Heparan Sulfate Regulates Fibrillin-1 N- and C-terminal Interactions*

Fibrillin-1 N- and C-terminal heparin binding sites have been characterized. An unprocessed monomeric N-terminal fragment (PF1) induced a very high heparin binding response, indicating heparin-mediated multimerization. Using PF1 deletion and short fragments, a heparin binding site was localized within the domain encoded by exon 7 after the first hybrid domain. Rodent embryonic fibroblasts adhered to PF1 and deletion fragments, and, when cells were plated on fibrillin-1 or fibronectin Arg-Gly-Asp cell-binding fragments, cells showed heparin-dependent spreading and focal contact formation in response to soluble PF1. Within domains encoded by exons 59–62 near the fibrillin-1 C terminus are novel conformation-dependent high affinity heparin and tropoelastin binding sites. Heparin disrupted tropoelastin binding but did not disrupt N- and C-terminal fibrillin-1 interactions. Thus, fibrillin-1 N-terminal interactions with heparin/heparan sulfate directly influence cell behavior, whereas C-terminal interactions with heparin/heparan sulfate regulate elastin deposition. These data highlight how heparin/heparan sulfate controls fibrillin-1 interactions.

Fibrillin-1 is a large glycoprotein that assembles pericellularly into microfibrils that have a complex recurring bead construction and are widespread in elastic tissues, such as skin, lung, arteries, and ligaments (1,2). It contains 47 epidermal growth factor-like domains, 43 of which bind calcium, which are interspersed with 8-cysteine-containing TB modules (3). N and C termini undergo pericellular processing by proprotein convertases, which are considered to be a prerequisite for subsequent assembly involving N-and C-terminal interactions (4).
It is increasingly clear that fibrillin-1 is a major extracellular heparin/heparan sulfate-binding molecule (6 -8). We identified four high affinity heparin-binding regions on fibrillin-1 and partially localized three of these sites (6). Fibrillin-1 is likely to interact with heparan sulfate in the form of the cell surface heparan sulfate proteoglycans syndecans or glypicans (9). Basement membrane perlecan also interacts with fibrillin-1, mainly through protein-protein interactions with a central fibrillin-1 region but also through lower affinity interactions with perlecan heparan sulfate chain (10).
In this study, we have mapped an N-terminal fibrillin-1 terminal heparin binding site that induces heparin-inhibitable cellular responses. We have also identified a novel high affinity C-terminal heparin binding site that competes with tropoelastin. These data highlight the importance of heparan sulfate for fibrillin-1 interactions with cells and elastic fiber molecules.
Gel filtration chromatography was used to ensure that only monomeric recombinant fragments were used in all experiments. Recombinant fragments were concentrated after elution from nickel affinity chromatography, using Vivaspin columns (Sartorius), prior to separation using Superdex 200 10/300 GL (GE Healthcare), using an AKTA Purifier system. Monomeric protein was pooled for further experimentation. All recombinant fragments had the correct molecular mass as determined by SDS-PAGE. Fragments with N-glycosylation sites were N-glycosylated, as determined by treatment with PNGase F (New England Biolabs). In some experiments, PF12, PF13, and PF16 were pretreated with trypsin for 10 -120 min before SDS-PAGE analysis.
Our initial PF16-expressing 293-EBNA cultures produced two bands (see Fig. 3A), which were shown by mass spectrometry to correspond to intact PF16 plus a large cleavage fragment lacking some C-terminal sequence (designated PF16 trunc ). PF16 was separated from PF16 trunc using S200 size fractionation. Some transfected 293-EBNA cultures expressed recombinant PF16 trunc only, as judged by SDS-PAGE, immunoblotting, and mass spectrometry. Both PF16 trunc preparations gave similar results in binding assays and solution studies.
