Laminin-1 and Laminin-2 G-domain Synthetic Peptides Bind Syndecan-1 and Are Involved in Acinar Formation of a Human Submandibular Gland Cell Line*

The culture of human submandibular gland (HSG) cells on laminin-1 induces acinar differentiation. We identified a site on laminin involved in acinar differentiation using synthetic peptides derived from the C-terminal G-domain of the laminin α1 and α2 chains. The α1 chain peptide AG73 (RKRLQVQLSIRT) decreases the size of acini formed on laminin-1. Cells cultured with either AG73 or the homologous α2 chain peptide MG73 (KNRLTIELEVRT) form structures that appear acinar-like, but the cell nuclei are not polarized to the basal surface and no lumen formation occurs, indicating that additional sites on laminin are required for complete differentiation. The G-domain of laminin-1 contains both integrin and heparin binding sites, and anti-β1-integrin antibodies disrupt acinar formation. Cell adhesion to the peptides and to E3, an elastase digest fragment of laminin-1 containing AG73, is specific, since other laminin peptides or EDTA do not compete the binding. Heparin and heparan sulfate decrease cell adhesion to AG73 and MG73 but anti-β1-integrin antibodies have no effect. Treating the cell surface with heparitinase inhibits adhesion to both AG73 and MG73. We isolated cell surface ligands using both peptide affinity chromatography and laminin-1 affinity chromatography. Treating the material bound to the affinity columns with heparitinase and chondroitinase enriches for a core protein identified as syndecan-1 by Western blot analysis, thus identifying a syndecan-1 binding site in the globular domain of laminin-1 and laminin-2. In summary, multiple interactions between laminin and HSG cells contribute to acinar differentiation, involving both β1-integrins and syndecan-1.


From the Craniofacial Developmental Biology and Regeneration Branch, NIDR, National Institutes of Health, Bethesda, Maryland 20892-4370
The culture of human submandibular gland (HSG) cells on laminin-1 induces acinar differentiation. We identified a site on laminin involved in acinar differentiation using synthetic peptides derived from the C-terminal G-domain of the laminin ␣1 and ␣2 chains. The ␣1 chain peptide AG73 (RKRLQVQLSIRT) decreases the size of acini formed on laminin-1. Cells cultured with either AG73 or the homologous ␣2 chain peptide MG73 (KNRLTIELEVRT) form structures that appear acinarlike, but the cell nuclei are not polarized to the basal surface and no lumen formation occurs, indicating that additional sites on laminin are required for complete differentiation. The G-domain of laminin-1 contains both integrin and heparin binding sites, and anti-␤ 1integrin antibodies disrupt acinar formation. Cell adhesion to the peptides and to E3, an elastase digest fragment of laminin-1 containing AG73, is specific, since other laminin peptides or EDTA do not compete the binding. Heparin and heparan sulfate decrease cell adhesion to AG73 and MG73 but anti-␤ 1 -integrin antibodies have no effect. Treating the cell surface with heparitinase inhibits adhesion to both AG73 and MG73. We isolated cell surface ligands using both peptide affinity chromatography and laminin-1 affinity chromatography. Treating the material bound to the affinity columns with heparitinase and chondroitinase enriches for a core protein identified as syndecan-1 by Western blot analysis, thus identifying a syndecan-1 binding site in the globular domain of laminin-1 and laminin-2. In summary, multiple interactions between laminin and HSG cells contribute to acinar differentiation, involving both ␤ 1 -integrins and syndecan-1.
Laminins, a family of heterotrimeric glycoproteins found in basement membranes, promote cell adhesion, migration, differentiation, proliferation, neurite outgrowth, and tumor growth (1,2). Such diverse biological functions suggest multiple cellular interactions with the laminin molecule. Current reports identify 11 isoforms of laminin displaying tissue-specific expression patterns at different times during development (3,4). Cell type-specific interactions between different laminins and multiple receptors and signaling pathways provide mech-anisms for regulating the complex events of morphogenesis and the maintenance of differentiated adult tissues.
Integrins are well characterized laminin receptors (5,6). Different cell types can use different integrins to attach to laminin. Studies with recombinant and elastase fragments of laminin-1 indicate interactions of ␣6␤1, ␣3␤1, ␣7␤1 and ␣6␤4 within the C-terminal G-domain of the ␣1 and ␣2 chains (7). Laminin-1 also contains multiple heparin binding sites including one in the E3 fragment, although the exact epitope is undefined (2,8). The E3 fragment is important in early stages of branching morphogenesis of the salivary gland, since antibodies to E3 inhibit branching of salivary gland buds in culture (9).
