The LG3 Module of Laminin-5 Harbors a Binding Site for Integrin (cid:1) 3 (cid:2) 1 That Promotes Cell Adhesion, Spreading, and Migration*

Laminins are a family of extracellular matrix glycoproteins involved in cell adhesion and migration. A major obstacle to understanding their structure-function relationships is the lack of small laminin domains capable of replicating integrin-binding, cell-adhesive, and migratory functions of the intact molecule. Here, we show that the recombinant LG3 (rLG3) module (26 kDa) of laminin-5 (Ln-5) (cid:1) 3 chain replicated key Ln-5 activi- ties. rLG3 but not rLG1 or rLG2 supported cell adhesion and migration of at least two distinct cell lines, in an integrin (cid:1) 3 (cid:2) 1 -dependent manner. Cell adhesion to rLG3 was regulated by divalent cations and accompanied by cell spreading and tyrosine phosphorylation of FAK focal adhesion kinase. The integrin binding activity of rLG3 was confirmed by rLG3 affinity chromatography of detergent cell lysates, which resulted in specific purification of integrin (cid:1) 3 (cid:2) 1 . To our knowledge, this is the first report directly demonstrating that a recombinant laminin LG module is an active domain capable of supporting integrin-dependent cell adhesion and migration. rLG3. rLG3 or Ln5 were either coated to plastic wells or used in solution at concentrations of 25 and 1 (cid:5) g/ml, respectively. Cells were allowed to adhere to coated plastic or uncoated control or, alternatively, incubated in suspension with or without test proteins fo r 1 h at 37°C. Total FAK was immunoprecipitated and subjected to Western blotting analyses with antibodies to phosphotyrosine or FAK protein. The results are depicted by images from one representative Western blot and by a bar graph generated with results of three experiments (mean (cid:6) S.D. ( n (cid:7) 3)). The method used for calculating relative -fold induction of FAK phosphorylation is described in detail under “Experimental Procedures.”

Ln-5 (460 kDa) is a heterotrimer with the subunit composition ␣ 3 ␤ 3 ␥ 2 (8) and resembles a Y-shaped structure when viewed by rotary-shadowed electron microscopy (20). The short arms of the "Y" represent ␤ and ␥ chains. The long arm consists of heterotrimerization domains followed by a prominent globular structure, the ␣ chain G domain, composed of five tandem repeats, called laminin G-like, or LG repeats. While LG monomers and dimers are found in other proteins, a continuous string of five LG modules are only found in the laminin ␣ chains (21). The five modules are linked together by short (ϳ5-10 residues) intermodular sequences, except between the third and the fourth module, where the linker sequence is ϳ50 residues in length, essentially dividing the G domain into two separate structural components, the LG123 triplet and the LG45 pair (19,21).
Here, we report expression of the proximal LG123 modules as monomeric, soluble fusion proteins. In functional analyses, purified recombinant LG3 (rLG3), not rLG1 or rLG2, replicated key Ln-5 activities including cell adhesion, spreading, and migration. Antibody inhibition assays, cation studies, and rLG3 affinity chromatography demonstrated that rLG3 is a ligand for integrin ␣ 3 ␤ 1 . To our knowledge, this is the first report directly demonstrating that the LG3 module is an integrinbinding, cell-adhesive, and migratory domain of Ln-5. Furthermore, this is the first recombinant LG domain from any laminin isoform that displays integrin-dependent activities.
Induction of protein expression in Escherichia coli JM109 strain grown to midlog phase in Luria-Bertani medium was carried out with 1.0 mM isopropyl-␤-D-thiogalactopyranoside (Roche Molecular Biochemicals). After protein induction for 5 h at 34°C, cells were harvested by centrifugation at 6000 rpm for 10 min (Sorvall, GSA rotor). Cell pellets were stored at Ϫ80°C until use. For protein purification, pellets were thawed and resuspended in ice-cold lysis buffer (20 mM Tris-HCl, pH 7.8, 10% glycerol, 0.15 M NaCl, 5 mM imidazole, 1% Triton X-100) containing 1 mM phenylmethylsulfonyl fluoride (Sigma). Cell lysis was carried out by passage of cell suspension through French Press (SLM Instruments, Urbana, IL) at 1500 p.s.i. twice. Clarified cell lysates were applied to a Ni 2ϩ -nitrilotriacetic acid (Qiagen, Valencia, CA) column. The column was washed with wash buffer (20 mM Tris-HCl, pH 7.8, 10% glycerol, 0.5 M NaCl, 60 mM imidazole). Protein elution was carried out with 300 mM imidazole and 1 mM dithiothreitol (Sigma) in wash buffer. Histidine-tagged (i.e. without glutathione S-transferase (GST)) LG proteins were expressed using the pET-30a(ϩ) expression vector (Novagen, Madison, WI) in E. coli. Recombinant His-tagged LG3 were also expressed in High Five TM insect cells (Invitrogen, Carlsbad, CA) under serum-free culture conditions using Pharmingen's Baculovirus Expression Vector System. Circular Dichroism Spectroscopy-Protein samples at 0.2-0.45 mg/ml were prepared in Tris-buffered saline containing 10% glycerol and 1 mM dithiothreitol. CD spectra were recorded on an AVIV model 202SF stop-flow circular dichroism spectrometer (Lakewood, NJ), using a thermostatted cell of 0.1-mm path length. Protein samples were analyzed at 23°C from 200 to 300 nm with a bandwidth of 1 nm. Three repetitive scans were averaged and smoothed by binomial curve smoothing. The molar ellipticity (in degrees cm 2 dmol Ϫ1 ) was calculated on the basis of protein concentration and molar mass for each rLG.
