Homo- and heterotypic fibrillin-1 and -2 interactions constitute the basis for the assembly of microfibrils.

Fibrillin-1 and fibrillin-2 constitute the backbone of extracellular filaments, called microfibrils. Fibrillin assembly involves complex multistep mechanisms to result in a periodical head-to-tail alignment in microfibrils. Impaired assembly potentially plays a role in the molecular pathogenesis of genetic disorders caused by mutations in fibrillin-1 (Marfan syndrome) and fibrillin-2 (congenital contractural arachnodactyly). Presently, the basic molecular interactions involved in fibrillin assembly are obscure. Here, we have generated recombinant full-length human fibrillin-1, and two overlapping recombinant polypeptides spanning the entire human fibrillin-2 in a mammalian expression system. Characterization by gel electrophoresis, electron microscopy after rotary shadowing, and reactivity with antibodies demonstrated correct folding of these recombinant polypeptides. Analyses of homotypic and heterotypic interaction repertoires showed N- to C-terminal binding of fibrillin-1, and of fibrillin-1 with fibrillin-2. The interactions were of high affinity with dissociation constants in the low nanomolar range. However, the N- and C-terminal fibrillin-2 polypeptides did not interact with each other. These results demonstrate that fibrillins can directly interact in an N- to C-terminal fashion to form homotypic fibrillin-1 or heterotypic fibrillin-1/fibrillin-2 microfibrils. This conclusion was further strengthened by double immunofluorescence labeling of microfibrils. In addition, the binding epitopes as well as the entire fibrillin molecules displayed very stable properties.

Microfibrils are extracellular supramolecular aggregates found in many elastic and non-elastic tissues (1). Ultrastructurally, they appear as beaded filaments with a periodicity of 50 -55 nm (2). The backbone of microfibrils are constituted by fibrillins, a family of large extended proteins (3,4). Other proteins such as microfibril-associated glycoprotein (MAGP) 1 -1 and -2 (5, 6), fibulin-2 (7), versican (8), and latent transforming growth factor ␤-binding protein (LTBP)-1 and -2 (9, 10) were found associated with microfibrils. Although it is clear that one of the basic functions of fibrillins is the formation of the microfibrillar backbone through a complex multistep assembly mechanism, the functional importance for the associated proteins is presently obscure.
Fibrillins consist of characteristic extracellular repetitive domains such as the calcium-binding epidermal growth factorlike domains (cbEGF) also found in many other extracellular proteins, and the 8-cysteine-containing domains (8-CYS) found exclusively in fibrillins and LTBPs (4,(11)(12)(13). Fibrillin-1 and fibrillin-2 share 100% homology on the domain level and ϳ68% homology on the overall amino acid residue level. Based on this homology, one would predict a similar mechanistic basis and architecture for the supramolecular assembly of both fibrillins. Whether fibrillin-1 and fibrillin-2 have intrinsic properties to self-assemble into homotypic or heterotypic microfibrils or both is not clear at present. Based on developmental studies, it is intelligible that adult tissues chiefly contain homotypic fibrillin-1 microfibrils, because fibrillin-2 generally is expressed early during mammalian embryogenesis and tends to disappear later (14). However, during development fibrillin-1 and fibrillin-2 often coincide in many tissues such as skin, lung, heart, aorta, central nervous system anlage, but are individually expressed in certain regions of kidney, liver, rib anlagen, and elastic cartilage (4,15). In situations where both fibrillins are expressed simultaneously, it is theoretically possible that each fibrillin isoform forms separate homotypic fibrils or both isoforms together form heterotypic microfibrils. Different organization of fibrillins in tissues have different functional consequences that may be relevant in pathological situations.
Mutations in the genes for fibrillin-1 (FBN1) on chromosome 15 and fibrillin-2 (FBN2) on chromosome 5 are responsible for the genetic disorders Marfan syndrome (MFS, MIM no. 154700) and congenital contractural arachnodactyly (MIM no. 121050), respectively (for recent reviews, see Refs. 16 and 17). Although these autosomal dominant disorders share some skeletal complications such as arachnodactyly, scoliosis, and chest deformities, the cardiovascular and ocular features characteristic for MFS are typically absent in congenital contractural arachnodactyly. How mutant fibrillin-1 and -2 molecules exert dominant negative effects in these disorders is presently unknown. Certain lines of evidence point to the possibility that fibrillin assembly mechanisms are compromised. For instance, pulse-chase and immunofluorescence experiments have revealed reduced amounts of mutant fibrillin-1 deposited into the extracellular matrix of many fibroblast strains obtained from individuals with MFS (18 -21). These data suggest that many mutations in fibrillin-1 impair the ability to assemble into microfibrils. Similar analyses of fibrillin-2 mutations in cell culture are presently lacking.
