The Cellular Interactions of Laminin Fragments CELL ADHESION CORRELATES WITH TWO FRAGMENT-SPECIFIC HIGH AFFINITY BINDING SITES*

The molecular interactions of laminin with several tumor cell lines and skin fibroblasts were investigated by radioligand binding studies and cell attachment as- says using laminin, the laminin-nidogen complex, and laminin fragments as substrates and also domain-spe- cific antibodies as inhibitors of cell attachment. The majority of cells showed a dual binding pattern for fragments 1 and 8 which originate from short-arm or long-arm structures of laminin, respectively. Both of these fragments in solution bind to suspended cells with high affinity (KO = 1-10 nM), with the receptor num- bers for each fragment depending on the cell type. Competition studies and independent variation of re- ceptor numbers demonstrated that the cell-binding structures on each fragment are different, implicating the existence of two distinct cellular receptors for laminin. The ability of these fragments to act as substrates for cell adhesion correlated with the presence of high affinity binding sites on the cells. However, only antibodies to fragment 8 were able to block cell adhesion to laminin, despite the presence of binding sites for fragment 1. A few cells had very low numbers of high affinity receptors for either fragment 1 or 8. The latter cell type was used to demonstrate that complex for- mation between laminin and nidogen, which binds to fragment 1 structures, reduces the potential of laminin for cell binding.

cell type was used to demonstrate that complex formation between laminin and nidogen, which binds to fragment 1 structures, reduces the potential of laminin for cell binding.
The ability of the basement membrane protein laminin to promote the adhesion, spreading, proliferation, and differentiation of cells is related to its potential for cell binding (reviewed in Refs. 1 and 2). Laminin-binding proteins have been isolated from the plasma membranes of various tumor cells and muscle (3)(4)(5), cloned (6), and shown to have molecular mass of about 68 kDa. Such putative laminin receptors are thought to mediate the various biological activities expressed by cells when in contact with laminin. Other studies, however, have shown binding between laminin and sulfatides (7), gangliosides (8), a fibronectin-binding protein (9), and heparan sulfate (lo), which may be alternative mediators of laminin-cell interactions.
In attempts to determine which domains of the laminin molecule are responsible for its cellular interactions, several proteolytic fragments of the protein have been shown to act as substrates for cell attachment (see Fig. 1). They include the disulfide-rich fragment 1, which consists of three rod-like segments originating from the center of the short arms of laminin (11,12). This fragment was found to compete with the laminin-mediated cell attachment to collagen IV substrates (13,14), to promote directly cell adhesion (15), and to bind to cells with an affinity comparable to that of laminin (16). Fragment 1 also decreased the metastatic potential of tumor cells (16,17) presumably due to blockade of the laminin receptors thought to be required for invasion through basement membranes. In addition, fragment 1 displayed metabolic effects similar to those of intact laminin in that it enhanced the proliferative response of adherent cells in culture (18) and stimulated the synthesis of a metalloproteinase which cleaves specifically basement membrane collagen (19).
Another cell-binding fragment of laminin, fragment 8, was originally identified in studies with neural cells. This fragment stimulated neurite outgrowth and potentiated the survival of cultured neurons (20) and was also implicated in the specific increase in levels of catecholamine synthetic enzymes seen when chromaffin cells are cultured on laminin substrates (21). Fragment 8 consists of a rod-like segment and a complex globular domain, representing the terminal half of the long arm of laminin (22,23). It also contains the heparin-binding domain of laminin (fragment 3), although it could not be demonstrated that this domain was directly involved in the interaction of laminin with neurons (20). Other, small globular laminin fragments (fragments 5 and 6) were found to promote adhesion of hepatocytes (15) but have not yet been mapped within the laminin structure.
Although antisera against these fragments could block cell attachment to the fragments themselves, no inhibition of cell binding to laminin was seen, indicating that fragments 5 and 6 do not correspond to the major cell-binding sites of the intact laminin molecule and may have arisen de nouo due to the proteolysis.
