Molecular modelling and simulation of retroviral proteins and nanobiocomposites

List of contents

List of tables

List of figures

General introduction

Motivation for present research

1 Structural and docking analysis of HIV-1 integrase and Transportin-SR2 interaction: Is this a more general and specific route for retroviral nuclear import and its regulation?

1.1 Overview

1.2 The problem to solve

1.3 Computational methods

1.3.1 Structural analysis software

1.3.2 Sequence analysis software

1.3.3 Docking programs

1.4 Results and discussion

1.4.1 Highly similar sequences of TR-SR2 and TR-SR1 have different 3D folding domain structures

1.4.2 Where does HIV-1 integrase bind to TR-SR2?

1.4.3 Hydrogen bonds involved in the binding between HIV-1 IN NLS and the H8-loop of TR-SR2 Ran-GDP binding domain

1.4.4 Role of highly accessible and hydrophilic amino acids in HIV-1 IN and TR-SR2 interaction

1.4.5 Is this a general type of viral transport interaction?

1.5 Conclusions

2 Role of the central polypurine tract in retroviral nuclear import (analysis of the HIV-1 central polypurine tract in a foamy virus vector background)

2.1 Overview

2.2 A molecular biology challenge

2.3 Materials and methods

2.3.1 Materials and solutions

2.3.2 General molecular genetics methods

2.3.3 General cell biology methods

2.4 Results and discussion

2.5 Conclusions

3 Lead expansion and virtual screening of Indinavir derivate HIV-1 protease inhibitors using pharmacophoric - shape similarity scoring function

3.1 Overview

3.2 The strategy

3.3 Computational metho ds

3.3.1 HIV-1 subtype C protease and Indinavir structures

3.3.2 Protease active site detection

3.3.3 Compound library generation

3.3.4 ADME/Tox Studies

3.3.5 Protein-ligand docking

3.3.6 Construction of pharmacophore models

3.4 Results and discussion

3.5 Conclusions

4 Molecular dynamics simulation of POPC and POPE lipid membrane bilayers enforced by an intercalated single-wall carbon nanotube

4.1 Overview

4.2 The simulation and its goals

4.3 Computational methods

4.4 Results and discussion

4.5 Conclusions

Concluding discussion

Summary

Zusammenfassung

Acknowledgments

List of references

Publications related to this work

Conference contributions and participations

Training courses, workshops and lectures

List of tables

Table 1.1 Protein-protein docking

Table 1.2 Docking clustering

Table 1.3 Docking conformations

Table 1.4 Close contact residues that could be involved in the H-bond formation

Table 1.5 Residue's accessibility area and hydrophobicity index of docked molecules

Table 1.6 Different viral nuclear localization signals contain different kinase binding mo- tifs

Table 2.1 Main characteristics of the inserts using for the construction of modified expression vectors

Table 3.1 Comparative characteristics of IDV and novel hit

List of figures

Figure 1.1 Three-domain structure of HIV-1 IN. The sequence of the 161-173 region of the HIV-1 virus is shown and the two residues essential for NLS function are indicated in black (figure modified after Bouyac-Bertoia et al., 2001)

Figure 1.2 Sketch of the computational algorithm implemented in the CAVER. The black bold circle represents the starting point. The protein is visualized by gray circles with Van der Walls atom radii mapped on a discrete grid (black dots). The solid line represents the boundary between the protein (convex hull) interior and its surroundings. Empty circles represent the maximally inscribed balls on the probable route (dashed line)

Figure 1.3 Sequence alignments and superimposition of TR-SR1 and TR-SR2 molecules. The root mean square (RMS) value calculated with the UCSF Chimera v.1.3 (see section Materials and Methods) is 2.0 Ä between the two structures: crystal TR-SR1 and TR- SR2 (homology model). The superimposed regions are marked in red

Figure 1.4 (A) TR-SR1 and TR-SR2 superimposition, coloured in gray and blue respectively. Superimposition was performed on the basis of Needleman-Wunsch-like alignment algorithm. Scores for aligned characters were specified by similarity substitution matrix (BLOSUM-62), which is widely used to score alignments between evolutionary divergent protein sequences (Needleman and Wunsch, 1970; Henikoff et al., 1992). Sequences were iterated by pruning long atom pairs until no pair exceeded 2.0 Ä RMSD. (B) The rigid docking of HIV-1 IN CCD and TR-SR2. Protein domains painted in different colours: HIV-1 IN CCD (162 aa) is green; NLS (13 aa) is red; Ran-GDP BD (1-303 aa) is grey; Acidic loop (304-378 aa) is yellow; CBD (379-

aa) is blue, RMSD=1.6 Ä

Figure 1.5 (A) H8-loop binding center detected by CAVER module. (B) The second confor-mation of HIV-1 IN NLS and TR-SR2 H8-loop flexible docking. Labelled amino ac-ids are forming H-bonds between two molecules (see text for details). Molecules are coloured according to their atom composition. Wire frame spheres are displayed on the atoms with the hydrogen bonds.

