The Molecular Modelling Laboratory

The availability of the complete genomes of a variety of living organisms is changing the way biomedical research is performed. Already now there are 3 dozen fully sequenced genomes available, including human. The importance of the discovery of a new protein sequence will fade, however, it will become essential to find or predict the function of genes and the way they interact, in order to understand how cells and organisms work on a molecular basis. Knowing the detailed functions that proteins can perform will enable us to find genes and proteins, that can be used as targets for drugs against diseases like cancer.

It is interesting that despite all the recent methodological advances in combinatorial chemistry and high-throughput screening by pharmaceutical companies world-wide, by the end of 2000 very few new drugs have been found using these methods. A more rational and broader approach based on the structure and function of genes and proteins is more promissing. This is why computational and molecular modelling techniques have been a standard part of drug design and development in pharmaceutical companies for more than 10 years.

For the last 7 years, we have been tackling the problem of how to use computational methods in order to understand or predict the functions of proteins, as well as their interactions with other proteins and small molecules. Most of our methods are based on the three-dimensional structures of these molecules. Our projects can be roughly classified into two groups: (1) the prediction of protein structures and structures of protein complexes, and (2) the design of new therapeutic agents. We have successfully applied our methods to proteins and molecules of interest to the Institute. In particular the development of models for the EGF receptor complex formed by two EGF and two EGF receptor molecules have led to a new understanding of the binding mechanisms of EGF mutants and the activation of the EGF receptor. The advantage of our models is the fact that our predictions can be tested experimentally and that the results of the experiments can directly feed back into the refinement of the models. Similarly, our models of the VEGF receptor and G-CSF receptor complexes are also helping in collaborations with other groups of the Institute to shed light into the function of two important molecular systems.

In 1999 we started our first antibody humanization project in collaboration with the Tumour Targeting Laboratory. We are already now in a situation where our antibody constructs can be tested. We have also reported the development of a new method for inhibitor design. Our first peptides and small molecules that were designed to inhibit the Ras - Raf interaction have now been tested experimentally. The results are very promising as some of our peptides and small molecules have been shown to inhibit the interaction between Ras and Raf. While the affinity of our compounds is still low, they bind with higher binding affinity than all other published Ras-Raf inhibitors.

We continued the investigation of the conformational properties of the small cyclic peptide cyclosporin, which is an important clinically used immunosupressant. We are now studying the interactions of cyclosporin with the cell membrane. These calculations are very time consuming as they involve more than 30,000 atoms. We therefore had to improve our computer hardware and are now building our own "Beowolf"-cluster, i.e. a parallel computer with 20 Pentium III processors running the Linux operating system. All 20 processors can work together to simultaneously perform large computations. The PC-based computer clusters are now becoming an extremely economical way to make the power of a supercomputer affordable for smaller laboratories.

The three-dimensional structure of the EphA3 receptor
H. Treutlein, J. Zeng in collaboration with M. Lackmann (Epithelial Biochemistry Laboratory) and A. Boyd (Queensland Institute for Medical Research, QLD)

Members of the protein family of "Eph-like" receptors are expressed on migrating cells in particular during embryogenesis. These receptors have been found to be overexpressed in some forms of cancer. By binding to specific protein ligands on the cell membrane of stationary cells the receptors become clustered and transmit signals into the cells in order to direct the movement of the migrating cell.

Eph receptors are membrane-bound proteins. The extracellular part of the Eph receptors with which the ligand interacts consists of a ligand binding domain, a cysteine rich ("EGF-like") domain and two fibronectin type III domains. While the ligand binds to the binding domain, the EGF-like domain seems to be important for the dimerisation or clustering of the receptor. We created models of the combined binding and EGF-like domains of the EphA3 receptor in order to investigate the atomic details of the ligand binding and of the dimerisation site. The X-ray structure of the related EphA2 receptor was used to model the EphA3 ligand binding site, while the model of the EGF-like domain is based on available X-ray structures of cysteine-rich domains and our experiences with the modelling of the EGFR S1 and S2 domains. Our model suggests a particular orientation between the binding and the EGF-like domains and is currently being tested using site-directed mutagenesis. Crude models of possible configurations of the EphA3 dimer have also been built. Our aim is to use our model to aid the design of inhibitors of ligand binding or dimerization of Eph receptors, which may lead to a novel class of anti-cancer drugs.

Structure of the G-CSF receptor complex
H.R. Treutlein and N. Hall, in collaboration with J.E. Layton (Cytokine Biology Laboratory) and A. Hammacher (Epithelial Biochemistry Laboratory)

The binding of granulocyte colony stimulating factor (G-CSF) to its cell surface bound receptor induces receptor oligomerisation and initiates intracellular signalling. The knowledge of the full three-dimensional structure of the G-CSF receptor complex is therefore highly desirable and important for our understanding of signal transduction across cell membranes.

We have now finished a model of the full receptor complex consisting of two G-CSF and two G-CSFR molecules. Since experimental evidence suggests that the complex can be formed without the three membrane proximal fibronectin type III domains of G-CSFR, our model of the G-CSFR molecules only includes the Ig-like domain and the cytokine-binding domain of the receptor. Our model agrees well with results obtained from site-directed mutagenesis.

Electrostatics of vascular endothelial growth factor receptor complexes
N.E. Hall and H.R. Treutlein in collaboration with M. Achen and S. Stacker (Angiogenesis Laboratory)

Vascular Endothelial Growth Factor (VEGF), VEGF-B, VEGF-C, VEGF-D and placental growth factor (PlGF) are proteins of immense biological interest due to the importance of vascular development in cancer growth and proliferation. In addition to the five known VEGF family ligands, there are three homologous VEGF receptors, each binding a different combination of VEGF ligands. The VEGF ligands have a different range of biological functions depending upon which receptors are recruited for signalling.

