Group Title: computational atomistic model of radiation damage to DNA
Title: A computational atomistic model of radiation damage to DNA
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00100741/00001
 Material Information
Title: A computational atomistic model of radiation damage to DNA
Physical Description: Book
Language: English
Creator: Aydogan, Bulent, 1969-
Publisher: University of Florida
Place of Publication: Gainesville Fla
Gainesville, Fla
Publication Date: 2001
Copyright Date: 2001
 Subjects
Subject: Nuclear and Radiological Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Nuclear and Radiological Engineering -- UF   ( lcsh )
Genre: government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )
 Notes
Summary: ABSTRACT: A review of past and current biophysical models of DNA damage reveals that current DNA damage models have become increasingly complex in their attempts to model the full 3D structure of the nucleosome and chromatin fiber. As such, many of the finer details of direct, quasi-direct, and indirect action on DNA become difficult to study in isolation. Also, experimental comparisons that seek to validate these models become increasingly difficult to make. A better approach may be to perform the atomistic modeling of direct, indirect, and quasi-direct effects in total isolation from considerations of the macroscopic conformation of the DNA target. This would permit the highly detailed atomistic modeling to be performed only once in order to produce a database of outcome probabilities that can then be used in radiation chemistry modeling of different and more complex conformations of double-stranded DNA. This work is performed to establish the groundwork to accomplish this goal. A system of Monte Carlo computer codes that model radiation damage to DNA at the atomistic level is developed and used to predict the radiation damage to a 167-bp DNA molecule. A database of the ·OH attack outcomes is generated for a 167-bp DNA molecule and used in the prediction of radiation-induced damage to DNA. D0 (the dose required to create, on average, one single strand break per 167-bp DNA molecule) is calculated to be 69.9 Gy.
Summary: ABSTRACT (cont.): There is no experimental study found in the literature that studied small DNA molecules like the one used in this study. Nevertheless, the results from this computational study can be compared to experimental studies preformed with larger DNA molecules such as plasmids when DNA concentrations are scaled. The 'concentration scaled D0 (ssb)' values from Klimczak et al. 1993 and Tomita et al. 1998 were approximately 65 and 80 Gy, respectively. These experimental results compare favorably with the computational value of 69.9 Gy calculated in this study. With the future development and evolution of this modeling scheme, it is hoped that the mechanisms of DNA-radiation interactions and the influence of variations in the microenvironment of the DNA may be better understood.
Thesis: Thesis (Ph. D.)--University of Florida, 2001.
Bibliography: Includes bibliographical references (p. 284-293).
System Details: System requirements: World Wide Web browser and PDF reader.
System Details: Mode of access: World Wide Web.
Statement of Responsibility: by Bulent Aydogan.
General Note: Title from first page of PDF file.
General Note: Document formatted into pages; contains xviii, 294 p.; also contains graphics.
General Note: Vita.
 Record Information
Bibliographic ID: UF00100741
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 50742222
alephbibnum - 002766284
notis - ANP4323

Downloads

This item has the following downloads:

Aydogan ( PDF )


Full Text











A COMPUTATIONAL ATOMISTIC MODEL
OF RADIATION DAMAGE
TO DNA


















By

BULENT AYDOGAN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2001




























Copyright 2001

by

Bulent Aydogan

















































To my mom, for reminding me I have what it takes, and my wife, who has shown

me how to appreciate what life offers.















ACKNOWLEDGMENTS


I am indebted to Dr. Wesley E. Bolch, the chairperson of my PhD committee, for

his guidance and patience throughout my doctorate study. This learning experience could

have not been possible without his constant scientific and personal interactions. I would

like to give a special thanks to Dr. David T. Marshall for being there whenever I needed

help for the last two years. His constant encouragement and insightful comments

throughout this research and preparation of this dissertation made this study complete.

His ability to maintain a broad perspective on the study was extremely beneficial, as was

his clever assistance in solving the problems. Dr. James E. Turner, the great scientist

who inspired me and who made me love radiation physics, provided significant help in

refining the focus of this study, and he continued to be an excellent resource throughout

the course of this research. Dr. Steven G. Swarts guided me with his expertise in

radiation chemistry. I am very thankful for his scholarly advice and willingness to help.

I also appreciate all the help received from Dr. Edward T. Dugan. In addition, I am

grateful to Dr. Dietmar W. Siemann and Dr. Emmett W. Bolch for agreeing to serve on

my supervisory committee and for reading and commenting on my dissertation. Their

comments and suggestions truly made this work complete. I received assistance from

Amy Boone, graduate student at the Chemistry Department, in creating atomistic DNA

models. Without her help, it would have taken me longer to learn and perform this task.









On a personal note, I am very grateful to my parents, who taught me by example

to strive for excellence in my work. Near or far, I am thankful to my parents Anahanim

and Ahmet Aydogan, my sister Fatma Nur, my brother Halit, and close friends for

cheering me on and encouraging me to keep going. Special thanks go to my dear friends

in St. Petersburg whose example and encouragement has kept me healthy since I met

them. I am blessed to have crossed paths with them. I'm thankful for the one special

person, Nouraline, my eternal soul mate, the light of my eyes, who has been closer to my

heart more than anyone else and who has helped me persist during the past two years.

Finally, I give thanks and praise to Almighty God, the source of every good and perfect

gift, for His eternal faithfulness and unmerited blessings.
















TABLE OF CONTENTS

Page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ........................... .. ......... ............ ........ ........ x

LIST O F FIGU RE S .... ............................ ...................... .......... ............. xiv

A B STR A C T .................................................... xvii

CHAPTERS
1 IN T R O D U C T IO N ................. ................................ ........ ........ .. ............ .

D N A .................. .................................................... 2
W atson and C rick M odel .................................................. ................ .............. 3
Hydration Shell ........................................ .............. 8
H history of D N A D am age M odeling................................................................................ 9
Current Atomistic Models of Radiation Damage to DNA....................................... 10
G SF atom istic m o d els .......... .......... .................. .............................. .............. 1 1
University of Tokyo atomistic models.............................. ..... ............ 13
Summary of Past Work and the Significance of This Work ........................................ 14

2 RADIATION CHEMISTRY OF PURE WATER ....................................................18

In tro d u ctio n .................................................................................... 1 8
M materials and M ethods........................................................ ........... ... 20
Physical Transport of Electrons in Liquid Water Using OREC ............................. 20
Free Radical Formation and Radiation Chemistry of Electrons in Liquid Water
U sin g R A D L Y S .......................................................................... ..................... 2 2
Results............................................ .............. 24

3 DEVELOPMENT OF NEAR-APPROACH MODEL AND STRUCTURAL
INFLUENCES OF SITE-SPECIFIC *OH ATTACK TO THE SUGAR MOIETY ....31

In tro du ctio n ................. .... .. ............ ... ........ ..... ...... ...................... 3 1
Previously Developed Biophysical Models of Radiation Damage to DNA............. 31
Previous Models of Indirect Action of Radiation................................. ............... 32
Experimental Data on Site-Specific *OH Attack .............................................. 35
M materials and M ethods................... ............................................. .................... ...... 37









Construction of the DNA Molecular Target Model............................................ 37
Inclusion of the Hydration Layer...... ............................................ .. .............. 39
R action Sites........................................ ............... ..... 39
D eoxyribose sugar m oiety .............. ........................................................... 39
N ucleobases ........................................ 4 1
R action R adii .................................... ............................ 42
Monte Carlo Algorithm for *OH Diffusion and Reaction...................................... 45
Calculations of % SA SA ............................................................................ ........ 51
Determination of optimal model parameters ............................................. .. 53
R esults...................................... ... ... ................ ... ....... 54
Percent Solvent Accessible Surface Area (%SASA).............................................. 54
Sensitivity of %SASA with Variations in Probe Size ......................................... 57
Considerations of Geometrical Optimization via Energy Minimization .............. 57
Sensitivity of *OH Attack Probabilities to the Inclusion of Counter Ions.............. 60
Sensitivity of *OH Attack Probabilities with Variations in Diffusion Jump Time.. 60
Sensitivity of *OH Attack Probabilities with Variations in ROH ............................. 61
Sensitivity of *OH Attack Probabilities with Variations in R.OH + H ....................... 64
Sensitivity of *OH Attack Probabilities with Considering or not Considering
Steric Hindrance .......................... ... .... .... ........... ........... 65
Model Predictions and Comparison to Experimental Data.................................. 67
C onclusions............................... ........... .......... 70

4 STRUCTURAL INFLUENCES OF SITE-SPECIFIC WATER RADICAL
ATTACKS TO THE BASE M OIETY...................................... ........................ 72

Introduction ....................... ......... ............. 72
H ydroxyl and H ydrogen R adicals................................................... ... ................. 73
G uanine ................................................................................... .. ...... ....................... 73
A d e n in e .......................................... ........................................... 7 4
T hy m in e .................................................................................. .............. 74
C y to sin e .............................................................................. 7 5
Hydrated Electron .............. ......... ........ ...................... 75
Material and Methods ............................ ...... ........ 76
*OH and Ho Base Damage Modeling .................................... 78
Hydrated Electron Base Damage Modeling ............................................... 78
R e su lts..................................... .............. ...... 8 0
*OH Attack to Bases...................................... ............ 80
Hydrated Electron Attack to Nucleobases ....................................................... 85
Ho attack to 10-bp DNA Molecule ............................ ......... ................. 87
C o n c lu sio n .............................................................................. 8 7









5 VARIATIONS OF *OH ATTACK TO THE SUGAR AND BASE MOIETIES
WITH CHANGES IN DNA FORM AND STRANDEDNESS.............................. 89

In tro d u ctio n ......................................................... ............... 8 9
M materials and M ethods................... ............................................. .......................... 92
R e su lts ...................... ... ......................................................................................... . ..... 9 4
Effect of the DN A Strandedness....................................................... .......... .... 94
E effect of Interaction Site........................................................................ 95
Effect of DN A Conform ation ........................................................ .............. 96
Conclusion ............ ................................ ............... 103

6 *OH ATTACK DATABASE FOR A 167 BASE-PAIR DNA MOLECULE........... 105

In tro d u ctio n ................................................................................... 10 5
M materials and M ethods............................................. ............................. .............. 106
Construction of the 167 base-pair DNA Molecular Target Model......................... 106
Construction of the Computer Model for the Database Calculation....................... 106
R e su lts ............................................................................................. 1 10
Conclusion ............ ................................ ............... 118

7 MODELING OF THE PHYSICAL STAGE OF DNA DAMAGE AND THE
CALCULATION OF DIRECT AND QUASI-DIRECT EFFECTS...........................123

Intro du action ................................................. ...... ......... ...... 12 3
Materials and Methods.... ....................................... 124
M C N P Sim ulation s .................... .................. ... ..................... .... .................. .. 12 5
Electron Transport with OREC and Energy Deposition in the Target Volumes.... 128
R results ......... ...... ........ ...................................... ........................... 138
Conclusion ............ ................................ ............... 138

8 MODELING OF THE PRECHEMICAL AND EARLY CHEMICAL STAGES
OF DNA DAMAGE AND THE CALCULATION OF INDIRECT EFFECTS ........140

Intro du action ................................................. ...... ......... ...... 14 0
Materials and Methods........................................... 142
Conclusions and Experimental Comparisons ..................................................... 147

9 CONCLUSIONS AND FUTURE WORK ......................... ................. 155

C o n c lu sio n .................................................................................................................. 1 5 5
F utu re W ork ............................ ............. ......................................... 160

APPENDICES
A OREC INPU T AND OU TPU T....................................................... ............... 164

B RADLYS INPUT AND OUTPUT....................... ... ..................186









C WATER RADICAL ATTACK CODES........................................... ....................190

D OH RADICAL ATTACK DATABASE OF 167-BP DNA MOLECULE................226

E A D D IT IO N A L C O D E S ...................................................................... ..................275

L IST O F R EFER EN CE S ........................................................................... ..............284

BIOGRAPH ICAL SKETCH ............................ ......... ..................... 294
















LIST OF TABLES


Table Page

2.1. CSDA ranges for energies ranging between 500 eV and 1 MeV as calculated with
OREC before and after corrections were made. ................... .......................... 26

2.2. Additional changes proposed to improve pure water calibration. ................................27

2.3. Revised partitioning coefficients of excitation in RADLYS ......................................28

3.1. The positions of each tightly bound water molecule within the first hydration layer. .40

3.2. *OH reaction rate constants and reaction radii for with deoxyribose and the four. ......43

3.3. van der W aals radii. ..................................................... .. .. ....... ............... 44

3.4. Diffusion coefficients and jump distances for the OH............................................ 47

3.5. Sensitivity of %*OH attack probabilities with variations in diffusion jump time........62

4.1. Reaction rate constants and reaction radii ........................................ ............... 77

4.2. Diffusion constants of the water radicals............................ ...... ... .............. 78

4.3. Preferential electrons adduct sites in four nucleobases......................................79

4.4. Distribution of *OH attack among the four nucleobases.............................................83

4.6. Comparison of %*OH attack as calculated for each nucleobases individually ...........84

4.7. Percent hydrated electron attacks........................................................ ............... 85

4.8. Site specific hydrated electron attacks in nucleobases.. ............................................. 86

4.9. Percent hydrated electron contribution of for nucleobases in 10-bp DNA molecule ....88

5.1. Distribution of *OH attack between the base and sugar moieties. ............................95

5.2. Comparison of the %*OH attack with changes in DNA strandedness ..........................96

5.3. Variations of *OH attack with changes in DNA form.................................................96









5.4. Comparison of %SASA and %*OH attack probabilities ...........................................101

6.1 Database section 1. 3rd through 8th base-pair.. ................ ........ ............................ 120

6.2 Database section 2. 3rd through 11th base-pair.. .................................. ............... 121

6.3 Database section 3. 5th through 14th base-pair.. .................................... ...............122

7.1. Energy required to deposit a dose varying between 2 and 12.4 Gy in the target cube..132

8.1. An example output of damage analysis program (DAP). ...........................................146

8.2. Average survival fractions for ssb and dsb. .........................................................148

D.1. Database section 1. 3rd through 8th base-pair.. .............................................................227

D .2. D database section 2. 3rd through 11th base-pair.. ........................................ ................228

D.3. Database section 3. 5th through 14th base-pair. ................................. .................229

D .4. D database section 4. 11th through 17th base-pair ........................................ ............... 230

D .5 D database section 5. 5th through 14th base-pair. ............................ ........................... 231

D.6 Database section 6. 17th through 24th base-pair. ...................................................232

D.7. Database section 7. 19th through 28th base-pair. ........................................ ............... 233

D.8. Database section 8. 23rd through 33rd base-pair. ............................... .....................234

D.9. Database section 9. 25th through 35th base-pair. ........................................ .............. 235

D .10. D database section 10. 30th through 39th base-pair.. ....................................................236

D.11. Database section 11. 34h through 42nd base-pair ............... ................................. 237

D .12. D database section 12. 38h through 46th base-pair.. ........................................................238

D.13. Database section 13. 40h through 50th base-pair.. ........................................ ....... 239

D.14. Database section 14. 44h through 54th base-pair. ........................................ ......... 240

D.15. Database section 15. 48h through 56th base-pair. .............................................. 241

D.16. Database section 16. 34h through 42nd base-pair............... ............. ............... 242

D.17. Database section 17. 55h through 64th base-pair.............. ............. ............... 243

D.18. Database section 18. 59th through 67th base-pair. ......................... ............... 244









D.19.