Adhesion Assays-Cell adhesion assays were performed in 96-well GSH-coated plates (Pierce) that were pretreated with SuperBlock TM , a proprietary blocking reagent containing a mammalian derived protein (Pierce). The GSH plates were first coated with rLGs overnight at 4°C. Unbound protein was removed with phosphate-buffered saline washing. Cells were detached with 0.025% trypsin and 0.53 mM EDTA (Life Technologies). Serum-free trypsin-neutralizing solution (Clonetics, San Diego, CA) was used to neutralize trypsin. Cells were then washed twice with serum-free DMEM containing 25 mM HEPES, resuspended in serum-free DMEM/HEPES at a cell density of 8.0 ϫ 10 5 cells/ml. For each well, 100 l of cell suspension was seeded and allowed to attach for 1 h at 37°C. Nonadherent cells were removed; adherent cells were fixed with glutaraldehyde (3% (v/v) in phosphate-buffered saline) and stained with crystal violet (3% (w/v) in methanol); and absorbance was measured at 595 nm. Cell spreading analysis was performed using the public domain NIH Image program (version 1.61) available on the Internet at rsb.info.nih.gov/nih-image/. For adhesion inhibition assays, cells were preincubated with antibodies at room temperature for 15 min prior to seeding. For cation studies, cells were washed with Ca 2ϩ /Mg 2ϩ -free Hanks' balanced salt solution (Life Technologies) containing 2 mM EDTA to deplete preexisting cations prior to trypsinization. Cells were then trypsinized, washed twice with cation-free Hanks' balanced salt solution, and resuspended in cation-free Hanks' balanced salt solution at 8.0 ϫ 10 5 cells/ml (100 l per well was used). Cations (i.e. CaCl 2 , MgCl 2 , and MnCl 2 ) at the indicated concentrations were preincubated with the cells for 15 min at room temperature prior to seeding.
Migration Assays-Cell migration assays were carried out as previously described (31). Briefly, Transwell filters (Costar, Cambridge, MA) were coated on the lower side with purified rLGs at 3.5 g/ml. The lower side of the filter was then blocked with 5% milk in phosphate-buffered saline for 2 h. Cells were prepared as in adhesion assays. 8.5 ϫ 10 4 cells in 600 l of serum-free DMEM containing 0.1% bovine serum albumin were added to the upper chamber and allowed to migrate for 18 or 36 h at 37°C under tissue culture conditions. Cells that migrated to the bottom surface of the filter were fixed with methanol, stained with crystal violet, and counted. Each substrate was repeated in duplicate wells, and within each well counting was done in four randomly selected microscopic fields (ϫ 300 magnification).
FAK Phosphorylation Assays-FAK phosphorylation assays were performed as previously described (32). In brief, HT-1080 cells were trypsinized, washed, and kept in suspension in serum-free DMEM for 30 min at 37°C to reduce background FAK phosphorylation level. For the adherent FAK assay, cells were plated onto uncoated, Ln-5-coated (1 g/ml), or LG3-coated (25 g/ml) wells and allowed to attach for 1 h. For the suspension FAK assay, the same number of cells were kept in suspension in the presence of various test proteins at the same concentration used for the adherent assay. At the end of 1 h of incubation, cells were lysed with lysis buffer (40 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 3 mM EDTA, 10% glycerol, 1% Triton X-100, 50 mM NaF, 1% deoxycholate, 0.1% lauryl sulfate, 1 mM Na 2 VO 4 ; 25 mM Na 4 P 2 O 7 ). Total FAK protein was immunoprecipitated with anti-FAK pAb that was immobilized onto protein G-conjugated Sepharose beads (Amersham Pharmacia Biotech). Immunoprecipitated FAK protein was subjected to Western blotting analyses with phosphotyrosine antibody PY20 to assess the extent of tyrosine phosphorylation and anti-FAK mAb to determine the amount of FAK. The relative -fold induction of FAK phosphorylation, which is defined as the increase in FAK phosphorylation of a sample relative to its corresponding background level (i.e. cells plated on plastic or kept in suspension) was determined as followed. First, densitometry intensity of phosphorylated FAK (p-FAK) and total FAK (t-FAK) signals from Western blots were measured with a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA). The ratio of p-FAK to t-FAK (p-FAK/t-FAK) for each sample was then calculated to correct for differences in protein loading. This ratio was then normalized to the p-FAK/t-FAK ratio of the negative controls to obtain the relative -fold induction of FAK phosphorylation.
