Mapping the heparin-binding site on the 13-14F3 fragment of fibronectin.

Fibronectin, a multifunctional glycoprotein of the extracellular matrix, plays a major role in cell adhesion. Various studies have revealed that the human 13th and 14th fibronectin type III domains (labeled (13)F3 and (14)F3 here) contain a heparin-binding site. Mapping of the heparin-binding sites of (13-14)F3, (13)F3, and (14)F3 by NMR chemical shift perturbation, isothermal titration calorimetry, and molecular modeling show that (13)F3 provides the dominant heparin-binding site and that the residues involved are within the first 29 amino acids of (13)F3. Predictions from earlier biochemical and modeling studies as well as the x-ray structure of (12-14)F3 were tested. It was shown that the positively charged residues that project into the solvent from the ABE face of the triple-stranded beta sheet on (13)F3 are involved in binding, but (14)F3 does not appear to contribute significantly to heparin binding.

Fibronectin, a multifunctional glycoprotein of the extracellular matrix, plays a major role in cell adhesion. Various studies have revealed that the human 13th and 14th fibronectin type III domains (labeled 13 F3 and 14 F3 here) contain a heparin-binding site. Mapping of the heparin-binding sites of 13-14 F3, 13 F3, and 14 F3 by NMR chemical shift perturbation, isothermal titration calorimetry, and molecular modeling show that 13 F3 provides the dominant heparin-binding site and that the residues involved are within the first 29 amino acids of 13 F3. Predictions from earlier biochemical and modeling studies as well as the x-ray structure of 12-14 F3 were tested. It was shown that the positively charged residues that project into the solvent from the ABE face of the triple-stranded ␤ sheet on 13 F3 are involved in binding, but 14 F3 does not appear to contribute significantly to heparin binding.
Fibronectin is a multifunctional extracellular glycoprotein found in plasma and the extracellular matrix. It participates in cell adhesion and migration processes in various physiological events, such as embryogenesis, wound healing, hemostasis, and thrombosis (1-7). Plasma fibronectin consists of two similar subunits of molecular mass 200 -250 kDa, held together near the C terminus by two disulfide bonds. Each subunit is composed of homologous repeating structural modules (type I, II, and III, referred to here as F1, F2, and F3) (3,8). Fibronectin F3 modules are known to be involved in several interactions including heparin binding and cell binding via integrins.
Heparin and heparan sulfate glycosaminoglycans are acidic complex polysaccharides found outside cells (9,10). The most common repeat structure unit is the trisulfated disaccharide, IdoUA2S-GlcNS6S, 1 but a number of structural variations of this disaccharide exist, leading to microheterogeneity. Heparin-binding proteins play important roles in various biological processes; it is therefore important to understand the molecular basis of heparin-protein interactions (11)(12)(13).
The interaction of fibronectin with cell surface heparan sulfate proteoglycans is important for inducing reorganization of the cytoskeleton and the assembly of focal adhesions. Physically and structurally distinct heparin-binding sites have been identified on fibronectin. One of these, called Hep-1, is in the N-terminal region, and a recent study identified a novel heparin-binding site in the alternatively spliced IIICS region (14). The major site, however, lies at the C terminus and is designated as Hep-2. Both Hep-1 and Hep-2 promote cell adhesion and attachment, which are independent of integrin receptors (15)(16)(17) and are important for fibronectin fibrillogenesis and extracellular matrix formation (18 -21). The Hep-2 domain is specifically implicated in focal adhesion formation (22), and it plays a major role in determining the compact solution conformation of fibronectin (23). From various studies it has been established that the heparin binding activity of the Hep-2 region is 10 -25-fold stronger than Hep-1 (24,25) and is similar to that of intact fibronectin (25). The specificity of heparin binding to the Hep-2 region is greater than that of Hep-1, which has also been shown to bind chondroitin/dermatan sulfate (26).
Analysis of deletion mutants indicated the presence of a major heparin-binding site on 13 F3 (the 13th type III module) with a possible contribution from 14 F3 (27,28). The binding of heparin to the Hep-2 domain is mainly electrostatic in nature, which is evident from the fact that the complexes can be disrupted by increasing ionic strength (24,29). The positively charged basic residues, Arg and Lys, play an important role in heparin binding; their chemical (30,31) or recombinant (27,32,33) modification leads to a reduction in heparin affinity. Molecular modeling studies of 13 F3 in conjunction with mutational analysis suggest that six positively charged residues, remote in sequence, form a "cationic cradle"-binding site for heparin (34). The crystal structure (35) of the Hep-2 region ( 12-14 F3) confirmed the existence of this positively charged cluster; the six basic residues of 13 F3 were found to project into the solvent from the ABE face of a ␤-sheet. The crystal structure also suggested a possible heparin-binding site on 14 F3 (35). Studies have been carried out with peptides to explore heparin binding activity on 14 F3 but have not been tested in the context of the folded domain (28).