D.20.

D.21.

D.22.

D.23.

D.24.

D.25.

D.26.

D.27.

D.28.

D.29.

D.30.

D.31.

D.32.

D.33.

D.34.

D.35.

D.36.

D.37.

D.38.

D.39.

D.40.

D.41.

D.42.

D.43.


Database section 19.

Database section 20.

Database section 21.

Database section 22.

Database section 23.

Database section 24.

Database section 25.

Database section 26.

Database section 27.

Database section 28.

Database section 29.

Database section 30.

Database section 31.

Database section 32.

Database section 33.

Database section 34.

Database section 35.

Database section 36.

Database section 37.

Database section 38.

Database section 39.

Database section 40.

Database section 41.

Database section 42.

Database section 43.


63h through 71st base-pair. ....................................................245

65h through 74th base-pair.. ........................... .....................246

69h through 77th base-pair.. ........................... .....................247

74h through 81st base-pair. ....................................................248

76h through 86th base-pair. ...................................... ........... 249

80h through 88th base-pair. ................................................. 250

83rd through 82nd base-pair............... ............. .....................251

86th through 96th base-pair.. ............................. ................252

91st through 98th base-pair. ............... ... ........................... 253

96th through 102nd base-pair. ........................... ..................254

97th through 106th base-pair................ .... ............... 255

101st through 109th base-pair. .............................................256

106th through 113th base-pair.................... ...............257

108th through 117th base-pair...................... ........ .......258

112th through 119th base-pair ............. ...........................259

116th through 124th base-pair.......................... ..... ..........260

119th through 128th base-pair.................. ...............261

122nd through 130th base-pair. ............................ .................262

127th through 134th base-pair...................... ...............263

129th through 138th base-pair...................... ...............264

133rd through 140th base-pair.............. .......... ................ 265

137th through 144th base-pair. ..........................................266

140th through 148th base-pair ............... ...... ...............267

145th through 151st base-pair. ......................................268

148th through 156th base-pair. ..........................................269









D.44. Database section 44. 151st through 160th base-pair. .............................................270

D.45. Database section 45. 154th through 161st base-pair. ............................................271

D.46. Database section 46. 159th through 166th base-pair. ............................................272

D.47. Database section 47. 161st through 169th base-pair. ............................................273

D.48. Database section 48. 164h through 169th base-pair. ............................................274
















LIST OF FIGURES


Figure Page

1.1. Nucleotides of DNA: Adenine, Guanine, Thymine, and Cytosine.............................4

1.2 D eoxyribose Sugar............................... .......... ... ........ ................. .5

1.3. DN A base-pair bonding: A T; G C ........................................ ....................... 6

1.4 W atson and C rick M odel .................................................................... .....................7

2.1. Time dependence of G(*OH) as generated by RADLYS......................................29

2.2. Time dependence of G(e-aq) as generated by RADLYS ........................................ 30

3.1. Stick model representation of the 10-bp DNA molecule ...........................................38

3.2. Schematic geometry of the near approach model .....................................................46

3.3. A schematic representation of *OH diffusion and near-approach interaction
with a DNA molecule in its geometrically optimized configuration. ...................49

3.4. A magnified representation of this same radical as it approaches the sugar moiety ....50

3.5. Illustration of the SASA and the van der Waals surface ............................................52

3.6. Comparison of %SASA for the DNA structures considered in the study ...................55

3.7. Variations in site-specific %SASA with changes in the spherical probe radius............56

3.8. Variations in site-specific %SASA and %*OH attack probabilities following
geometrical optimization via energy minimization.............. ................................. 58

3.9. Effect of the presence ofNa+ counter ions.in the %*OH ...........................................59

3.10. Hydroxyl attack probabilities calculated with and without steric hindrance ..............66

3.11. Comparison of the %*OH calculated in this study to DNA cleavage rates from
B alasubram anian et al ...................... ................ ................... ......... 68

5.1. Ball and stick representations of A-, B-, Z-DNA, and ssDNA................................93









5.2. The percent *OH attack to the sugar and base moieties when changing the site of
attack from hydrogens to carbons in sugar moiety ..................................................98

5.3. Relative *OH attack distribution between the sugar and base moieties with and
without steric hindrance in ds B-DN A .............. .................................. ............... 102

6.1 171 base-pair DNA molecule with the Na ions................................. ............... 107

6.2. Schematic representation of the database geometry for thel67 -bp DNA molecule.....108

6.3. Cylindrical volumes representing each of the 48 database sections.............................109

6.4 % *OH attack at each of the 48 database sections........................................................112

6.5. Distribution of the *OH attack in 1st through 40th base-pairs ..................................... 114

6.6. Distribution of the *OH attack attack in 41st through 80th base-pairs.......................... 115

6.7. Distribution of the *OH attack in 81st through 120th base-pairs ..................................116

6.8. Distribution of the hydroxyl attack in 121st through 167th base-pairs ...........................117

6.9. Shcematic representation of how to read the database tables............... .................. 119

7.1. Schematic representation of the irradiation vial used in the companion irradiation
experim ent ............................................................................ ....... ....... 125

7.2. Experimental design for irradiation of DNA sample with a clinical Co-60 unit...........126

7.3. Photon Spectrum of a clinical Co-60 ................................................. ...... ......... 127

7.4. Electrons spectrum of from a clinical Co-60 machine inside the irradiated vial...........30

7.5. Pictorial illustration of the target cube with a 167-bp DNA............... ................ ...131

7.6. Geometry of the energy deposition model introduced into OREC (not scaled)............133

7.7. Energy distribution in the ten slabs through the cubical target volume.........................135

7.8. Average energy deposited and their probability of occurrence among the one
thousand target volumes for the 4 Gy dose.................................... ............... 136

7.9. Two pictorial representations of the cubical target volume with a 167-bp.................. 137

8.1. Schem atic representation of the *OH interaction ......................................................... 144

8.2. Ssb survival fraction as a function of Dose (Gy). ................................................149

8.3. D sb survival fraction as a function of Dose (Gy).. ....................................................... 150









8.4. Calculated dose-effect relationship for ssb as reproduced from the plot provided by
Tom ita et al. [1998] ............. .. ................ ................ .. ............. 153















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

A COMPUTATATIONAL ATOMISTIC MODEL
OF RADIATION DAMAGE TO DNA

By

Bulent Aydogan

August 2001


Chairman: Dr. Wesley E. Bolch
Major Department: Department of Nuclear and Radiological Engineering

A review of past and current biophysical models of DNA damage reveals that

current DNA damage models have become increasingly complex in their attempts to

model the full 3D structure of the nucleosome and chromatin fiber. As such, many of the

finer details of direct, quasi-direct, and indirect action on DNA become difficult to study

in isolation. Also, experimental comparisons that seek to validate these models become

increasingly difficult to make.

A better approach may be to perform the atomistic modeling of direct, indirect,

and quasi-direct effects in total isolation from considerations of the macroscopic

conformation of the DNA target. This would permit the highly detailed atomistic

modeling to be performed only once in order to produce a database of outcome

probabilities that can then be used in radiation chemistry modeling of different and more

complex conformations of double-stranded DNA. This work is performed to establish









the groundwork to accomplish this goal. A system of Monte Carlo computer codes that

model radiation damage to DNA at the atomistic level is developed and used to predict

the radiation damage to a 167-bp DNA molecule. A database of the *OH attack

outcomes is generated for a 167-bp DNA molecule and used in the prediction of

radiation-induced damage to DNA. Do (the dose required to create, on average, one

single strand break per 167-bp DNA molecule) is calculated to be 69.9 Gy. There is no

experimental study found in the literature that studied small DNA molecules like the one

used in this study. Nevertheless, the results from this computational study can be

compared to experimental studies preformed with larger DNA molecules such as

plasmids when DNA concentrations are scaled. The 'concentration scaled Do (ssb)'

values from Klimczak et al. [1993] and Tomita et al. [1998] were approximately 65 and

80 Gy, respectively. These experimental results compare favorably with the

computational value of 69.9 Gy calculated in this study. With the future development

and evolution of this modeling scheme, it is hoped that the mechanisms of DNA-radiation

interactions and the influence of variations in the microenvironment of the DNA may be

better understood.


xviii














CHAPTER 1
INTRODUCTION


One of the most confounding issues in the treatment of cancer is why two patients

with the same diagnosis respond differently to the same treatment. One reason this may

occur is that the environments in which their tumors reside may be different. For

example, oxygen gradients may exist across a tumor volume. Conventional radiotherapy

treatment planning, which is based on delivering a maximally uniform absorbed dose (as

described by rad or Gy) to a tumor volume, does not take into account variations in the

tumor microenvironment (TM) such as hypoxia. However, absorbed dose only describes

energy deposition per unit mass of tissue. Variations in the TM induce corresponding

variations in radiosensitivity. Factors in the TM that can alter biological response to

radiotherapy include oxygen levels, pH, free radical scavengers, and chemical modifiers

(radiation sensitizers/protectants). An improved paradigm for radiotherapy treatment

planning would be to deliver the radiation required to produce a uniform radiobiological

response throughout the volume of interest. To achieve this paradigm of basing radiation

therapy treatment planning on uniform response rather than absorbed dose, two

objectives must be met: (1) there must be a way to measure variations of the factor of

interest (oxygen concentration, pH, chemical modifier concentration) throughout the

tumor volume, and (2) a technique is needed to quantify the impact these variations have

on biological response to radiotherapy. Many researchers are working on a variety of

techniques to quantify the tumor microenvironment using both invasive (e.g., Eppendorf









electrodes, fiberoptic probes) and noninvasive techniques (e.g., Magnetic Resonance

Imaging, Magnetic Resonance Spectroscopy and radio- labeled compounds) vivo

[Aboagye et al. 1998; Hunjan et al. 1998; Luyten et al. 1990; Mason et al. 1994; Mason

et al. 1998; Yang et al. 1999]. Currently, the second requirement quantifying the

radiobiological effect of variations in the tumor microenvironment is not well addressed.

Understanding the basic mechanism of radiation damage to biological systems has

long been of interest. When a cell is subjected to ionizing radiation, many chemical

reactions are induced, eventually leading to a variety of biologically significant end

points. Depending on the type of damage studied, there will be more than one target at

the cellular level. In the mid-1960s, microdosimetry studies provided evidence that there

are very little lethal damages as long as the radiation is absorbed outside the nucleus in

eukaryotic cells, which contain their DNA mostly within the nucleus. A drastic decrease

in the cell survival curves is observed as soon as the radiation reaches the nucleus or

DNA [von Sonntag 1987]. Since DNA is recognized as a critical target for radiation

induced stochastic effects, the development of computational models to quantify

radiation induced DNA damage has been a challenging task for many researchers. A

brief discussion on DNA structure might be helpful before we further articulate the DNA

damage mechanisms and computational models in detail.



DNA

DNA is recognized as the most important biological structure since it serves as

the "Master Copy" for genetic information. DNA replicates itself before the cell divides,

ensuring the genetic information in the next cell generation is identical. DNA also









maintains all information needed for protein synthesis. DNA is defined as a long double-

stranded polymer, where the nucleotides are the main building blocks of its structure.

Each nucleotide consists of three tightly joined components: a nitrogenous base

known as a purine or pyrimidine, a deoxyribose sugar and a phosphate group. Figures

1.1 and 1.2 show the atomic models of nucleotides of DNA and the deoxyribose,

respectively. The number notation for each atom is also shown. Considering all the

components of DNA (sugar, phosphate, and bases), DNA molecules are composed of

phosphorous, oxygen, hydrogen, nitrogen, and carbon atoms. The purines are adenine

(A) and guanine (G) and the pyrimidines are cytosine (C) and thymine (T). The bases can

only hydrogen bond in a very specific pairing scheme: A always bonds to T and G always

bonds to C (see Figure 1.3).


Watson and Crick Model

DNA was first identified in 1868 by a Swiss biologist, Friedrich Meischer. He

discovered DNA in the nuclei of pus cells obtained from discarded surgical bandages and

called the substance nuclein. In 1953, Watson and Crick introduced a model of DNA that

exists as a regular two-chain structure with H-bonds formed between opposing bases on

the two chains as shown in Figure 1.4.

This model illustrates that the hydrogen bonding is possible only when the directional

senses of the two interacting chains are antiparallel. This structure can adopt two

different types of conformations. One possibility is a stepladder-like structure in which

the chains lie straight in a fully extended distance of 0.68 nm between residues in the

direction of the long axis. However, this was not an energetically favorable structure so

Watson and Crick converted the stepladder to a helix structure by a simple right-hand
















Pyrimidines


Thymine


Cytosine


Purines


Adenine Guanine

Figure 1.1. Nucleotides of DNA: Adenine, Guanine, Thymine, and Cytosine











































Figure 1.2 Deoxyribose Sugar
































Guanine-Cytosine




















Adenine-Thymine


Figure 1.3. DNA base-pair bonding: A T; G C

































Base-pair
spacing 6.8 A


(a) Stepladder


Figure 1.4. Watson and Crick Model


Base-pair
spacing 3 4 A


1 1 1 !


i/

,.;:


: :.. 1


(b) Hel x


L


p.









twist forming a -2 nm diameter double helix. A distance of 0.34 nm between the base-

pairs allows a close contact between adjacent base-pairs. This structure repeats itself

after 10 residues or once every 3.4 nm along the long axis. Watson and Crick were also

the first to appreciate the significance of strong 0.34 nm and 3.4 nm spacings and the

central cross-like pattern that reflects a helix structure in the X-ray diffraction pattern of

DNA. As indicated by Watson and Crick, X-ray diffraction patterns suggested that B-

DNA is the native form of DNA. The "ideal" B-DNA helix has 10 base-pairs (bp) per

turn. Each base-pair is stacked with a helical rotation of 36 degrees.


