Laminin-6 Is Activated by Proteolytic Processing and Regulates Cellular Adhesion and Migration Differently from Laminin-5*

Laminin-6 (LN6) and laminin-5 (LN5), which share the common integrin-binding domain in the laminin α3 chain, are thought to cooperatively regulate cellular functions, but the former has poorly been characterized. Human fibrosarcoma HT1080 cells expressing an exogenous α3 chain were found to secrete LN6 with the full-length α3 chain and a smaller amount of its processed form lacking the carboxyl-terminal G4-5 domain, besides mature LN5 without G4-5 (mat-LN5). We prepared the unprocessed LN6 and mat-LN5, as well as LN6 mutants without G4-5 (LN6ΔG4-5) or G5 (LN6ΔG5). These laminins supported attachment of HT1080 cells and human keratinocytes (HaCaT) through integrins α3β1 and/or α6β1. LN6ΔG4-5, LN6ΔG5, and mat-LN5 promoted rapid cell spreading, whereas LN6 did hardly. A purified G4-5 fragment of the laminin α3 chain supported cell attachment through interaction with heparan sulfate proteoglycans and promoted cell spreading in combination with mat-LN5 or LN6ΔG4-5. These results imply that the G4-5 domain within the LN6 molecule suppresses cell adhesion, while the released G4-5 promotes it. The presence of G5 rather than the heparin-binding domain G4 was responsible for the impaired cell spreading activity of LN6. However, the unprocessed LN6 promoted cell spreading in the presence of mat-LN5. Unlike mat-LN5, both LN6ΔG4-5 and LN6 did weakly or did not stimulate cell motility. These findings demonstrate that LN6 and LN5 have distinct biological activities, but they may cooperatively support cell adhesion. The proteolytic processing of the α3 chain seems to regulate the physiological functions of LN6.

The components of basement membranes not only support tissue architectures but also regulate various cellular functions, such as adhesion, migration, differentiation, growth, and apoptosis. The regulatory functions of basement membranes can be largely attributed to the interaction of laminins with their receptors. Genetically distinct five ␣, three ␤, and three ␥ subunits of laminin form specialized heterotrimers (laminins-1 to -15) that are expressed in a tissue-specific manner during embryonic development as well as in the adult (1). All laminin ␣ subunits share a large globular domain at their carboxylterminal region (G domain), which consists of five homologous globular subdomains (or modules) of about 200 amino acids each (G1-G5 or LG1-LG5). This region contains binding sites for extracellular matrix proteins (e.g. perlecan and fibulin-1), as well as cellular receptors including integrins, syndecans, and ␣-dystroglycan (1).
LN5 is synthesized and secreted in a precursor form containing a 190-kDa ␣3, a 140-kDa ␤3, and a 150-kDa ␥2 chains, but the ␣3 and ␥2 chains undergo proteolytic processing to smaller species of 160 kDa and 105 kDa, respectively (8). Recent studies suggest that post-translational processing of LN5 molecules modulates their function. Proteolytic cleavage of the ␥2 chain by MMP-2 (gelatinase A) or MT1-MMP increases the cell motility activity of LN5 (9, 10) but decreases the cell adhesion activity (11). Similarly, it has been reported that the processing of the 190-kDa ␣3 chain to the 160-kDa one by plasmin converts the precursor LN5 from a migration ligand to an anchorage ligand in hemidesmosomes (12). However, the mature LN5 with the processed ␣3 chain has been shown to strongly promote cell motility or cell scattering via integrin ␣ 3 ␤ 1 (5,6). It was recently found that the cleavage of the ␣3 chain occurs between the G3 and G4 domains, releasing the G4-5 fragment (13), and that the G3 domain is essential for the potent activity of LN5 to promote cell adhesion and migration (14). It has also been found that the G2 domain contains an integrin ␣ 3 ␤ 1 binding sequence (15). Taken together, these past studies indicate that the G1-3 domain of the ␣3 chain is the primary site to bind cell surface receptors, but the short arms of the ␥2 and possibly ␤3 chains are likely to affect the biological activity of LN5.
To clarify the structure and function relationship of LN5, it seems important to characterize LN6, because these two laminins share the common ␣3 subunit and are often produced by the same cell types. LN6 was first identified as k-laminin in culture media of human keratinocytes and a squamous cell carcinoma line (8,16). In human amnion, about half-amount of LN5 exists in complexes with LN6 or LN7 (17). The rotaryshadowed image analysis of the complexes suggests that LN5 binds with LN6 or LN7 through the interaction of their short arms. It is assumed that the complex formation allows stable association of LN5 with the basement membrane in the amnion (17). However, no previous studies have shown the biological activity of LN6 or LN7 presumably because of the difficulty in isolating these laminins in LN5-free forms. In the present study, we isolated LN6 as an unprocessed single protein and as truncated forms lacking the G4-5, G3-5, or G5 domain of the ␣3 chain. This report describes the biological activities of the four forms of LN6, as well as their differences from LN5.
