctions (12,13). Similarly, targeted disruption of the laminin-5 genes or integrin ␣ 6 ␤ 4 genes in mice causes severe junctional blisters and abnormal hemidesmosomes, resulting in neonatal lethality (14 -16). On the other hand, the potent cell motility activity of laminin-5 has been suggested to contribute to wound healing (17) and tumor invasion (18).
For understanding the molecular basis for the unique bifunctional properties of laminin-5, the stable adhesion and motility, it seems important to clarify its structural and functional relationship. All laminin ␣ chains have a large carboxyl-terminal globular domain consisting of a tandem array of five small globular domains (or modules) (G1 to G5) (1,2). These G domains are autonomous folding units (19). They contain binding sites for ␤ 1 integrins (20) and heparin (21), as well as ␣-dystroglycan in some laminin isoforms (22,23). Our previous study with recombinant G domains of the laminin ␣3 chain has shown that the G2 domain contains an integrin ␣ 3 ␤ 1 -binding site, and the G4 and G5 domains weakly interact with heparan sulfate proteoglycans (24). To clarify the functions of the G domains of the laminin ␣3 chain, we have prepared recombinant laminin-5 proteins serially lacking G domains of the ␣3 chain, and we examined their activities to promote cell adhesion and motility.

MATERIALS AND METHODS
Antibodies-Mouse monoclonal antibodies against human laminin ␣3 chain were established in our laboratory with the support of Eiken Chemical (Tokyo, Japan). These antibodies were raised against the glutathione S-transferase fusion protein of the amino-terminal region of human laminin ␣3 chain (amino acid residues 109 -331). One of these antibodies, LS␣3c4, was used for immunoaffinity purification of recombinant laminin-5 proteins. Monoclonal antibody against human laminin ␥2 chain (D4B5) was described previously (11,25). Rabbit polyclonal antibody (␣3G4) was raised against the glutathione S-transferase fusion protein of the G4 domain of human laminin ␣3 chain (amino acid residues 1335-1552), which included the carboxyl-terminal 19 amino acid residues of the G3 domain besides the complete G4 sequence. Monoclonal antibody against human laminin ␤3 chain was purchased from Transduction Laboratories (Lexington, KY). Function-blocking antibodies against integrins used are anti-␣ 2 -integrin antibody P1E6, anti-␣ 3 -integrin antibody P1B5, and anti-␤ 1 -integrin antibody P4C10 from Life Technologies, Inc., (Gaithersburg, MD) anti-␣ 5 -integrin antibody P1D6 from Chemicon (Temecula, CA), and anti-␣ 6 -integrin antibody G0H3 from PharMingen (San Diego, CA).
Cell Culture-Human fibrosarcoma cell line HT1080 and human tongue squamous adenocarcinoma cell line C-4I were obtained from Japanese Cancer Resources Bank. Buffalo rat liver-derived epithelial cell line has been used in previous studies (8,9). These cell lines were cultured in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F12 medium (DME/F12) 1 (Life Technologies, Inc.) supplemented with 10% fetal calf serum (FCS) (JRH Biosciences, Lenexa, KS), penicillin, and streptomycin sulfate.
Electrophoretic Analyses-Northern blotting analysis of the laminin ␣3, ␤3, and ␥2 chains were performed by the method described previously (26). SDS-polyacrylamide gel electrophoresis (PAGE) was performed on 6% gels under nonreducing or reducing conditions. Separated proteins were stained with Coomassie Brilliant Blue R-250. For immunoblotting, proteins separated by SDS-PAGE were transferred onto nitrocellulose membranes. The three subunits of laminin-5 were detected by the alkaline phosphatase method with chain-specific antibodies.
Assay of Cell Adhesion-Adhesion of BRL cells or C-4I cells to purified laminin-5 proteins and extracellular matrices deposited by various HT1080 transfectants was assayed as described previously (24). To prepare the matrices, each transfectant clone of HT1080 cells (5 ϫ 10 3 cells) was inoculated and incubated on 96-well plastic plates (Sumibe Medical, Tokyo, Japan) in DME/F12 medium containing 10% FCS at 37°C for 4 days. Then the transfectant cells were detached from the plates by incubating with 100 l of 10 mM EDTA in Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline. Complete detachment of cells was confirmed under a microscopy. The plates were washed twice with phosphatebuffered saline and blocked with 200 l of 1.2% (w/v) bovine serum albumin in phosphate-buffered saline at 37°C for 1.5 h. These plates were used as the matrix-coated plates for the assay of cell adhesion. Cells adhered to the laminin-5 proteins or the HT1080 matrices were stained with Hoechst 33342 for 1.5 h, and the fluorescent intensity of each well was measured using a CytoFluor 2350 fluorometer (Millipore, Bedford, MA).
