• fibroblast growth factor;
  • heparin-like hexasaccharide;
  • protein–carbohydrate complex


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

The 3D structure of a complex formed by the acidic fibroblast growth factor (FGF-1) and a specifically designed synthetic heparin hexasaccharide has been determined by NMR spectroscopy. This hexasaccharide can substitute natural heparins in FGF-1 mitogenesis assays, in spite of not inducing any apparent dimerization of the growth factor. The use of this well defined synthetic heparin analogue has allowed us to perform a detailed NMR structural analysis of the heparin–FGF interaction, overcoming the limitations of NMR to deal with the high molecular mass and heterogeneity of the FGF-1 oligomers formed in the presence of natural heparin fragments. Our results confirm that glycosaminoglycans induced FGF-1 dimerization either in a cis or trans disposition with respect to the heparin chain is not an absolute requirement for biological activity.


fibroblast growth factor


fibroblast growth factor receptor




molecular dynamics


restrained minimization

The human and mouse fibroblast growth factor (FGF) family consists of 18 members that share a common homologous core [1–3]. FGF-1 and -2 properties have often been considered paradigmatic for the whole family. Proteins including FGF-like domains have also been detected in invertebrates [2]. FGFs are involved in a wide variety of physiological processes besides the control of cell division [4,5]. They are also involved in many developmental steps during embryogenesis [6,7]. The 3D structure of several vertebrate FGFs have been determined using X-ray diffraction and NMR spectroscopy [8–11]. The FGF structure is homologous to that first described for soybean trypsin inhibitor [12]. The fold, known as β-trefoil motif, consists of six β-strand pairs, which define an β-barrel capped at its base by three of the pairs. Five of the pairs have a hairpin structure (β2-β3, β4-β5, β6-β7, β8-β9, β10-β11). β1 and β12 are separate strands, but sometimes are also designed as the sixth hairpin because of their topological equivalence to the other strand pairs. There is some controversy about whether β-strand 11 really adopts such a canonical secondary structure in FGF-1, and whether it is more flexible than the rest [10,13–17]. FGF signaling begins at the target cell with the recognition of these growth factors by specific transmembrane tyrosine kinase receptors (fibroblast growth factor receptors; FGFRs [18]).

FGFs strongly bind to glycosaminoglycans (GAGs) of the type of heparin and heparan sulfate [2,19,20]. Sulfated GAGs seem also essential for the appropriate assemblage of an effective signaling complex between FGF and their cell membrane kinase receptors. In vitro assays have shown that FGF-2-induced mitogenesis is highly dependent on cell-surface heparan sulfate GAGs, although it can be replaced by soluble heparin [21–23]. On the other hand, heparin itself is a near absolute requirement for FGF-1-driven mitogenesis. Nevertheless, in cells with heparan sulfate GAGs at their surface, heparin can be replaced by some nonphysiological compounds of relatively low molecular mass [14,15]. There is a certain controversy about the minimal sulfated GAG size required for this set of processes [18].

There is convincing evidence that FGF-1 and FGF-2 oligomerize to a certain extent in the presence of sulfated GAGs [24,25]. This oligomerization has been proposed to constitute a key step in physiological FGF-induced mitogenesis [21,26]. According to the X-ray solved structures of FGFs complexed to heparin [27,28], and the NMR studies of FGFs bound to heparin functional analogs [14], it has been concluded that the main heparin binding site is located near the C-terminus, and that it involves residues belonging to the loop connecting strands 11 and 12, and to strands 10 and 11. The 3D structures determined by X-ray diffraction show striking discrepancies in the proposed topologies of the GAG–protein interactions in both the binary FGF–GAG and ternary FGF–GAG–FGFR complexes [27–31]. Because some of the studies have been carried out with FGF-1 and others with FGF-2, it cannot be ruled out that the discrepancies arise from the specific protein used in each case. There is an ongoing discussion on this point, although recent data better support the symmetric two-end model than the trans one [32].

Synthetic GAG-type oligosaccharides may allow one to circumvent some of the problems that the inherent heterogeneity of the natural GAG fragments constitutes for high-resolution structural characterizations. We have recently pioneered the use of those oligosaccharides for the structural characterization of the dependence of FGF-1-induced mitogenesis on heparin. We have shown that hexasaccharide 1(Fig. 1), which displays all sulfate groups only on one side of its helical structure, can substitute natural heparins in FGF-1 mitogenesis assays in spite of not inducing any apparent dimerization of the growth factor [33]. In contrast, for synthetic GAG-type oligosaccharides displaying the sulfate groups oriented on both sides of the helix as in natural heparin, and for natural heparin fragments [29], octasaccharide is the minimum size that can fully replace native heparin in the in vitro mitogenesis assays. Herein, we describe the 3D structure of the biologically active monomeric complex between FGF-1 and this heparin-like hexasaccharide, as determined by NMR. As far as we know, this is the first complete NMR study of FGF-1 in complex with a bioactive sulfated oligosaccharide [27]. Previous studies have been carried out with complexes of FGF-1 with small noncarbohydrate molecules able to functionally emulate natural heparins [14], because the oligomerization of FGFs induced by these compounds precluded detailed NMR structural studies due to the broadening of the signals [10]. The results presented here may constitute a further step toward the understanding of the structural basis of the recognition and activation of FGF-1 by GAGs.