Analysis of PF16 by Analytical Ultracentrifugation-Sedimentation velocity experiments were performed, as reported (15), on PF12 (4.5 M), PF13 (3.4 M), and PF16 (9.1 M) in 0.15 M NaCl, 10 mM HEPES, 1 mM CaCl 2 , pH 7.4, and wavelength 230 nm, using an Optima XL-A analytical ultracentrifuge (Beckman Coulter, Inc.). Analyses of the raw data were performed using the distribution of Lamm equation solutions software Sedfit version 9.4 (16). The correction of s 20,w from standard conditions along with an estimation of the frictional ratio (f/f 0 ) and hydrodynamic radius (R h ) was performed using Sednterp with a partial specific volume of 0.71 ml/g, as published previously (17,18). Hydropro (19) was used to derive shell-based bead models around the coordinates obtained from ab initio modeling from small angle x-ray scattering (20). This approach gave theoretical sedimentation coefficients and hydrodynamic radius measurements that were compared with those derived from analytical ultracentrifugation.
Samples eluting from the column passed through a Wyatt EOS 18-angle light scattering detector fitted with a 688-nm laser and an Optilab r-EX refractometer. The solute M r and R h values were determined using in-line multiangle laser light scattering attached to a (quasielastic) QELS and a differential refractometer (Wyatt Technology Corp.).
Analysis of the relationship between R h and solute molecular mass provided details of the size and shape of C-terminal fibrillin-1 and how these parameters are affected by fragment mass. For a single molecule, a rodlike conformation gives an approximated ␣ value of 1, whereas a compact symmetrical structure gives an approximated ␣ value of 0.33. For each fibrillin-1 mass value, the slope ␣ was calculated from the coordinates of the tangent fitted by regression analysis of the 10 surrounding mass values.
Modeling-Fibrilllin-1 oligomers were represented as an assembly of spheres using the equation set out by Bloomfield et al. (21), which allows the calculation of theoretical friction values (f) for molecules of known shape. Other molecules modeled with good correlation in this way include phycocyanin and fibrinogen (21).
Modeling was performed using the assembly of 2-nm diameter spheres, and the theoretical S values were calculated for measurements performed in 0.1 M NaCl, 50 mM Tris, 1 mM CaCl 2 , pH 8.0. We used 0.71 ml/g for the partial specific volume of fibrillin-1 (17,18). Fibrillin-1 dimensions used as constraints in our modeling were as follows: (i) fibrillin-1 domain widths of ϳ2 nm based on solution NMR and crystallography, electron microscopy, and previous modeling (17,22,23); (ii) 90-nm solution length of fibrillin-1 determined by small angle x-ray scattering (20); (iii) the 150 -180-nm extended length of fibrillin-1 as seen on mica (24); and (iv) the 50 -60-nm periodicity of isolated and tissue microfibrils (18,22,24,25). Calculated S values from the model were compared with experimental S values; the sphere arrangement in the model was modified accordingly and in all cases was limited by the above constraints and available structural information.
PF12 and PF13 were modeled using shapes provided by small angle x-ray scattering (20), as described (17). The calculated S values from the model were compared with experimental S values. Then PF12 and PF13 models were combined, taking into account their 3-nm overlap, which approximates to one TB domain and one cbEGF domain.
Biotinylation of Heparin Saccharide-The heparin used was a defined sized heparin saccharide dp24, which contains 24 monosaccharides (6) (kindly provided by Prof. J. Gallagher, University of Manchester). This fragment is a chemical analogue of the sulfated domains (S domains) of heparan sulfate. Biotinylation involved coupling via the heparin reducing end, in two stages (6,26,27). First, reductive amination with ammonia and then biotin was coupled to the free amines. Heparin species dp24 (0.1 mg) were dissolved in 2 M NH 4 Cl, in a volume of 100 l. 2 mg of NaCNBH 3 was added, and the mixture was heated at 70°C for 2 days. After cooling, the mixture was dialyzed extensively into Dulbecco's phosphate-buffered saline (Cambrex). 10 l of 3 mg/ml sulfosuccinimidyl-6-(biotinamido) hexanoate (Sulfo-NHS-LC-Biotin) (Pierce), was added and incubated

Fibrillin-1 Interactions with Heparin
overnight at 4°C. The nonreacted biotin was then removed by further dialysis into 0.1 M sodium acetate, pH 5.5.