The work in this paper describes the mapping of the Hep-2 domain on fibronectin. There are relatively few x-ray structures of heparin-protein complexes available because heteroge-neity and polydispersity of the heparin can cause problems in crystallization. Here NMR methods in conjunction with isothermal titration calorimetry (ITC) and molecular modeling have been used for mapping the heparin-binding site.
Bacterial culture was grown in M9 minimal media. Cells were harvested by centrifugation, and the cell pellet was resuspended in chilled MTPBS (16 mM Na 2 HPO 4 , 4 mM NaH 2 PO 4 ⅐H 2 O, pH 7.3) containing lysozyme (1 mg/ml), MgSO 4 (10 mM), and DNase I (Sigma) (20 g/ml). After sonication, Triton X-100 (Roche Molecular Biochemicals) was added to a final concentration of 0.1% (v/v), and the supernatant was loaded over glutathione-Sepharose 4B (Amersham Biosciences) and purified according to the instructions provided in the product manual. Fusion protein was cleaved away from glutathione S-transferase by digesting the fusion protein to completion with glutathione S-transferase fusion protease 3C pro overnight at 4°C (36). Fibronectin modules were purified by reverse phase high performance liquid chromatography using a C8 column (DYNAMAX, Rainin Instruments Co.) for 13 F3 and 14 F3 and a C4 column (Jupiter, Phenomenex) for [13][14] F3. The purified sample was lyophilized and stored at Ϫ20°C. The identity and purity of all the recombinant proteins were confirmed by electrospray mass spectrometry and N-terminal protein sequence analysis.
Preparation of Heparin Oligosaccharide IV-61 (a Pentasaccharide)-Oligosaccharide IV-61 was prepared by gel filtration chromatography on Sephacryl S-100HR of a heparinase (Seikagaku Co., Tokyo) digest of heparin (LAOB). Four fractions (I-IV) were collected and concentrated before re-chromatography using the same gel filtration column. Each tube fraction was desalted through a column of cellulofine GCL-25 and lyophilized. The sample name IV-61 indicates that the sample eluted at the tube number 61 of gel filtration fraction IV. Other three fractions of oligosaccharides were of high molecular weight (M r Ͼ 3500).
Preparation of Hexasaccharide and Octasaccharide from Bovine Lung Heparin-Oligosaccharides from a heparinase (the gift of Leo Pharmaceutical Products, Ballerup, Denmark) digest of bovine lung heparin (material remaining from the 2nd international standard heparin (37); a particularly homogeneous sample of heparin) were separated by gel permeation chromatography on Bio-Gel P-10 column (Bio-Rad) using a volatile salt eluant (2% ammonium bicarbonate in water. Fractions corresponding to resolved oligosaccharide peaks were pooled, repeatedly evaporated to remove eluant, and checked for the degree of polymerization using high performance gel permeation chromatography with refractive index detection. Concentration of the oligosaccharides was determined by measuring the area under the chromatography peaks compared with a standard curve prepared using known concentrations of heparin. Various disaccharides (summarized in Table I) as well as low molecular weight heparin (M r ϳ 3000) were also purchased from Sigma and used without further purification.
Analytical Ultracentrifugation Experiments-Sedimentation equilibrium experiments were performed with an OptimaXL-I analytical ultracentrifuge (Beckman Instruments) equipped with a UV-visible optical scanner with 12-mm 6-channel centerpieces at 35°C at 15,000, 27,000, and 36,000 rpm. Equilibration was established by comparing scans after 29 and 33 h. Samples of 13 F3 and 13-14 F3 were dissolved in a 50 mM acetate buffer containing 150 mM NaCl, pH 4.5, and 100 mM phosphate-buffered saline, pH 7.4, and dialyzed against the respective buffers overnight to attain dialysis equilibrium. The samples were diluted to obtain an A 280 nm of 0.5 (ϳ40 M of 13 F3 and ϳ20 M of 13-14 F3). 110 l of each sample solution was introduced into the respective centrifugation cell, and the corresponding equilibration buffers (125 l) were introduced in the reference compartment of each cell. The An60 Ti rotor was used for the experiments. All the data were analyzed using MicroCal Origin software (MicroCal Software Inc.) supplied with the XL-I centrifuge. A value for partial specific volume was calculated for each module at 20°C using the software Ultrascan version 5.0 (University of Texas Health Science Center at San Antonio), and a temperature correction was performed using the formula v T ϭ v 20 ϩ 4.25 ϫ 10 Ϫ4 (T Ϫ 298.5) (38). Values of 0.732752 and 0.741248 ml/g for 13 F3 and 13-14 F3, respectively, were used in all calculations. The apparent weight average molecular weight (M r app ) was determined from the equilibrium distribution using a least squares fitting procedure for single ideal species.
