Computational determination of radiation damage ...

CEJP 1 (2003) 179{190

Computational determination of radiation damage e® ects on DNA structure Miroslav Pinak¤ Japan Atomic Energy Research Institute, Shirakata, Shirane 2-4, 319-1195 Tokai-mura, Ibaraki-ken, JAPAN Received 25 June 2002; revised 7 December 2002 Abstract: Molecular dynamics (MD) studies of several radiation originated lesions on the DNA molecules are presented. The pyrimidine lesions (cytosinyl radical, thymine dimer, thymine glycol) and purine lesion (8-oxoguanine) were subjected to the MD simulations for several hundred picoseconds using MD simulation code AMBER 5.0 (4.0). The simulations were performed for fully dissolved solute molecules in water. Signi cant structural changes in the DNA double helical structure were observed in all cases which may be categorized as: a) the breaking of hydrogen bonds network between complementary bases and resulted opening of the double helix (cytosinyl radical, 8-oxoguanine); b) the sharp bending of the DNA helix centered at the lesion site (thymine dimer, thymine glycol); and c) the ®ippingout of adenine on the strand complementary to the lesion (8-oxoguanine). These changes related to the overall collapsing of the double helical structure around the lesion, are expected to facilitate the docking of the repair enzyme into the DNA in the formation of DNA-enzyme complex. The stable DNA-enzyme complex is a necessary condition for the onset of the enzymatic repair process. In addition to structural changes, speci c values of electrostatic interaction energy were determined at several lesion sites (thymine dimer, thymine glycol and 8-oxoguanine). This lesion-speci c electrostatic energy is a factor that enables repair enzyme to discriminate lesion from the native site during the scanning of the DNA surface. c Central European Science Journals. All rights reserved. ®

¤

Keywords: molecular dynamics, DNA lesions, repair enzymes PACS (2000): 31.15.Qg

E-mail: [email protected], tel.: 81-29-282-5583, fax: 81-29-282-6768 Unauthenticated Download Date | 9/25/15 1:29 AM

180

1

M. Pinak / Central European Journal of Physics 1 (2003) 179{190

Introduction

Recognition and binding at speci¯c sites on DNA by regulatory enzymes is essential for speci¯c DNA transcription to occur. In addition to speci¯c DNA transcription, the functioning of repair enzymes in removing the damaged DNA parts is very important to ensure correct cell proliferation and to eliminate potential mutagenic cells. Several nucleotide sequences of speci¯c DNA binding sites involved in gene transcription regulation have been known to suggest the existence of a code for recognition of DNA regulatory and repair enzymes and DNA sites [1, 2, 3, 4]. Considerable information on enzyme-DNA interaction has been obtained from biological experiments. In several of these systems, both prokaryotic and eukaryotic, a DNA recognition alpha helix within the enzyme’s DNA binding domain has been observed [5, 6]. It is known that sequence-speci¯c DNA binding by repair and regulatory enzymes occurs as a result of multistage hydrogen bonding and Van der Waals interactions between the amino acid chains of enzyme and nucleotide base sites of DNA. The underlying mechanism by which enzymes recognize speci¯c or damaged sites on DNA is, however, not well established. [e.g., 7, 8]. The present paper is a report on the investigation of several lesioned DNA molecules pertaining to the enzyme-DNA interactions between amino acids and nucleotides. The method used in the study is molecular dynamics (MD) simulation. The object of our study is radiation damage, as for example 8-oxoguanine, thymine dimer, thymine glycol and cytosinyl radical, and their potential impact on the DNA structure. In particular, Van der Waals and electrostatic interaction energies are calculated. The DNA{enzyme interactions may induce breakage of the Watson-Crick nucleotide base pairing hydrogen bonds, further resulting in the bending of DNA, strand elongation and its unwinding. The formation of stable DNA-enzyme complex at the onset of repair process is also studied.

2

Method

The MD technique is the main tool utilized in the present investigation. Since DNA molecule is not a rigid and static structure, the x-ray di®raction and NMR results show only the average structural parameters. In reality, every DNA molecule is under constant thermal °uctuations, which lead to local twisting, stretching, bending and unwinding of the double helix. In this sense, MD as a simulation technique that yields static and dynamic properties of a molecular system may provide useful scienti¯c data showing the DNA in its dynamical mode. The classical method of MD is based on solving Newton’s equations of motion and thus capable of simulating the behavior of a system consisting of N atoms. Solving these equations produces new atomic coordinates that are used to calculate a new set of forces. Static and dynamic properties of the system are then obtained as time averages over the trajectory. The simulations were accomplished using the MD program package AMBER 5.0 (AMBERUnauthenticated 4.0 in the case of the cytosinyl radical) [9]. Download Date | 9/25/15 1:29 AM