Purification of LN5 and LN6 -Various recombinant LN5 forms have been isolated from the conditioned media of HT1080 transfectants (14). In the present study, LN5 and three forms of LN6 were purified by basically the same method as before. Briefly, the serum-free conditioned medium of HT1080/WT cells was fractionated by molecular-sieve chromatography on a Sepharose-4B column (Amersham Biosciences) and then by a heparin-Sepharose column. Proteins bound to the heparin column were eluted with 0.5 M NaCl and then with 1.0 M NaCl. Each fraction was passed through a gelatin-Sepharose column to remove fibronectin, and then subjected to two kinds of immunoaffinity chromatographies. LN5 was purified from the 0.5 M NaCl eluate using a LS␣3c4-Sepharose (anti-␣3 chain antibody) column and a D4B5-Sepharose (anti-␥2 chain antibody) column as reported previously (14). To isolate LN6, the 1.0 M NaCl eluate from the heparin column was diluted 10-fold to decrease the NaCl concentration, and then applied to a Q-Sepharose HPLC column (Amersham Biosciences). The materials eluted between 0.3 and 0.4 M NaCl were passed through the D4B5-Sepharose column to remove LN5. Finally, the unbound materials were applied to the LS␣3c4-Sepharose immunoaffinity column. Bound LN6 was eluted with 0.05% trifluoroacetic acid and immediately neutralized. LN6 mutants without the G5 domain of laminin ␣3 chain (LN6⌬G5), without the G4-5 domain (LN6⌬G4-5), and without the G3-5 domain (LN6⌬G3-5) were purified from the conditioned media of HT1080/⌬G5, HT1080/⌬G4-5, and HT1080/⌬G3-5 cells, respectively, by essentially the same procedure as above.
Purification of G4-5 Fragment of Laminin ␣3 Chain-The G4-5 fragment of laminin ␣3 chain (␣3G4-5) was isolated according to the recently published procedure with some modifications (13). Briefly, the concentrated conditioned medium of HT1080/WT cells was fractionated on a Sepharose 4B column and then on a heparin-Sepharose column. Proteins bound to the heparin column at 0.1 M NaCl were eluted sequentially with 0.4, 0.6, and 0.8 M NaCl. The ␣3G4-5 fragment was mostly eluted at 0.8 M NaCl, and this fraction was finally applied to a cation-exchange Mono S HPLC column at 0.1 M NaCl. The ␣3G4-5 fragment was eluted from the column at 0.5ϳ0.6 M NaCl.
Electrophoretic Analyses-SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting were performed as described previously (14). Two-dimensional SDS-PAGE analysis was performed according to the method reported before (20). The first dimensional SDS-PAGE was carried out on a 4% polyacrylamide disc gel (2 mm in diameter and 70 mm in length) under non-reducing conditions. After the electrophoresis, the gel was incubated in a SDS sample buffer containing 5% 2-mercaptoethanol at room temperature for 15 min, placed on a 6% polyacrylamide slab gel (85-mm wide, 1-mm thick, and 70-mm long), and run for the second dimensional SDS-PAGE. Proteins separated on the slab gel were subsequently subjected to the immunoblotting procedure.
Assay of Cell Attachment-The cell attachment assay was performed as described previously (14). 96-well microtiter plates (Corning Costar, Acton, MA) were coated with various substrates in Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline at 4°C overnight, and then blocked with phosphate-buffered saline containing 1.2% (w/v) bovine serum albumin at 37°C for 1.5 h. Cells were suspended in serum-free medium at a density of 2ϳ4 ϫ 10 5 cells/ml, and 100-l aliquots were inoculated onto the plates. After incubation at 37°C, adherent cells were fixed with 2.5% glutaraldehyde and stained with 0.0005% Hoechst 33342, 0.001% Triton X-100 for 1.5 h. The fluorescent intensity of each well of the plates was measured using a CytoFluor 2350 fluorometer (Millipore, Bedford, MA). A blank value corresponding to an empty well was automatically subtracted.
Assay of Cell Scattering and Cell Migration-Cell scattering activity of various forms of laminins toward BRL cells was assayed as reported previously (14). Cells were plated at a density of 7,000 cells/well onto 24-well plates (Sumibe Medical, Tokyo, Japan) containing 0.5 ml/well of Dulbecco's modified Eagle's medium/F12 plus 1% fetal bovine serum. In the standard assay, test samples were directly added into the culture and incubated at 37°C. After 2 days, cell scattering was judged by microscopic observation. Alternatively, test samples were coated on 24-well plates as described above, and the plates were used for assays of cell scattering and cell migration. Cell migration on these substrates was monitored at 37°C with a time-lapse video. The length of cell migration was measured with a video micrometer (VM-30; Olympus, Tokyo, Japan).

