Fibulin-2 binds to the short arms of laminin-5 and laminin-1 via conserved amino acid sequences.

Epithelial cell-specific laminin-5, consisting of three chains, α3, β3, and γ2, is a component of the anchoring filament that traverses the lamina lucida beneath the hemidesmosomes of epidermal cells and functions to link these cells to the basement membrane. We have studied the molecular interaction between laminin-5 and extracellular matrix proteins using recombinant proteins and synthetic peptides. Affinity chromatography assays with recombinant fragments of the laminin γ2 short arm identified a 195-kDa binding protein in the conditioned media from the mouse epidermal cell line Pam 212 and from primary dermal fibroblasts. This molecule was identified by Western blotting as fibulin-2, a recently identified extracellular matrix protein. Using deletion mutants and various synthetic peptides in competition assays, the 9-amino acid sequence SADFSVHKI (residues 199-207) in domain IV of the γ2 chain was defined as a critical site for fibulin-2 binding. An anti-γ2 antibody co-immunoprecipitated fibulin-2 from the conditioned media, further confirming the interaction of fibulin-2 with laminin-5. Fibulin-2 was also found to interact with laminin-1 (α1β1γ1) through a region (residues 654-665) of the α1 chain short arm whose sequence is similar to that of the fibulin-2 binding site of the γ2 chain. Together these results suggest that fibulin-2 functions to bridge laminin-1 and laminin-5 with other extracellular matrix proteins, providing a linkage between the cell surface and the basement membrane.

At the dermal-epidermal junction, there is stable attachment of epithelia to the underlying stroma through various proteinprotein interactions. Electron microscopy and immunohistochemical studies have defined the topographical linkage between hemidesmosomes on the basal surface of epithelium, anchoring filaments, and anchoring fibrils. These structures form an extended network, which surrounds the stromal fibers and inserts into the basement membrane. Hemidesmosomes of basal keratinocytes contain several molecules including BP180, BP230, HD1, and integrin ␣6␤4. The anchoring filaments contain laminin-5 (kalinin/nicein, epiligrin) and colocalize with hemidesmosomes at the supra basement membrane. The basement membrane components laminin-1, type IV collagen, and nidogen/entactin and the anchoring fibrils consisting of type VII collagen are located on the dermal side of the basement membrane (1)(2)(3).
Mutations in the genes for the components of the dermalepidermal junction in human patients with skin blister-form-ing disease have revealed the importance of these protein linkages in maintaining the structural stability of the dermalepidermal junction. Mutations in BP180 (4), integrin ␤4 (5), laminin-5 (6 -9), and type VII collagen (10 -12) have been identified. Acquired skin blister-forming diseases have also been shown to be due to autoantibodies to BP180 (13), laminin-5 (14), and type VII collagen (15). Basement membrane components have been shown to interact with each other and self-assemble to form a supramolecular network. Laminin-1 polymerizes through interactions at the N-terminal short arms of the monomeric molecules to form a hexagonal array of molecules (16). Nidogen is also a crucial molecule required for network formation since it binds several components of basement membrane including laminin-1, type IV collagen, and perlecan (17)(18)(19). Fibulin-1 (BM-90) and fibulin-2 were recently identified as a family of extracellular matrix proteins that interact with the laminin-1-nidogen complex, type IV collagen, and fibronectin (20 -23). The interaction of fibulins with multiple components of the extracellular matrix suggests that they function as mediators of supramolecular assembly at the basement membrane.
In order to study the interaction between laminin-5 and other extracellular matrix components, proteins in the culture medium of Pam 212 epidermal cells were screened for binding recombinant laminin chains by affinity chromatography. Fibulin-2, which is prominently expressed in skin and heart, was found to bind laminin-5 through the short arm of the ␥2 chain and we have identified a 9-amino acid sequence in domain IV of the ␥2 chain critical for this binding. We have also found that fibulin-2 binds to laminin-1 via the N terminus of the ␣1 chain, a site showing sequence homology to the 9-amino acid sequence of the ␥2 chain. Together these results suggest that fibulin-2 functions in assembling the laminin network in the basal lamina at the dermal-epidermal junction in bridging laminin-1 and laminin-5 with other matrix proteins.