LG3 Affinity Chromatography-The rLG3 affinity column was prepared with 1 ml of GSH-Sepharose (Amersham Pharmacia Biotech) and 5 mg of purified rLG3. Protein immobilization efficiency was ϳ80%, as determined by absorbance readings (at 280 nm) of applied and flowthrough fractions. Since the binding of GSH to GST is high affinity, no cross-linking was performed. The column was equilibrated with cell lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100) until use.

Expression, Purification, and Characterization of Recombinant
LG Modules of Ln-5-To identify integrin-binding, adhesive domains of Ln-5, the three proximal LG modules (i.e. LG1, -2, and -3) of Ln-5 ␣ 3 chain ( Fig. 1A and 2A) were expressed as monomers in E. coli. Amino acid sequence analyses of Ln-5 LG3 module orthologues (i.e. LG3 from rats, humans, and mice) revealed that the rat LG3 sequence was considerably more hydrophilic in three major regions (Fig. 1, B and C, highlighted by bars) than the corresponding human sequence. This obser-Laminin-5 LG3 Module Mediates Cell Adhesion via Integrin ␣ 3 ␤ 1 vation may partially explain why the human Ln LG3 modules (derived from Ln ␣ 2 and ␣ 3 chains) failed to express recombinantly in previous attempts (15,28) and suggested that the Ln-5 LG3 module from rat may be more amenable to recombinant expression. In addition to switching species, three additional strategies were used to facilitate recombinant protein expression: 1) each LG construct was extended by 10 amino acids into neighboring modules (Figs. 1C and 2A, top panel) to facilitate protein folding; 2) the N-terminal nonbonding cysteine residue of each module was converted to serine, while the two C-terminal pairing cysteines were left intact (Figs. 1C and 2A, top panel) to favor correct disulfide bond formation; and 3) the recombinant LG modules were expressed as double-tagged fusion proteins, containing a GST moiety at the N terminus and a His 6 tag at the C terminus ( Fig. 2A, top panel) to provide convenient handles on protein purification and functional assays.
rLG modules were purified to near homogeneity with Ni 2ϩnitrilotriacetic acid-agarose under nondenaturing conditions as determined by Coomassie staining of SDS-polyacrylamide gel ( Fig. 2A). Purification with GSH affinity column provided further improvement in purity ( Fig. 2A). The expression levels for all three rLGs were similar; on average, about 0.5 mg of purified rLG protein was obtained from 1 liter of bacterial culture (data not shown).
To determine whether or not an intramolecular disulfide bond formed in the rLGs, we subjected the purified recombinant proteins to SDS-PAGE under reducing or nonreducing conditions and looked for mobility differences. Treatment of rLGs with ␤-mercaptoethanol (10%, v/v) prior to SDS-PAGE caused a small but reproducible reduction in gel mobility, suggesting that intramolecular disulfide bonds are present in all three recombinant proteins (Fig. 2B).
The secondary structure of rLGs was assessed by CD spectroscopy. With the GST-tagged rLGs, obtaining CD data of the LG portion of the fusion proteins requires CD spectral subtraction between the two recombinant proteins (i.e. GST-LGs and GST). This extra step requires protein concentrations to be precisely determined in order to avoid introducing errors due to over-or undersubtraction of CD data. To more directly and accurately determine the CD spectra of rLGs, they were subcloned into a pET vector, expressed as His 6 -tagged recombinant proteins (without the N-terminal GST moiety), and used in CD studies. CD spectra of recombinant His 6 -LG1, recombinant His 6 -LG2, and recombinant His 6 -LG3 showed ellipticity minima at 217, 215, and 214 nm, respectively (Fig. 2C), characteristic of proteins rich in ␤ structure (33,34). Our findings are in good agreement with previously reported CD values for other LG homologs (27,35) and are consistent with the crystal structures of Ln-2 LG5 monomer and LG45 dimer, which adopt an overall fold of 14-stranded ␤ barrel (21,36).
In summary, these results suggest that the bacterially expressed LG modules from Ln-5 are sufficiently well folded to support potential functional activity.