Figure 1.6 H8-loop of TR-SR2 (A) and the HIV-1 IN NLS amino acids (B) are represented. There is a correlation between residue's accessibility and hydrophobicity of the amino acids with tendency to H-bond formations. Highly accessible amino acids have very low hydrophobicity index

Figure 1.7 Scheme of the integrase - transportin comp lex. (A) integrase - H8-loop (HIV-1 integrase-yellow, NLSs-white, H8-loops-magenta), (B) integrase - TR-SR2s. The TR- SRs are blue and red respectively

Figure 2.1 (A) mature and fully assembled retroviral virion. (B) The complex organization of lentiviral preintegration complex (PIC). It includes the viral integrase (IN), reverse transcriptase (RT), matrix part (MA) of Gag and some auxiliary proteins, such as Vpr. The PIC exceeds more than two times the size of the nuclear pore central channel within the nuclear pore complex yet is able to successfully negotiate into the nucleus. (C) In brief the HIV life cycle comprises cognate co-receptor initiating fusion, uncoat-ing, reverse transcription, nuclear import, virion production and budding (figure modified after Sherman and Greene, 2002)

Figure 2.2 Different reverse transcription outcomes in lenti- and foamy viruses

Figure 2.3 Structural elements of pWPXL and pMD9 parental plasmid backbones are shown. Foamy viral central polypurine tract (cPPT) was replaced with the HIV cPPT in different variations

Figure 2.4 Timescale representation of the transduction experiments with Aphidicolin sub-stance to stop a cell cycle (G1/S phase). DMSO was added to observe its effect upon the A549 cell-line in pure solution and as a solvent for the drug

Figure 2.5 Summarized statistics of retroviral replication efficiency in A549 cells. pMD9 mutants were tested to efficiently transduce non-dividing cells (G1/S phase of cell cycle). Three conditions were analyzed in the experiment: pure medium (A); DMSO

(B); Aphidicolin + DMSO

(C.)

Figure 3.1 Crystallographic structures of the HIV-1 protease. The HIV-1 protease labelled according to its resemblance to an English bulldog (figure modified after Perryman et al., 2004). The blue and cyan-green ribbons depict the peptide backbone of a wildtype and a mutant structure, respectively

Figure 3.2 Chemical structure of IDV and its 'derivate'- novel hit (N'-cyano-N-{3-[(1S, 2S)-1,2-diamino-2-(4-oxo-3,4-dihydroquinazolin-7-yl)ethyl]-5-[(2R)-oxolan-2-yl]pyridin- 2-yl} benzenecarboximidamide)..

Figure 3.3 Outline of the evolutionary algorithm

Figure 3.4 (A) HIV protease catalytic tunnel (binding site) was predicted by the PyMol CAVER module. IDV (B) and novel hit (C) interactions with the HIV-1 protease are shown. H-bonds are depicted as dashed lines. IDV - protease complex was analyzed as a crystal structure. 3D alignment and 'fuzzy' model (D) of the hit 'native' conforma-tion (gray) together with its 'functional' conformation (yellow) at 2.0 Ä RMSD. The shown potential pharmacophore points are colour-coded as follows: lipophilic areas are green, H-bond donors and acceptors are coloured in blue and red, respec- tively.57 Figure 3.5 (A) Indinavir and (B) novel hit docking profiles calculated by the AutoDock software. Both ligands are coloured in gray. Chains A and B of the HIV protease and their amino acids are painted in green and magenta respectively. The hydrogen bonds are shown for each of both molecules.