Figure 1.  (Top) Schematic view of the structure and NMR conformation (in surface representation) of the heparin-like hexasaccharide analogue (1) used in this study. For the sake of simplicity, the l-iduronate units are all shown in the 1C4 conformation. Glucosamine-iduronate interglycosidic NOE crosspeaks are labelled in blue whereas iduronate-glucosamine interglycosidic NOE cross peaks are labelled in red. (Bottom) Expansion of the anomeric region of the 600 MHz double 13C filtered NOESY (100 ms mixing time) of the complex of hexasaccharide 1 and FGF-1, showing the assignment of the interglycosidic NOE peaks for the carbohydrate (G-6, I-5, G-4, I-3, G-2) moieties.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

Chemical shift perturbation

The first step in the interaction study was a chemical shift perturbation analysis of the FGF-1 1H and 15N NMR resonances upon addition of hexasaccharide 1. Figure 2A shows the superimposition of the 1H-15N HSQC spectra for free and bound FGF-1. A relatively high number of NH groups are indeed affected by sugar binding. A plot of the weighted average (of 15N and 1H) chemical shift perturbation [Δδ =(Δinline image + 0.2Δinline image)1/2] along the protein sequence is depicted in Fig. 2B. The residues that show the largest changes are N32, H55, L87, H116, G124, G129, K132, G134, R136, H138, Y139, and I144. It can be observed that the higher shifts are located at the C-terminus. The observed perturbations are fairly similar to those reported for the interaction of FGF-1 with sucrose octasulfate [34] and with a natural heparin-derived saccharide [10]. In the latter case, the signal intensities of FGF-1 in the 1H-15N HSQC spectrum were significantly reduced, due to the FGF-1 oligomerization upon complex formation. Because overall signals were extremely broadened, according to the authors [10], it was not possible to connect the main chain assignments to those of side chains and therefore the 3D structure of the complex could not be obtained. However, the chemical shift perturbation analysis could be carried out, and according to the published data [10], the more shifted residues were S31, G34, H35, L87, E105, N106, K127, G124–G134 and R136–A143. These residues are in the same region of the protein that experiences the largest shifts in our complex and it is worth noting that in both complexes K132 is the residue more affected upon sugar binding.


Figure 2.  Chemical shift perturbation analysis. (A) Overlaid 1H-15N HSQC spectra of free (blue) and bound FGF-1 (red). The experiments were acquired at 17.6 T. Residues K132, G134, H138 and I144 display the highest chemical shift variations between the free and bound states. (B) Weighted average (of 15N and 1H) chemical shift perturbation (Δδ = (Δinline image + 0.2Δinline image)1/2) of FGF-1 residues upon complex formation.

Download figure to PowerPoint

Assignments of the FGF-1 resonances in the complex

The protein resonance assignments were obtained using standard strategies based on triple-resonance NMR experiments. FGF-1 was prepared in the uniformly labeled 15N and 15N, 13C double-labeled forms, as shown below.

The 1H-15N HSQC spectrum of FGF-1 in the complex with hexasaccharide 1 showed 119 backbone amide NMR resonances out of the expected 132. Most of the residues that could not be found in the HSQC were later identified as residues located in high flexible regions of the protein; six of them belonged to the N-terminus region of FGF-1. The good quality of the 1H-15N HSQC spectrum and the line width of the signals obtained for this complex provided further arguments in favor of the monomeric state of FGF-1 in the presence of hexasaccharide 1. Amide proton and nitrogen resonances of Lys142 could not be identified in the 1H-15N HSQC of the complex.

Then, 90% of the backbone amide resonances in the 1H-15N HSQC spectrum and their α carbons were sequentially assigned based on the analysis of the correlations observed in HNCA and HN(CO)CA experiments. The assignment of the remaining amide resonances in the 1H-15N HSQC was obtained from the 3D 15N-edited NOESY-HSQC and homonuclear 2D NOESY experiments. The backbone carbonyl resonance assignments were obtained from the HNCO and HN(CA)CO experiments. The α and side chain proton resonances were assigned combining the information of the 3D 15N-edited TOCSY-HSQC, 3D 15N-edited NOESY-HSQC and 2D NOESY experiments.

The NH2 groups of Gln and Asn residues were connected to their side chain γ and β protons with the 15N-edited NOESY-HSQC. The aromatic proton resonances were connected to their β protons with the 2D NOESY experiment. The chemical shift assignments were deposited at BioMagResBank, BMRB accession number 6875. The hexasaccharide signals in the bound state were assigned using double filtered 13C TOCSY and double filtered 13C NOESY experiments [35].

Determination of the structure of FGF-1–hexasaccharide 1 complex

A set of 1564 intramolecular protein–protein NOEs were assigned from the analysis of the 15N-edited NOESY-HSQC, as well as 2D NOESY experiments, and converted into 1384 relevant distance restraints. In addition, and only after a well defined protein structure was obtained, 23 hydrogen bonds, identified from deuterium exchange experiments, were also used as distance restraints. Thus, a total of 1407 distance restraints were used for the final structure calculations (Table 1). Initially, only the structure of the protein in the complex was addressed. No protein–protein NOEs that could be attributed to a dimeric complex were found. In addition, the line shapes of the NMR crosspeaks in all experiments carried out for the protein in the presence of the saccharide were similar to those obtained in the free state. Moreover, recent 15N NMR relaxation data [36] have shown that the global motion correlation time of FGF-1 when free or bound to hexasaccharide 1 are very similar (less than 1 ns increase upon binding). All these data point to the biologically active complex between hexasaccharide 1 and FGF-1 being a monomer, as also deduced by a detailed equilibrium sedimentation analysis [33]. Due to the absence of experimental data for the six first residues at the N-terminus of the protein, this region was not included in the structure calculations.

Table 1.   Statistics from dyana and amber restrained molecular dynamics (MD) calculations for the FGF-1–hexasaccharide 1 complex. Limits of variation interval are given in brackets.
Structural statistics from dyana and amber restrained MD calculations
Protein distance restraints
 Intra-residue 261
 Sequential 472
 Medium range 1 < |i-j| < 5 182
 Long range |i-j| ≥ 5 469
H bond restraints  23
Structural statistics for the best 40 dyana conformers
Average target function value0.59 ± 0.05 (0.45–0.67)
Restraints violated > 0.2 Å3
No. structures2, 1, 4
Average rmsd (residues 26–150)
 Backbone rmsd to mean (Å)1.05 ± 0.14 (0.67–1.49)
 Heavy atom rmsd to mean (Å)1.83 ± 0.17 (1.38–2.26)
Average rmsd (residues 26–31, 35–49, 55–72, 77–104, 108–127, 130–150)
 Backbone rmsd to mean (Å)0.91 ± 0.12 (0.61–1.28)
 Heavy atom rmsd to mean (Å)1.67 ± 0.15 (1.31–2.06)
Structural statistics for the best 20 amber conformers
Average rmsd (residues 26–150)
 Backbone rmsd to mean (Å)1.13 ± 0.15 (0.76–1.48)
 Heavy atom rmsd to mean (Å)1.96 ± 0.21 (1.40–2.40)
Average rmsd (residues 26–31, 35–49, 55–72, 77–104, 108–127, 130–150)
 Backbone rmsd to mean (Å)0.96 ± 0.12 (0.69–1.34)
 Heavy atom rmsd to mean (Å)1.76 ± 0.19 (1.27–2.23)
Ramachandran analysis (residues 22–154)
 Residues in most favoured regions70.1%
 Residues in additional allowed regions26.4%
 Residues in generously allowed regions1.9%
 Residues in disallowed regions1.6%

Starting with 200 randomized conformations and applying the dyana program [37], a group of 40 high quality NMR structures was obtained (Table 1).

