The Self-association and Fibronectin-binding Sites of Fibulin-1 Map to Calcium-binding Epidermal Growth Factor-like Domains*

Fibulin-1 is a modular glycoprotein with amino-terminal anaphylatoxin-like modules followed by nine epidermal growth factor (EGF)-like modules and, depending on alternative splicing, four possible carboxyl termini. Fibulin-1 has been shown to self-associate as well as to bind calcium, fibronectin (FN), laminin, nidogen, and fibrinogen. To map ligand-binding sites within fibulin-1, polypeptides corresponding to various regions of fibulin-1 were expressed recombinantly and evaluated for their capacity to bind calcium, FN, or fibulin-1. A calcium-binding site(s) was mapped to EGF-like modules 5–9. A fibulin-1 self-association site was localized to EGF-like modules 5 and 6 (amino acid residues 356–440), as was a binding site for FN. The self-association interaction mediated by this pair of modules involved calcium since divalent cation chelators reduced the binding affinity of the interaction. By contrast, FN binding to EGF-like modules 5 and 6 was unaffected by the presence of divalent cation chelators. It can be concluded that EGF-like modules 5 and 6 bind calcium and mediate homotypic interaction between EGF-like modules 5 and 6 present in different fibulin-1 molecules and heterotypic interaction between EGF-like modules 5 and 6 and type III repeats 13 and 14 in FN. While additional binding sites for calcium or FN were not detected, another fibulin-1 self-association site was found within amino acid residues 30–173. However, unlike the self-association site in EGF-like modules 5 and 6, which was functional in the native protein, the amino-terminal site was cryptic and revealed only after the protein was denatured.

Fibulin-1 is an extracellular matrix and plasma glycoprotein that belongs to an emerging gene family with three members designated fibulin-1-3 (1)(2)(3). While the function(s) of fibulin-1 is not known, fibulin-1 has been shown to bind calcium (4,5); the extracellular matrix proteins fibronectin (FN), 1 nidogen, and laminin (6,7); and the coagulation protein fibrinogen (8). In addition, fibulin-1 is capable of self-association (6,7). These interactions, individually or in combination, may account for the observed association of fibulin-1 with basement membranes (5, 9 -11), connective tissue elastic fibers (12), and fibrin clots (8). The molecular basis for the protein interactions of fibulin-1 as well as its binding to calcium have not been fully established. It has been determined that fibulin-1 binds to FN within type III repeats 13 and 14 (6), to a site within the amino-terminal G1-G2 domains of nidogen (7), to a site contained within the carboxyl terminus of the ␣ chain of Engelbreth-Holm-Swarm tumor laminin (␣1␤1␥1) (7,13), and to a site within the carboxyl-terminal region of the fibrinogen B␤ chain (8). However, localization of the binding sites within fibulin-1 for these ligands has not been determined. Similarly, the binding site(s) for calcium has not been localized within fibulin-1, although four of its nine EGF-like modules contain the consensus sequence for post-translational hydroxylation of asparagine, and such hydroxylation has been associated with calcium-binding EGF-like modules in a number of proteins (14,15).
In an effort to determine the location of ligand-binding sites within fibulin-1, we have expressed various portions of fibulin-1 in stably transfected eucaryotic cells, purified the expressed proteins, and used them in solid-phase binding assays to evaluate their capacity to bind calcium, fibulin-1, and FN.

MATERIALS AND METHODS
Antibodies-The mouse monoclonal fibulin-1 antibodies 3A11 and 5D12 and the rabbit polyclonal fibulin-1 antiserum 1323 have been described previously (1,4). The epitope for the 3A11 antibody maps to the amino-terminal region of fibulin-1 within residues 30 -153, whereas the epitope for the 5D12 antibody maps to the carboxyl-terminal region of fibulin-1C within residues 567-683 (data not shown). The 3A11 and 5D12 IgGs were purified by protein G-Sepharose (Pharmacia Biotech Inc.) chromatography. Mouse monoclonal fluorescein isothiocyanate (FITC) antibody (FL-D6) was purchased from Sigma.
Proteins-Fibulin-1 was isolated from human placenta by immunoaffinity chromatography using 3A11 IgG-Sepharose as described previously (1). Human FN was purified as described by Miekka et al. (16). The 30-kDa heparin-binding fragment of FN (30-kDa FN) that contains type III modules 12-14, also known as the HEP-2 fragment, was provided by Dr. Kenneth Ingham (American Red Cross, Rockville, MD). Ovalbumin was purchased from Sigma.