MATERIALS AND METHODS
Cell Lines, Antibodies, and Reagents-The murine epidermal cell line Pam 212 (33) and the human epidermoid carcinoma cell line A431 were obtained from ATCC (Bethesda, MD). Primary cultures of mouse dermal fibroblasts were isolated from newborn FVB/N mouse. All cells were maintained with 10% fetal bovine serum/Dulbecco's modified Eagle's medium (Life Technologies Inc.). Anti-mouse fibulin-2 antiserum was made by immunizing rabbits with recombinant fibulin-2 (34) kindly provided by Dr. R. Timpl (Max-Planck-Institut fü r Biochemie, Munich, Germany). Anti-laminin ␥2 chain antiserum was prepared as * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. described previously (35). Anti-mouse type IV collagen antibody was made with Engelbreth-Holm-Swarm-derived type IV collagen in rabbits and purified using type IV collagen-coupled Sepharose column. Antihuman thrombospondin antibody was a generous gift from Dr. D. D. Roberts (NCI, NIH, Bethesda, MD) (36). Laminin-1 was prepared from the Engelbreth-Holm-Swarm tumor and purified as described previously (37). Bacterial collagenase form III was purchased from Advanced Biotechnologies Inc.
Synthetic peptides, ␥2-pC and ␥2-pN, were synthesized with a peptide synthesizer (Applied Biosystems, model 431A) by the t-butoxycarbonyl-based solid-phase strategy (40). All other synthetic peptides were manually synthesized by the Fmoc (9-fluorenylmethoxycarbonyl)-based solid-phase strategy and prepared as the C-terminal amide form as described previously (41). All synthetic peptides were purified by reverse phase high performance liquid chromatography. The purity and identity of the synthetic peptides were confirmed by analytical reverse phase high performance liquid chromatography and amino acid analysis. The location and amino acid sequences of the synthetic peptides from the ␥2 chain are listed in Table I and Fig. 3. Synthetic peptides listed in Fig. 7B include: pN-21, 9 amino acids (residues 199 -207) of the ␥2 chain; p␣1-654, 12 amino acids (residues 654 -665) of the ␣1 chain; p␣1-279, 9 amino acids (residues 279 -287) of the ␣1 chain; p␤1-461, 9 amino acid (residues 461-469) of the ␤1 chain with two arginine residues added to increase the solubility of the peptide; p␥1-587, 9 amino acids (residues 587-595) of the ␥1 chain with two added arginine residues to increase the solubility of the peptide; and p␣2-287, 9 amino acids (residues 287-295) of the ␣2 chain.
Affinity Chromatography and Competition Assays-The recombinant proteins fused to glutathione S-transferase (GST) 1 were bound to glutathione-agarose (Pharmacia) at 0.2 mg/ml. Laminin-1 prepared from Engelbreth-Holm-Swarm tumor was coupled to CNBr-activated Sepharose beads (1 mg/ml) (Pharmacia). Pam 212 cells and murine dermal fibroblasts were labeled with 50 Ci/ml [ 35 S]methionine (ICN, Costa Mesa, CA) in methionine-free Dulbecco's modified Eagle's medium (ICN) for 4 h. The conditioned media were adjusted to 2 mM phenylmethylsulfonyl fluoride and centrifuged at 3,000 revolutions/min for 10 min. These supernatants were precleared with 30% v/v Sepharose CL-4B (Pharmacia) by incubation for 30 min at 4°C, and these precleared supernatants were stored at Ϫ80°C until use. The conditioned media (500 ml) were incubated with 30 ml of affinity beads for 2 h with rotary shaking at 4°C. Following three washings with 1 ml of 0.1% Triton X-100, phosphate-buffered saline, 2 mM phenylmethylsulfonyl fluoride, the proteins bound to the affinity beads were extracted with SDS-sample buffer. The samples were boiled with or without 100 mM dithiothreitol, analyzed on 4 -12% SDS-PAGE, and treated with Enlightning (DuPont NEN) prior to autoradiograph. For competition assay, the synthetic peptides were added to the mixture during the incubation period.