At this stage, the carbohydrate structure was introduced into the calculation. The starting sugar conformation was that previously determined for the free carbohydrate by Angulo et al. [33]. Indeed, it was concluded, by performing 13C double filter NOESY experiments, that the hexasaccharide conformation in the bound state was very similar to that observed for the free sugar. Moreover, the pattern of the intra and interglycosidic NOE crosspeaks and their relative intensities were strikingly similar for both the free and the FGF-1 bound states [35]. Twelve carbohydrate–carbohydrate distances derived from a double 13C filtered NOESY experiment were introduced in the structure calculation. These constraints involved H1′-H4 and H1′-H3 proton pairs for the GlcN-IdoA linkages and H1′-H4 and H1′-H6proR proton pairs for the IdoA-GlcN linkages (Fig. 1). The presence of strong NOEs between these protons indicated that the glycosidic linkages adopt a syn-ψ type conformation. The conformations of these linkages define the global shape of the molecule, and the FGF-1 bound hexasaccharide keeps the typical helical structure present in heparin (Fig. 3). A second essential aspect of the interaction of heparin with FGFs is the role of the conformational equilibrium of the iduronate rings in the binding. In this case, the iduronate conformation was not restricted to any of the 1C4 or 2SO forms in the calculations because, according to the NOE data, the iduronate rings of the bound oligosaccharide displayed conformational flexibility with both the 1C4 and the 2SO forms being present on the bound state [35].


Figure 3.  Representative structures for hexasaccharide 1, l-iduronate units are in the chair (A) and boat (B) conformation.

Download figure to PowerPoint

The unambiguous assignment of the weak protein–carbohydrate intermolecular NOEs was not straightforward (hexasaccharide protons are surrounded by the bulky NMR-silent sulfate groups). Nevertheless, six intermolecular NOE crosspeaks could be detected in the 2D-NOESY spectrum and were used as additional weak distance restraints, in ambercalculations, to hold the hexasaccharide on the protein binding site. These intermolecular NOE crosspeaks involved sugar protons and the side chain protons of Lys127, Ser130, Arg133 and Lys142 (Table 2).

Table 2.   Hexasaccharide–FGF-1 intermolecular NOEs.
Intermolecular NOEs
CH3 Isop. IdoA-1Hβ3 (δ 3.67) S130
H2 IdoA-1Hδ2 (δ 3.15) R133
H2 GlcN-4Hε2 (δ 3.04) K127
H5 GlcN-4Hε2 (δ 3.04) K127
H5 GlcN-4Hα (δ 3.61) K127
H1 GlcN-6Hε2 (δ 2.93) K142

The 30 protein structures with the lowest values of the dyana target function were selected for the complex structure calculation. Initial models for the complex structure were built within autodock 3.0 [38,39] (see below) starting from the dyana structures and the hexasaccharide conformation reported by Angulo et al. [33]. These models were used as initial structures for a simulated annealing protocol performed with the amber 5.0 package that finally yielded the 3D solution structure of the hexasaccharide–protein complex.

The final ensemble comprised 20 structures with very small deviations from ideal covalent geometry and no nonbonded contacts. The average rmsd value for the backbone atoms N, Cα and C between the individual conformers within this ensemble and their average coordinates was 1.13 Å in the structured region of the protein (residues 26–150). This rmsd value was lower (0.96 Å), when some of the loop and turn motifs were excluded from the calculation. The quality of these structures was evaluated using the program procheck-nmr[40]. Most of the residues are restricted to favored or additionally allowed regions of the (φ,ψ) space (Table 1). Indeed, among all nonglycine residues, only His107 consistently fell into a disallowed region. It is noteworthy to mention that this residue also presents anomalous torsion angles in other FGF-1 structures described so far [28,41]. This structural feature can be due to the fact that this residue is located in a small turn (not strictly well defined) that connects two β-strands of the protein. Also, it is spatially close to Tyr108, and the deduced distortion of the angles could be a way of minimizing the steric hindrance between the lateral chains of His107 and Tyr108.

The Ramachandran plot for all residues in the minimum energy structure obtained in the restrained MD simulations is shown in the supplementary material. FGF-1 (fragment 22–154) comprises 113 nonglycine and nonproline residues. The φ,ψ plot shows 84 residues (74.3%) in most favored regions, 27 residues (23.9%) in additional allowed regions, and two residues (1.8%) in generously allowed regions. In the NMR structure described by Ogura et al. [10] for free FGF, only 52.6% of the residues were located in the most favored regions. This is very probably due to the high flexibility of FGF-1 in the free state, as also deduced by 15N relaxation measurements [36].

The FGF-1–hexasaccharide 1 final coordinates have been deposited in the Protein Data Bank (PDB accession code 2ERM).

Description of the 3D structure of the complex

FGF-1 forms a monomeric complex in the presence of hexasaccharide 1, with the protein conformation remaining fairly similar to the structure of the free protein [11]. Figure 4A shows the backbone superposition of the 20 structures of the complex derived from the restrained molecular dynamics calculations while Fig. 4B presents the superposition of one of these structures with the highest resolution X-ray structure reported for the free protein (PDB accession code 1RG8). The backbone rmsd (residues 26–150) for the superposition of the NMR structures with the X-ray one is within the range 1.05–1.47 Å.


Figure 4.   (A) Stereo view of the superimposed backbones of the 20 structures derived from the restrained molecular dynamics calculations of the complex between hexasaccharide 1 and FGF-1. (B) The FGF-1 structure in the hexasaccharide complex (red) superimposed onto the X-ray structure of the free protein (blue, PDB code 1RG8).