NMR Spectroscopy-All NMR experiments were acquired at 1 H fre-quencies of 500.  14 F3, and 13-14 F3, respectively. The mixing times in the three-dimensional NOESY-HSQC experiments were 125, 150, and 130 ms for 13 F3, 14 F3, and [13][14] F3, respectively. A two-dimensional NOESY spectrum for 13 F3 was also recorded with a mixing time of 150 ms. 1 H-15 N TOCSY-HSQC spectra (39) were recorded with the same acquisition times, except in the case of [13][14] F3, where the acquisition time was 40.96 ms in the direct 1 H dimension; the total experimental time was approximately 3 days. The isotropic mixing times in the three-dimensional TOCSY spectra were 53, 50, and 45 ms for 13 F3, 14 F3, and 13-14 F3, respectively. The two-dimensional TOCSY spectrum for 13 F3 was recorded with a mixing time of 58 ms. Data were processed using the FELIX 2.3 software package (Molecular Simulation Inc.), and spectra were analyzed with the aid of the XEASY (40) program on Sun and SGI workstations. The spectra of 13 F3, 14 F3, and 13-14 F3 were well resolved, allowing the assignment of amino acid spin systems and sequential connectivities using well established procedures.
HSQC Titrations-Titration experiments were carried out by adding heparin in samples comprising protein concentrations ϳ0.15 mM (calculated by measuring A 280 and extinction coefficients determined by quantitative amino acid analysis). The 15 N-labeled 13 F3, 14 F3, and 13-14 F3 proteins were dissolved in 0.5 ml of 90% H 2 O, 10% D 2 O containing 50 mM sodium acetate buffer and 150 mM NaCl, pH 4.5. Titrations were performed by removal of the sample from the NMR tube and thorough mixing with a concentrated solution of heparin in water. Except for hexasaccharides and octasaccharides concentrations, which were calculated by the method described above, the heparin concentration was calculated by weight.
Isothermal Titration Calorimetry-Protein samples were dialyzed overnight against the titration buffer (50 mM acetate, pH 4.5, 150 mM NaCl). The protein concentration was determined from the A 280 nm and adjusted to appropriate concentrations (between 40 and 60 M) by dilution with titration buffer. Low molecular weight heparin (average M r ϭ 3000) was dissolved in titration buffer at 50 times higher concentration relative to the protein. The experiment was conducted with the Microcal VP-ITC calorimeter at 35°C. Protein solutions were titrated by the addition of aliquots of heparin solution (25 ϫ 5 l for 13 F3 and 35 ϫ 5 l for 13-14 F3) while stirring at 320 rpm. The time differences between each successive injection were 5 min. To measure heats of dilution, the above procedure was repeated but with the buffer in the sample cell.
A model corresponding to single binding site was used to simulate the binding isotherms with manufacturer-supplied software (ORIGIN version 5.0, MicroCal, Inc.). Integration of heat signal as well as nonlinear regression analysis to fit the simulated binding isotherms to the experimental ⌬q i was done by floating n, K A , and ⌬H 0 as fitting parameters for the minimization of the sum of squared residuals, 2 . Heats of dilution were subtracted from heats of binding before performing the fit.
Molecular Modeling-Docking of heparin oligosaccharide ligands to protein structures was performed with Autodock, version 2.4 (41). This program allows for flexibility in the ligand structure but uses a rigid body protein approximation to speed up the calculation.
Pentasaccharide model ligands consisted of three GlcNS6S residues separated by IdoUA2S residues (OMe-GlcNS6S-IdoUA/2S-GlcNS6S-IdoUA2S-GlcNS6S-OMe). Because the IdoUA2S residue can adopt different ring conformations, two model ligands were used, one in which all IdoUA2S residues were in the 1 C 4 ring form and another in which the 2 S 0 ring form was adopted. The model ligands had fixed glycosidic torsion angles, taken from a reported heparin NMR structure (53).