Hydration Shell

The distribution of water molecules around the four bases, phosphate groups and

sugar moieties are distinct and known to play an important role in the three dimensional

structure of DNA [Goodfellow et al. 1994; Nikjoo et al. 1994]. The conformation that

DNA adopts is primarily determined by this solvent structure. For instance, at high

relative humidity or water activity, there are enough water molecules to hydrate all

phosphate group atoms individually. This causes the DNA to adopt the B-form. In

contrast, DNA fibers transform from B to A form in the presence of low relative

humidity. The primary hydration shell consists of 20-22 water molecules per nucleotide,

12-17 of which are bonded directly to the DNA1 [Umrania et al. 1995]. These water

molecules bind in decreasing order of strength to the anionic oxygens of the phosphate

group, to the ester oxygens of the phospodiester linkage, and to the electronegative atoms

of the base-pairs. The hydration water molecules are proven to be distinct from normal

water by their different infrared spectral characteristic [Wolff et al. 1980]. Non-mobile


1 Personal communication with Dr. S.G. Swarts, January 2000.









water molecules act as part of the atomistic structure and thus are target for direct action

by incident radiation. Consequently, the structure of the water hydration shell may

contribute significantly to the formation and reaction of the radiation-induced radicals in

and around DNA by altering their absolute number and spatial distribution at times less

than 10-12 S.



History of DNA Damage Modeling

Understanding the biochemical mechanisms associated with the production of

DNA strand breaks by ionizing radiation has been of great interest to the scientific world.

Until the late 1980s most of our knowledge on radiation-induced biological damage was

from experimental studies [Holley and Chatterjee 1994]. One of the recent challenges of

radiation research is to develop computational tools to attain better quantitative

understanding about how ionizing radiation induces damage in critically important

biological systems. Computational modeling became an important tool in providing

means to explore molecular events occurring during the time period of radiation

interactions and the biomolecular damages are subsequently produced. Furthermore, this

identifies where and how energy is imparted and how these changes lead to biological

effects.

Various biophysical models have been proposed to describe the action of damage

from ionizing radiation to DNA. Some of the earlier models analyzed radiation action on

DNA in terms of energy deposition and/or the number of ionizations (direct effect) within

cylindrical volumes on nanometer dimensions representing the biological targets

including a DNA segment, a nucleosome, a chromatin fiber [Charlton and Humm 1988;









Charlton et al. 1989; Goodhead and Nikjoo 1989; Hamm et al. 1989; Nikjoo et al. 1994].

To quantify the production of ssb and dsb, Charlton et al. constructed a DNA double

helix as an inner cylinder representing the bases around which a sequence of stacked

rotated arches was overlaid representing the two sugar-phosphate strands [Charlton and

Humm 1988]. The size of an energy deposition within the target volume was used to

decide which of various possible DNA lesions was produced. Using

an empirical threshold energy of 17.5 eV, the calculated number of ssb was found to be in

good agreement with experimental results in cell nuclei [Charlton et al. 1989]. These

modeling efforts, however, provided only crude estimates of radiation-induced DNA

damage occurring in the early stages of radiation interactions with DNA. To extend the

modeling to the chemical stage of interaction, one needs to know the spatial distribution

of radiolysis products relative to DNA nucleotide [Terrisol 1994]. Other attempts have

been made to expand the early computational models including quasi-direct action to the

DNA hydration layer [Henss and Paretzke 1992; Nikjoo and Charlton 1995; Ottolenghi et

al. 1995] and indirect effects from *OH formed in the surrounding water [Nikjoo et al.

1994; Nikjoo et al. 1997; Terrisol 1994; Zaider et al. 1994].


Current Atomistic Models of Radiation Damage to DNA

More recent biophysical models of radiation action on DNA have focused on both

the atomic structure of the DNA molecule as well as higher-order structures of DNA

conformation. These atomistic models are based on simple double-helical structures

found in plasmids [Bardash and Zaider 1994; Chatterjee and Holley 1990; Pomplun and

Terrissol 1994] and nucleosomes [Michalik and Begusova 1994; Pomplun and Terrissol

1994]. Chromatin fibers have been introduced in a "beads-on-a-string" form [Chatterjee









and Holley 1990], a solenoid form [Chatterjee et al. 1994], a crossed-linker form

[Friedland et al. 1996], and a stochastic configuration [Jacob et al. 1997; Woodcock

1994]. In addition, looped supercoiled structures and hinged minisolenoids have also

been considered [Chepel et al. 1994].


GSF atomistic models

Potentially the most complex atomistic model of DNA damage produced to date

is that by Friedland and colleagues at GSF-National Research Center for Environment

and Health, Neuherberg, Germany [Friedland et al. 1998]. In their model, the DNA

target molecule is constructed at five levels of organization: nucleotide pairs, double

helix, nucleosome, chromatin fiber, and chromatin fiber loop [Friedland et al. 1998].

Deoxynucleotide pairs are stacked in either a preselected or random sequence within the

B-DNA form [Chandrasekaran and Arnott 1989]. Each atom in the structure is modeled

as a sphere with radii taken as the van der Waals radius for use in scoring direct

ionizations [Saenger 1984]. The hydration shell is modeled by increasing each atom

radius by a factor of two; consequently, the individual water molecules of the hydration

layer are not explicitly modeled. The simulation of the nucleosome core protein is based

on the model of Pomplun and Terrisol [Pomplun and Terrissol 1994]. Four chromatin

fiber structures are then considered: zigzag, solenoid, cross-linker, and stochastic.

Finally, each fiber structure is used to form a looped segment of chromatin ranging from

17,716 base-pairs using the zigzag structure to 89,688 base-pairs using the stochastic

structure.

In their 1998 study, the new DNA target model was incorporated within the

PARTRAC photon-electron transport code also developed at GSF. The chromatin fiber









loops were randomly placed within a 5-jPm spherical nucleus located within a 10-apm

spherical cell. The cell was then assumed to be irradiated isotropically by monoenergetic

electrons and a 220 kVp filtered x-ray spectrum. Tracks of secondary electrons were

started randomly throughout a larger volume of water and the overlap of energy

deposition events with the DNA structure of the loops was tabulated. Ssb and dsb were

recorded as a function of two assumed parameters: the energy threshold for ssb

production and the base-pair separation of two ssb required to form a dsb. A parameter

combination of 10.5 eV and 4 base-pairs led to the smallest deviation from experimental

data on human fibroblasts [Lobrich et al. 1996].

In 1999, the GSF group expanded their DNA model by adjoining many non-

overlapping chromatin fiber loops in order to simulate an entire DNA chromosome

[Friedland et al. 1999, 1999]. DNA damage via quasi-direct action was added to the

modeling scheme such that events inside the water shell were classified into three

categories: (1) 60% of events involving water associated with the phosphate groups are

scored as a ssb with 40% treated in the radiation chemistry module; (2) 100% of events to

water associated with the sugar moiety were treated in the radiation chemistry module;

and (3) 100% of events involving water associated with the base moiety were scored as

base damage. A radiation chemistry module was thus incorporated in which reaction

radii were calculated from reaction rate constants given in Buxton et al. Free radical

creation, diffusion, and chemical reactions are based upon the models put forward by

Turner et al. in their codes OREC and RADLYS [Hamm et al. 1989; Turner et al. 1988].

The steric hindrance by non-reacting atoms are not accounted for in the radiation









chemistry portions of their simulations2. Although detailed distributions of DNA

fragment patterns are obtained through model simulations, experimental comparisons are

few and in many cases nonexistent against which to verify their modeling approach. At

the 6h International Workshop on Radiation Damage to DNA (Chapel Hill, NC, April

1999), comparisons presented by Friedland against cellular data from other laboratories

were inconclusive [Friedland et al. 1999].


University of Tokyo atomistic models

Another recent atomistic model of radiation damage to DNA, and one that more

closely approaches the present proposed research, is that of Tomita et al. for the pBR322

plasmid [Tomita et al. 1998]. In their study, the equilibrium structure of the plasmid

DNA was established by minimization of strand elastic energy on structures generated

through the algorithm of Vologodskii et al. [Vologodskii et al. 1992]. In this model, the

4,362 base-pair structure of pBR322 was assembled as a polygon of 148 cylindrical

segments. The 3D atomic coordinates of B-DNA using the base-pair sequence of

pBR322 was fit spirally within the various cylindrical segments. Track structure

simulations were superimposed upon the plasmid model simulating 60Co gamma-ray

irradiation of the solution with a DNA concentration of -30 Pg/cm3. The solution was

further assumed to be fully aerated and to include 1 mM Tris, 5 mM NaC1, and 0.1 mM

EDTA to simulate physiologic conditions of pH and *OH scavenging capacity. The free

radical chemistry of the system has been summarized in their previous work [Tomita et

al. 1995]. In contrast to the Friedland et al. model, the Tomita et al. chemistry model

includes specific atomic sites of free radical attacks based upon those summarized by von


2 Personal communication between Dr. J.E. Turner and Dr. W. Friedland









Sonntag [von Sonntag 1991]. For example, *OH adduct formation is assumed to occur at

C(5) and C(6) of thymine with H-abstraction at the CH3 group. Relative probabilities of

single atom attack are given, but details as to how these probabilities are computationally

implemented within the simulation model are not discussed. Also in contrast to the GSF

model is the lack of detailed simulations of free radical diffusion and intratrack chemistry

prior to radical approach. In Tomita et al., a time-integrated probability of interaction

with DNA is used based upon the initial displacement of each free radical in the

simulated track from each reacting atom of the DNA structure [Tomita et al. 1995]. In

their 1998 study, Tomita et al. compared their simulated yields of ssb and dsb to data

obtained in their laboratory for 60Co gamma-ray irradiations of pBR322. Best agreement

between the measured and simulated data was found assuming (1) an *OH attack

efficiency of 50% for producing ssb and (2) a separation of ssb by at most 10 base-pairs

is required to form a dsb. No differentiation between sugar damage and base damage was

reported.



Summary of Past Work and the Significance of This Work

A review of past and current biophysical models of DNA damage reveals that the

current DNA damage models have become increasingly complex in their attempts to

model the full 3D structure of the nucleosome and chromatin fiber. As such, many of the

finer details of direct, quasi-direct, and indirect action on DNA have become difficult to

study in isolation. Also, experimental comparisons that seek to validate these models

have become increasingly difficult to make.









A better approach might be to perform the atomistic modeling of direct, indirect,

and quasi-direct effects in total isolation from considerations of the macroscopic

conformation of the DNA target. This would permit the highly detailed atomistic

modeling to be performed only once in order to produce a database of outcome

probabilities that can then be used in radiation chemistry modeling of different and more

complex conformations of double-stranded DNA. This work is performed to establish

the groundwork to accomplish this goal. A system of Monte Carlo computer codes that

model radiation damage to DNA at the atomistic level is developed and used to predict

the radiation damage to a 167-bp DNA molecule. This Ph.D. research project is

conducted under the following tasks:


1. To reassess OREC/RADLYS fits to current pulse radiolysis data on G(OH) and G(e-

aq). The computer codes OREC and RADLYS have been used successfully to

simulate the irradiation of both pure water systems and aqueous solutions of a

biological target molecule such as glycylglycine [Bolch et al. 1990; Turner et al.

1991]. In this task, these two codes are employed to fit the experimental data on time

dependence of G(OH) and G(e-aq) obtained by Pimblott & LaVerne (1991) which

were calculated by inverse Laplace transform of radical scavenger data [Pimblott and

LaVerne 1990, 1992]

2. To develop a near-approach radiation chemistry model. This model simulates, at an

atomistic level, the close proximity radiation chemistry of radiolytic free radicals in

their reactions with the sugars, bases and the tightly bound water molecules of the

first hydration layer.









3. To construct a geometrical DNA model, at the atomistic level, to be used with the

computational scheme developed in Task 2 to predict DNA damage.

4. To perform model simulations to construct a database of outcome probabilities for

indirect radical attack for all damage types as needed in the computational model of

the 167-bp DNA molecule.

5. To define the radiation field that is "seen" by the DNA target molecule after the

irradiation with a clinical Co-60 machine. The MCNP code is used to determine the

initial energy distribution of photoelectrons and Compton electrons created within the

solution.

6. To construct an atomistic computational radiation chemistry model of a 167-bp

double-stranded DNA molecule. An outcome database of *OH constructed in Task 4

is used in the determination of subsequent DNA damage calculations. The

computations include (1) detailed simulations of secondary electron production and

energy loss within the aqueous medium, (2) the production, reaction, and chemical

diffusion of radiolytic free radicals, (3) sites of initial direct, indirect, and quasi-direct

attacks single-strand breaks (ssb), double-strand breaks (dsb) scoring.



Chapter 2 presents the basics of the radiolysis of water and reassessment of

OREC/RADLYS code to fit the experimental data on time dependence of G(OH) and

G(e-aq) obtained by Pimblott and LaVerne (1991).

The next goal of this work is to develop a near-approach radiation chemistry

model to simulate the radiation chemistry of water radicals, at an atomistic level, in their

reactions with the sugars, bases and the first hydration layer. Chapter 3 explains the









important aspects of the near-approach chemistry model developed to simulate the

radiation chemistry of *OH s with DNA. A comparison of the model results of OOH

attack probabilities to the individual reaction sites with the experimental is concluded in

Chapter 3.

In Chapter 4, the near approach model is modified to simulate the interaction of

other water radicals including hydrated electrons and hydrogens.

Chapter 5 deals with the variations of radical attack to the sugar and base moieties

with changes in DNA form, the choices of the reaction sites in DNA, and the

strandedness.

A database of hydroxyl interaction for the same 167-base pair DNA molecule

irradiated in the companion UF experiment is constructed and presented in Chapter 6.

The physical stage of DNA damage modeling, direct and quasi-direct action of

radiation is studied in Chapter 7 while Chapter 8 discusses the modeling of the

prechemical and early chemical stages of DNA damage. The comparison of the results

from this study with the available experimental data is also is presented in Chapter 8.

Chapter 9 presents the conclusions and future work.














CHAPTER 2
RADIATION CHEMISTRY OF PURE WATER



Introduction

Due to the large proportion of the water in biological organisms, water became

one of the earliest molecules that were studied for radiation-induced damage in biological

systems. It is essential to first understand the mechanisms of water radiolysis before

further discussing radiation damage mechanisms in DNA. The study of radiation

chemistry requires a determination of what species are formed by radiation, how different

chemical systems alter the evolution of the primary species and what various chemical

reactions can occur [Jonah et al. 1976; Jonah and Miller 1977]. The interaction of

ionizing radiation with water precipitates physical and chemical changes. These changes

can be described in four stages: physical, prechemical, early chemical, and late chemical

stages [Bolch et al. 1998]

In the physical stage, the primary charged particles transfer their energy through

elastic and inelastic collisions. Inelastic collisions of charged particles with water

molecule result in ionization or excitation, leaving behind ionized (H20+) and excited

(H20*) water molecules, and unbound subexcitation electrons (e-sub). Some electrons

that are produced in ionizations may have sufficient energy to cause additional electronic

transitions. These electrons lose energy through inelastic collisions eventually degrading

in energy below the threshold energy necessary for further electron excitations of water









molecule. All these events occur within -10-15 s after the initial passage of the primary

particle.

In the prechemical stage, from 1015 s to about -10-12 s, some rearrangements and

initial conversions of the products produced in the previous stage occur. No significant

diffusion or chemical reaction takes place, since the time elapsed is not large enough for

these species to react [Bolch et al. 1988]. It is believed that H20 first diffuses some

fraction of a nanometer and reacts with a neighboring water molecule by about ~10-12 s in

the following manner:



H20 +H20 H30 +20H Eq. 2.1.