Cell Adhesion, Spreading, and Migration Activities of rLGs-The adhesive activity of GST-tagged rLGs was evaluated in adhesion assays. Purified rLGs proteins were adsorbed onto GSH-coated microtiter wells via their N-terminal GST moieties LG3 module indicate that rat LG3 is more hydrophilic and polar than human sequence. A, subunit and domain organization of Ln-5. A unique feature of the Ln-5 ␣ 3 chain is the presence of a globular structure (G domain) at the end of the long arm. The G domain is comprised of five tandem laminin G-like (LG) repeats, the first three (i.e. LG1, -2, and -3) of which were implicated in integrin-dependent cell adhesion and migration. B, upper panels, hydrophilicity (64) and polarity (65) plots of rat and human LG3. Bars (in dark blue) indicate regions where the rat sequence is considerably more hydrophilic or polar than the human one. Lower panel, comparison of overall hydrophilicity and polarity scores of rat and human LG3. *, net hydrophilicity represents the sum of hydrophilicity scores (positive and negative values) of all of the amino acid residues in LG3; **, average polarity represents the sum of all polarity scores (all positive values) divided by the total number of residues. C, sequence alignment of LG3 modules of rats, mice, and humans. Rat LG3 is 81 and 93% homologous to human and mouse LG3, respectively. The sequences within LG3 corresponding to the marked regions in B are also highlighted in bars. Conserved cysteine residues predicted to form a disulfide bond are highlighted in yellow. The nonbonding cysteine, indicated with a light blue dot, was mutated to serine. Each recombinant LG construct was extended by 10 amino acids into neighboring modules to facilitate protein folding (as marked by arrows).
LG3 sequence is shown here as an example.
in order to favor uniform protein orientation and minimize denaturation. rLG3 showed unequivocal adhesion activity with two distinct cell lines, the fibrosarcoma HT-1080 and the hepatoma HLF (Fig. 3, A (left of dashed line) and B). In contrast, rLG1 and rLG2 displayed minimal or no adhesion activity under our conditions, even at high coating concentrations, compared with the GST control (Fig. 3, A (left of dashed line) and B). HT-1080 adhesion on GST-tagged rLG3 was dose-dependent with maximal adhesion occurring at 25 g/ml or 0.5 M coating concentration. At this or higher coating concentrations, the adhesion activity of GST-tagged rLG1, rLG2, and GST remained negligible (Fig. 3B). Adhesion assays were also performed with His 6 -tagged rLGs (without the N-terminal GST fusion) to rule out possible effects on biological activities by the presence of GST, if any, on the LG modules. The adhesion activity profile of His 6 -tagged rLGs was very similar to those of GST-tagged counterparts (Fig. 3A, right of dashed line), suggesting that in the LG fusion proteins the GST moiety is functionally inert. In addition, His 6 -rLG3 was expressed in insect High Five TM cells, purified to near homogeneity with Ni 2ϩnitrilotriacetic acid-agarose (data not shown), and tested in adhesion assays. Insect expressed His 6 -rLG3 was active in promoting cell adhesion of HT-1080 cells (Fig. 3A, right of dashed line), suggesting that the adhesion activity observed with the bacterially expressed rLG3 is preserved in the rLG3 obtained from an eukaryotic expression host. The higher adhesion activity observed with the insect-derived rLG3, however, is probably due to more efficient and complete intramolecular disulfide bond formation in eukaryotic cells. Since protein yields were the highest with the bacterially derived, GSTtagged rLG fusion proteins (data not shown), and no gross functional difference was observed between the rLGs generated LG modules (rLG) were purified first over nickel (Ni) and then GSH columns. Eluted material was subjected to SDS-PAGE analysis (11% acrylamide, reducing) and visualized by Coomassie staining. In the gel shown, purified rLGs were compared with GST, purified over a GSH column, and compared with a total cell lysate of the rLG3 clone before purification. B, gel mobilities of purified rLGs treated with ␤-mercaptoethanol (␤-MeOH) were compared with nontreated rLGs under 10% SDS-PAGE conditions. Reduction of rLGs prior to electrophoresis resulted in reproducible decrease in gel mobility. C, circular dichroism analyses of rLGs in Tris-buffered saline, pH 7.5, at 23°C. The CD spectra of all three rLGs showed ellipticity minima at around 210 -220 nm, indicating that the purified rLGs consist of predominantly ␤ structure.