RESULTS
Unprocessed and Processed Forms of LN6 -It was previously found that when a cDNA for human laminin ␣3 chain is transfected into human fibrosarcoma cell line HT1080, the exogenous ␣3 chain is assembled with the endogenous ␤3 and ␥2 chains to produce the LN5 hetrotrimer of ␣3␤3␥2 (14). In that study, the conditioned medium of the HT1080 transfectant (HT1080/WT) contained both the 190-kDa and 160-kDa ␣3 chains, but only the LN5 with the 160-kDa ␣3 chain, which had been proteolytically cleaved between the G3 and G4 domains, was purified from the conditioned medium. Antibodies to the laminin ␤3 chain and to the ␥2 chain precipitated the 160-kDa ␣3 chain but not the 190-kDa one from the conditioned medium, suggesting the presence of ␣3-containing laminin(s) other than LN5. In the present study, the conditioned medium of HT1080/WT cells was analyzed by two-dimensional SDS-PAGE and the following immunoblotting with the antibodies to the laminin ␣3, ␥2, ␤1, and ␥1 chains. On the first dimensional non-reducing SDS-PAGE, the ␣3 chain was separated into at least four different molecular sizes, over 1,000 kDa (the top of gel), 600, 450, and 400 kDa (Fig. 1A). The 450-and 400-kDa ␣3 chains were associated with the ␥2 chain, showing that they were the LN5 forms with the 150-and 105-kDa ␥2 chains, respectively (Fig. 1B). This demonstrated that the 190-kDa ␣3 chain had completely been converted to the 160-kDa mature form in LN5 molecule. On the other hand, the ␣3 chain in the 600-kDa complex was not associated with the ␥2 chain ( Fig.  1B). On the second dimensional reducing SDS-PAGE, the ␣3 chain at this position was separated into a 190-kDa major spot and a 160-kDa minor one. Laminin ␤1 and ␥1 chains were separated as broad bands, showing that they existed as different complexes including the ␤1-␥1 heterodimer ( Fig. 1, C and D). The ␤1 and ␥1 chains were found at the position of the 600-kDa ␣3 complexes. These results suggested that the 190-kDa ␣3 chain might exist as LN6 (␣3␤1␥1). The results also imply that the proteolytic processing of the laminin ␣3 chain occurs preferentially in LN5 in the HT1080/WT cells. The highmolecular weight aggregate of the ␣3 chain on the top of gel was not further investigated in this study.
To characterize the 600-kDa complexes containing the laminin ␣3 chain, we attempted to isolate the laminin isoforms from the HT1080/WT conditioned medium. The conditioned medium was fractionated by the molecular-sieve chromatography, followed by the heparin affinity chromatography. Both the 190and the 160-kDa ␣3 chains bound to the heparin column, but the former was eluted at 1.0 M NaCl, whereas the latter was mainly eluted at 0.5 M NaCl (data not shown). The difference in the affinity to heparin was consistent with our previous observation that the G4-5 fragment of the ␣3 chain tightly binds to a heparin column (13). The eluted 190-kDa ␣3 chain was further purified by an anion-exchange HPLC and then passed through the anti-laminin ␥2 antibody column to remove LN5. Finally, the material was bound to an immunoaffinity column conjugated with the anti-laminin ␣3 antibody and eluted therefrom.
The purified material was analyzed by SDS-PAGE and immunoblotting with five kinds of antibodies. The isolated protein migrated as a single band of about 600 kDa slightly beneath mouse LN1 under non-reducing conditions ( Fig. 2A). Reducing SDS-PAGE resolved the 600-kDa laminin into three bands with molecular masses of 220, 210, and 190 kDa (Fig. 2B), and these bands were identified as the laminin ␤1, ␥1, and ␣3 chains, respectively (Fig. 2C). Neither anti-␤3 nor anti-␥2 antibody reacted with any of the three chains (Fig. 2C). These results indicate that the purified 600-kDa laminin is LN6. The LN5 purified by the immunoaffinity chromatography with the anti-laminin ␥2 antibody contained only the 160-kDa form of the ␣3 chain, in addition to the 135-kDa ␤3 chain and the 150-/105-kDa ␥2 chain (hereafter referred as mature LN5; mat-LN5) (Fig. 2, B and C).
We also purified a small amount of LN6 with the 160-kDa ␣3 chain from the 0.5 M NaCl eluate from the heparin column. The 160-kDa ␣3 chain in this preparation was not reactive with the antibody against the G4 domain (data not shown). This indicated that the 160-kDa ␣3 chain had been cleaved between the G3 and G4 domains just like the 160-kDa ␣3 chain in mat-LN5. Therefore, we decided to purify this laminin at a higher yield from the conditioned medium of HT1080/⌬G4-5 cells, which had been transfected with the cDNA for the laminin ␣3 chain lacking both G4 and G5 domains (14). By the same purification procedure as above, a LN6 mutant, which lacked the G4 and G5 domains, was purified. This preparation, named LN6⌬G4-5, contained a 220-kDa ␤1 chain, a 210-kDa ␥1 chain, and a 160-kDa ␣3 chain as analyzed by SDS-PAGE under reducing conditions (Fig. 2, B and C). Furthermore, we prepared a LN6 mutant, which lacked the G3, G4, and G5 domains, named LN6⌬G3-5, from the HT1080/⌬G3-5 cells (14).
