Fibrillin-1 and fibulin-2 interact and are colocalized in some tissues.

Microfibrils 10-12 nm in diameter are found in elastic and non-elastic tissues with fibrillin as a major component. Little is known about the supramolecular structure of these microfibrils and the protein interactions it is based on. To identify protein binding ligands of fibrillin-1, we tested binding of recombinant fibrillin-1 peptides to different extracellular matrix proteins in solid phase assays. Among the proteins tested, only fibulin-2 showed significant binding to rF11, the N-terminal half of fibrillin-1, in a calcium-dependent manner. Surface plasmon resonance demonstrated high affinity binding with a Kd = 56 nM. With overlapping recombinant fibrillin-1 peptides, the binding site for fibulin-2 was narrowed down to the N terminus of fibrillin-1 (amino acid positions 45-450). Immunofluorescence in tissues demonstrated colocalization of fibrillin and fibulin-2 in skin, perichondrium, elastic intima of blood vessels, and kidney glomerulus. Fibulin-2 was not present in ocular ciliary zonules, tendon, and the connective tissue around kidney tubules and lung alveoli, which all contain fibrillin. Immunogold labeling of fibulin-2 on microfibrils in skin was found preferentially at the interface between microfibrils and the amorphous elastin core, suggesting that in vivo the interaction between fibrillin-1 and fibulin-2 is regulated by cellular expression and deposition as well as by protein-protein interactions.

Microfibrils 10 -12 nm in diameter are found in elastic and non-elastic tissues with fibrillin as a major component. Little is known about the supramolecular structure of these microfibrils and the protein interactions it is based on. To identify protein binding ligands of fibrillin-1, we tested binding of recombinant fibrillin-1 peptides to different extracellular matrix proteins in solid phase assays. Among the proteins tested, only fibulin-2 showed significant binding to rF11, the N-terminal half of fibrillin-1, in a calcium-dependent manner. Surface plasmon resonance demonstrated high affinity binding with a K d ‫؍‬ 56 nM. With overlapping recombinant fibrillin-1 peptides, the binding site for fibulin-2 was narrowed down to the N terminus of fibrillin-1 (amino acid positions 45-450). Immunofluorescence in tissues demonstrated colocalization of fibrillin and fibulin-2 in skin, perichondrium, elastic intima of blood vessels, and kidney glomerulus. Fibulin-2 was not present in ocular ciliary zonules, tendon, and the connective tissue around kidney tubules and lung alveoli, which all contain fibrillin. Immunogold labeling of fibulin-2 on microfibrils in skin was found preferentially at the interface between microfibrils and the amorphous elastin core, suggesting that in vivo the interaction between fibrillin-1 and fibulin-2 is regulated by cellular expression and deposition as well as by protein-protein interactions.
Microfibrils have been identified as small diameter (8 -12 nm) extracellular matrix fibrils found in close proximity to basement membranes (Low, 1962) and elastic fibers (Cleary and Gibson, 1983). Microfibrils can be distinguished from banded collagen fibers since microfibrils display no clear banding pattern, have relatively uniform diameters, are found in loose bundles or in bundles aggregated by elastin, and often appear hollow in cross sections. Elastic fiber microfibrils and basement membrane-associated microfibrils share at least one major component, fibrillin (Sakai et al., 1986).
Although both fibrillins have been immunolocalized to microfibrils (Sakai et al., 1986;Zhang et al., 1994a), the specific functions of these proteins are unknown. Indications for similar but distinct functions for fibrillin-1 and fibrillin-2 in extracellular matrices have emerged from mutation analysis of their genes. Mutations in FBN1 cause the Marfan syndrome (Dietz et al., 1994), whereas FBN2 is linked to congenital contractural arachnodactyly (Lee et al., 1991;Tsipouras et al., 1992). Recently, the first fibrillin-2 mutations have been described in this disease (Putnam et al., 1995).