FIG. 3. Recombinant
LG3 supports cell adhesion and migration. A, purified GST-tagged rLG proteins and GST were coated onto 96-well GSH plates at 50 g/ml and Ln-5 at 1 g/ml. Purified histidinetagged (His 6 ) rLG proteins were adsorbed onto 96-well, non-tissue culture-treated polystyrene plates at 50 g/ml. Blotto (dry milk in phosphate-buffered saline) was used as a negative control. HT-1080 and HLF cells were allowed to adhere for 1 h in serum-free medium. Unbound cells were washed off; adherent cells were fixed, stained (crystal violet), and solubilized; and absorbance was read at 595 nm. Only rLG3 supported cell adhesion, ϳ3-fold lower than Ln-5 under these conditions. In the bar graph, results are expressed as mean Ϯ SD (n ϭ 3). B, dose-dependent HT-1080 adhesion to immobilized rLG3. Purified recombinant proteins were coated at indicated concentrations. Similar dose dependence was observed with HLF cells (data not shown). The adhesion assay conditions were same as described in A. Results represent mean Ϯ S.D. (n ϭ 3). C, haptotactic migration of HT-1080 cells toward rLG3-coated Transwell filters. Cells that migrated through the filter to the lower side were fixed and stained (crystal violet). Cell migration was quantified by counting migrated cells in eight microscope fields. Results are expressed as mean Ϯ S.D. (n ϭ 3). by these three expression approaches, all of the remaining assays in this paper were performed with the GST-tagged fusion proteins expressed in E. coli.
To further evaluate the adhesion activity of rLG3, we examined adherent cells to distinguish between cell tethering to the substratum and cell spreading. HT-1080 and HLF cells adherent on rLG3 were microphotographed in the adhesion assay wells after washing, fixation, and staining with crystal violet. On rLG3, 64% of HT-1080 and 53% of HLF cells displayed spreading morphology; i.e. they adopted a flattened, polygonal shape, with filopodia-and lamellipodia-like extensions (Fig. 4). The remaining nonspread cells on rLG3 resisted washing and remained tethered to the plate surface (Fig. 4). As expected, no cell spreading was observed on either rLG1-or rLG2-coated plates (Fig. 4). These results indicate that rLG3 is similar to Ln-5 in that it supports cell spreading.
Since Ln-5 is an adhesive as well as migratory substrate (4, 37), we evaluated haptotactic migration of rLGs in standard Transwell assays. Coating the underside of Transwell filters with rLG3 promoted migration of HT-1080 and HLF cells (Figs. 3C and 5B), whereas no cell migration was detected in rLG1-, rLG2-and blotto-coated wells after 18 h of incubation (Fig. 3C). Prolonging the incubation time to 36 h caused ϳ30% increase in cell migration number with the rLG3 substrate (data not shown), suggesting that the majority of cell migration is completed by 18 h. Under chemotactic migration conditions where soluble rLGs were added in solution to the lower chamber of Transwell devices, no cell migration was detected for any of the substrates tested (data not shown), suggesting that rLG3 functions mainly as a haptotactic migration stimulus.
In summary, the rLG3 module displayed functional properties similar to its parent molecule Ln-5 because it promoted adhesion, spreading, and migration of two distinct cell lines.
Integrin-dependent Cell Adhesion and Migration on rLG3-The integrin dependence of rLG3 activities was evaluated in adhesion and migration assays in the presence of agents that modify integrin functions.
A similar integrin dependence profile was observed with rLG3-stimulated cell migration. In brief, migration toward the rLG3 substrate was inhibited by ␣ 3 ␤ 1 blocking antibodies, while the ␣ 6 blocking antibody caused a small reduction in cell migration with HLF but not HT1080 cells (Fig. 5B). These patterns of inhibition by integrin antibodies are consistent with those previously observed with Ln-5-induced cell migration (8,43,44), suggesting that the LG3 module is a key integrin binding, migratory domain of Ln-5.
Integrins exhibit multiple affinity states that can be modulated by extracellular activators, such as divalent cations Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ (38,39). We next examined how cell adhesion to rLG3 responded to these cations at various doses. Among the three cations, Mn 2ϩ was the strongest stimulator of cell adhesion, causing maximal adhesion already at 0.5 mM, the lowest concentration used (Fig. 5C). Stimulation with Mg 2ϩ exhibited a dose dependence response, although the overall effect was less compared with Mn 2ϩ (Fig. 5C). In contrast, the Ca 2ϩ effect was the weakest (Fig. 5C), showing a small dose-dependent increase up to 3 mM. These cation effects on rLG3 adhesion activity are comparable with those obtained with Ln-5 (17), 2 further confirming integrin involvement of rLG3 cell adhesion activity.