Syndecans, a family of transmembrane heparan sulfate proteoglycans, bind a wide range of components through their heparan sulfate side chains. All cells and tissues except B stem cells express syndecans in cell-specific, tissue-specific, and developmentally specific patterns (10,11). Syndecans-1 and -4 mediate cell-cell adhesion, which occurs by interaction with a heparan sulfate-binding ligand present on an adjacent cell surface (12). Antisense experiments show that maintenance of an epithelial phenotype requires syndecan-1 (13). Syndecan-1 expressed by transfected NIH-3T3 cells binds to laminin-1. The E3 fragment from the G-domain of laminin-1 revealed the highest binding for syndecan-1 compared with the other elastase fragments (14). Syndecans have been proposed to function as co-receptors for components in the extracellular microenvironment, the syndecan utilizing a heparan sulfate-protein interaction and the other receptor a classical protein-protein interaction (10).
We have identified biologically active sites on the laminin molecule using a synthetic peptide approach. YIGSR on the ␤1 chain promotes cell adhesion and migration and inhibits angiogenesis and tumor metastasis (15). IKVAV, on the ␣1 chain, promotes cell adhesion, neurite outgrowth, experimental metastasis, collagenase IV secretion, angiogenesis, and tumor growth (16,17). Systematic screening of the G-domain of the laminin ␣1 chain has identified five active peptides with cell adhesion and spreading activities (18,19). Testing these peptides for activity with neuronal cell lines identified peptide sequences from corresponding regions of the ␣1 and ␣2 laminin chains with cell type specificity for neurite outgrowth (20). These two active sequences, AG73 (RKRLQVQLSIRT) from laminin-1 and MG73 (KNRLTIELEVRT) from laminin-2, are in the E3 fragment of the G-domain and are highly conserved between murine laminin ␣ chains (Table I). Conservation of these sequences among different species suggests a critical biological role (18,19). AG73 promotes malignant behavior of melanoma cells in vitro and tumor growth and metastases in vivo (21). AG73 also inhibits branching morphogenesis of cultured embryonic mouse submandibular glands (22).
We have shown that laminin-1 promotes acinar differentiation of the human submandibular gland (HSG) 1 cell line, and other components such as transforming growth factor-␤3 also contribute to acinar formation (23). Acinar differentiation is a multistep process involving multiple cell-matrix interactions, but here we focus on the interactions between HSG cells and laminin. Integrins clearly play an important role, since function-perturbing anti-␤ 1 -integrin antibodies disrupt acinar formation. We have used synthetic peptides from the G-domain of laminin ␣1 and ␣2 chains to identify sites involved in acinar formation. We have identified the heparan sulfate proteoglycan syndecan-1 as a cell surface ligand for one of the peptides and for laminin-1. Our results suggest multiple interactions between laminin-1 and HSG cells occur during acinar formation involving both ␤ 1 -integrins and syndecan-1.
Acinar Formation Assay-Laminin-1 and growth factor-reduced Matrigel were prepared as described previously (25)(26)(27). Either Matrigel (135 g in 50 l of media) or laminin-1 (155 g in 50 l of media) was added to 96-well tissue culture dishes. HSG cells (1.0 ϫ 10 4 cells in 100 l of media) were added and incubated for 48 h and stained with Diff-Quick (American Scientific Products, McGaw Park, IL). An image analysis program (NIH Image) was used to measure the number and area of acini formed in the matrix. We assayed three fields/well in triplicate wells (70 -150 acini total) and determined the percentage of acini with a surface area either less than 800 m 2 or greater than 800 m 2 .
Preparation of Peptides-All peptides were manually synthesized using the Fmoc (9-fluorenylmethoxycarbonyl)-based solid phase strategy with a C-terminal amide form and purified by reverse-phase high performance liquid chromatography as described previously (18).
Inhibition of Acinar Formation and Culture of Cells with Peptides-Peptides (100 g/ml) and cells (1.0 ϫ 10 4 cells) were added to wells containing laminin-1 (155 g in 50 l of serum-free media), and after 48 h the cells were stained and the sizes of acini were measured. HSG cells were also cultured with the laminin-derived peptides (100 g/ml) in 96-well dishes for 48 h. The peptides were incubated in the wells with 50 l of media for 2 h before the cells (in a volume of 150 l) were added; unbound peptide was not removed from the wells. Cell proliferation was measured using a nonradioactive cell proliferation assay (Cell Titer 96; Promega, Madison, WI), according to the manufacturer's instructions.
Cell Adhesion Assays-A 50-l volume containing either laminin-1 (0.1 g/well), AG73 (0.05 g/well), MG73 (2 g/well), or E3 (1.5 g/well) was dried overnight onto round-bottomed 96-well plates. The wells were blocked with 3% BSA for 1 h at 37°C and then washed twice with 0.1% BSA. 25,000 cells in 100 l of 0.1% BSA in Dulbecco's modified Eagle's medium/F-12 were added per well for 45 min at 37°C. The medium was gently removed from the wells, and the cells were stained with crystal violet for 10 min and washed twice with water. The cells were lysed with 50 l of 10% SDS, and the optical density (600 nm) was measured.