Cell Attachment and Spreading Activities of LN6 -To characterize LN6, we compared some biological activities of the three forms of LN6 (LN6, LN6⌬G4-5, and LN6⌬G3-5), mat-LN5 and some other matrix proteins using three different cell lines: human keratinocyte cell line HaCaT, human fibrosarcoma cell line HT1080, and the rat liver cell line BRL. First, attachment of HaCaT cells was examined by inoculating these cells onto plastic plates precoated with different concentrations of the protein substrates (Fig. 3A). mat-LN5, LN6, and LN6⌬G4-5 efficiently supported the cell attachment, but LN6⌬G3-5 did hardly. Compared on a molar basis assuming a similar coating efficiency, the cell attachment activities of mat-LN5, LN6, and LN6⌬G4-5 were comparable, but higher than those of LN1, LN10/11, and fibronectin. The marked difference of the cell attachment activity between LN6⌬G4-5 and LN6⌬G3-5 suggests that the G3 domain is essential for the high affinity binding of LN6 to integrins. Essentially the same dose-response curves were obtained with HT1080 cells (data not shown). In the case of the BRL cells, LN6⌬G3-5 showed a cell attachment activity similar to those of LN1 and LN10/11 (Fig. 3B).
Morphology of HaCaT cells on the different substrates was examined (Fig. 4, a-d). Despite the apparently similar cell attachment activity of mat-LN5, LN6, and LN6⌬G4-5, there was a marked morphological difference between the two LN6 forms. Rapid and extensive spreading of HaCaT cells was induced on mat-LN5 or LN6⌬G4-5, whereas on the LN-6 substrate they barely spread but exhibited some small spikes (or projections). HaCaT cells displayed more flattened morphology on LN10/11 than on mat-LN5 or LN6⌬G4-5. Poor cell spreading on LN6 was reproduced in HT1080 cells (Fig. 4, e-h) and BRL cells (data not shown). These results demonstrate that the proteolytic processing of laminin ␣3 chain converts LN6 from the inactive form to the active form regarding its cell spreading activity.
To compare integrin requirement of mat-LN5 and two forms of LN6 (LN6 and LN6⌬G4-5), effects of function-blocking antiintegrin antibodies on the attachment of HaCaT cells were examined. The cell attachment activity of LN6 was completely blocked by the anti-integrin ␣ 3 or the anti-integrin ␤ 1 antibody (Fig. 5A). Almost the same results were obtained for LN6⌬G4-5 and mat-LN5 (data not shown). HaCaT cells are known to express both integrins ␣ 3 ␤ 1 and ␣ 6 ␤ 4 (21). Therefore, our results imply that in HaCaT cells integrin ␣ 3 ␤ 1 is the primary receptor for these laminins, and integrin ␣ 6 ␤ 4 is not functional at least as the initial receptor. On the other hand, attachment of HT1080 cells to LN6 was partially blocked by the antiintegrin ␣ 3 antibody but completely by the anti-integrin ␤ 1 antibody or the anti-integrin ␣ 3 plus anti-integrin ␣ 6 antibodies (Fig. 5B). Essentially identical results were obtained for LN6⌬G4-5 and mat-LN5 (data not shown). Since HT1080 cells express integrins ␣ 3 ␤ 1 and ␣ 6 ␤ 1 but not integrin ␣ 6 ␤ 4 (3,22), LN6, and LN6⌬G4-5, as well as mat-LN5, are thought to recognize both integrins ␣ 3 ␤ 1 and ␣ 6 ␤ 1 . Integrin ␣ 6 ␤ 1 may work as a secondary (or supplementary) receptor for these laminins. It is also noted that the presence or absence of the G4-5 domain in the ␣3 chain has practically no effect on the integrin binding specificity of LN6.
LN5 is known to require a heat-sensitive helical conformation for the expression of the biological activity (5,23). When the heat stability was examined for LN6, its cell attachment activity was completely abolished after heating at temperature at 80°C for 10 min as in the case of mat-LN5 (data not shown). In contrast, mouse LN1, which contains additional cell-binding sites on the amino-terminal portions of the molecule, maintained ϳ70% of the full cell attachment activity even after heating at 90°C.
Cell Motility Activity of LN6 -LN5 promotes not only cell adhesion but also cell scattering and migration (5, 6). Both cell  scattering and migration on LN5 are though to reflect the enhanced cellular motility. Although most of adherent cultured cell lines efficiently adhere to LN5, a limited number of cell lines are responsive to the cell-scattering and cell migration activities of LN5 (5). Because HaCaT and HT1080 were poorly responsive to the cell motility activity of LN5, BRL cells were used to compare this activity between LN5 and LN6. When each laminin was directly added into the culture of BRL cells in medium containing 1% fetal bovine serum, the typical cell scattering was observed only with mat-LN5 (Fig. 6A). mat-LN5 showed cell scattering at concentrations less than 10 ng/ml, whereas neither LN6 nor LN6⌬G4-5 did even at 200 ng/ml (Fig.  6B). LN10/11 showed no cell scattering activity toward BRL cells (data not shown).