Figure 4.1 Snapshots, from the top and the side, of the membrane-CNT unit cell consisting of 256 DMPC lipids, 2560 coarse grain water sites representing 7680 water molecules, and one 10-ring narrow intercalated single-wall carbon nanotube. The six inner hy-drophobic CNT rings are coloured white whereas the hydrophilic rings are coloured blue. The lipid tails are shown in yellow and the headgroups in red, purple, and green. The water, which is suppressed in the top view, is coloured in blue (figure modified after Nielsen et al., 2009)

Figure 4.2 Self-assembly of DPPC and single-wall CNT for 46 DPPC molecules using GROMOS96 parameters. Front view of the simulation system configuration at 15 ns (figure modified after Wang et al., 2009)

Figure 4.3 (A) Armchair carbon nanotube formed from a graphite sheet that is rolled up so that the edge is in the shape of armchairs. High symmetry armchair CNTs occur for m = n, where m and n are chiral indices, respectively. (B) The construction of the unit cell for a CNT (figure modified after Dresselhaus et al., 1995)..

Figure 4.4 Schematic representation of the lipid membrane bilayer stabilized by a single-wall carbon nanotube: (A) Lipid membrane; (B) CNT structure; (C) Membrane-CNT complex. CNT (hydrogen atoms removed), water and lipid molecules are given in 'space - Alling' and 'steak' representations, respectively

Figure 4.5 (A) Constant temperature parameters and (B-F) RMSD values are shown during 1000 ps evolution time. All RMSD values of investigated MD systems are represented with respect to their initially minimized structures at different temperature lev- els.

Figure 4.6 Visualization of the molecular dynamics trajectories (multiple frames) at different temperature parameters (300, 300-400 and 400 K): (A1-A3) Single-wall carbon nanotube; (B1-B3) POPC membrane; (C1-C3) POPE membrane; (D1-D3) POPC-CNT complex; (E1-E3) POPE-CNT complex. Images of every hundredth frame are shown simultaneously to make the large-scale motion of the system more apparent. Molecules are represented as carbon frameworks

Figure 4.7 Comparative characteristics of the average root mean square deviation (RMSD) values of different simulated structures at different temperature parameters (300, 300400 and 400 K)

Figure 4.8 Root mean square fluctuations (RMSF) of carbon atoms at different temperature parameters (300, 300-400 and 400 K) are represented for: (A) Single-wall carbon nanotube; (B) POPC membrane; (C) POPE membrane; (D) POPC-CNT complex; (E) POPE-CNT complex during molecular dynamics simulation. The periodic pattern shows the position of carbon atoms in CNT structure as sharp peaks interspaced by low fluctuating atoms. The calculated RMS fluctuations (fluctuations of the water molecules are not shown) show that amplitude is minimal at the ends of the nanotube. The peaks of increased flexibility are represented in the nanotube 'body' due to the low binding frequency motion of the CNT. Peaks at identical position relate to the corresponding atoms in different models. Atom numbering is from one end of the CNT to another

Figure 4.9 Comparative characteristics of the average root mean square fluctuation (RMSF) values of different simulated structures and substructures at different temperature parameters (300, 300-400 and 400 K): (A) RMSF average values of 'native' CNT system and the CNT substructures from different membrane-CNT systems; (B) RMSF average values of 'native' POPC, POPC-CNT systems and POPC substructure; (C) RMSF average values of 'native' POPE, POPE-CNT systems and POPE substructure. All substructural average RMSFs were calculated with respect to initial RMSF values of represented substructures, extracted from the corresponding dynamically simulated systems

General introduction

Molecular modelling and simulation of versatile biological systems is a very powerful toolbox in modern bioinformatics, and enables to follow and understand structure and dynamics with extreme detail on scales where motion of individual atoms can be tracked. Different biological systems such as lipid membrane bilayers and retroviral proteins play an im-portant role in physiology and pathogenesis of living cells. The availability of high-resolution crystallographic structures, together with the development of detailed atomic models and molecular dynamics algorithms, provide a unique opportunity to refine our understanding of these systems. Although the complexity of biological structures does present a formidable challenge to theoretical studies, even with modern computational resources, it is particularly encouraging to note that many of the recent results from simulations have been consistent with the information emerging from higher resolution structural data. This relative success relies for a large part on computational strategies involving docking studies and molecular dynamics simulations.

This thesis focuses on the most commonly used methods, namely, molecular dynamics simulations, homology modelling and molecular docking which, respectively, optimize structures and simulate the natural motion of biological macromolecules with various materials and dock ligands to protein targets. The common theoretical framework based on statisti-cal mechanics is covered briefly as well as limitations of the computational approach, for instance, the lack of quantum effects and limited timescales.