181

2.1 Molecular dynamics protocol The simulated molecules were subject to several hundred picoseconds (ps), up to a maximum of 2 nanosecond (ns), of simulation under MD protocol consisting of the following sequential steps: (1) Preparation of solute molecule(s): Solute molecules were DNA segments having certain parts replaced by lesion (e.g. 8-oxoguanine). The structural and chemical parameters of the lesion were de¯ned prior to the insertion of modi¯ed parts into the solute molecule. Parameters such as lengths of chemical bonds, angles and charges, were taken from existing experimental data when available, or for small molecules were calculated by quantum chemistry methods. The structure of modi¯ed solute molecules were then optimized in order to achieve stable molecular con¯guration with minimal potential energy. (2) Locating the solute molecule in the simulation cell. (3) Neutralization of the negative charges of DNA phosphates by adding sodium counterions Na+ at the initial positions bisecting the O-P-O angle at a certain distance (¹ 5 º A) from each phosphorus atom. (4) Dissolving the solute molecules in water (several tens of thousands of water molecules were used to dissolve the solute molecule). (5) Minimization of the potential energy of the system. (6) Heating to a required temperature (e.g. 310K (36.85o C), human body temperature) during sequential MD runs. (7) Stabilization of the density of the system during constant pressure MD runs. (8) Production MD with constant volume.

2.2 Computational details Dissolving the solute molecule requires a large number of water molecules that increase the requirements on the capacity of RAM and CPU time. To be able to handle such large systems, the original AMBER 5.0 code was partly vectorized and parallelized, and installed on FUJITSU VPP5000 vector/parallel supercomputer using auto-vectorizing compiler. Its sequential and parallel °ags were modi¯ed in order to compile the program on the VPP5000 computer. After these modi¯cations and the necessary resizing, the program was capable of dealing with a system consisting of up to 100,000 atoms within a reasonable CPU time. Production MD simulations were performed on the Fujitsu VPP5000 supercomputer or on the Hitachi SR8000 parallel supercomputer. Preparatory steps such as formation of solute molecules, minimization, heating and density stabilization were performed on the scalar workstation (SUN). Supercomputers used in simulations are at the Center for Computational Science and Engineering of the Japan Atomic Energy Research Institute. The samples of CPU simulation time required to accomplish 1 ps of MD are shown in Table 1. In MD simulation, a constant dielectric Unauthenticated function was used and 1-4 electrostatic inDownload Date | 9/25/15 1:29 AM

182


Machine Execution type

SUN BLADE 1000 scalar

CPU (750MHz), scalar

248 sec.

SR8000 scalar-parallel

VPP5000 vector-parallel

1 CPU (VU-9.6G®ops, 333MHz), scalar

2434 sec.

1 CPU, vector*

286 sec.

4 CPU, vector*-parallel

91 sec.

1 CPU (1.56G®ops, 375 MHz), scalar

1914 sec.

1 node (8 CPU), parallel***

316 sec.

4 node (32-CPU), parallel***

179 sec.

Table 1 Execution time** required to accomplish 1 ps of MD simulation of the system comprising of a total of approximately 40,000 atoms. * Vector mode means execution by auto-vectorized compilation, vectorization ratio is 96% ** Execution time is the elapsed time *** Pseudo-vectorization function, i.e. fast supply of data from memory for the CPU processing

teractions (electrostatic interactions separated by only three bonds), were scaled by a factor 1.2, the recommended value for AMBER 5.0 force ¯eld. Particle Mesh Ewald Sum technique [10], was implemented in AMBER 5.0. In this method a Gaussian charge distribution of opposite sign is superimposed upon the original point charges, producing a screened charge distribution, with consequent short-range electrostatic interaction. The original distribution is recovered by adding a second Gaussian charge distribution identical to the ¯rst, but of opposite sign. No cut-o® distance was applied in the calculation of electrostatic interactions and thus all water molecules in the system were included. The Van der Waals interactions were calculated within the de¯ned cut-o® distance (usually 10-12 º A). Periodic Boundary Conditions were applied throughout the entire simulation to eliminate undesirable edge e®ects.