The conversion of excited water molecules to chemical species is rather complex,

depending upon the six different presumed excitation states [Bolch et al. 1988]. The

excited water molecule is subject to either dissociation reactions or a delayed ionization

reaction:

Dissociation:

H20* H + OH Eq. 2.2.



H20* + H20 H2 + 20H Eq. 2.3.



Autoionization:


H20* H20+ + eub


Eq. 2.4.









Subexcitation electrons loose their kinetic energies to the surrounding medium

through both vibrational and rotational excitations, eventually reaching thermal

equilibrium within the water medium. These electrons are then called "hydrated or

aqueous electrons, e-aq, since a hydration layer of five to seven water molecules form a

hydration layer around these low energy electrons. At the end of the prechemical stage,

at -10-12 s, there are 5 species left in a charged particle track: *OH, H, e-aq, H30 and H2.

These chemical species then diffuse and react with each other and with other reactive

molecules by about -10-6 s in a single electron track. After this time, at low dose rate,

further intertrack reactions between any remaining species are unlikely. From this time

onward, the water radiolysis phase is referred as the late chemistry stage and the

remaining reactive species are considered to be homogenously distributed. Consideration

of this bulk phase chemistry thus allows the calculation of further product yields in water

radiolysis [Bolch et al. 1998].



Materials and Methods


Physical Transport of Electrons in Liquid Water Using OREC

OREC is a Monte Carlo electron transport code which follows electrons through

liquid water on an event-by-event basis. In the computational algorithm employed,

detailed cross sections are first sampled to determine both the distance of travel for a

primary electron (its "flight distance") as well as the type of interaction it will undergo.

A total of 11 different individual energy loss events are possible which include six

excitation states (H20*) and five ionization states (H20+) of water molecule.

Alternatively, the primary electron may undergo elastic scattering from oxygen or









hydrogen atoms in water molecule. If the first interaction is an excitation the OREC

selects the scattering angle of the primary electron and a residual energy transferred to the

excited water molecule. This excitation energy is subsequently subtracted from the

kinetic energy of the primary electron and both a new flight distance and a new

interaction type are selected. If an ionization occurs, both a scattering angle and energy

loss of the primary electron are determined. The energy loss is further subdivided

through additional cross sectional sampling into the residual energy of the ionized water

molecule and the initial kinetic energy of the secondary electron. Finally, an initial

trajectory of the secondary electron is selected within the 3D space based upon

predetermined relative cross sectional probabilities. At this point, the code stores the

energy, 3D location, and flight direction of the primary electron, while it continues the

transport and energy loss history of the newly created secondary electron. Once the

secondary electron or any tertiary or quaternary electrons that the secondary electron may

have created, become all fully depleted in their residual kinetic energy in further

ionizations and excitations of the water medium, the primary electron is then

reestablished as the current electron to be transported. In this manner, the 3D branching

of the real electron track is fully and explicitly simulated. This technique of simulation is

in stark contrast to other electron simulation codes such as EGS4 or ITS which employ

approximations to energy loss (e.g., the continuous slowing down approximation or

CSDA) and do not explicitly simulate all energy loss events occurring during particle

transport. In our transport algorithm, the primary electron, and all its secondaries, are

followed down to subexcitation energies (< 7.4 eV). A history file is subsequently

generated indicating the (x,y,z) coordinates of each H20*, H20 and the subexcitation









electron that are created, as well as the energy loss associated with each species. The

physics of the electron transport is thus completed at a relative time of 10-15 s.


Free Radical Formation and Radiation Chemistry of Electrons in Liquid Water
Using RADLYS

RADLYS reads the OREC history file and converts all prechemical species to an

initial spatial distribution of chemical reactants and free radicals. Three conversions are

simulated. Excited water molecules (H20*) are allowed to dissociate into hydroxyl

(*OH) and hydrogen (H*) radicals. They are placed at a separation distance of one water

molecule and in a random orientation centered at the original site of the excited water

molecule. Ionized water molecules (H20 ) are allowed to migrate in a random direction

about 0.75 nm prior to reacting with a neighboring water molecule through a hydrogen

abstraction reaction which results in the formation of a hydronium ion (H30 ) and a OH

radical (*OH). Finally, subexcitation electrons are allowed to thermalize within the

liquid water medium through a distance which is dependent upon their residual kinetic

energy. In this process, they reach thermal energies through rotational and vibrational

excitations of water molecules eventually forming a hydration layer. At this stage they

are referred to as hydrated or aqueous electrons (e- ). The prechemical stage of electron

track development is thus completed at a relative time of 10-12s.

In the final stage of the computations, these various chemical reactants are

allowed to undergo intratrack chemical reactions and thermal diffusions, all simulated via

Monte Carlo techniques. At one picosecond, the RADLYS code checks each pair of

reactant against a predetermined list of potential reactions and reaction distances. These

reaction distances represent the maximum spatial separation necessary for a given









chemical reaction to be allowed. They are based upon the combined diffusion

coefficients of the two reactants and the reaction rate constants. For example, if two

hydroxyl are found to be within a distance of 0.6 nm from one another, they are allowed

to interact and form a single molecule of hydrogen peroxide (e.g., OH + OH H202).

Once all possible pairs of species are allowed to undergo their individual reactions,

RADLYS permits all remaining reactants to take a diffusive jump in time. The diffusion

distance for each specific reactant is determined by both their respective diffusion

coefficients and the jump in time. Initially, all time jumps are set at 3 picoseconds so as

to avoid two reactants from jumping through each other in the simulations [Hamm et al.

1989]. This alternating pattern of chemical reaction simulation and diffusion simulation

is permitted out to a total elapsed time of one microsecond. At this time, any remaining

reactants are too far apart from one another to permit any other reactions within a single

electron track. Intertrack reactions (those between reactants from different electron

tracks) may now be considered through deterministic solutions of the coupled differential

equations governing the various chemical reactions in bulk solution. The code also

permits the simulations of reactions with molecular oxygen and other chemical free-

radical scavengers. In such cases, total track development may be complete by a much

earlier time (e.g., one nanosecond following the initial creation of the electron track).

The computer codes OREC and RADLYS have been used successfully to simulate the

irradiation of both pure water systems and aqueous solutions of a biological target

molecule such as glycylglycine [Bolch et al. 1990; Turner et al. 1991]. In the

glycylglycine work, the model has been shown to accurately predict the production of

free ammonia and other radiation products via gamma-ray radiolysis [Bolch et al. 1998].









Results

Electron version of OREC is used to transport fast electrons in straight path to

simulate the experimental conditions performed to determine the time decay of G- values

of the water radicals.

Prior to the utilization of the OREC code for these studies, several verifications

steps were implemented. These steps included the validation of electron ranges and the

generation of time-dependent G values for hydroxyl and hydrated electrons as inferred

through radical scavenging experiments. G value is defined as the yield per 100 eV

energy deposition in a medium. A full review of the code indicated that previous uses of

the OREC code had resulted in unintended corruption of counters used to report energy

loss during particle transport. Revisions to subroutines for determining the mean free

path of the particles (CALFD) and the energy loss following inelastic scattering

(ENLOSS) are given in Appendix A. After these changes were implemented, sets of

monoenergetic electron tracks ranging from 0.5 keV to 1 MeV were created and their

mean pathlengths were estimated.

Shown in Table 2.1 are the average electron ranges calculated with OREC before

and after the corrections as well as the published electron ranges calculated using the

continuous slowing down approximation, CSDA [ICRU Report 56 1997]. In this method

energy-loss fluctuations are neglected, and the negatron or positron is assumed to loose

energy along its track according to the mean energy loss per unit path length given by the

stopping power. The CSDA range for an electron of initial kinetic energy Eo slowing

down to rest is then evaluated according to:









1 -Li 1 Eq. 2.1
rcsD = 1P S(E) dE




where S(E) is the stopping power for a given energy E. Bethe stopping power theory is

not adequate for use at low energies. ICRU Report 37 suggests the use of a simple

approximation to determine the stopping power of electrons with initial energies less than

lkeV [ICRU Report 37 1984b].

RADLYS has been used, without much change, to calculate time-dependent yield

of H2, H202, *OH, and e-aq in pure water by several scientists in the past [Bolch et al.

1998; Turner et al. 1983]. One major change introduced to RADLYS, since it was

developed, was the modification on the channels through which Type 2 excitation

disassociates. The revised Type 2 excitation indicates two *OH that are formed instead

of one oxygen atom [Stabin et al. 1997].

This work adopted the latest version of RADLYS, which includes the changes

introduced by Stabin et al.[1997], to fit the current time-decay of hydrated electrons and

*OH in pure water. Most of the RADLYS parameters are fairly well known and are

agreed upon, such as diffusion coefficients and reaction rate constants. However, there

are some parameters whose values are only approximately known and can be modified to

improve model predictions. For example, the fractional partitioning coefficients, which

determine the fate of certain electronic excited states in water radiolysis during the

prechemical stage, are not well known. The revised and the previous values of

partitioning coefficients are summarized in Table 2.2.
















Table 2.1. CSDA ranges for energies ranging between 500 eV and 1 MeV as calculated
with OREC before and after corrections were made to subroutines CALFD and ENLOSS.
RCSDA values are from two sources; (1) Berger [1973] for electron energies less than 10
keV and (2) ICRU Report 56 [1997] for electron energies greater than 10 keV.


ELECTRON RANGE (gim)


ENERGY (keV)



0.5

1

10

50

100


400

800

1000


OREC+
(Before Correction)

4.25E-2

8.34E-2

3.02E+0

6.46E+0

2.91E+2


2.79E+3

5.17E+3

7.88E+3


OREC*
(After Correction)

3.88E-2

7.61E-2

2.63E+0

4.34E+1

1.43E+2

1.25E+3

3.28E+3

4.37E+3


* Coefficients of variation associated with calculations are less than 1%.
+Coefficients of variations associated with calculations are more than 10%.


RCSDA

2.27E-2

6.01E-2

2.49E+0

4.40E+1

1.40E+2

1.26E+3

3.25E+3

4.30E+3









Table 2.2. Additional changes proposed to improve pure water calibration.


Lavern and
Parameter Stabin et al. Present Study Pimblott
Pimblott
D(e-aq) [10-5cm2/s] 5.0 4.5 4.5

D(*OH) [10-cm2/s] 2.0 2.8 2.8
H20DIFF (apm) 0.00125 0.002




Slight adjustments were also made in some other parameters to attain a better

model fit. For instance, diffusion coefficients (D) for the reactants involved in the short-

time radiolysis of water have been modified according to published data [Pimblott and

LaVerne 1997, 1998]. Positive improvements in the model fits are observed when

D(*OH) and D(e-aq) are modified. Another adjustable parameter in the model is H2ODIF,

the mean distance an ionized water molecule travel prior to its interaction with a water

molecule to form H30 and two *OH. Dramatic changes in the model response are

observed when changing this parameter. Nevertheless, no study was found to discuss

potential values of the parameter H2ODIF. Table 2.3 summarizes the additional changes

made in the model parameters. A copy of the RADLYS code and input file that lists all

the model parameters and their current values are given in Appendix B.

The OREC code has been modified to move electrons in straight paths and then

energy loss 100 jam track segments from one hundred 1.MeV electron tracks are

obtained. The for a 100 jam track segment is calculated to be about 1% the initial energy.

Each electron track produces a large number of species within the track, with great

variations (-1,000-5,000), due to the nature of stochastic processes explained previously









when discussing radiolysis of water and OREC/RADLYS codes. The statistical

uncertainty, however, is found to be less than 1% with the G(*OH) and G(e-aq) when 100

electron track segments are used. Copies of OREC and RADYLS outputs are presented

in the Appendix C.




Table 2.3. Revised partitioning coefficients of excitation in RADLYS, with comparison
to previously used values


Partitioning Coefficients (%)


Type Reaction Boch et al. Stabin et Present
Bolch et al.
al. Study


H,0* H + OH 75 14 12

Type 1
Type H,0* HO +AE 25 86 88


HO* H2 +O 100 0 0
0 62
Type 2 HO* + HO H+ 20H 0 60 50
0 40
HO* HO +AE 0 40 50


H20* -- H20+ + eub
57 47 33
Type 3-6 H2O* -H+ OH 20 4 1

H,2 0* HO +AE 23 49 57


















Time dependence of OH radical decay


6.0

5.0

4.0

3.0

2.0

1.0

0.0


1.E-12 I.E-11 i.E-10 i.E-09 i.E-08 i.E-07 i.E-06

Time (s)


Figure 2.1. Time dependence of G(*OH) as generated currently by RADLYS and the
G(*OH) calculated via the inverse Laplace transform of scavenging capacity as
performed by LaVerne and Pimblott [1991].




















Time dependence of hydrated electron decay


6.0

5.0

4.0

3.0

2.0

1.0

0.0


1.e-12 .e-11 1.e-10


l.e-9


1.e-8


l.e-7


l.e-6


Time (s)



Figure 2.2. Time dependence of G(e-aq) as generated currently by RADLYS. Shown in
comparison are the result of inverse Laplace transforms of G(e-aq) calculated from
scavenging capacity as performed by LaVerne and Pimblott [1991].


0 RADLYS----
---LaVeme & Nnblott 1991














CHAPTER 3
DEVELOPMENT OF NEAR-APPROACH MODEL AND STRUCTURAL
INFLUENCES OF SITE-SPECIFIC *OH ATTACK TO THE SUGAR MOIETY



Introduction

Understanding the basic physical and chemical mechanisms of radiation action on

nuclear DNA is one of the fundamental goals in radiation research. Through the late

1970s, the majority of data on radiation-induced DNA damage had been derived from

experimental studies [Holley and Chatterjee 1994]. Within the last two decades,

however, computational modeling has become an increasingly important tool in

understanding the molecular events that take place between the initial deposition of

energy within the biological system and experimental observations of DNA damage.


Previously Developed Biophysical Models of Radiation Damage to DNA

Various biophysical models have been proposed which seek to simulate the

production of lesion formation within DNA by ionizing radiation. Many of the radiation

damage models developed in the late 1980s were based in terms of energy deposited or

the number of ionizations (direct effect) created within cylindrical nanometer-scale

volumes representing the biological structure of the DNA double helix, the nucleosome,

and chromatin fiber [Charlton and Humm 1988; Charlton et al. 1989; Hamm and Turner

1992; Nikjoo et al. 1991; Ottolenghi et al. 1995]. More recent biophysical models of

radiation action on DNA have focused on both the atomic structure of the DNA molecule

as well as higher-order structures of DNA conformation. These atomistic models are









based on simple double-helical structures found in plasmids [Bardash and Zaider 1994;

Chatterjee and Holley 1990; Pomplun and Terrissol 1994] and nucleosomes [Michalik

and Begusova 1994; Terrisol 1994]. Chromatin fibers have been introduced in a "beads-

on-a-string" form [Chatterjee and Holley 1990], a solenoid form [Charlton et al. 1994], a

crossed-linker form [Holley and Chatterjee 1996], and a stochastic configuration [Jacob

et al. 1997]. Other efforts have been devoted to exploring quasi-direct action to the DNA

hydration layer [Friedland et al. 1999; Nikjoo and Charlton 1995; Zaider et al. 1994] and

high-resolution atomistic models for radiation damage simulation [Friedland et al. 1999;

Friedland et al. 1998; Michalik et al. 1995; Michalik et al. 1995; Tomita et al. 1994;

Tomita et al. 1998].