BIAcore Analysis of Heparin Interactions with Fibrillin-1
Fragments-For kinetic binding studies of dp24 heparin saccharides with fibrillin-1 by surface plasmon resonance, a Biacore biosensor was used (Biacore 3000; GE Healthcare). Biotinylated heparin fragments were immobilized onto commercially prepared SA sensor chips, which have preimmobilized streptavidin, to allow biotin capture. Using heparin concentrations of 1 M, typically 150 -200 response units of biotinylated heparin samples were immobilized, which was at a saturation level. Samples were applied to the sensor chip surface in 0.1 M sodium acetate, pH 5.5. All subsequent binding experiments were performed in 10 mM HEPES, pH 7.4, 0.1 M NaCl, 1 mM CaCl 2 , and 0.005% surfactant P20 (designated HBS-Ca). Protein fragments were injected at concentrations ranging from 1 to 20 g/ml at a flow rate of 12 l/min. Samples were injected for 12 min and dissociated for 10 min, before the chip was regenerated using 5 mM NaOH, 1 M NaCl twice for 1 min each and then stabilized for 10 min using HBS-Ca, before the next injection. After subtraction of each response value from the blank cell, association and dissociation rate constants were determined by global data analysis. Initially, all curves were fitted using a 1:1 Langmuir association/dissociation model (BIAevaluation 4.1; GE Healthcare). This model was found to fit the data for all of the protein fragments very well, apart from PF1 and PF4, with low 2 values. 2 values are a standard statistical measure of the closeness of fit (mean square of the signal noise).
Because of very high binding responses, PF1 and PF4 binding to heparin was calculated independently using equilibrium analysis. The equilibrium response was plotted against concentration, and nonlinear regression using the equation for onesite binding used to calculate K D .
To investigate the kinetics of interactions between the fibrillin protein fragments and MAGP-1, MAGP-1 was immobilized on the surface of a CM5 sensor chip via amine coupling, as described (12); PF1 and deletion constructs Ex1F-11, Ex3-11, and Ex4 -11 were then injected at concentration ranges of 1-20 g/ml at a flow rate 12 l/min. To calculate the equilibrium constant, equilibrium response was plotted against concentration, in the same method as described above for heparin interactions. Each binding interaction was performed in triplicate.
Solid Phase Binding and Inhibition Assays-Flat bottomed microtiter plates were coated with tropoelastin or with PF12, PF13, or PF16 at 100 nM in 50 mM Tris, pH 7.4, 0.1 M NaCl, 1 mM CaCl 2 (TBS-Ca), overnight at 4°C, as described (12). Nonspecific binding sites were then blocked with TBS-Ca containing 4% BSA at room temperature for at least 2 h. The plates were washed three times with TBS-Ca, 0.1% BSA and incubated with 100 nM biotinylated PF12, PF13, or PF16, with or without preincubation with heparin in TBS-Ca, overnight at 4°C. 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 a colorimetric assay using 40 mM 2,2Ј-azino-bis(3-ethyl-benzthiazoline-6-sulfonic acid) solution (Sigma) for 10 -20 min at room temperature. Plates were read at a wavelength of 405 nm. Control wells with the first nonbiotinylated soluble ligand not added were incubated in TBS-Ca only. For both methods, wells were then washed four times, and the color was developed. Any nonspecific fibrillin-1 binding was detected by blocking wells before incubation with BSA only. All experiments were performed in triplicate, and all assays were repeated at least twice to confirm observed results.

