Computer simulations of human interferon gamma ...

Computer simulations of human interferon gamma mutated forms E. Lilkova∗ , L. Litov∗ , P. Petkov∗ , P. Petkov† , S. Markov∗∗ and N. Ilieva‡ ∗

Department of Atomic Physics, Faculty of Physics, University of Sofia, Bulgaria † Faculty of Chemistry, University of Sofia, Bulgaria ∗∗ Institute for Parallel Processing, Bulgarian Academy of Sciences, Bulgaria ‡ Institute for Nuclear Research and Nuclear Energy, Bulgarian Academy of Sciences, Bulgaria Abstract. In the general framework of the computer-aided drug design, the method of molecular-dynamics simulations is applied for investigation of the human interferon-gamma (hIFN-γ ) binding to its two known ligands (its extracellular receptor and the heparin-derived oligosaccharides). A study of 100 mutated hIFN-γ forms is presented, the mutations encompassing residues 86–88. The structural changes are investigated by comparing the lengths of the α -helices, in which these residues are included, in the native hIFN-γ molecule and in the mutated forms. The most intriguing cases are examined in detail. Keywords: Virtual Drug Design, Molecular Dynamics Simulations, Human Interferon Gamma, Human Interferon Gamma Receptor, Heparin Derived Oligosaccharides, Secondary Structure, Mutagenesis PACS: 36.20.-r, 87.10.Tf, 87.14.E-, 87.15.A-, 87.15.ap, 87.15.bd, 87.15.bg, 87.15.hg, 87.15.hp, 87.15.km, 87.15.kp

INTRODUCTION Drug design and development is an interdisciplinary, expensive, time-consuming and full of risks process. During this process, a huge number of biologically active compounds (ligands) are tested to form a stable complex with the active site of the target structure, responsible for the attacked disease. All but a few of them are sampled out at different stages of the investigation. The estimated investment in a single new drug from the concept to the market — from the disease target identification and validation, through the discovery and optimization of the lead structures, to the clinical tests and regulatory approval – ranges from $150 to $880 million and 9–15 years of research [1]. Roughly 75% of these costs are attributed to failures along this way [2]. A method to predict these failures at an early stage is highly desirable on account of decreasing of both the drug discovery costs and duration. In addition to the experimental in vitro and in vivo approaches, computer simulations (coined as in silico) are now routinely used as a tool to prioritize experiments at each stage of the process [3]. One of the methods for investigation of biological macromolecule interactions is Molecular Dynamics (MD). In this paper we investigate the interactions of human interferon gamma (hIFN-γ ) with its extracellular receptor and with heparin-derived oligosaccharides by means of MD simulations. This method is used also for prediction of the effects of the mutations of three amino-acid residues on the structure of the hIFN-γ molecule.

COMPUTER-AIDED DRUG DESIGN In silico methods are one of the few techniques that have the potential to improve significantly the drug discovery process [4]. In biomedical arena, computer-aided or in silico design is being utilized to expedite and facilitate the hit identification and the hit-to-lead selection, to optimize the absorption, distribution, metabolism, excretion and toxicity profile and to reduce seriously the necessary safety checks. Commonly used computational approaches include ligandbased drug design (oriented on the molecular framework that carries the essential features responsible for a drugs biological activity, or pharmacophore), structure-based drug design (drug-target docking), and quantitative structureactivity (QSAR) and quantitative structure-property relationships [5]. Most of these studies use the modified “lock-and-key” approach where the conformational diversity of the small molecule ligand is taken into account, but the “high molecular weight” ligand binding site (e.g. enzyme, protein, receptor) is considered rigid. The eventual conformational change upon ligand binding is fundamental in protein behavior as most enzymes and receptors are flexible and contribute to ligand binding with a significant conformation

CP1203, 7th International Conference of the Balkan Physical Union, edited by A. Angelopoulos and T. Fildisis © 2009 American Institute of Physics 978-0-7354-0740-4/09/$25.00

914 Downloaded 26 Nov 2010 to 93.152.157.93. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions

FIGURE 1.