Electron micrographs of basement membrane regions in skin (Tsuji, 1980), lung (Low, 1961), kidney (Farquhar et al., 1961), and lens capsule (Raviola, 1971) reveal small connective tissue spaces with bundles of microfibrils that appear to intersect the lamina densa. These microfibrils intersecting basement membranes lack a visible amorphous elastic component. In some tissues like skin, continuous tracts of microfibrillar bundles may extend between basement membranes and adjacent elastic fibers. These electron microscopic observations suggest that microfibrillar proteins may interact with diverse types of extracellular matrix (ECM) molecules: components of basement membranes as well as elastic fiber components. In addition, microfibrils may interact with other components that are present in the ECM. Immunolocalization experiments have demonstrated that fibrillin is periodically arranged in microfibrils (Sakai et al., 1986) and that fibrillin periodicity in microfibrils liberated from their tissue environment is variable and extendable (Keene et al., 1991). In relaxed tissues, the fibrillin periodicity appears to match the periodic banding pattern of collagen fibers (Sakai et al., 1986). These observations suggest that fibrillin microfibrils may interact with molecules, which in turn interact with collagen fibers to stabilize the ECM.
The structure of the fibrillins contains 43 calcium-binding (cb) EGF-like domains. This type of module has been shown to mediate certain specific protein-protein interactions, for exam-* This work was supported by Fellowship Re 1021/1-1 (to D. P. R.) and Sonderforschungsbereich 266 from the Deutsche Forschungsgemeinschaft, by Grant AR38923 (to M. L. .C) from the National Institutes of Health, and by the Shriners Hospitals. 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.
** To whom correspondence should be addressed: Shriners Hospital For Crippled Children,Portland, ple the clotting activity of human coagulation factor IX (Rees et al., 1988) or the interaction of protein S with complement C4b-binding protein (Dahlbäck et al., 1990). Similar tandem arrays of cb EGF-like repeats exist in the extracellular matrix proteins fibulin-1 and fibulin-2 and are likely to be involved in calcium-dependent binding to other extracellular ligands such as fibronectin and nidogen (Sasaki et al., , 1995b. For Drosophila Notch, which is composed of 36 EGF-like repeats (25 of which contain consensus sequences for calcium binding) (Wharton et al., 1985), calcium-dependent binding to Delta and Serrate is mediated by two tandem cb EGF-like repeats (EGF repeats 11 and 12) (Rebay et al., 1991). Therefore, it has been proposed that Notch may serve as a multifunctional molecule composed of a tandem array of discrete ligand-binding units (Rebay et al., 1991). Similarly, fibrillins may be multifunctional molecules composed of tandem arrays of discrete ligand-binding domains.
To test this hypothesis as well as to specifically test whether known candidate ECM molecules may interact with fibrillin to stabilize the integrity of certain connective tissue zones, we have performed ligand-binding assays using recombinant peptides of fibrillin-1 and determined that fibulin-2, a novel ECM molecule (Pan et al., 1993), binds to fibrillin-1. Moreover, we show that fibulin-2 is localized to some microfibrils, demonstrating that specific binding may also occur in vivo. However, not all fibrillin-containing microfibrils are labeled with antibodies to fibulin-2, suggesting that functional differences exist between microfibrils in different connective tissues.

Expression of Recombinant Proteins
Recombinant mouse fibulin-1 variants C and D and mouse and human fibulin-2 were prepared as described (Pan et al., 1993;Sasaki et al., , 1995b. Recombinant fibrillin-1 peptides rF6, rF11, and rF20 were produced and purified as described (Reinhardt et al., 1996). An additional recombinant fibrillin-1 construct (rF23), coding for S 19 -I 489 , was prepared by PCR amplifying template HFBN29 (Corson et al., 1993) with oligonucleotides N1540S (5Ј-CAACAAGCTGTGCTCTGTT CCTATGG-3Ј) and DR51 (5Ј-ATAGTTTAGCGGCCGCTAGTGATGGT-GATGGTGATGAATACACTCCCCACGGAGG-3Ј), which introduced the coding sequence for six histidine residues, a stop codon, and a NotI site at the 3Ј end. The 74-base pair AgeI-NotI fragment of the PCR product was then fused with the AgeI-NotI restricted expression plasmid pCis-rF11H (Reinhardt et al., 1996). A 1441-base pair NheI-NotI fragment was then fused with the NheI-NotI restricted pCEP4/␥2III4 (Mayer et al., 1995). The resulting plasmid was designated pCEPSP-rF23H. The correct insertion of the insert and the sequence of the PCR-amplified DNA was confirmed by automated DNA sequencing following the manufacturer's instructions (Applied Biosystems). Transfection of 293/EBNA cells (Invitrogen), selection of clones, and purification of the recombinant peptide was done as described for construct rF20 (Reinhardt et al., 1996).