Previous Models of Indirect Action of Radiation

Indirect mechanisms of radiation-induced DNA damage are particularly relevant

for low-LET radiations. In models of indirect radiation action, detailed information is

needed on both the mechanisms of free radical attack to the DNA molecule and the

spatial distribution of radiolytic products relative to the DNA target. The following is a

brief review of previous DNA damage models highlighting only their treatment of OOH

diffusion and chemical attacks to the DNA target molecules.

Chatterjee et al. (1985-1994). Chatterjee et al. pioneered the study of chemical

pathways for radiation-induced DNA damage and developed one of the first

computational models of indirect action of radiation with DNA [Chatterjee and Magee

1985]. Their models were subsequently extended to include radical damage to chromatin

[Chatterjee and Holley 1990, 1991; Chatterjee et al. 1994]. In these studies, the water

radicals, whose initial positions are determined via Monte Carlo sampling from Gaussian









spatial distributions, are then followed in space and time through simulations of their

diffusion motion. Survival of any given *OH was assumed to decrease exponentially

according to the scavenging capacity of the medium. Radical reactions were then scored

when a surviving *OH was found to be within the reaction radius of either the sugar or

base moiety. The positions of both the bases and sugars, modeled as individual spheres,

were obtained from x-ray diffraction studies.

Terrisol and Pomplun (1994). Terrisol and Pomplun [1994] used Monte Carlo

simulations of Auger electron tracks from localized decay of 125I within the DNA

molecule to develop models of both direct and indirect radiation damage to DNA. Both

duplex DNA and nucleosome DNA were considered in their studies. Atomic coordinates

were obtained from x-ray diffraction studies [Chandrasekaran and Arnott 1989]. Direct

hits were scored during the physical phase of electron transport, after which reactions

between free radicals and the sugar and base moieties, and with background scavenger

molecules, were considered.

Tomita et al. (1994). Tomita et al. [Tomita et al. 1994; Tomita et al. 1998]

developed a model to compute *OH damage to several DNA structures including one

turn of the duplex DNA, a nucleosome, a solenoid conformation of chromatin fiber, and

the plasmid pBR322 [Tomita et al. 1994; Tomita et al. 1998]. In the latter study, the

atomic coordinates of standard B-DNA [Saenger 1984] were spirally fit within the

supercoiled structural conformation of the plasmid. Due to computational time

constraints, explicit treatment of *OH diffusion was not considered. Instead, time and

position-dependent reaction probabilities were utilized to score *OH interactions with the

DNA at specific sites [von Sonntag 1987] (i.e., C(5) and C(6) of cytosine with the









relative probabilities of 0.4 and 0.6, respectively. Only the C(4') of the sugar moiety was

considered the site of *OH attack.

Michalik et al. (1995). Stochastic models of *OH reaction with DNA have been

developed by Michalik et al. [1995]in which structural influences of DNA were

investigated in simulations of *OH attack to DNA in its A, B, and Z conformations. The

sequence dependency of *OH attack probabilities was additionally studied. Short DNA

molecules were downloaded from Protein Data Bank and each atom was modeled

explicitly by its van der Waals radius. Sites of *OH attack within the sugar moiety were

explicitly modeled at all carbon atoms; sites of *OH attack to the base moiety were

modeled at reaction sites given by von Sonntag [1987].

Nikjoo et al. (1997). Nikjoo et al. [1997] developed a DNA target structure

consisting of a cylinder divided into two regions: one for the sugar-phosphate groups and

one for the nucleobases. Diffusion of the *OH was simulated until they reach the

cylindrical volume representing the DNA molecule. The interaction of *OH was

determined using the relative probabilities of 20% and 80% for *OH attack to the sugar

and base moieties, respectively, as referenced by Scholes and Simic [Scholes 1983;

Scholes and Simic 1968].

Friedland et al. (1998-1999). Friedland et al. have recently published one of the

more complex structural models of DNA in which atomic-level definition is retained

through the simulation of an entire chromosome [Friedland et al. 1999; Friedland et al.

1998]. Detailed information is given regarding the simulation of both direct and quasi-

direct interactions with the DNA structure, as well as methods by which intratrack free

radical chemistry is treated. Reaction radii from Buxton et al. [Buxton et al. 1988] are









utilized for *OH interactions with the DNA. The atomic coordinates of the DNA

molecule were obtained from X-ray diffraction studies [Chandrasekaran and Arnott 1989]

and were utilized within the structural model of the chromatin fiber and nucleosomes.

One feature of DNA structure that is missing from this model, and the other

models discussed above, is the influence non-reacting atoms could have on the reaction

of the hydroxyl with specific sites on the DNA molecule. The volume that these non-

reacting atoms occupy along the surface of the molecule is expected to sterically hinder

the accessibility of the hydroxyl from some reactive sites on both the DNA sugars and

bases. Consequently, the steric hindrance exhibited by these non-reacting atoms may

play an important role in determining the distribution of hydroxyl-mediated DNA

damage and, thus should be included when modeling hydroxyl and other radical reactions

with DNA.


Experimental Data on Site-Specific *OH Attack

In a recent study, Balasubramanian et al. investigated site-specific hydroxyl

interactions with DNA. In their studies, five sets of 19-bp duplex DNA molecules were

created via PCR in which deuterium was selectively incorporated at one of the five

carbon sites of the sugar moiety within a given deoxynucleotide of the molecule. These

DNA solutions were then subjected to hydroxyl cleavage reactions. Strand break

products observed on the denaturing gels were DNA fragments having free phosphate

ends. In principle, these strand breaks could result from initial hydrogen abstractions at

any of the several different deoxyribose hydrogens [von Sonntag 1987]. However, it has

been proven that hydrogen abstractions at C(5') and C(4') give rise, under oxygenated

conditions, to chemically unique products, specifically 5'-aldehyde and 3'-









phosphoglycolate, respectively. Nevertheless, several initial hydrogen abstraction

reactions can, in theory, lead to the production of 3' phosphate [Balasubramanian et al.

1998]. Balasubramanian et al. [1998] measured the intrinsic isotope effects for C(5')

and C(4') hydrogen abstractions by quantifying these two unique chemical products.

Apparent isotope effects for the other deoxyribose carbon sites were determined by

detecting 3' phosphate. The fractional contributions of each deoxyribose carbon site to

the total DNA cleavage were estimated by using the intrinsic isotope effect for the C(5')

and C(4'), and the apparent isotope effect for the remaining deoxyribose carbon sites.

Their result concluded that the hydrogens of C(5') and C(4') are the primary sites of

hydroxyl attack. Furthermore, these authors showed that their experimental data were in

general agreement with the accessibility of the sugar hydrogens to approaching *OH.

The present study had three aims: (1) to develop a detailed Monte Carlo model to

simulate site-specific *OH reactions with duplex DNA in which preferences at the

different reactions sites are accommodated through both steric hindrance by non-reacting

atoms of the structure and the overall geometrical configuration of molecule; (2) to make

systematic sensitivity studies of the various Monte Carlo model parameters and study

their effects with regard to site-specific *OH attack probabilities; and (3) to compare our

detailed computational model predictions with the experimental data of Balasubramanian

et al. [1998] to obtain more detailed information on the mechanism of hydroxyl attack

with DNA duplex before and after energy minimization, respectively, as created within

HyperChem. A space-filled model of the energy-minimized form is shown in Figure 3.1

C.









Materials and Methods


Construction of the DNA Molecular Target Model

The accessibilities of the various radical reaction sites in DNA are dependent not

only on their 3D spatial orientation relative to other reactive atoms, but also on the

intermediate positions of non-reactive atoms within the molecule. The latter can

sterically hinder the approach of diffusing reactants in reaching potentially reactive sites

interior to molecular structure. In this study, atomistic models of duplex DNA molecules

are constructed which explicitly incorporate steric hindrance. Additional features include

the tightly bound water molecules of the first hydration shell, the presence of counter

ions, and 3D geometrical optimization of the molecular structure via energy minimization

algorithms. DNA structures are created using two software packages: MacroModel

version 6.5 [Mohamadi et al. 1990] and HyperChem version 6.0 [HyperCube Inc. 2000].

Both codes allow the user to construct and graphically manipulate simple as well as

complex chemical structures. Algorithms for molecular mechanics and dynamics are

available within each to evaluate bond energies and produce energy-minimized

geometries for various molecules in aqueous solution. HyperChem additionally allows

the user to consider the positions of counter ions.

In this study, the self-complementary decamer duplex d(CCAACGTTGG) is

created using both HyperChem and MacroModel. Balasubramanian et al. [1998] used

this molecule in looking at solvent accessible surface areas (SASA) for comparison to

their experimental data on site-specific *OH cleavage. In both programs, the base-pair

sequence is created in both its standard B-DNA conformation, as well as in an energy-

minimized configuration using the AMBER algorithm (Figures 3.1.A and 3.1.B).





















I,




*SL.,.

-J .
-lt


- y


ij~l-- 4--fl


,


Figure 3.1. Stick model representation of the self-complementary decamer duplex
d(CCAACGTTGG) shown with counter ions (o) in the B-DNA configuration either (A)
without geometrical optimization or (B) with geometrical optimization via energy
minimization. Both configurations were generated with HyperChem v6.0. (C) A space-
filled representation of the geometrically optimized structure along with bound water
molecules and counter ions. Color code: violet counter ions, dark small gray -
hydrogen, light gray water molecules, blue nitrogen, cyan carbon, and red oxygen.









Inclusion of the Hydration Layer

The number and the coordinates of the tightly bound water molecules of the first

hydration layer are determined using the model of Umrania et al. [1995] as further

modified by S.G. Swarts Table 3.1 summarizes the location of each water molecule by

the cylindrical coordinates relative to DNA atom to which it is bound. A total of 17

water molecules around each C-G and A-T base-pair are manually added to the DNA

atomic structure as given by HyperChem and/or MacroModel. The first hydration layer

is thus defined in the model as five water molecules positioned around each phosphate

group, one about each deoxyribose moiety, three about adenine, three about guanine, two

about thymine, and two about cytosine. While these waters play only a minor role in the

indirect action of *OH attack, they will be used in subsequent studies for consideration of

quasi-direct action by incident charged species (e.g., electron-deficient "holes" and

electrons).


Reaction Sites


Deoxyribose sugar moiety

*OH attack via an abstraction reaction with the various hydrogen atoms of the

sugar moiety is assumed to be random and nonselective [von Sonntag 1987] despite

experimental studies that have suggested attack at the C(4') hydrogens is more likely than

at C(3') or C(5') [von Sonntag 1987]. This latter assumption is based on estimated

reaction probabilities only and does not take into effect steric hindrance provided by the

entire atomic structure of the molecule [von Sonntag 1987]. Other studies conclude that


'Personal communication between B. Aydogan and S.G. Swarts, June 2000.













Table 3.1. The positions of each tightly bound water molecule within the first hydration
layer of duplex DNA listed according to their relative position, in cylindrical coordinates,
from the specified reference atom in the model. For example, the first entry indicates that
a bound water is located 2.91 A from the 1st oxygen atom of each phosphate group at a
location given by a polar angle of 530 and an azimuthal angle of 1450, the latter measured
from the central axis of the DNA helix.




Origin r (A) (0) (0)

Phosphate
01P 2.91 53 145
01P 2.63 32 -103
02P 2.95 48 179
02P 2.76 46 -74
02P 2.95 48 -179
Sugar
04 2.84 53 -174
Cytosine
02 2.94 22 -160
N4 3.07 63 9
Thymine
02 2.77 22 -123
04 2.68 45 15
Adenine
N3 2.84 43 23
N6 3.18 41 176
N7 2.70 45 -7
Guanine
N7 2.70 43 -176
N3 2.91 45 7
06 2.77 46 -9









since *OH are very reactive, they must, in principle, be equally selective among all

hydrogens within the sugar and the extent of *OH interaction with any given hydrogen is

governed by its accessibility to the solvent [Balasubramanian et al. 1998].

Experimentally, evidence exists to support this hypothesis as damage products have been

measured for all hydrogen abstraction sites except at C(2') [Cadet et al. 1999]. The

product of *OH attack at C(1') has been found to be 2-D-deoxyribonolactone [von

Sonntag 1987]. While potentially all hydrogen atoms are subject to H-abstraction by

approaching *OH, only attacks at C(5'), C(4'), and C(3') are recognized as sites for

inducing DNA strand breaks [Cadet et al. 1999; von Sonntag 1987]. In this study,

hydrogen-abstraction reactions by *OH are considered at all hydrogens of the

deoxyribose moiety using constant-value reaction radii for each site. Preferences for

abstraction at C(5'), C(4'), and C(3') are accommodated through steric hindrance by non-

reacting atoms within the target structure.


Nucleobases

Modified purine and pyrimidine bases comprise one of the major classes of

hydroxyl-mediated DNA damage. Detailed information on hydroxyl interactions with

free nucleobases can be found for both thymine and guanine, whereas data are not readily

available for adenine and cytosine [Douki et al. 1998; von Sonntag 1987]. *OH adduct

formation at the nucleobases is shown to be somewhat selective. For example, guanine is

widely accepted as the most susceptible DNA target, due in part to a wide variety of

observed oxidation-mediated damage products for this base [Douki et al. 1998; Scholes

1983; Scholes and Simic 1968; von Sonntag 1987]. Recent studies report that the

contribution of total guanine base damage is -60% at C(4) and -25% at the C(8) position









[Cadet et al. 1999]. The total contribution of the C(2) and C(5) positions accounts for the

remaining 15% in aerated aqueous solutions [von Sonntag 1987]. For thymine, *OH add

preferentially to C(5) (60%) and then to C(6) (40%) base [Cadet et al. 1999; Douki et al.

1998; von Sonntag 1987].

In the present computational model, *OH adduct formation is considered at C(2),

C(4), C(5), and C(8) for both adenine and guanine, and C(5) and C(6) for both cytosine

and thymine [Cadet et al. 1999; Douki et al. 1998; von Sonntag 1987]. Again,

preferences for abstraction at the specific reaction sites are accommodated through steric

hindrance by non-reacting atoms. H-abstraction by the *OH from the three hydrogen

atoms of the methyl group of thymine is treated as well. The model can additionally

consider reactions by both Ho and the hydrated electron. Ho are permitted to undergo the

same reactions as *OH with DNA but at reduced rates (smaller reaction radii), while

hydrated electrons are allowed to form electron adducts uniformly at all atoms within the

nucleobases [von Sonntag 1987].