Fibrillin-1 Interactions with Heparin
Cell Attachment Assays-Cell attachment assays were performed using 24-well plates coated with recombinant fibrillin-1 PF1 fragment, as previously reported (9,28). Bound cells were quantified by incubating with 15 M the fluorescent dye Calcein (Invitrogen) and then measuring in a Fluostar Galaxy plate reader (BMG Labtechnologies) at 485-nm excitation and 520-nm emission. Relative cell attachment was determined by comparison with known numbers of cells added to uncoated, unblocked wells. In all experiments, triplicate wells were used. Data were statistically analyzed using unpaired Student's t tests (GraphPad Prism 2.0). Error bars represent S.D. values of the three experiments. Results were statistically significant when p was Ͻ0.05 (*, p Ͻ 0.05; **, p Ͻ 0.001; ***, p Ͻ 0.0001).
Immunofluorescence Microscopy-Spread cells were fixed after 2 h or 2 h 30 min in 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Cells to be stained for focal adhesion kinase were fixed and permeabilized in ice cold methanol. Primary antibodies used were a monoclonal paxillin antibody Z035 (Zymed Laboratories Inc.), monoclonal anti-Tyr(P) antibody 4G10 (Upstate Biotechnology), monoclonal vinculin antibody hVin1 (Sigma), and polyclonal anti-focal adhesion kinase antibody BC3 (Upstate Biotechnology). Double fluorescence microscopy was performed using the described antibodies, followed by incubation with appropriate fluorochrome-conjugated secondary antibodies and phalloidin (Alexa Fluor 488/568; Molecular Probes, Inc., Eugene, OR). Samples were analyzed on an Olympus AX70 fluorescence microscope (Olympus; objectives, UPlanApo ϫ40, 1.0 numerical aperture oil iris, UPlanApo ϫ60, 1.4 numerical aperture

Fibrillin-1 Interactions with Heparin
oil). Images were captured with a SPOT Insight Mono digital camera and processed using Adobe Photoshop 7.0.
Focal plaque formation was analyzed using immunofluorescence microscopy. Human dermal fibroblasts, REFs (29), wildtype and syndecan-4 null MEFs, and CHOK1 and CHO761 cells, all plated on the fibrillin-1 ArgGlyAsp-containing (RGD) fragment PF8 or on fibronectin (FN) or the 110-kDa integrinbinding domain of fibronectin (110FN), were analyzed for focal plaques and actin filaments as reported (9,29), in the presence or absence of PF1, PF13, or hepII. Cells with five or more distinct vinculin containing plaques were deemed positive.

N-terminal Heparin Binding Sites
We previously reported that heparin binds with high affinity to the N-terminal fibrillin-1 fragment PF1 (6). Here, a series of PF1 deletion fragments has been generated to localize the heparin binding site(s) and define the binding kinetics (Fig. 1).
Heparin Binding to PF1 and Deletion Fragments-The association (k a ) and dissociation (k d ) rate constants of the molecular interactions between a polymeric heparin saccharide (dp24) and the fibrillin-1 fragments and their dissociation constants (K D ) were determined by surface plasmon resonance (Fig. 2, A and B, and Table 1). Heparin strongly bound to PF1 with a K D of 27.3 Ϯ 0.9 nM. It also interacted with all the PF1 deletion fragments, but some of these interactions had markedly lower affinities for heparin than PF1. The affinities of the deletion PF1 fragments for heparin ranged from 116 Ϯ 4.5 nM for the Ex1F-11 fragment to 248 Ϯ 32 nM for Ex5-11. A short form of PF1, designated PF4 (11), which lacks the proline-rich region and flanking domains, bound heparin with an affinity of 104 Ϯ 12 nM. Fragment PF2 overlaps PF1 at the proline-rich region and flanking domains but did not bind heparin.
Although binding kinetics for intact PF1 and PF4 could be calculated using global fitting and the 1:1 binding equation, the resulting fit was poor compared with fitting for all of the PF1 deletion fragments, the three-domain fragment, and PF2. To calculate the K D for the PF1-heparin interaction more accurately, equilibrium data were obtained by plotting a saturation binding curve (Table 1). By this method, the K D was calculated to be 27.3 Ϯ 0.9 nM, which is lower affinity but on the same order as 14.4 Ϯ 2.9 nM, which was calculated by global fitting.