Proteins from Tissue or Cell Culture Sources and Antibodies
Mouse laminin-1 nidogen complex (Paulsson et al., 1987), human recombinant BM-40 (Nischt et al., 1991), and pepsin-solubilized bovine or human collagens I, II, III, IV, and V (Miller and Rhodes, 1982) were prepared following established procedures. Bovine collagen XI was a gift of Dr. David Wolley, and human plasma fibronectin was of commercial origin (Behringwerke).
For extraction of extracellular matrix proteins from cell culture sources, confluent layers of normal human skin fibroblasts were first washed two times with 50 mM Tris and 150 mM NaCl, pH 7.4 (TBS) including 2 mM phenylmethylsulfonyl fluoride and 5 mM N-ethyl-maleimide (protease inhibitors). The cells and the residual extracellular layer were then sequentially extracted with 0.1 ml/cm 2 TBS, including protease inhibitors, and 0.5% (w/v) Triton X-100 (cells) or 10 mM EDTA (extracellular layer) for 10 min at 25°C.

Binding Assays
The binding of different proteins to recombinant peptides was investigated by an enzyme immunoassay or a blot overlay assay. For the enzyme immunoassay, multiwell plates (Costar, 96 wells) were coated with purified protein (10 g/ml, 100 l/well) in 15 mM Na 2 CO 3 and 35 mM NaHCO 3 , pH 9.2, for 16 h at 4°C. For the blot overlay assay, recombinant fibulin-2 (1 g) and EDTA extract of the extracellular layer of skin fibroblasts (2 ml, precipitated with trichloroacetic acid) were separated by SDS-gel electrophoresis (5% w/v acrylamide) and then transferred onto nitrocellulose membrane (Bio-Rad) in 10 mM sodium borate, pH 9.2 (0.4 A for 45 min). Each of the following incubations was performed at 25°C and was followed by washing the wells or the nitrocellulose membrane three times with TBS including 0.025% Tween 20. Nonspecific binding sites were blocked with 5% nonfat dry milk in TBS for 1 h. The wells were then incubated with serial dilutions of the soluble ligand, and membranes, with a fixed concentration of 100 g/ml recombinant protein for 3 h, in TBS/2% dry milk including 2 mM CaCl 2 . Incubation (1.5 h) with monoclonal or polyclonal antibodies against the soluble ligand (1:100 -1:2000) in TBS/2% dry milk was then followed by incubation (1.5 h) with a peroxidase conjugate of goat anti-mouse or goat anti-rabbit immunoglobulin (Bio-Rad, 1:200 -1:800) in the same buffer. The color reaction of the enzyme immunoassay was performed with 1 mg/ml 5-aminosalicylic acid (Sigma) in 20 mM phosphate buffer, pH 6.8, containing 0.1% H 2 O 2 (100 l/well) for 3-4 min and stopped by adding 2 M NaOH (100 l/well). Color yields were determined at 492 nm. Nitrocellulose membranes were developed in a solution of 100 ml TBS/0.02% (v/v) H 2 O 2 plus 20 ml of 3 mg/ml 4-chloro-1-naphthol in methanol (Bio-Rad) until bands were clearly visible.
For kinetic binding studies, the BIAcore biosensor (Pharmacia Biotech Inc.) was used. Biotinylated fibulin-2 (50 g/ml) was coupled in 0.1 M sodium acetate, pH 4.0, to sensor chip SA5 (streptavidin chip), which resulted in 5500 resonance units equivalent to about 5 ng/mm 2 immobilized on the chip. Binding studies with soluble recombinant fibrillin-1 peptides rF11 in the range of 0.27-1.5 M and rF6 (0.5 M) were performed in TBS and 0.05% P20 (Pharmacia) containing either 2 mM CaCl 2 or 3.4 mM EDTA. Kinetic rate constants were calculated according to Fä gerstam et al. (1992) with BIAevaluation software version 2.1 (Pharmacia).