Reaction Radii

Radii for *OH reaction are assigned to each reacting atom in the target molecule

based upon the observed reaction rate constants for free moieties and the Smoluchowski

diffusion-controlled rate equation [Smoluchowski 1916]:


kd= 47r(D, +Db )r


Eq. 3.1









where kd is the diffusion-controlled reaction rate constant, Da and Db are the diffusion

constants for the two interacting species (*OH and the DNA atom in question), and rab is

their reaction radius. Diffusion constants for the different DNA moieties can be

assumed to be zero (stationary target); hence, the interaction radius is given as:


Eq. 3.2


kd
4ab rD
OH


where D*OH = 2.8 x 10-5 cm2 -1 [Pimblott and LaVerne 1998]. Reaction rate constants

used in this study are taken from Buxton et al. (1988). In the cases where multiple values

are reported, the average value is taken. Table 3.2 gives the hydroxyl reaction rate

constants and their corresponding reaction radii for the different DNA constituents.




Table 3.2. Reaction rate constants and reaction radii for *OH with deoxyribose and the
four nucleobases.


Reaction Site for *OH


Adenine

Deoxyribose

Cytosine

Guanine

Thymine


Reaction Rate
Constant
(109 M-1s1)

6.10

2.50

6.10

9.20

6.40


Reaction Radius
(A)

2.88

1.18

2.88

4.34

3.02









In the Smoluchowski formalism, one treats all chemical species as points within

the solution such that the likelihood of a reaction is determined by comparing their

separation distance against the reaction radius of Eq. 3.2. In the present study, non-

reacting atoms, the bound water molecules, and the counter ions are all considered as

spheres of size given by their van der Waals radii. These values are shown in Table 3.3

and are taken from Saenger [1984]. In this same manner, the approaching hydroxyl is

modeled as a sphere with an effective van der Waals radius of 1.2 A. As such, the

Smoluchowski formalism for radical reactions is modified in this study so that all

reacting atoms of the target are treated as spheres with a size given by their reaction radii

with *OH (Eq. 3.2). In this manner, *OH reactions are permitted to occur if the spherical

volume of the approaching *OH overlaps the spherical volume of a reacting atom (e.g.,




Table 3.3. Van der Waals radii for non-reacting atoms, the bound water molecules, and
counter ions within the model of the DNA target molecule.



van der Waals
Atom, Molecule, or Ion
radius
(A)
H 1.20

C 1.70

N 1.55

0 1.40
P 1.80
Water molecules of the 1
1st hydration layer
Na+ counter ions 1.02









hydrogens within both deoxyribose and the methyl group of thymine for abstraction

reactions, or the carbons of the nucleobases for *OH adduct formation).


Monte Carlo Algorithm for *OH Diffusion and Reaction

In simulating indirect attack by radiolytic free radicals to pure DNA in solution,

the initial radical interactions with the target may be considered separate from the

previous diffusion history of the radicals within the surrounding aqueous solution.

Consider, for example, a single *OH that has diffused to within the first hydration

layer of a given nucleotide within the target. Its interactions with atoms at either the

sugar or base moieties are independent of both its origin within the track of the secondary

electron which created it as well as its previous random walk through the medium. In

this study, we focus only on proximal interactions of the hydroxyl with DNA atomic

structure.

As shown in Figure 3.2 an elliptical starting volume for *OH is created whose

dimensions are determined by the outer-most extent of the hydration layer and whose

thickness is equal to two times the van der Waals radius of an *OH. The height of the

cylinder is determined by the dimensions of the DNA molecule along its long axis. A

second, larger elliptical buffer zone is located beyond the inner starting volume to allow

the return of an outwardly diffusing *OH back within the starting volume. The radical in

the starting volume is first allowed to undergo a constant diffusive jump in an isotropic

direction. The diffusive jumps in time are held to within 3 picoseconds [Hamm et al.

1998]. The root-mean-square jump distance / is given by the following expression:





















Furthest extend of
Bound Water
(Hydration Layer)












Starting Volume for
Diffusion Buffer Zone Approaching Radical
(3 x van der Waals Radius) (2 x van der Waals



Figure 3.2. Schematic geometry (top view looking down the long axis of the double-
helix) of a 3-bp segment of the DNA molecule in its B-configuration (without energy
minimization). The starting volume for the approaching radical is represented as an
elliptical cylinder whose inner boundary is defined as the furthest extent of the first
hydration shell. A diffusion buffer zone is also defined from which the diffusing radical
may re-enter the starting volume. If the radical leaves the diffusion buffer zone, the
radical is scored as an escaped reactant re-entering the bulk solution. Color code: red -
atoms of the DNA molecule, dark green water molecules, and yellow counter ions.









Table 3.4. Diffusion coefficients and jump distances for the OH. Two jump times are
considered: 3 and 0.5 ps.




D (A) / (A)
(1O- cm2 s-) z =3ps z =0.5ps

2.5 2.12 0.87


2.8 2.14 0.92






1= 6 DOH Eq.

3.3



where D*OH is the diffusion coefficient of the *OH and z is the jump time.

Table 3.4 displays two reported values of the diffusion coefficient of the

hydroxyl, along with their corresponding jump distances for jump times of either 3 or 0.5

picoseconds.

There are five possible outcomes after a given diffusive jump of the OH. These

are: (1) the radical is found to have jumped beyond the buffer region and is no longer

considered; (2) the radical has jumped to a location such that a volume overlap with a

bound water molecule has occurred; (3) the radical has jumped to a location such

that a volume overlap with a non-reactive atom has occurred; (4) the radical has jumped

to a location such that a volume overlap with a reacting atom has occurred; or (5) the

radical has diffused to any other location within the DNA target volume.









If the outcome of the diffusive jump is Case (1), the event is scored as no

interaction. If the outcome is Case (2), then a swap of the positions of the radical and the

bound water molecule is permitted and additional reactions are evaluated. Under Case

(3), the radical is allowed to deflect or "bounce off' the non-reactive atom using an

algorithm discussed below. For Case (4), the chemical reaction (H-abstraction or adduct

formation) is allowed to occur and the type and location is scored. Finally, if the

diffusive jump results in Case (5), an additional diffusive jump is permitted and the

resulting new position of the radical is evaluated as before.

In some instances, more than one case will be under simultaneous consideration.

For example, a diffusive jump might place the radical within the reaction radii of more

than one reactive atom. In this situation, the selection of the reaction is determined via

Monte Carlo sampling, based upon both the relative sizes of the reaction radii in question

and the proximity of the radical to each reacting atom. Another example might entail a

diffusive jump that places the radical to within both the reaction radius of a reacting atom

(or atoms) and the van der Waals radius of a non-reacting atom (or atoms). In its present

form, the computational model assumes that steric hindrance will redirect the radical in

its random diffusive walk prior to a possible reaction with the adjacent reacting atoms.

Another important aspect of this near-approach model is radical "bounce off"

from non-reacting atoms as described above in Case (3). This is the central feature in the

model where steric hindrance of the approaching radicals is explicitly treated. In the

present model, when a diffusing radical moves to within the van der Waals radius of a

non-reacting atom, the jump is not permitted. Instead, the radical is allowed another

attempt at an isotropic diffusive jump, but over a smaller time interval (0.5 ps or 1 ps).

























_ I____ajl.


o \ T *v
',_


Figure 3.3. A schematic representation of *OH diffusion and near-approach interaction
with a DNA molecule in its geometrically optimized configuration. The cylinder surface
represents the inner surface of the starting volume in Figure 3.2. A single *OH is shown
entering the hydration layer. After several diffusive jumps, it exits the hydration layer
only to re-enter at a position higher within the structure.






50



















[I. k., F


I3'


I J i I I "
I '. .- ," .



'..11. :- -{ .._) ,
"--"









Figure 3.4. A magnified representation of this same radical as it approaches a
deoxyribose sugar. The radical interacts with both the oxygen atom adjacent to the C3'
carbon, and then with the C3' carbon itself, undergoing "bounce-off' deflections with
both non-reacting atoms. The .OH continues its diffusion path behind the C4' carbon,
where it eventually undergoes a hydrogen abstraction at the first hydrogen of the C5'
carbon atom.









We refer to this reduced time interval as TSH denoting the jump time utilized for modeling

steric hindrance (SH). The diffusive jumps this particular radical are continued at a jump

time of TSH until the radical either undergoes an H-abstraction or adduct formation

reaction, or escapes from the buffer region of the target molecule. Additional checks are

made to avoid jumping through non-reacting atoms which might impose a narrow

channel of access to reacting atoms interior to the structure. Figures 3.3 and 3.4 display

pictorial examples of the diffusion and bounce-off algorithms applied with in the Monte

Carlo simulations.


Calculations of %SASA

In molecular modeling of chemical reactions, the van der Waals surface is defined

as the molecular envelope encompassing all spherical atomic volumes within the target

molecule. The solvent accessible surface area (SASA) of individual atoms or groups of

atoms corresponds to the fraction of that total surface area contributed by each atom or

atom group. Operationally, it is determined by rolling a spherical probe of radius Rw=

1.4 A, simulating a single water molecule, along the van der Waals radius surface of a

target molecule [Saenger 1984]. This process is demonstrated in Figure 3.5. In this

study, the SASA for approaching *OH is calculated by using the HyperChem software

with a spherical probe of radius R.OH = 1.2 A. Consequently, SASA is used in the present

study to define the accessible surface of any atom or atom group to the *OH diffusing

through the bulk solvent enveloping the DNA target. Accessible surfaces for the *OH

were determined for all seven hydrogen atoms of the five carbon sites in a sugar moiety

including 1H5', 2H5', H4', H3', 1H2', 2H2', and HI'. HyperChem [HyperCube Inc.

2000] employs the grid method that is computationally slower but yields greater accuracy




























iIO








Figure 3.5. Illustration of the solvent-accessible surface area (SASA) and the van der
Waals surface. A spherical probe is moved around the van der Waals molecular
envelope, its center describing the perimeter of the accessible surface. The SASA
approaches the van der Waals surface as the probe size is decreased toward zero radius.









compared to other algorithms for a given set of atomic radii [Bodor et al. 1989]. The

percent solvent accessible surface area (%SASA) for hydrogen abstractions at the

deoxyribose sugar gives the contribution of each hydrogen to the total solvent accessible

surface area of the sugar moiety. In this study, values of %SASA are obtained by

averaging this quantity over the central eight base-pairs of the decamer duplex

d(CCAACGTTGG).


Determination of optimal model parameters

Several model sensitivity studies were carried out in order to determine best-fit

reaction parameters and modeling assumptions for Monte Carlo simulations of OOH

attack to DNA. These included: (1) the presence versus the absence of geometrical

optimization via energy minimization, (2) the presence versus the absence of counter

ions, (3) the diffusive jump time for approaching *OH, (4) the effective van der Waals

radius of the OH, (5) the reaction radii of the deoxyribose hydrogen atoms, and

(6)explicit consideration versus no consideration of steric hindrance by non-reacting

atoms of the DNA. In each case, the selection of a best-fit parameter value or model

assumption was based on the model's ability to report site-specific *OH attack

probabilities at the deoxyribose hydrogen atoms that matched %SASA values given for

our benchmark DNA model (geometrically optimized B-DNA decamer duplex with

counter ions present as given by HyperChem). In this study set, only sugar reactions are

considered.

A second set of analyses was made which compared the output of the

computational model with experimental *OH cleavage rates as measured by

Balasubramanian et al. [1998]. In these studies, *OH reactions with both the sugars and









nucleobases were permitted. These simulations were run using the optimal parameters

determined from the sensitivity analyses of %SASA discussed above. A set of four

different models of the decamer duplex d(CCAACGTTGG) was created, two with

HyperChem and two with MacroModel, in which geometry optimization via energy

minimization was or was not considered. A fifth model, serving as our benchmark DNA

model for the entire series, was geometrically optimized using HyperChem and included

the presence of Na+ counter ions. Energy minimization was achieved with the AMBER

algorithm [Weiner et al. 1984] using the distance-dependent dielectric constant to mimic

the screening effect of solvation. Consequently, inclusion of the tightly bound water

molecules of the first hydration shell was added following geometry optimization. One

million *OH were randomly introduced within the starting volume of the model resulting

in statistical uncertainties of the *OH attack probabilities below 1%. Calculations were

performed using a Dell Precision 420 Dual 950-MHz processor computer.



Results


Percent Solvent Accessible Surface Area (%SASA)

Values of %SASA are shown in Figure 3.6 for each of the five DNA structures

considered. In each calculation, a probe radius of 1.2 A is used. The two commercially

available software packages calculated %SASA values to within 5% for all hydrogen

sites. Interestingly, the optimized structure from MacroModel more closely agrees with

the %SASA predictions of the benchmark model than those from the optimized

HyperChem model without counter ions. For both codes, little variation in %SASA is

indicated for hydrogen atoms associated with C(3'), C(2'), and C(1'). In contrast, the




















Non-optimized (HC)
-* Non-optimized (MM)
C0 Optimized (HC)
0 Optimized (MM)
Benchmark DNA Model (HC)


1H5'+2H5'


H4'


H3'


1H2'+2H2'


Hydrogens of the Sugar Moiety


Figure 3.6. Comparison of %SASA for each of the five DNA structures considered in the
study. The various models either consider or do not consider geometrical optimization
and are created with either the HyperChem (HC) or MacroModel (MM) software
package. A probe radius of 1.2 A is used equal to the effective van der Waals radius of
an OH. The Benchmark DNA Model is equivalent to the optimized HC structure, but
additionally includes the Na+ counter ions.


HI'


















A R(OH) = 0.0
R(OH) = 1.2
R(OH) = 1.4
R(OH) = 2.5



A



A A


t


1H5'+2H5'


1H2'+2H2'


Hydrogens of Sugar Moiety


Figure 3.7. Variations in site-specific %SASA among the deoxyribose hydrogen atoms
with changes in the radius of the spherical probe used to make the estimates (see Figure
3.5)









single hydrogen atom at C(4') is shown to capture -39% of the solvent accessible surface

area in the non-optimized B-DNA configuration, while the two hydrogen atoms at C(5')

contribute only -31% to the total SASA. When energy minimization of the molecule is

considered, however, the two hydrogen atoms at C(5') become the dominate reaction sites

within the structure.


Sensitivity of %SASA with Variations in Probe Size

Figure 3.7 shows variations in the %SASA with changing probe size. Only small

differences are noted in %SASA values for probes of 1.2 A and 1.4 A in size. These

sizes represent the effective van der Waals radii of *OH and the water solvent molecules,

respectively. The largest difference is seen when the probe is reduced to a single point as

suggested in the Smoluchowski reaction scheme. An effective radius for the OH, ROH, is

thus taken as 1.2 A for all %SASA values used in the following parameter sensitivity

analyses.