Assay of Cell Migration and Cell Scattering-Migration and scattering of BRL cells on purified laminin-5 proteins and extracellular matrices deposited by various HT1080 transfectants were assayed as reported previously (9). For the cell migration assay, the matrices of HT1080 transfectants were prepared in 25-cm 2 tissue culture flasks (Beckton Dickinson, Franklin Lakes, NJ) as described above. Cell migration on these substrates was monitored at 37°C with time-lapse video equipment in which a video camera (MKC-385; Olympus, Tokyo) was mounted on an inverted microscope (IX-70; Olympus) and connected to a time-lapse control unit (LVR-3000AN; Sony, Tokyo, Japan). The length of cell migration was measured with a video micrometer (VM-30; Olympus).
Purification of Recombinant Laminin-5-Four transfectants of HT1080 cells were grown to confluence in roller bottles (Beckton Dickinson) with DME/F12 medium containing 10% FCS (200 ml/bottle). The confluent cultures were washed twice with and incubated in serum-free DME/F12 medium. The serum-free conditioned medium was harvested every 2 days, clarified by centrifugation, and subjected to protein precipitation with 80%-saturated ammonium sulfate. The precipitated proteins were dissolved in and dialyzed against 20 mM Tris-HCl buffer containing 0.5 M NaCl, 0.01% (w/v) Brij35, and 0.1% (w/v) CHAPS and then applied to molecular sieve chromatography on a Sepharose 4B (Amersham Pharmacia Biotech) (8). Fractions containing laminin-5 were pooled and subjected to immunoaffinity chromatography with the anti-laminin-␣3 monoclonal antibody (LS␣3c4). Bound proteins were eluted from the affinity column with 0.05% (v/v) trifluoroacetic acid and immediately neutralized with 1/3 volume of 1 M Tris-HCl (pH 7.5). The eluted laminin-5 fractions were further purified by immunoaffinity chromatography with the anti-laminin-␥2 monoclonal antibody (D4B5). The recombinant laminin-5 proteins thus purified were stored at 4°C in the presence of 0.005% Brij35 and 0.1% CHAPS.

Expression of Full-length and Carboxyl-terminally Deleted
Derivatives of Laminin ␣3 Chain in HT1080 Cells-To assess the contribution of the five G domains of the laminin ␣3 chain to the cell adhesion and cell motility activities of laminin-5, we tried to prepare recombinant laminin-5 variants serially lacking G domains. Human fibrosarcoma cell line HT1080, which expresses the laminin ␤3 and ␥2 chains but not the ␣3 chain (26), was chosen as the recipient cells to transfect the ␣3 chain cDNAs.
When cell morphology and growth rate were compared among the HT1080 transfectants, the difference is not evident between the control and cDNA transfectants under usual serum-containing culture conditions. However, when incubated in serum-free culture medium, HT1080 transfectants of WT, ⌬G 5 , and ⌬G 4 -5 survived much longer than the other cDNA transfectants and the control transfectant (data not shown). This suggested that the laminin-5 forms with G1-3 might prevent HT1080 cells from apoptosis in serum-free medium.
Next, we examined gene expression of the three laminin-5 subunits in the control and cDNA transfectants of HT1080 cells by Northern blotting. As expected, all of the five transfectants of the ␣3 chain cDNAs expressed the ␣3 chain mRNA with different sizes at comparable levels, whereas the control transfectant expressed no positive signal ( Fig. 2A). In contrast, the laminin ␤3 and ␥2 chain mRNAs were expressed in the control transfectant as well as the cDNA transfectants (Fig. 2, B and C). The transfection of the laminin ␣3 chain cDNAs did not affect the transcriptional levels of the ␤3 and ␥2 chain genes in any clones.