Inhibition of Cell Adhesion-Antibodies and cells were rotated for 15 min at 4°C and then added to the adhesion assay. Function perturbing antibodies to the ␣ 6 , (GoH3), ␤ 1 (M13), ␣ 5 (M16), ␣ 2 (P1E6), and ␣ 3 (P1B5) integrin subunits were used. Nonimmune rat IgG and mouse ascities (Sigma) were used as negative controls. Antibodies were used in a range of dilutions (1:20, 1:33, and 1:100). Cells were also preincubated for 10 min at room temperature with peptides (1 g/well), EDTA (2 mM), or GAGs (50 g/ml) and then added to the adhesion assay. Heparin; de-N-sulfated heparin; heparan sulfate; chondroitin sulfates A, B, and C; keratan sulfate; and hyaluronic acid (Sigma) were all tested. In other experiments, heparin was incubated with the peptide-coated wells for 30 min, the wells were washed twice, and the cells were added. Conversely, heparin was incubated with the cells for 30 min, and the cells were washed twice and then added to the peptide-coated wells.
Treatment of the Cell Surface with GAG-degrading Enzymes-Cells were cultured for 2 h with 2 g/ml cycloheximide to inhibit new protein synthesis (23), washed, and resuspended in 100 l of medium consisting of Dulbecco's modified Eagle's medium/F-12, 0.1% BSA, 2.0 g/ml cycloheximide, 2.0 mM aminoethylbenzesulfonyl fluoride, 1.5 mM aprotinin, and either 0.05 units/ml heparitinase or 0.02 units/ml heparinase (Seikagaku, Rockville, MD). Cells were incubated at 37°C for 90 min, with mixing every 15 min. After enzyme treatment, the cells were used in a cell adhesion assay as described above. Cells were also treated with inactivated enzymes (100°C for 10 min and the addition of 1 mM ZnCl 2 ) as negative controls.
Peptide and Laminin-1 Affinity Chromatography-Affinity columns (1 ml) were prepared using Affi-Gel 10 (Bio-Rad) according to the manufacturer's instructions. Laminin-1, AG73, MG73 affinity columns, and negative control columns were run in parallel. An AG73T column was used as a peptide negative control, a BSA column was used as a protein negative control column, and an Affi-Gel control column was used to detect nonspecific binding to the resin. The columns were equilibrated in running buffer containing 6.0 M urea, 1% Triton X-100, 2.0 mM phenylmethylsulfonyl fluoride in TBS, pH 7.4. Cells were surface-biotinylated using sulfo-NHS-biotin (Pierce, IL) as described in the manufacturer's instructions. A crude cell membrane fraction was prepared by hypoosmotic lysis in 10 mM KCl, 20 mM Tris, pH 7.4, 0.1% ␤-mercaptoethanol, 1 mM EDTA. After Dounce homogenization, the nuclei were removed by centrifugation (1500 ϫ g for 5 min). The NaCl concentration of the remaining supernatant was increased to 150 mM, and the cell membranes were pelleted at 50,000 ϫ g for 30 min. The cell membrane pellet was solubilized in 2 ml of 8.0 M urea, 1% Triton X-100, 0.5 M KCl, 2.0 mM phenylmethylsulfonyl fluoride in TBS, pH 7.4, and insoluble material was removed by centrifugation at 14,000 ϫ g for 20 min. The volume was increased to 10 ml with running buffer. A 500-l aliquot of the crude cell membrane fraction (ϳ700 g of total protein) was incubated with the peptide affinity column for 2-4 h at 4°C. The columns were washed with running buffer and then sequentially eluted with 2-ml aliquots of running buffer containing either 20 mM EDTA, 250 mM NaCl, 1.0 M NaCl, or 2.0 M NaCl. Bound material was precipitated with acetone, washed in 80% ethanol, and air-dried. Samples were separated by SDS-polyacrylamide gel electrophoresis (4 -20% gels) and transferred to nitrocellulose filters (Novex, San Diego, CA). The filters were blocked in 3% nonfat milk in PBS-T (Tween 20, 0.1%), washed, incubated with streptavidin-horseradish peroxidase in PBS-T for 1 h and then washed again three times for 10 min in PBS-T. The biotinylated material was visualized by ECL (Amersham Pharmacia Biotech). In separate experiments to try to inhibit syndecan binding to laminin-1, the crude cell membrane fraction was preincubated with 1 mg/ml AG73 or AG73T for 1 h before incubating with laminin-1 affinity columns. The columns were then eluted, and the fractions were analyzed as described above.