Considerations of Geometrical Optimization via Energy Minimization

In this study, the computational model of *OH attack was used to determine the

percentage of total hydrogen abstractions which were seen at each of the five hydrogen

sites within the deoxyribose sugar. As stated earlier, these percentages were averaged

over the central eight base-pairs of the decamer duplex. To permit comparisons latter

with experimental data, the percent attacks at the two hydrogen atoms of the C(5') and at

the two hydrogen atoms of C(2') were summed for each carbon site. The data are shown

in Figure 3.8 indicate that for both the geometrically optimized and non-optimized

configurations (created via HyperChem), the Monte Carlo predictions of *OH attack



















% SASA (optimized)
S0% OH Attack (optimized)
0% SASA (non-optimized)
___ % OH Attack (non-optimized)





8 *

-------^- 8---


1H5'+2H5'


1H2'+2H2'


Hydrogens of Sugar Moiety


Figure 3.8. Variations in site-specific %SASA and % *OH attack probabilities for the
decamer duplex DNA molecule under both standard B-DNA configuration and under B-
DNA configuration following geometrical optimization via energy minimization.






















60


50 %OH Attack Without Counterion
O %SASA Without Coutenon
____ /oOH Attack With Countenon
S40
O %SASA With Counterion

30


020


S10


0
1H5'+2H5' H4' H3' 1 2'+2H2' HI'
Hydrogens of Sugar Moiety


Figure 3.9. Variations in site-specific %SASA and %*OH attack probabilities using the
geometrically optimized structure with or without the presence ofNa+ counter ions.









track closely with the percent solvent accessible surface area (1.2 A probe radius). The

greatest disagreement between the Monte Carlo calculations and the %SASA is seen at

C(2') for the non-optimized configuration (14.4% versus 9.1%). As seen previously, a

preference for *OH attack at the C(5') hydrogen sites is shown with the geometrically

optimized structure. All Monte Carlo simulations were performed with an *OH radius of

1.2 A, a diffusion jump time of 3 ps, a "bounce off' correction jump time of 0.5 ps, and a

reaction radius for *OH hydrogen abstraction of 1.18 A at all hydrogen sites in the sugar.


Sensitivity of *OH Attack Probabilities to the Inclusion of Counter Ions

Two DNA molecules, one with sodium counter ions and one without counter ions,

were created and geometrically optimized with the HyperChem software. These ions

were modeled as non-reacting spheres of radius equal to 1.02 A (see Table 3.3). OOH

attack probabilities for the hydrogens with the sugar moiety were calculated and

compared with their corresponding %SASA values. As shown in Figure 3.9, steric

hindrance by the counter ions played a very minor role in attack probabilities at C(2') and

at C(3'), with a slightly greater role seen at C(3') and at C(5'). The largest impact shown

was at C(4') where the % *OH attack at changed decreased from 28% to 22% with the

addition of the Na.


Sensitivity of *OH Attack Probabilities with Variations in Diffusion Jump Time

*OH attack probabilities in this study were initially determined using a steric

hindrance jump time TSH of 0.5 ps within the bounce-off algorithm. In addition, the

primary jump time, t, of the approaching radicals was systematically varied from

between 1 ps and 4 ps. As shown in Table 3.5, the various site-specific attack









probabilities were found to vary only slightly with variations T. The best agreement

between the %SASA and *OH attack probabilities is observed for an initial jump time to

3 ps. Interestingly, this was the jump time suggested by the Hamm et al. [1989] when

considering "jump through" during *OH chemistry in aqueous solution. The

corresponding jump distance for the *OH is thus 2.2 A. This value is larger than all radii

of atoms found in our DNA model. In treating steric hindrance, however, the

corresponding jump distance for TSH = 0.5 ps is -0.9 A, a value smaller than all atomic

radii (see Table 3.3). *OH attack probabilities changed negligibly when tSH was varied

between 0.1 ps and 1 ps. Further model simulations were thus made using values of 3 ps

and 0.5 ps for T and TSH, respectively.


Sensitivity of *OH Attack Probabilities with Variations in ROH

Table 3.6 displays *OH attack probabilities from the computational model with

variations in effective radius, ROH, of the approaching hydroxyl. Ratios of the OOH

attack probabilities to their corresponding % SASA, in which a probe sphere of

equivalent radius was utilized, is also shown in Table 3.6. Reported values for the

reaction radius for the reaction *OH + *OH -- H202 have varied between 1.2 A and 1.4

A [Buxton et al. 1988], and these radii can also be taken as values for the effective van

der Waals radius of the hydroxyl. The results show that the closest agreement between

%*OH attack and %SASA that the C(5') and C(4') hydrogens is found when the *OH is

modeled as a sphere of radius 1.2 A (percent errors of 4.2% and 2.5%, respectively). The

agreement at the C(3') and C(2') hydrogens is shown to be 11.5% and 8.8% for R.OH =

1.2 A, respectively, and to be 7.1% and 13.7% for R.OH = 1.4 A. The agreement between


























Table 3.5. Sensitivity of %*OH attack probabilities with variations in diffusion jump
time. %SASA is displayed for comparison purposes


%SASA


%oOH Attack
T=4ps

T=3ps

T=2ps


1H5'+2H5'


47.0



49.5

49.2

49.3


H4'


22.8



23.3

22.4

19.9


H3'


15.3



11.6

13.8

14.3


1H2'+2H2'


11.1



10.3

10.1

10.7


51.6 18.7 15.2


HI'


3.8


9.1 5.4


z=lps


















Table 3.6. Percent *OH attack to hydrogen atoms of sugar moiety with variations in the
effective radius, ROH, of the approaching hydroxyl. Also shown is the ratio of the % OOH
attack to the corresponding %SASA value for an equivalent spherical probe size.




% OH Attack


R.OH(A)

0.0

1.0

1.2

1.4

1.5


1H5'+2H5
31.4

47.0

49.0

50.9

51.9


H4'
13.3

20.6

22.2

22.5

22.8


H3' 1H2'+2H2'


14.3

14.2

13.6

13.0

12.5


25.1

12.2

10.1

9.0

8.4


HI'
15.9

6.0

5.1

4.7

4.4


Ratio of %eOH Attack to % SASA


SASA Probe Radius


1H5'+2H5


1.07

1.06

1.04

1.05


H4'
0.90

1.05

0.98

0.91


H3' 1H2'+2H2'


0.98

0.89

0.88

1.07


0.94

0.91

0.91

0.86


HI'
1.08

0.88

1.35

1.04


1.06 0.74 1.31


1.12 1.44









%OH attack and %SASA at the C(1') hydrogen is the relatively poor for R.OH = 1.2 A

(35% error), while it is very good for R.OH = 1.4 A (4.4% error). When considering the

approaching radical as a point, generally good agreement is seen at the C(3') to C(1')

hydrogens, although the agreement at C(5') and C(4'), which represent a larger total

percentage of all sugar attacks, is not as good as that for R.OH = 1.2 A. In the final model

of *OH attack, the effective hydroxyl radius was fixed at this latter value.


Sensitivity of *OH Attack Probabilities with Variations in R.OH + H

Two values have been reported in the literature for the diffusion coefficient for

the hydroxyl, 2.5 x 10- and 2.8 x 10- cm2 s1 Using the reaction rate constant 2.5 x 109

M1 s1 and Eq. 3.2, the corresponding reaction radii for *OH hydrogen abstraction

reactions, ROH+H, with deoxyribose C-H bonds are thus 1.32 A and 1.18 A, respectively.

Monte Carlo simulations were performed using both values of the reaction radius while

keeping the effective *OH radius constant at 1.2 A. Percent *OH attack probabilities

were noted to have changed less than 2% when the reaction radius was increased from

1.18 to 1.32 A. In the present study, a value of D.OH= 2.8 x 10-5 cm2 S-1 is assumed for

diffusion jump estimates of the approaching radical, and thus a reaction radius, ROH+H, of

1.18 A is utilized within the final Monte Carlo model.

With the completion of this sensitivity study, a final set of model parameters was

thus established. These parameters include ROH+H = 1.18 A, R.OH = 1.2 A, T = 3 ps, and

TSH = 0.5 ps.









Sensitivity of *OH Attack Probabilities with Considering or not Considering Steric
Hindrance

One of the central features of the present model is the explicit consideration of the

non-reacting atoms within the structure. These atoms or groups of atoms provide zones

of spatial exclusion through which approaching hydroxyl may not pass nor penetrate.

The Monte Carlo model, with its optimized set of reaction parameters, was subsequently

run twice: once with and once without steric hindrance via radical bounce-off from non-

reacting atoms in the molecule. In the latter case, the *OH were permitted to freely

penetrate the volumes occupied by non-reacting atoms and only the sugar reactions sites

(e.g., the seven deoxyribose hydrogen atoms) were considered as space-filled spheres

with radius ROH+H. As before, all base reactions were turned off so that comparisons of

site-specific sugar attack could be compared in confidence with %SASA values.

The results are shown in Figure 3.10 in which excellent agreement is seen

between the model predictions of site-specific radical attack and the solvent-accessible

surface area as expressed as a percentage basis of total sugar hydrogens. When the

presence of non-reacting atoms is ignored, relatively poor agreement is seen at all

hydrogen sites. The data show that %*OH attacks at C(5') and C(4') decreased by 15%

and 41%, respectively, with corresponding increases in the attack probabilities at C(3'),

C(2'), and C(1') (35%, 44%, and 99% increases, respectively). This study thus shows that

explicit treatment of steric hindrance is necessary in modeling the near-approach of

radiolytic free radicals to duplex DNA.



















60


5 50 %SASA

0 %OH Attack (w/ steric hindrance)
^ 40
40 %OH Attack (w/o steric hindrance)


0 30

o
S20


10


0
1H5'+2H5' H4' H3' 1H2'+2H2' HI'
Hydrogens of the Sugar Moiety


Figure 3.10. Hydroxyl attack probabilities calculated with (0) and without (*) steric
hindrance. Percent solvent accessible surface area for a probe radius of 1.2 A is shown
for comparison.









Model Predictions and Comparison to Experimental Data

Once the optimized set of reaction parameters and model assumptions was established,

the Monte Carlo simulation model for *OH attack was expanded to consider not only

hydrogen abstraction reactions at the sugars, but *OH adduct formation and H-

abstractions at the nucleobases. The percent *OH attack probabilities to the various sugar

hydrogens are shown in Figure 3.11 for model calculations in which sugar only or sugar

plus base reactions are considered. In both cases, the total numbers of sugar attacks are

normalized to 100%. Also plotted in Figure 3.11 are %SASA values (Rprobe = 1.2 A and

normalized only to the sugar hydrogens) and the experimental data on site-specific

cleavage rates by Balasubramanian et al. Several conclusions may be drawn from the

data of Figure 3.11. First, the inclusion of base reactions does not greatly influence the

%*OH sugar attacks at C(4') and at C(3'); however, inclusion of the base reactions

increases the percentage of attacks at C(5') with a corresponding drop in attacks at C(2')

and C(1') hydrogen atoms. Due to their proximity to the nucleobases, the latter carbon

sites in the sugar moiety must compete for approaching hydroxyl with interaction sites in

the nucleobases (see Table 3.2 for a comparison of reaction radii).

It has been shown that the hydroxyl reactions at C(5') and C(4') lead to unique

chemical products that can be experimentally detected [Balasubramanian et al. 1998;

Cadet et al. 1999; Scholes 1983; Scholes and Simic 1968; von Sonntag 1987]. These

chemical products are 5'-aldehyde and 3'-phosphoglycolate for hydroxyl abstraction

reactions at the C(5') and C(4'), respectively [Balasubramanian et al. 1998; von Sonntag

1987]. Several initial hydrogen abstraction reactions can, in principle, lead to production

of 3'-phosphate [Balasubramanian et al. 1998]. There is, however, little information as to






















- % Cleavage rate (Balasubramanian et al.)

0 % OH Attack (sugar reactions only)

* % OH Attack (both sugar and base reactions considered)

0 %SASA (HyperChem)


H4'
Hydrogens


H3' 1H2'+2H2'
of the Sugar Moiety


Figure 3.11. Normalized site-specific %*OH attack probabilities for sugar hydrogens as
calculated for the benchmark DNA model in which sugar only (o) or sugar plus base
reactions (*) are considered. Un-normalized site-specific percent DNA cleavage rates
from Balasubramanian et al. are shown, along with %SASA values from HyperChem
using a probe radius of 1.2 A.


1H5'+2H5'









which hydrogen abstraction reaction is responsible for this product [Balasubramanian et

al. 1998; von Sonntag 1987]. Balasubramanian et al. estimated fractional contributions

of each deoxyribose carbon site to the total DNA breaks by using the intrinsic isotope

effects measured for the C(5') and C(4') and apparent isotope effect for the other

deoxyribose carbon sites. Our calculated %*OH attack probabilities track closely with

the experimental values of percent DNA cleavage rates at the C(5') and C(4') carbons,

and to some extent at C(3'), as determined by Balasubramanian et al. Less agreement is

seen at the C(2') and C(1') carbons in the model calculations which permit competitive

base reactions. Several explanations might be proposed for this discrepancy. First, given

that several initial hydrogen abstraction reactions might lead to the production of 3'-

phosphate, an accurate estimate of the relative extent of abstraction of each deoxyribose

hydrogen atom is not possible. This situation holds true especially for the C(3'), C(2'),

and C(1') carbon sites for which there are no unique products that can be observed

experimentally after *OH attack. Secondly, the mean values of % cleavage rates

determined by Balasubramanian et al. do not sum to 100%, but rather to 120%. The

authors emphasized that this was not unexpected since three different nucleobases,

deoxyadenosine, deoxycytidine, and deoxythmidine were deuterated at a different

positions within their experimental 19-bp model. This 20% discrepancy is almost

exactly attributed to the differences between experimental values and model predictions

at the C(2') and C(1') sites, the two sites within the sugar moiety where experimental

measurements of product yields is the most difficult to quantify.









Conclusions

In this study, we have shown that both computational model predictions and

%SASA are very sensitive to the structural properties of the DNA molecule being

studied. This study suggests that DNA molecules generated with commercially available

software should be used cautiously for this purpose. DNA molecules should be created

as realistically as possible by using all the tools provided with these software packages,

including energy minimization. In addition, Monte Carlo model parameters and their

effect on the determination of *OH attack probabilities were discussed in detail. The use

of an effective *OH radius to simulate chemical reactions of *OH with both the Monte

Carlo computational models and the percent accessibility calculations was validated.

The relative *OH attack probabilities at the different deoxyribose hydrogen atoms

and their relationship to OH-accessible surface were explored. It is observed that the

%*OH attacks to the different hydrogens of the deoxyribose followed the trend dictated

by their percent accessibility to the *OH when only the sugar reactions were allowed in

our computational model. When both sugar and base reactions were allowed, the

agreement between the %SASA and %*OH attack was not as good as before. The reason

is that the reactive sites in the base moieties compete effectively with the hydrogens of

the deoxyribose. However, the results were in good agreement with the experimental

strand break formation data by Balasubramanian et al.