The relative binding responses of all N-terminal fibrillin-1 fragments (PF1, PF1 deletion fragments, PF2, and PF4) to heparin were also plotted, for an analyte concentration of 10 g/ml, after 12 min of association (Fig. 2B). Both PF1 and PF4 had a highly elevated response compared with all of the other fragments, with response units of 5315 and 4250 compared with the PF1 deletion fragments, which had response values ranging from 624 for Ex1F-11 to 44 for Ex5-11.
The enhanced binding response of PF1 and PF4 to immobilized heparin was not due to PF1 being aggregated prior to BIAcore analysis. Multiangle laser light scattering and analytical ultracentrifugation results confirmed that the PF1 fragments were monomeric (not shown). PF1 differs from Ex1F-11 only by the additional 17-amino acid N-terminal sequence that precedes the proprotein convertase cleavage site (AGNVKET-RASRAKR), which thus contains a heparin-mediated fibrillin-1 multimerization signal.

TABLE 1 Analysis of interactions between heparin saccharides and fibrillin-1 protein fragments by surface plasmon resonance
Surface plasmon resonance binding was performed as described under "Experimental Procedures." The kinetic parameters were derived from typical sensorgrams shown in Fig. 2 Table 2). N-terminal heparan sulfate binding is thus likely to be a critical determinant of microfibril assembly and of MAGP-1 binding during subsequent elastic fiber assembly.
Heparin Binds a Three-domain N-terminal Region-Heparin binding to the PF1 deletion fragments and to PF4 indicated that a heparin binding site is present within the cbEGF domains encoded by exons 7 and 8. To confirm this localization, a recombinant fragment comprising the hybrid domain within PF1 and flanking cbEGF domains (see Fig. 1) was tested for heparin binding by BIAcore analysis ( Fig. 2A). Ex5-7 was found to bind with a high affinity of 6.0 Ϯ 1.8 nM (Table 1). Together, these data localize this heparin binding site within the cbEGF domain encoded by exon 7.

C-terminal Heparin Binding Sites
Heparin has been shown to bind to the last 17 residues of the C-terminally furin-processed fibrillin-1, which is encoded by exon 64 and ends with RKR (7). Here, we have identified a novel upstream high affinity fibrillin-1 heparin binding site within a new recombinant fibrillin-1 fragment, designated PF16, which is encoded by exons 50 -65 and comprises overlapping PF12 and PF13 fragments (6) (Fig. 1).
SDS-PAGE analysis of recombinant PF16 fragment revealed two bands with molecular mass values of 89 and 63 kDa (Fig. 3A). The lower mass band, designated PF16 trunc , had 26 kDa less mass than PF16. Mass spectrometry analysis of PF16 and PF16 trunc confirmed that the 89-kDa fragment corresponded to furin-processed PF16, since the last detected peptide was INGYPKR, which immediately precedes the RKRR furin cleavage site (3). The PF16 trunc fragment was C-terminally truncated, and the last detected PF16 trunc tryptic peptide was GFSLDQTGSSCED-VDECEGNHR, which is within the cbEGF encoded by exon 62.

Heparin-inhibitable PF1-mediated Cell Attachment and Spreading
Although PF1 does not contain known integrin adhesion motifs, REFs adhered to PF1 (Fig. 4A) and to all the PF1 deletion fragments (not shown). When REFs were plated on PF1, attachment was significantly inhibited by heparin (1-100 g/ml). Chondroitin sulfate had no effect on adhesion. REFs spread poorly on all of the PF1 fragments (Fig. 4B).