As described above, the HT1080 transfectants of WT, ⌬G 5 , and ⌬G 4 -5 secreted apparently the same size (160 kDa) of the ␣3 chain. It is known that the mature ␣3 chain of 190 kDa is proteolytically processed to the 160-kDa form (28). When the conditioned media of the control HT1080, HT1080/WT, and HT1080/⌬G 5 were analyzed by Western blotting with a rabbit polyclonal antibody against a recombinant G4 protein, the 190- kDa band in HT1080/WT and the 175-kDa band in HT1080/ ⌬G 5 were clearly detected, whereas the 160-kDa band in both transfectants was scarcely detected by the antibody (Fig. 4). Therefore, we judged the 160-kDa band as the ␣3 chain that had been proteolytically cleaved between the G3 and G4 domains.
Western blotting analysis with the anti-␤3 chain antibody detected almost a single band of the 135-kDa mature ␤3 chain in the conditioned media of all transfectants (Fig. 3B). On the other hand, analysis with the anti-␥2 chain antibody detected the mature ␥2 chain of 150 kDa and its proteolytically processed form of 105 kDa at relatively irregular intensity (Fig.  3C). It is noteworthy that the control transfectant secreted significant amounts of the ␤3 and ␥2 chains. This indicates that the secretion of the laminin ␤3 and ␥2 chains does not depend on the co-expression of the ␣3 chain.
To confirm the secretion of laminin-5, immunoprecipitation with the anti-laminin ␥2 chain monoclonal antibody was carried out with the conditioned media of some HT1080 transfectants. The anti-␥2 chain antibody precipitated both the laminin ␣3 and ␤3 chains, as well as the laminin ␥2 chain, from the conditioned media of transfectants of the laminin ␣3 cDNAs, indicating the formation of the heterotrimer of the ␣3, ␤3, and ␥2 chains (data not shown). Interestingly, the non-processed forms of the laminin ␣3 chains were hardly precipitated from the conditioned media of HT1080/WT and HT1080/⌬G 5 , as compared with their processed forms (data not shown).
Cell Adhesion Activity of Matrices Deposited by HT1080 Transfectants-To examine the cell adhesion activity of the laminin-5 deletion forms, the matrices deposited by HT1080 transfectants were tested with Buffalo rat liver cell line BRL, which has been used for the assays of cell scattering activity and cell adhesion activity of laminin-5 (8,9). This cell line adheres not only to laminin-5 but also to fibronectin. When the relative amounts of the laminin ␣3 chain proteins deposited by HT1080 transfectants were determined using the anti-laminin-␣3 antibody, all of the matrices except for the control and ⌬G 1-5 contained the laminin ␣3 chains at similar levels (data not shown). When the cell adhesion activity was assayed at 20 min after seeding, BRL cells attached and spread on the matrices of HT1080/WT, HT1080/⌬G 5 , and HT1080/⌬G 4 -5 but not on the matrices of the other transfectants (Fig. 5, closed columns). When assayed at 1 h after seeding, all of the matrices supported the adhesion of BRL cells at almost the same activity (Fig. 5, open columns). However, there was a morphological difference of BRL cells between the matrices from the transfectants of laminin ␣3 cDNAs with and without the G3 domain. BRL cells spread more on the matrices of WT, ⌬G 5 , and ⌬G 4 -5 than on those of the others (data not shown). It has previously been reported that the RGD peptide inhibits adhesion of BRL cells to fibronectin by about half but does not inhibit that to laminin-5 at all (9). When BRL cells were treated with a RGDcontaining peptide (GRGDSP) or a control peptide (GRGESP), the cell adhesion to the ⌬G 3-5 matrix, but not that to the ⌬G 4 -5 matrix, was effectively inhibited by the GRGDSP peptide (Fig.  6). This suggests that BRL cells preferentially attach to and spread on laminin-5 in the matrices of ⌬G 4 -5 , ⌬G 5 , and WT, whereas they slowly attach to fibronectin or other cell adhesion proteins in the matrices of the other transfectants including ⌬G 3-5 . This possibility was confirmed using the adhesion assay with human cervix epidermoid carcinoma cell line C-4I, which is able to adhere to laminin-5 but not fibronectin, vitronectin, or laminin-1. 2 When C-4I cells were seeded on the matrices of HT1080 transfectants, they effectively attached to and spread on only the matrices of WT, ⌬G 5 , and ⌬G 4 -5 even at 1.5 h after seeding (Fig. 7). When effect of function-blocking anti-integrin antibodies was examined, antibodies to integrins ␣ 3 and ␤ 1 strongly inhibited the cell adhesion to the ⌬G 4 -5 matrix and to purified laminin-5, indicating that C-4I cells had adhered to the laminin-5 deposited on the matrix through integrin ␣ 3 ␤ 1 ( Table I). The anti-␣ 6 -integrin antibody weakly inhibited both cell adhesion to the ⌬G 4 -5 matrix and to the purified laminin-5, but antibodies to integrins ␣ 2 and ␣ 5 rather stimulated the cell adhesion. All these results strongly suggested that the G3 domain of the laminin ␣3 chain is essential for the potent cell adhesion activity of laminin-5.
