Rolf Minkwitz

FGF Binding and FGF Receptor Activation by Synthetic Heparan-Derived Di- and Trisaccharides

Possible binding sites of 2 disaccharides and 1 trisaccharide were identified on the surface of basic Fibroblast Growth Factor (bFGF) via Monte Carlo docking simulation. These binding sites were compared to the crystal structures of one of the disaccharides with bFGF by Ornitz¹ and to competitive binding experiments with ¹²⁷I-labeled heparin by Svahn.²

Introduction

FGFs regulate a diverse range of physiologic processes such as cell growth and differentiation as well as pathologic processes involving angiogenesis, wound healing, and cancer.³ FGFs use a dual receptor system to activate signal transduction pathways.^4-7 The primary component of this system is a family of signal-transducing FGF receptors (FGFRs) that contain an extracellular ligand-binding domain and an intracellular tyrosine kinase domain.³ The second component of this receptor system consists of HS proteoglycans or heparin-like molecules that are required in order for FGF to bind to and activate the FGFR.^5,6 Although the mechanism by which heparin/HS activates FGF is unknown, heparin, FGF, and the FGFR can form a trimolecular complex.⁵ Heparin/HS may interact directly with the FGFR linking it to FGF.⁸ Furthermore, heparin/HS can facilitate the oligomerization of two or more FGF molecules, which may be important for receptor dimerization and activation.⁵ There are no pharmacologic agents that modulate the activity of FGFs.

Heparin and HS are heterogeneously sulfated glycosaminoglycans that consist of a repeating dissacharide unit of hexuronic acid and D-glucosamine. At a minimum, highly sulfated octa⁴ or decasaccharide⁹ fragments derived from heparin are required for FGF to bind to the FGFR. However, preparation of these heparin fragments produces mixtures of isomers and chemically modifies the oligosaccharide ends.¹⁰ Furthermore, size-fractionated heparin may contain individual molecules with distinct biological properties. To overcome these limitations and to address the question of whether nonsulfated oligosaccharide sequences, which are abundant in HS, are involved in FGFR activation, Svahn et al.²synthesized di-, tri- and tetrasaccharides that correspond to structures found in heparin/HS. In our modeling project, we picked the strongest binding di – and trisaccharids(see below). Two of those were disaccharides, which are enantiomers and differ only by the relative configuration of the carboxyfunction to the rest of the molecule. The other two were trisaccharides - derivatives of the disaccharides – with the same sugar molecule attached to them.

Competitive binding experiments were undertaken in order to evaluate the saccharides strenghts as heparin mimiks.² These experiments revealed that some of the structures had a better binding towards FGF than heparin had.

Figure 2: Disaccharide 2(Dis-2)

Of the aforementioned structures, Dis-1was not bound to bFGF (around 0%), Dis-2 was bound to 10%, while Tri-1 and Tri-2 were up to 90% and 80% in the competitive binding experiment.

All molecular modeling was done using the Insight II software on Silicon Graphics O₂ workstations. The Oligosaccharides were built using the Biopolymer-Module in Insight II and minimized using the steepest descent method in a cff91 forcefield to a RMS deviations of 0.01 kcal. From the 26 different FGF crystal structures in the PDB, the most appropriate and interesting to do the calculations on seemed to be the basic FGF of Homo Sapiens without any mutations (PBDid: 1BFG).¹¹ Ornitz et al.¹ published the crystal structures of Di-2 bound to bFGF, unfortunately not mentioning the exact nature of the FGF.

The saccharides were docked on the protein using the Autodock software Version 2.4. (For further details see appendix) by G.M. Morris, D.S. Goodsell, R.Huey and A.Olson. Obtained binding sites and binding energies were compared to the crystal structures in Figure 5 and the ¹²⁷I labeled binding experiments.