Radioiodination and Fluorescein Labeling of Proteins-Fibulin-1 (50 g) was radioiodinated using 20 g of IODO-GEN (Pierce), 0.5 mCi of Na 125 I (Amersham Corp.), and 0.25 mM NaI in 100 l of phosphatebuffered saline. Radiolabeled fibulin-1 was separated from the unincorporated iodine by gel filtration chromatography using a Sephadex G-25M column (Pharmacia Biotech Inc.). The specific activity achieved ranged from 1 to 10 Ci/g of protein.
Proteins (e.g. placental fibulin-1, recombinant fibulin-1 polypeptide Fib5-9C, and 30-kDa FN) were fluorescein-labeled using a method described by Busby and Ingham (42). Briefly, each protein was mixed with a 40-fold molar excess of FITC (Molecular Probes, Inc., Eugene, OR) in 0.1 M NaHCO 3 , pH 9.5. After a 4-h incubation in the dark at room temperature, FITC-labeled protein was separated from unincorporated FITC by gel filtration on a Sephadex G-25M column. The degree of labeling was determined optically as described by Ingham and Brew (17). A typical labeling efficiency was 4 -5 mol of FITC/mol for fibulin-1, 30-kDa FN, and FN and 0.75 for Fib5-9C.
Recombinant Expression of Various Portions of Fibulin-1-Plasmid constructs were designed to express full-length fibulin-1C and six permeations of fibulin-1 that are depicted in Fig. 1. The eucaryotic expression vector pcDNA3 (Invitrogen, San Diego, CA) was used to place the fibulin-1 cDNAs under the transcriptional control of the human cytomegalovirus promoter. The design of the full-length fibulin-1C expression construct designated pcDNAINeoFibC and its transfection into human fibrosarcoma HT1080 cells (ATCC CCL-121) have been described previously (3).
The constructs FibE1-9C, FibE5-9C, and FibE7-9C were generated by ligating a PCR fragment encoding the fibulin-1 signal sequence (PCR fragment 1) (Table I) to PCR fragments 2, 3, and 4, respectively, and subcloning each into pcDNA3. The construct designated FibA1-3C was made by ligating PCR fragments 5 and 6 and subcloning into pcDNA3. FibA1-3E1-4 was made by subcloning a single PCR fragment (fragment 7) into pcDNA3. FibA1-3E1-8 was made using site-directed mutagenesis by overlap extension (18) of two PCR fragments (fragments 8 and 9) to convert Asp 525 to a stop codon. All of the PCR fragments described were made using the pairs of primers indicated in Tables I and II and pcDNAINeoFibC (3) as template. Fibulin-1 plasmid constructs were individually introduced into a non-fibulin-1-expressing cell line, human fibrosarcoma HT1080 cells, using calcium phosphate transfection with reagents supplied in a kit (Life Technologies, Inc.). Cells were grown in complete medium (minimal essential medium/Earle's balanced sodium salt (Hyclone Laboratories, Logan, UT), 100 units/ml penicillin, and 100 g/ml streptomycin (Life Technologies, Inc.)) containing 0.6 mg/ml Geneticin (Life Technologies, Inc.). Colonies of resistant cells were isolated after 4 -6 weeks using a sterile cotton swab. For immunological evaluation, the cells were grown in serum-free medium. Aliquots of the conditioned culture medium were screened by immunoblot analysis using fibulin-1 antibody 1323. Those cell lines that were found to express relatively high levels of each of the recombinant forms of fibulin-1 were selected for largescale protein purification.
Purification of Fibulin-1 Permeations Expressed by Transfected Cells-Stably transfected cell lines were grown to confluence in 2-liter roller bottles (Corning Inc., Corning, NY) in complete medium containing 0.3 mg/ml Geneticin. The medium was replaced with serum-free medium, and the cells were grown for 2 days. The medium was collected and clarified by centrifugation at 5000 ϫ g. The supernatant was supplemented with phenylmethylsulfonyl fluoride (final concentration of 1 mM) and EDTA (final concentration of 5 mM) and preabsorbed on a column of Sepharose CL-4B. The flow-through fraction was applied to one of two types of anti-fibulin-1 IgG-Sepharose affinity columns. 5D12 IgG-Sepharose was used to purify those fibulin-1 polypeptides that contained the carboxyl-terminal fibulin-type module (e.g. FibE1-9C, FibE5-9C, and FibE7-9C fragments), whereas 3A11 IgG-Sepharose was used to purify those fibulin-1 polypeptides that contained the amino-terminal repeated anaphylatoxin domain. Bound protein was eluted with 4 M KSCN. After dialysis against Tris-buffered saline, the fibulin-1 polypeptide preparations were absorbed on heparin-Sepharose and gelatin-Sepharose to remove any contaminating FN, as described previously (19).