Immunofluorescence
Fetal bovine tissues were obtained from a local slaughterhouse. Human tissues were from a 16-week-old fetus, obtained according to human use protocols. The tissues were frozen in hexanes (EM sciences) and embedded in OCT (Miles). Cryosections (7 m) were prepared using a Kryostat 1720 (Leica). The sections were air-dried, fixed in acetone (10 min, Ϫ20°C), and then rehydrated in PBS. The sections were then incubated with a mixture of a 1:100 dilution of antiserum to mouse fibulin-2 and 100 g/ml mAb 201 or mAb 69 to fibrillin. After washing with PBS, the tissue sections were incubated with 1:50 dilutions of either phycoerythrin-conjugated goat anti-mouse IgG and fluoroblueconjugated goat anti-rabbit IgG (Biomeda) or fluorescein isothiocyanate-conjugated goat anti-mouse IgG and rhodamine-conjugated goat anti-rabbit IgG (Sigma). The sections were mounted in Gel mount (Biomeda) and photographed using an Axiophot (Zeiss) or Leitz DMRB (Leica) microscope equipped for epifluorescence.

Electron Microscopy
Section Surface Immunolabeling-Skin from the back of a 12-yearold male was obtained during surgery and processed for electron microscopy as described previously (Sakai and Keene, 1994). Fresh tissue was fixed in ice-cold 0.1% glutaraldehyde, rinsed in buffer and Tris-HCl, dehydrated in ice-cold ethanol, and embedded in Lowicryl K4 M. Polymerization in a CO 2 atmosphere was accomplished using UV light at Ϫ20°C. Ninety-nm-thick sections, which included the epithelium, papillary, and reticular dermis, were mounted onto Formvar-coated slot grids and exposed to a polyclonal antiserum against human fibulin-2 diluted 1:5 in PBS, followed by a goat anti-rabbit 10-nm colloidal gold conjugate (Amersham Corp.; diluted 1:3 in Tris/HCl with 1% bovine serum albumin, pH 8.0). Grids were exposed to the electron beam prior to staining in uranyl acetate and lead citrate. Controls included samples where the primary antibody was omitted and where a primary antibody of irrelevant specificity was substituted for the anti-fibulin-2 antibody.
Enbloc (Diffusion) Immunolabeling-Fresh neonate foreskin was immunolabeled as described previously (Sakai and Keene, 1994). Fresh tissue was rinsed in PBS, then submersed overnight in polyclonal antiserum against human fibulin-2 diluted 1:5 in PBS. Following an extensive rinse in PBS, the tissue was submersed overnight in goat anti-rabbit 5-nm colloidal gold conjugate (Amersham), rinsed extensively in PBS and 0.1 M cacodylate buffer, then fixed, dehydrated, and embedded in Spurrs epoxy. Ninety-nm-thick sections, which included the epithelium, papillary, and reticular dermis, were stained in uranyl acetate and lead citrate and examined in a Philips EM 410 transmission electron microscope. Controls were as described above.

Miscellaneous Methods
Protein concentrations were determined in triplicate after hydrolysis (6 M HCl, 24 h, 110°C) on a Beckman 6300 amino acid analyzer. For N-terminal sequencing, rF23 was analyzed on an automated protein sequencer (Hewlett Packard G1000S).

Protein Ligand Binding of Recombinant
Fibrillin-1 Peptides-To identify potential binding ligands of fibrillin-1, a variety of extracellular matrix proteins including fibronectin, laminin-1, BM40 (SPARC), fibulin-1C and fibulin-1D, fibulin-2, and collagen types I, II, III, IV, V, and XI were tested in solid phase assays for binding to recombinant fibrillin-1 peptides rF11 (amino acid positions 45-1527) and rF6 (amino acid positions 1487-2871), which together span the whole fibrillin-1 molecule (Reinhardt et al., 1996). Among the proteins tested, only fibulin-2 showed significant binding to the N-terminal peptide rF11 but not to the C-terminal peptide rF6 (Fig. 1). All other ligands, including the isoforms fibulin-1C and fibulin-1D, had no or only minor affinities to the fibrillin-1 peptides. Nearly identical binding profiles of human and mouse fibulin-2 to rF11 indicate a conserved binding epitope between species, suggesting an important function of the fibrillin-1/fibulin-2 binding (Fig. 1). The binding was reduced at least 10-fold when the ionic strength of the binding buffer was increased from 0.15 to 1 M NaCl, suggesting, at least partially, an electrostatic character of the interaction (Fig. 1).