hIFN-γ - hIFN-γ Rα complex.

change (induced fit). Thus, the large scale collective movements of the proteins upon binding of substrates and/or inhibitors are also very important for the understanding of the binding mechanism itself and eventually, of the structurefunction relationships [6]. Molecular-dynamics simulations provide a powerful tool for investigation of protein bindings. In this framework, many useful computational methods have been developed particularly for the purposes of the structure-based drug design [7]. The first clinically useful drugs to emerge from MD simulations (used both in the refinement of the crystallographic structures and in the computational docking of the model compounds to the target structures) are the HIV protease inhibitors [8].

HUMAN INTERFERON GAMMA Human Interferon gamma (hIFN-γ ) is a pleotropic cytokine which plays crucial role in regulating host defense and immunopathologic processes [9]. The first step in eliciting the cell response is the specific high-affinity (Kd = 1010 M) interaction of hIFN-γ with its cell-surface receptor IFNγ R (Fig. 1). In addition to its receptor, hIFN-γ binds with high-affinity (Kd = 1.5 × 109 M) also to heparan sulfate (HS) [10, 11, 12]. Aberrant hIFN-γ expression is associated with a number of autoinflammatory and autoimmune diseases [13]. hIFN-γ accomplishes its multiple biological effects by activating STAT transcription factors, which are translocated to the nucleus through a specific nuclear localization sequence (NLS) in the hIFN-γ molecule. Two putative NLS have been pointed out in the hIFN-γ , one of which comprises residues 83-89 [14].

hIFN-γ - hIFN-γ Rα interaction The search for a mechanism for inhibition of hIFN-γ activity requires detailed information about the receptor identification by the cytokine, which we attempted to gain by means of MD simulations of the dynamics of hIFNγ binding to its extracellular receptor [15]. We used the 3D structures of hIFN-γ and the α -subunit of its extracellular receptor, published in Protein Data Bank [16] with ID 1fg9. As the important for our investigations C-termini are missing there, we added them, taking into account the fact that they do not adopt any particular secondary structure [17]. For the resulting object, full solvent MD simulations with GROMOS based force fields were performed, using GROMACS 3 simulation package [18] with implemented parallelization employing MPI at a multi-CPU facility. Our results point at the high charge density of the C-termini as the main reason for the activity of the hIFN-γ . However, the sole electrostatic interaction between them and the receptor units is not sufficient for proper formation of


FIGURE 2. Mutation site - residues Lys86 − Lys87 − Lys88 are in yellow.

the complex. Most likely, another negatively charged molecule, located on the cell surface, is involved in the process by attracting the C-termini and thus providing the fine tuning of the hIFN-γ position.

hIFN-γ and heparin derived oligosaccharides As suggested in Refs. [10, 11, 12], HS may play a key role in the formation of the hIFN-γ – hIFN-γ Rα complex. We developed a computer model of hIFN-γ homodimer interacting with two heparin-derived octasaccharide molecules (dp81 ) and the receptor units, in order to investigate the interactions of hIFN-γ with heparin-derived oligosaccharides with different molecular weights. The heparin molecules are located between the receptor units on the cell surface and attract the C-termini stronger than the receptors. As a result, the cytokine is pulled downwards to the cell surface, which allows it to adopt the correct position for proper formation of the hIFN-γ - hIFN-γ Rα complex [15].

Mutated forms of hIFN-γ Another possibility for inhibition of the biological activity of hIFN-γ is to find a mutation in the protein, which still allows it to bind to the receptors but not to induce a biological response in the cell. The NLS-sequence in the molecule of such a mutant, which is involved in cell signaling and induces a biological response, must be inactivated. To ensure the proper binding to hIFN-γ R, the mutagenesis should not affect the configuration of the active sites. We investigated the changes in the secondary structure of 100 mutants of hIFN-γ with random mutations of residues Lys86 − Lys87 − Lys88 by means of molecular dynamics simulations (Fig. 2) [19]. The alterations in tertiary structure, caused by mutation of amino acid residues 86-88 in the molecule of native human interferon gamma, were also studied. For each of the investigated 100 different mutants, the trajectory, generated by a 10ns MD simulation, was compared to the reference trajectory of the native hIFN-γ molecule. We applied two different methods for analysis of the possible secondary structure changes in the α -helix, comprising residues 86–88 — the Ramachandran plot [20, 21] and the Kabsch–Sander method [22]. The simulations of the mutated hIFN-γ forms were carried out with the NAMD 2 MD simulation package [23] with the CHARMM 22 force field [24] in simulation box with dimensions 80 × 80 × 80 A˚ 3 with periodic boundary conditions. The solvent was described by a modified TIP3P water model. The system was then coupled to a Nosè-

1

dp - degree of polymerization


FIGURE 3. Secondary structure of the helix, comprising residues 86-88 in the native form of hIFN-γ and in mutant id with Lys86 − Lys87 − Lys88 substituted by Ser86 − Leu87 − Phe89

FIGURE 4. Secondary structure of the helix, comprising residues 86–88 in the native form of hIFN-γ and in mutant id with Lys86 − Lys87 − Lys88 substituted by Cys86 − Ala87 − Pro89

Hoover thermostat to ensure a NTV ensemble. The calculations were made on 512 processing units of the IBM BlueGene/P system at Bulgarian Supercomputing Center [25]. It was found that the residues, which substitute the Lysine in position 88, have the greatest impact on the stability of the α -helix, comprising the mutated residues [19]. Depending on the secondary-structure alterations, the investigated mutations can be divided in several types. Mutations of the first type do not lead to any changes in the structure of the α -helix, comprising residues 86–88. Particularly mutations with Leucine, Threonine, Phenylalanine and Valine at position 88 do not change the structure. Moreover, the helix, comprising the mutated residues, remains stable along the whole trajectory (Fig. 3). There are two types of mutations which cause helix-breaking. In the first case, Lysine88 is substituted by a “helixdisruptor” — Proline or Glycine. The alterations in the secondary structure of the helix, comprising residues 86–88, in these mutated forms are visible already after the 2nd nanosecond of the simulation with both structure analysis methods (Fig. 4).


FIGURE 5. Secondary structure of the helix, comprising residues 86–88 in the native form of hIFN-γ and in mutant id with Lys86 − Lys87 − Lys88 substituted by Ile86 − Ser87 − Asp89

In the other case, there is no ”helix-disruptor” at position 88, but the helix still unfolds partially to residue 88. The changes in the secondary structure of these mutants occur after the 5th − 6th nanosecond of the simulation (Fig. 5). We observed also a kind of ”oscillations” in some of the mutants — the end of the helix folds and unfolds at regular intervals with significantly longer period (∼200 ps) than the one of the thermal motion in the helix.

CONCLUSIONS The basic idea in controling the hIFN-γ activity is to block either the hIFN-γ molecule, or its receptors. One way to achieve this is to single out a suitably mutated form of hIFN-γ (mutating residues from the NLS), which binds to the receptor, but lacks biological activity. By means of MD simulations encompassing 10 nanoseconds of the system evolution, we showed that the changes of the Lysine in position 88 have the greatest impact on the stability of the α -helix, comprising the mutated residues. In particular, substitution with Proline and Glycine always leads to helix-breaking, while Leucine, Threonine, Phenylalanine and Valine at the same position do not change the structure of the investigated random mutants. The observed in some of the mutants ”oscillations” — longer-period (∼200 ps) regular folding and unfolding — indicate on the possible impact of larger-scale processes on the investigated features. The whole simulations occupied about 2000 h CPU time on 512 processing units of the IBM BlueGene/P at Bulgarian Supercomputing Centre.

ACKNOWLEDGMENTS We thank Ivan Ivanov and Genoveva Nacheva for the consultations and stimulating discussions. This work was supported by Bulgarian Ministry of Education and Science under contracts BY-MI-201/2006, BY205-2006 and DO02-115/15.12.2008.


REFERENCES 1. DiMasi, J.A. et al. (2003) The price of innovation: new estimates of drug development costs. J. Health Economics 22, 151-185. 2. Tollman, P. et al. (2001) A Revolution in R&D. How Genomics and Genetics are Transforming the Biopharmaceutical Industry, The Boston Consulting Group 3. Clark, D.E. (2006) What has computer-aided molecular design ever done for drug discovery? Expert Opin. Drug Discov. 1, 103-110. 4. Markus H.J. Seifert, Kristina Wolf and Daniel Vitt, Virtual high-throughput in silico screening, BIOSILICO Vol. 1, No. 4, 2003. 5. Lipinski, C.A. et al. (1997) Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 23, 3 6. James Garner, John Deadman, David Rhodes, Renate Griffith, Paul A. Keller. A new methodology for the simulation of flexible protein-ligand interactions, Journal of Molecular Graphics and Modelling 26 (2007) 187-197 7. C.F. Wong, J.A. McCammon, Protein flexibility and computer-aided drug design, Annu. Rev. Pharmacol. Toxicol. 43 (2003) 31-45. 8. C.N. Hodge, T.P. Straatsma, J.A. McCammon, A. Wlodawer, Rational design of HIV protease inhibitors, in: W. Chiu, R.M. Burnett, R. Garcea (Eds.), Structural Biology of Viruses, Oxford Univ. Press,1997, pp. 451-473. 9. Farrar, M. A. & Schreiber, R. D. (1993) Annu. Rev. Immunol. 11:571- 611. 10. Lortat-Jacob, H., et al. (1991) J. Clin. Invest. 87, 878-83. 11. Lortat-Jacob, H., et al. (1995) Biochem. J. 310, 497-505. 12. Vanhaverbeke, C., et al. (2004) Biochem J. 384(Pt 1): 93-99. 13. Skurkovich, B., Skurkovich, S., Anti-interferon-gamma antibodies in the treatment of autoimmune diseases, Current Opinion in Molecular Therapeutics 5:52-57, 2003. 14. Zu. X.W., Jay F., The E1 functional epitope of the human interferon gamma is a nuclear targeting signal-like element. Mapping of the E1 epitope., J. Biol. Chem., 266, 6023-6026, 1991. 15. L. Litov, P. Petkov, P. Petkov, S. Markov, N. Ilieva, Understanding of Interferon-Gamma Binding, Proc. of International Scientific Conference Computer Science 2008, ISBN: 978-054-580-254-6, p. 37-42. 16. http://www.pdb.org 17. Grzesiek, S., et al. (1992) Biochemistry 31, 8180-8190. 18. Van Der Spoel D, et al. (2005), J Comput Chem 26 (16): 170118 19. E. Lilkova, L. Litov, P. Petkov, P. Petkov, S. Markov, Computer simulations of human interferon gamma mutants, accepted for printing in Meetings in Physics at University of Sofia, vol. 9, 2008. 20. Ramachandran, G. N., Ramakrishnan, C. and Sasisekharan, V., Stereochemistry of polypeptide chain configurations, J. Mol. Biol. vol. 7, p. 95-99, 1963. 21. Ramachandran, G. N. and Sasiskharan, V., Conformation of poly-peptides and proteins., Adv. Protein Chem. 23, 283-437, 1968. 22. Kabsch, W. and Sander, C., Dictionary of Protein Secondary Structure: Pattern Recognition of Hydrogen-Bonded and Geometrical Features., Biopolymers, Vol. 22,2577-2637, 1983. 23. Nelson, M., Humphrey, W., Gursoy, A., Dalke, A., Kale, L., Skeel, R. D., Schulten, K., NAMD - a Parallel, Object-Oriented Molecular Dynamics Program., J. Supercomputing App., 10, pp. 251-268, 1996. 24. MacKerell, A. D., Bashford, D., Bellott, M., Dunbrack, R. L. et. al., All-Atom Empirical Potential for Molecular Modeling and Dynamics Studies of Proteins., J. Phys. Chem. B, 102 (18), 3586-3616, 1998. 25. http://www.scc.acad.bg