Gel Blot Overlay Assay-Recombinant fibulin-1 polypeptides were transferred to nitrocellulose membranes after SDS-PAGE. The membranes were treated with 3% nonfat dry milk in Tris-buffered saline and 5 mM CaCl 2 and incubated for 18 h at 4°C with 125 I-labeled fibulin-1 (20 nM) in the same buffer containing 0.05% Tween 20. Following incubation, the filters were washed with Tris-buffered saline and 0.05% Tween 20 and used to expose Kodak X-Omat AR film at Ϫ70°C.
Solid-phase Binding Assays-Enzyme-linked immunosorbent assays (ELISA) were performed as described previously (6). Homologous and heterologous ligand displacement assays were carried out in a manner similar to that described before (20). However, FITC-labeled proteins were used instead of radiolabeled proteins, and binding was measured by ELISA using monoclonal FITC antibody and goat anti-mouse IgG conjugated to alkaline phosphatase (Bio-Rad) and the substrate pnitrophenyl phosphate (disodium; Sigma). The relative efficiency of recombinant fibulin-1 derivatives to bind microtiter plastic was evaluated by ELISA, and based on the results, subsequent coating concentrations were adjusted to achieve equimolar coatings for each protein.
45 Ca 2ϩ Binding Assay-Evaluation of the ability of recombinant fibulin-1 subfragments to bind calcium was carried out using the method of Maruyama et al. (21). Recombinant fibulin-1 fragments were separated by electrophoresis on 10% acrylamide gels under nonreducing conditions and transferred to nitrocellulose membranes. The membranes were washed with 60 mM KCl and 10 mM imidazole HCl, pH 6.8, and incubated for 10 min at room temperature with 45 Ca 2ϩ (1 Ci/ml) in the same buffer. Unbound 45 Ca 2ϩ was removed by washing the membrane twice (for 5 min each) with 50% ethanol. The membranes were dried and used to expose Kodak X-Omat AR film at Ϫ70°C. erythro-␤-Hydroxyaspartate/erythro-␤-Hydroxyasparagine Analysis-Analysis for the presence of hydroxyasparagine and/or hydroxyaspartate in fibulin-1 was carried out as described by Przysiecki et al. (22). Briefly, 100 g of fibulin-1 or bovine protein S (Enzyme Research Laboratories, Inc., South Bend, IN) were dried and resuspended in 50 mM Tris, pH 7.5 (2 g/l final concentration). Pronase (Calbiochem) was added to a final concentration of 0.16 g/l and incubated for 8 h at 37°C. An additional aliquot of Pronase was added to make the concen- 1313-2200 441-683 1658-2200 555-683 The indicated primer designations refer to those described in Table  II.
b The construct name refers to the plasmid construct into which the indicated fragment, either alone or in combination with another fragment, was incorporated.
c In order to generate the AatII cloning site located at the 3Ј-end of PCR fragment 1, Ala 29 (within the signal peptide) was converted to Val. d Required ligation of PCR fragments 1 and 2. e Required ligation of PCR fragments 1 and 3. f Required ligation of PCR fragments 1 and 4. g Required ligation of PCR fragments 5 and 6. h Contained a mutation that introduced a stop codon at positions 1583-1585.
i Required annealing of fragments 8 and 9 and subsequent fill-in according to the overlapping extension PCR procedure described by Ho et al. (18).  CTCATCAATtAaTTGGCAGTTGCG 1571-1594 b a The numbers indicated correspond to residues within pcDNAINeo (Invitrogen).
b The numbers indicated correspond to residues within human within fibulin-1C cDNA (1). tration 0.28 g/l and incubated for 5 h at 37°C. Following the Pronase digestion, aminopeptidase M (Calbiochem) was added to a concentration of 0.01 g/l and incubated for 10 h at 37°C. The sample was adjusted to pH 2.0 with 98% formic acid and subjected to cationexchange HPLC analysis. Chromatography was carried out using a 10 ϫ 0.46 cm (inner diameter) column packed with SS/D-X8.25 resin (Sierra Separations, Reno, NV) and a 3 ϫ 0.32-cm guard column of Aminex A-9 (Bio-Rad). Elution buffers and post-column derivatization conditions were as described by Przysiecki et al. (22).