The interaction of fibulin-2 with rF11 was also analyzed by a surface plasmon resonance method to determine kinetic rates in real time (Fä gerstam et al., 1992). Since the binding profiles for mouse and human fibulin-2 were nearly identical (Fig. 1) and the availability of human fibulin-2 was limited, we used mouse fibulin-2 for these experiments. Soluble rF11 in the concentration range of 0.27-1.5 M bound in a dose-dependent manner (14 -75 response units) to fibulin-2 fixed on a sensor chip (Fig. 2). The C-terminal fibrillin-1 peptide rF6 was inactive. Kinetic rate constants for the association (k ass ϭ 57.8 ϫ 10 3 Ϯ 3.7 ϫ 10 3 M Ϫ1 s Ϫ1 ) and dissociation (k diss ϭ 3.255 ϫ 10 Ϫ3 Ϯ 0.35 ϫ 10 Ϫ3 s Ϫ1 ) were calculated for each concentration and resulted in a dissociation constant of K d (k diss /k ass ) ϭ 56.3 Ϯ 3.5 nM for the fibrillin-1/fibulin-2 interaction. Binding was abolished when CaCl 2 in the binding buffer was replaced by EDTA (Fig. 2). However, when fibulin-2 was preincubated with rF11 in the CaCl 2 -containing binding buffer, the complex could not be dissociated by EDTA, indicating that after binding of the proteins, the calcium ions are held firmly within the binding site.
The interaction was also tested with authentic fibulin-2 and fibrillin from cell culture sources. We performed an overlay assay where one ligand is bound to nitrocellulose after electrophoresis under nonreducing conditions, and the other ligand is soluble. rF11 in the soluble phase bound to a ϳ600-kDa band of an EDTA extract of human skin fibroblasts and to recombinant fibulin-2, which comigrated with this band (Fig. 3, left panel). No binding was observed when rF11 was omitted (Fig. 3, middle panel). The ϳ600-kDa band was identified as authentic fibulin-2 by reaction with a 1:10,000 diluted polyclonal antiserum specific for fibulin-2 (Fig. 3, right panel). A second band (ϳ290 kDa), which reacts with rF11, represents possibly another fibrillin-binding protein, which is not yet identified. When rF11 and authentic fibrillin from serum-free medium of skin fibroblasts were immobilized onto nitrocellulose, no binding could be observed with soluble recombinant fibulin-2 (data not shown). These data suggest that the binding site in fibrillin-1 is inactivated after SDS treatment for gel electrophoresis, whereas the binding site in fibulin-2 is not affected by SDS.
For solid phase assays with authentic fibrillin, we purified small quantities (Ͻ10 g) from 1 liter of serum-free cell culture medium by affinity chromatography on mAb 26. The eluate of the affinity column was concentrated to 1 ml and then used in the binding assay. A typical binding profile was observed with immobilized recombinant fibulin-2, whereas the control without fibulin-2 did not show binding (data not shown). However, the absolute numbers of these experiments were relatively low (maximum absorbance at 492 nm ϭ 0.15), probably due to the low amounts of fibrillin present in the soluble phase or alternatively to partial inactivation of the fibulin-2 binding site by proteolysis. Similar results were obtained with inversed orientation of the ligands in the assay.
Localization of the Fibulin-2 Binding Site on Fibrillin-1-To narrow down the binding site for fibulin-2 on fibrillin-1, we produced recombinantly a subdomain of fibrillin-1 (rF23) ( Table I), which together with the previously described subdomain rF20 spans the whole N-terminal fibulin-2-binding peptide (rF11). The episomal expression plasmid designed was pCEPSP-rF23 coding for S 19 -I 489 of fibrillin-1 plus additional N-terminal (APLA; Mayer et al., 1995) and Cterminal (HHHHHH) amino acids. The peptide was expressed in stably transfected 293/EBNA cells in amounts of 10 -15 g/ml/day. After a one-step purification on a cobaltloaded chelating column (Reinhardt et al., 1996), the peptide showed Ͼ95% homogeneity on SDS gel electrophoresis (data not shown) with an apparent molecular mass of 63 kDa, which is slightly higher than expected from the cDNA ( Table  I). The observed difference is likely due to the occupation of one N-glycosylation site predicted in this peptide (position 448 -450). Edman degradation of rF23 revealed a sequence 30 amino acid residues shorter than expected (R 45 GGGGHDALKGP) (Table I). This N-terminal sequence, also observed previously for rF11 (Reinhardt et al., 1996), occurs after a consensus processing sequence for many propeptides (RXK/RR; Hosaka et al., 1991), suggesting that fibrillin-1 undergoes intracellular N-terminal processing. FIG. 4. Localization of the fibulin-2 binding site on fibrillin-1 in a solid phase binding assay. Fibrillin-1 peptides rF11 (f), rF20 (q), and rF23 (å) were used as soluble ligands with immobilized human fibulin-2. Similar binding profiles were obtained with inversed orientation of the ligands. Data points are averages of duplicates. For clearer graphs, only the positive part of the error bars are indicated. The experiment was repeated three times.  hardt et al., 1996). b Molecular mass determined by SDS electrophoresis.