A comparative analysis of the cell-binding potentials of the major laminin fragments 1 and 8 has so far not been undertaken. This study has become feasible after the development of a new method to isolate the laminin-nidogen/entactin complex (23) that allows the preparation of biologically active fragment 8 in a reproducible fashion. The data presented here demonstrate that many cell types can adhere to fragment 1 and/or fragment 8. The ability of these fragments to act as substrates for cell attachment correlates with the presence of two distinct laminin receptors, demonstrated by the high affinity binding of radiolabeled ligands to cells in suspension.

MATERIALS AND METHODS
Preparation of Laminin, Laminin Fragments, and the Laminin-Nidogen Complex-Laminin was purified from the mouse Engelbreth-Holm-Swarm tumor as previously described (24). Extraction and purification of the laminin-nidogen complex followed a more recent 11532 H 20 nm procedure (23). Dissociation of the complex for the preparation of individual components was achieved with 2 M guanidine HCl. The intact complex was, however, used for the purification of laminin fragment E8 (23). Fragments P1, El-4, E3, and E4 and 25-kDa fragment were prepared from conventionally purified laminin following previously published procedures (1 1,15,22,25). ~dioimmunoassays and electrophoresis demonstrated a high purity of fragments (contamination <1-2%).
Antisera and Immunoadsorption-Rabbit antisera were raised against laminin (24) and the laminin fragments PI, El-4, E3, E4 (25), 25-kDa fragment andE8 (22) as previously described. Antibodies were affinity-purified from anti-laminin antisera by successive passages over immunoadsorbents prepared from iaminin fragments P1, E4, El-4, and E3 and from laminin. Preparation of the agarose conjugates and elution of column-bound antibodies with 3 M KSCN followed standard protocols (26). Purified antibodies were concentrated by ultrafiltration to 0.3-1.2 mg/ml, and their specificity was evaluated by radioimmunoassays. Immunological Assays-Antigens were labeled with '"I by the chloramine-T method (specific radioactivity, 5,000-20~000 cpm/ng) and were used for binding assays with antisera or purified antibodies following established procedures (26). Ant~gen-binding capacity (ABC-33) was determined from these binding data according to Minden and Farr (27). Specificity of antibodies and purity of laminin fragments were also determined by radioimmunoinhibition assays. have been characterized previously (28). Human embryonic skin fibroblasts were those used in previous studies (29). Cells were grown to confluence in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100-400 units/ml penicillin, 50-90 pg/ml streptomycin, 50 pg/ml ascorbate, and 300 pg/ml glutamine.
Coating of Cuiture Dishes-Tissue culture plastic dishes (24 multiwell plates with 2-cm2 surface area/well, from Costar, Cambridge, MA) were coated by air-drying protein solutions overnight at 37 "C (29). Protein substrates were diluted in 1% BSA dissolved in distilled water prior to coating. In a second procedure, wells were coated by adsorbing proteins from solution in distilled water overnight at 4 "C, followed by blocking with 1% BSA for an additional 24-h adsorption The abbreviations used are: CHO, Chinese hamster ovary; BSA, bovine serum albumin; KRH, Krebs-Ringer-Henseleit.
period. Wells coated with BSA alone served as negative controls.
Amounts of protein substrates present in the wells were determined by adding trace amounts of '"I-labeled proteins to the unlabeled substrate solution before coating. The substrates were washed with Dulbecco's modified Eagle's medium, solubilized with 2 M NaOH, and counted. The actual amounts of coat obtained by both adsorption procedures was in the range 7-10% of total protein added for laminin and 3 4 % for fragments 1 and 8 in the concentration range used.
Celt Attachment and Inhibi~wn Assays-Cell monolayers were detached by exposure to 0:05% trypsin, 0.02% EDTA in phosphatebuffered saline, pelleted by low speed centrifugation, and suspended in plain Dulbecco's modified Eagle's medium (4-8 X 1 0 5 cells/ml). Aliquots of the cell suspension were plated onto the coated dishes at a density of 5 X IO4 cells/cm2 and incubated at 37 "C in a humidified incubator. Maximal attachment occurred after 20-60 min depending on the cell line. At the end of this period, medium was removed, and adherent cells were washed once with phosphate-buffered saline and detached with trypsin/EDTA (see above). Their number was then determined in a Coulter Counter. Background values of cells attached to BSA-coated wells (10% of added cells) were subtracted from the number of cells attached in the presence of various other proteins.