Laminin Peptides Decrease the Size of Acini Formed on
Laminin-1-We examined the effects of various laminin-derived synthetic peptides on HSG cells cultured in an acinar formation assay on laminin-1 (Table II). The assay is a competitive inhibition assay where peptides compete for cell binding with intact laminin-1. The addition of AG73 resulted in a significant reduction in the size of acini formed (Fig. 1a). AG73 had the greatest effect, whereas AG73T, a scrambled version of AG73, had no effect (Fig. 1b) and was used as a negative control. Other active peptides from the G-domain of the laminin ␣1 chain and their ␣2 homologues (19) as well as other previously identified sequences such as IKVAV, YIGSR, RGD, and SINNNR (28) had no effect on the size of acini formed except for AG10, which had a small but consistent effect.
Laminin Peptides Promote Morphological Organization of HSG Cells-Most laminin peptides tested had no effect on the morphological organization of HSG cells, and the cells grew in a monolayer (Fig. 2a). Surprisingly, cells incubated with AG73 ( Fig. 2b) and MG73 (Fig. 2c) formed multicellular three-dimensional structures. Hematoxylin-and eosin-stained sections of the structures formed with the peptides (Fig. 2, d and e) revealed they were not well organized. The cell nuclei were not  polarized to the basal surface of the structures and lumens did not form. In contrast, cells incubated with laminin-1 formed acinar structures with basally polarized nuclei, lumens, and cystatin immunoreactivity (23). When the peptides were added to collagen gels, acinar formation was not observed (data not shown). The major problems of culturing cells with peptides were that cell adhesion, spreading, and formation of multicellular structures all occur in the same well. Also, high concentrations of peptides were required to see morphological effects. We therefore used suspension cultures to study the effect of the peptides in solution utilizing a rotary cell culture system. Acinar formation in the bioreactor occurred within 24 h in the presence of laminin-1 (Fig. 3, a and e). The cells formed both spherical and multilobulated structures (Fig. 3a) as observed by light microscopy. In cross-sections, stained with hematoxylin and eosin (Fig. 3e), the cell nuclei were clearly located at the basal surface of the structures and lumen formation occurred. Similar to our previous results (23), the formation of acinar structures was a cell density-dependent event occurring within 24 -48 h. When cells were cultured for 24 h in a bioreactor either with AG73T (Fig. 3, b and f) or without peptide (not shown), the cells remained mainly as a single cell suspension. In contrast, both AG73 and MG73 stimulated HSG cells to form large multicellular structures with a smooth lobulated surface, appearing similar by light microscopy to those formed with laminin-1 (Fig. 3, c and d). However, in hematoxylin and eosinstained cross-sections of the structures that formed with AG73 and MG73 (Fig. 3, g and h), it is apparent that they do not organize morphologically as well as those that formed with laminin-1 (Fig. 3e). The structures that formed with the peptides do not have cell nuclei polarized to the basal surface, and there was no lumen formation. These results suggested that the peptides may be providing some but not all of the signals for the morphological changes in the process of acinar differentiation. It is likely that multiple sites on laminin, including inte- grin and nonintegrin interactions and/or conformational requirements of the intact laminin molecule are required for complete differentiation. Since integrins are well characterized laminin receptors, we wanted to define their role in HSG cell acinar formation. We also wanted to determine whether the mechanism of action of the peptides involved integrin or nonintegrin interactions.
␤ 1 -Integrins Are Involved in Acinar Formation and Adhesion to Laminin but Not in Adhesion to the Synthetic Peptides-Immunofluorescent analysis of integrins on the cell surfaces of acini showed differential distribution (Fig. 4a). Anti-␤ 1 and ␣ v stained the cell matrix surfaces and the cell-cell contacts, whereas anti-␣ 5 only stained the cell-cell contacts. Anti-␣ 6 and anti-␣ 2 stained mainly the cell matrix surface, although ␣ 6 also weakly stained the cell-cell contacts and ␣ 2 had mainly an intracellular localization. When function-blocking antibodies were tested with HSG cells on laminin-1, only the anti-␤ 1integrin antibody inhibited acinar formation. Paraffin sections of these cells stained with hematoxylin and eosin showed a lack of organization into acinar structures and no polarization of cell nuclei (Fig. 4b).