Western Blotting and Immunoprecipitation Analysis-For Western blotting, the conditioned media were prepared from early confluent Pam 212 cells in 80-cm 2 flasks by incubation with 8 ml of Dulbecco's modified Eagle's medium supplemented with 2% fetal bovine serum for overnight and treated as mentioned above. The proteins bound to the affinity beads were electrophoresed through 4 -12% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane (Millipore, Marlborough, MA). The membrane was blocked with 5% milk, Tris-buffered saline, pH 7.4, 0.1% Tween 20 for 30 min and incubated with antibody either overnight at 4°C or for 2 h at room temperature. Bound antibody was detected by peroxidase-conjugated anti-rabbit IgG antibody (Pierce) and antibody reactivity followed by ECL (Amersham Life Science). The anti-fibulin-2 antiserum and anti-laminin ␥2 chain antiserum were diluted at 1:20,000 and at 1:5,000, respectively. For immu-noprecipitation, the antibody to the ␥2 chain was diluted at 1:100. Conditioned media (1 ml) were incubated with the antibody and protein G beads (Pharmacia) for 2 h at 4°C. After washing three times with 1 ml of 0.1% Triton X-100, phosphate-buffered saline, 2 mM phenylmethylsulfonyl fluoride, the bound proteins were eluted by slow agitation in washing buffer containing 5 mM EDTA for 20 min at room temperature. The eluted proteins were precipitated with 10% trichloroacetic acid for 30 min on ice with 10 mg of bovine serum albumin added as a carrier. After centrifugation at 30,000 ϫ g for 15 min at 4°C, the pellets were washed with 1 ml of 0.5% trichloroacetic acid once and then washed two times with 1 ml of acetone. These pellets were resuspended in sample buffer and separated by SDS-PAGE.
Sequence Analysis-Protein sequence analysis were performed using a software package from the University of Wisconsin Genetics Computer Group; optimal alignment was provided by the program BESTFIT.

Screening of Extracellular Proteins
Bound to the Short Arms of Laminin-5-Since the short arms of laminins have been shown to interact with other basement membrane proteins, we examined whether additional extracellular matrix proteins bind to the short arms of the ␥2 and ␤3 chains of laminin-5 using recombinant laminin chains. [ 35 S]Methionine-labeled conditioned media from murine epidermal Pam 212 cells were incubated with various recombinant laminin proteins-coupled to agarose beads. After washing, bound proteins were eluted and analyzed on SDS-PAGE. The N terminus of the ␥2 chain short arm bound a protein with an apparent molecular mass of 195 kDa, while there was no protein binding to either the short arm of the ␤3 chain of laminin-5 or the ␣2 chain of laminin-2 ( Fig. 1A, lanes 1-3). Binding of a 195-kDa molecule to the ␥2 chain short arm was also observed with the conditioned media from mouse dermal fibroblasts (Fig. 1A, lane 5). Since 5 mM EDTA abolished the binding, this interaction was likely dependent on a divalent cation (Fig. 1A, lane 4). Electrophoresis containing 6 M urea showed a single protein band, strongly suggesting that this 195-kDa band consists of a single molecule (Fig. 1B). Furthermore, the shift of the molecular size from 195 kDa to ϳ600 kDa under non-reducing conditions suggested that the 195-kDa molecule might form a disulfide-bonded homotrimer (Fig. 1B). Digestion of this 195-kDa protein with collagenase did not cleave this protein, suggesting it does not contain a collagenous domain (Fig. 1C, lanes 3 and 4).
Fibulin-2 Binding to the Recombinant ␥2 Chain Short Arm-Judging from its molecular size and its ability to form a homotrimer, we speculated that this non-collagenous 195-kDa extracellular protein might be either fibulin-2 (21) or thrombospondin (36). To test these possibilities, we used antibodies to both fibulin-2 and thrombospondin in Western blotting. The 195-kDa protein bound to the recombinant ␥2 chain short arm (␥2-r1) was recognized by an antibody against mouse fibulin-2 ( Fig. 2A, lane 2), but not by an antibody to thrombospondin (Fig. 2B). Furthermore, this protein that reacted with the antifibulin-2 antibody formed a trimer under non-reduced conditions ( Fig. 2A, lane 3) and exhibited divalent cation-dependent binding to ␥2-r1 ( Fig. 2A, lanes 4 and 5). These data confirm the identity of the 195-kDa protein as fibulin-2.
These results indicate that the active region for fibulin-2 bind-ing is located within residues 197-247 in domain IV of the ␥2 chain.
Delineation of the Fibulin-2 Binding Sequence of the ␥2 Chain-To delineate the fibulin-2 binding sequence of the ␥2 chain, synthetic peptides derived from the ␥2 chain were tested for their ability to compete for binding to fibulin-2 ( Table I).

FIG. 3. Deletion analysis for the active site of fibulin-2 binding.
A, a schematic illustration of the laminin ␥2 chain. Domains V and III contain several epidermal growth factor-like repeats (boxes), and domain IV is a globular region (oval) of the short arm region. Domains II/I indicate ␣-helical-rich long arm region forming a triple-stranded coiled-coil structure with other laminin-5 chains. A set of deletion mutant recombinant proteins, ␥2-r1 to ␥2-r5, and synthetic peptides, ␥2-pN and ␥2-pC, are shown with the amino acid residue numbers of the murine laminin ␥2 chain. B, affinity chromatography assays were performed with ␥2-r1-GST to ␥2-r5-GST agarose beads using [ 35 S]methionine-labeled Pam 212 conditioned media and followed by fluorography.
suggesting that either Ile-207 or Ser-199 was required for activity. Furthermore, none of three scrambled peptides containing the pN-4 residues were active (Fig. 4C, lanes 16 -18), indicating that the activity depends on the specific sequence of amino acids and not on the amino acid composition.