When plated on PF8, CHO761 mutant cells, which do not produce glycosaminoglycans, did not induce focal plaques in response to PF1 or hepII, unlike control CHOK1 cells (Fig. 4C). However, when we analyzed syndecan-4 null MEFs, it was evident that although syndecan-4 null MEFs on PF8 had poorly organized cytoskeletons, the addition of PF1 still induced some organized plaques comparable with wild-type MEFs (58 and 44%, respectively; Fig. S1). Interestingly, when syndecan-4 null MEFs were plated on PF8 and soluble hepII was added, the cells could hardly form any plaques (8%). Thus, although PF1-induced focal plaques on PF8 are heparin/heparan sulfate-inhibitable, perhaps more than one syndecan is involved in this process.
REFs plated on PF8 ( Fig. 5A; Fig. S2) or on 110FN (Fig. 5B) and incubated with PF1 deletion mutants were compared for focal plaques. Abundant plaques were detected on 110FN in the presence of all PF1 deletion fragments. For REFs on PF8, fewer

Fibrillin-1 Interactions with Heparin
adhesions were detected with Ex5-11, and none were detected with Ex6 -11 (Fig. S2). Similar results were obtained using wildtype MEFs (not shown). Thus, 110FN induces stronger spreading than PF8, and heparin/heparan sulfate-inhibitable effects of PF1 do not involve the N-terminal prefurin cleavage site sequence that induces a high binding response (see Fig. 2B) ( Table 1). Focal plaques induced for cells on PF8 by PF1 appeared less mature than fibronectin-mediated adhesions, since they contained vinculin (Fig. 5, A and B), paxillin (not shown), and some phospho-Tyr but little focal adhesion kinase (Fig. 5C).
Together, these experiments show that PF1 induces heparin/ heparan sulfate-inhibitable focal plaques, probably through the heparin binding site in cbEGF encoded by exon 7.

Heparin Inhibits C-terminal Interactions with Tropoelastin
BIAcore analysis of tropoelastin binding to fibrillin-1 revealed that PF16 bound tropoelastin very strongly, with a K D of 1.5 Ϯ 0.4 nM ( Fig. 6A and Table 3). PF16 trunc also bound immobilized tropoelastin with very high affinity (K D ϭ 1.7 Ϯ 0.3 nM). Solid phase binding assays allowed comparison of tropoelastin interactions with PF16, PF16 trunc , PF12, or PF13, using wells coated with tropoelastin (12.5-200 nM). PF16 interacted most strongly with tropoelastin, PF16 trunc , and PF13 showed similar strong levels of binding, but PF12 showed little binding (Fig. 6B). We also showed that REFs on PF8 in the presence of soluble PF13 formed focal plaques (Fig. 5D). These data extend our previous report that PF13 (but not PF12) binds tropoelastin (12); PF16 trunc allows us to localize the site within domains encoded by exons 59 -62 and also indicates that high affinity binding of tropoelastin is conformation-dependent.
Biotinylated PF16 (100 nM) pretreated for 30 min at room temperature with or without 100 nM heparin (Fig. 6C) was incubated in wells precoated with tropoelastin (12.5-200 nM). At equimolar concentrations, preincubation of PF16 with heparin significantly inhibited PF16 binding to tropoelastin. Thus, tropoelastin and heparin binding sites on PF16 may overlap.