We studied the effect of anti-integrin antibodies on cell adhesion to peptides and to laminin-1 using a solid phase adhesion assay (Fig. 4c). Dose-response curves and time courses of adhesion determined the amount of each substrate and the time of the assay. We coated the wells with the least amount of substrate that gave a maximal cell adhesion response. The cells attached to the peptides more quickly than to laminin-1; i.e. 50% of total adhesion to AG73 occurred in 10 min, whereas 50% of total adhesion occurred to laminin-1 in 30 min (data not shown). None of the integrin antibodies used inhibited cell adhesion to AG73 or MG73. The anti-␤ 1 -integrin antibody inhibited adhesion of HSG cells to laminin-1. The ␣-subunit antibodies alone (or in combinations, data not shown) were unable to inhibit adhesion to either laminin-1 or the peptides. Taken together, these data suggest that ␤ 1 -integrins were important for laminin-1-mediated acinar formation but did not bind to AG73 and MG73.
HSG cells interact with multiple sites on the laminin-1 molecule, and adhesion to a peptide sequence may be affected by the conformation of the surrounding protein. We compared the ability of AG73 to inhibit cell adhesion to laminin-1, AG73, MG73, and the E3 fragment, which contains the AG73 sequence. AG73 slightly inhibited HSG cell adhesion to laminin-1 and inhibited adhesion to E3, AG73, and MG73 (Fig. 5). Therefore, AG73 may be an important adhesion site in the E3 fragment. AG73T did not inhibit cell adhesion to any substrate.
The addition of EDTA inhibited cell adhesion to laminin-1 and partially reduced cell adhesion to E3, AG73, and MG73 (Fig. 5). Taken together with the anti-␤ 1 -integrin results (Fig.  4c), these data suggest that HSG cells attach to laminin-1 by multiple interactions, including integrin, divalent cation-independent, and heparin-mediated interactions. The data suggest that cell adhesion to AG73 and to E3 is by a similar mechanism.
Adhesion of HSG Cells to Laminin Peptides and E3 Is Inhibited by Heparin-Since the E3 fragment contains a heparin binding region, we tested the effects of GAGs on cell adhesion to the peptides, E3, and laminin-1. None of the GAGs tested (all at 50 g/ml) inhibited cell adhesion to laminin-1 (Fig. 6a). Heparin inhibited cell adhesion to E3, and chondroitin sulfate C had a slight inhibitory effect, suggesting that cell adhesion to E3 was mediated by both GAGs. Heparin inhibited cell adhesion to both AG73 and MG73. Heparan sulfate inhibited cell FIG. 4. Localization and role of integrins in acinar formation. a, immunostaining showed differential localization of integrin subunits within acini. Acini were stained with antibodies against integrin subunits ␤ 1 (M13), ␣ 5 (M16), ␣ 2 (P1E6), ␣ v (VNR147), and ␣ 6 (GoH3). Immunofluorescence was detected with 1-m optical sections using a confocal laser-scanning microscope. b, anti-␤ 1 -integrin antibodies (M13) disrupted acinar formation. Antibody or nonimmune rat IgG was incubated with cells for 48 h, and the cells were sectioned and stained with hematoxylin and eosin. c, integrin antibodies do not inhibit cell adhesion to the peptides. Antibodies and cells were preincubated for 15 min at 4°C, added to the wells coated with either laminin-1 or the peptides, and incubated for 30 min at 37°C. The cells were stained and lysed, and the A 560 was measured. Nonimmune rat IgG and mouse ascities were used as negative controls. *, p Ͻ 0.005. adhesion to AG73 but only partially to MG73 (Fig. 6, c and d). At higher doses (100 g/ml), heparan sulfate inhibited adhesion to both MG73 and AG73 (data not shown). De-N-sulfated heparin; chondroitin sulfates A, B, and C; keratan sulfate; and hyaluronic acid had no effect on cell adhesion to the peptides. These data suggest that the adhesion to the peptides may be mediated by regions of the GAG chains with different sulfation patterns and that N-sulfation may be involved.
Cell adhesion to the peptides still occurred when cells were preincubated with heparin and washed before being used in an adhesion assay. However, cell adhesion did not occur when heparin was preincubated in the peptide-coated well and then washed out of the well before the addition of the cells (Table  III). Thus, heparin was not binding to the cell surface but was binding to the peptide and blocking cell adhesion to the peptide.
We added heparin (0.1 mg/ml) to the bioreactors in combination with laminin-1 and the peptides. Heparin inhibited the morphological organization of the cells cultured with AG73 and MG73 (data not shown); they appeared as a single cell suspension, similar to the cells cultured with the scrambled peptide (Fig. 3b). The addition of both heparin and laminin-1 in the bioreactor had no effect on the formation of acinar-like structures (similar to Fig. 3a). These results are consistent with the acinar formation assay, where heparin did not affect the size of acini formed on laminin-1 (data not shown), and with the finding that heparin did not inhibit cell adhesion to laminin-1 (Fig. 6a). As a control, we added heparin to cells in the bioreactor without peptides or laminin and also found no effects on the morphology of the cells.