Disaccharide 1

For the first disaccharide, we obtained 3 binding sites (see picture 1). The major binding site is in accordance with the nomenclature of Ornitz et al. (see figure 5) site 1. At this site, we found 92 % of the docked molecules in two states, in a twisted conformations of the rings, displayed in a bright brownish color, and a more linear conformation, displayed in the atom colors. 26% of all docked molecules found themselves in this twisted conformation, which had a binding energy of 48.29 (± 0.75) kcal. The other conformation in site 1 was the most stable found and 66% of the Dis-1 made their way into this energy minima at 53.99 (± 0.5) kcal. Minor binding site was site 2`,where 2% of the molecules were found in an enery of –46.52 kcal. In site 2, we yielded 6% of the docked saccharides in an energy of 48.05 (± 0.1) kcal. All the obtained binding energies are displayed in table 1(see list of figures).

Disaccharide 2

For the 2nd disaccaride only 2 binding sites were found at position 1 and 2(see picture 2). 86 % of all molecules docked at site 1 with an energy of 51.23 (± 0.5) kcal. 2% docked twisted at position 1 with a binding energy of 49.09 kcal and are displayed in a brownish color. The other molecules had an average binding energy of 45.53 (± 0.5) kcal. All binding energies and conformations of Dis-2 are displayed in Table 2.

Trisaccharide 1

The Autodock simulations of the first trisaccharide did not work and not a single molecule docked. From the final picture (see picture 3), one can conclude though, that the saccharides had enough space around the molecule to move and find any docking position.

Trisaccharide 2

With Tri-2(see picture 4), site 1 was again the major binding pocket with 90 % of the docked molecules and an energy of 58.07 (± 2) kcal. Unlike the disaccharides, the trisaccharides are not bound exactly at the same coordinates, but are more scattered at the binding site. 4% of Tri-1 were bound on top of the molecule with an energy of 53.38 kcal and 6% at site 2 with 56.17 (± 0.1) kcal. Obtained binding energies and conformations are displayed in table 3.

With the autodock calculations performed, we could identify site 1 as the major binding pocket. We investigated the binding position of the three oligosaccharides at site 1, drawing enlarged pictures (4-7) of the protein`s electrostatic potential of the surface at this site and one bound molecule. Blue colored are the partial positive charged zones up to a value of four and red colored are the partial negative zones up to a value of –4. From these images, which are representative for all of the saccharides bound in site 1, one can derive a major difference in the binding habit. While Dis-2 and Tri-2, with S-configuration at the adjacent carbon atom next to carboxyfunction in the a -ring, stick their negatively charged carboxyfunction in the positively charged hole of the site, Dis-1 does not so. Instead, the Dis-1 molecule is horizontally 180° C tilted, pointing with the carboxygroup away from the binding pocket. This could explain the better binding of Dis-2 in contrast to Dis-1 in the competitive binding experiments.

Surprisingly, the binding energy is higher for Dis-1 than for Dis-2. This is in contrast to the experimental results of the competitive heparin binding. An approach to explain our deviations from the experiment are the limitations of an Autodock simulation. While in reality, the protein can change its conformation upon the influence of the docking ligand, this is not possible in an Autodock simulation. We also consider only the energetic aspects of the binding and no entropic factors.

Future work

We will recalculate the docking for trisaccharide 3.

References

List of figures

Table 1: Binding sites and energies of Dis-1 molecules

In the Autodock simulation the saccharides were placed 15Å away from the outer surface of the protein. A grid of 120x120x120 was chosen for the disaccharides and 127x127x127 for the trisaccharides. The point spacing in the grid was 1Å . Our annealing parameters for the Monte Carlo procedure are given below and in brackets are the suggestions from the program authors for typical automatic dockings.

Surfaces were drawn a distance of 1Å to the atoms. A mistaken happened, when the saccharides were built. Instead of a methylgroup at the acetat rest of nitrogen, only a hydrogen was attached to the carbonylgroup. This had no effect on the outcome of the calculations though, as an evalution of the calculated structures revealed. Like the hydrogen, the butyl group will not be involved in any hydrogen bonding and does not interact with the protein.

Send comments to rminkwit@indiana.edu
Last updated: April 20, 1999
Return to MolViz Home