Flexibility was allowed for all exocyclic torsion angles. Partial atomic charges required for the docking calculation were obtained by ab initio quantum chemistry calculations using the Jaguar program (Schrodinger Inc., Portland, OR) on 1-OMe4-OMe-substituted monosaccharides. Other oligosaccharides were built by manipulating the pentasaccharide using the program INSIGHT II. Grids of probe atom interaction ener-gies were computed first; 84-Å side grids were used for all the modules with a spacing of 0.7 Å. The ligand probes were then docked by simulated annealing according to the protocol developed by Mulloy and Forster (54) for bovine fibroblast growth factor (54). All the rotatable bonds were allowed to rotate freely except the glycosidic linkages, which were kept fixed (54). Typically, the ten lowest energy coordinate sets were extracted for each ligand type and used for visualization in Web-Lab Viewer (Molecular Simulation Inc.).

F3 Constructs in Solution
Although the structure of 12-14 F3 has already been determined by crystallography, it has not so far been possible to produce co-crystals of heparin and 12-14 F3. Our primary goal here is to define the site of heparin binding rather than redetermine the structure in solution.
Recombinant proteins were prepared corresponding to individual 13 F3 and 14 F3 modules as well as the 13-14 F3 module pair. The 15 N and 1 H NMR resonances of backbone amide protons of 13 F3, 14 F3, and 13-14 F3 were assigned using threedimensional 15 N-edited NOESY and TOCSY experiments. The observed pattern of nuclear Overhauser effects involving amide protons in the individual modules 13 F3 and 14 F3 (data not shown) indicated that all these proteins are folded, with a predominantly ␤-sheet structure, typical of a fibronectin type-III module (35,55). Information about the module interface can be obtained by comparing chemical shifts of individual modules with chemical shifts in the module pair. Only residues at the module interface exhibit chemical shift differences between the single modules and the module pair. There is also a striking correlation between residues that experience shifts and residues that have a different calculated solvent exposure between individual modules and the module pair. These data encourage the assumption that the solution structures of 13 F3, 14 F3, and 13-14 F3 are very similar to their structure in the crystallized 12-14 F3 fragment.

Sedimentation Equilibrium Experiments
The apparent molecular masses of 9933 Ϯ 390 and 18658 Ϯ 349 Da for 13 F3 and 13-14 F3 at pH 4.5 agree well with the calculated molecular masses of 10224.4 and 20147.9 Da. However, the observed molecular masses at pH 7.4 are 13,122 Ϯ 882 and 21,718 Ϯ 1263 Da for 13 F3 and 13-14 F3, suggesting that there is a tendency to aggregate at higher pH. Precipitation of NMR samples of 13 F3 was indeed observed at higher pH; therefore, all the NMR experiments were recorded at pH 4.5. To map these effects on the protein backbone, the sum of the magnitude of the changes in 1 H and 15 N chemical shifts was calculated with a weighting factor of 1/8 applied to 15 N chemical shifts to compensate for the larger 15 N chemical shift range (59). In the case of glycine a weighting factor of 1/6 was applied because glycine amide groups form a distinct subset with a chemical shift spread different from those of the rest of the amino acid residue types (60). Therefore, ⌬( 1 H, 15 Table I) differing in their sulfation pattern; heparin-derived oligosaccharides (IV-61, highly sulfated), hexasaccharide and octasaccharide (prepared from bovine lung heparin, which is known to have a high degree of sulfation in the heparin chain), and low molecular weight heparin (LMWH) (from porcine in-  The heparin-binding site on 13 F3 was investigated by titrating 15 N-labeled 13 F3 with heparin oligosaccharides while monitoring chemical shift changes in 1 H-15 N HSQC spectra. Of four disaccharides, only one, I-S, showed binding to 13 F3, whereas the other three did not bind even at very high concentrations. I-S did not bind, however, in the presence of 150 mM NaCl. All other oligosaccharides including LMWH induced significant chemical shift changes at physiological ionic strength. In all HSQC spectra, a single set of resonances was obtained, and shifts were observed at each addition of heparin until saturation, indicating that binding was on the fast exchange time scale relative to the 1 H and 15 N resonance frequency differences between free and bound 13 F3. The results are consistent with earlier studies that showed that trisulfated oligosaccharide binds best to the Hep-2 region (57, 58).