For inhibition assays, coated dishes were incubated with various dilutions (1:10-1:6~) of antisera or purified antibodies (250 pl/well) for 60 min at 37 "C. The attachment assay was then done in the presence or after removal of the antibodies.
Radioligand Binding Assay for Cells-Laminin and the laminin fragments E8 and P1 were iodinated by the lactoperoxidase method (30) to a specific activity of 3-10 mCi/pg. The solution containing the iodinated protein was supplemented with BSA (1 mg/ml) to reduce background adsorption and separated from unreacted iodine by gel filtration on Bio-Gel P-30. The labeled material was stored no longer than 4 days; and immediately before use, it was freed from aggregated material by centrifugation (6 min, 20,000 x g), with the concentrations corrected for losses due to aggregation.
Growing cells (50-70% confluence) were harvested after incubation with 2 ml of calcium-and magnesium-free Krebs-Ringer-Henseleit (KRH) buffer with 1 mM EDTA (5 min, 37 "C) and gentle trituration. The resulting cell suspension was then added to 5 ml of KRH containing 0.1 mg/ml BSA (KRHfBSA) supplemented with 5.5 units/ ml DNase I (550 units/mg, Sigma) to .degrade any DNA from disrupted cells, which interferes with the binding assay. The cells were then centrifuged (5 min, 800 X g), resuspended in 5 ml of KRH/BSA, and recentrifuged through 8 ml of KRH containing 5% BSA (5 min, 100 X g) to remove cell debris, which remained in the upper phase. The pellet of intact cells was resuspended in KRH/BSA, and the cell concentration was determined with a hemocytometer. Cell viability ranged between 85 and 95% as determined by trypan blue exclusion.
The binding assays were performed essentially according to Herrup and Thoenen (31). Assays were performed with 300-p1 suspensions of cells (lo5 cells/tube) to which 1261-laminin or its fragments (ranging from 0.1 to 40 nM) was added with or without 100-fold excess unlabeled ligand. Incubation continued for 2 h at 0 'C without shaking, which allows maximal specific and nonspecific ligand binding to the cells.' To separate free from unbound labeled laminin, the cells were centrifuged through a two-step sucrose gradient in a Beckman Microfuge tube. The bottom layer consisted of 75 pl of 0.3 M sucrose in KRH/BSA. Onto this was overlayered 100 p l of 0.15 M sucrose in KRH/BSA. Finally, 100 pl of the sample to be assayed was layered on top. The tubes were spun for 3 min in a Beckman Microfuge (Model 152,10,000 rpm) and frozen in dry ice. The bottom 2 mm of the tube, containing the pelleted cells, was then cut off, and the "' 1 was counted. Total and unspecific binding were determined in triplicates at each ligand concentration. Competition assays were performed by adding increasing amounts of nonlabeled components to the labeled ligand prior to incubation with the cells.

RESULTS
Comparison of Cell Adhesion to Substrates of Laminin, Laminin-Nidogen Complex, and Laminin Fragments-The tumor cell lines and embryonic skin fibroblasts that were used to examine dose-response profiles of attachment on air-dried substrates had been previously shown to adhere to laminin substrates (28, 29). Many cell lines (including astrocytoma 251 MG, rhabdomyosarcoma RD, CHO, and fibroblasts, not shown) behaved similarly to the fibrosarcoma H T 1080, displaying maximal attachment on 2-5 pg of laminin added per well (Fig. 2). Plateau values for attachment varied between 25 and 75% of total seeded cells, depending on the cell line.
The rat schwannoma cell line RN 22 reached higher plateau values of 80-90% attachment when only 1-2 pg of laminin had been used to coat the wells, whereas the rhabdomyosarcoma cell line A 204 attached relatively poorly and did not reach plateau values even after using 20 pg of laminin/well in the same incubation period of 30 min (Fig. 2). Increasing the time of incubation did not change the plating efficiencies (not shown).
The cells were also compared in their ability to adhere to the laminin-nidogen complex. The complex showed generally higher activity than laminin at low concentrations, but there was no effect on the final plateau values of attachment obtained. The only exception was cell line A 204, which did not attach to the laminin-nidogen complex (Fig. 2). Components of the complex were separated from each other after dissociation in 2 M guanidine HCl (23) and were then examined in adhesion assays. Laminin obtained from the complex showed activity, whereas no adhesion-promoting activity could be detected with nidogen dissociated from the complex (data not shown).