From examining the sequences of the VEGF family of ligands and receptors it is impossible to understand the binding preferences of each ligand to the three receptors. To improve the knowledge of receptor ligand complex formation, the crystal structures of VEGF and VEGF bound to domain 2 of the VEGF receptor 1 have been investigated. Utilising the MODELLER program, homology models of the remaining ligands and receptors, currently of unknown structure, have been created in order to understand binding specificities of the remaining ligand-receptor complexes.

We are evaluating the approach of analysing electrostatic complimentarities between ligand and receptor to explain the binding specificities of the family of VEGF-VEGF receptor complexes.

Simulating the partitioning of cyclosporin A through a cell membrane
M.F. O'Donohue and H.R. Treutlein in collaboration with A.W. Burgess (Epithelial Biochemistry Laboratory)

Cyclosporin A (CsA) is an important therapeutic agent for the prevention of graft rejection in organ transplantation. CsA is an eleven residue cyclic peptide, which is found in a variety of backbone conformations.

In order to reach its receptor, cyclophillin, CsA needs to pass through the membrane in order to reach the cell interior. The cell membrane is formed by a lipid bilayer that separates the cell cytosol from the extracellular region. Due to its highly hydrophobic nature, the lipid bilayer serves as a relatively impermeable barrier to the passage of most water-soluble molecules. The ability of CsA to undergo a conformational change may support its passage from a hydrophilic extracellular region though the hydrophobic cell membrane into a hydrophilic interior of the cell.

The knowledge of the conformations that CsA can adopt and of the mechanisms for its partitioning through the cell membrane could help the design of more effective immunosupressant analogues.

We have performed various computer simulations of a CsA molecule in a 60Å cubed area containing 144 lipid molecules (1,2 dilauroyl-DL-Phosphatidylethanolamine) arranged in a bilayer. The 60Å box is also packed with water molecules to 1 atmosphere pressure. During the simulations, directed forces are applied to the CsA to drive it through the membrane. We are currently working on extended simulations that apply weak forces over an extended period of time in order to sample possible molecular conformations of CsA outside, on top of, and inside the membrane.

Computational combinatorial peptide design of Ras inhibitors
J. Zeng & H.R. Treutlein in collaboration with E. Nice, H. Maruta (Epithemal Laboratory)

We developed a new inhibitor design approach based on a computational combinatorial method. Our method has allowed us, for the first time, to design an inhibitor targeted to the surface of a protein unlike other design methods, which are only considering well defined binding pockets. When we applied our method to the Ras protein, our results identified a previously unnoticed potential interaction region between Ras and other proteins. This interaction region was later confirmed by the X-ray structure of the Ras/Sos complex. We also used our method to design novel peptides that could bind to Ras. These peptides resembled the hydrophobic, hydrophilic and charged pattern observed in peptides experimentally identified from the Ras effectors. Moreover, some of our peptides have been shown to inhibit Ras-Raf binding in vitro using biosensor and radioassay analysis, as well as enzyme-linked immunosorbent assay techniques. In addition, we are extending the method to derive a pharmacophore for the Ras surface. With the rapid development of computer technology, this approach could automate structure based drug design and can significantly reduce the costs of the drug discovery process.

Computational design of MHC-bound peptides
J. Zeng & H.R. Treutlein in collaboration with G.B. Rudy (Walter and Eliza Hall Institute of Medical Research, Melbourne)

Proteins encoded by the major histocompatibility complex (MHC) bind peptides derived from foreign proteins, such as viral products, and present them to T cells in order to initiate an immune response. To understand what characteristics determine peptides binding to MHC molecules is therefore of central importance to the immune system. We applied our computational combinatorial ligand design method, developed initially for the Ras inhibitor, to predict which peptides bind to MHC molecules. The probabilities for specific amino acids to occur at certain positions of a peptide that can bind to a specific MHC molecule were calculated and showed good agreement with experimentally determined amino acid distributions. Moreover, we also demonstrated the possibility for using our method to dock flexible peptides into protein binding sites.

Modelling studies of the EGF receptor
R.N. Jorissen and H.R. Treutlein in collaboration with A.W. Burgess (Epithelial Biochemistry Laboratory)

The Epidermal Growth Factor (EGF) receptor is a cell surface receptor tyrosine kinase which mediates the biological effects of ligands such as EGF and TGF. Little is known about the molecular basis of the interactions between the extracellular portion (ectodomain) of the EGF receptor and its ligands. A molecular model of the EGF receptor ectodomain has been previously constructed in our laboratory. This structure was modelled as two components: the L1 and S1 (first and second) domains and the L2 and S2 (third and fourth) domains. At the end of the S1 domain is a putative "hinge" about which rotation of the two halves is proposed to occur.

A proteolytically generated fragment, which includes all of the L2 domain, is able to bind one molecule of EGF or TGF-a. Using a protein-protein docking algorithm and manual filtering of the resultant structures, a model of EGF bound to the L2 domain of the EGF receptor was generated.

Using the above complex as a basis, a crude model of the 2:2 EGF-EGF receptor ectodomain was built using a molecular graphics package. In this model, each EGF ligand molecule is in contact with the L2 domain of one of the EGF receptor molecules and with the L1 domain of the other EGF receptor molecule. This model provides an explanation for the ability of ligand-bound EGF receptor to associate with other EGF receptor homologues, including ErbB-2, which has no known high affinity ligand.