It has been demonstrated that integrins function not only as adhesion receptors but also as signal transducers by activating intracellular associated protein kinases (45,46). One such upstream linker protein is FAK (47,48). To determine whether or not integrin ␣ 3 ␤ 1 -mediated adhesion on rLG3 induced tyrosine phosphorylation of FAK, cells were allowed to adhere to immobilized substrates (i.e. rLG3 or Ln-5) or incubated with soluble substrates in suspension for 1 h. Total cellular proteins were then extracted with detergent cell lysis followed by immunoprecipitation and Western blotting analyses of FAK to assess the level of tyrosine phosphorylation under various incubation conditions. Adhesion of HT-1080 cells on plates coated with rLG3 and Ln-5 resulted in a 4-and 12-fold increase, respectively, in FAK phosphorylation levels relative to background (Fig. 5D). These levels of induction are consistent with previous findings with Ln-5 (32) and other extracellular matrix substrates. 3 The extent of FAK phosphorylation on Ln-5 was higher than that of FIG. 5. Adhesive and migratory activities of rLG3 are dependent on integrin ␣ 3 ␤ 1 . A, cell adhesion to rLG3 is blocked by antibodies to integrin ␣ 3 ␤ 1 . Adhesion assays were carried out as described in the legend to Fig. 3A, except that cells were preincubated with integrin antibodies for 15 min at room temperature prior to seeding. Anti-integrin antibodies used (10 g/ml) were as follows: ASC-1, anti-␣ 3 ; P4C10, anti-␤ 1 ; GoH3, anti-␣ 6 ; TS2/16, anti-␤ 1 (activating). B, cell migration on rLG3 is blocked by antibodies to integrin ␣ 3 ␤ 1 . Migration assays were performed as described in the legend to Fig. 3C. All antibodies used were the same as in A except for the integrin ␤ 1 antibody. Here, 6S6 was used in place of P4C10 because P4C10 was only available as ascitic fluids, which contain serum components that stimulated cell migration on rLG3 as well as Ln-5 (not shown). See A for specificity of other antibodies. C, divalent cations modulate cell adhesion to rLG3. To evaluate the effect of a particular cation, cells were depleted of extracellular cations by washing with EDTA prior to trysinization. The cations at the indicated concentrations were added to cells in cation-free Hanks' balanced salt solution. Other assay conditions were same as described in the legend to Fig. 3A. D, tyrosine phosphorylation of FAK was induced by HT-1080 adhesion to immobilized rLG3. rLG3 or Ln5 were either coated to plastic wells or used in solution at concentrations of 25 and 1 g/ml, respectively. Cells were allowed to adhere to coated plastic or uncoated control or, alternatively, incubated in suspension with or without test proteins for 1 h at 37°C. Total FAK was immunoprecipitated and subjected to Western blotting analyses with antibodies to phosphotyrosine or FAK protein. The results are depicted by images from one representative Western blot and by a bar graph generated with results of three experiments (mean Ϯ S.D. (n ϭ 3)). The method used for calculating relative -fold induction of FAK phosphorylation is described in detail under "Experimental Procedures." Laminin-5 LG3 Module Mediates Cell Adhesion via Integrin ␣ 3 ␤ 1 rLG3, consistent with the higher adhesion activity of Ln-5 described earlier (Fig. 3A). Interestingly, incubation of Ln-5 with cells in suspension also resulted in FAK phosphorylation, although at a much reduced level compared with coated Ln-5, whereas soluble rLG3 did not. This finding indicates that solid phase presentation of rLG3 is required to induce FAK phosphorylation.
Taken together, these results strongly indicate that rLG3mediated cell adhesion and migration are integrin ␣ 3 ␤ 1dependent.
Purification of Integrin ␣ 3 ␤ 1 by rLG3 Affinity Chromatography-To confirm the integrin binding specificity of rLG3, we performed rLG3 affinity chromatography with detergent cell lysates of HT-1080 and HaCaT cells. rLG3 affinity columns were constructed with Ni 2ϩ -nitrilotriacetic acid-purified GST-rLG3 and glutathione-Sepharose as the column matrix. After application of clarified cell lysates, the columns were washed stringently with a NaCl gradient up to 1 M in a Tris-based buffer. Elution of bound materials with EDTA (10 mM in Tris buffer) resulted in a distinct protein peak in the first three elution fractions (Fig. 6A). The identity of eluted materials was determined by SDS-PAGE under nonreducing conditions followed by Western blotting with integrin antibodies. Antibodies to integrin subunits ␣ 3 and ␤ 1 specifically recognized two bands with apparent molecular masses of 165 and 110 kDa, respectively, indicating that integrin ␣ 3 ␤ 1 was present in the eluted fractions (Fig. 6B). In comparison, antibody to integrin-␣ 5 subunit only recognized a band in the total cell lysate lane but not the eluted fraction lanes (Fig. 6B), demonstrating that the binding of integrin ␣ 3 ␤ 1 to rLG3 affinity column is receptor-specific.
These results further substantiate the notion that rLG3 supports integrin-dependent functions and indicate that integrin ␣ 3 ␤ 1 bind to rLG3 with sufficient affinity to allow for biochemical purification. DISCUSSION It has long been recognized that the primary adhesion site of laminins is localized in the C-terminal G domain region of laminin ␣ chains. However, further definition of cell-binding site(s) to smaller, more manageable laminin fragments had not been achieved (14,22,26,49). In this paper, we directly demonstrate that a small recombinant domain of Ln-5 can reproduce, to a large extent, the activities of the intact molecule by mapping an integrin binding, cell-adhesive domain of Ln-5 to the third module (LG3) of the G domain. This finding, along with the recently published crystal structure of Ln-2 LG45 pair (21), should be useful toward unraveling the structural basis of adhesion activities associated with the G domain modules of laminins.