Computational techniques provide other options for understanding chemical systems, which yield information that is difficult, if not nearly impossible, to obtain in laboratory analysis. The advent of large genome sequencing reinforced the observation that structural information is needed to understand the detailed function and mechanism of biological molecules such as enzyme reactions and molecular recognition events. Molecular modelling of proteins aims primarily at establishing sequence-structure-function relationships for biological molecules using in silico techniques. This discipline emerged about 40 years ago (Levitt, 2001) and has made much progress in the past decade. Knowledge of the three-dimensional (3D) structures of proteins provides invaluable insights into the molecular basis of their func-tions. Computational methods for predicting the 3D structures of proteins enjoy a high degree of interest and are the focus of many research and service development efforts. Despite an intensive effort in molecular modelling to account for the extensive thermodynamics data on proteins in pure water and solutions, some protein structures remain incompletely understood at the molecular level. Sometimes, there is no substitute for practical laboratory experience, but computer modelling methods play an important role both as an aid in interpreting experi-mental results and as means of explaining these results. Molecular modelling is now used not just in chemistry but in a wide range of subjects such as pharmacology, biology and biophys-ics. One example of the protein computer modelling approach is to provide a reasonably ac-curate first guess at a structure which can then be used in methods such as X-ray diffraction of powders which, unlike the X-ray diffraction of single crystals, do not provide enough information to determine the total structure from scratch.

In this thesis, my own results and relevant literatures on different important aspects of relevant molecular modelling methods are reviewed and discussed. The physiological function and chemical properties of different biological systems might be modelled with various sets of following techniques such as homology modelling, docking studies, molecular dynamics etc.

The remaining chapters of the thesis are organized as follows. In Chapter 1, detailed results and review of relevant literature on different important aspects of structural and docking analysis of HIV-1 integrase and Transportin-SR2 interaction, modelling and simulation techniques are described. Adopted experimental methodology to analyze the role of the central polypurine tract in retroviral nuclear import in context of a foamy virus vector back-ground is described in Chapter 2. Chapter 3 is devoted to the investigation of the HIV protease inhibitor Indinavir derivatives using pharmacophoric - shape similarity scoring function. Chapter 4 focuses on the CNT interactions and molecular dynamics simulation of palmitoy-loleoylphosphatidylcholine and palmitoyloleoylphosphatidylethanolamine membrane bilayers enforced by an intercalated single-wall carbon nanotube. Finally, concluding remarks from all the studies along with recommendations for future research are summarized in concluding discussion section.

Motivation for present research

Molecular modelling and simulation are powerful methods in providing important information on different biological systems to elucidate their structural and functional properties, which cannot be determined in experiment. These methods are applied to analyse the versatile biological systems: POPC/POPE lipid membrane bilayers stabilized by an interca-lated single wall carbon nanotube and retroviral proteins such as HIV protease and integrase.

HIV-1 integrase has nuclear localization signals (NLS) which play a crucial role in nuclear import of viral preintegration complex (PIC). However, the detailed mechanisms of PIC formation and its nuclear import transport are still unknown. Previously it was shown that NLSs bind to the cell transport machinery e.g. proteins of nuclear pore complex such as transportins. I investigated the interaction of this viral protein HIV-1 integrase with proteins of the nuclear pore complex such as transportin-SR2. I showed the possible reasons for role of the transportin-SR2 in a nuclear import via its interaction with the HIV-1 integrase. I ana-lyzed key domain interaction and hydrogen bond formations in transportin-SR2. These results were discussed in comparison to other retroviral species such as foamy viruses to better understand this specific and efficient retroviral trafficking route.

Experimentally, the retroviral nuclear import was analyzed by investigating the retroviral ability to infect nondividing cells. To accomplish gene transfer task successfully, retro-viruses must efficiently transduce different cell cultures at different phases of cell cycle. However, promising and safe foamy viral vectors used for gene transfer therapy are unable to efficiently infect quiescent cells. This drawback was due to their inability to create a preinte-gration complex (PIC) for nuclear import of retroviral DNA. On the contrary, the lentiviral vectors are not dependant on cell cycle. In the course of reverse transcription the polypurine tract (PPT) is believed to be crucial for PIC formation.

In this thesis, I compared the transduction frequencies of PPT modified FV vectors with lentiviral vectors in nondividing and dividing adenocarcinomic human alveolar basal epithelial cells (A549) by using molecular cloning, transfection and transduction techniques and several other methods. In contrast to lentiviral vectors, FV vectors were not able to efficiently transduce nondividing cell in my hands. Despite these findings, which support the use of FV vectors as a safe and efficient alternative to lentiviral vectors, there is still a big limita-tion in terms of foamy-based retroviral vector gene therapy in quiescent cells.