FIG. 3. Binding of recombinant fibrillin-1 peptide rF11 to authentic fibulin-2 in a blot overlay assay. EDTA extract (2-ml aliquots) of the extracellular layer of human skin fibroblasts (a) and recombinant fibulin-2 (1 g) (b) were separated under nonreducing conditions by SDS gel electrophoresis and then transferred onto nitrocellulose. The soluble binding ligand was recombinant rF11 (100 g/ ml), which was detected with mAb 201 (left panel). In the control, rF11 was omitted (middle panel). Identification of the ϳ600-kDa rF11-binding protein in the EDTA extract as fibulin-2 was demonstrated by reaction with a polyclonal antiserum, specific for fibulin-2 (1:10,000 diluted) (right panel). The additional ϳ290-kDa band in the EDTA extract that reacts with rF11 (left panel) was not identified. The positions of marker proteins are indicated.
Binding of mAb 26, which is dependent on intact disulfides, to rF23 in Western blot analysis (data not shown) demonstrated the correct folding of this peptide and, furthermore, confirmed that the epitope for mAb 26 is located at the amino terminus of fibrillin-1, as suggested previously (Reinhardt et al., 1996).
Recombinant peptides rF20 (Reinhardt et al., 1996) and rF23 were used as soluble ligands in a solid phase binding assay with immobilized human fibulin-2 (Fig. 4). Nearly identical binding profiles of rF11 and rF23 and the complete lack of binding of rF20 clearly demonstrate that the binding site for fibulin-2 is located at the N-terminal end of fibrillin-1 (amino acid positions 45-450). Whether the binding site involves the two cb EGF-like motifs present in rF23 (positions 246 -329) remains to be established.
Electron Microscopic Immunolocalization-Ultrastructural localization of fibulin-2 is shown in Fig. 6. In "enbloc" immunolocalization experiments of fibulin-2, the antibody-directed gold labeling was to microfibrils at the outer periphery of the amorphous elastin core. No labeling of microfibrils positioned away from the periphery of the elastin core was noted (Fig. 6E). Immunolocalization of fibulin-2 to the surfaces of cross-sections of elastin fibrils of lightly fixed skin demonstrated that the majority of gold particulates delineate the periphery of the elastin fibrils and are located at the intersection of microfibrils with elastin (Fig. 6, A-D). The relatively low amount of labeling within the elastin core is approximately equal to that seen randomly distributed on the section surface and, therefore, represents background labeling. By either the diffusion or section surface methods, elastin fibers in any location of the deep reticular dermis or close to the dermal-epidermal junction were labeled as described above.
Extensive structural studies of fibrillin-1 recombinant peptides (Reinhardt et al., 1996) have shown that the peptides utilized in these studies are fully functional and equivalent to native peptides. To demonstrate the biological relevance of these in vitro studies with recombinant peptides, we performed immunolocalization studies. Ultrastructural investigations of skin revealed fibulin-2 primarily at the interface of amorphous elastin cores and microfibrils, even though fibrillin-1 is a component of all microfibrils. These data suggest that fibulin-2 may also bind to a component of the amorphous elastin core. A similar distribution was recently demonstrated for emilin (Bressan et al., 1993). In contrast, fibulin-1 is located within the amorphous core of elastic fibers, but not in the fibrillincontaining, elastin-associated microfibrils (Roark et al., 1995). Although more peripheral microfibrils were not labeled by fibulin-2 antibodies, these antibodies did label microfibrils at the dermal-epidermal junction. However, antibody labeling of these microfibrils was not very extensive.