Cell Migration Activity of Matrices Deposited by HT1080 Transfectants-Laminin-5 has potent cell scattering-and cell migration-stimulating activities toward BRL cells (8,9). The cell scattering activity of the matrices deposited by each HT1080 transfectant was analyzed using BRL cells. BRL cells showed marked cell scattering on the matrices of WT, ⌬G 5 , and ⌬G 4 -5 but not on the others (data not shown).
To compare the cell migration activity of laminin-5 deletion forms, BRL cells were incubated on the matrix deposited by each HT1080 transfectant. The cell migration on the matrix was monitored for 12 h using a time-lapse video recorder (Fig.  8). The migration speed was about 5-10 times higher on the matrix of WT than on the control matrix, suggesting that laminin-5 was responsible for the high motility of BRL cells. Similar elevated cell motility was observed on the matrices of ⌬G 5 and ⌬G 4 -5 , whereas this activity was remarkably decreased by losing the G3 domain. The matrix of ⌬G 3-5 slightly stimulated the migration of BRL cells as compared with those of the control transfectants, ⌬G 2-5 and ⌬G 1-5 . These results indicate that the G3 domain of the laminin ␣3 chain is indispensable for the strong cell motility activity of laminin-5 and that the G2 domain also has a low cell motility activity.
Purification of Recombinant Laminin-5 Deletion Forms and Their Subunit Composition-HT1080 cells are expected to secrete some cell adhesion proteins such as fibronectin and collagens. To rule out the effects of these intrinsic matrix proteins, we purified recombinant laminin-5 variants and investigated their biological activities. Four types of laminin-5, WT, ⌬G 5 , ⌬G 3-5 , and ⌬G 2-5 , were prepared from the conditioned media of 2 Y. Kikkawa and K. Miyazaki, unpublished data.

TABLE I Effect of function-blocking antibodies specific to various integrin subunits on adhesion of C-4I cells to matrix of HT1080/
⌬G 4 -5 and to laminin-5 C-4I cells were incubated with the indicated anti-integrin antibodies (1:100 dilution), mouse IgG (15 g/ml), or PBS at 37°C for 30 min, and then they were seeded on plates precoated with the ⌬G 4 -5 matrix or with 0.3 g/ml of laminin-5, which had been purified from STKM-1 cells (26). The numbers of adherent cells were measured after a 1.5-h incubation. The averaged value of the PBS controls was taken as 100%. Each value represents the mean Ϯ S.D. for triplicate assays. the respective cDNA transfectants. Each recombinant laminin-5 was separated by molecular sieve chromatography followed by immunoaffinity chromatography using the anti-laminin-␣3 monoclonal antibody LS␣3c4. The laminin-5 preparations slightly contained the laminin ␤1 chain (220 kDa) and the laminin ␥1 chain (210 kDa), besides the laminin-5 subunits, suggesting that they contained laminin-6. Therefore, these preparations were finally applied to an anti-laminin-␥2 antibody (D4B5) column to remove laminin-6. The total amount of laminin-5 and laminin-6 in conditioned medium was estimated to be about 100 g/liter in WT, 60 -80 g/liter in ⌬G 5 and ⌬G 3-5 , and less than 50 g/liter in ⌬G 2-5 . The recovery of laminin-5 ranged between 25 and 50% in the final laminin-5 preparations. The purified materials contained the laminin ␣3 chains (160 -120 kDa), the ␤3 chain (135 kDa), the ␥2 chain (150 kDa), and the proteolytically processed laminin ␥2 chain (105 kDa) (Fig. 9, left). The approximate size of the laminin ␣3 chain was 160 kDa in WT and ⌬G 5 , 140 kDa in ⌬G 3-5 , and 120 kDa in ⌬G 2-5 ( Fig. 9, right). The laminin-5 forms of WT and ⌬G 5 were considered to be essentially identical to the laminin-5 form of ⌬G 4 -5 (see Figs. 3A and 4). When the laminin-5 of WT was analyzed by Western blotting with the anti-G4 antibody, no immunoreactive band was detected at a low molecular weight region, indicating that the cleaved fragment of G4 -5 was not associated with the purified laminin-5 (data not shown). The fractions eluted from the anti-laminin-␣3 antibody column slightly contained the non-processed ␣3 chains (190 kDa in WT and 175 kDa in ⌬G 5 ) and the laminin ␤1 and ␥1 chains, but these proteins passed through the anti-laminin-␥2 antibody column. This suggested that the non-processed ␣3 chains belonged to laminin-6.
