Fibrillins constitute the backbone of extracellular multifunctional assemblies present in elastic and non-elastic matrices, termed microfibrils. Assembly of fibrillins into microfibrils and their homoeostasis is poorly understood and is often compromised in connective tissue disorders such as Marfan syndrome and other fibrillinopathies. Using interaction mapping studies, we demonstrate that fibrillins require the complete gelatin-binding region of fibronectin for interaction, which comprises domains FNI6–FNI9. However, the interaction of fibrillin-1 with the gelatin-binding domain of fibronectin is not involved in fibrillin-1 network assembly mediated by human skin fibroblasts. We show further that the fibronectin network is essential for microfibril homoeostasis in early stages. Fibronectin is present in extracted mature microfibrils from tissue and cells as well as in some in situ microfibrils observed at the ultrastructural level, indicating an extended mechanism for the involvement of fibronectin in microfibril assembly and maturation.
Fibrillin-1, fibrillin-2 and fibrillin-3 are extracellular matrix glycoproteins constituting the fibrillin protein family. Each member of the family has a typical modular organization primarily composed of calcium-binding epidermal growth factor-like domains, TGFβ (transforming growth factor-β)-binding domains and hybrid domains [1,2]. The three fibrillin isoforms are highly homologous with each other on the amino acid sequence (~60–70%). They differ in their spatio-temporal expression patterns with fibrillin-1 being expressed throughout life, whereas fibrillin-2 and fibrillin-3 are mainly developmentally expressed [3–5]. Fibrillins form the backbone of highly ordered extended structures termed microfibrils, which have a characteristic bead-on-a-string ultrastructure with a 50–55 nm periodicity after extraction from tissues or cell culture sources [6–8]. Microfibrils are found in many tissues including the cardiovascular system, bones, eyes, skin and other tissues, where they fulfil a wide range of physiological functions . They form the scaffold for elastic fibre assembly, act as stress-bearing entities (i.e. in ciliary zonules of the eye) [10–12], and serve as reservoirs and regulators for growth factors of the TGFβ superfamily [13–15]. Proper microfibril assembly and function is crucial as demonstrated by the wide range of clinical symptoms associated with fibrillinopathies [2,16]. Mutations in fibrillins can result in several connective tissue disorders including Marfan syndrome, stiff skin syndrome, autosomal dominant Weill–Marchesani syndrome, autosomal dominant geleophysic dysplasia and acromicric dysplasia, all caused by mutations in fibrillin-1, and congenital contractural arachnodactyly caused by fibrillin-2 mutations [17–21].
Fibronectin is an extracellular matrix protein which is involved in fundamental processes such as cell adhesion, migration and proliferation during development and physiological processes (for a review, see ). Like fibrillins, fibronectin is a modular protein, but with a different set of domains called type I, II and III repeats (FNI, FNII and FNIII). Fibronectin domains confer self-assembly and ligand-binding properties to a number of proteinaceous and non-proteinaceous ligands [22,23]. Two forms of fibronectin exist: cellular fibronectin is secreted by mesenchymal cells and assembled into an insoluble fibrillar network; and soluble plasma fibronectin is synthesized by hepatocytes and secreted into the blood where it polymerizes during blood clotting following vascular injury. Both forms are secreted as soluble dimers linked by two disulfide bonds [24,25]. On interaction with α5β1 integrin and other surface components, the protein is extended through forces originating from the intracellular contractile actin cytoskeleton . This extension causes the exposure of cryptic self-assembly sites which enables the fibronectin dimers to multimerize and to form an extracellular fibronectin network . Fibronectin network is formed and disassembled at the same rate as shown in vivo and in vitro. Therefore the fibronectin network is constitutively turned-over [28,29].
In the recent years, a number of studies provided evidence that fibronectin is a ‘master organizer’ of extracellular matrix assembly. Several studies demonstrated the requirement of a continuous polymerization and supply of fibronectin for the assembly of extracellular proteins such as collagen type I, thrombospondin-1, LTBP-1 (latent TGFβ-binding protein-1) and fibulin-1 [30–34]. In previous studies, we and others demonstrated an essential role of fibronectin in fibrillin-1 network assembly by human dermal fibroblast cultures [35,36]. In our previous study, we have also shown that the C-terminal halves of fibrillin-1, fibrillin-2 and fibrillin-3 as well as the N-terminal half of fibrillin-1 can interact directly with fibronectin in solid-phase binding assays. More precisely, the relevant fibrillin fragments interact strongly with the collagen/gelatin-binding domain of fibronectin. These interactions are inhibited by gelatin. We have demonstrated further that cell-associated multimerization of the fibrillin C-termini generate high-affinity binding sites for both fibronectin and the fibrillin N-termini [36,37].
Remodelling of the extracellular matrix is an important process during development, wound healing and pathological processes. Therefore it is important to understand how matrix deposition, homoeostasis and degradation are regulated. Improper matrix remodelling by preventing turnover of collagen I or by alteration of regulation of matrix-degrading proteases and their inhibitors result in fibrosis, arthritis and developmental abnormalities [38–41].
Sottile and Hocking  demonstrated the requirement of a continual fibronectin matrix polymerization for the retention of collagen I and thrombospondin-1 in fibrillar structures. These authors have shown that an intact fibronectin matrix is essential to maintain the composition of cell–matrix adhesion sites, whereby integrin α5β1 is unable to localize to cell–matrix adhesion sites in the absence of fibronectin polymerization. These data suggest that fibronectin is not only a ‘master organizer’ of extracellular matrix, but also a ‘master stabilizer’. In order to do so, it is essential that fibronectin polymerization is maintained as inhibition of fibronectin assembly causes destabilization of fibronectin matrix .
The present study aims to further characterize the interaction of fibrillins with fibronectin and the role fibronectin plays in microfibril homoeostasis. We show that the entire collagen/gelatin-binding domain of fibronectin is required for optimal interaction with fibrillins, but that this interaction is not necessary for fibrillin-1 network assembly. We demonstrate that the fibronectin network is required for homoeostasis of an immature, but not a mature, fibillin-1 network, even though fibronectin is present in mature microfibrils.
MATERIALS AND METHODS
The production and characterization of rabbit polyclonal antibodies against the N-terminal and C-terminal halves of human fibrillin-1 (anti-rFBN1-N and anti-rFBN1-C antibodies) [42,43], fibrillin-2 (anti-rFBN2-N and anti-rFBN2-C antibodies)  and fibrillin-3 (anti-rFBN3-C antibody)  have been described previously. The mouse monoclonal antibody (mouse ascites, anti-FN clone 15 antibody) against human fibronectin was purchased from Sigma (product F7387). The mouse monoclonal antibody against 5C3 (anti-FN 5C3 antibody) was generated against the recombinant fibronectin fragment FN70K which includes the fibrin/heparin- and collagen/gelatin-binding domain and the epitope was mapped previously to FNI9 . The rabbit anti-human fibronectin polyclonal antibody (anti-hFN poly IgG antibody) was made in-house to plasma fibronectin isolated by affinity chromatography on gelatin–agarose: the antibodies recognized only fibronectin in immunoblots of plasma or fibroblast lysates. The rabbit anti-mouse fibronectin polyclonal antibody (anti-mFN poly IgG antibody) was purchased from Millipore (product AB2033). Cy3 (indocarbocyanine)-conjugated AffiniPure goat anti-(rabbit IgG) (H+L; heavy chain+light chain) and Alexa Fluor® 488-conjugated AffiniPure goat anti-(mouse IgG) (H+L), horseradish peroxidase-conjugated AffiniPure goat anti-rabbit and anti-(mouse IgG) (H+L), 12-nm colloidal gold-AffiniPure donkey anti-(mouse IgG) (H+L) and 18-nm colloidal gold-AffiniPure donkey anti-(rabbit IgG) (H+L) antibodies were purchased from Jackson ImmunoReseach Laboratories.
The production and the purification of the N- and C-terminal halves of fibrillin-1 (rFBN1-N and rFBN1-C) , fibrillin-2 (rFBN2-N and rFBN2-C)  and fibrillin-3 (rFBN3-C)  have been described previously in detail. An overview of these fragments is presented in Figure 1(A). Fibronectin was purified from human plasma as previously described . Recombinant FNI4–FNII2 (amino acid residues 184–467), FNI6–FNII2 (amino acid residues 291–467), FNI6–FNI9 (amino acid residues 291–608) and FNI7–FNI9 (amino acid residues 468–608) were expressed as secreted His-tagged proteins using the baculovirus vector pAcGP67.coco (COCO) and purified from conditioned medium as described previously [45,48]. Residues are numbered beginning with the start methionine residue of the fibronectin signal peptide. The purity of the recombinant fibrillin and fibronectin fragments is shown in Supplementary Figure S1 (at http://www.biochemj.org/bj/456/bj4560283add.htm). Porcine gelatin (Sigma; product G1890) was resuspended in 1% (v/v) acetic acid at 37°C. For some experiments, gelatin was labelled with FITC (Sigma; product F7250) according to the manufacturer's protocol followed by extensive dialysis against TBS (50 mM Tris/HCl, pH 7.4, and 150 mM NaCl).
Characterization of the fibrillin–fibronectin interaction
HSFs (human skin fibroblasts) were isolated from foreskin obtained from standard circumcision procedures in agreement with local ethics regulations (ethics approval PED-06-054, Montreal Children Hospital, QC, Canada). The tissue donors were healthy individuals aged between 2 and 14 years. Consent was obtained from their parents. HSFs between passage 2 and 10 were used in experiments. Cells were cultured at 37°C in a 5% CO2 atmosphere, in DMEM (Dulbecco's modified Eagle's medium) supplemented with 2 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 10% FBS (Wisent Bioproducts). Depletion of fibronectin from FBS was performed by loading FBS twice on a gelatin Sepharose 4B column according to the manufacturer's instructions (GE Healthcare). The flow-through consisting of FBS depleted of fibronectin was used for fibrillin-1 assembly inhibition studies in cell culture.
Solid-phase binding assays
Solid-phase binding assays were performed as described previously . In brief, 10 μg/ml TBS (100 μl) of full-length or recombinant fragments of fibronectin were immobilized for 16 h at 4°C in 96-well plates (Maxisorp; Nalge Nunc International). Wells were blocked with 5% (w/v) non-fat dried skimmed milk powder in TBS including 2 mM CaCl2 for 1 h at room temperature (21°C). Washing was performed with TBST (TBS/0.05% Tween-20). Serial dilutions (1:2 starting at 100 μg/ml) of fibrillin fragments were prepared in 2% (w/v) non-fat dried skimmed milk powder and 2 mM CaCl2 in TBS (binding buffer) and were incubated with immobilized proteins for 2 h. Similarly, for ionic strength dependency experiments, fibrillin fragments were incubated in binding buffer containing various concentrations of NaCl as indicated in Figure 1(B). Three 5-min washes with TBST were then performed between each subsequent step. Bound ligands were detected by incubation with polyclonal antisera against rFBN1-N, rFBN1-C, rFBN2-N, rFBN2-C and rFBN3-C, at a dilution of 1:1000 in binding buffer for 90 min, followed by incubation with horseradish peroxidase-conjugated AffiniPure goat anti-(rabbit IgG) (H+L) antibody (1:800 dilution) for 90 min and was treated with 5-aminosalicylic acid for colour reaction.
For inhibition experiments, the above procedure was used with modifications of the incubation with fibrillin fragments. After blocking non-specific binding sites, serial dilutions (1:2 starting at 1:400) of anti-FN 5C3 or anti-FN clone 15 diluted in binding buffer containing constant concentrations (50 μg/ml) of fibrillin fragments were incubated with immobilized full-length fibronectin for 2 h. For inhibition experiments with the FUD (functional upstream domain) peptide from Streptococcus pyogenes F1 adhesin protein , after the blocking step, constant concentrations (50 μg/ml) of fibrillin fragments were incubated with immobilized full-length fibronectin for 90 min with or before incubation with serial dilutions (1:2 starting at 50 nM in binding buffer) of FUD in binding buffer.
For double immunofluorescence experiments, HSFs were seeded at 7.5×104 cells/well into eight-well chamber slides in DMEM supplemented with fibronectin-depleted FBS. Cells were grown for 4 days until robust fibronectin and fibrillin-1 networks developed. Cells were washed twice with 137 mM NaCl, 207 mM KCl, 4.3 mM Na2HPO4 and 1.47 mM KH2PO4, pH 7.4 (PBS, standard washing buffer). Cells were then fixed with ice-cold 70% methanol/30% acetone for 5 min, followed by three washes with PBS. Cells were blocked for 30 min with PBS-G (10% normal goat antiserum in PBS, Jackson ImmunoReseach Laboratories) and incubated for 90 min with the primary antibodies anti-rFBN1-C and anti-FN clone 15 in PBS-G (1:500 dilution). Three washes were performed followed by a 60-min incubation with secondary Cy3-conjugated AffiniPure goat anti-rabbit and Alexa Fluor® 488-conjugated AffiniPure goat anti-mouse, or Cy3-conjugated AffiniPure goat anti-mouse antibodies (1:100 dilution in PBS-G). Cells were washed thrice. Cell nuclei were counterstained with DAPI (1 μg/ml in water) for 5 min before the slides were washed and sealed with coverslips. Fluorescent images were recorded with an Axioskop 2 microscope equipped with an Axiocam camera and AxioVision software version 188.8.131.52 (Zeiss), or in some cases with an Axiovert 135 microscope (Zeiss) equipped with a Retiga EXI camera and the Northern Eclipse imaging software.
Gelatin inhibition of fibrillin-1 network formation
HSFs were seeded (7.5×104 cells/well) into eight-well chamber slides in DMEM supplemented with fibronectin-depleted FBS in the presence of 100 μg/ml gelatin, FITC-gelatin or equivalent volumes of TBS. Cells were grown for 5 days and immunofluorescence was performed as described above.
FUD pulse–chase experiment in cell culture
For short-term experiments, HSFs were prepared as described above. Following 4 days of growth, the cells were then pulsed with 500 nM FUD peptide for 2 days. A chase was then performed with the same medium in the absence of the FUD peptide for 2 and 4 additional days. Indirect immunofluorescence was performed as described above. For long-term experiments, HSFs were cultivated for 3 weeks. A pulse was performed with 2 μM FUD for 3 days and the cells were analysed by indirect immunofluorescence.
Microfibrils were extracted from HSFs or mouse tissues using collagenase or guanidine extraction as described previously with small modifications [50,51]. For microfibril extraction from cells, HSFs were grown for 3 weeks for subsequent collagenase extraction and for 1 month for subsequent guanidine extraction on a 1000 cm2 total cell culture area. Cells were scraped off in 50 mM Tris/HCl, pH 7.4, 250 mM NaCl, 5 mM CaCl2 (extraction buffer), followed by centrifugation at 6000 g for 15 min. The pellet was then resuspended in extraction buffer containing either 1 mg/ml crude collagenase (from Clostridium histolyticum, Sigma; product C0130), 2 mM phenylmethylsulfonyl fluoride and 5 mM N-ethylmaleimide; or 6 M guanidine, followed by incubation for 4 h at 4°C. The extract was centrifuged at 7000 g for 20 min. The supernatant was then separated on a Sephacryl S-500 HR gel filtration column (void volume=47.52 ml) equilibrated with extraction buffer at a flow rate of 0.5 ml/min using an ÄKTApurifier 10 chromatography system (GE Healthcare). For microfibrils extracted from mouse tissue, eight 1-month-old mice were anaesthetized. The McGill University Animal Care Committee approved all procedures. Intracardiac perfusion was performed with PBS followed by extraction buffer. The lungs, aorta and dorsal skin were dissected, homogenized in extraction buffer and centrifuged at 6000 g for 15 min. The pellet was divided into two equal amounts and extraction buffer with either 1 mg/ml crude collagenase, 2 mM phenylmethylsulfonyl fluoride and 5 mM N-ethylmaleimide, or 6 M guanidine was added followed by incubation for 3 days at 4°C. The samples were then processed as described above for microfibril extraction from cells.
Monoclonal anti-FN clone 15 (2.9 mg) antibody was coupled with 2.5 ml of cyanogen bromide-activated Sepharose (GE Healthcare) according to the manufacturer's instructions and packed into a 2 ml chromatography column. High-molecular-mass microfibril and intermediate-molecular-mass gel filtration fractions from HSF guanidine microfibril extractions were extensively dialysed against TBS and 2 mM CaCl2. Aliquots (780 μl) of high-molecular-mass microfibril and intermediate-molecular-mass gel filtration fractions from the HSF collagenase and guanidine microfibril extractions were loaded under low flow (10 μl/min) on to the antibody column equilibrated in TBS and 2 mM CaCl2. The unbound material was collected. The bound material was eluted with 0.1 M glycine, pH 3.0, at 0.5 ml/min and immediately neutralized with 2 M Tris.
Each major peak fraction (100 μl) from the microfibril extraction was diluted in 400 μl of TBS, 2 mM CaCl2 and dot-blotted on to a 0.45 μm nitrocellulose membrane (Bio-Rad Laboratories). Non-specific binding sites were blocked with 5% (w/v) non-fat dried skimmed milk powder in TBS for 1 h at room temperature. Blots were then incubated with primary antibodies (1:500 dilution in 5% BSA in TBS) overnight at 4°C. After three 10-min washes with TBST, blots were incubated with horseradish peroxidase-conjugated AffiniPure goat anti-rabbit antibody (1:800 dilution in TBS) for 1.5 h at room temperature. Colour reaction was performed with TBS, including 0.5 mg/ml 4-chloro-1-naphthol, 17% (v/v) methanol and 0.02% H2O2.
For the immunoprecipitation experiments, the immunoblotting procedure was performed with slight modifications. To compensate for the differences in absorbance at 280 nm of the different peaks, 5 μl of start material, 50 μl of unbound and 200 μl of the bound fractions were dot-blotted on to a 0.45 μm nitrocellulose membrane (Bio-Rad Laboratories). The membrane was incubated with horseradish peroxidase-conjugated AffiniPure goat anti-rabbit antibody (1:3500 dilution) and then developed with ECL (Thermo Scientific) using Hyblot CL autoradiography film.
Immunogold staining and electron microscopy
HSFs were prepared as described above and grown for 7 days. Cells were washed three times with PBS and then fixed for 1 h on ice with 3% paraformaldehyde in PBS, followed by three washes with PBS. Cells were blocked for 1 h with 5% normal donkey antiserum in PBS (Jackson ImmunoResearch Laboratories). The primary anti-rFBN1-C (1:100 dilution) and anti-FN clone 15 (1:100 dilution) antibodies were diluted in PBS and incubated overnight at 4°C. Following three washes with PBS, 12-nm and 18-nm gold-conjugated secondary antibodies were used at a dilution of 1:20 in PBS. Cells were washed with 0.1 M sodium cacodylate (cacodylate buffer) and then fixed with 2% glutaraldehyde in cacodylate buffer. Cells were washed four times with cacodylate buffer, fixed for 20 min with 1% OsO4 in cacodylate buffer. Cells were dehydrated and embedded in EPON. Ultra-thin sections were processed and grids were contrasted with 1% uranyl acetate and enhanced with Reynold's lead for 3 min. Sections were then examined with a FEI Tecnai 12, 120 kV electron microscope equipped with a Gatan 792 Bioscan 1k×1k Wide Angle Multiscan CCD camera.
Characterization of the fibrillin–fibronectin interaction
We have shown previously that fibrillin-1, fibrillin-2 and fibrillin-3 C-terminal halves and the fibrillin-1 N-terminal half interact directly with fibronectin in solid-phase binding assays . To determine whether the fibrillin–fibronectin interaction is of ionic nature, various fibrillin fragments were tested for binding to immobilized full-length fibronectin in the presence of increasing NaCl concentrations (Figure 1B). The presence of up to 1 M NaCl did not decrease the fibrillin interaction with fibronectin. Instead, the interactions increased slightly. These data indicate that the fibrillin–fibronectin interaction is of non-ionic nature. In control experiments, we verified that high NaCl concentrations did not affect the multimerization state of fibrillin C-terminal fragments, which is a pre-requisite for the interaction with fibronectin (results not shown).
Characterization of the fibrillin-binding site in fibronectin
We have reported previously that the fibrillin-1, fibrillin-2 and fibrillin-3 C-terminal halves and the fibrillin-1 N-terminal half interact with the collagen/gelatin-binding domain of fibronectin comprising FNI6–FNI9 . To map the fibrillin-binding site in this region, we performed inhibition experiments with a monoclonal antibody against FN 5C3, where the epitope is located in the FNI9 domain of fibronectin . This antibody inhibited the interaction of rFBN1-C, rFBN2-C and rFBN3-C with full-length fibronectin, whereas the control anti-FN clone 15 antibody, where the epitope is located in the region between FNIII1 and the C-terminus of fibronectin, did not exert an inhibitory effect (Figure 2A). Several other monoclonal antibodies, where epitopes are located outside of the collagen/gelatin-binding domain of fibronectin, also did not inhibit this interaction (results not shown). These data show that FNI9 is a critical determinant of the fibrillin–fibronectin interaction in the solid-phase binding assay.
Mapping of the fibrillin-binding site on fibronectin
To map further the fibrillin-binding site on fibronectin, we used recombinant fibronectin subfragments spanning the collagen/gelatin-binding domain in solid-phase binding assays with fibrillin fragments. Proper folding of those recombinant fragments was verified through interaction with monoclonal antibodies sensitive to conformational changes ( and results not shown). We previously reported that rFBN1-C, rFBN2-C, rFBN3-C and rFBN1-N interacted strongly with a fibronectin proteolytic fragment FN40K spanning the entire collagen/gelatin-binding domain FNI6–FNI9 . In the present study, we demonstrated that fibrillin fragments interact with recombinant FNI6–FNI9 in a manner similar to that of FN40K (Figure 2B). However, interaction with smaller subfragments of this region was either greatly reduced or completely absent (Figure 2B). From these experiments, we conclude that the fibrillin C-terminal halves and the fibrillin-1 N-terminal half require the complete collagen/gelatin-binding domain of fibronectin for efficient interaction with FNI9, potentially providing initial protein–protein contacts.
The collagen/gelatin-binding domain of fibronectin is not involved in fibrillin network assembly
We have shown previously that gelatin inhibits fibrillin interaction with fibronectin in solid-phase binding assays . In the present study, we tested whether gelatin inhibits the fibrillin-1 network assembly by skin fibroblasts. HSFs were grown for 4 days in the presence of various concentrations of gelatin (100 μg/ml is shown in the Figure) or TBS as a control (Figure 3A). The formation of both networks, fibrillin-1 and fibronectin, by HSFs was not affected by the addition of gelatin. To ensure that gelatin was not endocytosed by the cells and thus was available to interact with fibronectin in the experimental system, FITC-labelled gelatin was added to HSFs under identical conditions. FITC-gelatin co-localized with the fibronectin network, indicating that it occupies the binding sites on fibronectin within the collagen/gelatin-binding region (Figure 3B) as previously observed . FITC-gelatin also co-localized with the fibrillin-1 network, which in turn co-localizes with fibronectin as was previously shown  (Figure 3B). We verified by solid-phase binding assays that gelatin does not directly interact with fibrillin-1 (results not shown). Several concentrations of gelatin as well as addition of gelatin at different time points were tested, but did not result in any effect on fibrillin-1 network formation. The monoclonal anti-FN 5C3 antibody, which inhibits the fibrillin–fibronectin interaction in solid-phase binding assay (see Figure 2A), and the FNI6–FNI9 fibronectin fragment also did not affect fibrillin-1 network formation when added to HSFs (results not shown). From these experiments, we conclude that the interaction of fibrillin-1 with the collagen/gelatin-binding domain of fibronectin, that dominates the solid-phase binding assay, is not involved in fibrillin-1 network assembly. However, these experiments do not rule out involvement of other fibrillin–fibronectin interactions in fibrillin-1 assembly.
Analysis of gelatin treatment of fibrillin-1 network formation by HSFs
Fibronectin network is required for fibrillin-1 network homoeostasis
We have shown previously that the presence of a fibronectin network is required for the assembly of fibrillin-1 into microfibrils . The FUD peptide from the bacterial adhesin F1 protein has been shown to block fibronectin network formation . As the fibronectin network is constitutively turned over [28,29], we used the FUD peptide as a tool to study the role of the fibronectin network in microfibril homoeostasis. The fibronectin network and microfibrils were first allowed to be assembled by HSFs for 4 days (Figure 4A), before the FUD peptide was added for 2 days to initiate fibronectin fibre disassembly (Figure 4B). As expected, the fibronectin network readily disassembled after FUD treatment. Similarly, microfibrils also disassembled after the addition of FUD, suggesting that microfibrils at this early stage require the presence of a fibronectin network for homoeostasis. The FUD peptide does not affect expression of fibronectin and fibrillin-1 proteins in HSFs . After the FUD peptide was removed and replaced by fibronectin-depleted normal cell culture medium for 2 days (Figure 4C) and 4 days (Figure 4D), both networks reassembled and co-localized. These data demonstrate that fibrillin-1 homoeostasis is directly coupled with fibronectin homoeostasis.
Microfibrils require fibronectin network for homoeostasis
FUD interferes with the fibrillin–fibronectin interaction
To understand the mechanism by which the FUD peptide affects the fibrillin-1 network homoeostasis, we performed inhibition experiments in solid-phase assays. The FUD peptide interacts with FNI2–FNI5 domains of the fibrin-binding N-terminal fragment and FNI8 and FNI9 of the collagen/gelatin-binding domain of fibronectin . Interaction of FUD with fibronectin causes significant conformational changes in fibronectin, exposing the Arg-Gly-Asp integrin-binding sequence in FNIII10 . The FUD peptide does not interact directly with fibrillin-1 (Supplementary Figure S2 at http://www.biochemj.org/bj/456/bj4560283add.htm). However, FUD inhibits the interactions of the rFBN1-N and rFBN1-C with fibronectin when added simultaneously with the fibrillin fragments (Figure 5A). FUD also partially displaces bound rFBN1-N and rFBN1-C from fibronectin, probably by competing with fibrillin for interaction with the FNI8 and FNI9 domains of the collagen/gelatin-binding region of fibronectin (Figure 5B).
FUD inhibits fibrillin-1 interaction with fibronectin
FUD does not affect fibronectin and fibrillin-1 networks in long-term culture
To investigate whether the fibrillin-1 network requires fibronectin for homoeostasis when both networks are mature, long-term cultures of HSFs were analysed for their sensitivity to FUD treatment (Figure 6). HSFs were grown for 3 weeks (Figure 6A), before FUD was added for 3 days to the culture medium (Figure 6B). After 3 weeks in culture, HSFs assembled dense fibronectin and fibrillin-1 networks. A higher concentration (2 μM) and a longer pulse of the FUD peptide (3 days) were used to compensate for the higher number of fibronectin fibres present in the long-term culture compared with short-term cultures. The FUD peptide did not disrupt the fibronectin or fibrillin-1 networks (Figure 6B), suggesting that both networks are more stabilized in long-term cultures. Similar results were observed with cultures cultivated for 2 weeks (results not shown).
FUD does not affect fibronectin and fibrillin-1 networks in long-term culture
Fibronectin is present in extracted microfibrils from cells and tissues
We observed previously that both fibrillin-1 and fibronectin are present at the ultrastructural level in some extracellular fibres produced by HSFs . To investigate whether fibronectin is present in mature ‘beads-on-a-string’ microfibrils, microfibrils were extracted by collagenase digestion and by guanidine extraction from 3-week-old and 1-month-old HSF cultures respectively and separated by gel filtration chromatography (Figures 7A and 7C). Microfibrils were also extracted by collagenase and guanidine extraction from 1-month-old mouse lung, skin and aorta. Microfibrils from lung collagenase extraction (Figure 7B) and skin guanidine extraction (Figure 7D) are shown. Fractions from each peak were dot-blotted to determine the presence of fibrillin-1 and fibronectin. Note that the intensity of the dots for both fibrillin-1 and fibronectin do not correlate with absorbance peak heights as other proteins present in those fractions also account for the absorbance level. As expected, on the basis of previously published results, a strong signal for fibrillin-1 was detected in the peak containing the void volume (50–60 ml) for all extractions, indicating the presence of high-molecular-mass microfibrils [51,54]. However, fractions from larger elution volumes, which were never analysed before, also contained detectable levels of fibrillin-1 potentially representing smaller forms of microfibrils. Fibronectin was present in the microfibril fractions in all extractions (50–60 ml). It was also present in higher elution volume fractions. We observed similar results with collagenase-extracted microfibrils from 1-, 2- and 9-week-old HSFs cultures and from mouse skin and aorta, as well as with guanidine-extracted microfibrils from mouse aorta (results not shown).
Fibronectin is present in various size extracted microfibrils
We investigated whether fibronectin and fibrillin-1 are interacting in high-molecular-mass microfibrils and in the intermediate-molecular-mass peaks. Samples of those peaks from collagenase and guanidine extractions from HSFs were immunoprecipitated with an anti-fibronectin antibody affinity column. To compensate for the differences in absorbance at 280 nm of each peak after the affinity column, 5 μl of start material, as well as 50 μl of unbound and 200 μl of bound fractions were dot-blotted with anti-fibronectin and anti-fibrillin-1 antibody (Figure 8). Fibronectin immunoprecipitated under all the conditions tested. Fibrillin-1 was also co-immunoprecipitated in the high-molecular-mass microfibrils and in the intermediate-molecular-mass peaks regardless of the extraction method used (Figure 8). Therefore we conclude from these experiments that fibronectin interacts directly or indirectly with large mature microfibrils as well as with fibrillin-1 in smaller intermediate-level microfibrils.
Fibrillin-1 co-immunoprecipitates with fibronectin from the microfibril and intermediate-molecular-mass gel filtration fractions
Fibrillin-1 and fibronectin can co-localize to the same extracellular matrix fibres or to distinct fibres
In our previous study, we observed frequent co-localization of fibrillin-1 and fibronectin to the same extracellular fibrils produced by dermal fibroblasts through double-immunogold labelling . In the present study, whereas the majority of fibrillin-1 co-localizes with fibronectin in immunofluorescence, we noticed some distinct fibrillin-1 and fibronectin fibres (Figures 4A and 6A). Considering our previous electron microscopy analysis and current immunofluorescence results, we investigated the co-localization of both proteins at the ultrastructural level in more detail. Double-immunogold labelling was performed on extracellular fibrils produced by HSFs cultured for 7 days. As previously discussed, we observed many fibres labelled with both 12-nm gold particles representing fibronectin and 18-nm gold particles representing fibrillin-1 (Figures 9B–9D, asterisks). However, we also noticed some fibres labelled solely for fibrillin-1 or for fibronectin (Figures 9A and 9B). These results corroborate our immunofluorescence, microfibril extraction and immunoprecipitation results where fibrillin-1 and fibronectin can be present in the same extracellular fibrils or can be localized to fibrils independently of the other protein.
Fibrillin-1 and fibronectin localize to the same and to different fibres
Although a number of studies have investigated microfibril assembly, the mechanism of microfibril formation, as well as the components required to maintain their assembly, are still obscure. In the present study, we show that fibrillin interaction with fibronectin in a solid-phase assay is non-ionic and sensitive to inhibition with FUD and monoclonal antibody to FNI8–FNI9. We also demonstrate that the entire collagen/gelatin-binding domain of fibronectin, and not just FNI8–FNI9, is required in the solid-phase assay for optimal interaction with fibrillins. This latter finding is reminiscent of the multi-module interaction of type I collagen with fibronectin . However, we uncovered the enigma that both FUD (Figure 5) and gelatin  block the interaction between fibronectin and fibrillin in the solid-phase assay, whereas only FUD blocks deposition of fibrillin during short-term culture of HSFs. Using pulse treatment with FUD demonstrated that fibronectin and, in turn, fibrillin networks in short-term cultures are FUD-sensitive, whereas these networks were unaffected in long-term cultures.
Increasing NaCl concentration did not inhibit the interaction of fibrillins with fibronectin, indicating that the dominant interaction between the proteins is not electrostatic. A covalent interaction appears unlikely as the FUD peptide (shown in the present study) and gelatin  dissociate bound fibrillin-1 from fibronectin. Therefore the data indicate that the interaction between fibrillins and fibronectin is primarily of a hydrophobic nature.
All six modules of the collagen/gelatin-binding domain of fibronectin are required for full-affinity interaction with gelatin [55,56]. In the present study, we observed that the fibrillin C-terminal halves and the fibrillin-1 N-terminal half interact less well with smaller subfragments of the collagen/gelatin-binding domain of fibronectin. We also demonstrate that a monoclonal antibody that binds to FNI9  inhibits interaction of the fibrillins to the collagen/gelatin-binding domain. Gelatin inhibits the fibrillin interaction with fibronectin in vitro . On the basis of these results, we propose that fibrillins are similar to type I collagen in that they require all six modules of the collagen/gelatin-binding domain for complete interaction in vitro. However, the individual residues involved in both interactions must be different as the gelatin interaction with fibronectin was found to require essential charged residues , whereas we observed that the fibrillin interaction with fibronectin was insensitive to salt concentration.
Gelatin did not block assembly of fibronectin or early assembly of fibrillin. Sottile et al.  have shown that collagen type I fibre assembly requires the collagen/gelatin-binding domain of fibronectin. Therefore fibrillin assembly must be linked to fibronectin assembly via fibrillin–fibronectin interaction independently of those involving the collagen/gelatin-binding region that dominates in the solid-phase assay. Such interactions could be via other parts of fibronectin or unknown adapter molecules. A potential adapter could be heparan sulfate since both fibrillin and fibronectin interact directly with heparan sulfate and since the addition of exogenous heparin/heparan sulfate inhibits fibrillin-1 network assembly by HSFs [42,59,60]. However, several experiments support the idea of another fibrillin-binding site in fibronectin. In our previous study, we observed that the N-terminal half of fibrillin-1 interacted with the FNIII1-C proteolytic fragment of fibronectin . Kinsey et al.  observed that two recombinant fibrillin-1 fragments, which are both contained in our rFBN1-N construct, interact in a region between FNIII12 and FNIII14 of fibronectin, as well as between FNI1 and FNI5. These authors also observed interactions of N-terminal recombinant fibrillin-1 fragments with a fibronectin region between domains FNIII7 and FNIII11. Further experiments are needed to investigate whether those interactions outside the collagen/gelatin-binding region are involved in early fibrillin-1 network assembly. Moreover, more experiments should define the functional significance of fibrillin interaction with the fibronectin collagen/gelatin-binding region (also see below).
In contrast with the results with gelatin, the FUD polypeptide disrupted formation of the fibronectin network and early fibrillin-1 network. FUD peptide has the potential to act through two distinct mechanisms. On the one hand, FUD prevents polymerization of fibronectin [49,61]. On the other hand, FUD inhibits the direct interaction between fibrillin-1 and fibronectin as shown in the present study by solid-phase inhibition assays (Figure 5). FUD forms a β-zipper structure with fibronectin FNI2–FNI5 modules of the fibrin-binding N-terminal domain and FNI8 and FNI9 of the collagen/gelatin-binding domain . FUD binding to fibronectin disrupts fibronectin homotypic interactions and induces important conformational changes in fibronectin which extend as far as FNIII10 and expose the Arg-Gly-Asp integrin site in this domain . As the FUD peptide does not interact directly with fibrillin-1 (Supplementary Figure S2), we suggest that on binding to fibronectin, FUD induces conformational changes in fibronectin disrupting the interactions with fibrillin-1. However, as the fibronectin network disappeared after FUD treatment, FUD must act primarily through the former mechanism in our short-term cell culture experiments. Regardless of which mechanism acts, we can conclude that the early fibrillin-1 network requires a stable fibronectin network for homoeostasis. Remarkably, FUD did not inhibit fibronectin polymerization after pre-culturing of HSFs for 3 weeks (Figure 6). We conclude that fibronectin fibres in these long-term cultures are more stable and resistant to FUD effects, probably due to more extensive cross-linking.
The finding that fibrillin–fibronectin dynamics are different after prolonged culture of HSFs prompted us to examine the localization of these proteins in vivo and in vitro in more detail. Following the extraction of microfibrils by two methods from HSF cultures and mouse tissues, we observed fibronectin co-eluting with fibrillin-1 in intermediate-molecular-mass fractions. Through immunoprecipitation, we demonstrated that fibronectin interacts directly or indirectly with fibrillin-1 in those fractions extracted from HSFs. We hypothesize that those fibronectin-containing fractions are intermediate or immature (nascent) fibrillin-1 fibres. In the same experiments, we demonstrated that fibronectin co-eluted in the void volume with the extractable mature microfibril peak from cell cultures as well as mouse tissues regardless of the extraction method used. Immunoprecipitations demonstrated that fibronectin interacts with these more mature microfibrils. Those results are supported by our electron microscopy data in which we observed fibrillin-1 and fibronectin frequently (but not always) co-localizing to the same extracellular fibres produced by HSFs.
The results of the present study favour the model shown in Figure 10. We propose that fibrillins require the assistance of fibronectin for initial early assembly and for the homoeostasis of immature fibres, which probably corresponds to the intermediate-molecular-mass fractions (Figure 7). As microfibril maturation proceeds, the fibres are progressively larger in size, such as the microfibrils present in the void volume of the extraction column (Figure 7) and observed by electron microscopy (Figure 9). These maturing microfibrils lose the requirement for a fibronectin network even though fibronectin remains associated with fibrillin. One transglutaminase-mediated cross-link has been identified in fibrillin-1 from isolated microfibrils and the study suggested that many more (10–15%) of the lysine residues in fully developed microfibrils are involved in cross-links . The exact time course for this cross-linking is currently unknown. Interestingly, tissue transglutaminase interacts with fibronectin via the collagen/gelatin-binding region from which it acts upon glutamine residues at the N-terminus of fibronectin . It is possible that a complex between fibronectin and tissue transglutaminase serves to mediate cross-links in fibrillin-1 organized in mature microfibrils. Given the ability of fibrillin-1 to strongly interact with fibronectin's collagen/gelatin-binding region, we speculate that fibrillin-1 may be able to compete with the binding site of tissue transglutaminase on fibronectin to regulate cross-link formation in microfibrils. As maturation through cross-link formation proceeds, microfibrils become stable enough to become independent of the fibronectin network and constitute the microfibrils only labelled with fibrillin-1 (Figures 4A, 6A, 9A and 9B). With the present study, we provide new evidence of the complex roles of fibronectin as a ‘master stabilizer’ of early fibrillin-1 extracellular matrix and contribute to the understanding of fibrillin matrix turnover and remodelling.
Hypothetical model describing the stages of microfibril assembly that depend on the presence of fibronectin (involvement of glycosaminoglycans is omitted for simplicity)
Dulbecco’s modified Eagle’s medium
functional upstream domain
heavy chain+light chain
human skin fibroblast
10% normal goat antiserum in PBS
transforming growth factor-β
Laetitia Sabatier designed, executed and analysed the experiments and wrote the paper. Jelena Djokic designed, executed and analysed the revision experiments and revised the text. Christine Fagotto-Kaufmann and Marian Chen performed original and revision experiments respectively. Douglas Annis and Deane Mosher contributed reagents, provided experimental input and helped to finalize the paper. Dieter Reinhardt conceived and supervised the project, analysed the data and finalized the paper.
We thank Dr Jean-Martin Laberge (Montreal Children's Hospital, Montreal, QC, Canada) for providing clinical samples, and Ms Amelie Pagliuzza for critically reading the paper before submission.
This work was supported by the Canadian Institutes of Health Research [grant number MOP-106494], the Natural Sciences and Engineering Research Council of Canada [grant number RGPIN 375738-09], the Canada Foundation for Innovation, the National Institutes of Health [grant number HL021644] and the Network for Oral and Bone Health Research (Ph.D. student scholarship to L.S.).