The differential chemical shift changes experienced by the 13 F3 backbone amide resonances in the presence of various oligosaccharides or low M r heparin were very similar. The results are illustrated in Fig. 1 for oligosaccharide IV-61, LMWH, and hexasaccharide. To map these effects on the protein backbone, the sum of the magnitude of the changes in 1 H and 15 N chemical shifts was calculated with a weighting factor applied to 15 N chemical shifts to compensate for the larger 15 N chemical shift range (59,60). Oligosaccharides of higher molecular mass than IV-61 were also tested, as stronger binding is expected because of the increased charge of the ligand, but higher oligosaccharides caused protein precipitation possibly because of oligosaccharide-mediated protein aggregation.
Commercially available heparins are typically heterogeneous with respect to degree of sulfation and molecular weight (61,62). The hexasaccharide prepared from bovine lung heparin is the most homogeneous preparation among all the oligosaccharide used in this study. LMWH, IV-61, and bovine lung hexasaccharide were from different sources and of different homogeneity, but each binds to 13 F3, affecting the same residues. Therefore, the heterogeneity within each sample has no effect on binding.
Most of the protein residues experience no chemical shift perturbation, indicating that heparin binding does not alter the overall protein structure (Fig. 2). The following residues in 13 F3 exhibit large changes in their backbone 1 H, 15 N chemical shifts upon addition of heparin: Ser-3, Arg-9, Arg-23, Thr-24, Lys-25, Thr-26, Glu-27, Ile-29, and Ser-55 (⌬␦ Ͼ 0.1). Ala-2, Arg-6, Arg-7, Ala-8, Val-10, Ser-55, Tyr-72, Ser-81, and Val-83 also show changes in chemical shifts, albeit smaller (0.05 Յ ⌬␦ Ͻ 0.1) (the above list is based on the results of all three oligosaccharides, LMWH, IV-61, and bovine lung hexasaccharide; see Fig. 1). All the affected residues lie in the triple-stranded ␤-sheet portion of 13 F3. Of the six basic residues (Arg-6, Arg-7, Arg-9, Arg-23, Lys-25, and Lys-54), which have been proposed to be involved in heparin binding, we found that only three of them, Arg-9, Arg-23, and Lys-25 exhibit strong perturbation. The slight perturbation observed for Arg-6 and Arg-7 contrasts with mutagenesis (34) data, showing that these residues are critical for heparin binding. It is worth noting that chemical shift mapping of backbone resonances as done here may not give a representative view of the role of side chains in ligand binding (see for example Ref. 63).
The largest chemical shift perturbations are observed for residues near the N terminus and the loop between strands B and C. Residues that are present in ␤-strands do not show a large shift (Fig. 3). Ser-3, which is in the loop region, shows the strongest perturbation, whereas Ser-21 and Ser-55, which are present in ␤-strands, show very weak perturba-  tion. The polar side chain of the residues Thr-24, Thr-26, and Glu-27 may form hydrogen bonds with the oligosaccharide as seen in several heparin-fibroblast growth factor complex structures (64,65). Under the same conditions as used for 13 F3, 15 N and 1 H chemical shifts of residues in 14 F3 were not significantly perturbed by the heparin oligosaccharides or disaccharides tested, even at very high concentrations (residues 90 -179, Fig. 4).
The NMR titration with [13][14] F3 was performed similarly to those with individual 13 F3 and 14 F3 modules; the results are shown in Fig. 4a. Titration of 13-14 F3 did not show any significant change in backbone chemical shift compared with residues from module 14 F3 (Fig. 4b). The residues of the 14 F3 module in the pair do not show any significant change as observed for the isolated 14 F3 module. The residues in 13 F3 region are, however, affected in a very similar way to those in the individual 13 F3 module. These data, therefore, indicate the presence of one dominant binding site within the first 29 residues of 13 F3. Fig. 5 shows the fractional change in the chemical shifts of backbone amide proton of affected residues upon the addition of heparin. We can, thus, conclude that the dominant heparin binding site is on module 13 F3 and that, under the experimental conditions used, binding to 14 F3, whether on its own or attached to 13 F3, is insignificant.
The salt-dependent binding of heparin (the 13-14 F3-heparin complex can be disrupted by increasing ionic strength) supports the notion that the charge density is an important requirement for interaction. Because the low molecular weight heparin gives the same binding site as hexasaccharide and octasaccharides, there is no evidence for an extended site to which longer oligosaccharides bind, indicating that the heparin sequence that binds to the protein is no more than a hexasaccharide in length.