Expression and Isolation of Recombinant Fibulin-1 in HT1080
Cells-A series of plasmid constructs were generated so as to permit expression in eucaryotic cells of six permeations of human fibulin-1. Fibulin-1 polypeptides were expressed that contained, either alone or in combination, each of the major structural elements found in fibulin-1, including the aminoterminal anaphylatoxin-like region (designated A), the EGFlike region (designated E), and the carboxyl-terminal fibulintype module (designated C). The portion of the fibulin-1 protein that each construct encoded is schematically depicted in Fig. 1. Each plasmid was stably transfected into HT1080 cells, and the cDNA insert-encoded protein was isolated by anti-fibulin-1-Sepharose affinity chromatography. Shown in Fig. 2 is the SDS-PAGE analysis of the resulting preparations of recombinant fibulin-1 polypeptides. The typical yields ranged from 0.48 to 1.8 mg of protein/liter of conditioned culture medium. The results indicate that chemical amounts of fibulin-1 polypeptides corresponding to various regions of fibulin-1 could be derived from cell lines stably transfected with the fibulin-1 expression constructs.
Of the nine EGF-like modules present in fibulin-1, four (EGF-like modules 5-8) bear a consensus sequence for asparagine ␤-hydroxylation. Such hydroxylation has been associated with calcium-binding EGF-like modules in several proteins (14,15). To determine whether fibulin-1 contains ␤-hydroxyasparagine, cation-exchange HPLC analysis of proteolytic digests of fibulin-1 was performed. As shown in Fig. 4, the HPLC profile of proteolyzed fibulin-1 contains a peak corresponding to that of the erythro-␤-hydroxyasparagine standard (Fig. 4, traces C and  A, respectively). Analysis of bovine protein S, which is known to contain hydroxylated asparagine (23), produced a similar peak in its HPLC profile (Fig. 4, trace B). Because data from the In the nomenclature used for each polypeptide (indicated at the left), the letter A refers to the anaphylatoxin-like modules (the numbers 1-3 refer to the presence of three A modules); E refers to EGF-like modules, with the numbers that follow defining which of the nine EGF-like modules are included; and C refers to the carboxyl-terminal module, also known as the fibulin-type module. The dotted lines indicate that the intervening sequence has been deleted. enzymatic digestion of fibulin-1 could not be used to make quantitative estimates of the amount of ␤-hydroxyasparagine in fibulin-1, HPLC analysis of acid-hydrolyzed fibulin-1 was performed. The results of this analysis (data not shown) indicated that fibulin-1 contains 3 mol of ␤-hydroxyaspartate/mol of protein. Since the acid hydrolysis converts asparagine to aspartic acid and each of the consensus ␤-hydroxylation sites present in EGF-like modules 5-8 contains asparagine, the results can be interpreted to indicate that there are 3 mol of ␤-hydroxyasparagine in fibulin-1.