Biological Activity of Purified Recombinant Laminin-5 Deletion Forms-The purified recombinant laminin-5 deletion forms were assayed for cell adhesion activity using BRL cells as the indicator. When each recombinant laminin-5 preparation was precoated on plastic plates at different concentrations, WT and ⌬G 5 promoted the cell adhesion in a concentration-dependent manner, but neither ⌬G 3-5 nor ⌬G 2-5 supported the cell adhesion even at the maximum concentration tested (Fig. 10). This confirmed that the G3 domain is essential for the potent cell adhesion activity of laminin-5. As expected from the above SDS-PAGE analysis, WT and ⌬G 5 showed essentially the same activity.
The purified recombinant laminin-5 forms were also assayed for cell-scattering activity using BRL cells (Fig. 11). When BRL cells were incubated with each recombinant protein in 1% FCS-containing medium for 2 days, typical cell scattering was observed with WT and ⌬G 5 . ⌬G 3-5 induced weak cell scattering of BRL cells as compared with the negative control, but ⌬G 2-5 did not at all. These results were very consistent with the previous experiments with matrices deposited by HT1080 transfectants (Fig. 8). It is clear that the G3 domain of the laminin ␣3 chain is indispensable for the potent cell motility activity of laminin-5, as well as its cell adhesion activity.

DISCUSSION
The present study demonstrated that the laminin-5 lacking both the G4 and G5 domains in the ␣3 chain had potent activity to promote adhesion and migration of BRL cells, but deletion of the G3 domain caused marked decrease of the biological activity of laminin-5. This implies that the G3 domain plays an indispensable role in the expression of biological activity of laminin-5. HT1080 cells expressing laminin-5 forms with the G3 domain were more resistant to apoptosis in serum-free culture medium than those without the G3 domain. This is consistent with the previous report that laminin-5-deficient keratinocytes exhibit reduced survival as compared with normal cells (14).
Laminin-1 has been reported to have several integrin-binding sites in the G domains of the laminin ␣1 chain. Two or more integrin recognition sequences, which are adjacently located in distinct G domains, are likely to cooperate in ligand binding (20,29). However, our previous study with four recombinant G domains of the laminin ␣3 chain showed that only the G2 domain contains an integrin ␣ 3 ␤ 1 binding activity, although the activity of the G3 domain was not examined (24). The recombinant G2 protein and the integrin ␣ 3 ␤ 1 -binding peptide (␣3G2A) have a very low cell adhesion activity compared with intact laminin-5 (24). In addition, the G2 domain and intact laminin-5 lose their cell adhesion activity by heating (8,24). These imply that a specific conformation of the G2 domain produced by the interaction with the G1, G3, and some other parts of the ␣3, ␤3, and ␥2 chains might be important for the high affinity binding of laminin-5 to integrins. Correspondingly, Talts et al. (19,23) have recently shown that recombinant G1-3 protein of laminin ␣2 chain has cell adhesion activity similar to that of native laminin-2/4, although each of G1, G2, and G3 recombinant proteins does not. However, any natural or recombinant laminin-2/4 forms without G4 and G5 domains has not been reported.