The mass and shape of PF12, PF13, and PF16 were also determined biophysically (Table 4). PF12 and PF13 have similar R h values (3.7 and 3.4 nm, respectively), whereas PF16 has a higher FIGURE 6. Binding of C-terminal fibrillin-1 recombinant protein fragments to tropoelastin. A, Biacore analysis of the interactions of PF16 and PF16 trunc with tropoelastin. Fibrillin-1 fragments at a concentration of 1-5 g/ml were injected over tropoelastin-immobilized sensor chip surfaces. One representative experiment is shown. Response difference (Resp. Diff.) is the difference between experimental and control flow cells, in response units (RU). Time is shown in seconds (s). B, using solid-phase binding assays, tropoelastin (12.5-200 nM) was plated at 4°C prior to blocking with BSA and subsequent binding of biotinylated C-terminal fibrillin-1 fragments, as described under "Experimental Procedures." Nonspecific binding to BSA was subtracted from all data points. Results shown are the means Ϯ S.E. of triplicate values of a single experiment, with each experiment repeated at least twice. Binding of each fibrillin-1 fragment to tropoelastin (relative to BSA control) was highly significant, with unpaired Student's t tests at 100 nM tropoelastin all giving p Ͻ 0.0001. Nonlinear regression analysis of PF16 binding tropoelastin was calculated (see left panel), giving a K D value of 19.61 nM (R 2 ϭ 0.9848). Nonspecific binding to BSA was subtracted from all data points. Results are the means Ϯ S.E. of triplicate values of a single experiment, with each experiment repeated at least twice. C, solid phase assays revealed that heparin competes with PF16 to bind tropoelastin. Biotinylated PF16 at 100 nM was pretreated for 30 min at room temperature with 100 nM heparin, whereas a corresponding PF16 sample was left without heparin. These samples were incubated with preplated tropoelastin at increasing concentrations. A biotinylated heparin control was also included. Nonspecific binding was subtracted from all data points. Results are the means Ϯ S.E. of triplicate values of a single experiment, with each experiment repeated at least twice. ***, unpaired Student's t test at 200 nM tropoelastin shows that binding of PF16 to tropoelastin in the presence of heparin is significantly reduced compared with binding in the absence of heparin (***, p Ͻ 0.0001, R 2 ϭ 0.9909). OCTOBER 3, 2008 • VOLUME 283 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 27025
R h value of 5.1 nm. Furthermore, the frictional ratios (f/f 0 ) of PF12 and PF13 are 1.6 and 1.5, respectively, whereas the frictional ratio of PF16 is 1.8. Spherical structures have a frictional ratio of 1, and the value rises as the molecule elongates. PF12 and PF13 are relatively elongated in solution, but PF16 is even more elongated (Fig. 7). Bead modeling (21) predicts that PF16 has an S-shaped structure, with the binding region for heparin and tropoelastin at a loop adjacent to the PF12-PF13 overlap (Fig. 7). This result accords well with small angle x-ray scattering analysis, which showed nonlinear solution shapes for the smaller fragments, PF12 and PF13 (20).

DISCUSSION
Although it is now clear that fibrillin-1 is a major extracellular heparin/heparan sulfate binding molecule, the conse-quences of heparin/heparan sulfate binding are less apparent. Heparan sulfate may regulate fibrillin-1 multimerization and microfibril assembly, interactions with cell surfaces through syndecan or glypican receptors, and fibrillin-1 interactions with other elastic fiber molecules during elastic fibrillogenesis. Here, we analyzed heparin interactions with fibrillin-1 N-and C-terminal regions, which are considered to regulate linear assembly. We have defined novel N-terminal heparin binding sites at both termini, shown that heparin may support N-terminal multimerization, demonstrated that the PF1 fragment directs heparin-inhibitable cell spreading and focal plaque formation, and identified a novel high affinity, conformation-dependent tropoelastin binding site close to the C terminus that is inhibited by heparin.