ITC Studies-The binding of low M r heparin to 13 F3, 14 F3, and 13-14 F3 was also analyzed by ITC. Although 13 F3 and 13-14 F3 showed binding with low M r heparin, 14 F3 did not under these conditions. Representative titrations of low M r heparin with 13 F3 and [13][14] F3 are presented in Fig. 6. Although an interaction with heparin is observed with 13 F3, the plot of heat evolved against molar ratio cannot be confidently defined by a simple model corresponding to a single binding site, preventing an accurate determination of enthalpy of binding, stoichiometry, and affinity. This complexity may result from processes additional to the direct protein heparin interaction and is supported by the observation of precipitation of the 13 F3 sample at the end of the experiment when conditions were explored. In contrast, the binding of low M r heparin to [13][14] F3 exhibited a good distribution, and the data fit to a single binding site model. Heparin binding was characterized by a K A of 3.5 Ϯ 0.4 ϫ 10 5 M Ϫ1 (K D 2.9 M), ⌬H 0 of Ϫ4079 Ϯ 66 cal/mol, ⌬S 0 of 12.13 cal/mol/K, and stoichiometry of 1.1 Ϯ 0.01. This binding is enthalpically dominated, and generally, a positive ⌬S 0 is a strong indication that water molecules have been expelled from the complex interface. The main limitation in determination of exact thermodynamic parameters is uncertainty in the measurement of heparin concentration; the quoted K D is therefore an apparent value. Table II summarizes previously reported binding constants of heparin with Hep-2 region of fibronectin. The values are all of a similar order of magnitude, as observed with 13-14 F3. Interaction of heparin with 13 F3 and 13-14 F3 but not 14 F3 in these experiments supports the results of the NMR analyses that the heparin-binding site is limited to the 13 F3, although [13][14] F3 may be more stable in solution than 13 F3 on its own.
Modeling-The heparin-binding site of the Hep-2 region was predicted by using systematic docking calculations. The lowest energy coordinate sets of predicted complexes led to predicted binding sites as shown in Fig. 7 and summarized in Table III. The heparin-binding site inferred from the NMR study is shown in Fig. 7 for comparison. The binding site predicted by docking calculations matches well with the results obtained by NMR. A 3.5-Å cut-off was used to determine the protein residues in close contact with oligosaccharides. For  pentasaccharide ligand, no difference in binding was observed between the 1 C 4 and 2 S 0 forms of IdoUA2S residues. All the other oligosaccharide model ligands were in the 2 S 0 conformation. Also no conformational change of the Hep-2 domain has been reported on binding of heparin (binding of a heparin pentasaccharide (66) to antithrombin III causes a conformational change, making it a much more efficient inhibitor). Basic residues are present on 13 F3 as well as on 14 F3, but in the case of 13 F3 all the basic residues suspected to be involved in binding are clustered, whereas this is not true for 14 F3. Predictive docking calculations on 13 F3 and 13-14 F3 indicate that the oligosaccharide binds to 13 F3 and not to 14 F3, which we also observe from the NMR and ITC study.
Because the results obtained from predictive docking were consistent with experimental data, the same calculation protocol was also applied to all the other fibronectin modules whose x-ray or NMR structure are known, using a pentasaccharide as a probe molecule. The results are summarized in Table IV. No binding was predicted between heparin and fibronectin modules except for 1 F1 modules in the N-terminal domain of fibronectin, a region shown by various biochemical methods to be involved in heparin binding. Comparing the total interaction energy of the complex of 1 F1 with pentasaccharide with that of 13 F3, the binding interaction is very weak. It is already established that the Hep-1-binding site is weaker than the Hep-2-binding site. The result obtained by docking was further verified by NMR titration of 1 F1 with octasaccharide in which it was found that 1 F1 did show binding with octasaccharide (data not shown). CONCLUSION Based on the results of NMR and ITC it can be concluded that the dominant heparin-binding site is on module 13 F3 and that, under the experimental conditions used, binding to 14 F3, whether on its own or attached to 13 F3, is insignificant. The 14 F3 module does not appear to form a synergistic binding site with 13 F3. The heparin binding is not affected by the heterogeneous nature of heparin. The degree of sulfation seems to have a major effect as has been shown previously by Lyon et al. (58). Docking calculations closely predict the experimentally determined Hep-2 heparin-binding site. Also, similar calculations applied to other fibronectin modules of known structure were consistent with other experimental results. These results encourage the use of this type of calculation for the prediction of heparin binding.