Many attempts have been made in the recent years to search for the potential new drugs to treat HIV infection. These molecules can be retrieved from chemical libraries or can be designed on a computer screen and then synthesized in a laboratory. Most notably, one could use the computerized structure as a reference to determine the types of molecules that might block the enzyme. Such structure-based drug design strategies have the potential to save off years and millions of dollars compared to a more traditional trial-and-error drug development process.

After the crystal structure of the HIV-encoded protease enzyme had been elucidated, computer-aided drug design played a pivotal role in the development of new compounds that inhibit this enzyme which is responsible for HIV maturation and infectivity. Promising repre-sentatives of these compounds have recently found their way to patients. Protease inhibitors show a powerful sustained suppression of HIV-1 replication, especially when used in combi-nation therapy regimens. However, these drugs are becoming less effective to more resistant HIV strains due to multiple mutations in the retroviral proteases.

In computational drug design I used molecular modelling methods such as lead expansion algorithm (Tripos®) to create a virtual library of compounds with different binding affinities to protease binding site. In addition, I heavily applied computer assisted combinato-rial chemistry approaches to design and optimize virtual libraries of protease inhibitors and performed in silico screening and pharmacophore-similarity scoring of these drug candidates. Further computational analyses revealed one unique compound with different protease binding ability from the initial hit and its role for possible new class of protease inhibitors is dis-cussed in the appropriate chapter.

A number of atomistic models were developed to elucidate the nanotube behaviour in lipid bilayers. However, none of them provided useful information for CNT effect upon the lipid membrane bilayer for implementing all-atom models that will allow us to calculate the deviations of lipid molecules from CNT with atomistic precision. Unfortunately, the direct experimental investigation of nanotube behaviour in lipid bilayer remains quite a tricky problem opening the door before the molecular simulation techniques. In this regard, more detailed multi-scale simulations are needed to clearly understand the stabilization characteristics of CNTs in hydrophobic environment.

The phenomenon of an intercalated single-wall carbon nanotube in the center of lipid membrane was extensively studied and analyzed. The root mean square deviation and root mean square fluctuation functions were calculated in order to measure stability of lipid mem-branes.

The results indicated that an intercalated carbon nanotube restrains the conformational freedom of adjacent lipids and hence has an impact on the membrane stabilization dynamics. On the other hand, different lipid membranes may have dissimilarities due to the differing abilities to create a bridge formation between the adherent lipid molecules. The results de-rived from this thesis may be of importance in developing stable nanobiocomposites for con-struction of novel biomaterials and delivery of various biomolecules in the field of materials science.

Chapter 1

1.1 Overview

HIV-1 integrase has NLS (nuclear localization signals) which plays an important role in internuclear transport of viral PIC (preintegration complex). The exact mechanisms of PIC formation and its internuclear transport are not known. It was shown that NLSs bind to the cell transport machinery e.g. proteins of nuclear pore complex such as transportins. I investi-gated the interaction of this viral protein with proteins of the nuclear pore complex (trans-portin-SR2). I showed reasons why transportin-SR2 is the nuclear import protein for HIV-1 integrase and not transportin-SR1: (i) 3D alignments identify differences between transportin-SR1 and transportin-SR2. (ii) Rigid protein-protein docking showed key domain interactions and hydrogen bonds available to transportin-SR2. (iii) Flexible receptor-ligand docking was performed to reveal crucial amino acid residues involved in this hydrogen bond formation. These results are discussed to better understand this specific and efficient retroviral transport route comparing the interactions of related retroviruses (SIV, HIV-2, PFV etc.) with their cognate transport proteins, NLS sequences and kinase binding motifs.

1.2 The problem to solve

In contrast to other viruses, retroviruses generate complex and structured preintegration complexes (PICs) to transport reverse transcribed RNA genomes into the host nucleus. In addition to cellular proteins, HIV-1 PIC consists of viral proteins such as reverse transcriptase (RT), integrase (IN), matrix protein (MP), auxiliary protein Vpr and viral DNA with a flap region. The PIC has the ability to bind proteins of the nuclear pore complex (NPC) to enter the nucleus and integrate into the host genome (Bushman et al., 1990; Engel-man et al., 1991). To penetrate through nuclear membrane PIC proteins must bear nuclear localization signals (NLS) for recognition by the proteins of the nuclear pore complex.

[...]