Download figure to PowerPoint

The FGF-1 bound hexasaccharide adopts a helical structure. In 15 of the 20 structures, the iduronate rings moved to the 2SO conformation during the molecular dynamics protocol. This observation is consistent with a previous study that showed the presence of a chair-skew boat equilibrium for the hexasaccharide iduronate rings in the bound state [33].

The site of interaction between the hexasaccharide and FGF-1

The intermolecular NOEs that allowed locating of the hexasaccharide at the protein surface are described in Table 2. These intermolecular NOE crosspeaks involved the side chain protons of positively charged amino acid residues, Lys127, Ser130, Arg133 and Lys142 (Fig. 5), suggesting a electrostatically driven interaction between the protein and the negatively charged sulfate groups of the hexasacharide. A close view of the electrostatic interactions in the binding site is given in Fig. 6. Consistently, similar NOEs (involving Lys127, Arg133 and Lys142) have been detected in previous NMR studies of FGF-1 with nonoligosacharidic molecules [14]. The protein residues giving intermolecular NOEs are in the protein region that experiences the largest chemical shift perturbation upon complex formation. Moreover, 15N relaxation studies have shown that the order parameter substantially increases in this region from that observed in the free state upon complex formation [39].


Figure 5.  Hexasaccharide structure (conformation refined with amber) at the binding site of FGF-1. Residues involved in intermolecular NOEs are given in yellow. The hexasaccharide protons involved in intermolecular NOEs are within blue circles.

Download figure to PowerPoint


Figure 6.  Close view of the binding site of FGF-1, the residues involved in the key electrostatic interactions with hexasaccharide 1 are marked with arrows. The lateral chains of these residues are shown as sticks.

Download figure to PowerPoint

A comparison of the heparin binding site of our NMR structure with the crystallographic ones of heparin–FGF-1 [23], and heparin–FGF-2–FGFR [26] was then performed. The overlays are shown in Fig. 7. The three complexes share a common FGF–carbohydrate interface, in which the hexasaccharide sulfate groups N-SO3(GlcN-6), 2-OSO3(IdoA-5) and 6-OSO3(GlcN-4) fit nicely with the negatively charged groups COO(IdoA)-N-SO3(GlcN) and 2OSO3(IdoA), which lie juxtaposed on the same side of the heparin helix (Fig. 7). This sequence GlcN(N-SO3)-IdoA(2-OSO3)-GlcN(6-OSO3) has been described as the minimal binding sequence for FGF-1 [42]. In the crystallographic models of FGF–heparin [27,28] and FGF–FGFR–heparin [30,31] complexes, the N-sulfate group of one glucosamine residue, and the 2-O-sulfate of the following iduronic acid pack closely against the protein backbone joining Q141 and K142 of FGF-1 (Q134 and K135 of FGF-2) and K126 and K127 of FGF-1 (K119 and R120 of FGF2), respectively, and are surrounded by the side chains of N32, Q141, K142, A143, K126, K127 and K132 of FGF-1 (N28, Q134, K135, A136, K119, R120, and K125 of FGF-2). A similar binding region is observed in our complex, where K142 is close to the N-sulfate group of GlcN-6 and K127 is close to the 2-O-sulfate group of IdoA-5 (Fig. 6). Residues N32, Q141, A143, and K132 are also directly involved in the interaction with this hexasaccharide region (GlcN6-GlcN4). Indeed, from the experimental data, they show important chemical shift changes upon complex formation (Δδ = 0.18, 0.14, 0.12 and 1.54 p.p.m., respectively). Moreover, R136, K126 and R133 are in close contact with the IdoA3-IdoA1 hexasaccharide region (Δδ = 0.22, 0.10 and 0.10 p.p.m., respectively) (Fig. 6). This sugar end is closer to the protein in our complex than in the complexes with natural heparin, because in our case, due to the differences in the sulfation pattern, there is an additional sulfate group in this sugar region that can interact with K126 and R133.


Figure 7.  Comparison of the heparin binding sites. (A) Crystal structure of an FGF2–FGFR1–heparin complex, with 2 : 2 : 2 stoichiometry (PDB entry 1FQ9) [31]. FGFR and FGF are in ribbon representation and heparin is coloured red. For the sake of simplicity, only one of the FGF2–heparin units (red) is overlaid with our complex (blue, PDB entry 2ERM). (B) Heparin–FGF-1 dimer (red, PDB entry 2AXM) [28] overlaid with our complex (blue). The three complexes share a common FGF–carbohydrate interface in which the hexasaccharide sulfate groups N-SO3(GlcN-6), 2-OSO3(IdoA-5) and 6OSO3(GlcN-4) fit very nicely with the negatively charged groups COO(IdoA)-NSO3(GlcN) and 2OSO3(IdoA) that lie juxtaposed on the same side of the heparin helix.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

The designed synthetic hexasaccharide heparin analogue 1 has allowed us to perform a detailed structural analysis of the heparin–FGF interaction by NMR and to explore those interactions that are established in the biologically active complex. Our work overcomes the limitations of oligomerization and heterogeneity that occur when natural heparins interact with FGFs. Thus, according to the data described above, GAG-induced FGF-1 dimerization either in a cis or trans disposition respect to the heparin chain in the culture media is not an absolute requirement for the growth factor to be able to trigger cell division [33]. The monomeric complex whose characterization is reported here shows a biological activity equivalent to that provided by natural heparin [33].

Crystallographic data have provided two different models of the FGF–heparin–FGFR complex. The assemblage of this complex at the cell membrane triggers the mitogenesis process. In one of the models (Pellegrini et al. [30]), heparin interacts in a so called trans orientation with two FGF polypeptides. This arrangement somehow resembles the heparin–FGF complex described by DiGabriele et al. [28] Activation of FGF by heparin was proposed by these authors to require the formation of this FGF dimer. Herein, a single heparin molecule is intercalated between two polypeptide entities that do not show reciprocal interactions. On the basis of this structure, the formation of the trans-like dimer could favour the assemblage of the FGF–heparin–FGFR ternary complex proposed by Pellegrini et al. [30], despite certain variations in the relative degree of alignment of the two FGF protomers. In principle, this arrangement could provide a reasonable explanation for the activating role of heparin in FGF-driven mitogenesis.