Laminin-5 is synthesized initially as a high molecular weight precursor protein that undergoes specific processing to smaller forms after being secreted (28,30,31). The size reduction is a result of proteolytic processing of the ␣3 and ␥2 subunits from 190 to 160 and from 150 to 105 kDa, respectively (8,28,31). In this study, we first demonstrated that the laminin ␣3 chain was proteolytically cleaved between the G3 and G4 domains in HT1080/WT cell line, producing the 160-kDa ␣3 chain. Both the naturally processed laminin-5 with the 160-kDa ␣3 chain and the recombinant laminin-5 lacking G4 and G5 (⌬G 4 -5 ) showed high cell adhesion and cell motility activities. This clearly indicates that the G4 and G5 domains are not essential for the stimulation of cell adhesion and motility. We have recently found that the G4-G5 fragment of laminin ␣3 chain is secreted from some human carcinoma cell lines. 3 The G4-G5 fragment appeared not to be associated with laminin-5 after the proteolytic cleavage. It has been reported that recombinant G4 and G5 domains bind to heparan sulfate proteoglycans as major cell surface receptors, and the G5 recombinant protein has some activity to stimulate cell migration (24). Therefore, it is conceivable that the secreted G4-G5 fragment acts on cells in cooperation with or independently of the laminin-5 with the 160-kDa ␣3 chain.
Goldfinger et al. (32) have compared biological activities of extracellular matrices containing different forms of laminin-5 and found that the laminin-5 with the unprocessed, 190-kDa ␣3 chain has high cell motility activity, whereas one with the processed, 160-kDa ␣3 chain supports stable cell adhesion. However, we have observed the high motility activity of the laminin-5 with the 160-kDa ␣3 chain toward BRL cells and many human carcinoma cell lines (8). It is evident that the laminin-5 with the 160-kDa ␣3 chain exhibits potent cell motility activity toward some cell types. It has also been reported that the processing of the laminin ␥2 chain by the matrix metalloproteinases gelatinase A and MT1-MMP regulate the cell motility activity of laminin-5 (33,34). For clarifying the structure-function relationship of laminin-5, it seems essential to isolate the laminin-5 forms with processed and unprocessed ␣3 and ␥2 chains. In this respect, it should be noted that HT1080/WT cells secreted both the 160-and 190-kDa ␣3 chains, but only the 160-kDa ␣3 chain was purified as a laminin-5 complex. Our recent attempt to isolate the 190-kDa ␣3 chain has shown that this ␣3 chain exists as a laminin-6 but not laminin-5 form. 4 This suggests that the proteolytic processing of the laminin ␣3 chain occurs specifically in laminin-5.
The present study showed that the control HT1080 cells, which do not express the ␣3 chain, secreted the laminin ␤3 and ␥2 chains into culture medium. Various laminin subunits are assembled in the rough endoplasmic reticulum. Several groups have proposed the mechanism for the assembly of laminin subunits. In laminin-1 and laminin-5, a disulfide-linked ␤␥ heterodimer is formed as a presumed intermediate, and ␣ chain is added at a subsequent stage (35)(36)(37). In laminin-1, the ␣1 chain can be secreted as a single subunit, whereas the ␤1 and ␥1 chains cannot (37). When the ␤1 and ␥1 chains are overexpressed separately or together, they remain intracellular as the disulfide-linked dimer of ␤1␥1 or ␤1␤1. We recently found that the laminin ␥2 chain is solely overexpressed at the invasion front of gastric carcinomas, and the ␥2 chain monomer is se- FIG. 11. Cell scattering of BRL cells on recombinant laminin-5 proteins. Five hundred microliters of BRL cell suspension (1.4 ϫ 10 4 cells/ml in DME/F12 medium containing 1% FCS) was inoculated into each well of 24-well plates, and a purified recombinant laminin-5 form of WT, ⌬G 5 , ⌬G 3-5 , or ⌬G 2-5 was added to a final concentration of 0.3 g/ml and incubated at 37°C for 2 days. The cell morphology was examined under a phase-contrast microscope. creted from gastric carcinoma cells in vitro (25). These results clearly indicate that the laminin-5 subunits are secreted differently from the laminin-1 subunits. Furthermore, the present study indicated that HT1080/⌬G [1][2][3][4][5] , which lacked all G domains, secreted only a trace amount of the laminin ␣3 chain into culture medium and contained little ␣3 protein in the cytoplasm despite the high expression of its mRNA. Therefore, the G domains appear to be essential for the formation of stable heterotrimer of laminin-5. The failure of subunit assembly may cause the prompt degradation of the ␣3 chain inside the cells.
In conclusion, we first demonstrated that the G3 domain, but not G4 and G5 domains, of the laminin ␣3 chain is required for the potent activity of laminin-5 to promote cell adhesion and migration. The physiological meaning of the proteolytic cleavage and the biological activity of the G4-G5 fragment are currently under investigation.