A Fibulin-1 Self-association Site Maps to EGF-like Modules 5 and 6 -To localize the regions of fibulin-1 that mediate its self-association, the various recombinantly derived fibulin-1 polypeptides were coated on microtiter wells and tested for their ability to promote binding of solution-phase FITC-labeled fibulin-1. As shown in Fig. 5A, fibulin-1 polypeptides containing EGF-like modules 1-9, 1-8, and 5-9 (e.g. FibE1-9C, FibE5-9C, and FibA1-3E1-8) promoted binding of FITC-labeled fibulin-1. Fibulin-1 polypeptides that lacked EGF-like modules 5 and 6 (e.g. FibA1-3C, FibA1-3E1-4, and FibE7-9C) showed little or no ability to bind FITC-labeled fibulin-1. Similar binding results were obtained when radioiodinated or digoxygenin-labeled fibulin-1 was used as a probe (data not shown). In addition, when a recombinant fibulin-1 polypeptide that contained EGF-like modules 5-9 (FibE5-9C) was FITClabeled and used as a probe in the binding assays, it was found to bind to those polypeptides that contained EGF-like modules 5 and 6, but not to the polypeptides that lacked the two modules (Fig. 5B). Taken together, the binding data indicate that a fibulin-1 self-association site is contained within EGF-like modules 5 and 6. Considering the above-mentioned finding that EGF-like modules 5-8 also bind calcium, we evaluated the effect of divalent cation chelators (EDTA and EGTA) on the fibulin-1/ fibulin-1 interaction. As shown in Fig. 5C, when the binding reactions were done in the presence of EDTA, there was a marked decrease in the binding of FITC-labeled fibulin-1 to fibulin-1 polypeptides bearing EGF-like modules 5-8. A similar magnitude of reduction in binding was obtained when EGTA was used instead of EDTA (data not shown). While the apparent affinity of fibulin-1 binding to polypeptides bearing EGF-like modules 5-8 was significantly reduced by either EDTA or EGTA, complete inhibition of binding was not achieved. The results suggest that calcium bound to EGF-like modules 5-8 plays an important role in the self-association interaction of fibulin-1.
A Cryptic Fibulin-1 Self-association Site Maps to the Aminoterminal Region-The ability of the recombinantly derived fibulin-1 polypeptides to promote binding of 125 I-labeled fibulin-1 in a gel blot overlay assay was next tested. As shown in Fig. 6, radioiodinated fibulin-1 bound only to those recombinantly derived fibulin-1 polypeptides that contained the aminoterminal anaphylatoxin-like domains (e.g. FibA1-3E1-4, FibA1-3E1-8, and FibA1-3C). Furthermore, a breakdown fragment present in the placenta-derived fibulin-1 preparation also did not bind 125 I-labeled fibulin-1 (see arrowheads in  Fig. 1 for nomenclature), placenta-derived fibulin-1, or ovalbumin (Oval). Bound FITC-labeled protein was detected with mouse monoclonal FITC antibody, alkaline phosphatase-conjugated goat anti-mouse IgG, and pnitrophenyl phosphate substrate. In C, the binding of FITC-labeled fibulin-1 to wells coated with the indicated fibulin-1 polypeptides, placentaderived fibulin-1, or ovalbumin was measured in the presence (open symbols) or absence (closed symbols) of 10 mM EDTA. The data shown are representative of four experiments, each performed in duplicate. 6). We subjected this fragment to amino-terminal sequence analysis and found that it began at Asp 173 , which occurs within the EGF-adjoining segment (data not shown). The fragment therefore lacks the anaphylatoxin-like domains and most of the EGF-adjoining segment and likely contains the nine EGF-like modules and the carboxyl-terminal fibulin-type module. The results indicate that a fibulin-1-binding site exists within the amino-terminal portion of fibulin-1 (residues 30 -173) containing the anaphylatoxin-like domains and the small EGF-adjoining segment.
The results from the gel blot overlay assays are contrary to those derived from the microtiter well binding assays, which showed that polypeptides containing the anaphylatoxin-like domains and the EGF-adjoining segment, but lacking EGF-like modules 5 and 6 (e.g. FibA1-3E1-4 and FibA1-3C), failed to promote fibulin-1 binding. Considering that the gel blot overlay assay involves SDS denaturation of the fibulin-1 polypeptides, whereas the microtiter assays do not, it can be concluded that the amino-terminal self-association site is not available in the native protein, but can be exposed following denaturation. In addition, these data suggest that the fibulin-1-binding site contained within EGF-like modules 5 and 6 is sensitive to denaturation and therefore can be considered to be a conformation-dependent binding site.