3

Results and discussions

3.1 Cytosinyl radical (5-hydroxy-6-cytosinyl) This is lesion produced by indirect radiation, i.e. by interaction of active OH radical with cytosine base [11] (Figure 1). This lesion is important in the study of strand break formation through the intramolecular process of H-abstraction form sugar (pentose) and emphasizes the importance of initial base damage in relation to strand breaks. The 200 ps of MD simulation of DNA dodecamer d(CGCGAATTC¤ GCG) 2 with cytosinyl radical at the position 9 - C¤ (9) revealed a strong bending at the A(6) and T(7) DNA segment originating from the lesioned Unauthenticated DNA. Since this large bending was not observed at the damaged site - C¤ (9), it suggests Download Date | 9/25/15 1:29 AM


183

Fig. 1 Structure of 5-hydroxy-6-cytosinyl radical (axial position of OH is marked by shadow).

intramolecular interactions among C¤ (9) and A(6) and T(7). In addition to the bending, large distortions and disruptions of hydrogen bonding network between bases of neighboring pairs was observed. In the identical simulation of non-lesioned DNA molecule of the same nucleotide sequence, bending was not observed and the network of hydrogen bonds was well preserved.

3.2 Thymine dimer (5,6 cis.sin cyclobuthane thymine dimer (TD)) This is photolesion produced by ultraviolet (UV) radiation in sunlight and is one major factor causing skin cancer. It is formed as the covalently bonded complex of two adjacent thymines on the single strand of DNA. This damage is very frequent but almost 90% of TD’s are repaired within a short time, order of minutes, and only a few are experimentally observable and produce future changes on cell level, [12]. This study was conducted with DNA dodecamer d(TCGCGT ^ TGCGCT) 2 , where T T indicates the thymine dimer. The results of 600 ps of MD simulation shows that this lesion does not disrupt the double helical structure and the hydrogen bonds are well preserved throughout the simulation. Thymine dimer-lesioned DNA, compared with the native one, has a sharp bending at the TD site which is originated by the two covalent bonds C(5)-C(5) and C(6)-C(6) between the adjacent thymine bases forming the thymine dimer, (Figure 2). Thymine dimer is repaired by the repairUnauthenticated enzyme T4 Endonuclease V that slides on Download Date | 9/25/15 1:29 AM

184


Fig. 2 Thymine dimer as a composition of two adjacent thymine bases covalently joined between C(5)-C(5) and C(6)-C(6) atoms of adjacent thymine bases.

non-target sequences and progressively incises at all dimers within the DNA molecule. This enzyme binds to the DNA double strand in a two-step process: at ¯rst it scans non-target DNA by electrostatic interactions in search of damaged sites, and then it sequentially and speci¯cally recognizes the dimer sites. The observed bending at the TD site facilitates the docking of enzyme into DNA. Docking is further facilitated by the complementary structural shapes of the repair enzyme and bent DNA (Figure 3). The dynamical process of binding of T4 Endonuclease V to thymine dimer-lesioned DNA was simulated with the MD method. Considering the limitations arising from the simulations of large systems and requirements for CPU time, only the catalytic center of the enzyme was subjected to the simulations. Glutamic acid 22 is the key amino acid of the enzyme, in which the carboxyl chain plays a crucial role in the cleavage of N-glycosyl bond in DNA (base excision repair). This amino acid, together with the surrounding 9 amino acids (8 of H1 and 2 of H2 helices) was selected to form the simulated part of the enzyme. After nearly 100 ps of the MD simulation, the catalytic part of the enzyme approached the DNA at the thymine dimer site, docked into it, and the molecular complex (DNA + catalytic center of enzyme) remained stable afterwards (the simulation was performed for 500 ps) (Figure 4). When the same simulation was performed with the non-lesioned native DNA molecule, the catalytic center did not dock into the DNA molecule and the molecular complex was Unauthenticated not formed. In further consideration of factors that caused the fusion of the DNA and Download Date | 9/25/15 1:29 AM


185

Fig. 3 Structure of T4 Endonuclease V and thymine dimer-lesioned DNA at 300 ps of MD. The arrows show the position of the catalytic center of the enzyme and the dimer on the DNA.

repair enzyme, the electrostatic interaction energy between the dimer lesion and catalytic center was calculated. It was found that while the electrostatic energy of thymine dimer was negative, about {10 kcal/mol, the electrostatic energy of glutamic acid 23 (the closest amino acid to the C5’ atom of phosphodiester bond of dimer) was positive, about +10 kcal/mol. The value of the electrostatic energy represents the total electrostatic interaction of the selected molecules calculated with Particle Mesh Ewald Sum technique for in¯nite simulated volume of repeating units through periodic boundary conditions, i.e., no cut-o® distance was applied. Since the electrostatic energy of the native thymine is nearly zero, the value of electrostatic energy represents a factor discriminating the thymine dimer lesion from the native thymine [13].