However, the data presented herein showing that dimerization is not required for FGF to show full mitogenic activity seem to go against this trans-type model of FGF–heparin–FGFR assemblage. In the trans-type structure, a heparin molecule showing negative charges all around both sides of the sugar chain is essential for the model consistency. In contrast, our data seem to fit better with the complex with a 2 : 2 : 2 stoichiometry as proposed by Schlessinger et al. [31]. In this model, each FGF molecule interacts with a different heparin moiety. Two heparin entities are now present in the complex, fitting in the canyon created by the assemblage of four polypeptides (two FGF plus two FGFR molecules). In our case, hexasaccharide 1 could also fit in this canyon (Fig. 7A), because now, the presence of negative charges at both sides of the sugar chain is not required.

Nevertheless, regarding the physiological significance of either complex, it has to be considered that there is a high concentration of heparan sulfate at the cell surface, in close vicinity to the FGFRs, and thus, the actual assemblage for the mitogenesis signaling complex may also involve the incorporation of this other molecule.

The comparison of the FGF binding sites in our complex with those previously reported for crystallographic complexes with fragments of natural heparin shows the same residues involved in carbohydrate recognition. Therefore, hexasaccharide 1 constitutes an excellent model to study FGF–heparin interactions. The high ability of this compound to activate FGF-1 can be due to its characteristic sulfation pattern. Probably, the existence of one additional sulfate group pointing toward the protein increases the stability of the FGF–carbohydrate complex. This feature can also explain why the activity of this compound in the biological assays is similar to the activity of a synthetic heparin regular region octasaccharide [33]. Recent investigations [32] have suggested that primary amino acid differences within the heparin binding sites of FGFs and FGFRs, together with ligand-induced variations in the orientation of the D2 receptor domain, lead to the formation of distinct positively charged canyons for individual FGF–FGFR combinations and that these canyons serve as molecular ‘sieves’ that select an optimal heparin sulfation pattern. Therefore, our future challenge is to use this well defined synthetic compound for the structural characterization of the ternary complex, FGF-1–FGFR–hexasaccharide to gain insights into the molecular basis of the highly cooperative FGFR dimerization process.

Experimental procedures

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References


Hexasaccharide 1 was synthesized as previously reported [43]. 15NH4Cl and 13C glucose were obtained from Cambridge Isotope Laboratories (Andoves, MA, USA) and Isotec (Miamisburg, OH, USA), respectively. Heparin-Sepharose was obtained from Pharmacia (Uppsala, Sweden).

Expression and purification of FGF-1

The FGF-1 gene was cloned between the NcoI and HindIII restriction sites in pRAT-4 [44]. Protein expression was carried out in Escherichia coli BL21(DE3). Uniformly labeled 15N and 15N, 13C double-labelled FGF-1 were obtained by growing the cells in M9 minimal media, using 15NH4Cl and 13C glucose as the nitrogen and carbon sources, respectively. The protocol for isotope labeling is as follows [45]. Cells were grown in 1 L of Luria–Bertani medium at 37 °C upon reaching optical cell densities at 600 nm (D600) ≈ 0.7, then the cells were pelleted. The cells were washed and centrifuged for 5 minutes at 4°C and 5500 g, using an M9 salt solution, excluding all nitrogen and carbon sources. The cell pellet was resuspended in 250 mL of isotopically labeled minimal media. Protein expression was induced after 1 h by addition of isopropyl thio-β-d-galactoside to a concentration of 0.8 mm. The cells were harvested after 12 h of induction, resuspended in 10 mm sodium phosphate (pH 7.2), containing 5 mm EDTA and disrupted by sonication. The protein was purified by affinity chromatography on a heparin-Sepharose column and was eluted at 1.5 m NaCl concentration.

NMR spectroscopy

For NMR studies, the buffer of the protein solution was changed by ultrafiltration to 10 mm sodium phosphate (pH 6.0) containing 150 mm NaCl and 1 mm hexasaccharide in the case of the complex samples, and 300 mm NaCl for the free protein. All samples were prepared in a mixture of 90% H2O/10%2H2O to give a final protein concentration of 1 mm in the NMR tube. In the case of amide proton exchange the hexasaccharide–15N-labelled FGF-1 sample was lyophilized and reconstituted with 2H2O. 3-Trimethylsilylpropionate was added to the samples to a final concentration of 100 µm. The 3-trimethylsilylpropionate signal was used as a direct reference for proton. Nitrogen and carbon chemical shifts were referenced in an indirect manner.

NMR spectra were performed at 298 K on Varian Inova-750, Bruker AVANCE-800, Bruker AVANCE-500 and 600 (cryoprobe) spectrometers.

2D NOESY were acquired at 800 MHz with a mixing time of 100 ms; the numbers of acquired complex points were 2048 × 512 in the t2 (1H) and t1 (1H), respectively. The spectral width was 11160 Hz (13.95 p.p.m). 2D 1H-15N HSQC spectra, used in the chemical shift perturbation analysis were performed at 750 MHz for both the free protein and the hexasaccharide–FGF-1 complex; the number of acquired complex points were 1024 × 256 in t2 (1H) and t1 (15N), respectively. 2D 1H-15N HSQC spectra, used to detect backbone amide resonances that slowly exchange with D2O, were performed at 500 MHz with the same size acquisition data matrix as in the 1H-15N HSQC spectra acquired at 750 MHz.

3D 1H-15N NOESY-HSQC and 1H-15N TOCSY-HSQC were performed at 750 MHz with mixing times of 150 and 60 ms, respectively, using standard pulse sequences [46]. The numbers of acquired complex points were 2048 × 96 × 32 in the t3 (1H), t2 (1H) and t1 (15N), respectively. Thirty-two scans (NOESY-HSQC) and 24 scans (TOCSY-HSQC) per increment were used. (3D) HNCA and (3D) HN(CO)CA were acquired at 800 MHz with 2048 × 64 × 64 complex points in the t3 (1H), t2 (15N) and t1 (13C), respectively. Thirty-two scans in HNCA, 48 scans in HN(CO)CA, 16 scans in HNCO and 32 scans in HN(CA)CO were used. Spectral widths were 5600 Hz (1H), 2800 Hz (15N) and 6500 Hz (13C).