Treatment of the Cell Surface with Glycosaminoglycan-degrading Enzymes-Heparitinase treatment slightly decreased cell adhesion to laminin-1 and to AG10 (p Ͻ 0.05) when compared with cells treated with inactive enzyme (Fig. 7). This suggested that cell adhesion to intact laminin may be in part mediated by heparan sulfate, although we could not significantly inhibit adhesion to laminin-1 with heparan sulfate in the previous adhesion assay. AG10, another active peptide from the G-domain, was included as a positive control for cell adhesion to a peptide. Cell adhesion to AG10 is mediated by ␣ 6 -integrin (18). The decrease in cell adhesion to AG10 with GAG removal suggests that GAGs may affect ligand binding properties of the integrin, as recently reported with the ␣ 4 ␤ 1integrin (29). Heparitinase treatment inhibited cell adhesion to AG73 and to MG73 compared with cells treated with inactive enzyme. However, heparinase treatment of the cell surface inhibited the adhesion to MG73 but did not significantly decrease the adhesion to AG73. Heparinase digests more highly sulfated or "heparin-like" regions of the GAG chain, whereas heparitinase digests less sulfated or "heparan sulfate-like" regions of the GAG chain. Our earlier finding showed that heparan sulfate decreased cell adhesion to AG73 more than to MG73 (Fig. 6, b and c). Taken together, these results suggested that AG73 and MG73 bind to regions of the heparan sulfate chains with different sulfation patterns i.e. MG73 binds to a more sulfated (or heparin-like) region of the heparan sulfate side chain than AG73. Also, these results suggested that the cell surface receptor for AG73 could be a heparan sulfate proteoglycan.
Laminin-1 and Peptide Affinity Chromatography and Western Blot Analysis with HSE-1, an Antiserum Recognizing Syndecan-1-The major species bound to the AG73 and MG73 peptide columns eluted with 1.0 M NaCl and appeared as a high molecular weight smear, with a molecular mass greater than 250 kDa ( Fig. 8a and b, lanes 5). Columns prepared with a scrambled peptide, AG73T (Fig. 8c), with BSA (not shown), or with uncoupled Affi-Gel resin (not shown) allowed identification of nonspecific interactions and gave similar results. As expected, laminin-1 affinity chromatography (Fig. 8d) identified multiple species binding to the column including a high molecular weight smear with a similar appearance to that eluted from the peptide columns. We were unable to inhibit adhesion of the high molecular weight smear to the laminin-1 column by preincubating the cell lysate with AG73 (data not shown). We interpret this result as meaning there are other syndecan binding sites on the laminin molecule. It is well documented that other heparin binding sites exist on laminin. This is also supported by our cell adhesion experiments to the intact laminin-1 molecule (Fig. 5), where we were unable to completely inhibit cell adhesion to laminin-1 with AG73. Another species (ϳ150 kDa) also eluted from the laminin-1 column and corresponded to the expected size of dystroglycan, another proteoglycan known to bind E3 (31). However, in a Western blot (data not shown) with an anti-␣-dystroglycan antibody (IIH6 C4, a kind gift of K. Campbell), the band did not stain. Our attempts to isolate an AG73 receptor using octyl glucoside or Triton X-100 alone were unsuccessful at identifying specific binding molecules. Syndecans are insoluble in these lysis buffers and precipitate with the cytoskeleton (30).
Treatment of the material bound to the peptide affinity columns with heparitinase and chondroitinase ABC resulted in a shift in molecular weight of the smear and enriched for a core species at ϳ67 kDa (Fig. 9). Enzyme treatment of the material eluted from the laminin-1 column also showed a similar sized core protein, although multiple other species were also present (data not shown). Western blot analysis with HSE-1, a polyclonal antibody to the core protein of human syndecan-1, revealed an antibody-reactive band that co-migrated with the major band that appeared after heparitinase and chondroitinase ABC treatment (Fig. 9). The other minor species that appeared after GAG removal have not yet been identified. These data demonstrate that syndecan-1 binds to the AG73 sequence in the G-domain of laminin-1 and to the homologous laminin-2 peptide, MG73.

DISCUSSION
Our previous studies identified a role for laminin-1 in the morphological differentiation of human salivary gland cells cultured on Matrigel, a basement membrane substrate (23,32). Others demonstrated that laminin-1 played an important role in salivary gland morphogenesis (9,33,34). The E3 fragment from the G-domain of the laminin ␣1 chain was involved, since salivary gland explants treated with an anti-E3 antibody were unable to form branches (9). We identified a number of biologically active peptides in the G-domain of laminin-1 (18,19,35). The AG73 peptide, from the E3 fragment of the G-domain, decreased the size of acini that formed on laminin-1, suggesting that this site of interaction was important for acinar formation. Furthermore, of the active peptides identified, only AG73 and the corresponding homologous peptide MG73 from the ␣2 laminin chain, stimulated multicellular morphological organization of HSG cells in suspension cultures. The peptides may bind to the cell surface and promote cell-cell adhesion and/or stimulate the cells to begin morphological organization. However, in hematoxylin and eosin-stained sections of the peptide-induced structures, the cells were not as well organized as those cultured with laminin-1. The cell nuclei were not polarized to the basal surface of the cells, and lumens within the structures were not apparent, suggesting that other sites on laminin-1 were necessary for complete differentiation.