3.3 Thymine glycol (5,6-dihydroxy-5,6-dihydro-pyrimidine) This is observed in DNA after irradiation in vitro as well as in vivo and after oxidation Unauthenticated by chemicals (Figure 5). Download Date | 9/25/15 1:29 AM

186


Fig. 4 Complex of thymine dimer lesioned DNA molecule with catalytic center of repair enzyme T4 Endonuclease V formed during 100 ps of MD simulation.

Fig. 5 Molecule of the thymine glycol (5,6-dihydroxy-5,6-dihydrothymidine). Unauthenticated Download Date | 9/25/15 1:29 AM


187

Thymine glycol is known as lesion-causing Cockayne Syndrome, an inherited disorder in which people are sensitive to sunlight, have short stature and have the appearance of premature aging. It is repaired with the repair enzyme Endonuclease III, which removes a number of damaged pyrimidines from DNA via its glycosylase activity and also cleaves the phosphodiester backbone at apurinic/apyrimidinic sites via a ¼-elimination mechanism. To study the time evolution of the recognition processes of TG-lesioned DNA by repair enzyme Endonuclease III, the MD simulation of the following molecules was performed: DNA 30-mer d(CCAGCGCACGACGCA’TG’GCACGACGACCGGG) 2 where `TG’refers to thymine glycol; and repair enzyme Endonuclease III [14, 15]. Analysis of the results of 1 ns MD simulation showed that the double helical structure and hydrogen bonding were well preserved through the simulation (except the base pair of cytosine C5’ { guanine C3’ end, in which hydrogen bond pairing collapsed after 850 ps). DNA began to bend at the thymine glycol site after 500 ps of MD and bending continued until simulation was terminated. A kink was observed at the TG site. The bending associated with the kink-dislocated glycosyl bond at C5’ atom closer to the DNA surface, enables it to be eventually accessed by repair enzyme (Figure 6).

Fig. 6 Snapshots of DNA molecule during the course of MD simulation. DNA molecule is shown from the same side and angle with respect to the simulation box. The cytosine C5’ end and guanine C3’ end of DNA molecule are shown. It is seen that the molecule at 600, 800 and 1000 ps stimulation is bent and kinked at the thymine glycol site (shown as a Connolly surface). Bending is expressed as the value of the angle measured between phosphates of the guanine (position 41), thymine glycol (position 16) and guanine (position 13) (numbers in degrees). Unauthenticated Download Date | 9/25/15 1:29 AM

188


3.4 8-oxoguanine (7,8-dihydro-8-oxoguanine). This is formed by oxidation of a guanine base in DNA, (Figure 7). It is considered to be one of the major endogenous mutagens contributing broadly to spontaneous cell transformation and its frequent mis-pairing with adenine during replication increases the number of G-C ! T-A transversion mutations. This mutation is one of the common somatic mutations found in human cancers. The 8-oxoguanine is recognized and subsequently repaired by the DNA glycosylase (hOGG1 in humans). DNA glycosylases acting on single-base lesions use an extrahelical repair mechanism during which the enzyme recognizes oxidative damaged guanines and excludes normal DNA bases. The study of the 8-oxoguanine (8-oxoG) lesion enables us to describe structural and energetic changes in the DNA molecule which are caused by this lesion, and to discuss how these changes may be signi¯cant in the formation of a complex repair enzyme. The MD simulations (2 ns) of two B-DNA molecules - native DNA 15-mer, d(GCGTCCAGGTCTACC)2 and 8-oxoG lesioned DNA 15-mer, d(GCGTCCA’8-oxoG’GTCTACC)2 , were performed.

Fig. 7 Molecule of the 8-oxoguanine (7,8-dihydro-8-oxoguanine)

The disruptions of weak hydrogen bonds between respective bases in the 8-oxoG lesioned DNA molecule caused locally collapsed B-DNA structure. While the hydrogen bonds between 8-oxoG and opposite cytosine 23 were well preserved, the neighboring base pairs (adenine 7 { thymine 24, and guanine 9 { cytosine 22) were broken. The hydrogen bonding of base pair thymine 10 { adenine 21 ceased to exist very early (after about 50 ps of MD simulation). In the case of the native DNA, the B-DNA structure around native guanine 8 was well preserved. Adenine 21 on the complementary strand (separated from 8-oxoG by 1 base pair) was completely °ipped-out of the DNA double helix (Figure 8). This extrahelical position was caused by the disrupted hydrogen bonds and by the strong electrostatic repulsion between the atoms in the region after 1 ns of MD. The cytosine 22 was also severely dislocated form its intrahelical position and its hydrogen bonding to guanine 9 was absent. The extrahelical position of adenine 21 formed a hole in the double helix that may favor docking of repair enzyme into the DNA during the repair process. Unauthenticated The °ipped-out base may also be inserted into the enzyme cavity further ensuring the Download Date | 9/25/15 1:29 AM


189

stability of DNA-enzyme complex [16].