Double filtered 13C TOCSY (60 ms) and NOESY (50–200 ms) were recorded on a Bruker Avance 600 (cryoprobe) spectrometer at 298 K with 1 mm FGF-1 solution, 150 mm NaCl, 10 mm sodium phosphate in D2O.

NMR restraints

2D NOESY and 3D 1H-15N NOESY-HSQC crosspeaks were integrated and classified as strong, medium and weak with corresponding upper limits of 2.6, 3.5 and 5.5 Å. The lower limit for proton–proton distances was set as the sum of van der Waals radii of the protons. In the case of methyl groups, nonstereoassigned pairs, or equivalent aromatic ring protons, the appropriate pseudo atom corrections were applied.

Structure calculation and refinement

Structure calculations were performed by torsion angle dynamics as implemented in the dyana package [47] using the distance restraints derived from NOESY cross peaks. Structure refinement proceeded in an iterative manner in which distance restraints were added or modified following analysis of the previous ensemble of structures. The first well defined ensemble of structures was used to obtained the stereospecific assignment of prochiral protons with GLOMSA module [37] implemented in dyana, followed by the application of the ‘distance modify’ module, also implemented in dyana, which removes redundant and meaningless distance restraints and adjusts distance limits with diastereotopic groups. A set of 1384 distance restraints derived from protein–protein NOEs were used in the calculations. In the final round of structure calculations, 23 hydrogen bond constraints were included for those backbone HN groups whose signals were observed to exchange slowly when the sample buffer was exchanged for D2O, a distance constraint of 2.5 Å was used for the (D)H-O(A) partners.

The 30 best dyana structures in terms of target function in complex with the hexasaccharide as given by autodock 3.0 were subjected to restrained molecular dynamics [48] with the amber force field [49]. At this step, six additional intermolecular distance restraints were included in the calculations.

The hexasaccharide amber parameters were those already applied for molecular dynamics simulations of the free sugar in explicit solvent [33,50]. The amber input files were prepared by using the X-LEAP module of the amber package.

After an initial short restrained minimization (REM) to generate a starting complex structure devoid of intermolecular clashes one cycle of 35 ps of restrained simulated annealing was run with the amber package. During the first 5 ps, the temperature was gradually raised from 100 K to 600 K. Then, the temperature was held constant at 600 K for 15 ps to efficiently explore the available conformational space. Finally, the temperature was cooled down gradually to 100 K for 15 ps. The final production run comprises 5 ps of RMD/REM. Temperature regulation was achieved by coupling the system to a thermal bath using the Berendsen algorithm [51]. Structures with no distance restraint violation greater than 0.25 Å were chosen as representative of the solution structure of the complex.

Docking calculations

As mentioned above, docking simulations were performed to obtain a model of the hexasaccharide in the binding site of the protein. Docking simulation of hexasaccharide to FGF-1 was performed by using the Lamarckian genetic algorithm of the autodock 3.0 program [39].

Solutions that were within 3 Å RMS deviations of each other belonged to the same cluster, and the clusters were ranked according to their lowest energy member. After different trials, a dielectric constant of 78 Debye was selected for the calculations. The starting ligand conformation was that previously obtained for the free carbohydrate [33]. Partial charges were assigned to the hexasaccharide atoms according to the Pérez-Imberty-Mazeau (PIM) forcefield [52,53]. In the starting structure, the iduronate rings were in the 1C4 conformation. For the protein coordinates, we used the structures obtained in the dyana restrained molecular dynamics. Grids of probe atom interaction energies and electrostatic potential were generated around the whole protein by the autogrid program present in autodock 3.0. For each calculation, one job of 100 docking runs was performed using a population of 200 individuals and an energy evaluation number of 3 × 106. First, a global search, in which the whole protein was searched for ligand binding sites, was performed. The hexasaccharide was placed 10 Å away from the surface of the protein in a 126 × 126 × 126 Å box of 0.375 Å grid spacing. In this first docking simulation, flexibility for the ligand torsions was not allowed. In a second step, a local search was performed with smaller grid spacing, 0.275 Å. In this calculation, only potential interaction sites detected in the preceding global search were searched and the 10 hexasaccharide glycosidic torsion angles were allowed to rotate freely.

In all the sets of structures obtained during the global search, the hexasaccharide was located at the same protein region, close to 10 and 11 β-strands, and to the loop connecting β11 and β12. Therefore, in a second step, a local search centered in this protein region was performed. The solutions obtained in the local search can be grouped in three clusters. In two of them, those with less docking energy, the hexasaccharide adopts an orientation with respect to the protein in which Lys127 and Lys142 are close to GlcN-6 (Fig. S2A). In contrast, in the third cluster, the carbohydrate adopts the opposite orientation, and in this case, Lys142 is close to IdoA-1 (Fig. S2B). The more populated clusters with Lys142 close to GlcN-6 were also those in agreement with the experimental intermolecular NOEs. Therefore, the coordinates of the minimum energy structure of these clusters were chosen as the starting coordinates for the restrained molecular dynamics calculations.


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

We thank Mrs Mercedes Zazo for their help during the expression and purification of labelled FGF-1. This work was supported by the Dirección General de Investigación Científica y Técnica (Grants BIO2002-0374, BQU2003-03550-C03-01 and BQU2003-0374). We also thank Comunidad de Madrid, Fundación Ramón Areces, and Fundación Francisco Cobos for fellowships to JA, and RO, respectively. We thank Dr Olson for the use of autodock. We also thank Dr M. Gairí (Parc Cientific de Barcelona) and Dr Martín-Pastor (Universidad de Santiago de Compostela) for access to the 800 and 750 MHz spectrometers, respectively.