Several lines of evidence suggest that fibulin-2 is a microfibril-associated protein rather than an integral structural component of microfibrils. Fibulin-2 labeling of microfibrils is not linearly periodic. Some, but not all, tissues containing microfibrils are labeled by fibulin-2 antibodies. Fibulin-2 can be extracted from tissues or cell layers with EDTA, indicating that it is not a covalently cross-linked component of microfibrils, and in vitro binding studies with fibrillin-1 peptides showed strongly reduced binding affinities in the presence of 1 M NaCl. Beaded microfibril structures, on the other hand, have been identified after repeated extractions of tissues including 1 M NaCl washes, collagenase digestions, and final extraction with guanidine-HCl (Keene et al., 1991), conditions which would be expected to dissociate fibulin-2 from the microfibrils. These data, together with the noted absence of fibulin-2 in tissues that are subject to strong tensional forces (e.g. tendon and ciliary zonule), suggest that the fibrillin-1/fibulin-2 interaction is not required for the structural/mechanical integrity of microfibrils.
Immunohistochemical studies and in vitro binding data demonstrate that fibrillin-1 interacts with fibulin-2 in certain tissues (skin, perichondrium, kidney glomerulus, and the elastic intima of blood vessels) and not in others (tendon, cartilage, ciliary zonule, other areas of connective tissue in the kidney, and lung). Since fibulin-2 is able to interact with multiple ECM molecules in in vitro binding experiments, a combination of cellular expression, cellular deposition of molecules into the ECM, and the multifunctional nature of fibulin-2 must regulate the in vivo protein-protein interactions of these molecules. In basement membrane regions, like the kidney glomerulus and the dermal-epidermal junction, which does not contain visible amorphous elastin cores, fibulin-2 may bind to fibrillin-1 and basement membrane proteins (e.g. nidogen or type IV collagen) to stabilize the interaction between microfibrils and the lamina densa. The interaction of microfibrils with the lamina densa is likely not stabilized by direct binding between fibrillin-1 and the major basement membrane components (type IV collagen, laminin-1, and nidogen), since no binding between these molecules was detected in our studies.
In addition to a potential role in stabilizing interactions between structural elements of the ECM (microfibrils and elastic cores; microfibrils and basement membranes), the fibrillin-1/fibulin-2 interaction may play an important role in developmental processes. Both molecules are highly expressed in the developing heart. At early developmental stages, the most striking expression of fibrillin-1 and fibulin-2 is in endocardial cushion tissue. In the developing mouse embryo (days 8.5-9.0), the endocardial tissue was the only place where Fbn1 could be detected by in situ hybridization (Yin et al., 1995a), and at day 11, fibulin-2 expression was found exclusively in the endocardial cushion (Zhang et al., 1995a). In the developing chick, fibrillin was detected by immunofluorescence in the endocardial layer at stage 11 and in the developing cushion mesenchyme from stage 20 on (Hurle et al., 1994). The development of the heart from a muscular tube into a chambered structure involves the transformation of endocardial cells into mesenchymal cells that form the endocardial cushion tissue; the endocardial cushion then contributes to the formation of the valves and septa of the chambered heart (Little and Rongish, 1995). The prominent coexpression of fibrillin-1 and fibulin-2 in a transitional developing structure, the endocardial cushion, suggests that the interaction of these molecules may promote the transition of endocardial cells to mesenchyme, contribute to a specialized cushion matrix, and also permit or promote the formation of valves and septa.
The calcium-dependent nature of the fibrillin-1/fibulin-2 interaction suggests that cb EGF-like repeats in one or both binding ligands contribute to the binding site. Since binding to fibulin-2 was demonstrated with rF23, only two cb EGF-like repeats in fibrillin-1 (nos. 1 and 2) could be responsible for the active binding site. In fibulin-2, the binding site has not been narrowed down. Therefore, all of the nine cb EGF-like repeats in human fibulin-2 are possible candidate domains for the binding site.
A Marfan mutation has been identified in most of the 43 cb EGF-like repeats present in fibrillin-1 (Dietz and Pyeritz, 1995). However, to date, mutations have not been found to occur in cb EGF-like repeats nos. 1 and 2. If these repeats contribute to the binding site for fibulin-2, these data might again indicate that binding of fibulin-2 to fibrillin-1 is important to human development. Mutations in these domains might lead to embryonic death. In the two fibrillins, cb EGF-like repeats nos. 1 and 2 are practically identical at the level of primary structure. Therefore, it is possible that fibrillin-2 may also bind fibulin-2. This interaction may be important particularly during early embryogenesis when fibrillin-2 is more highly expressed than fibrillin-1 (Zhang et al., 1995b).