When the cell scattering activity was assayed on the plastic plates precoated with various concentrations of each laminin, mat-LN5 again promoted prominent cell scattering (Fig. 6C). LN6⌬G4-5 induced weak cell scattering, but scarcely LN6 (Fig.  6C). We also analyzed the migration of BRL cells on the laminin-coated plates by the video-microscopy (Fig. 6D). mat-LN5 supported the highest cell migration, whereas LN6 did not stimulate the cell migration. LN6⌬G4-5 stimulated the cell migration dose-dependently, but to a lesser extent than mat-LN5. These results demonstrate that LN6⌬G4-5 has a lower cell motility activity than mat-LN5 though they share the common ␣3 subunit. The results also imply that a relatively low cell motility activity is acquired in LN6 by the loss of the G4-5 domain. The complete lack of the cell motility activity in LN6 may be related with its inability to support cell spreading.
Roles of ␣3G4-5 Domain and Its Released Fragment-It has been reported that bacterially expressed recombinant G4 and G5 domains of laminin ␣3 chain show cell adhesion activity that is inhibited by heparin (15,24). LN5-producing cell lines secrete the G4-5 fragment of ␣3 chain (␣3G4-5) into the culture medium (13). In the culture of HT1080/WT cells, ␣3G4-5 was seen in the extracellular matrix fraction, as well as in the conditioned medium (data not shown). It indicates that a part of the released G4-5 fragment is assembled into the matrix. This was true in the culture of HaCaT cells.
To analyze the biological activity of the G4-5 domain, we purified ␣3G4-5 from the conditioned medium of HT1080/WT cells. The purified ␣3G4-5 protein exhibited a single band of ϳ45 kDa on SDS-PAGE under both non-reducing and reducing conditions (Fig. 7A). Plastic plates precoated with ␣3G4-5 supported the attachment of HaCaT cells (Fig. 7B). The effective concentration for cell attachment was similar in a weight concentration (g/ml) between ␣3G4-5 and LN6, indicating that the specific activity in a molar concentration of ␣3G4-5 was less than one-tenth that of LN6 (Fig. 7B). Although ␣3G4-5 supported cell attachment, it did not promote the spreading of HaCaT and HT1080 cells (data not shown). The cell spreading activity of ␣3G4-5 was absent even at the maximum concentration tested (2 g/ml). However, a low concentration of ␣3G4-5 (0.125 g/ml) promoted the spreading of HaCaT and HT1080 cells in the presence of a low concentration of LN6⌬G4-5 or mat-LN5 (Fig. 7C). These results suggest that the ␣3G4-5 fragment supports cell adhesion independently or in cooperation with mat-LN5 or LN6⌬G4-5. This activity of ␣3G4-5 is in contrast to the predicted role of the G4-5 domain in LN6.
We also investigated the cell surface receptors of ␣3G4-5.
The attachment of HaCaT cells to ␣3G4-5 was almost completely inhibited by either EDTA or heparin (Fig. 8A), but not by the anti-integrin ␤ 1 antibody (data not shown). Pretreatment of HaCaT cells with heparitinase also inhibited their attachment to ␣3G4-5 (Fig. 8A). These results indicate that the cell attachment to ␣3G4-5 is mediated by heparin-like nonintegrin receptors, presumably heparan sulfate proteoglycans (HSPGs). Some divalent cations seem to be required for the active structure of the receptors or ␣3G4-5. Furthermore, heating at 90°C almost completely abolished the cell attachment activity of ␣3G4-5 (data not shown), indicating that it requires the heat-sensitive conformation.
To show the interaction of unprocessed LN6 with HSPGs, the inhibitory effects of heparin and heparitinase treatments were examined. The attachment of HaCaT cells to LN6 was blocked completely by EDTA but only slightly by the heparin or heparitinase treatment of cells (Fig. 8B). A similar weak inhibition of cell attachment by heparin was obtained when LN6⌬G4-5 was used as the substrate (data not shown). When HT1080 cells were used, neither heparin nor heparitinase treatment inhibited the cell attachment to LN6 (Fig. 8C). These results demonstrate that the interaction between the G4-5 within LN6 and cell surface HSPGs is not involved in the cell attachment to LN6. We also found that the treatment of HaCaT cells with heparitinase or heparin did not induce cell spreading on LN6 (data not shown). This also implies that the impaired cell spreading on LN6 is not due to the interaction between its G4-5 domain and cell surface HSPGs.