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References
  • 1
    Zhang X, Ibrahimi OA, Olsen SK, Umemori H, Mohammadi M & Ornitz DM (2006) Receptor specificity of the fibroblast growth factor family. The complete mammalian FGF family. J Biol Chem 281, 1569415700.
  • 2
    Itoh N & Ornitz DM (2004) Evolution of the Fgf and Fgfr gene families. Trends Genet 20, 563569.
  • 3
    Olsen SK, Garbi M, Zampieri N, Eliseenkova AV, Ornitz DM, Goldfarb M & Mohammadi M (2003) Fibroblast growth factor (FGF) homologous factors share structural but not functional homology with FGFs. J Biol Chem 278, 3422634236.
  • 4
    White KE, Evans WE, O'Riordan JL, Speer MC, Econs MJ, Lorenz-Depiereux B, Grabowski M, Meitinger T & Strom TM (2000) Autosomal dominant hypophosphataemic rickets is associated with mutations in FGF23. Nat Genet 26, 345348.
  • 5
    Gimenez-Gallego G & Cuevas P (1994) Fibroblast growth factors, proteins with a broad spectrum of biological activities. Neurol Res 16, 313316.
  • 6
    Nishimura T, Nakatake Y, Konishi M & Itoh N (2000) Identification of a novel FGF, FGF-21, preferentially expressed in the liver. Biochim Biophys Acta 1492, 203206.
  • 7
    Powers CJ, McLeskey SW & Wellstein A (2000) Fibroblast growth factors, their receptors and signalling. Endocr Relat Cancer 7, 165197.
  • 8
    Zhu X, Komiya H, Chirino A, Faham S, Fox GM, Arakawa T, Hsu BT & Rees DC (1991) Three-dimensional structures of acidic and basic fibroblast growth factors. Science 251, 9093.
  • 9
    Moy FJ, Seddon AP, Bohlen P & Powers R (1996) High-resolution solution structure of basic fibroblast growth factor determined by multidimensional heteronuclear magnetic resonance spectroscopy. Biochemistry 35, 1355213561.
  • 10
    Ogura K, Nagata K, Hatanaka H, Habuchi H, Kimata K, Tate S, Ravera MW, Jaye M, Schlessinger J & Inagaki F (1999) Solution structure of human acidic fibroblast growth factor and interaction with heparin-derived hexasaccharide. J Biomol NMR 13, 1124.
  • 11
    Bernett MJ, Somasundaram T & Blaber M (2004) An atomic resolution structure for human fibroblast growth factor 1. Proteins 57, 626634.
  • 12
    McLachlan AD (1979) Three-fold structural pattern in the soybean trypsin inhibitor (Kunitz). J Mol Biol 133, 557563.
  • 13
    Arunkumar AI, Srisailam S, Kumar TK, Kathir KM, Chi YH, Wang HM, Chang GG, Chiu I & C (2002) Structure and stability of an acidic fibroblast growth factor from Notophthalmus viridescens. J Biol Chem 277, 4642446432.
  • 14
    Pineda-Lucena A, Jimenez MA, Nieto JL, Santoro J, Rico M & Gimenez-Gallego G (1994) 1H-NMR assignment and solution structure of human acidic fibroblast growth factor activated by inositol hexasulfate. J Mol Biol 242, 8198.
  • 15
    Pineda-Lucena A, Jiménez MA, Lozano RM, Nieto JL, Santoro J, Rico M & Gimenez-Gallego G (1996) Three-dimensional structure of acidic fibroblast growth factor in solution: effects of binding to a heparin functional analog. J Mol Biol 264, 162178.
  • 16
    Moy FJ, Seddon AP, Campbell EB, Bohlen P & Powers R (1995) 1H, 15N, 13C and 13CO assignments and secondary structure determination of basic fibroblast growth factor using 3D heteronuclear NMR spectroscopy. J Biomol NMR 6, 245254.
  • 17
    Chi YH, Kumar TK, Kathir KM, Lin DH, Zhu G, Chiu IM & Yu C (2002) Investigation of the structural stability of the human acidic fibroblast growth factor by hydrogen-deuterium exchange. Biochemistry 41, 1535015359.
  • 18
    Faham S, Linhardt RJ & Rees DC (1998) Diversity does make a difference: fibroblast growth factor–heparin interactions. Curr Opin Struct Biol 8, 578586.
  • 19
    Capila I & Linhardt RJ (2002) Heparin–Protein interactions. Angew Chem Int Ed Engl 41, 391412.
  • 20
    Conrad HE (1998) Heparin Binding Proteins. Academic Press, San Diego.
  • 21
    Ornitz DM, Yayon A, Flanagan JG, Svahn CM, Levi E & Leder P (1992) Heparin is required for cell-free binding of basic fibroblast growth factor to a soluble receptor and for mitogenesis in whole cells. Mol Cell Biol 12, 240247.
  • 22
    Yayon A, Klagsbrun M, Esko JD, Leder P & Ornitz DM (1991) Cell surface, heparin-like molecules are required for binding of basic fibroblast growth factor to its high affinity receptor. Cell 64, 841848.
  • 23
    Rapraeger AC, Krufka A & Olwin BB (1991) Requirement of heparan sulfate for bFGF-mediated fibroblast growth and myoblast differentiation. Science 252, 17051708.
  • 24
    Venkataraman G, Shriver Z, Davis JC & Sasisekharan R (1999) Fibroblast growth factors 1 and 2 are distinct in oligomerization in the presence of heparin-like glycosaminoglycans. Proc Natl Acad Sci USA 96, 18921897.
  • 25
    Waksman G & Herr AB (1998) New insights into heparin-induced FGF oligomerization. Nat Struct Biol 5, 527530.
  • 26
    Spivak-Kroizman T, Lemmon MA, Dikic I, Ladbury JE, Pinchasi D, Huang J, Jaye M, Crumley G, Schlessinger J & Lax I (1994) Heparin-induced oligomerization of FGF molecules is responsible for FGF receptor dimerization, activation, and cell proliferation. Cell 79, 10151024.
  • 27
    Faham S, Hileman RE, Fromm JR, Linhardt RJ & Rees DC (1996) Heparin structure and interactions with basic fibroblast growth factor. Science 271, 11161120.
  • 28
    DiGabriele AD, Lax I, Chen DI, Svahn CM, Jaye M, Schlessinger J & Hendrickson WA (1998) Structure of a heparin-linked biologically active dimer of fibroblast growth factor. Nature 393, 812817.