Our results support the following conclusions: 1) small globular domains of Ln-5 ␣ 3 chain can be expressed as soluble recombinant fusion proteins in a prokaryotic expression system and retain sufficient tertiary structure to support function; 2) the recombinant ␣ 3 LG3 (rLG3) module supports cell adhesion, spreading, and migration; 3) rLG3-mediated adhesion and migration is mediated by integrin ␣ 3 ␤ 1 ; 4) divalent cations, including Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ , influence cell adhesion to rLG3; 5) cell adhesion to immobilized rLG3 is accompanied by tyrosine phosphorylation of FAK; and 6) the binding of integrin ␣ 3 ␤ 1 to rLG3 is of sufficient affinity to allow purification of ␣ 3 ␤ 1 from total cell lysates on a rLG3 affinity column.
Previous attempts to express human rLG3 in a soluble form have not been successful (15,28), so this domain has never been directly investigated for adhesive functions. We circumvented this problem by expressing the rat Ln-5 LG3, since both hydrophilicity and polarity plots of the rat sequence suggested a more hydrophilic protein surface compared with the human one and thus a greater chance of obtaining recombinant proteins in a soluble form. This approach indeed yielded soluble LG modules that were readily purified to greater than 90% homogeneity by a one-step affinity chromatography, utilizing either the C-terminal histidine (His 6 ) tag or N-terminal GST fusion moiety. CD spectroscopy, a sensitive method for determining protein secondary structures, was used to assess the folding of purified rLGs. CD spectra of rLGs (with His 6 tag but not GST) are characteristic of proteins rich in ␤ structure. Our findings are in good agreement with secondary structure predictions performed with Ln- 5 LGs, CD results previously obtained with Ln-1 LG modules (27), and crystal structures of Ln-2 LG4 monomer and LG45 dimer (21,36).
The expression host we used here, E. coli, may raise concerns as to the lack of disulfide bonding machinery in prokaryotic cells and its potential negative effects on rLG structure. However, several considerations may dampen such concerns. First, we chose a prokaryotic expression system for the following well known advantages: shorter protein production time, higher yields, lower cost, more convenient to set up, etc. Second, the tertiary structure Ln-5 rLGs are predicted to adopt (i.e. a Laminin-5 LG3 Module Mediates Cell Adhesion via Integrin ␣ 3 ␤ 1 14-stranded ␤ barrel) has not been shown to be incompatible with prokaryotic expression hosts. A number of proteins with this fold were obtained recombinantly from prokaryotic hosts and retained functional activities (50 -52). Third, since each LG module contains one intramolecular disulfide bridge, and disulfide bond formation may occur outside of the E. coli cytoplasm (i.e. during cell lysis and protein purification). Indeed, gel mobility up-shifts observed with reduced rLGs relative to the nonreduced proteins indicate that the purified rLGs are indeed disulfide-bonded. Fourth, the adhesion-promoting activity of bacterial derived rLG3 was confirmed with the insectderived protein, further strengthening our conclusion that the recombinant rLG fusion proteins purified from E. coli are sufficiently well folded to support biologically activities.
rLG3 replicated Ln-5 activities clearly in an integrin ␣ 3 ␤ 1dependent manner. This conclusion is based on results obtained with cell adhesion assays, in which rLG3, but not rLG2 or rLG1, supported dose-dependent adhesion of both HT-1080 and HLF cells, and in cell migration assays, in which rLG3 stimulated HT-1080 and HLF haptotactic cell migration. Furthermore, the adhesive activity of rLG3 was associated with cell spreading and activation of focal adhesion kinase, via tyrosine phosphorylation. Cations that are known modulators of integrin affinity states (i.e. Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ ) regulated cell adhesion on rLG3 in a cation-specific and concentrationdependent manner. Of particular importance is the general pattern of cation effects, which showed an overall order of Mn 2ϩ Ͼ Mg 2ϩ Ͼ Ͼ Ca 2ϩ , in good agreement with the pattern observed with Ln-5 (17). Antibody blocking experiments demonstrated that cell adhesion and migration on rLG3 were mediated by integrin ␣ 3 ␤ 1 . This result was confirmed by rLG3 affinity chromatography of detergent lysates of two cell lines, which specifically eluted integrin ␣ 3 ␤ 1 but not the other integrins from the affinity column.