As a reason for the impaired cell spreading on LN6, we supposed a mechanism in which G4-5 in LN6 might interfere with the interaction between G1-3 and cell surface integrins by a steric hindrance. To examine this possibility, we prepared a LN6 mutant without G5, named LN6⌬G5, from the conditioned medium of HT1080/⌬G5 cells (14), and compared the cell spreading activity toward HaCaT cells among the three LN6 forms (LN6, LN6⌬G4-5, and LN6⌬G5). LN6⌬G5 promoted spreading of HaCaT cells at almost the same level as LN6⌬G4-5 (Fig. 9A). In addition, there is no significant difference in the cell attachment activity toward HaCaT cells among the three LN6 forms (Fig. 9B). Similar results were obtained with HT1080 cells (data not shown). These results clearly indicated that the presence of G5 in LN6, rather than the heparin-binding domain G4, suppresses the cell spreading. LN6⌬G5 showed essentially the same cell scattering activity as LN6⌬G4-5 (data not shown).  2 and 3) conditions. Proteins were stained with silver (lanes 1 and 2), or detected by immunoblotting with a polyclonal antibody against the G4 domain (␣3G4) (lane 3). Panel B, cell attachment activity of LN6 (open circle) and ␣3G4-5 fragment (filled circle) toward HaCaT cells. In this experiment, cells were trypsinized, recovered in 10% fetal bovine serum containing Dulbecco's modified Eagle's medium for 30 min at 37°C followed by washing with serumfree medium, and used. The other experimental conditions are the same as described in Fig. 3. Panel C, effect of combination of ␣3G4-5 fragment with mat-LN5 or LN6⌬G4-5 on spreading of HaCaT (a-f) and HT1080 (g-l) cells. 96-well plates were coated with 0.125 g/ml mat-LN5 (a, d, g, and j) or 0.25 g/ml LN6⌬G4 -5 (b, e, h, and k) in the absence (a-c and g-i) or presence (d-f and j-l) of 0.125 g/ml ␣3G4-5 fragment. After blocking with bovine serum albumin, the cells were plated on each substrate and incubated at 37°C for 1 h. Neither mat-LN5 nor LN6⌬G4-5 supported cell spreading by itself at the concentration chosen, whereas they promoted cell spreading by the simultaneous coating with ␣3G4-5 of the plates .   FIG. 8. Effects of EDTA, heparin, and heparitinase treatment on attachment of HaCaT and HT1080 cells to ␣3G4-5 fragment or LN6. Plastic plates were coated with 0.5 g/ml ␣3G4-5 (panel A) or 1 g/ml LN6 (panels B and C). HaCaT cells (panels A and B) or HT1080 cells (panel C) were incubated on the coated plates in the absence (None) or presence of 5 mM EDTA or 10 g/ml heparin at 37°C for 1 h. To see the effect of heparitinase treatment, cells were pretreated with 2 milliunits/ml heparitinase (100703, Seikagaku Corp., Tokyo, Japan) at 37°C for 30 min (Hep.ase). As a negative control, cells were treated with the same enzyme inactivated by incubation with 1 mM ZnCl 2 at 100°C for 10 min (None).
Functional Interplay of LN5 and LN6 -LN6 was originally found as the covalent complex with LN5 (17). Since these laminins are expected to cooperatively function in vivo, we examined cooperative effect of unprocessed LN6 and mat-LN5 on the adhesion of HaCaT cells (Fig. 10). LN6 did not support cell spreading by itself even at 1 g/ml, but it stimulated cell spreading in the presence of a low concentration of mat-LN5 that did not support cell spreading. This suggests the possible cooperative action of LN6 and mat-LN5 in vivo. mat-LN5 at a concentration higher than 0.5 g/ml supported cell spreading within 20 min by itself. An excess amount of LN6 did not have any significant effect on the spreading of HaCaT cells induced by mat-LN5 alone.