  • 29
    Stauber DJ, DiGabriele AD & Hendrickson WA (2000) Structural interactions of fibroblast growth factor receptor with its ligands. Proc Natl Acad Sci USA 97, 4954.
  • 30
    Pellegrini L, Burke DF, von Delft F, Mulloy B & Blundell TL (2000) Crystal structure of fibroblast growth factor receptor ectodomain bound to ligand and heparin. Nature 407, 10291034.
  • 31
    Schlessinger J, Plotnikov AN, Ibrahimi OA, Eliseenkova AV, Yeh BK, Yayon A, Linhardt RJ & Mohammadi M (2000) Crystal structure of a ternary FGF-FGFR-heparin complex reveals a dual role for heparin in FGFR binding and dimerization. Mol Cell 6, 743750.
  • 32
    Mohammadi M, Olsen SK & Goetz R (2005) A protein canyon in the FGF-FGF receptor dimer selects from an a la carte menu of heparan sulfate motifs. Curr Opin Struct Biol 15, 506516.
  • 33
    Angulo J, Ojeda R, de Paz JL, Lucas R, Nieto PM, Lozano RM, Redondo-Horcajo M, Gimenez-Gallego G & Martin-Lomas M (2004) The activation of fibroblast growth factors (FGFs) by glycosaminoglycans: influence of the sulfation pattern on the biological activity of FGF-1. Chembiochem 5, 5561.
  • 34
    Chi Y, Kumar TK, Chiu IM & C (2000) 15N NMR relaxation studies of free and ligand-bound human acidic fibroblast growth factor. J Biol Chem 275, 3944439450.
  • 35
    Canales A, Angulo J, Ojeda R, Bruix M, Fayos R, Lozano R, Gimenez-Gallego G, Martin-Lomas M, Nieto PM & Jimenez-Barbero J (2005) Conformational flexibility of a synthetic glycosylaminoglycan bound to a fibroblast growth factor. FGF-1 recognizes both the C-1 (4) and S-2 (0) conformations of a bioactive heparin-like hexasaccharide. J Am Chem Soc 127, 57785779.
  • 36
    Canales A, Fayos R, Angulo J, Ojeda R, Martín-Pastor M, Nieto PM, Martín-Lomas M, Giménez-Gallego G & Jiménez Barbero J (2006) Backbone dynamics of a biologically active human FGF-1 monomomer, complexed to a hexasaccharide heparin-analogue, by 15N NMR relaxation methods. J Biomol NMR 35, 225239.
  • 37
    Guntert P, Braun W & Wuthrich K (1991) Efficient computation of three-dimensional protein structures in solution from nuclear magnetic resonance data using the program DIANA and the supporting programs CALIBA, HABAS and GLOMSA. J Mol Biol 217, 517530.
  • 38
    Goodsell DS & Olson AJ (1990) Automated docking of substrates to proteins by simulated annealing. Proteins 8, 195202.
  • 39
    Morris GM, Goodsell DS, Halliday RS, Huey R, Hart WE, Belew RK & Olson AJ (1998) Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. J Comp Chem 19, 16391662.
  • 40
    Laskowski RA, Rullmannn JA, MacArthur MW, Kaptein R & Thornton JM (1996) AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR. J Biomol NMR 8, 477486.
  • 41
    Lozano RM, Pineda-Lucena A, Gonzalez C, Angeles Jimenez M, Cuevas P, Redondo-Horcajo M, Sanz JM, Rico M & Gimenez-Gallego G (2000) 1H NMR structural characterization of a nonmitogenic, vasodilatory, ischemia-protector and neuromodulatory acidic fibroblast growth factor. Biochemistry 39, 49824993.
  • 42
    Pellegrini L (2001) Role of heparan sulfate in fibroblast growth factor signalling: a structural view. Curr Opin Struct Biol 11, 629634.
  • 43
    Ojeda R, Angulo J, Nieto PM & Martín Lomas M (2002) The activation of fibroblast growth factors by heparin: Synthesis and structural study of rationally modified heparin-like oligosaccharides. Can J Chem 80, 917936.
  • 44
    Peränen J, Rikkonen M, Hyvönen M & Kääriänien L (1996) T7 Vectors with a modified T7lac promoter for expression of proteins in E. coli. Anal Biochem 236, 371373.
  • 45
    Marley J, Lu M & Bracken C (2001) A method for efficient isotopic labeling of recombinant proteins. J Biomol NMR 20, 7175.
  • 46
    Cavanagh J, Fairbrother WJ, PalmerAG III & Skelton NJ (1996) Protein NMR Spectroscopy. Principles and Practice. Academic Press, Inc, San Diego.
  • 47
    Guntert P, Mumenthaler C & Wuthrich K (1997) Torsion angle dynamics for NMR structure calculation with the new program DYANA. J Mol Biol 273, 283298.
  • 48
    Scheek RM, van Gunsteren WF & Kaptein R (1989) Molecular dynamics simulation techniques for determination of molecular structures from nuclear magnetic resonance data. Methods Enzymol 177, 204218.
  • 49
    Pearlman DA, Case DA, Caldwell JW, Ross WS, Cheatham TEI, DeBolt S, Ferguson D, Seibel G & Kollman P (1995) AMBER, a package of computer programs for applying molecular mechanics, nomal mode anylisis, molecular dynamics and free energy calculations to simulate the structural and energetic properties of molecules. Comp Phys Commun 91, 141.
  • 50
    Angulo J, Nieto PM & Martín Lomas M (2003) A molecular dynamics description of the conformational flexibility of the 1-iduronate ring in glycosaminoglycans. Chem Commun 13, 15121513.
  • 51
    Berendsen HJC, Postma JPM, van Gunsteren WF, DiNola A & Haak JR (1984) Molecular dynamics with coupling to an external bath. J Chem Phys 81, 36843690.
  • 52
    Sadir R, Baleux F, Grosdidier A, Imberty A & Lortat-Jacob H (2001) Characterization of the stromal cell-derived factor-1alpha-heparin complex. J Biol Chem 276, 82888296.
  • 53
    Imberty A, Bettler E, Karababa M, Mazeau K, Petrova P & Pérez S (1999) PIM Parameters, Perspectives in Structural Biology (Vijayan M, Yathindra N & Kolaskar AS, eds) pp. 392409. Indian Academy of Sciences and Universities Press, Hyderabad, India.