Aggregation of small N-terminal regions of fibrillins in recombinant systems has suggested that homotypic dimerization is involved in the multistep assembly mechanism of fibrillins (22)(23)(24). The static organization of fibrillin-1 in microfibrils has been examined by several groups and various techniques. Labeling of microfibrils with specific antibodies, high resolution structure of cbEGF modules, and analysis of transglutaminase cross-links have led to various models of fibrillin alignment in microfibrils (25)(26)(27)(28)(29). Despite the unresolved controversy whether fibrillin molecules are arranged in a nonstaggered or in a staggered fashion, common to all models is a head-to-tail arrangement of fibrillin molecules originally proposed by Sakai and co-workers in 1991 (25). Mapping of monoclonal antibody epitopes in fibrillin-1 molecules and correlation with the epitopes in microfibrils revealed that the N-and the C-terminal ends of the fibrillin molecules are located in or close to the bead structures (25,26). However, it is not clear whether microfibrillar backbone formation requires fibrillin-1 and fibrillin-2 to directly interact with themselves and with each other, or requires adapter molecules to connect fibrillin molecules in a head-to-tail fashion.
Here, we have analyzed in detail the spectrum of homotypic and heterotypic molecular interactions of fibrillin-1 and fibrillin-2. The results demonstrate direct head-to-tail interactions of fibrillin-1 alone, and of fibrillin-1 with fibrillin-2. However, fibrillin-2 alone was not able to self-interact in a N-to Cterminal fashion, indicating a different assembly mechanism for fibrillin-2. The results presented suggest that fibrillins are able to form microfibrils in a direct head-to-tail fashion without the aid of adapter molecules.

Construction of Expression Plasmids
Fibrillin-2 Constructs-To construct an expression plasmid for the C-terminal half of human fibrillin-2 (rFBN2-C, positions 1531-2771), the NheI-NotI 9028-bp fragment from pBS-rFBN2full was subcloned into the NheI-NotI-restricted expression vector pcDNA3.1 (Invitrogen), resulting in plasmid pcDNA-rFBN2full. To add a sequence for a 6-histidine tag and a stop codon at the 3Ј end, template rFBN2-1g was amplified by PCR using oligonucleotides 7s (5Ј-TTTTGGGTCCTAT-GAATGCACG-3Ј) and 14as (5Ј-ATCCGAATCAGCGGCCGCTCAC-TAGTGATGGTGATGGTGATGCTTAGAATAGCCGTTGATTTTGC-3Ј). The 1483-bp Bsu36I-NotI fragment of the resulting PCR product was ligated into the Bsu36I-NotI-restricted pcDNA-rFBN2full, resulting in plasmid pcDNA-rFBN2full-his. The 6-histidine tag and the stop codon replaced the sequence for part of the C-terminal unique domain from position 2772 to 2911. To modify the 5Ј end of the construct, template rFBN2-4a was amplified by PCR using oligonucleotides 15s (5Ј-CGT-AGCTAGCAGATATTGATGAGTGTGCAGATCC-3Ј) and 15as (5Ј-CTG-CTACTCGAGAGCGGCCGCTCACTAGTGATGGTGATGGTGATGCA-TGCAGTTGTGGCCTCCATTGACC-3Ј). A 460-bp BbvCI-NheI fragment of the PCR product was ligated into the BbvCI-NheI-restricted pcDNA-rFBN2full-his, resulting in plasmid pcDNA-rFBN2-C-his. The XbaI-NheI-restricted sequence from plasmid pBS-BM40 coding for the signal peptide of the human BM40 protein, which confers efficient expression and secretion of the expressed protein, was fused via the NheI site at the 5Ј site of the construct. The resulting plasmid was designated pDNSP-rFBN2-C. Expression and processing resulted in the secretion of a protein (rFBN2-C) with additional 4 N-terminal and 6 C-terminal amino acid residues preceding and following the authentic domains (Ala-Pro-Leu-Ala-Asp 1531 -Lys 2771 -(His) 6 ).
To prepare the plasmid coding for the N-terminal half of human fibrillin-2 (rFBN2-N, position 1-1732), the PCR product described above for the C-terminal construct was restricted by XhoI (479 bp) and inserted into the XhoI-restricted plasmid pcDNA-rFBN2full-his, resulting in pcDNA-rFBN2-N. Expression of this plasmid resulted in a protein (rFBN2-N) that includes the sequence for the authentic fibrillin-2 signal peptide and a 6-histidine tag at the C-terminal end (Met 1 -Met 1732 -(His) 6 ). The most likely prediction for the cleavage site of the signal peptide in fibrillin-2 is between position 28 and 29 (Thr-Ala-Gly 28 2Gln 29 -Pro) (32). Therefore, the secreted protein likely starts at position 29. Correct ligation of all constructs have been verified by DNA sequencing.
Fibrillin-1 Constructs-The expression plasmids to express the Nterminal half (pDNSP-rF16) and the C-terminal half (pcDNA-rF6H) of human fibrillin-1 have been described in detail previously (33). Both plasmids have been utilized to construct an expression vector for fulllength human fibrillin-1 as follows. A 3869-bp NotI-ApaI fragment was excised from plasmid pcDNA-rF6H and subcloned into NotI-ApaI-restricted pDNSP-rF16. A 320-bp duplicated fragment was removed by restriction with PmlI, followed by religation of the plasmid. The resulting plasmid was designated pcDNA-rFBN1 and produces a secreted polypeptide (rFBN1) with the sequence Ala-Pro-Leu-Ala-Ser 19 -Lys 2725 -(His) 6 . The correct DNA sequences at the restriction sites utilized have been verified by DNA sequencing.

Transfection of Cells and Identification of Recombinant Clones
For stable expression, the expression plasmids were linearized by PvuI restriction and transfected into human embryonic kidney cell 293 (American Type Culture Collection) using established methods (34). 10 h after transfection, selection was started with 500 g/ml G418 (Calbiochem) and continued for 1 week. Thereafter, the concentration of G418 was reduced to 250 g/ml. After ϳ4 weeks, individual clones were transferred into 24-well plates and propagated to confluence. After washing with phosphate-buffered saline (PBS), the wells were incubated with serum-free culture medium (1 ml) for 2 days. Identification of clones expressing the recombinant proteins was achieved by conventional gel electrophoresis, followed by Coomassie Blue staining and Western blotting with specific monoclonal antibodies against the expressed proteins. Typically, 50 -80% of the analyzed clones stably expressed the recombinant proteins.

Production of Recombinant Proteins
Stably transfected cell clones were routinely cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, and 10% (v/v) fetal calf serum at 37°C in a 5% CO 2 atmosphere. For large scale production, the clones were propagated in triple layered flasks (Nalge Nunc International) to confluence; washed twice with 20 mM HEPES, pH 7.4, 150 mM NaCl, 2.5 mM CaCl 2 ; and incubated with serum-free culture medium for 48 h. The conditioned medium (2-2.5 liters) was harvested, supplemented with 1 mM phenylmethylsulfonyl fluoride (Fluka), and concentrated to ϳ50 ml by ultrafiltration (YM30 mem-branes, Millipore). After dialysis against 20 mM phosphate, pH 7.2, 1 M NaCl (equilibration buffer), the medium was passed over a High-Trap chelating column loaded with Co 2ϩ (1-ml column size, Amersham Biosciences) equilibrated in the same buffer. After washing the column with equilibration buffer, bound proteins were eluted with a linear imidazole gradient in equilibration buffer (0 -250 mM imidazole in 30 ml) and fractioned in 1-ml aliquots. The fractions were analyzed by gel electrophoresis followed by Coomassie Blue staining and by Western blot analyses. Fractions containing the recombinant proteins were pooled and dialyzed against 50 mM Tris-HCl, pH 7.4, 150 mM NaCl (TBS).

Quantification of Protein Concentrations
Aliquots (50 l) were supplemented with 450 l of 6.67 M guanidine HCl in TBS and incubated at room temperature for 30 min. The absorbance at 280 nm was determined on a Ultrospec 3000 spectrophotometer (Amersham Biosciences). Calculation of the molar extinction coefficient followed an established method (⑀ ϭ n Trp ϫ 5500 ϩ n Tyr ϫ 1490 ϩ n Cys-S-S-Cys ϫ 125 [M Ϫ1 ϫ cm Ϫ1 ]) (35).
Alternatively, protein concentrations were determined by amino acid analyses. The proteins were hydrolyzed in 6 N HCl under N 2 for 24 h at 110°C, and the amino acid composition was determined on a Biochrom 20 analyzer (Biochrom).

Circular Dichroism Measurements
The purified recombinant proteins in TBS were diluted to a concentration of 0.5 mg/ml. Spectra from 190 to 260 nm were recorded at 20 and 100°C in a 1-mm quartz cuvette on a Jasco J-715 instrument.

Electron Microscopy
Purified recombinant proteins were dialyzed against 100 mM NH 4 HCO 3 and adjusted to concentrations of 0.25 mg/ml. The samples were diluted with 0.05% (v/v) acetic acid to a final concentration of 60 g/ml and mixed with glycerol to a final concentration of 50% (v/v) glycerol. The samples were sprayed onto freshly cleaved mica from a distance of 25 cm and dried under vacuum for 2-3 h in an Edwards Auto 306 vacuum coater (Edwards). Rotary shadowing was performed by platinum evaporation for 15 s, 50 mA, 2.5 kV at an angle of 5°and a distance of 12 cm, followed by carbon evaporation for 2 s, 100 mA, 2.5 kV at an angle of 90°. The replicas were floated onto a very clean surface of distilled water and then supported using 400-mesh copper grids. Replicas were examined at 100 kV in a transmission electron microscope (Zeiss TEM 109).

Antibodies
Polyclonal antisera were produced in rabbits commercially (Biotrend, Cologne, Germany) against the recombinant polypeptides rFBN2-N, rFBN2-C, and an ϳ110-kDa N-terminal fragment of human fibrillin-2, rF37, described previously (30). The antisera were characterized by a standard ELISA technique. Generation and properties of anti-human fibrillin-1 polyclonal antisera B9543 and ␣-rF6H, as well as the specificity of monoclonal anti-human fibrillin-2 antibody mAb 48 have been described elsewhere (36 -38). B9543 and mAb 48 were generous gifts from Dr. Lynn Y. Sakai (Shriners Hospital for Children, Portland, OR).

Protein Interaction Assays
Solid Phase Binding Assay-Multiwell plates (96 wells, MaxiSorp, Nalge Nunc International) were coated overnight with purified recombinant proteins (10 g/ml, 100 l/well) in TBS at 4°C. Nonspecific binding sites were blocked for 1 h with 5% (w/v) nonfat milk in TBS. Each of the following incubations was performed in TBS/5% nonfat milk including either 5 mM CaCl 2 or 10 mM EDTA at room temperature (ϳ 20°C) and was followed by three washes with TBS including 0.05% (v/v) Tween 20. Coated proteins were incubated with serial dilutions (1:3) of the soluble ligands (0.14 -100 g/ml) for 2 h. Incubation (1.5 h) with the primary polyclonal antibodies (1:200 -1:1250 diluted) against the soluble ligands was followed by incubation with horseradish peroxidaseconjugated secondary goat anti-rabbit antibody (1:800 diluted; Bio-Rad) for 1.5 h. Color development was performed with 1 mg/ml 5-aminosalicylic acid (Sigma) in 20 mM phosphate buffer, pH 6.8, including 0.1% H 2 O 2 for 3-5 min and stopped by adding 2 M NaOH. Color yields were determined at 492 nm using a Microplate EL310 (Bio-Tek Instruments).
To determine the stability of fibrillin interaction epitopes, the immobilized or the soluble binding ligands were heat-treated in discrete steps between 20 and 95°C for 10 min and then used in the above described solid phase binding assay.
Kinetic Binding Studies-For kinetic binding studies of recombinant fibrillin-1 and fibrillin-2 fragments by surface plasmon resonance, a Biacore biosensor was used (Biacore 3000; Biacore AB). Purified recombinant fibrillin-1 and fibrillin-2 fragments were biotinylated using activated NHS-LC-biotin as instructed by the supplier (Pierce). Biotinylated fragments were coupled in TBS to a streptavidin sensor chip SA (Pioneer), which resulted in 1000 -2500 response units. Binding studies were performed with soluble recombinant fibrillin-1 or fibrillin-2 fragments in concentrations of 12.5-1000 nM in TBS at flow rates of 10 l/min. The binding sites of the immobilized ligands were regenerated by injection of a mixture of detergents (0.075% (v/v) each of CHAPS, Zwittergent 3-12, Tween 80, Tween 20, Triton X-100) after each cycle. After subtraction of the blank curves, representing binding to bovine serum albumin, the association and dissociation rate constants were determined by separate k a /k d fitting all curves at once with the 1:1 Langmuir association/dissociation model (BIAevaluation software version 3.0, Biacore AB). Although this model produced the best fits, the observed interactions probably diverge somewhat from the 1:1 binding model because k a values slightly decreased with increasing concentrations of the soluble ligands. Mass transfer limitations were not apparent.

Immunofluorescence
Primary dermal fibroblasts from a 1-year-old individual and primary osteoblasts from a 42-year-old individual were cultivated in the first (fibroblasts) or fourth (osteoblasts) passage using culture conditions as described above for the recombinant cell clones. Confluent cells were trypsinized and seeded at densities of 7.5 ϫ 10 4 cells/well of a 8-well chamber slide (Permanox; Nalge Nunc International). After 4 days, the cells were washed in PBS, and fixed in 70% (v/v) methanol, 30% (v/v) acetone for 5 min, and rehydrated in PBS. The nonspecific binding sites were blocked with 1:10 diluted normal goat serum for 30 min. The cells were incubated with monoclonal anti-fibrillin-2 antibody mAb 48 (1: 200) and the polyclonal anti-fibrillin-1 ␣-rF6H (1:400) in PBS for 1 h, followed by three washes with PBS. After incubation with 1:200 diluted goat anti-rabbit fluorescein conjugate (Jackson ImmunoResearch) and Cy3-conjugated Affinipure goat anti-mouse IgG (Jackson ImmunoResearch), the fibrillin-1 and fibrillin-2 networks were visualized by fluorescence microscopy with a Axioplan microscope (Zeiss). Digital images were recorded using a 3CCD color video camera (Sony) and AxioVision software version 3.0 (Zeiss).

RESULTS
To generate recombinant human fibrillin-2 fragments, the human fibrillin-2 cDNA was cloned from MG-63 cells. The cloned sequence was compared with the published sequence for human fibrillin-2 cDNA, and 19 differences have been identified (Table I, GenBank TM accession no. NM_001999; Refs. 4 and 31). Based on sequence homology of the cloned and published sequence for human fibrillin-2 cDNA with the sequences for human fibrillin-1 (GenBank TM accession no. NM_000138; Refs. 11-13) and human fibrillin-3 (GenBank TM accession no. AB053450; Ref. 39), we conclude that all of the observed differences are correct in the sequence presented here. In some instances they likely represent polymorphisms ( Table I). Some of the observed variations have been reported previously (30,40).
To analyze the mechanisms of how fibrillins assemble into higher ordered structures, we have generated new recombinant fragments of fibrillin-2 as well as a recombinant full-length fibrillin-1 polypeptide. For fibrillin-2, we produced the N-terminal half rFBN2-N (position 1-1732) and the C-terminal half rFBN2-C (position 1531-2771) in human 293 cells (Fig. 1). rFBN2-N and rFBN2-C span the entire fibrillin-2 amino acid sequence except part of the C-terminal unique domain (position 2772-2911), which in analogy to fibrillin-1 processing presumably is cleaved by furin-type proteases (41,42). Similarly, the recombinant fibrillin-1 full-length construct rFBN1 comprises the entire fibrillin-1 sequence except the processed C-terminal domain (position 2726 -2871; Fig. 1). Additionally, two recombinant halves of fibrillin-1, rF16 and rF6H, described previously were used in this study (33). For clarity, these fragments have been renamed for this study to rFBN1-N and rFBN1-C (Fig. 1).
The recombinant fragments rFBN2-N and rFBN2-C were synthesized and secreted into the culture medium of recombinant cell clones at concentrations of ϳ5 g/ml/day. Full-length rFBN1 was produced in significantly lower amounts of ϳ0.5 g/ml/day. All recombinant polypeptides were purified to homogeneity by chelating chromatography (Fig. 2). The molecular masses for rFBN2-N (ϳ210 kDa nonreduced and ϳ225 kDa reduced), rFBN2-C (ϳ175 kDa nonreduced and ϳ185 kDa reduced), and rFBN1 (ϳ322 kDa nonreduced and ϳ345 kDa reduced) corresponded well with the expected masses for these polypeptides. Freshly prepared and purified rFBN1 resulted in single bands in Coomassie-stained gels (Fig. 2). However, rFBN1 tended to precipitate in solution after a few days, or upon repeated freezing/thawing cycles (data not shown). The recombinant halves of fibrillin-1 and fibrillin-2 were soluble and did not precipitate in solution. All recombinant polypeptides reacted in Western blot analyses with specific mono-and polyclonal antibodies, which are dependent on correct disulfide bonds, indicating correct three-dimensional structures of the polypeptides (data not shown).
The new recombinant polypeptides were visualized by electron microscopy after rotary shadowing. rFBN2-N (Fig. 3A) and rFBN2-C (Fig. 3B) showed extended shapes similar to what was observed for the corresponding recombinant fibrillin-1 polypeptides (26). Occasionally, kinks and bends have been observed in both molecules. The length of the molecules were 74.9 Ϯ 4.1 nm (rFBN2-N, n ϭ 62) and 68.2 Ϯ 4.8 nm (rFBN2-C, n ϭ 30). The shape of the rFBN1 molecules was also threadlike and extended (Fig. 4). Again, kinks and bends could be observed within the molecules (Fig. 4). Length measurements of these molecules showed that 43.1% were in the range of monomers between 100 and 180 nm (139 Ϯ 24 nm), 13.8% in the range of dimers between 240 and 320 nm (274 Ϯ 21 nm), and 3.4% in the range of trimers between 380 and 460 nm (416 Ϯ 27 nm). The remaining particles likely represented proteolytically truncated products of monomers, dimers, and trimers. These The sequence used for comparison were GenBank NM_001999 (human fibrillin-2). For homology comparisons of amino acid sequences, GenBank NM_000138 (human fibrillin-1) and AB053450 (human fibrillin-3) have been used (shaded letters). data suggest that monomeric fibrillin-1 can associate to multimers in TBS buffer. Interestingly, the molecules appeared to connect at their ends with each other, because no significant overlap between two molecules have been observed in dimers and trimers (Fig. 4). Occasionally, kinks or globules were detected in the region where two molecules are in contact with each other. These results suggested that fibrillin-1 molecules can interact with each other without the support of other molecules.
To further analyze the self-interaction properties of fibrillin-1 and fibrillin-2, as well as the ability of fibrillin-1 to interact with fibrillin-2, binding activities of various combinations of the recombinant halves of fibrillin-1 and fibrillin-2 have been analyzed by solid phase binding assays (Fig. 5). Strong selfinteractions were observed between the N-and the C-terminal halves of fibrillin-1 (Fig. 5A). Surprisingly, the corresponding constructs of fibrillin-2 did not show significant self-interaction properties (Fig. 5B). Fibrillin-1 also interacted strongly with fibrillin-2. The N-terminal half of fibrillin-1 clearly showed dose-dependent binding to the C-terminal half of fibrillin-2 (Fig. 5C), and the N-terminal half of fibrillin-2 interacted with the C-terminal half of fibrillin-1 (Fig. 5D). All of the observed interactions were dependent on calcium (Fig. 5).
The association (k a ) and dissociation rate (k d ) constants of the homotypic and heterotypic fibrillin binding interactions, and the dissociation constants (K D ), have been determined by surface plasmon resonance (Fig. 6). The kinetic data obtained from such experiments are summarized in Table II. In these experiments, high affinity self-interaction of fibrillin-1 has been observed between the rFBN1-N and -C ( Fig. 6A; K D ϭ 3.8 -25.6 nM), whereas the corresponding fragments of fibrillin-2 did not significantly interact (Fig. 6B). Heterotypic interaction between fibrillin-1 and fibrillin-2 was observed for combinations of rFBN1-N with rFBN2-C ( Fig. 6C; K D ϭ 4.3-20.5 nM), and rFBN2-N with rFBN1-C ( Fig. 6D; K D ϭ 23.1-82.0 nM). These results correlated well with the data obtained by solid phase binding assays and demonstrated that the homotypic interaction of fibrillin-1 and the heterotypic interaction of fibrillin-1 with fibrillin-2 are of high affinity.
The stabilities of the interacting epitopes have been studied by heat inactivation experiments (Fig. 7). The recombinant fibrillin polypeptides were incubated at increasing tempera- tures followed by analyses of the homo-and heterotypic interactions in solid phase binding assays. Most binding epitopes were inactivated only when the fragments were treated with high temperatures above 80°C. In these cases, the inactivation temperatures resulting in 50% inhibition of binding were in the range between 83 and 87°C. Only when the C-terminal fragments of fibrillin-1 or fibrillin-2 were heat-treated and used as soluble ligands with immobilized rFBN1-N did the inactivation temperatures resulting in 50% binding inhibition decrease somewhat to 68°C (rFBN1-N/heat-treated rFBN2-C) and 77°C (rFBN1-N/heat-treated rFBN1-C). These data demonstrate that the binding epitopes for homo-and heterotypic fibrillin N-to-C interactions represent very stable regions.
These results prompted us to analyze the stability of the entire recombinant fragments by circular dichroism spectra. The spectra of the recombinant polypeptides were recorded at low temperatures (20°C) and compared with spectra recorded at high temperatures (100°C) (Fig. 8). Spectra of all fragments at low temperature showed molar ellipticities similar to what has been observed for shorter fragments of fibrillin-1 (43). These spectra are characteristic of large amounts of ␤-structures. The spectra of the polypeptides recorded at 100°C only marginally differed from the spectra recorded at 20°C. These results demonstrate that the polypeptide chains maintained much of their secondary structures even at 100°C. In addition, denaturation experiments using guanidinium hydrochloride demonstrated similar stabilities of the recombinant proteins (data not shown). When the binding inactivation experiments and the analyses by circular dichroism spectra at different temperatures are taken together, we conclude that fibrillins are very stable proteins.
The ability of fibrillin-1 and fibrillin-2 to interact in microfibrils has been examined by double-immunofluorescence experiments (Fig. 9). First, the polyclonal anti-fibrillin-1 antiserum (␣-rF6H) was tested for potential cross-reactivity with fibrillin-2 by ELISA (Fig. 9A). Only minor cross-reactivity of ␣-rF6H raised against the C-terminal half of human fibrillin-1 was observed with the C-terminal half of human fibrillin-2 at high concentrations (1:50 -1:200 dilutions). This cross-reactivity was negligible at dilutions of 1:400 and higher. Even less crossreactivity of the ␣-rF6H antiserum was observed with the N-terminal halves of human fibrillin-2 and -1 (Fig. 9A). The specificity of the monoclonal mAb 48 antibody exclusively for fibrillin-2 has been reported elsewhere (38). In dermal fibroblasts from a 1-year-old donor, fibrillin-1 (Fig. 9B, green signal) and fibrillin-2 (Fig. 9C, red signal) were detectable in a microfibrillar network. Superimposition of both fluorescence signals clearly demonstrated that both fibrillins were localized to the same microfibrils (Fig. 9D). On the other hand, osteoblasts from a 42-year-old donor showed a fibrillin-1 microfibrillar network (Fig. 9E), but fibrillin-2 could not be detected (Fig. 9F). Consequently, superimposition of both signals only resulted in a green signal exclusively representing the presence of fibrillin-1 (Fig. 9G). These results demonstrate that fibrillin-1 alone can form microfibrillar structures, and that fibrillin-1 and fibrillin-2 can occur in the same microfibrils. These findings  6. Quantification of homo-and heterotypic fibrillin-1 and -2 binding strengths by surface plasmon resonance. Association and dissociation curves for the homotypic N-to C-terminal interaction of fibrillin-1 (A), and fibrillin-2 (B), and of N-to C-terminal heterotypic interactions between fibrillin-1 and fibrillin-2 (C and D) have been analyzed by real time kinetic studies using a Biacore 3000 instrument. The reaction partners were rFBN1-N and rFBN1-C (A), rFBN2-N and rFBN2-C (B), rFBN1-N and rFBN2-C (C), and rFBN2-N and rFBN1-C (D). Association time was 300 s, and dissociation time was 500 s. Concentrations of soluble ligands used are indicated in nM. RU, response units.
clearly correspond with the in vitro binding interaction of fibrillins described above. DISCUSSION Microfibrils 10 -12 nm in diameter are supramolecular aggregates in the extracellular matrix consisting of fibrillins and other matrix proteins. Fibrillins are repetitively aligned within microfibrils and constitute their structural backbone. At present, the mechanistic basis for formation of the fibrillin back-bone is unknown, as is whether fibrillins exclusively form homotypic microfibrils consisting of only one isoform or heterotypic microfibrils containing both isoforms. Because impaired assembly mechanisms may precipitate the pathogenetic pathways of genetic disorders caused by mutations in fibrillin-1 and fibrillin-2, it is important to answer these questions. In this study, we have analyzed in detail homotypic and heterotypic interaction repertoires and stabilities of fibrillin-1 and fibrillin-2.
Fibrillins cannot be extracted from tissues in their native form because they are heavily cross-linked by reducible and nonreducible cross-links (28,44). A feasible alternative to obtain fibrillin for mechanistic studies is its recombinant expression in mammalian cells. Previously, we have recombinantly produced two halves of fibrillin-1 and demonstrated correct folding and functional properties (7,26). Here, we have produced two corresponding halves of fibrillin-2 spanning the entire processed fibrillin-2 molecule. These constructs resembled the corresponding counterparts of fibrillin-1, as judged by gel electrophoretic analysis and electron microscopy after rotary shadowing. The extended shapes together with the reactivity with antibodies requiring native epitopes, stabilized by disul-  7. Temperature stability of homo-and heterotypic fibrillin interaction epitopes. Fibrillin fragments have been incubated for 10 min at the temperatures indicated, and then used in homo-and heterotypic fibrillin solid phase interaction assays. Heat-treated or nontreated immobilized rFBN1-N (A) or rFBN2-N (B) were analyzed with soluble heat-treated or nontreated rFBN1-C or rFBN2-C. The combination of ligands for each analysis are indicated on top with shaded areas representing heat-treated proteins. The soluble ligands were used in serial dilutions (1:3) starting at 100 g/ml. The entire binding curves for each temperature were numerically integrated using the trapezoidal rule, and the numerical integral for each analysis at 20°C was set to 100% binding. fide bonds, suggested correct folding of these recombinant polypeptides. A similar rationale applies to rFBN1, a new recombinantly expressed full-length fibrillin-1. We anticipate that this full-length construct will be also, beyond the scope of this study, a useful tool for the analysis of fibrillin assembly mechanisms. rFBN1 as well as the recombinant halves of fibrillin-2 showed several bends and sharp kinks within the molecules, indicating the potential to fold back on themselves. In rFBN1, for instance, kinks were obvious close to one end and in the center of the recombinant molecules. Potentially, these sites correspond to regions containing the proline-rich domain and the third or fourth 8-CYS domain. A folding mechanism in these regions has been hypothesized to facilitate condensation of the fibrillin molecules within a microfibril (26,29).
It has previously been established that fibrillins are oriented in microfibrils in a head-to-tail fashion and that the N-and C-terminal ends of the molecules are located either in or very close to the bead structures (25,26). Theoretically, the connections between fibrillin molecules could be mediated by adapter molecules or, alternatively, by direct fibrillin interactions in a head-to-tail manner. It has been observed by immunogold localization that MAGP-1 is located at or close to the bead structures, and thus it potentially could function as a fibrillin bridging molecule in microfibrils (5). However, it has been demonstrated that MAGP-1 does only interact with fibrillin-1 via an epitope located at the N-terminal region of fibrillin-1, whereas the C-terminal region of fibrillin-1 does not harbor a binding site for MAGP-1 (33). Therefore, MAGP-1 alone cannot function as an adapter to connect fibrillin molecules in the bead regions of microfibrils. Other fibrillin-binding ligands such as fibulin-2 (7), LTBP-1 (45), or versican (8) have been described, but they are present only on some but not all microfibrils and they are typically not periodically aligned along microfibrils.
In this study, we have found for fibrillin-1 very strong interactions between the N-terminal and the C-terminal halves with dissociation constants in the low nanomolar range. Surprisingly, for fibrillin-2, this type of interaction could not be observed, although various protein interaction assays have been employed. Self-interaction properties in a head-to-tail fashion were also observed for the recombinant full-length fibrillin-1 polypeptide, which represented the C-terminally processed form of fibrillin-1, because it lacks the small C-terminal portion usually processed by furin-type proteases (41,42). This polypeptide clearly showed the tendency to precipitate in physiological buffer solutions, indicating self-interaction properties. When rFBN1 molecules were analyzed ultrastructurally by electron microscopy after rotary shadowing, monomers, dimers, trimers, and, very occasionally, tetramers were observed. Because the analyzed samples originated from the cleared supernatant of a rFBN1 preparation, we suspect that the precipitate of this preparation is formed by even higher molecular aggregates. The multimers of rFBN1 found among the population of molecules appeared often as continuous strings without obvious overlapping regions of the molecules (Fig. 4). In some instances globules or kinks indicated the end of one and the beginning of the next molecule. These results suggest that the interaction epitopes are located relatively close to the processed ends of the molecules. We have found that recombinantly expressed fibrillin proteins representing N-terminal parts of the molecules are not straight as suggested from the domain structure, but often display a curved or globular shape (data not shown). Thus, the globules occasionally found between two rFBN1 molecules potentially could repre- FIG. 10. Co-organization of fibrillins in microfibrils. Based on the results presented in this study, fibrillin-1 (red arrows) alone (A), or fibrillin-1 together with fibrillin-2 (green arrows) (B) can assemble via direct N-terminal (arrowhead) to C-terminal (arrowtail) interactions into microfibrils. For clarity, only the previously proposed parallel nonstaggered alignment model is shown (26). However, the interaction mechanisms described are also applicable to all other alignment models proposed (27)(28)(29). For further clarity, only 3 molecules are shown between each globular domain. They represent 6 -8 fibrillin monomers in cross-sectional diameter (29,46). Globular bead regions are shown as gray circles. In B it is not clear whether fibrillin-1 and fibrillin-2 are regularly alternating in the microfibrils or whether they are arranged in clusters. The ␣-rF6H antiserum was diluted in 1:2 serial dilutions. Note that cross-reactivities of the antiserum with fibrillin-2 are negligible at dilutions of 1:400 and higher. The specificity of mAb 48 for fibrillin-2 has been described elsewhere (38). Dermal fibroblasts from a 1-year-old donor (B-D) or osteoblasts from a 42-year-old donor (E-G) were simultaneously labeled with ␣-rF6H (green signal, B and E) and mAb 48 (red signal, C and F). Superimposition of both labels are shown in panels D and G. Note that the yellow signal in D indicates the presence of fibrillin-1 and fibrillin-2 within the same microfibrils. The bar represents 25 m. sent the N-terminal region. In this light, it is possible that the N-terminal interaction epitope is not located strictly at the N terminus of the fibrillin-1 molecule but somewhat further Cterminal, for instance, in the region of the proline-rich domain.
To further understand the mechanism of fibrillin-2 assembly, which appears not to involve homotypic self-interaction, we determined the ability of fibrillin-2 and fibrillin-1 to interact heterotypically with each other. Interestingly, we have found strong interaction properties between fibrillin-1 and fibrillin-2 with dissociation constants again in the low nanomolar range. The N-terminal half of fibrillin-2 interacted with the C-terminal half of fibrillin-1, and, vice versa, the C-terminal half of fibrillin-2 interacted with the N-terminal half of fibrillin-1. Because the recombinant halves of the fibrillins used in these assays are relatively large (185-225 kDa), we cannot draw a conclusion as to which domains are responsible for mediating the fibrillin homo-and heterotypic interactions. Interestingly, when smaller and well established recombinant fragments of fibrillin-1 were utilized to narrow down the binding epitopes for fibrillin-2, no binding activity could be observed (data not shown). One interpretation of such results is that the binding epitopes are stabilized by regions of the molecules that are not in the immediate vicinity of the binding epitope, e.g. by stabilizing long range structural effects. Previously, long range structural effects have been reported in fibrillin-1 (43). Another interpretation is that binding epitopes are located in regions at the ends of the recombinant subfragments used and thus are potentially truncated.
In addition, we have used a functional assay to assess the stability of the fibrillin interaction epitopes. Heat denaturation of recombinant proteins prior to the established interaction assay showed that the interaction epitopes are very stable regions of the molecules. Only treatment with high temperatures resulted in loss of interaction properties, e.g. in denaturation of the binding sites, an observation that corresponds well with the overall structural stability determined. The basis for this high temperature stability of fibrillin proteins very likely resides in the many stabilizing disulfide bonds. The cbEGF-like domain as well as the 8-CYS domains are stabilized by three and four disulfide bonds, respectively (26,27). This stability is probably a prerequisite to serve as components of systems underlying mechanical forces such as for instance elastic tissues.
In conclusion, the presented data suggest mechanisms for homo-and heterotypic fibrillin assembly, which are discussed in the following and schematically visualized in Fig. 10. The schematic representation in Fig. 10 is shown for the parallel nonstaggered model previously proposed (26), but the mechanisms are applicable to all other fibrillin alignment models (27)(28)(29). First, fibrillin-1 alone can homotypically form the backbone of a microfibril in the absence of adapter molecules and in the absence of fibrillin-2 (Fig. 10A). This type of mechanism could be further confirmed by the observation that osteoblasts from a 42-year old donor deposited a microfibrillar network consisting of fibrillin-1 in the absence fibrillin-2 ( Fig.  9). Generally in mammalian development, fibrillin-1 expression persists for longer periods of time, whereas fibrillin-2 tends to disappear earlier (14). Based on this observation, homotypic fibrillin-1 microfibrils are likely the predominant form in adult organisms. The data presented here provide an explanation of how fibrillin-1 can form microfibrils in the absence of fibrillin-2. Second, the data strongly suggest that fibrillin-1 and fibrillin-2 can co-polymerize in a head-to-tail fashion to form heterotypic microfibrils (Fig. 10B). Based on the high homology of the domain structures of fibrillin-1 and -2, it is conceivable that both fibrillins fit into the same geometric constraints of a single microfibril. This mechanism is supported by our observation that dermal fibroblasts from a 1-year-old donor deposited a microfibrillar network containing both fibrillin-1 and fibrillin-2 within the same microfibrils (Fig.  9). Ultrastructural immunogold co-localization of fibrillin-1 and -2 within the same microfibril is presented elsewhere by Charbonneau et al. (38). Early in mammalian development, fibrillin-1 and fibrillin-2 are co-expressed in most tissues (4,15). Based on the ability of fibrillin-1 and fibrillin-2 to interact with each other with very high affinity, we predict in situations where both fibrillins are expressed simultaneously, they will form heterotypic microfibrils. Third, the data presented also predict that fibrillin-2 alone, in the absence of fibrillin-1, cannot form homotypic microfibrils by corresponding N-to Cinteracting mechanisms. It is clear that fibrillin-2 is expressed in some tissues during development where fibrillin-1 is not expressed (4,15). In these situations, fibrillin-2 may assemble into homotypic microfibrils by alternative mechanisms. Such mechanisms may require molecular adapters, or a new member of the fibrillin family, fibrillin-3, may play a critical role (39). Alternatively, fibrillin-2 assembles into another, not yet recognized form of supramolecular aggregates.