We further examined the direct interaction of LN6 and mat-LN5. Simple mixing of purified LN6 and mat-LN5 in solution did not lead to the formation of LN5-LN6 complex as analyzed by immunoprecipitation (data not shown). In addition, we could not identify the LN5-LN6 complex in the conditioned medium of the HT1080/WT cells. DISCUSSION The laminin ␣3 chain forms the trimeric assemblies of ␣3␤3␥2 (LN5), ␣3␤1␥1 (LN6), ␣3␤2␥1 (LN7), or ␣3␤2␥3 (LN13). Of these laminins, LN5 is known to promote both cell adhesion and migration. The unique activities of LN5 are likely to be in large part mediated by the interaction of the carboxylterminal globular domain with cell surface receptors. This predicts that the other three ␣3-laminins (LN6, LN7, and LN13) may have a similar activity to LN5. In the present study, we found that the HT1080 cell line transfected with the laminin ␣3 chain cDNA (HT1080/WT) secreted LN6 with a 190-kDa ␣3 chain and a small amount of its processed form with a 160-kDa ␣3 chain lacking the G4-5 domain, as well as the mature LN5 with the 160-kDa ␣3 chain (mat-LN5). We isolated the unprocessed LN6 from HT1080/WT cells, as well as three recombinant LN6 forms lacking the G4-5 domain (LN6⌬G4-5), the G3-5 domain (LN6⌬G3-5), or the G5 domain (LN6⌬G5) from different cDNA transfectants of HT1080 cells. The comparative analysis of the biological activities of these LN6 forms and mat-LN5 disclosed interesting differences. First, the unprocessed LN6 and LN6⌬G4-5 showed a marked difference in capability to promote cell spreading. LN6 and LN6⌬G4-5, as well as mat-LN5, showed an apparently similar and high cell attachment activity through integrins ␣ 3 ␤ 1 and ␣ 6 ␤ 1 . This indicates that LN6 and LN5 have a common receptor-binding specificity, and that the presence of G4-5 domain has practically no effect on the integrin-binding specificity. However, morphological examination revealed that LN6⌬G4-5 and mat-LN5, both of which lack the G4-5 domain in the ␣3 chain, promoted cell spreading efficiently, but the unprocessed LN6 did not. This implies that the G4-5 domain of LN6 negatively regulates cell spreading. Second, there is a clear difference in stimulation of cell migration between LN6 and LN5. Especially when added into culture medium, the two forms of LN6 showed no cell scattering activity toward BRL cells, whereas mat-LN5 strongly stimulated cell scattering.
Recent studies have suggested that LN5 interacts with integrins through the G1-3 domain of the laminin ␣3 chain (14,15). Especially, G3 seems to play an important role in regulating cell adhesion and migration (14,25). In this study, we found that LN6⌬G4-5 promoted cell adhesion much more efficiently than LN6⌬G3-5, which lacked G3-5, in good agreement with our previous findings on the role of the G3 domain in LN5. On the other hand, the function of the G4-5 domain has not been understood even in LN5. Bacterially expressed recombinant G4 and G5 have been reported to show a weak cell attachment activity, presumably binding to cell surface HSPGs (15). More recently, the recombinant G4 protein of the ␣3 chain, but not G5, was shown to bind heparin and syndecans-2 and -4 (24). We have found that the recombinant G4 protein binds syndecans-1 and -4 extracted from the membrane fraction of HT1080 cells. 2 Furthermore, the natural G4-5 fragment released from the precursor LN5 has a strong heparin binding activity and stimulates cell scattering in the presence of a low concentration of mat-LN5 (13). There are reports showing that the G4 or G5 domains of other laminin ␣ chains contain binding sites for 2 T. Hirosaki, K. Suzuki, and K. Miyazaki, unpublished data. syndecans and ␣-dystroglycan, despite their relatively low sequence homology among different ␣ chains (26 -29).
In the present study, the secreted G4-5 fragment supported cell attachment by binding cell surface HSPGs, and in combination with a low concentration of LN6⌬G4-5 or mat-LN5 it promoted even cell spreading. It is very likely that the G4-5 fragment induces an intracellular signaling to promote cell adhesion through interaction with cell surface HSPGs such as syndecans. We also found that the G4-5 fragment was deposited on the matrix produced by the LN5-secreting cells. These results strongly suggest that in vivo the G4-5 fragment released from LN5 contributes to cellular adhesion, presumably cooperating with mat-LN5 and other cell adhesion molecules. This biological activity of the G4-5 fragment is apparently contradictory to the finding that the G4-5 domain in the LN6 molecule rather suppresses the cell spreading and motility activities of LN6. Although the cell adhesion to the G4-5 fragment was completely blocked by heparin or heparitinase treatment of the target cells, the impaired cell spreading activity of LN6 was not rescued by either treatment. We also found that not only LN6⌬G4-5 but also LN6⌬G5 supports cell spreading. The major heparin-binding site in the G domain of the ␣3 chain has been shown to be located in G4 but not G5 (24). All these facts clearly indicate that the lack of cell spreading activity in LN6 is not due to the interaction between G4-5 and HSPGs. The presence of the G5 domain of the ␣3 chain rather than the heparin-binding domain G4 is responsible for the impaired cell spreading activity of LN6. A recent study has demonstrated the crystal structure of the G4-5 domain of laminin ␣2 chain (30). Based on the analysis, a model of the entire G domain of laminin ␣ chain has been proposed (31). This model predicts that the G1-3 domain of laminins has a shape of a cloverleaf in contact with the rod domain, and G5 is located closer to the G1-3 cloverleaf than G4. In this model, G5 is very likely to reduce or interfere with the interaction between the G1-3 domain and integrins by a steric hindrance. If this is true in LN6, deletion of G5 or G4-5 from the entire G domain should allow efficient G1-3 binding to integrins. Our experimental results agree well with this model, strongly suggesting that G4-5 in LN6 partially masks the integrin-binding site in the G1-3 domain of the ␣3 chain. It is also noted that the proteolytic processing of the ␣3 chain occurs in LN5 and LN6 with different efficiency. In HT1080/WT cells, the ␣3 chain of LN5 was completely processed to the 160-kDa form, whereas that of LN6 remained in large part intact. It is very likely that the cleavage site between G3 and G4 is exposed in LN5 but masked in LN6. This also suggests a special conformation of the G domain in LN6. On the other hand, it is well known that integrins are colocalized or associated with not only syndecans (32,33) but also other membrane proteins including tetraspanins (transmembrane-4 superfamily proteins) (34,35). We cannot exclude the possibility that the interaction of G5 with such integrinassociated proteins other than HSPGs suppresses the integrininduced cytoskeletal changes.