AG73 and MG73 have been tested in both in vitro and in vivo assays with a number of cell types. They played a role in melanoma cell adhesion and metastases (21) and promoted neuronal cell line adhesion and neurite outgrowth (20). Other peptides from the G-domain of laminin-1 were active with neuronal cells in a cell type-specific manner distinct from that observed with HSG cells. A high degree of homology exists between AG73, MG73, and the corresponding regions in the ␣3a, ␣3b, ␣4, and ␣5 laminin chains (Table I). These conserved sequences may have an important role in cell surface receptor interactions. Taken together, these studies in a number of biological systems and the highly conserved sequence suggest an important role for AG73 and for MG73 in mediating some of the biological functions of laminin. AG73 is located in E3, a fragment of laminin-1 that mediates cell binding and has heparin binding activity. AG73 inhibited cell adhesion to itself and to E3 but not to laminin-1. Interestingly, the addition of heparin (0.1 mg/ml) inhibited the morphological effects of the peptides in the bioreactor, and the cells remained as a single cell suspension. However, heparin did not inhibit the morphological effects of the entire laminin-1 molecule. Clearly, this site is not the only active site on laminin for HSG cells, but it does influence cell-cell interactions and morphological organization.
Our studies begin to define the role of certain cell-matrix interactions in acinar differentiation. The inhibition of acinar formation on laminin-1 with ␤ 1 -integrin antibodies suggests an important role for ␤ 1 -integrins in the multistep process of acinar differentiation. Furthermore, integrin subunits showed differential localization during acinar formation. Inhibition of HSG cell adhesion to laminin-1 with the ␤ 1 -integrin antibody and with EDTA indicated that integrins were important HSG cell laminin receptors. However, neither integrin antibodies nor EDTA inhibited cell adhesion to the active peptides. Rather, heparin and heparan sulfate inhibited cell adhesion to the peptides and to E3. Furthermore, treatment of the cell surface with heparitinase reduced cell adhesion to the peptides but not to laminin-1. These data indicate that the cell surface ligand for the peptide contains heparan sulfate. AG73 peptide affinity chromatography of cell membrane extracts identified a high molecular weight species comparable in size with the syndecans. Heparitinase and chondroitinase treatment of this material resulted in a major core protein (ϳ64 kDa), which was identified by subsequent Western blot analysis as syndecan-1. We were unable to identity the other minor species at ϳ36 kDa (Fig. 9, lane 3). Our data suggest that syndecan-1 is the major AG73 and MG73 binding ligand on HSG cells, although other heparan sulfate-containing ligands may exist. Laminin-1 affinity chromatography also showed a similar high molecular weight species bound to intact laminin-1. We attempted with no success to address the issue of binding specificity by eluting syndecan from the laminin-1 and peptide columns with peptides. These types of experiments have been used to elute To determine the mechanism of GAG-mediated inhibition, heparin (50 g/ml) was incubated with the peptide-coated wells for 30 min, the wells were washed twice, and then the cells were added. The converse experiments were also done where heparin was incubated with the cells for 30 min and the cells were washed twice and then added to the peptide-coated wells. The cell adhesion assay is described under "Materials and Methods."  7. Effect of heparitinase or heparinase treatment of the cell surface on cell adhesion to AG73 and MG73. Heparitinase treatment decreased cell adhesion to AG73 and MG73 by 75% but decreased adhesion to laminin-1 by only 20%. Heparinase treatment inhibited cell adhesion to MG73 but not to AG73. Enzyme treatments are described under "Materials and Methods." Cells were also treated with inactivated enzymes as negative controls, and the percentage of cell adhesion and p values were calculated by comparing adhesion to cells treated with inactive enzymes. Triplicate wells were used for each condition, and the graph is representative of at least three similar experiments. S.E. values are indicated. *, p Ͻ 0.05; **, p Ͻ 0.001. integrins from RGD columns with RGD-containing peptides. The interaction between integrins and RGD is relatively weak, and integrins are eluted with EDTA or 1 mM peptide. With AG73 columns, however, the interaction of syndecans with the peptide column is mediated by heparan sulfate and is relatively strong, as evidenced by the need to elute material with more than 250 mM NaCl.