Since previous studies with tumor cells (13,15,32) and neurons (20) have implicated fragments 1 and 8 of laminin (Fig. I), respectively, as cell binding structures, we compared cefl adhesion to both of these fragments (Fig. 2) laminin and its fragments 1 and 8, showing that different coating procedures do not influence cellular recognition of laminin structures. Antibody Inhibition of Cell Attachment-Eight different rabbit anti-laminin antisera of comparable radioimmunoassay titers were compared in attachment inhibition at a fixed dilution (1:25). Seven antisera showed significant inhibition of cell attachment which varied between 37-100% inhibition on an optimal laminin coat, depending on the particular antiserum and cell line used. Antibodies of anti-laminin antisera were then fractionated by immunoadsorption on various laminin fragments in order to study the domain specificity of their inhibitory effect (Table I). Immunoadsorption of the antiserum on fragments 1, 4, and 1-4, which are all structures of the short arms of laminin (25) (see Fig. l), yielded antibodies which failed to inhibit HT 1080 cell attachment to laminin. Subsequent purification of the remaining antibodies by immunoadsorption on fragment 3 and laminin resulted in antibodies which reacted strongly with the long-arm fragments 3 and/or 8 of laminin (22) but failed to bind significantly shortarm fragments ( Table I). Both of these antibody fractions inhibited attachment of HT 1080 cells on laminin (Table I).
Similar results were obtained when the inhibition of adhesion of a variety of cells (HT 1080, 251 MG, RD, CHO, and RN 22) was examined using rabbit antisera raised directly against fragments 1, 1-4, 3, 4, and 8. These antisera showed radioimmunoassay binding for the fragment used for immunization comparable to that for laminin but failed to react significantly with unrelated fragments (antigen-binding capacities usually 100-fold lower, not shown). Again, only antisera against fragments 3 and 8 showed strong inhibition of cell attachment on laminin. Antisera against 25-kDa fragment or fragments 1, 1-4, or 4 were either inactive or of lower potency (Table 11). The low activity of the latter antisera indicates the presence of some antibodies recognizing epitupes near the active site of fragment P1. Similar antibodies are apparently missing in antisera against laminin (see Table I). The same set of antisera was, in addition, used to examine attachment of HT 1080 cells on the laminin-nidogen complex (Table 11). Here again, only antisera to fragment 8 were inhibitory, whereas other fragment-specific antisera and antisera to nidogen were inactive.
Adhesion of HT 1080 cells to fragment 8 could be inhibited in an equivalent manner by antisera against fragments 8 and 3, whereas antisera against fragment 1-4 or 25-kDa fragment

39
Rabbit anti-laminin antiserum was successively passed over immunoadsorbents prepared from laminin fragments and eventually from laminin. Antibodies eluted from the respective column were then used for determining domain specificity by radioimmunoassay and for adhesion inhibition.
*HT 1080 cells adhering on a 5 -~g laminin coat. Coats were preincubated with antibody solutions diluted to antigen-binding capacity for laminin of 1 pg/ml (30 min) prior to the addition of cells. 44% of cells attached to the coat in the presence of normal rabbit IgG (0.1 mg/ml).   and El-4 (0). Antigen-binding capacities were determined by radioimmunoassays with fragment E8 (antisera against fragments E8 and E3, and 25-kDa fragment) or fragment El-4 (antiserum against fragment El-4) and were in the range 12-60 pg/ml antiserum. Numbers of attached cells in the negative controls were similar to that in Table 11. were either inactive or of low inhibitory capacity (Fig. 3). No inhibition of cell adhesion on fragment 1 was seen with any of the antibodies used in this study; in particular, three different antisera raised against either fragment 1-4 or 1 directly were essentially noninhibitory (maximal 10% inhibition) even at the highest concentration used, which was 1 pg/ ml antigen-binding capacity (Table 11). Also, three different antisera against laminin which blocked H T 1080 cell adhesion on a laminin substrate failed to inhibit attachment of these cells on a fragment 1 substrate.