Fig. 8 Flipped-out adenine 21 on the complementary strand-to-strand with 8-oxoG. The gure also indicates the absence of hydrogen bonds between guanine 9 and cytosine 22, since the cytosine 22 is severely dislocated from its intrahelical position.

4

Conclusions

The present paper is a report on the results of MD simulation of several di®erent radiation lesions on the DNA molecule. The studied lesions were three pyrimidine base lesions { cytosinyl radical, thymine dimer and thymine glycol, and one purine lesion { 8-oxoguanine. Except for thymine glycol, the other three lesions are believed to produce neoplasic transformation of the cell and are frequently found in human cancers. The common features observed in all lesions are the speci¯c e®ects present at the lesion site, like disruption of hydrogen bonding networks (cytosinyl radical, 8-oxoguanine), sharp bending at the lesion site (thymine dimer, thymine glycol), °ipping-out of the base on the strand complementary to the lesion, and speci¯c values of the electrostatic interaction energy at the lesion (thymine dimer, 8-oxoguanine). The most important among these changes is the °ipping-out of the base, since it creates a hole in the DNA double strand, which may serve as a template for the docking of the enzyme and for the formation of the DNA-enzyme complex. The strong bending that was observed in the thymine dimer-lesioned DNA molecule forms a complementary shape in respect to the repair enzyme T4 Endonuclease V and further facilitates the formation of the complex. The electrostatic interaction energy at several lesion sites di®er from the values at the native DNA sites (thymine dimer, thymine glycol, 8-oxoguanine), which facilitates the proper recognition of the respective lesion by discriminating the lesion from theUnauthenticated native site. This recognition is important Download Date | 9/25/15 1:29 AM

190


during the electrostatic scanning of the DNA surface by repair enzyme. It is hoped that the results of MD simulations, in conjunction with the existing crystallographic and molecular biology techniques may augment the investigations of DNA damage and its repair by the dynamical description of the structural and chemical processes that are undergoing at the lesioned DNA molecule. This investigation based on MD simulations is also intended to determine key factors in the process of lesion recognition by the repair enzyme.

Acknowledgments The author wishes thank to Mr. Toshiyuki Nemoto of The Research Organization for Information Science and Technology for the installation, maintenance and adjustment of the AMBER 5.0 code on supercomputers VPP5000 and SR8000. The valuable support from the all members of The Radiation Risk Analysis Laboratory, JAERI Tokai Research Establishment is also gratefully acknowledged.

References [1] Harrison, S. and Aggarwal, A., Annu. Rev. Biochem. 59 (1990) 933. [2] Gicquel-Sanzey, B. and Cossart, P., EMBO J. 1 (1982) 591. [3] Ham, J., Thompson, A., Nedham, M., Webb, P. and Parker, M., Nucleic Acid Res. 16:12 (1988) 5263. [4] Beato, M., Cell 56 (1989) 335 [5] Harris, L, Sullivan, M. and Hickok, D., Computers and Mathematics with Applications 20 (1990) 25. [6] Marx, J., Science 229 (1985) 846. [7] Matthews, B., Nature 335 (1988) 294. [8] Harris, L, Sulliwan, M. and Hickok, D., Proc. Natl. Acad. Sci. USA 90 (1993) 5534. [9] Case, D.A., Pearlman, D.A., Caldwell, J.W., Cheathman III, T.E., Ross, W.S., Simmerling, C.L., Darden, T.A., Merz, K.M., Stanton, R.V., Cheng, A.L., Vincent, J.J., Crowley, M., Ferguson, D.M., Radmer, R.J., Seibel, G.L., Weiner, P.K. and Kollman, P.A., AMBER 5.0, (1997) University of California San Francisco. [10] Smith, P.E. and Petit, B.M., J. Chem. Phys. 105 (1996) 4289. [11] Pinak, M., Yamaguchi, H. and Osman, R., J. Radiat. Res. 37 (1996) 20. [12] Pinak, M., J. Mol. Struct.: THEOCHEM 466 (1999) 219. [13] Pinak, M., J. Mol. Struct.: THEOCHEM 499 (2000) 57. [14] Pinak, M., JAERI-research 2001-038, (2001). [15] Pinak, M., J. Comput. Chem. Vol. 22, Iss.15 (2001) 1723. [16] Pinak, M. J. Mol. Struct.: THEOCHEM 583/1-3 (2002) 189. Unauthenticated Download Date | 9/25/15 1:29 AM