Syndecans, by way of their GAG side chains, have been proposed to function as co-receptors with integrins for fibronectin (10). Laminin has multiple heparin binding sites, which are found at the ends of the molecule and co-localize with integrin binding sites. It has been suggested that integrin binding and cell surface proteoglycan binding may be involved in laminin-mediated cell adhesion and cell signaling (7). Our data are consistent with this hypothesis, since both integrin-and syndecan-mediated adhesion are involved in acinar formation. Recently, a direct interaction between cell surface chondroitin sulfate GAG and ␣ 4 ␤ 1 -integrin was shown to affect the ligand binding properties of the integrin (29). Whether a direct interaction occurs between syndecan-1 GAG and an integrin subunit on the HSG cell surface remains to be determined.
Interaction between laminin and syndecans has been previously shown. NIH-3T3 cells transfected with syndecan-1 produced a laminin binding form of syndecan-1 (14). The E3 fragment of laminin-1 had the highest binding to syndecan-1 among the elastase-derived fragments of laminin. Furthermore, overexpression of syndecan-1 resulted in an increase in cell adhesion to laminin. Analysis of two cell lines with identical amounts of syndecan-1 revealed that the syndecan-1 from one cell line does not bind collagen, whereas the other does (36). Analysis of the side chains showed that the fine structure of heparan sulfate differed and that these differences could control fundamental cell properties such as cell matrix adhesion. Studies of the syndecan-1 side chain variability between mammary gland cells and cell lines revealed differences in the number of highly sulfated domains in the GAG side chains (13). These data support the concept that variability of syndecan-1 GAG side chains may affect the interaction between different laminin isoforms and syndecan-1.
Our data show differences between the interaction of HSG cells with AG73 and MG73. Although both peptides bound syndecan-1, they had different effects on HSG cells. Our in vitro data are supported by collaborative studies with Kadoya's group (22) that showed AG73 inhibited branching morphogenesis of day 13 embryonic salivary gland explants but MG73 had no effect. In our in vitro acinar formation assay, AG73 decreased the size of the acini formed, but MG73 had no effect. Furthermore, both heparin and heparan sulfate inhibited HSG cell adhesion to AG73, but heparan sulfate decreased cell adhesion to MG73 by ϳ50%. Heparitinase treatment of HSG  6). Bound material was blotted and detected with streptavidin-horseradish peroxidase/ECL. AG73 (a) and MG73 (b) affinity columns gave similar results; the major species bound appeared as a high molecular weight smear Ͼ 250 kDa that was eluted with 1.0 or 2.0 M NaCl (arrow). AG73T (c) and BSA (not shown) affinity columns were similar to each other. Laminin-1 affinity chromatography (d) revealed multiple components bound the column including a high molecular weight smear (arrow), which was also eluted in the 1.0 NaCl eluate. cells, which digests less sulfated regions of heparan sulfate, inhibited cell adhesion to AG73. However, heparinase treatment, which digests more sulfated, "heparin-like" domains of heparan sulfate, did not inhibit cell adhesion to AG73. Taken together, these data suggest that MG73 might bind to a more highly sulfated region of heparan sulfate than AG73. We have not analyzed the syndecan-1 heparan sulfate side chains that bind AG73 as compared with those that bind MG73. The differential adhesion to heparan sulfate would allow cells to respond distinctly to the heparan sulfate binding effectors (in our case the laminin isoforms) in the cellular microenvironment (13).
During early stages of salivary gland morphogenesis (embryonic day 13), laminin ␣1 chain was found in the basement membrane surrounding the entire submandibular gland rudiment (22). By day 17, the staining pattern of ␣1 was still faintly detectable around the ducts, while ␣2 staining was found around the terminal tubules and developing acinar cells. In situ analysis of laminin ␣ chain expression in the developing mouse salivary glands showed barely detectable ␣1, ␣2, and ␣4 at day 15.5, compared with high levels of ␣3 and ␣5 (4). We have not yet studied the conserved ␣3 and ␣5 laminin chain homologues to AG73, but they may also be important in salivary gland development. These results suggest that different laminin ␣ chains may be involved in different morphological events during development.
In summary, here we have identified a syndecan-1 binding site in the G-domain of laminin. The laminin ␣1 chain peptide AG73 (RKRLQVQLSIRT) and its laminin ␣2 chain homologue MG73 (KNRLTIELEVRT) are biologically active in a variety of in vitro and in vivo systems. The identification of their cell surface ligand as the heparan sulfate side chains of syndecan-1 provides new insights into the mechanism of their activity. The interaction between laminin and heparan sulfate side chains of syndecans could provide another mechanism for laminin isoform involvement in specific biological processes during development.