Fibulin-1/Fibulin-1 Binding Is Mutually Exclusive of Fibulin-1/FN Binding-Since the fibulin-1 self-association site and FN-binding sites were found to map to EGF-like modules 5 and 6, we were interested in determining whether binding was mutually exclusive. As shown in Fig. 8A, solution-phase FITClabeled fibulin-1 binding to immobilized fibulin-1 could be blocked by solution-phase FN. As an experimental control, fibulin-1 and fibulin-1 polypeptide-containing EGF-like modules 5 and 6 (FibE5-9C) competed for the solution-phase FITClabeled fibulin-1 binding to immobilized fibulin-1, whereas a fibulin-1 polypeptide lacking EGF-like modules 5 and 6 (FibE7-9C) did not inhibit the binding. It was also found that solution-phase FITC-labeled fibulin-1 binding to immobilized 30-kDa FN was inhibited by solution-phase fibulin-1 or fibulin-1 polypeptide-containing EGF-like modules 5 and 6 (FibE5-9C) (Fig. 8B). The results indicate that the interaction of fibulin-1 with EGF-like modules 5 and 6 is mutually exclusive of the binding of FN to EGF-like modules 5 and 6. This may mean that fibulin-1, once bound to FN via EGF-like modules 5 and 6, would not support binding to another fibulin-1 via that site. DISCUSSION We have described the use of recombinantly derived fibulin-1 polypeptides to map three major ligand-binding sites within fibulin-1. We showed that the calcium binding activity of fibulin-1 is contained within EGF-like modules 5-9 and that EGFlike modules 5 and 6 (amino acid residues 356 -440) bind fibulin-1 and FN. Furthermore, we showed that the aminoterminal region of fibulin-1 (amino acid residues 30 -173) contains a cryptic fibulin-1-binding site that can be exposed by detergent denaturation.
The function of EGF-like modules 5 and 6 of fibulin-1 in mediating self-association and FN binding is fitting with the FIG. 6. Gel blot overlay assay reveals that 125 I-labeled fibulin-1 binds to denatured fibulin-1 polypeptides that contain the amino-terminal region. Shown in A is a Coomassie Blue-stained gel containing recombinant fibulin-1 polypeptides, placenta-derived fibulin-1, and molecular mass standards (in kDa) that were electrophoresed on an SDS-10% polyacrylamide gel (nonreducing conditions). Shown in B is an autoradiograph of a membrane containing proteins from a duplicate gel as in A that had been electrophoretically transferred to a nitrocellulose membrane and probed with 125 I-labeled fibulin-1 (20 nM). Arrowheads indicate the position of a fibulin-1 breakdown product (with residue 173 at its amino terminus) that does not bind to 125 Ilabeled fibulin-1. fact that EGF-like modules within a number of other proteins have been shown to mediate protein/protein interactions. For example, the binding of Notch protein to Delta protein involves EGF-like modules 11 and 12 within Notch (24). An EGF-like module within laminin mediates its binding to nidogen (25). A single EGF-like module within the urokinase-type plasminogen activator mediates binding to the 60-kDa urokinase receptor (26). A pair of EGF-like modules in thrombomodulin mediate binding to thrombin (27). In addition, genetic diseases such as hemophilia and Marfan's syndrome result from mutations that affect calcium-binding EGF-like domains within coagulation factor IX (28) and fibrillin-1 (29,30), respectively. In the case of the mutations of fibrillin-1 that have been implicated in Marfan's syndrome, it has been speculated that they disrupt EGF-like domain interactions within fibrillin-1 or between fibrillin-1 monomers, thereby resulting in defective assemblages of connective tissue microfibrils (31,32). Similarly, mutations of the first EGF-like module of factor IX perturb the ability of the protein to mediate factor VIIIa-dependent activation of factor X. Common to both genetic disorders is that mutations that inhibit calcium binding to individual EGF-like modules can result in the destabilization of helical structure adopted by calcium-binding EGF-like modules (28). The binding studies presented here indicate that the divalent cationstabilized structure of EGF-like modules 5 and 6 is necessary for optimal fibulin-1/fibulin-1 binding, but not for fibulin-1/FN binding.
␤-Hydroxylated asparagine/aspartic acid residues are present within calcium-binding EGF-like modules of a number of proteins including fibrillin-1 (33) and vitamin K-dependent coagulation proteins such as protein C (34), factor IX (35), and factor X (36). While a correlation exists between the presence of ␤-hydroxylated asparagine/aspartic acid residues and calcium within EGF-like modules, experiments with factor IX indicate that ␤-hydroxylation is not necessarily required for calcium binding (37). Consistent with the fact that four fibulin-1 EGFlike modules (EGF-like modules 5-8) have consensus asparagine ␤-hydroxylation sites (14), we found that fibulin-1 indeed contains ␤-hydroxylated asparagine. Since we detected only 3 mol of ␤-hydroxylated asparagine/mol of fibulin-1, it is apparent that not all of the EGF-like modules of fibulin-1 that contain the consensus hydroxylation sequences are substituted. Our data indicate that the calcium binding activity of fibulin-1 is likely contained within EGF-like modules 5-8, the same modules that have consensus asparagine ␤-hydroxylation se-quences. It remains to be determined whether the EGF-like modules that bind calcium in fibulin-1 correspond with the three that contain ␤-hydroxylated asparagine.