Many previous models have been based on the assumption that C(4') is the

preferred site for the *OH mediated DNA damage. However, this study has shown that

the *OH exhibit preferential attack toward the different deoxyribose hydrogens in the

order 1H(5')+2H(5') > H(4') > H(3') > 1H(2')+2H(2') > H(1'). The geometrically









optimized structure of our decamer duplex discloses that hydrogens of C(5') are located

on the outside of the DNA backbone, presenting the largest accessibility to diffusing

*OH. This information is in good agreement with our computational model predictions

and the experimental DNA strand break formation data by Balasubramanian et al.

Consequently, the latest experimental and computational data agree that the *OH reacts

with various deoxyribose hydrogen atoms in the order suggested by their accessibilities to

the *OH diffusing to the DNA. We are devoting current efforts to modeling site-specific

attacks to the various reacting atoms of the nucleobases, not only for the *OH, but also

for the Ho and hydrated electron as well.














CHAPTER 4
STRUCTURAL INFLUENCES OF SITE-SPECIFIC WATER RADICAL ATTACKS
TO THE BASE MOIETY



Introduction

Many of the mutagenic or lethal effects of ionizing radiation can be linked to

structural and chemical alterations of DNA. The latter are often explained by the

hydrogen abstraction by the hydroxyl or to a lower extent the hydrogen radical. In

addition, electron adducs may be formed via the interactions of the hydrated electron with

nucleobases. Most of the degradation products have been identified to arise from

reactions with hydroxyl [Douki et al. 1999]. Nevertheless, it is believed that thermalized

electrons (sub-ionization electrons) can also efficiently disassociate organic molecules in

either gas or the condensed phase and thus their biological role in radiation damage must

also be evaluated [Huels M. A. 1998]. However, the quantitative aspect of the formation

of most of the base lesions induced by water radicals remains to be established mostly

because of a lack of reliable methods of measurements. In addition, although we know

the final products from experiments, the detailed mechanistic descriptions of reactions

remains somewhat [Ferradini and Jay-Gerin 1993].

Although there is a collection of experimental studies of radiation-induced base

damage products, essentially no detailed computational models of based damage by water

radicals have been found in the literature. This study is performed to increase our

knowledge on the mechanistic description of base reactions. The most likely base sites









are identified and attack probabilities for each of these reaction sites are computed The

radiation chemistry of purines (adenine and guanine) is only poorly understood, whereas

a great deal of improvement has been made on both structural and mechanistic aspects of

water radical-induced degradation pathways of pyrimidines (thymine and cytosine) [von

Sonntag 1987]. Due to many problems associated with the determination of *OH-

mediated base damage to DNA, no reliable and complete data are available for double-

stranded DNA or short oligonucleotides [Cadet et al. 1999]. Detailed information

regarding to mechanistic aspects and measurement of base lesions may be found in the

work of Cadet et al. (1999).


Hydroxyl and Hydrogen Radicals

*OH is the most extensively studied water radical in regards to its reactions with

DNA. A wealth of information has been accumulated concerning the reaction of

hydroxyl with isolated DNA and model compounds [Cadet et al. 1999]. Because of its

electrophilic nature, the hydroxyl preferentially attacks the carbon atoms with the highest

electron density [von Sonntag 1987]. The hydrogen radical reaction mechanisms are the

same as those for the hydroxyl although at much lower rates of reaction.


Guanine

Quantification of base damages has shown that guanine is the preferential target

[Douki et al. 1999]. This generally is explained by its lower oxidation potential and

charge transfer phenomena [von Sonntag 1987]. There are two main degradation

pathways of the guanine moiety of isolated DNA and model compounds in aerated

solutions that have been identified. These are *OH addition to C4 (-60%) and to C8









(-25%) [Cadet et al. 1999]. The *OH additions to the C5 and C6 of the guanine moiety

are believed to account for the remaining 15% [von Sonntag 1987]. Detailed information

on the final products of purines and their characterization can be found in the work of

Cadet J. et al. [1999], and von Sonntag [1987].


Adenine

The information on the radiation chemistry of adenine is very little and not

complete, as data from experimental studies are not as clear as in the case of pyrimidines.

Several computational modelers, however, have extrapolated from the guanine data and

have modeled the *OH adduct reaction at the same reaction sites as in guanine with equal

attack probabilities [Tomita et al. 1994; Tomita et al. 1998].


Thymine

A comprehensive mechanism of *OH-mediated thymine degradation has been

proposed based on both the determination of the structural and redox properties of

thymine and the identification through pulse radiolysis of 18 final products formed in its

reactions with *OH. Based on these studies, the following pathways and their percent

contribution to total *OH-mediated thymine degradation are suggested Cadet et al.

[1999].


1. *OH addition to C-5 (-60% ): yields C-6 centered reducing radical pyrimidinee

radical)

2. *OH addition to C-6 (35%) : accompanied by the formation of the oxidizing

radicals pyrimidinee radical)









3. H abstraction from the methyl group (-5%):) : followed by the formation of

exocyclic radicals pyrimidinee radical)

All these initially produced pyrimidine radicals are, then, converted to peroxyl

radical as a result of a fast reaction with oxygen.


Cytosine

The *OH mediated degradation pathways of cytosine system is not so well

understood because of the lack of reliable experimental methods [von Sonntag 1987].

Nevertheless, studies have concluded that C5 and C6 are most susceptible sites for OOH

attack [Douki et al. 1996; von Sonntag 1987]. Final products studied in cytosine systems

conclude that *OH adduct reaction is more likely with C5 of cytosine than C6 [Douki et

al. 1996; von Sonntag 1987].


Hydrated Electron

There are two types of electrons, the 'dry', otherwise referred to as unthermalized

electron and the hydrated or thermalized electron. The dry electron is a secondary

electron with enough kinetic energy to further react with a water molecule through

ionization (>-10 eV) and can be treated as a negative point charge in its interactions with

water. On the other hand, the hydrated electron has a solvent cage (6-12 water

molecules)1 around it and it reacts through diffusion-controlled processes with molecules

and other chemical species. It is believed that the hydrated electrons interact with

nitrogen or carbon atoms on both purine and pyrimidine rings. Such interactions result in

transient radical anions on that particular base [Ferradini and Jay-Gerin 1993]. At this


1 Personal communication between B. Aydogan and Dr. S.G. Swarts (August 2000)









stage, after the formation of the transient anion, the hydrated electron is still mobile and

can leave this initial site and migrate through the base stack until it localizes at a

particular base site via an irreversible protonation of the transient anion. Migration

generally occurs faster than the irreversible protonation, thus several transient anions

might be formed during the migration of electrons. As such, mechanistic descriptions of

hydrated electron reactions with DNA remain somewhat uncertain [Ferradini and Jay-

Gerin 1993]. As a first approximation, the locations of only the formation of the first

transient anion can be considered as the site of an electron adduct reaction1.



Material and Methods

Base damage modeling is performed using the computational model presented in

Chapter 3. The reaction constant of each water radical is taken from Buxton et al. [1998].

Reaction constants are only available for free bases. In assigning reaction rate constant to

individual atomic sites, therefore, total reaction rate constants are used. Shown in Table

4.1 are the reaction constants and reaction radii for hydrogen, hydroxyl and hydrated

electron with the four bases and sugar moiety. The reaction radii are calculated using the

Smoluchowski equation (see Chapter 3). The water radical diffusion constants are

tabulated in Table 4.2 [Buxton et al. 1988]. The reaction constants of hydrogen with the

DNA constituents are much smaller than the hydroxyl and hydrated electron reaction

constants. For instance, the reaction constant of the *OH with the sugar moiety is almost

100 times more than that of hydrogen. On the other hand, the reaction constant of

hydrated electrons with the sugar moiety is considerably smaller. Both hydroxyl and

hydrated electrons are very reactive with nucleobases.

















Table 4.1. Reaction rate constants and reaction radii for water radicals with sugar moiety
and the four nucleobases [Buxton et al. 1988].



*OH


Moiety


Deoxyribose
Adenine
Cytosine
Guanine
Thymine


Moiety
Deoxyribose
Adenine
Cytosine
Guanine
Thymine


Reaction Rate Constant
(109 M-s-1)
2.50
6.10
6.10
9.20
6.40


Reaction Rate Constant
(109 M-s-1)
0.03
0.09
0.50
0.50
0.10


Reaction Radius
(A)
1.18
2.88
2.88
4.34
3.02


Reaction Radius
(A)
0.01
0.02
0.13
0.13
0.02


Hydrated electron


Moiety


Deoxyribose
Adenine
Cytosine
Guanine
Thymine


Reaction Rate Constant
(109 M-is1)
0.01
13.00
17.00
14.00
9.00


Reaction Radius
(A)
0.00
3.44
4.50
3.70
2.38









Table 4.2. Diffusion constants of the water radicals [Buxton et al. 1988].



Water Radical ( 2 1
(10-1 cm s )
*OH 2.8
H* 6.0
eaq 5.0





*OH and He Base Damage Modeling

In the present computational model, *OH adduct formation is considered at C(2), C(4),

C(5), and C(8) for both adenine and guanine, and C(5) and C(6) for both cytosine and

thymine [Cadet et al. 1999; von Sonntag 1987]. Preferences for abstraction at the

specific reaction sites are accommodated through steric hindrance by non-reacting atoms.

Hydrogen abstraction reaction by hydroxyl from the three hydrogen atoms of the methyl

group is also explicitly modeled. Details of *OH and H* base damage modeling are

presented in Chapter 3.


Hydrated Electron Base Damage Modeling

All nitrogen or carbon atoms of purine and pyrimidine ring are considered as the

potential sited of hydrated electron attack. Again, preferences for electron adduct at the

specific reaction sites are accommodated through steric hindrance by non-reacting atoms.

Shown in Table 4.3 are all the preferential sites for electron adduct reactions. It is now

known that a hydrated electron has a solvent cage consisting of six to twelve water

molecules. This makes the computational modeling of hydrated electron interactions

very difficult. There might be several possibilities for its mechanism of interaction: (1)

an hydrated electron might tunnel through the water molecules; if so, it could be









Table 4.3. Preferential electrons adduct sites in four nucleobases.


Preferential electrons adduct sites


Thymine


Cytosine


C5M*


Guanine


Adenine


*Carbon atom of Methyl group









considered as a point entity, or (2) it could be considered as a spherical volume whose

size is determined by the effective size of the water molecules around it. Keeping in

mind that a water molecule has an effective radius of 1.4A [Saenger 1984], the latter

approach would reduce the accessibility of any atom in a molecule by an approaching

hydrated electron considerably. In this study, the hydrated electron is modeled as a

spherical volume with an effective radius changing between 1 and 1.6 A in order to

determine the effect of size of the hydrated electron on the attack probabilities. Due to

edge effects explained in Chapter 3, only the central 8 bp segment of the 10-bp DNA

molecule is considered in the determination of water radical attack. The central 8-bp

segment of the DNA molecule consists of four A-T and four C-G pairs. Appendix C

presents the subroutine 'Interaction' that is used in this study to calculate hydrated

electron attack probabilities for individual reaction sites in four nucleobases.



Results


*OH Attack to Bases

Table 4.4 displays the distribution of *OH among the four nucleobases and

%SASA calculated when including and not including the methyl hydrogens in the

%SASA computations. In the former case, % SASA of the three hydrogens of methyl

group make up almost 65% of the total SASA. When the reaction radii of these

hydrogens are modeled as being equal to that of the thymine moiety, more than half of

the *OH attacks to nucleobases are calculated to occur with the hydrogens of the methyl

groups of the thymine moiety. Cadet et al. (1999) observed the percent contribution of

the methyl groups to the total *OH mediated thymine damage to be only 5%. Therefore,









Table 4.4. Distribution of *OH attack among the four nucleobases when RMH=RThymine
and RMH=1.2 A and %SASA calculated including and not including methyl
hydrogens.(MH=Methyl hydrogens of thymine)




%SASA(With the MH) %SASA(Without the MH)
Attack Site % *OH Attack (RMH=RThymine) % *OH Attack (RMH=1.2A)

% SASA % *OH Attack % SASA % *OH Attack
Guanine 15 27 40 51

Adenine 14 13 36 25

Cytosine 7 4 18 13

Thymine 64 56 5 11





reaction radii of methyl hydrogens, in this study, are modeled as being equal to the van

der Waals radius of hydrogen (1.2 A). This application decreased the number of

interactions with the hydrogens of the methyl groups considerably and the percent

attackto the thymine moiety is reduced to 11%. When the hydrogens of methyl group

are not included in the %SASA calculations, the % *OH attack and the %SASA are

observed to be in very good agreement. This could be a correct approach since the global

accessibility of these hydrogens are reduced by assigning them much smaller reaction

radii compare to the other reaction sites. No quantitative experimental data are found in

the literature to compare the calculated distribution of *OH attack to nucleobases. This

study is performed with a 10-bp DNA molecule that is used to verify the *OH attack

probabilities, at the individual sugar reaction sites, reported y Balasubramanian et al. (see

Chapter 3). Therefore, the results form this study is strictly valid for the specific 10-bp









DNA molecule used herein. Qualitatively, it is known that guanine is the most

susceptible among the four nucleobases to radiation damage [Cadet 1994; von Sonntag

1987]. This study concludes that 51% of all *OH attacks in base moieties occur within

the guanine.

Shown in Table 4.5 is the comparison of percent *OH attack and percent SASA

for the individual reaction sites for the central 8-bp segment of the 10-bp DNA molecule

defined in Chapter 3. Table 4.6 compares the calculated %*OH attack for individual

reaction sites to the experimental base damage data for radical attack to the entire free

base [Cadet et al. 1999; Douki et al. 1999; von Sonntag 1987]. The percent attack values

are calculated for each type of nucleobases individually, since the experimental data are

only available for free floating base. The comparison of *OH attack among the

individual reaction sites with %SASA showed fairly good agreements for all nucleobases

except guanine (see Table 4.5). The %SASA calculations indicate that no other sites but

C8 introduce accessible surface area to an approaching *OH. On the contrary, the

computational model predicts equal attack probabilities of 21% to C8 and C4 and 4% to

C4 and 5% to C6. As shown in Table 4.6 model predictions of%*OH attack are not in

good agreement with the experimental data. There might be several reasons for this

disagreement. First, the major source of disagreement might be the nature of the

experimental data. The only quantitative data available are for free floating bases. There

is still a paucity of data on the formation of base lesions within double stranded and

cellular DNA [Cadet et al. 1999]. Secondly, this model only predicts the *OH attack

probabilities, there might be some other chemical mechanisms to create a type of damage

that is not known yet or the final damage site or product might be different than the




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs