Molecular basis of elastic fiber formation. Critical interactions and a tropoelastin-fibrillin-1 cross-link.

We have investigated the molecular basis of elastic fiber formation on fibrillin microfibrils. Binding assays revealed high affinity calcium-independent binding of two overlapping fibrillin-1 fragments (encoded by central exons 18-25 and 24-30) to tropoelastin, which, in microfibrils, map to an exposed "arms" feature adjacent to the beads. A further binding site within an adjacent fragment (encoded by exons 9-17) was within an eight-cysteine motif designated TB2 (encoded by exons 16 and 17). Binding to TB2 was ablated by the presence of N-terminal domains (encoded by exons 1-8) and reduced after deleting the proline-rich region. A novel transglutaminase cross-link between tropoelastin and fibrillin-1 fragment (encoded by exons 9-17) was localized by mass spectrometry to a sequence encoded by exon 17. The high affinity binding and cross-linking of tropoelastin to a central fibrillin-1 sequence confirm that this association is fundamental to elastic fiber formation. Microfibril-associated glycoprotein-1 showed calcium-dependent binding of moderate affinity to fibrillin-1 N-terminal fragment (encoded by exons 1-8), which localize to the beads. Microfibril-associated glycoprotein-1 thus contributes to microfibril organization but may also form secondary interactions with adjacent microfibril-bound tropoelastin.

Elastic fibers are one of the major insoluble fiber systems of connective tissues, providing elasticity and resilience to elastic tissues such as blood vessels, lungs, skin, and ligaments (1,2). They morphologically comprise an elastin core surrounded by a mantle of fibrillin-rich microfibrils, and their distinct tissuespecific arrangements reflect different biomechanical requirements. Elastic fiber formation in the extracellular space is a complex developmentally regulated process that has been visualized by electron microscopy studies as the accretion of tropoelastin on preformed bundles of fibrillin microfibrils in the pericellular space. The microfibril template thus profoundly influences tropoelastin deposition and the organization of mature elastic fibers as well as their biomechanical properties (3).
Understanding of elastic fiber formation requires delineation of the molecular basis of the critical early stage of tropoelastin deposition on microfibrils.
Fibrillin-1 is a large cysteine-rich multidomain glycoprotein that polymerizes in the extracellular space in a head-to-tail manner to form microfibrils that provide a force-bearing structural framework for dynamic connective tissues (2,4,5). It contains 47 epidermal growth factor (EGF) 1 -like domains: 43 calcium-binding epidermal growth factor (cbEGF)-like domains, seven 8-cysteine (TB) modules, two hybrid motifs, and a proline-rich region that may act as a hinge region. Mutations in fibrillin-1 cause Marfan syndrome, a heritable disease associated with severe aortic, ocular, and skeletal defects due to defective elastic fibers (6). This linkage to Marfan syndrome and its developmental distribution (7) confirm fibrillin-1 as the major fibrillin isoform in elastic fibers.
In addition to fibrillin-1, other microfibril-associated molecules have been identified (2). The best candidate for an integral structural component is microfibril-associated glycoprotein-1 (MAGP-1), which was first identified as a microfibrilassociated molecule in reductive denaturing tissue microfibril extracts (8) and which routinely co-localizes with microfibrils in elastic and nonelastic tissues (2,9). MAGP-1 was shown to bind to an N-terminal fibrillin-1 peptide (10), to tropoelastin (11), and to extracellular matrix through interactions of its C-terminal matrix binding domain (12). An antibody inhibition study suggested that the interaction between tropoelastin and microfibrils might be mediated by the N-terminal half of MAGP-1 (13). Tropoelastin was shown to bind a fibrillin-1 fragment (encoded by exons 10 -17) through interactions involving its lysine side chains (14). Ternary complexes of MAGP-1, fibrillin, and decorin and of MAGP-1, tropoelastin, and biglycan have been identified in vitro (15,16). Whereas small leucine-rich proteoglycans may contribute to elastic fiber formation, available evidence indicates that the key molecules in this process are tropoelastin, fibrillin-1, and MAGP-1.
The molecular basis of how tropoelastin binds microfibrils during elastic fiber formation remains unclear, and no comprehensive screening of interactions with fibrillin-1 has been reported. To resolve this critical issue, we have defined the interactions between tropoelastin and MAGP-1 with fibrillin-1 and their binding affinities. We have also identified and localized a novel specific transglutaminase cross-link between tro-poelastin and fibrillin-1. The data confirm that tropoelastin strongly binds and is cross-linked to a central fibrillin-1 sequence that is exposed on microfibrils, whereas MAGP-1 binds N-terminal fibrillin-1 and contributes to the microfibril template. These data thus provide essential new insights into the molecular basis of elastic fiber formation.
Recombinant MAGP-1-Full-length human MAGP-1 was expressed in the same mammalian episomal expression system and 293-EBNA described above. After harvesting medium from transfected cells by centrifugation (4000 rpm for 5 min), protease inhibitors (protease inhibitor mixture; Sigma) were added along with 0.5% Triton X-100. Medium was dialyzed overnight into 50 mM Tris/HCl, pH 8.0, 250 mM NaCl, 0.5 mM CaCl 2 , 10 mM imidazole, 0.5% Triton X-100, and 20 mM ␤-mercaptoethanol (start buffer) and then overnight again into start buffer containing 8 M urea; these conditions allowed subsequent elution of MAGP-1 from the nickel column as a single purified product. The dialyzed medium was then pumped over a Ni 2ϩ -charged 1-ml HiTrap Chelate HP column (Amersham Biosciences) using an ⌬KTAprime system (Amersham Biosciences) at 1 ml/min. The column containing bound protein was then washed with 30 ml of start buffer containing 8 M urea. To remove the Triton X-100, the column was then washed with 10 ml of start buffer with 8 M urea but without Triton X-100. MAGP-1 was then eluted from the column using a 10-ml gradient of 10 -400 mM imidazole in start buffer with 8 M urea but without Triton X-100 at 0.7 ml/min and collected in 0.5-ml fractions. Eluted MAGP-1 was refolded slowly by dialysis into 50 mM Tris/HCl, pH 7.4, 150 mM NaCl, 10 mM CaCl 2 (TBS/CaCl 2 ) containing 2 mM reduced glutathione, 0.2 mM oxidized glutathione, and 0.005% Tween 20, followed by dialysis into TBS. Purified MAGP-1 monomers exhibited a molecular mass of 31 kDa on reducing SDS-PAGE (Fig. 1D). In nonreducing conditions, MAGP-1 appeared both as 31-kDa monomers and as a range of high M r multimers (Fig. 1D).
For tropoelastin binding assays, tropoelastin was adsorbed onto wells that were then blocked prior to the addition of soluble fibrillin-1 or MAGP-1 ligands. For studying MAGP-1 interactions with fibrillin-1, fibrillin-1 fragments were bound to the wells prior to the addition of soluble MAGP-1. The soluble ligands were either biotinylated or detected using antibodies, and the binding assays were conducted at 37°C. The block for biotinylated ligands was bovine serum albumin (BSA), and the block for nonbiotinylated ligands was milk protein.
For assays using biotinylated ligands, each fibrillin-1 fragment or MAGP-1 was rotated at room temperature for 30 min with an approximate 10-fold molar excess of 10 mg/ml solution of Immunopure sulfo-N-hydroxysuccinimide ester-biotin (Pierce) diluted in PBS. Each mixture was then dialyzed against several changes of 0.02 M Tris/HCl, pH 7.4, containing 0.1 M NaCl and 0.001 M CaCl 2 (TBSa/CaCl 2 ) to remove excess biotin. Flat bottomed microtiter plates (ThermoLabsystems; Franklin, MA) were coated with MAGP-1 or tropoelastin at 5 g/ml in TBSa/CaCl 2 , overnight at 37°C. The plates were then incubated with TBSa/CaCl 2 and 5% BSA for at least 2 h at room temperature to block nonspecific binding sites, followed by three washes with TBSa/CaCl 2 containing 0.1% BSA. Initially, 5 g/ml fibrillin-1 protein fragments spanning the length of the molecule were screened for binding to tropoelastin and MAGP-1 by incubating in TBSa/CaCl 2 for 3 h or overnight at 37°C. For detailed binding curves of the fibrillin-1 fragments that were found to bind, biotinylation was carried out before the addition at concentrations from 0 to 50 g/ml in TBSa/CaCl 2 for 3 h or overnight at 37°C. Plates were then washed three times before detection of bound fibrillin-1. Biotinylated ligands were quantified by incubating with 1:200 dilution of ExtraAvidin-peroxidase conjugate (Sigma) in TBSa/ CaCl 2 at room temperature for 10 -15 min. For nonbiotinylated fibrillin-1 binding experiments, bound ligand was detected by incubation with anti-His 6 antibody (Novagen) diluted 1:1000 in TBSa/CaCl 2 for 3 h followed by anti-mouse peroxidase-conjugated secondary antibody (Jackson ImmunoResearch) diluted 1:2000 in TBSa/CaCl 2 . For both methods, wells were then washed four times, and the color was developed using 40 mM 2,2Ј-azino-bis(3-ethyl-benzthiazoline-6-sulfonic acid) solution (Sigma) for ϳ10 -20 min at room temperature. Plates were read at a wavelength of 405 nm. Any nonspecific biotinylated MAGP-1 or fibrillin-1 binding was detected by blocking precoated wells before incubation with biotinylated ligand. For the antibody detection assay, nonspecific binding was detected using immobilized tropoelastin and antibody only. All assays were performed in triplicate and repeated at least twice to confirm observed results. We used biotinylation as it is a widely used and sensitive molecular detection system. In addition, where antibodies were available, we also monitored binding assays immunologically. In all cases, similar patterns of binding were recorded, so biotinylation did not affect the binding interactions.
Competition Binding Assays-Competition binding assays were also conducted using both nonbiotinylated and biotinylated soluble proteins. Flat bottomed microtiter plates were coated with tropoelastin at 5 g/ml in TBSa/CaCl 2 , overnight at 37°C. Nonspecific binding sites were then blocked with TBSa/CaCl 2 containing 5% BSA at room temperature for at least 2 h. The plates were washed three times with TBSa/CaCl 2 and incubated with 5 g/ml nonbiotinylated MAGP-1 or fibrillin-1 fragments in TBSa/CaCl 2 overnight at 37°C. The plates were then washed again three times, and then 5 g/ml appropriate second biotinylated protein in TBSa/CaCl 2 was added overnight at 37°C. Control wells with the first nonbiotinylated soluble ligand not added were incubated in TBSa/CaCl 2 only. After a further three washes, plates were incubated with 1:200 dilution of extravidin peroxidase conjugate at room temperature for 15 min. Bound protein was quantified after four more washes by the colorimetric assay described above, using 2,2Ј-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) solution for 10 -20 min at room temperature. Plates were read at a wavelength of 405 nm. All experiments were done in quadruplicate wells.
MAGP-1 binding analysis was performed for fibrillin-1 protein fragments encompassing full-length fibrillin-1, as analytes, at two concentrations (40 and 200 nM) and a flow rate of 20 l/min for 3 min. After 5 min of dissociation, the chip was regenerated in HBS-Ca containing 400 mM NaCl for 60 s. The chip was then stabilized for 20 min using HBS-Ca before the next injection was carried out. The analyte was simultaneously passed over a blank capped flow cell, and this base line was subtracted from the experimental flow cell. The maximum relative response value for each injection was calculated using the binding assay result wizard (Biacore control software 3.0).
Kinetic analysis was then performed using the MAGP-1 bound chips and the fibrillin-1 fragments PF1 and PF4 that were shown in analyte

Fibrillin-1 Interactions with Tropoelastin and MAGP-1 23750
screening to bind MAGP-1. PF9 and PF12 were also tested as negative controls. Protein fragments were injected at concentrations ranging from 0.25 to 25 g/ml at a flow rate of 30 l/min. Samples were injected for 3 min, dissociated for 15 min, regenerated for 1 min using HBS-Ca containing 400 mM NaCl, and then stabilized for 20 min before the next injection. After subtraction of each response value from the blank cell, association and dissociation rate constants were determined by separate k a /k d fitting to the steepest parts of the binding and dissociation curves. All curves were fitted using 1:1 Langmuir association/dissocia-tion model (BIAevaluation 3.0; Biacore AB). A saturation binding curve was also plotted by performing a four-parameter mathematical fit to calculate the top of the sensorgram curve at infinite time (R i ); this was possible because all of the sensorgrams leveled off before the injection end. The R i value was then plotted against concentration, and the K D and B max values could be calculated using nonlinear regression (one-site binding (hyperbola) GraphPad Prism version 2.0).
To investigate the effects of EDTA on binding of PF1 to MAGP-1, kinetic analysis was also performed using 40 nM PF1 and 0.01-10 mM FIG. 2. Solid phase binding assays of soluble fibrillin-1 fragments to immobilized tropoelastin. Tropoelastin was coated to the plastic surface of multiwell plates at 5 g/ml and incubated with fibrillin fragments at 5 g/ml at 37°C. The wells were then blocked using 4% milk in TBSa/CaCl 2 . A, binding of fibrillin-1 fragments spanning the length of the molecule to tropoelastin-coated wells. B, binding of fibrillin-1 fragments PF2 and PF3⌬pro and truncated variants to tropoelastin-coated wells. C, binding of fibrillin-1 fragments PF2 and truncated PF2 to tropoelastin-coated wells. D, binding of fibrillin fragments PF2, PF5, PF7, and PF13 to tropoelastin with (gray) or without (black) incubation with 10 mM EDTA. The negative control was incubated with buffer only. Detection of fibrillin-1 fragments was performed using an anti-His 6 antibody. Results are shown as the mean Ϯ S.E. of triplicate values. Extremely significant difference from negative control is shown as follows: ***, p Ͻ 0.001 (unpaired t test).

Fibrillin-1 Interactions with Tropoelastin and MAGP-1
EDTA in HBS-Ca. An inhibition curve was then calculated using the R i value of each curve, as described above. Similar experiments were conducted to assess the potential effects of EDTA on binding of PF5 and PF7 to tropoelastin. To determine whether direct binding of MAGP-1 to the chip influenced its ability to bind fibrillin-1, MAGP-1 was also bound to the surface of a CM5 sensor chip, at saturating levels, using an immobilized polyclonal antibody (PR315; Elastin Products Inc.). Fibrillin-1 fragments were passed over the antibody-bound MAGP-1 at concentrations of 200 and 1000 nM, as described for the directly immobilized MAGP-1.
Using CM5 chips with tropoelastin immobilized, kinetic analysis of tropoelastin interactions were performed using all of the fibrillin-1 fragments and MAGP-1, as analytes, at concentrations ranging from 0.5 to 60 g/ml at a flow rate 30 l/min. Samples were injected for 3 min and dissociated for 15 min, as described for MAGP-1, except that the regeneration step used two 30-s injections of 1 M NaCl, 50 mM NaOH, and then stabilization was for 20 min before the next injection. The blank cell response was subtracted, and association and dissociation rate constants were calculated separately, as described for MAGP-1.
Cross-linking of Tropoelastin and Fibrillin-1 Using Transglutaminase-1 g of tropoelastin was incubated with 1 g of fragments scanning the entire fibrillin-1 molecule, with and without 0.1 g of transglutaminase (guinea pig liver; Sigma) for 2 h at 30°C. This was performed in a volume of 10 l to give protein concentration of 2 M for both fibrillin and tropoelastin. Samples were then resolved by SDS-PAGE and transferred onto nitrocellulose. Tropoelastin was detected using a mouse monoclonal antibody BA-4 (Sigma).
Mass Spectrometry Analysis of Transglutaminase-cross-linked Tropoelastin-PF2-To analyze the cross-linked complexes of tropoelastin and fibrillin-1 by mass spectrometry, 10 g each of fibrillin-1 fragment PF2 and tropoelastin were incubated with 1 g of transglutaminase in a volume of 30 l (12.5 M for each protein) for 2 h. Ni 2ϩ -chelate resin was then added in order to purify the PF2 and PF2-tropoelastin heterodimers and larger cross-linked complexes from the mixture. The resin was then washed twice in 200 l of 50 mM Tris, pH 7.4, 150 mM NaCl, 10 mM CaCl 2 (TBS/CaCl 2 ) followed by centrifugation to remove unbound noncross-linked tropoelastin and the transglutaminase; bound fibrillin-1 fragments, with or without cross-linked tropoelastin, were then eluted in 8 M urea, 400 mM imidazole. Samples were reduced by the addition of 9 mM dithiothreitol and incubated at 50°C for 30 min, followed by alkylation at room temperature by the addition of 20 mM imidazole for 15 min. Samples were then resolved using SDS-PAGE. Bands were visualized by colloidal Coomassie stain (Sigma), excised using a clean scalpel, and cut into 1-mm sections. The gel slices were washed twice in 200 l of H 2 O, followed by dehydration using 200 l of 50% acetonitrile. The slices were quickly washed in 40 mM ammonium bicarbonate before the addition of 20 l of 40 mM ammonium bicarbonate, 0.2 g of sequencing grade trypsin (Promega). The gel slices were incubated at 37°C overnight. For digestion using chymotrypsin after dehydration, gel slices were washed and then incubated in 100 mM Tris, pH 7.8, 10 mM CaCl 2 , 0.2 g of chymotrypsin (Roche Applied Science) and incubated at 25°C overnight. Digestion using endoproteinase Asp-N was performed in 10 mM Tris/HCl, pH 7.5, 0.2 g of endoproteinase Asp-N (Roche Applied Science). After digestion and centrifugation, the digestion buffer supernatant was removed and saved, whereas further tryptic peptides were extracted twice more in 20 l of 70% acetonitrile for 15 min. All supernatants were then pooled, and the acetonitrile was removed by speed vacuum, until the volume was reduced to 10 -20 l. 0.1% formic acid was then added, prior to analysis using a Q-TOF mass spectrometer (Micromass). Samples were also desalted using ZipTip C18 pipette tips (Millipore) prior to analysis by MALDI-TOF mass spectrometry.
To localize further the fibrillin-1 sequence within the PF2 fragment that binds to tropoelastin, deletion constructs were prepared (Fig. 2, B and C). PF2 fragments that lacked the proline-rich region (PF2⌬pro) or were truncated at the C terminus with removal of TB2 and with the proline region present or absent (PF2trunc and PF2⌬protrunc, respectively) (see Fig.  1, A and C) were used in solid phase binding assays. Binding of PF2⌬pro to tropoelastin was markedly decreased relative to the binding of PF2 to tropoelastin but was not ablated. However, neither PF2trunc or PF2⌬protrunc bound tropoelastin. Interestingly, PF3 (a larger fibrillin-1 fragment that encompasses PF2 and N-terminal upstream sequence encoded by exons 1-8; see Fig. 1) did not bind tropoelastin, but PF3⌬pro  Fibrillin-1 Interactions with Tropoelastin and MAGP-1 bound weakly to tropoelastin. These experiments show that fibrillin-1 TB2 (encoded by exons 16 and 17) is essential for tropoelastin binding to fragment PF2, that the proline-rich region influences but is not essential for binding, and that the additional presence of the N-terminal region (encoded by exons 1-8) completely inhibits tropoelastin binding. Any possible requirement for calcium in the strong interactions detected between fibrillin-1 fragments PF2, PF5, PF7, and PF13 with tropoelastin was tested in further solid phase assays. There were no apparent differences in the binding of these fibrillin-1 fragments to tropoelastin in the presence or absence of 10 mM EDTA (Fig. 2D), indicating that tropoelastin binding to these fibrillin-1 sequences is calcium-independent.
Biacore 3000 Analysis of Fibrillin-1 Binding to Tropoelastin-In order to compare the solid phase binding interactions with those detected by surface plasmon resonance and to determine binding affinities, potential binding of fragments covering the entire fibrillin-1 molecule to tropoelastin was investigated using surface plasmon resonance on a Biacore 3000 instrument. In these experiments, tropoelastin was immobilized on a CM5 sensor chip (12,100 resonance units), and soluble fibrillin-1 was the analyte. Kinetic analysis was performed using all fibrillin-1 fragments at a range of concentrations (0.5-60 g/ml) on the tropoelastin-immobilized chip. The association (k a ) and dissociation rate constants (k d ) of the binding interactions of fibrillin-1 fragments to tropoelastin and the dissociation constant (K D ) were determined for all fragments that exhibited any significant interaction. The kinetic data are summarized in Table I. Both PF5 and PF7 interacted very

Fibrillin-1 Interactions with Tropoelastin and MAGP-1
strongly with tropoelastin, with similar kinetic values (Fig. 4, A  and B). The average K D of three separate binding experiments for both fragments, using the BIAevaluation software analysis, was 4.1 Ϯ 0.6 and 5.0 Ϯ 1.5 nM, respectively. A three-domain fragment, PF6, which includes the two overlapping domains of PF5 and PF7, also bound tropoelastin strongly, albeit with lower affinity (K D of 151 Ϯ 18 nM) (Fig. 4C). PF13 (furincleaved) also bound tropoelastin, with lower affinity, with a K D of 51.9 Ϯ 17.4 nM. Fragment PF2, which bound strongly in solid phase assays, exhibited very low binding by surface plasmon The boxes indicate monomer (a), heterodimer (b), and multiple higher order species of PF2-tropoelastin that were subsequently analyzed by mass spectrometry. C, mass-spectrometry of PF2 from protease-digested peptides from gel bands. The tryptic peptides identified are highlighted in gray, and the peptides identified by chymotrypsin and endoproteinase Asp-N are shown underlined. The TB domains are shown in red, the prolinerich region is shown in green, and EGF/ cbEGF-like regions are shown in blue. The Asn from the putative N-linked glycosylation consensus sequence is shown in red boldface type, and the exon boundaries and corresponding exon numbers are indicated below the sequence. D, models of potential glutamine residues involved in transglutaminase cross-linking with tropoelastin, as determined by mass spectrometry analysis of proteolytic peptides. D, TB2 modeled on the known structures 1APJ (35) and 1KSQA (36). The potential cross-linking glutamines Gln 697 , Gln 702 , and Gln 708 are indicated.

Fibrillin-1 Interactions with Tropoelastin and MAGP-1
resonance, with a K D of 279 Ϯ 125, which may reflect differences in presentation of bound tropoelastin.
Similar Biacore binding experiments were conducted in the presence of up to 10 mM EDTA in order to determine whether the binding to PF5, PF6, and PF7 was calcium-dependent. Binding to these fragments was unaffected by EDTA (data not shown).
Biacore 3000 Analysis of MAGP-1 Binding to Tropoelastin-Kinetic analysis was performed using MAGP-1 as analyte at a range of concentrations (0.5-60 g/ml) and a tropoelastin-immobilized chip (Fig. 4D). The association (k a ) and dissociation rate constants (k d ) were determined using separate fits. The average dissociation constant (K D ) of the interaction between MAGP-1 and tropoelastin was 22.1 Ϯ 7.1 nM (Table II).
Transglutaminase Cross-linking of Fibrillin-1 to Tropoelastin-Co-incubation experiments were conducted between each of the fragments spanning the entire fibrillin-1 molecule with tropoelastin, in the presence or absence of transglutaminase, in order to determine whether any specific intermolecular transglutaminase-cross-linked complexes were formed. SDS-PAGE revealed that tropoelastin and PF2 formed unique nonreducible complexes, mainly heterodimers but also some higher order species (Fig. 5, A and B). None of these bands was present in transglutaminase-treated PF2 or tropoelastin alone. The higher order species became more prominent after longer incubations. No other fibrillin-1 fragments formed cross-linked complexes with tropoelastin.
Mass spectrometry analysis was conducted, after trypsinization, on the unique cross-linked heterodimers that had been purified by incubation with Ni 2ϩ -chelate resin. The His 6 tag on PF2 bound the resin, thereby facilitating removal of transglutaminase and uncross-linked tropoelastin. Both tropoelastin and PF2 tryptic peptides were detected in the heterodimers. No tropoelastin was seen in the tropoelastin-only control samples after Ni 2ϩ -chelate resin purification. Thus, the tropoelastin present in the transglutaminase-treated PF2 and tropoelastin preparations was cross-linked to PF2.
The tryptic peptides detected in the heterodimers indicate fibrillin-1 and tropoelastin sequences that are not involved in cross-linking, because, if they were, they would no longer have masses recognized by the data bases. Cross-linked fragments may also be too large for effective fragmentation and MALDI-TOF or Q-TOF analysis. Two exceptions are any tryptic peptides with carbohydrate attached to the NVT N-glycosylation consensus sequence at the end of the proline-rich region in PF2, which would not be detected due to increased mass, and the proline-rich region, which is not cleaved by trypsin. 2 Peptides were also generated from PF2-tropoelastin heterodimers by cleavage using chymotrypsin and endoproteinase Asp-N in order to generate different fragments. A drawback with these latter enzymes is that peptides generated are less detectable by mass spectrometry, since they do not have a charged Arg or Lys at their C termini, which would facilitate flight in the mass spectrometer.
Tryptic fragments detected by mass spectrometry of the PF2 monomer band alone accounted for 49% of the sequence of this fragment, and the PF2-tropoelastin heterodimer band also gave 49% coverage of the PF2 sequence (Fig. 5C). Tropoelastin, which is lysine-rich, is likely to contribute the lysine to the cross-link because all of the lysine residues in PF2 encoded by exons 12-17 are either included within peptides recognized by mass spectrometry or trypsin cleavage immediately after them; fibrillin-1 fragments encoded by exons 9 -11 can be excluded, because the fibrillin-1 fragment encoded by exons 1-11 does not form a cross-linked product with tropoelastin. No peptides encoding fibrillin-1 glutamines in exon 12 (QGSYCQ) and exon 17 (QPCPAQNSAEYQ) were detected in the heterodimers, so one or more of these glutamine residues are strong candidates for the cross-linking site (Fig. 5C). Since tropoelastin binding to PF2 is abolished when TB2 (encoded by exons 16,17) is deleted (Fig. 2, B and C), the cross-link is likely to be within TB2. Taken together, the data show that one or more glutamines encoded by exon 17 are cross-linked. Molecular modeling revealed that they are on the TB2 domain surface (Fig. 5D).
Solid Phase Binding of Fibrillin-1 to MAGP-1-Solid phase assays were used to identify any regions of fibrillin-1 that bind to MAGP-1. In initial experiments, wells were coated with 5 g/ml recombinant fibrillin-1 fragments and then incubated with 5 g/ml soluble recombinant MAGP-1. Fibrillin-1 fragment PF1 bound most strongly to MAGP-1, followed by PF5 and then PF7, but PF2 bound only weakly (Fig. 6, A and B). No other region of fibrillin-1 bound to MAGP-1. PF1 and PF5 both bound MAGP-1 in a dose-dependent and saturating manner (Fig. 6C).
MAGP-1 Biacore Analysis-Analyte screening of fibrillin-1 protein fragments spanning the length of the molecules for binding to MAGP-1 was performed using surface plasmon resonance (Biacore 3000). Fibrillin-1 fragments, at concentrations of 40 and 200 nM, were passed over MAGP-1, which was immobilized on the CM5 sensor chip. Both concentrations gave the same pattern of response (Fig. 7, A and B). Only the protein fragments PF-1, PF3, and PF4, all of which are at the N 2 S. A. Cain and C. M. Kielty, manuscript in preparation.
FIG. 6. Solid phase binding assay of soluble MAGP-1 to immobilized fibrillin-1 fragments. Fibrillin-1 fragments were coated to the plastic surface of multiwell plates at 5 g/ml, and the negative control was incubated with BSA at 5 g/ml. Incubation with biotinylated MAGP-1 at 5 g/ml followed as outlined under "Experimental Procedures." A, fibrillin-1 fragments spanning the N-terminal region of the molecule. B, fibrillin-1 fragments spanning the C-terminal region of the molecule. C, fibrillin-1 fragments PF1 (f), PF5 (OE), and BSA (ࡗ) were coated to the plastic surface of multiwell plates at 5 g/ml and incubated with biotinylated MAGP-1 at 0 -25 g/ml. Detection of biotinylated MAGP-1 was performed using streptavidin-horseradish peroxidase. Results are shown as the mean Ϯ S.E. of triplicate values. Extremely significant difference from negative control is shown as follows: ***, p Ͻ 0.001 (unpaired t test).
The association (k a ) and dissociation rate constants of these MAGP-1 and fibrillin-1 interactions and the dissociation constant (K D ) were determined by surface plasmon resonance. The kinetic data from these experiments are summarized in Table  II. Both PF1 and PF4 interacted with MAGP-1 with similar kinetic values. The K D of PF1 using the BIAevaluation software was 169 Ϯ 20 nM, and the K D value by saturation binding curve analysis was 242 Ϯ 24 nM, whereas the K D of PF4 was 171 Ϯ 44 and 139 Ϯ 39 nM. These data confirm that the main fibrillin-1 binding site for MAGP-1 is located within PF 4 (exons 1-8). Interestingly, no interaction was seen with PF5, although this interaction was observed in solid phase assays, possibly due to differences in presentation of bound MAGP-1 on chips or plastic wells.
To investigate whether a different MAGP-1 immobilization method would allow an interaction with PF5 using the Biacore, MAGP-1 was presented in saturating amounts on the sensor chip surface via an immobilized polyclonal antibody to MAGP-1. Fibrillin-1 fragments were then passed over the sensor chip at concentrations of 200 and 1000 nM. Using this method, some interaction was seen with PF5, but this was quantitatively less than that seen with PF1 (not shown).
To investigate the effects of EDTA on the binding of PF1 to MAGP-1, kinetic analysis was performed using 40 nM PF1 and 0.01-10 mM EDTA in HBS-Ca. An inhibition curve calculated using the R i value of each binding curve showed strong inhibition by EDTA with an IC 50 of 1.5 mM (not shown).
Competition Binding Assays-Having shown that fibrillin-1 fragment PF2, but not MAGP-1, specifically binds to tropoelastin, and since it has previously been reported that MAGP-1 binds to tropoelastin (10), we used competition binding assays to investigate whether MAGP-1 and PF2 have overlapping or separate binding sites for tropoelastin (Fig. 8). In these experiments, tropoelastin was first immobilized on the wells and then preincubated with MAGP-1 or PF2, prior to incubation with biotinylated PF2 or with MAGP-1, respectively. Biotinylated PF2 bound similarly to tropoelastin and to tropoelastin that been preincubated with MAGP-1. MAGP-1 bound simi-

Fibrillin-1 Interactions with Tropoelastin and MAGP-1
larly to tropoelastin and to tropoelastin that had been preincubated with PF2. These data showed that PF2 and MAGP-1 have separate binding sites on tropoelastin. DISCUSSION The molecular basis of how fibrillin microfibrils act as a template for tropoelastin during elastic fibrillogenesis is a critical yet still poorly understood process in the assembly of functional elastic tissues (2). In this study, we have screened the entire fibrillin-1 molecule for sites of interaction with tropoelastin and the major microfibril-associated molecule MAGP-1. Delineation of these interactions and their kinetics has provided important new insights into this complex process. We also addressed the major issue of how tropoelastin is stabilized on microfibrils following an initial molecular association. A novel specific transglutaminase cross-link between fibrillin-1 and tropoelastin was identified that could covalently stabilize newly deposited tropoelastin on microfibrils. Identification of the high affinity binding of tropoelastin to fibrillin-1 and the cross-link thus fill a major gap in our understanding of elastic fiber formation. These data, together with our previously determined structure of fibrillin-rich microfibrils (28), provide the basis for an updated model of elastic fiber assembly (Fig. 9).
The identification of novel high affinity interactions between tropoelastin and two overlapping fibrillin-1 fragments (PF5 and PF7) encoded by central fibrillin-1 sequences (exons 18 -25 and 24 -30) confirms this region of fibrillin-1 as the major tropoelastin binding site. Binding of tropoelastin to PF5 and PF7 is calcium-independent, suggesting that the binding site(s) may be in TB3, which is present in both fragments. Indeed, a smaller fragment (PF6) that includes the overlapping TB3 and flanking cbEGF domains was also found to bind tropoelastin strongly in a calcium-independent manner. Its relatively weaker affinity than PF5 implies that upstream cbEGF domains stabilize the TB3 conformation and possibly also provide additional weaker interactions. SDS-PAGE analysis revealed that the PF6 band was less compact than the longer PF5 band, which contains the PF6 sequence, implying that the threedomain fragment conformation was less ordered than that of the longer PF6 (Fig. 1B). Additional evidence comes from our cell adhesion studies of the homologous RGD-containing fibril-lin-1 TB4 motif. TB4 binds cells through integrins, but where TB4 is the first domain, the binding is much less than in fragments where the TB4 is preceded by six cbEGF domains (22). 3 We previously showed, from three-dimensional microfibril structure analysis and antibody epitope mapping, that a fibrillin-1 region encoded by exons 16 -22 locates to an exposed interbead "shoulder" feature comprising two prominent arms emerging from the bead. Our microfibril model from these data predicts that the domains encoded by exons 18 -25 and 24 -30 contribute to this interbead feature (28). We previously showed that mass accretes onto the interbead "shoulder" in developing elastic tissues (29), presumably reflecting tropoelastin deposition. Mutations that cause severe neonatal Marfan syndrome (6) map within this central fibrillin-1 region, so any of these mutant molecules that are secreted and assembled could affect tropoelastin deposition.
A further interaction of moderate affinity between tropoelastin and a fibrillin-1 fragment (PF2; encoded by exons 9 -17) that includes the proline-rich region has been identified by solid phase assays, although only a weak interaction was detected by Biacore. This interaction confirms an earlier report (14). Binding of tropoelastin to this fragment required the presence of the eight-cysteine motif TB2, so the binding site must be within this motif. The additional presence of N-terminal domains (encoded by exons 1-8) (PF3 fragment) ablates binding, so these domains must mask this binding site, possibly as a consequence of folding at the proline-rich region and direct interactions with TB2. Since removal of the proline-rich region (PF2⌬pro) does not ablate tropoelastin binding, it cannot be the primary interaction site. However, the reduced binding to this deletion fragment suggests long range conformational changes. Since deletion of the proline-rich region from the longer fragment (PF3⌬pro) does not fully rescue tropoelastin binding, there must be other upstream sequence effects. The physiological relevance of the tropoelastin interaction with C-terminal PF13 is unclear, since this fibrillin-1 sequence is almost certainly embedded within the bead (23, 28). Tropoelastin was coated to the plastic surface of multiwell plates at 5 g/ml, and then, after blocking with BSA, either unbiotinylated MAGP-1 or fibrillin-1 fragment PF2, respectively, was prebound at 5 g/ml before the addition of biotinylated PF-2 or MAGP-1, respectively (shown in gray). These wells were compared with other wells where no unbiotinylated protein was added after BSA blocking but before the addition of biotinylated protein (shown in black). The direct binding of biotinylated MAGP-1 to PF2 is also shown. Results are shown as the mean Ϯ S.E. of triplicate values.
FIG. 9. Model of elastic fiber formation in the extracellular space. I, microfibrils form from linear and lateral assembly of furinprocessed fibrillin-1. II, MAGP-1 associates with microfibrils at the bead surface. III, tropoelastin is deposited on microfibrils at the interbead "shoulder" feature adjacent to the beads. Tropoelastin is then cross-linked to microfibrils by transglutaminase. Tropoelastin and MAGP-1 may then interact on microfibrils. IV, further accretion of tropoelastin to microfibril-bound tropoelastin and MAGP-1 and then lysyl oxidase cross-linking. V, formation of mature elastic fiber. Predicted molecular folding not shown (28).

Fibrillin-1 Interactions with Tropoelastin and MAGP-1
We also identified a new specific transglutaminase cross-link between tropoelastin and fibrillin-1 that reinforces the critical importance of high affinity tropoelastin binding to the central region of fibrillin-1. Transglutaminases catalyze the formation of an isopeptide ␥-glutamyl-⑀-lysine bond either between or within polypeptide chains. Tissue transglutaminase (tTG, TG-2, TG C ) is important in stabilizing developing extracellular matrices including lung and bone and in wound repair (30,31). The cross-link site on fibrillin-1 was localized to three glutamines within the fibrillin-1 TB2 domain (exon 17), which are on the domain surface. This cross-link may covalently link newly bound tropoelastin on microfibrils during early elastic fiber formation. Other cross-links within microfibrils and between elastin molecules within the elastin core have previously been described, but hitherto it has not been known how newly deposited tropoelastin becomes stably associated with microfibrils. Lysyl oxidases catalyze the formation of lysyl-derived desmosine and isodesmosine cross-links between elastin molecules within the core of elastic fibers (32). Microfibrils are extensively stabilized by disulfide bonds, such that reducing conditions are needed to extract fibrillin-1 molecules from fetal issues (8). A fibrillin-1 transglutaminase-cross-linked fragment of two fibrillin-1 peptides with sequences starting at residues 580 (encoded within exon 14) and 2312 (encoded within exon 56) was identified in the interbead fraction of amnion microfibrils (33). MAGP-1 is also a substrate for transglutaminase (11), so it is possible that similar cross-links stabilize MAGP-1 on microfibrils. The novel transglutaminase cross-link between fibrillin-1 and tropoelastin defines a critical new elastic fiber role for this enzyme.
The principal MAGP-1 binding site is on fibrillin-1 sequence encoded by exons 1-8 (PF1), which fits a previous report (10) and its immunolocalization to microfibril bead surfaces (9,34). Since binding is strongly calcium-dependent, the only two cbEGF-like domains (exons 7 and 8) present in this fragment are implicated. Linear fibrillin-1 assembly involves interactions between furin-processed N and C termini, both of which are at the beads (23,28). MAGP-1 may bind a sequence close to the fibrillin-1 N-terminal sequence during or after microfibril assembly and contribute to the beads. Solid phase assays revealed that MAGP-1 can also bind overlapping PF5 and PF7 fragments, although binding is weaker than to the fibrillin-1 N-terminal region and is barely detectable by Biacore. These MAGP-1 interactions with central fibrillin-1 domains are also of much lower affinity than tropoelastin binding to the same fragments. Thus, tropoelastin binding to the exposed fibrillin-1 interbead sequence encoded by exons 18 -30 will take precedence. Further accretion of MAGP-1 and tropoelastin may occur on these microfibril-bound molecules, and MAGP-1 multimers might present an ensemble of binding faces that could further draw together nascent elastic fiber components.
In summary, we propose an updated model of elastic fiber formation based on our new data (Fig. 9). Microfibrils are first formed by linear and lateral fibrillin-1 interactions. MAGP-1 then accretes onto microfibril bead surfaces via an interaction with N-terminal fibrillin-1 domains. Tropoelastin is then deposited on an interbead region adjacent to the beads through strong interactions with the fibrillin-1 central sequence and subsequently becomes cross-linked to fibrillin-1. Juxtaposition of MAGP-1 and tropoelastin on the microfibril scaffold allows the possibility in situ of a ternary complex formation between the three molecules. This model thus provides a molecular explanation for the initial stages of elastic fiber formation.