EGF-like modules 5 and 6 of fibulin-1 appear to be quite versatile in that they can mediate self-association of fibulin-1 via homotypic EGF-like module/EGF-like module interaction as well as mediate fibulin-1 interaction with FN via a heterotypic interaction, presumably through the binding of EGF-like modules 5 and 6 to type III repeat(s) 13 and 14 of FN (6). Homotypic interaction involving EGF-like modules is likely similar to that which accounts for fibrillin-1/fibrillin-1 and Notch/Delta intermolecular interactions. Heterotypic interaction involving the binding of an EGF-like module to a non-EGFlike domain is similar to that which accounts for laminin/ nidogen (25), urokinase-type plasminogen activator/60-kDa urokinase receptor (26), thrombomodulin/thrombin (27), and factor IX/factor VIIIa intermolecular binding interactions. The fact that fibulin-1 uses EGF-like modules 5 and 6 to mediate both homotypic and heterotypic binding is a novel functional property, but it may turn out to be common to EGF-like modules in other proteins.
While a consequence of fibulin-1 binding to FN is that fibulin-1 becomes incorporated into FN-containing matrix fibers (3,19), the significance of fibulin-1/fibulin-1 interaction is unknown. In experiments in which fibulin-1 was added to or recombinantly expressed in cells that are incapable of assembling a FN matrix, fibulin-1 was not found to become incorporated into matrix fibers (3,19). However, with cells that are actively engaged in FN fiber assembly, fibulin-1 was found to incorporate into FN-containing matrix fibers. These data suggested that fibulin-1 associates with FN fibers, but, in the absence of FN, may not become assembled into fibrous fibulin-1 homopolymers. Rotary shadowing of fibulin-1 preparations shows dumbbell-shaped molecules (38) that fit a model of globular amino-and carboxyl-terminal domains separated by an extended stretch of repeated EGF-like modules. Whether these dumbbell-shaped molecules are monomers or dimers remains to be established. It is possible that the centrally located EGFlike modules 5 and 6 could mediate parallel or antiparallel alignments of fibulin-1 monomers that could produce dumbbell-shaped dimers. In sedimentation equilibrium experiments performed under physiological ionic strength conditions (phosphate-buffered saline), we determined a molecular mass of 150 kDa for fibulin-1, whereas under dissociative conditions (6 M guanidine HCl), a value of 60 kDa was derived. 2 Using laser desorption mass spectroscopy, the conditions of which are generally considered unfavorable for preservation of noncovalent intermolecular interactions, a value of 78,842 Da was determined for the molecular mass of fibulin-1 monomer (8). Taken together, it seems apparent that under physiological conditions, fibulin-1 can form noncovalently associated dimers. Whether fibulin-1 dimers can align with one another through lateral association of EGF-like modules 5 and 6 to form arrays of fibulin-1 and whether such arrays would have physiological relevance are not known.
In addition to EGF-like modules 5 and 6, another fibulin-1 self-association site was localized within the amino-terminal region. Unlike the site in EGF-like modules 5 and 6, the activity of the amino-terminal site was masked in the native protein and required detergent denaturation for it to be revealed. This site can therefore be considered a cryptic self-association site. A cryptic self-association site has been identified within the first type III module (III-1) of FN (39 -41). It has been proposed that unfolding of III-1 is a prerequisite to the polymerization of FN. Similarly, the amino-terminal cryptic site of fibulin-1 may be important for assembly of fibulin-1 into polymers. Rotary shadowing of fibulin-1 preparations has revealed the presence of spider-shaped multimers of fibulin-1 (38). These eight to ninelegged spider-shaped structures appear to contain the previously described fibulin-1 dumbbells, centrally interconnected through terminal globular domains. It is tempting to speculate that the self-association site contained within the amino-terminal region of fibulin-1 mediates the interaction involved in the formation of the spider-shaped multimers. In the case of FN, exposure of the III-1 cryptic site to initiate polymerization requires a cell-dependent action, perhaps involving integrin ␣ 5 ␤ 1 (39). The physiological mechanism that would lead to unfolding of the amino-terminal cryptic self-association site of fibulin-1 remains to be established.