Analysis with the truncated peptides described above demonstrated that the 9-amino acid sequence, SADFSVHKI (residues 199 -207), of the ␥2 chain was necessary for fibulin-2 binding. We introduced single amino acid substitutions (Fig.  4D, left panel) in this 9-amino acid peptide (pN-21) to identify residues important for the activity. pN-21A1 and pN-21A5 completely abolished the activity, indicating that Phe-202 and Lys-206 were critical for the activity (Fig. 4D, lanes 3 and 8). The Val-204 and Ile-207 were also important but not as important as Phe-202 and Lys-206, because pN-21A3 and pN-21A6 were more active than pN-21A1 and pN-21A5 (Fig. 4D, lanes 5  and 7). The significant loss of activity in pN-21A6 is consistent with a loss observed with pN-6 containing the deletion of Ile-207 of pN-5 (Fig. 4B). The fact that pN-21A2 and pN-21A4 peptides were active in binding suggests that Ser-203 and His-205 were not essential for binding (Fig. 4D, left panel). The change of Phe-202 to Leu (pN-21L1) completely abolished binding (Fig. 4D, right panel, lanes 2 and 3). Although a leucine substitution at Val-204 (pN-21L2) did not decrease its activity, a threonine substitution at Val-204 (pN-21T) abolished its activity (Fig. 4D, right panel, lanes 4 and 6). Furthermore, a glycine substitution at Asp-201 (pN-21G) did not reduce the activity, indicating that this residue was not essential for activity (Fig. 4D, right panel, lane 5). Taken together, these results demonstrate that the nonapeptide 199 -207 was the minimum active region with residues Ser, Ala, Phe, Val, Lys, and Ile, critical for the activity. The importance of Ala-200 was not confirmed by a deletion peptide since the peptide was not soluble after deletion of this residue from pN-13 (Table I).
Fibulin-2 Binds to Native Laminin-5-Binding of fibulin-2 to native laminin-5 was examined by immunoprecipitation assays. The conditioned media from Pam 212 cells were immunoprecipitated with the antibody to the ␥2 chain. The precipitates were analyzed by Western blotting with the anti-fibulin-2 antibody. The anti-␥2 chain antibody immunoprecipitated fibu-lin-2 (Fig. 5A). This co-immunoprecipitation of fibulin-2 was inhibited by peptide pN-4 (Fig. 5B, upper panel). pN-4 did not affect the amount of the ␥2 chain immunoprecipitated by the anti-␥2 antibody (Fig. 5B, bottom panel). These results suggest that fibulin-2 binds to native laminin-5 via the pN-4 site in domain IV of the ␥2 chain.

Peptide
Sequence Activity  Fig. 3A and Table I. The peptides were added to the mixture of ␥2-r1-GST and Pam 212 conditioned media to test whether they competed for fibulin-2 binding to ␥2-r1-GST. Bound substrates were analyzed on 4 -12% SDS-PAGE and detected by either fluorography (A) or Western blotting using anti-fibulin-2 antiserum (B-D). A, peptides depicted above each lane were added at 40 g/ml (lanes 2 and 4) or 200 g/ml (lanes 3, 5, and 7-9). B, peptides depicted above each lane were added at 200 g/ml. C, peptides depicted above each lane were added at 200 g/ml (lanes 1-9 and 15-18). In lanes 10 -14, the synthetic peptides were increased to 1 mg/ml. Molecular markers indicate 199 and 120 kDa. D, synthetic peptides shown in the boxes with alanine substitutions (left panel) or other single amino acid point mutations (right panel) were added at 200 g/ml.

␥2-pN
lin-2 binds to laminin-1 by affinity chromatography. The conditioned media from Pam 212 cell culture were applied on a native laminin-1-coupled Sepharose affinity column. Western blotting of the eluant from this column showed the presence of fibulin-2, indicating that fibulin-2 interacts with laminin-1 (Fig. 6A). Inclusion of 5 mM EDTA to the conditioned media completely eliminated the binding of fibulin-2 to laminin-1, suggesting that this interaction is cation-dependent. Addition of both Ca 2ϩ (8 mM) and Mn 2ϩ (5 mM) to the 5 mM EDTAcontaining conditioned media restored binding of the fibulin-2 to laminin-1 (Fig. 6B, upper panel). Mg 2ϩ alone also restored the binding activity, although to a lesser extent. Similar divalent cation dependence was also seen for the binding of fibulin-2 to the recombinant laminin ␥2 chain (Fig. 6B, bottom  panel). These results indicate that the fibulin-2 binding to laminin-1 and laminin-5 requires divalent cations.
The binding of fibulin-2 to laminin-5 appears to be a relatively strong interaction, since the anti-␥2 antibody co-immunoprecipitates the complex in the conditioned media. The finding that a synthetic peptide from the ␥2 chain could inhibit this complex formation suggests that the fibulin-2 binding site of the ␥2 chain is not cryptic and is active in the native laminin-5 molecule. A heptapeptide sequence within laminin ␥1 has been delineated for nidogen binding (42)(43)(44). As the nonapeptides from the ␥2 and ␣1 chains inhibit fibulin-2 binding, it is likely that fibulin-2 interacts with a small region in the laminins of similar size to that identified for nidogen. Since the 10 epidermal growth factor-like repeats of fibulin-2 have a calciumbinding motif (21), this region likely contains a site(s) for laminin binding. Consistent with these data is our finding that fibulin-2 binds to both laminin-1 and -5 and this binding is abolished by the addition of EDTA. Recombinant fibulin-2 has been shown to bind strongly to fibronectin in calcium-dependent manner by solid phase radioligand binding assays (23). It also binds to nidogen, although this interaction is only blocked partially by EDTA. However, little binding of fibulin-2 to laminin-1 was found in the solid phase binding assay system. This discrepancy of fibulin-2 binding to laminin-1 may be due to differences in the two assays.
Amino acid truncation and substitution analysis to delineate the region of the ␥2 chain responsible for binding to fibulin-2 suggested that residues Ser-199, Phe-202, Val-204, Lys-206, and Ile-207 within the nonapeptide sequence of the ␥2 chain (pN-21, residues 199 -207) were required. Although p␤1-461 from the ␤1 chain contains similar residues including Phe, Val, and Ile at the positions similar to pN-21, it was inactive in inhibiting fibulin-2 binding to laminin-1. The inactive peptide p␤1-461 also contains Leu at the position corresponding to Lys-206 in pN-21. This is also consistent with the result that Lys-206 was critical for the activity. The active peptide p␣1- 654 possesses Arg at the position of Lys-206 in pN-21, suggesting that a positively charged residue at this position is also involved in an ionic interaction with the binding site of fibulin-2. Moreover, p␣1-654 with Leu at the position of Ile-207 in pN-21 was active, and an Ala substitution at this position reduced the activity of pN-21, suggesting that a hydrophobic residue Leu or Ile at the position of Ile-207 in pN-21 was preferable for binding activity. Since an Ala-200 in pN-21 is not conserved in p␣1-654, this alanine residue does not seem to be essential for the activity. Together, it was concluded that consensus critical residues in laminin-1 and -5 for the fibulin-2 binding are F, V, (K/R), and (I/L). The laminin sequences for fibulin-2 binding defined in this report are not present in fibronectin and nidogen. Hence, fibulin-2 may interact with these molecules via different sites (23).
The biological importance of the short arm of the ␥2 chain of laminin-5 has been revealed by the finding of a ␥2 chain mutation in a human patient with junctional epidermolysis bullosa (7). This patient has an internal deletion of domains III and IV of the ␥2 chain short arm, suggesting that the short arm of the ␥2 chain is critical for the structural stability of the dermal-epidermal junction. Although the deleted region does not correspond exactly to the site for fibulin-2 binding, it is possible that the deletion perturbs the native conformation of domain IV, resulting in the masking or inactivation of the binding site. Since both molecular abnormalities in laminin-5 and autoantibodies specific to laminin-5 cause blister formation in skin, laminin-5 appears to be important for the integrity of skin. Since fibulin-2 binds to laminin-5, it is possible that it plays a critical role in stabilizing or organizing the epithelial basement membrane during development of skin or wound healing. The recent report that expression of fibulin-2 is markedly increased during skin repair supports this hypothesis (45). It will be of interest to examine whether the active peptide from the ␥2 or ␣1 chain can block formation of basal lamina in an in vitro reconstitution cell culture system (46). It is also interesting to examine whether blistering could be produced by subcutaneous-injection of the active peptide. Further studies will examine the significance of the fibulin-2 binding to laminins.