Cell-Binding of R~i~~b e~d
Laminin and Laminin Fragments-The assays were carried out with cells in suspension using an assay method which has been useful in cell-binding studies with nerve growth factor (31,33). A critical parameter of the assay was the use of freshly labeled ligands (within 4 days of labeling) which reduced background binding. Preliminary experiments showed maximal specific and nonspecific binding of radiolabeled laminin and its fragments after a 2-h incubation with the cells at 0 "C. This time point was therefore taken for the construction of equilibrium binding curves (Fig, 4). Such curves showed a sigmoidal shape extending above the inflexion point after plotting bound against free ligand in a semilogarithmic fashion (not shown). These data therefore appear suitable for analyzing affinity and receptor numbers involved in the cell binding of sohble ligands (341. Apparent dissociation constants (KO) and number of cellbinding sites for laminin and laminin fragments 1 and 8 were evaluated from Scatchard plots for which examples are shown in Fig. 4 (Table 111). Specificity of fragment binding by HT 1080 cells was shown by competition assays (Fig. 5). Both laminin and fragment 1 competed in equivalent manner, with the binding of labeled fragment 1 resulting in 50% displacement at equimolar ratios.
No displacement was seen with fragment 8 even at 125-fold molar excess. Conversely, laminin and fragment 8, but not fragment 1, were able to displace labeled fragment 8 from the cells. Thus, the cell-binding sites of the fragments are different in their ability to bind cells independently, consistent with the presence of two distinct cellular receptors.

DISCUSSION
This study shows that a variety of cells can adhere to fragment 1 or 8, or both, of laminin, reflecting the presence of distinct receptors for these fragments. The presence of such receptors was demonstrated directly by the high affinity binding of radiolabeled ligands and the specific competition of this binding. These observations imply that the ce~l-binding sites of laminin on fragments 1 and 8 are structurally different.
The reco~ition by cells of one or both topolo~cally separated domains of laminin seems to be a general phenomenon; cell interactions of fragment 1 have been reported previously for several metastatic tumor cell lines (13,14,(16)(17)(18)32) and hepatocytes (15), and cell interactions with fragment 8 have been reported with neurons (20) and chromaffin cells (21). Our present data show that fibroblasts and several cell lines (HT 1080, 251 MG, RD, and CHO) can adhere to both fragments 1 and 8 of laminin, showing that recognition of the latter is not a property restricted to neural cells. The dual recognition of laminin by cells does not appear, however, to be a uniform phenomenon; schwannoma RN 22 cells adhered more readily to fragment 8 substrate than to fragment 1, and similar observations have been made for myoblasts and several tumor cells lines? Conversely, the A 204 cell line behaved differently, adhering to laminin and fragment 1 substrates but not to fragment 8. Neurons may even differ slightly from this scheme since they respond to fragments 8 and 1-4 (20) but not to the fragment 1 substructure of fragment 1-4.* Whether these differences are due to a third cell-binding S. Goodman, R. Deutzmann, and K. von der Mark, personal D. Edgar, unpublished results.

various cells for lam~nin and laminin f r~~e n t s 1 and 8
Values were determined by radioligand binding. Data were analyzed by Scatchard plots (see Fig. 4).

Ce,ls
Binding Radioligand sequence only found on fragment 1-4 is not clear at present. The laminin-nidogen complex where nidogen is bound to fragment 1 structures (23) shows a comparable or even better adhesion activity than laminin alone (Fig. 2). We attribute the increased adhesive activity of the complex to the gentle isolation procedure used which may keep fragment 8 structures more intact, rather than to a modulatory effect of nidogen in the complex. Nidogen, however, can block fragment 1 cell-binding sites as shown for A 204 cells which did not adhere to the laminin-nidogen complex, presumably because they lack alternative receptors for fragment 8 structures. The contribution of fragment 1 to the adhesion of other cells could not be further evaluated by antibody inhibition experiments; neither antibodies raised against fragment 1 directly nor anti-fragment 1 antibodies purified from blocking antisera raised against laminin could inhibit adhesion to fragment 1 or laminin. The reason why antibodies to fragment 8 inhibit the adhesion of cells to laminin substrates even in varying amounts together with labeled ligands (5 nM) to HT 1080 cells. Competition range is defined between total binding and background binding (see Fig. 4) and corresponded to a difference of 6,000 cpm for fragment E8 and 12,700 cpm for fragment P1. 10 nM with 104-105 receptor sites/cell and are in good agreement with the previous observations. In addition, however, it is shown here that fragment 8 can also bind specifically to cells with an affinity similar to that of laminin and with comparable numbers of receptor sites. The ability of laminin fragments to bind to cells in suspension closely paralleled their ability to promote cell adhesion; the binding and adhesion to fragment 1 were low for RN 22 cells, and fragment 8 neither bound to A 204 cells nor stimulated their adhesion. There was also an inverse correlation between number of receptor sites for laminin or its fragments and the amount of laminin substrate required for optimal cell attachment (see Fig. 2 and Table 111). This suggests that the rate of cell adhesion to matrix molecules is determined by the number of receptors expressed. Whether individual cells have differing numbers of receptors for each fragment, as suggested by Fig.  2, remains to be determined. Several limitations should be considered in the interpretation of our binding data. The study was designed for the identification of high affinity binding sites; and any weaker interactions with laminin may therefore have escaped detection, although that is not to say that they are physioIogically irrelevant. The accuracy of receptor number determinations may also be influenced by endogenous ligands which may have been incompletely removed during detachment of cells with EDTA. Oligomers of laminin which persist in the presence of EDTA (36) could also contribute to errors in the determination of KO and receptor number. Very few and small oligomers are present in preparations of fragments 1 and 8 as shown by ultracentrifugation (11,22). The error introduced by laminin aggregation therefore does not appear to be very substantial since apparent KO and receptor numbers determined with laminin closely correspond to values determined for the laminin fragment binding most extensively to the cells.
The identification of two major cell adhesion sites on the laminin molecule that could be up to 80-100 nm distant from one another and different in structure predicts the existence of two distinct laminin receptors. This was clearly demonstrated by the lack of competition between fragments 1 and 8 in cell binding and because the numbers of each receptor vary independently between various cell types. It is not yet clear if the cells that possess the two receptors utilize them both for adhesion to laminin substrates. Different cell recognition sites have been identified on fibronectin (37), and it appears that at least two sites are necessary for the full interaction of fibroblasts with fibronectin substrates (38). A similar multiplicity of interactions might be required for cell adhesion to laminin.
Complexity of laminin receptors has also been indicated in previous studies which identified several cell-surface components as potential laminin-binding sites. A 68-kDa lamininbinding protein has been isolated from several sources (3-5).
Cells possessing this receptor were demonstrated to react with laminin fragment 1 in competition assays (14) and showed similar Krl values for laminin and fragment 1 in radioligand binding (16). This receptor has also been shown to be eluted from affinity columns by a synthetic peptide corresponding to a structure present in fragment 1 (39). It appears likely therefore that this 68-kDa protein corresponds to the fragment 1 receptors identified in this study.
A laminin receptor with similarly high affinity should be expected to exist for fragment 8, but the nature ofthe molecule responsible for binding is unclear. Attempts to identify this receptor by affinity chromatography and ligand binding to blots have yielded equivocal results in our hands. Nevertheless, an avian fibronectin cell receptor (CSAT) has been claimed to be a bifunctional cell receptor in that antibodies to CSAT are also able to block cell adhesion and neurite outgrowth on laminin and also because isolated CSAT binds laminin (9,40). The low affinity of this binding tends, however, to exclude CSAT as a binding protein for either fragment 1 or 8 structures of laminin. A more likely candidate for the second laminin receptor might be monogalactosyl sulfatides which, when purified from erythrocytes and brain (7), showed a high affinity for laminin (half-saturation of binding at 5 nM). Furthermore, these sulfatides do not bind laminin fragment 1. Cell membrane-bound forms of heparan sulfate proteoglycan (reviewed in Ref. 41) could also be involved in laminin binding since the major heparin-binding site of laminin has been localized to fragment 3, which is a globular domain of fragment 8 (22, 25). Whereas such proteoglycans might well be involved in cell interactions with fragment 8 (38), it is unlikely that these interactions alone are responsible for the high affinity since fragment 3 itself was found to be a poor adhesion substrate (see also Ref. 20). The ceilular receptor for the long arm of laminin therefore remains to be identified.