To reach half-maximal adhesion, 250-fold more (in molar terms) rLG3 than Ln-5 was required, assuming equal coating efficiency. Maximal adhesion and migration supported by rLG3 were about 30 and 10%, respectively, compared with Ln-5. The difference between adhesion and migration (30% versus 10%) may be technical, e.g. due to coating efficiency, since in adhesion assays rLGs were coated in an oriented fashion via the GST moiety to GSH plates, whereas in migration assays rLGs were directly coated (and hence randomly oriented) onto filter surfaces. It remains to be seen whether the activity differential between rLG3 and intact Ln-5 is entirely technical in nature or if it also reflects functional synergy of Ln-5 modules. The latter is not without precedent from other modular extracellular matrix adhesive proteins. In fibronectin, for example, the celladhesive fragment was mapped to the tenth FIII domain and later reduced to an RGD (Arg-Gly-Asp) sequence within that domain (53). However, the tenth FIII domain in its recombinant form showed a 100 -200-fold reduced adhesion activity compared with intact fibronectin (54,55). To fully reconstitute the adhesion activity of fibronectin, co-expression of the tenth FIII flanking domains was required, since they contribute synergistic cell binding sites (56 -58). If this scenario should be applicable to Ln-5 rLG3, then other LG domains adjacent to LG3 may participate in cell surface receptor binding and adhesion. Preliminary flow cytometry studies showed that both rLG1 and rLG2 bound to the cell surface in a dose-dependent and saturable manner, 3 although we have not yet been able to establish integrin dependence of these cell surface binding activities. When used alone or in trans with rLG3, neither rLG1 nor rLG2 had any effect on cell adhesion or migration, suggesting that the preservation of intermodular orientation may be important to detect synergistic activity of LG modules if any. Further studies, such as the co-expression of the three LG modules as a single recombinant protein, preferably in a eukaryotic expression host, may be necessary for detecting synergistic activities of LG1, LG2, or other LG modules.
An intriguing issue in the laminin field has been the conformation sensitive nature of laminin adhesion activity. While many results implied the localization of adhesion site(s) in the laminin G domain, attempts to replicate integrindependent adhesion activity with recombinant G domain fragments have not been successful (22,23,26,27). The underlying reasons may be due to the changes in G domain packing geometry in the context of intact or large laminin fragments versus smaller recombinant G domain fragments. A recently proposed structural model of LG123 modules predicts extensive interactions between LG1 and its immediately upstream trimerization domain (domain I), which would invariably impose conformation constraints on the LG123 trimer (21). This model also predicts that the absence of domain I may release some constrains on LG123, allowing them to adopt a different, presumably lower energy conformation. Indeed, Yurchenco et al. (27) reported a gain of additional heparin binding but a loss of integrin binding sites when the Ln-2 G domain tandem was expressed without domain I. Interestingly, when domain I was reconstituted back to the LG123 trimer, integrin binding activity was restored (26). Taken together, these observations suggest that changes in domain orientation probably took place, which resulted in the masking of some key functional sites and unmasking of some cryptic sites. Furthermore, the conformation-sensitive nature of laminin functions may not be a direct structural effect on LG conformation when expressed recombinantly, as previously believed, but an indirect consequence of changes in domain packing geometry. This hypothesis explains why we could detect integrin-dependent activities with monomeric recombinant LG modules, while multimeric LG modules failed to work (22,26). Our results are further supported by recent findings of Hirosaki et al. (18), who showed that deletion of the LG3 module from recombinant Ln-5 leads to an almost complete loss of cell adhesion and migration activities of recombinant Ln-5, strongly hinting that LG3 is an important domain of Ln-5. However, the underlying basis for the activity loss (i.e. functional and/or structural) was not clarified.
Ln-5 is an epithelial specific adhesion molecule critical for the establishment and maintenance of epithelial integrity via the formation of anchoring devices, the hemidesmosomes (20,59,60). A potential clinical application of Ln-5 is in keratinocyte sheet grafts to expedite re-epithelialization and wound healing (61,62). However, Ln-5 is a large (ϳ460 kDa) and mutimeric protein, and therefore a cost-effective, high yield supply of Ln-5 may not be easily obtained. In this respect, rLG3 may be an attractive alternative. Although the rLG3 we expressed here displayed reduced activity compared with Ln-5, future studies to optimize LG3 activity, by structure-based protein engineering for example, may yield high activity LG3 variants that may serve as a feasible substitute for Ln-5 in wound healing research and treatment.
In summary, we demonstrated that the LG3 module of Ln-5 ␣ 3 chain replicates key Ln-5 activities in an integrin ␣ 3 ␤ 1 -dependent manner. This is the first report providing direct evidence that a laminin LG module, expressed in a recombinant fashion, can bind specifically to an integrin and replicate key laminin functions. Furthermore, because our finding with Ln-5 may apply to other laminin ␣ chains, we suggest that the LG3 module of other laminin isoforms should be tested for activity as well.