The present study also demonstrated that LN6 and LN5 have distinct biological activities though they share the common ␣3 subunit. The poor cell motility activity of LN6⌬G4-5 as compared with mat-LN5 indicates that the activity of mat-LN5 depends not only on the ␣3 chain but also on the other two chains ␥2 and ␤3. Indeed, it has been reported that a specific cleavage of the ␥2 chain of mat-LN5 by matrix metalloproteinases increases the ability of mat-LN5 to stimulate cell migration (9,10). It was also reported that the mat-LN5 with the unprocessed ␥2 chain has a higher cell adhesion activity than that with the processed ␥2 chain (11). However, it remains to be clarified how the ␥2 and/or ␤3 chains affect the interaction of the G domain of the ␣3 chain with integrins. All of the LN5 subunits (␣3, ␤3, and ␥2) have the truncated amino-terminal structures, and the ␤3 and ␥2 chains are found only in LN5. This unique amino-terminal structure of LN5 may be responsible for the high cell motility activity.
Physiological roles of LN6 are mostly unknown. This laminin isoform has been found in the cultures of keratinocytes and squamous carcinoma (8,16), and in tissue extracts from the skin and the amnion (17). Approximately half of the LN5 extracted from the amnion and the skin is covalently associated with LN6 or LN7 (17). LN5 is thought to play an essential role in the epithelial-stromal attachment as a monomeric form and as a complex form with LN6 or LN7. Monomeric LN5 works as the primary bridge between integrin ␣ 6 ␤ 4 in the hemidesmosomes and type VII collagen in the stroma (36). LN6 cannot substitute for LN5 in stabilizing epithelial attachment because it does not bind to type VII collagen (36). This is also supported by the fact that genomic mutations in not only the laminin ␣3 but also the ␤3 or ␥2 chain result in Herlitz's junctional epidermolysis bullosa, which exhibits severe detachment of the epidermis from the dermis (37)(38)(39). On the other hand, LN6 and LN7 have a nidogen-binding site identified within the laminin ␥1 chain (40), and the VI domains of both the ␤1 (or ␤2) and ␥1 chains allow the assembly of the laminin network. Therefore, the formation of the LN5-LN6 or LN5-LN7 complex may be able to mediate the stable epithelial attachment and basement membrane assembly. In the complex, LN5 appears to associate with LN6 or LN7 through binding of the short arms from both laminins (17). In vivo the carboxyl-terminal globular domain of the LN5 ␣3 chain in the LN5-LN6 or LN5-LN7 complex is likely to bind to integrin ␣ 3 ␤ 1 or ␣ 6 ␤ 4 , while that of the LN6 ␣3 chain or LN7 ␣3 chain may be free (36). This model appears to be consistent with our finding that the unprocessed LN6 itself has a very poor cell spreading activity. Although mat-LN5, LN6, and LN6⌬G4-5 recognize both integrins ␣ 3 ␤ 1 and ␣ 6 ␤ 1 , the affinity of the unprocessed LN6 to integrins may be far lower than that of mat-LN5 due to the presence of G4-5. Therefore, we suppose a model that the G4-5 in LN6 may interact with HSPGs assembled into the matrix to stabilize the LN5-LN6 or LN5-LN7 complex. This model is supported by a recent study showing that in laminin-2 the G4-5 domain contributes to basement membrane assembly (41).
In the present study, we could not detect the LN5-LN6 complex in the conditioned medium of HT1080/WT cells. Even if the purified mat-LN5 and LN6 were mixed in solution, the LN5-LN6 complex was not detected, suggesting that they do not have a high affinity. Special microenvironment or conditions may be necessary for the complex formation. We also found that the monomeric LN6 poorly supported cell spreading by itself, but it promoted cell spreading in the presence of mat-LN5. This suggests a synergistic action of the two free laminins to support cell adhesion in vivo. On the other hand, the loss of G4-5 from LN6 by proteolytic processing leads to the production of the active LN6 and the G4-5 ligand. The processed monomeric LN6 (LN6⌬G4-5) stimulates epithelial cell adhesion and migration more effectively than the unprocessed LN6. Although it is unknown where and when LN6 undergoes the proteolytic processing of the ␣3 chain in vivo, this event may be important under some physiological and pathological conditions that induce expression of LN6 and LN7 (42).