Three regions of fibrillin-1 were originally shown to interact with heparin/heparan sulfate (7,8). Our previous studies identified four binding sites (6). Data in this study, combined with previous data, highlight that fibrillin-1 contains at least five high affinity heparin binding sites, confirming the critical role of heparan sulfate in fibrillin-1 biology. Mapping of N-terminal heparin binding sites revealed that the sequence preceding this convertase cleavage site plays a critical role in heparin binding. When present, monomeric N-terminal fragments exhibit a huge binding response to heparin-coated chips, indicating that this short sequence controls N-terminal fibrillin-1 multimerization. Thus, heparin-induced N-terminal multimerization and subsequent furin cleavage (which occurs extracellularly (4); the sequence preceding the cleavage site was not detected in mass spectroscopy of tissue microfibrils (30)) may control microfibril assembly. Interestingly, we mapped MAGP-1 binding to the adjacent unique region immediately after the cleavage site and preceding the first EGF-like domain (see Fig. 1), which accords well with reported MAGP-1 binding and competition with heparin (12, FIGURE 7. PF16 and PF16 trunc show conformational differences from PF12 and PF13. Shown is bead modeling of lateral and end-on views of PF12 and PF13 and PF16 in potential extended or S-shaped conformations; all lateral views have C-terminal regions on the right. Calculations predicting the shape of PF16 were performed using Hydropro software (19). Modeling was performed on coordinates derived from small angle x-ray scattering (20) to give a theoretical s 20,w 0 , which was compared with the measured s 20,w 0 from analytical ultracentrifugation (Table 4). Only the S-shaped conformation of PF16 fits the analytical ultracentrifugation data. The large gray oval in the bottom panel indicates the localized binding site for heparin and tropoelastin. Calculated M r was calculated from the amino acid sequence, including N-linked glycolysation on PF12 and PF16. Measured molecular mass was measured by multiangle laser light scattering. The measured s 20,w and the calculated R h were obtained by analytical ultracentrifugation. Calculated s 20,w was derived using bead modeling (21). s 20,w was also calculated for a fully extended bead model of PF16 (PF16 extended ). Fibrillin-1 Interactions with Heparin 31). N-terminal heparan sulfate binding is thus likely to be a critical determinant of fibrillin-1 polymerization during microfibril assembly and of MAGP-1 binding during elastic fiber assembly. Using deletion, overlapping, and short fragments, we have mapped a heparin binding site to the cbEGF encoded by exon 7, which contains three lysines. This site may bind cell surface heparan sulfate and induce a cellular response. Within assembled microfibrils, this N-terminal region may be at the beads (20). The ability of this N-terminal fibrillin-1 sequence to induce heparin-inhibitable focal plaques implies that it can associate with cell surface heparan sulfate proteoglycans. However, the efficient formation of N-terminal fibrillin-1-induced plaques in syndecan-4 null cells indicates the involvement of other syndecans or glypicans.

Calculated mass
High affinity heparin binding to a C-terminal site was localized within cbEGFs encoded by exons 59 -62, a region that contains seven lysines and two arginines. A high affinity C-terminal tropoelastin binding site identified within the same cbEGFs may be critical for elastic fiber formation. Since heparin competes with tropoelastin, these sites are probably overlapping. Heparan sulfate may thus regulate elastin deposition onto microfibrils.
This C-terminal tropoelastin binding site may be conformation-dependent, since the affinity of tropoelastin for PF16 is much higher than for PF13 (12). Bead modeling based on biophysical solution measurements predicted that the longer PF16 fragment has an S-shaped structure with a loop at the beginning of the shorter fragment PF13 that forms the tropoelastin and heparin binding sites. Perhaps newly secreted fibrillin-1 initially binds cell surface heparan sulfate through its C terminus, and subsequently tropoelastin binds this region following microfibril assembly. Within an assembled microfibril, this C-terminal region is likely to be adjacent to the bead (20).
Heparan sulfate is a critical determinant of microfibril assembly, since exogenous heparin, inhibition of sulfation, and inhibition of attachment of heparan sulfate to protein cores all block assembly in culture (7,8). Heparin does not impede fibrillin-1 secretion (8) but may compete with cell surface heparan sulfate to interact with fibrillin-1.
There are few reports of heparin binding to cbEGFs in extracellular matrix molecules. Heparin binding to fibrillin-1 through cbEGFs raises the intriguing possibility that this unusual feature may influence cell behavior through epidermal growth factor or related receptors in a manner similar to heparin-binding EGF-like growth factor (HB-EGF), which regulates cell signaling in cancer, wound healing, and other pathologies (32). In summary, we have provided new insights into novel heparin binding sites on fibrillin-1 and into the biological consequences of these strong interactions.