Molecular Modelling and NMR Studies of Multinuclear Platinum Anticancer Complexes

Transcription

1 Molecular Modelling and NMR Studies of Multinuclear Platinum Anticancer Complexes by Donald S. Thomas This thesis is presented for the degree of doctor of philosophy of The University of Western Australia School of Biomedical, Biomolecular and Chemical Sciences Perth, Australia August 2006

2 Declaration The work described in this thesis was performed between March 2001 and March 2005 in the School of Biomedical, Biomolecular and Chemical Sciences (formerly The Department of Chemistry) at the University of Western Australia. Unless otherwise stated, the experiments described were performed by the author. This work constitutes an original body of research that has not been submitted, either in whole or in part, for the purpose of obtaining any other degree. The following publications have been incorporated in part or whole within this thesis. Papers Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., Long Range 1,4 and 1,6-Interstrand Cross-Links Formed by a Trinuclear Platinum Complex. Minor Groove Preassociation Affects Kinetics and Mechanism of Cross-Link Formation as Well as Adduct Structure. Journal of the American Chemical Society 2004, 126, (7), Zhang, J.; Thomas, D. S.; Davies, M. S.; Berners-Price, S. J.; Farrell, N., Effects of Geometric Isomerism in Dinuclear Platinum Antitumor Complexes on Aquation Reactions in the Presence of Perchlorate, Acetate and Phosphate. JBIC, Journal of Biological Inorganic Chemistry 2005, 10, (6), Posters Thomas, D. S.; Davies, M. S.; Hegmans, A.; Berners-Price, S.; Farrell, N., Kinetic and Equilibria Studies of the Aquation of the Trinuclear Platinum Anticancer Agent BBR3464. (ICBIC 10) Journal of Inorganic Biochemistry 2001, 86, (1), Thomas, D. S.; Davies, M. S.; Zhang, J.; Berners-Price, S. J.; Farrell, N., NMR Studies Probing the Directionality of DNA Interstrand Cross-linking by the Trinuclear Platinum Complex BBR3464. (ICBIC 11) Journal of Inorganic Biochemistry 2003, 96, (1), 239. Thomas, D. S.; Berners-Price, S. J.; Davies, M. S.; Farrell, N., NMR Studies Probing the Directionality of DNA Interstrand Cross-linking by the Trinuclear Platinum Complex BBR3464. Australia and New Zealand Society for Magnetic Resonance Conference (ANZMAG 2004). Barossa Valley, Australia. February 15-19, The thesis is my own composition, all sources have been acknowledged and my contribution is clearly identified in the thesis. For any work in the thesis that has been co-published with other authors, I have the permission of all co-authors to include this work in my thesis, and there is a declaration to this effect in the front of the thesis, signed by me and also by my supervisor/s. Donald Thomas Prof. Sue Berners-Price ii

3 Abstract The trinuclear anti-cancer agent [(trans-pt(nh 3 ) 3 Cl) 2 {µ-trans-pt(nh 3 ) 2 (H 2 N(CH 2 ) 6 NH 2 ) 2 }] 4+ (BBR3464 or 1,0,1/t,t,t) is arguably the most significant development in the field of platinum anti-cancer agents since the discovery of cisplatin as a clinical agent more than 30 years ago. Professor Nicholas Farrell of Virginia Commonwealth University was responsible for the development of 1,0,1/t,t,t and an entire class of multinuclear platinum complexes. The paradigm shift that was required in the development of these compounds is based on a simple idea. In order to increase the functionality of platinum anti-cancer drugs a new way of binding to DNA must be employed. By increasing the number of platinum centres in the molecule and separating the binding sites, by locating them on the terminal platinum atoms, the result is a new binding motif that does not occur with cisplatin. The work described in this thesis involves the use of [ 1 H, 15 N] NMR spectroscopy combined with molecular modelling to investigate various aspects of the solution chemistry and DNA binding interactions of BBR3464 and the related dinuclear analogues [{trans-ptcl(nh 3 ) 2 } 2 (µ- NH 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/t,t) and [{cis-ptcl(nh 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/c,c). Chapter 2 contains detailed descriptions of the various methodologies used, including the molecular mechanics parameters that were developed for the various modelling studies described in this thesis. The initial parameter set that allowed the molecular modelling of 1,0,1/t,t,t or 1,1/t,t only included the chloro ligand as a possible substituent at the labile site. The addition of phosphate, acetate and aqua ligands along with the parameterization of the dinuclear complex 1,1/c,c, and its bridged phosphato and hydroxo derivatives, extended the useful range of complexes that could be modelled. The parameter set was also converted from the format used by HyperChem to the more up to date and useful format for use with Amber. In order to understand the way this new class of complexes functions questions must be asked about their basic solution behaviour. Chapter 3 describes a study of the aquation of two dinuclear platinum complexes in a variety of aqueous buffer conditions by [ 1 H, 15 N] HSQC spectroscopy. The complexes, 1,1/t,t and 1,1/c,c, differ only in the orientation of the labile chloro ligand on each of the platinum centres. The reactions of the 15 N-labelled complexes were followed in 15 mm perchlorate, acetate or phosphate solutions at 298 K using [ 1 H, 15 N] HSQC 2D NMR spectroscopy. Molecular models of three derivatives of the 1,1/c,c complex were constructed to aid in the assignment of peaks seen in the [ 1 H, 15 N] NMR spectra. All of the reactions are considered reversible and the rate and equilibrium constants for the initial aquation and the subsequent reactions with phosphate or acetate are reported. The rate constant for the first aquation step of 1,1/c,c (2.26 ± s -1 ) is half that of 1,1/t,t (4.77 ± s -1 ) while the anation rate constants are almost the same, resulting in the equilibrium favouring the iii

4 chloro form for the 1,1/c,c compound. The equilibrium constants for the first and second aquation steps are for 1,1/c,c pk 1 = 4.35 ± 0.03, pk 2 = 4.48 ± 0.19 and for 1,1/t,t pk 1 pk 2 = 3.96 ± A pk a value of 6.01 ± 0.03 was determined for the diaquated species [{cis- Pt(NH 3 ) 2 (H 2 O)} 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 4+ which is 0.4 units higher than that of the 1,1/t,t compound. The rate constants for the binding of acetate and phosphate to 1,1/t,t are similar, but the rate constant for the reverse reaction is close to 10-fold higher in the case of the phosphate reaction therefore equilibrium is reached more quickly (12 h as compared to 64 h). The rate constants for both forward and reverse reaction of 1,1/c,c with acetate and phosphate are quite similar resulting in a slow approach to equilibrium ( h) and a greater proportion of phosphate bound species are present. The reduced lability of the bound phosphate for 1,1/c,c is attributed to the formation of a macrochelate phosphate bridged species. The interaction of 1,0,1/t,t,t with DNA is of paramount importance. [ 1 H, 15 N] HSQC NMR has previously been used to study the kinetics of formation of long range 1,4 and 1,6 interstrand cross-links by the reaction of 15 N-1,0,1/t,t,t with 5 -{d(atatgtacatat) 2 } (14XL) and 5 - {d(tatgtatacata) 2 } (16XL) (Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, ). These experiments, conducted prior to my candidature, were designed to compare the kinetic profile of binding of 1,1/t,t and 1,0,1/t,t,t with the 14XL sequence and to investigate any differences between the formation of 1,4 and 1,6 interstrand cross-links by the longer 1,0,1/t,t,t. Chapter 4 describes the use of molecular modelling techniques to help rationalize the NMR evidence which indicated two conformations of the 1,4 interstrand cross-link formed by 1,0,1/t,t,t in the reaction with the 14XL duplex. Analysis of the initial 1 H spectrum obtained after the addition of 1,0,1/t,t,t to the 14XL and 16XL duplexes provided evidence of the localization of the central platinum atom of 1,0,1/t,t,t to the minor groove. In addition to providing a structural representation of the preassociated systems these models helped in the understanding of why the reaction with 16XL only resulted in a single product. The final products from the two reactions were purified by HPLC and subsequently verified to be cross-linked adducts of 1,0,1/t,t,t with either 14XL or 16XL by electrospray mass spectrometry. Chapter 5 describes a study of the interaction of 1,0,1/t,t,t with the duplex 5 d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 (dsgg) to investigate the possible preferential formation of different intra and interstrand cross-links. In other words does the DNA sequence or structure impose any preference on the final adduct formed in the reaction with 1,0,1/t,t,t, when compared to the dinuclear 1,1/t,t compound that had been studied previously (Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.-- Eur. J. 2003, 9, ). [ 1 H, 15 N] HSQC NMR spectroscopy experiments were first used to follow the kinetics of binding of the 15 N-labelled platinum complex to DNA. The rate constant iv

5 for the aquation step was 3.43 ± s -1, which is only slightly lower than that of 1,1/t,t with the same DNA duplex (4.00 ± s -1 ). The rate constant for formation of the monofunctional adduct (0.6 ± 0.1 M -1 s -1 ) was approximately half of the value for the reaction of 1,1/t,t with the same sequence. The rate constant for the closure to the bifunctional adduct (8.0 ± s -1 ) was two fold higher that of the 1,1/t,t reaction. Analysis of the 1 H NMR spectra of the final product showed that more than one adduct had formed and a detailed analysis of the initial 1 H spectrum indicated the existence of at least two preassociated states where the complex was electrostatically associated with the DNA duplex in the minor groove. Molecular modelling of these systems helped to demonstrate how these preassociated states could exist and lead to the multiple products seen in the reaction, which may indicate cross-links orientated in both the 3!3 and 5!5 directions. To further understand the role of preassociation in the mechanism of DNA binding by 1,0,1/t,t,t a series of in silico experiments were performed on three DNA duplexes, that had been previously employed by Viktor Brabec and co-workers in gel electrophoresis experiments that indicated a trend in the directionality of binding of 1,0,1/t,t,t based on the distance between the potential cross-linked guanine residues, (Kasparkova, J.; Zehnulova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2002, 277, ). The work described in Chapter 6 employed three duplexes; 5 -d(tctcctattcgcttatctctc)-3 5 d(gagagataagcgaataggaga)-3 (VB12), 5 -d(tctccttcttgttcttcctcc) d(ggattaagaacaagaaggaga)-3 (VB14) and 5 d(ctctctctattgttatctcttct)-3 5 -d(agaagagataactatagagagag)-3 (VB16). Two minor groove preassociated forms of 1,0,1/t,t,t with each duplex were created in which the complex was orientated in two different directions around the central guanine (labelled the 3!3 and 5!5 directions). The molecular dynamics simulations of these six systems indicated that each preassociated states was stable within the minor groove and could effectively support the formation of multiple interstrand cross-links. Subsequent investigations into the dynamic nature of the monofunctional adduct were conducted by the assembly of a single monofunctional adduct of the VB14 duplex with 1,0,1/t,t,t. Here it was found that the monofunctionally anchored 1,0,1/t,t,t adopted a position along the phosphate backbone of the duplex in the 5!5 direction. v

6 Acknowledgements The time I have spent working through my Ph.D. has been a most rewarding and enjoyable one. I would be remiss if I did not take some time to thank a large group of people without whom I would not have been able to get through the last few years no less actually finish. Possibly the most patient person during this whole process has been my wife. We both have been working towards our Ph.D.s at the same time. I believe that this fact saved us both since we understood the stress and strain that each was experiencing. Even with this I can t express how much Lea s love and support have meant to me during my candidature. A Ph.D. supervisor is possibly the critical element in any students progress through a postgraduate degree. I am truly grateful to have had the opportunity to work with Professor Sue Berners-Price. Her dedication and clarity of mind have been inspirational in many ways and at many times. A large number of people have helped me develop specific skills along this journey. When this long journey all started back at Griffith University in Brisbane I found a mentor and guide in Dr. Chris Brown whose help with Linux and Amber started me along this torturous path, Dr. Sue Boyd was an almost limitless source of information about nuclear magnetic resonance instrumentation be it Varian or Bruker initiated me into the sometimes arcane world of practical nuclear magnetic resonance spectroscopy. Finally but certainly not any less significant is Dr. Lindsay Byrne at the University of vi

7 Western Australia whose experience and knowledge about NMR will most likely dwarf my own for years to come. I would like to thank Professor Nicholas Farrell of Virginia Commonwealth University, without whose generosity and talent my entire project would not have existed. My time at Virginia Commonwealth University in his lab were productive and interesting. This was also when I was introduced to molecular modelling via John Cox, many thanks. Many people have come and gone through the Berners-Price group in my time. Murray Davies is with out a doubt the most significant of these people in terms of my work. Murray taught me the techniques that underpin the majority of my work. Joseph Moniodis who joined the Berners-Price lab shortly after our arrival at UWA has been a constant source of amusement and stimulation, the lab and the time would not have been nearly as much fun without him. In no particular order, except vaguely chronological: Anthony Humphries, Angela Ho, Junyong Zhang, Peter Barnard, Louise Wedlock and Scott McPhee have all in their own unique and individual way helped to make my time at UWA a wonderful experience. The support staff past and present in the Department of Chemistry at the University of Western Australia, too numerous to name. Whose help in those simple little things that suddenly aren t so simple when they stop working. Finally I would like to dedicate this thesis to the people who although not here to see me get to the end of this journey were those principally responsible for getting me here. In loving memory to my parents Robert and Shirley Thomas. vii

8 Table of Contents Declaration...ii Abstract...iii Acknowledgements... vi Table of Contents...viii List of Figures...xii List of Tables... xix Abbreviations... xx Abbreviations... xx 1 Introduction Background to Cisplatin Discovery, Efficacy, Side Effects Cisplatin: Mechanism of Action Aquation Getting to the DNA Binding to DNA Multinuclear Platinum Complexes: A New Paradigm Aquation Getting to the DNA Binding to DNA [ 1 H, 15 N] NMR Studies of Platinum Anticancer Drugs Molecular Modelling of Platinum-DNA Adducts Aims Aquation Studies Studies of the Preassociation Step in the Formation of 1,4- and 1,6- Interstrand Cross-Links by 1,0,1/t,t,t A Kinetic Study of the Competitive Formation of 5'-5' vs 3'-3' Crosslinks by 1,0,1/t,t Modelling Study of the Preassociation of 1,0,1/t,t,t with DNA References Experimental Methods and Conditions Nuclear Magnetic Resonance Techniques Molecular Modelling Background viii

9 2.2.1 Introduction Amber MOPAC General Modelling Procedures Parameterisation Charge Determination (MOPAC 2002) Introduction Method Results Molecular Dynamics Protocols Introduction Methods Analysis Visualization Rotation freedom at N7 of guanine Introduction Materials and Methods Results and Discussion Conclusions References Aquation of 1,1/t,t and 1,1/c,c in Various Buffered Environments Introduction Materials and Methods Aquation of 1,1/t,t and 1,1/c,c in Perchlorate Solution Aquation of 1,1/t,t and 1,1/c,c in Acetate Solution Aquation of 1,1/t,t and 1,1/c,c in Phosphate Solution pk a determination of 15 N-[{cis-Pt(H 2 O)(NH 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] Molecular Modelling Conclusions References Studies of 1,4- or 1,6-Interstrand Cross-links Formation by 1,0,1/t,t,t Introduction Materials and Methods Reaction of duplex 14XL with 15 N-1,0,1/t,t,t Reaction of duplex 16XL with 15 N-1,0,1/t,t,t ix

10 4.5 Discussion Conclusions References Binding of 1,0,1/t,t,t to a DNA Duplex Containing Multiple Guanine Binding Sites: A Competition Study Introduction Experimental Materials and Methods Results and Discussion Conclusions References Modelling Studies of 3'-3' vs. 5'-5' Interstrand Cross-linking Introduction Materials and Methods Sequence Generation Docking Equilibration Molecular Dynamics Production Molecular Dynamics Analysis Results Sequence Generation and Equilibration VB12 Sequence VB14 Sequence VB16 Sequence Monofunctional Adduct Production Molecular Dynamics Discussion Conclusions References A1 Chapter 1 Appendix A2 Chapter 2 Appendix Amber Parameter Sets frcmod.101ttt frcmod.11tt frcmod.11cc Acquisition Scripts x

11 2.2.1 Self submitting script for MD on APAC SC Docking Using the NMR Restraints Option in Sander Processing Scripts Renaming Equilibration Output Files Energy Output Analysis Geometry Output Analysis (Carnal) Geometry Output Analysis (ptraj) X3DNA Structural analysis script A3 Chapter 3 Appendix Scientist equation file for Model A fitting Scientist equation file for Model 2 fitting Scientist equation file for phosphate model fitting A4 Chapter 4 Appendix Selected Figures from Hegmans et al Mass Spectrometry of isolated Pt-DNA adducts A5 Chapter 5 Appendix Scientist input files xi

12 List of Figures Figure 1.1 Structural formula of cisplatin....2 Figure 1.2 Cisplatin analogues: carboplatin, nedaplatin and oxaliplatin...2 Figure 1.3 The speciation pathway of cisplatin including both aquation and hydrolysis steps...4 Figure 1.4 DNA binding sites of platinum drugs....7 Figure 1.5 Pathway for cisplatin binding to DNA ending in HMG domain protected bifunctional adduct...8 Figure 1.6 Structural formulae of 1,0,1/t,t,t, 1,1/t,t and 1,1/c,c...10 Figure 1.7 The scheme for the aquation/hydrolysis of multinuclear platinum complexes including the interactions with phosphate or other ligands Figure 2.1 The [ 1 H- 15 N] HSQC pulse sequence...26 Figure 2.2 Proton spectra of a platinated DNA oligonucleotide showing the Pt-NH 2 and -NH 3 regions...26 Figure 2.3 Diagram depicting the regions used in the HSQC spectra of cisplatin and multinuclear platinum complexes...27 Figure 2.4 Scheme depicting the pathway followed by 1,0,1/t,t,t from dissolution to the formation of a bifunctional adduct Figure 2.5 [ 1 H, 15 N] HSQC spectra from the reactions between a duplex DNA oligonucleotide and the multinuclear platinum complex 1,0,1/t,t,t...29 Figure 2.6 Watergate pulse sequence p3919 from a Bruker Avance 600 MHz spectrometer...29 Figure 2.7 Pulse sequence used in obtaining NOESY spectra on the Bruker Avance 600 MHz spectrometer...30 Figure 2.8 Molecular mechanics atom types assigned to 1,1/t,t, 1,0,1/t,t,t and 1,1/c,c...37 Figure 2.9 Proper and improper torsion angles assigned to 1,1/t,t, 1,0,1/t,t,t and 1,1/c,c...39 Figure 2.10 Atom types developed for phosphato, acetato and aqua ligands of both cis and trans platinum complexes...41 xii

13 Figure 2.11 Atom types assigned to the bridging µ-phosphate and µ-oh species of 1,1/c,c...43 Figure 2.12 Semi-empirical partial charges determined for 1,1/t,t using MOPAC Figure 2.13 The charge distribution of 1,1/t,t represented as a molecular surface Figure 2.14 Semi-empirical partial charges determined for 1,1/c,c using MOPAC Figure 2.15 The charge distribution of 1,1/c,c represented as a molecular surface...49 Figure 2.16 Semi-empirical partial charges determined using MOPAC 2002 for the complex 1,0,1/t,t,t...50 Figure 2.17 The charge distribution of 1,0,1/t,t,t represented as a molecular surface Figure 2.18 DNA nucleotide sequences used to generate the duplex structures studied in silico...54 Figure 2.19 Method used for the equilibration of oligonucleotide sequences for subsequent use in Amber MD simulations Figure 2.20 Structure for the C2'-endo and C3'-endo conformations of ribose...58 Figure 2.21 Syn and anti configurations of the guanine nucleoside...59 Figure 2.22 Pseudorotation wheel, used to describe the relationship between the pseudorotation angle and the sugar pucker of nucleosides...61 Figure 2.23 Example of initial conditions for torsion angle being studied...65 Figure 2.24 Final torsion angle dependency upon initial torsion angle for the G(7) or 5' system and the G(8) or 3' system...65 Figure 2.25 Two orientations of the platinum plane demonstrate an angular dependence upon the initial torsion angle while energetically indistinguishable...66 Figure 3.1 Two schemes representing the aquation/ligation pathways available to 1,1/t,t and 1,1/c,c in a variety of buffer solutions...70 Figure 3.2 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium perchlorate solution...75 xiii

14 Figure 3.3 Scheme representing the model used in the kinetic fit of the aquation of 1,1/t,t and 1,1/c,c in perchlorate solution...76 Figure 3.4 Time dependence plots of the species in the aquation of (a) 15 N- 1,1/t,t and (b) 15 N-1,1/c,c in 15 mm NaClO Figure 3.5 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium acetate solution...79 Figure 3.6 Scheme representing the aquation and ligand substitution (Model B) of 1,1/t,t and 1,1/c,c in acetate buffer...80 Figure 3.7 Plots of the time dependence of species in the aquation of (a) 15 N- 1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium acetate...81 Figure 3.8 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium phosphate solution Figure 3.9 Simplified scheme (Model B) used to model phosphate ligand substitution of both 1,1/t,t and 1,1/c,c...83 Figure 3.10 Plots of the time dependence of species in the aquation of (a) 15 N- 1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium phosphate solution, (Model B)...84 Figure 3.11 Plots of the time dependence of species in the aquation of 15 N- 1,1/c,c in 15 mm phosphate according to Model A...86 Figure 3.12 Plots showing the change in (a) 15 N and (b) 1 H chemical shifts with ph for Pt-NH 3 and Pt-NH 2 groups of the 1,1/c,c diaqua complex in 100 mm ClO Figure 3.13 Molecular model of the 1,1/c,c diaqua species suggesting possible hydrogen bonding between the two aqua ligands...91 Figure 3.14 Bridging hydroxide species observed in the 1,1/c,c aquation reaction in perchlorate solution...92 Figure 3.15 The aquaphosphato species of 1,1/c,c demonstrating the possible hydrogen bonding interaction between the aqua and phosphato ligands...93 Figure 3.16 The bridged 1,1/c,c phosphate species...94 Figure 4.1 The nucleotide sequence for the self complementary 12mer oligonucleotides, 14XL and 16XL...99 Figure 4.2 The NOESY spectrum of the 1,0,1/t,t,t-14XL DNA adduct xiv

15 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9 Figure 5.1 Figure 5.2 Figure 5.3 Figure 5.4 Figure 5.5 Figure 5.6 Figure 5.7 Figure 5.8 Figure 5.9 Aromatic region of the 1 H spectra from the 1,0,1/t,t,t-14XL DNA reaction Preparative scale separation of 1,0,1/t,t,t-14XL product mixture Molecular models depicting putative preassociation states of the 1,0,1/t,t,t-14XL adduct The NOESY spectrum of the 1,0,1/t,t,t-16XL DNA adduct H spectra of the aromatic region from the 1,0,1/t,t,t-16XL DNA reaction Preparative scale separation of the 1,0,1/t,t,t-16XL product mixture Molecular models depicting the putative preassociation state of the 1,0,1/t,t,t-14XL adduct The dsgg oligonucleotide duplex comprised of the complimentary single stranded oligonucleotides ssgg and sscc Scheme depicting the possible intermediates and products of 1,0,1/t,t,t in solution with dsgg A comparison of the imino region of the 1 H spectra (a) after and (b) before the addition of 1,0,1/t,t,t to the solution of duplex DNA The RMSD for the production MD simulation of (a) G(18) of 3'!3' and (b) G(25) 5'!5' Total energy plots from the preassociation production MD run of both (a) G(18) of 3'!3' and G(25) 5'!5' The 1,0,1/t,t,t complex aligned in the minor groove of dsgg in both the (a) 3'!3' and (b) 5'!5' D [ 1 H, 15 N] HSQC NMR spectra of 15 N-1,0,1/t,t,t during the reaction with the dsgg duplex at (a) 2.05 (b) 6.68 (c) and (d) hours. Peaks are assigned to the NH 3 and NH 2 groups of structures 1-5 as shown in Figure Kinetic fit of the combined model for 1,0,1/t,t,t forming bifunctional adducts with dsgg Comparison of the imino region of the 1 H spectra on completion of the reaction of (a) 1,0,1/t,t,t and (b) 1,1/t,t with dsgg xv

16 Figure 5.10 Figure 6.1 Figure 6.2 Figure 6.3 Figure 6.4 Figure 6.5 Figure 6.6 Figure 6.7 Figure 6.8 Figure 6.9 Figure 6.10 Figure 6.11 Figure 6.12 Comparison of the imino region of the 1 H spectra of the final products from the reaction of (a) 1,0,1/t,t,t and (b) 1,1/t,t with dsgg The VB14 DNA sequence depicting the 5'!5' and 3'!3' interstrand crosslinks possible The three "Viktor Brabec" sequences. The guanine residues highlighted in bold are those that took part in the interstrand crosslinking RMSD distance for the equilibration of (a) VB12 (b) VB14 and (c) VB16 sequences derived using ptraj Plots of the RMSD of distance for the pre-docking equilibration of VB12 with 1,0,1/t,t,t (a) in the 3'!3' direction and (b) in the 5'!5' direction derived from ptraj Total energy for pre-docking equilibration of the VB12 sequence (a) 3'!3' and (b) 5'!5' ,0,1/t,t,t docked in the minor groove of VB12 in the (a) 5'!5' orientation and (b) 3'!3' orientation. System prior to production molecular dynamics Plots of the RMSD of distance from the MD simulations of the VB12 DNA duplex with 1,0,1/t,t,t bound in (a) the 3'!3' orientation and (b) the 5'!5' direction Plots of the total energy sampled over the course of the production molecular dynamics simulation of the VB12 DNA sequence with 1,0,1/t,t,t bound in the minor groove in the (a) 3'!3' and (b) 5'!5' orientation Post molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB12 DNA duplex (a) in the 5'!5' orientation and (b) in the 3'!3' orientation Plots of the RMSD of distance for pre-docking equilibration of the VB14 DNA duplex in (a) the 3'!3' and (b) the 5'!5' orientations Plots of the total energy for pre-docking equilibration of VB14 sequences in the (a) 3'!3' and (b) in the 5'!5' orientations The docked species of VB14 in (a) the 3'!3' and (b) the 5'!5' orientations xvi

17 Figure 6.13 Figure 6.14 Figure 6.15 Figure 6.16 Figure 6.17 Figure 6.18 Figure 6.19 Figure 6.20 Figure 6.21 Figure 6.22 Figure 6.23 Figure 6.24 Figure 6.25 Figure A1.1 The RMSD distance plots for the production molecular dynamics run of the VB14 sequence in (a) the 3'!3' and (b) the 5'!5' system Total Energy for VB14 production molecular dynamics simulation of system seen in (a) the 3'!3' and (b) the 5'!5' system Post molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB14 DNA duplex (a) in the 5'!5' orientation and (b) in the 3'!3' orientation Plots of the RMSD of distance for the pre-docking equilibration of VB16 Brabec sequences of 2 preassociated states using ptraj Plots of the total energy for pre-docking equilibration of VB16 sequences in the (a) 3'!3' and (b) in the 5'!5' orientations The docked species of VB16 in (a) the 5'!5' and (b) the 3'!3' orientations Total Energy for VB16 production molecular dynamics simulation of system seen in (a) the 3'!3' and (b) the 5'!5' system The RMSD distance plots for the production molecular dynamics run of the VB16 sequence in (a) the 3'!3' and (b) the 5'!5' system Post molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB16 DNA duplex (a) in the 5'!5' orientation and (b) in the 3'!3' orientation Initial conformation of the 5' guanine N7 bound monofunctional adduct of VB14-1,0,1/t,t,t Plots of the RMSD of distance from the production molecular dynamics simulations of the monofunctionally bound 1,0,1/t,t,t adduct of VB Two structures that demonstrate the dramatic changes observed around 1600 ps of the molecular dynamics simulation The total energy plot for the production molecular dynamics simulation of 1,0,1/t,t,t bound monofunctionally to VB Monofunctional adduct of 1,1/t,t with dsgg sequence xvii

18 Figure A4.1 2D [ 1 H, 15 N] HSQC NMR (600 MHz) spectra at 298K of (a) duplex I and (b) duplex II after reaction with 15 N-1 for the times shown Figure A4.2 1 H NMR spectra (600 MHz) of the aromatic regions of (a) duplex I and (b) duplex II after reaction with 15 N-1 for between 0 and 47h Figure A4.3 1 H NMR spectra (600 MHz) of the imino regions of (a) duplex I and (b) duplex II after reaction with 15 N-1 for between 0 and 45 h Figure A4.4 Comparison of the 1 H spectra following the formation of (a) 1,4 - and (b) 1,6-interstrand cross-links by 15 N-1 showing the region of the CH 2 protons of the linker Figure A4.5 Plots of the relative concentration of species observed during formation of (a, b) 1,4- and (c) 1,6- interstrand cross-links by reaction of 15 N-1 with duplex I and II Figure A5.1 Contour plot of the 2D NOESY NMR spectrum of the dsgg duplex showing the H6/H8 aromatic ( ppm) to sugar ring H1 ( ppm) connectivities Figure A5.2 Contour plot of the 2D NOESY NMR spectrum of the dsgg duplex showing the H6/H8 aromatic ( ppm) to sugar ring H1 ( ppm) connectivities xviii

19 List of Tables Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 4.1 Table 5.1 Bond stretch parameters for the trans and cis multinuclear platinum complexes...37 Dihedral angle parameters for the trans and cis multinuclear platinum complexes...38 Torsion angle parameters for the trans multinuclear platinum complexes...39 Additional parameters required for aqua, acetato and phosphato ligands...42 Standard input parameters used when running MOPAC 2002 to optimize the geometry and determine the partial atomic charges of the platinum complexes studied Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm NaClO 4 in comparison to 1,0,1/t,t,t. a...77 Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm sodium acetate (at 298 K) according to the model shown as Figure Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm sodium phosphate at 298 K, according to the model shown in Figure Rate and equilibrium constants derived from the kinetic fit of the reaction 1,1/c,c and 1,0,1/t,t,t in phosphate buffer using Model A (Figure 3.1)...87 The gradient profile used for the elution of DNA adducts in the 14XL and 16XL experiments with 1,0,1/t,t,t Rate constants determined for the reaction of 1,0,1/t,t,t with the dsgg duplex according to the kinetic model shown in Figure xix

20 1,0,1/t,t,t 1,1/c,c Abbreviations [(trans-pt(nh 3 ) 3 Cl) 2 {µ-trans-pt(nh 3 ) 2 (H 2 N(CH 2 ) 6 NH 2 ) 2 }] 4+ (BBR3464) [{cis-ptcl(nh 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/c,c) 1,1/t,t [{trans-ptcl(nh 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/t,t) (BBR3005) 14XL 5 -{d(atatgtacatat) 2 } 16XL 5 -{d(tatgtatacata) 2 } 3DNA Nucleic Acid analysis software AM1 Austin Model 1 semi-empirical method Amber Assisted Model Building with Energy Refinement APAC Australian Partnership in Advanced Computing DFT Density functional theory DNA Deoxy-ribonucleic acid dsgg 5 -d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 ESP Electrostatic potential EXAFS HMG HPLC HSQC IVEC IXL MD MM MPICH nucgen nukit NOESY PDB perl PHP RESP RMSD Sander ssgg sscc TIP3 Extended X-ray absorption fine structure High mobility group protein High performance liquid chromatography Heteronuclear single quantum coherence Interactive Virtual Environment Centre Interstrand cross-link Molecular dynamics Molecular mechanic Message passing interface Subsidiary program of Amber used in creating DNA duplexes Subsidiary program of Amber used in creating DNA duplexes Nuclear Overhauser effect spectroscopy Protein Data Bank molecular file format Computer scripting language Computer scripting language Restrained electrostatic potential Root mean square deviation Energy calculation program in Amber Single stranded GG oligonucleotide Single stranded CC oligonucleotide Explicit water model used in Amber xx

21 UV/Vis VB12 VB14 VB16 VMD xleap Ultraviolet/Visible spectroscopy 5 -d(tctcctattcgcttatctctc)-3 5 -d(gagagataagcgaataggaga)-3 5 -d(tctccttcttgttcttcctcc)-3 5 -d(ggattaagaacaagaaggaga)-3 5 -d(ctctctctattgttatctcttct)-3 5 -d(agaagagataactatagagagag)-3 Visual Molecular Dynamics Subsidiary program of Amber used to generate input files xxi

22 1 Introduction 1.1 Background to Cisplatin Discovery, Efficacy, Side Effects Cisplatin (Figure 1.1) has been known since 1845 when Peyrone 1 first described the complex. However, until the accidental discovery by Rosenberg in 1965 its biological activity was unknown. 2, 3 Currently cisplatin is still in use as one of the most important anticancer agents available to the oncologist. The sales of cisplatin in 1998 were US $525 million 4 and carboplatin, the second generation analogue, had sales of US $775 million in Even when the substantial detractors of the therapy are considered cisplatin has been a very effective treatment especially for testicular cancer, with greater than 90% cure rate. 6 Additionally breast, ovarian, bladder, lung and head and neck cancers are also treatable with cisplatin. 7, 8 The three main drawbacks of cisplatin chemotherapy are its dose limiting toxicity, manifested particularly as nephrotoxicity, acquired resistance and intrinsic resistance. 9 It wasn't until clinical procedures were developed to diminish the overall toxicity of cisplatin that it was fully accepted as a chemotherapeutic agent. Diuresis has made the clinical application of cisplatin a reality but no method developed in the clinic will be able to surmount the more significant problem of cisplatin resistance, be it acquired or inherent. Many carcinomas do not respond to cisplatin, among these are colorectal and non small cell lung cancer, while ovarian and small cell lung cancer develop resistance to the drug. 10 1

23 Figure 1.1 Structural formula of cisplatin. Since the introduction of cisplatin into the market as an approved therapy an enormous amount of research time and money has been dedicated to developing more potent, less toxic, more specific, or orally-active variants of cisplatin that expand the activity profile of the parent complex. Unfortunately very few of the compounds that progressed to clinical trials have seen the light of day in the clinic. The most notable exception to this is carboplatin (Figure 1.2), which was introduced in 1986 and does exhibit reduced toxicity while maintaining the antitumour activity of cisplatin at much higher doses. However, carboplatin still acts on the same set of carcinomas and acquired resistance is still its limiting factor. 11 O O O Pt Pt O O O O carboplatin H 2 N O nedaplatin O Pt N H 2 O O oxaliplatin Figure 1.2 Cisplatin analogues: carboplatin, nedaplatin and oxaliplatin. 2

24 The basic difficulty with the nature of cisplatin research and drug development is embodied in the structure of the drug itself and as a consequence its mode of action. Historically, the development of new platinum complexes for clinical application employed the same structure-function paradigm that cisplatin has exploited for the last thirty years. Figure 1.2 demonstrates that the classical configuration of two am(m)ine groups in a cis arrangement around the platinum square plane is still the chosen form. The basic features of two cis leaving groups located trans to two am(m)ines is almost never altered. When combined with the requirement of complex charge neutrality the cisplatin dogma is complete. The resultant changes observed with modifications to cisplatin tend only to impact around the edges of platinum based drug functionality. That is to say reduced toxicity, slower/faster reactivity or uptake but only within a small range. Currently only three classic analogues have found their way into the clinic; carboplatin, 12 oxaliplatin and nedaplatin as shown in Figure 1.2. Nedaplatin 7 has only gained regulatory approval in Japan while oxaliplatin has only recently been approval for clinical use in France (1996), China (1998), the EU (1999) and in the USA (2002). 8,13 In order to obtain a new range of activity a new paradigm must be established. New functionality or activity can not be achieved with in this classical framework. To set the stage a brief review of the state of knowledge of the mode of action of cisplatin is required. The following sections will progress, as cisplatin does, from dissolution in aqueous media to transport through the media and to interaction ultimately with DNA. Following that the a new class of multinuclear platinum complexes will be introduced. 3

25 1.2 Cisplatin: Mechanism of Action Aquation It is imperative to gain a fundamental knowledge of the processes by which cisplatin interacts within the blood stream and within cells. The aquation profile of cisplatin has been extensively studied and the scheme detailing the various stages that cisplatin goes through when dissolved in aqueous solution is show in Figure Without this knowledge it would be impossible to develop a new approach to platinum based anticancer agents. The stages which cisplatin progresses through on its journey to the DNA of a tumour cell can be summarized by two major processes, aquation/ligation and transport. Transport occurs both in the venous system and within the cell. These environments are generally very different from each other and the cellular environment H 3 N Cl Pt NH 3 Cl K 1 1+ K Cl + H 2 O k 1 H 3 N NH 3 - Cl + H H 3 N NH 2 O k 2 3 Pt Pt k -1 k Cl OH -2 2 H 2 O OH 2 K a1 - H+ K a2 - H+ H 3 N Cl Pt NH 3 H 3 N NH 3 Pt OH H 2 O OH 1+ K a3 - H+ H 3 N HO Pt NH 3 OH Figure 1.3 The speciation pathway of cisplatin including both aquation and hydrolysis steps. 4

26 will vary greatly depending upon the cell type and its health. Upon entering the cancer cell cisplatin undergoes hydrolysis, one or both of the chloride ligands are replaced with an aqua ligand. The aquation step can occur at any time during the journey to the cell but is usually inhibited by the relatively high chloride concentration in blood plasma. 26 This is crucial since the aquated cisplatin is the active species 14 and persistence of the non-aquated species in the blood stream reduces unwanted side reactions albeit not completely Getting to the DNA The administration of cisplatin is via intravenous injection. The exact mechanisms by which entry into the cell is achieved is not known but both passive diffusion and active transport mechanisms have been proposed. 27 One school of thought suggests that a passive diffusion mechanism is the primary means by which cisplatin enters the cell. The limiting factor for the cellular uptake of cisplatin is its concentration and the process does not appear to have a saturation point It has been shown that the uptake of cisplatin is not inhibited by carboplatin or other structural analogues and also 32, 33 does not have a ph optimum. A new concept for understanding the uptake of platinum complexes into the cell has been put forward by Howell. A mechanism by which the platinum s passage into the cell is mediated by the copper influx transporter, human copper transport protein 1 (hctr1) has been proposed One significant factor in the efficacy of cisplatin, or any analogue, as a chemotherapy is the success the drug has in actually reaching its target, DNA. Glutathione (γ-l-glu-l- Cys-Gly) is a sulphur containing tripeptide that in the case of cisplatin can substitute for a chloro or aqua ligand. A number of glutathione containing complexes have been 5

27 observed, including a four-membered ring system, Pt 2 S 2, found as a product of the reaction of GSH with the cis-[pt(nh 3 ) 2 (H 2 O) 2 ] Monitoring of the direct reaction of 15 N-cisplatin with GSH using 15 N-{ 1 H} DEPT NMR revealed a mono-substituted sulphur containing complex at the early stages of the reaction. 39 A number of other complexes were evident during the reaction however, more interestingly 15 NH 3 was observed within 15 minutes of the start of the reaction and the final product did not contain 15 NH 3. It is thought that glutathione reacts with platinum drugs resulting in their deactivation. 40 The level of glutathione in cisplatin-resistant tumour cells has been found to be higher than in cisplatin sensitive cells. Additionally an active transport mechanism, the GS-X pump, acts to facilitate the removal of platinum drugs from within cells. 41 Interestingly, the concomitant administration of cisplatin and glutathione does not act to reduce the activity of cisplatin. 42 This observation does beg the question of how cisplatin actually gets to the target cell in an active form. The displacement of glutathione by nucleotides is theoretically possible even though the platinum sulphur bond is quite strong. In a competition experiment with methionine, histidine and 5 monophosphate nucleotides it was observed that the initial methionine bound Pt was displaced by 5 -GMP bound though the N7 atom. 43 The reaction of methionine with cisplatin is not completely analogous with glutathione as the reaction is faster and the reactions of thiols are generally not reversible. 44 6

28 1.2.3 Binding to DNA Once the cisplatin has been delivered to the cell and has undergone at least one aquation step it is able to interact with DNA. Figure 1.4 shows the principle binding site of cisplatin on DNA. The N7 positions of guanine and to a lesser extent adenine is the most electron rich and accessible regions on the DNA duplex. Both of these sites are located in the broad flat bottom of the major groove in B form DNA. H 2 N OH N N Adenine O - P O O O N N NH 2 O N Cytosine O - P O O N O O O N7 binding site O 7 N NH Guanine O - P O O O N N NH 2 O O - O P O HN CH 3 Thymine O O O N OH Figure 1.4 DNA binding sites of platinum drugs. Cisplatin forms preferentially GG or AG intrastrand cross-links 45, 46 and these lesions cause a pronounced kink to develop in the DNA helix of approximately at the binding site It has been postulated that this structural motif is recognized by high mobility group (HMG) domains of proteins and in addition to shielding the DNA from 7

29 excision repair mechanisms this protein binding also exaggerates the bending of the 51, 52 DNA to 86. H 3 N H 3 N Pt Cl Cl H 3 N H 3 N Pt OH 2 Cl H 3 N H 3 N Pt Cl H 3 N H 3 N Pt H 3 N H 3 N Pt Figure 1.5 Pathway for cisplatin binding to DNA ending in HMG domain protected bifunctional adduct. Figure 1.5 is a cartoon describing all of the stages that cisplatin passes through from injection to the binding of HMG1 protein to the 1,2 intrastrand cross-link. What is not demonstrated in Figure 1.5 is the biological outcomes that are proposed to accompany the binding of cisplatin to DNA. Two schools of thought exist on the importance of HMG binding to cisplatin-dna lesions. The first proposal suggests that the binding of HMG protein prevents the nucleotide excision repair mechanisms from functioning by 8

30 shielding the lesion. Thus the DNA can no longer replicate or transcribe which in turn prompts the cell to enter apoptosis or programmed cell death. The second idea concerning the role of HMG in cell death brought about by cisplatin binding to DNA centres around the native function that HMG proteins play in the body. Since the HMG proteins function natively as transcription factors it has been suggested that the hijacking of these proteins away from their predefined role sends the cell into apoptosis. 1.3 Multinuclear Platinum Complexes: A New Paradigm For new levels of activity to be realized from platinum based anticancer agents a new starting point was needed to direct development. Professor Nick Farrell of Virginia Commonwealth University in Richmond Virginia has created that new starting point and generated a series of multinuclear platinum based complexes that demonstrate unique and improved activity over cisplatin and cisplatin analogues Multinuclear platinum complexes make up a unique class of antitumour agents. In both structure and the biological effects observed after their application they differ markedly from cisplatin and the related mononuclear analogues developed over the last thirty years. Figure 1.6 shows the structure of three representative multinuclear platinum complexes. A number of structural motifs which distinguish these complexes from cisplatin are apparent. Each platinum atom (or each terminal platinum atom in the case of the trinuclear complex 1,0,1/t,t,t), has only a single labile ligand. The arrangement of this labile ligand is variable. The platinum coordination units are connected by an alkyl carbon chain. This linking chain can be varied in length and in the trinuclear examples the central linker platinum can be replaced by other groups such a spermine or 56, 57 spermidine. 9

31 Figure 1.6 Structural formulae of 1,0,1/t,t,t (BBR3464), 1,1/t,t and 1,1/c,c. These differences constitute a new paradigm for platinum anticancer chemotherapeutic agents. By changing the nature of the lesion on DNA the mechanism of cytotoxicity is necessarily going to change. This is supported by the fact that the range of cancers that have shown potential activity are distinctly different to those that cisplatin is effective against In addition it has been demonstrated that these complexes can be administered at much lower doses than cisplatin and have a broader profile of action 61, 62 including activity in cisplatin resistant cell lines. The most promising complex in this series is the trinuclear complex [(trans-pt(nh 3 ) 3 Cl) 2 {µ-trans- Pt(NH 3 ) 2 (H 2 N(CH 2 ) 6 NH 2 ) 2 }] 4+ (1,0,1/t,t,t or BBR3464) which has been investigated in Phase II clinical trials in small-cell lung, ovarian and gastric cancers. The effective does of 1,0,1/t,t,t is markedly lower than that of cisplatin, 1 mg/m 2 as opposed to mg/m As with cisplatin the path of a multinuclear platinum complex from injection to target involves transport and aquation/ligation. The following sections detail these steps and the work that has been done to date. 10

32 1.3.1 Aquation The aquation of both the trinuclear 1,0,1/t,t,t 64 and the dinuclear [{trans- PtCl(NH 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/t,t) 65 have been investigated using 2D [ 1 H, 15 N] HSQC spectroscopy (Figure 1.7). As with cisplatin the labile chloro ligands dissociate and are replaced with an aqua ligand. The physical separation of the two ends of the multinuclear complex are sufficiently large that each chloro ligand undergoes aquation independently of the other. Cl/Cl k -1 k 1 Cl - H 2 O Cl/HO H + K a1 Cl/H 2 O L H 2 O Cl/L k -2 k 2 Cl - H 2 O Cl - H 2 O HO/H 2 O H + K a2 H 2 O/H 2 O L H 2 O H 2 O/L - H 3 O + H 3 O µ-po 4 HO/HO H + K a3 k -7 k 7 + H 3 O - H3O µ-oh H 2 O L L/L Figure 1.7 The scheme for the aquation/hydrolysis of multinuclear platinum complexes including the interactions with phosphate or other ligands. The aquation rate constants of 1,0,1/t,t,t and 1,1/t,t are very similar, however the rate constant for the anation reaction was found to be less in the case of 1,0,1/t,t,t. 64 As a 11

33 consequence the proportion of aquated trinuclear complex at equilibrium is significantly greater. Hydrolysis of the aqua ligand is possible, the equilibria governing all possible acid dissociation steps are assumed to be the same. The pk a values of the aqua ligands of 1,1/t,t and 1,0,1/t,t,t have been measured in 100 mm NaClO 4 solution and were found to be identical (5.62). 64, 65 This value is much lower than for the cisplatin monoaqua complex (6.41) 23 and under physiological conditions the aquated species will be largely in the less reactive hydroxo form. The interactions of 1,0,1/t,t,t in a 15 mm phosphate buffer solution were studied using [ 1 H, 15 N] HSQC NMR. The experiments showed that the initial aquation reaction was not slower than in the absence of a competing ligand. 64 However, reversible reactions were seen between the aquated species and chlorophosphato or aquaphosphato species. The equilibrium conditions in the presence of a phosphate buffer were attained more rapidly than in a comparable cisplatin system Getting to the DNA Clinical trials of 1,0,1/t,t,t have been undertaken and to date Phase II trials have been completed. It was found that the interaction of 1,0,1/t,t,t within the blood plasma or with glutathione resulted in the ultimate degradation of the drug. The mechanism by which this occurs involves the binding of the glutathione through the sulphur to the terminal platinum atoms, replacing the chloro or aqua groups. This in itself would not be a real problem since the complex is still intact. However, the strong trans influence of the coordinated sulphur results in liberation of the alkyl ammine linker. The [{cis- 12

34 PtCl(NH 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ (1,1/c,c) complex with the change in orientation of the labile group has been brought forward as an alternative to the 1,0,1/t,t,t complex to overcome this problem Binding to DNA The DNA binding profile of multinuclear platinum complexes has been extensively studied in particular by Brabec and co-workers The sequence selectivity observed for cisplatin is not altered substantially in that the multinuclear platinum complexes principally bind to guanine, although adenine binding seems less pronounced than for 45, 46 cisplatin. However, there are several major differences between the DNA adducts formed by cisplatin and multinuclear platinum complexes. The first is that the di- and trinuclear complexes are able to form long range cross-links (both inter- and intrastrand). These interstrand cross-links can bridge up to four base pairs in the case of 1,1/t,t 71 and six base pairs in the case of 1,0,1/t,t,t. 72 These long range cross-links do not cause any major kinking in the DNA helix unlike the directed bend of the cisplatin intrastrand cross-link , 74 and thus the adducts are not recognized by HMG proteins. Finally, the 1,0,1/t,t,t complex has shown the unusual ability to form DNA interstrand cross-links in both the 5'!5' and the 3'!3' direction. Viktor Brabec has conducted a series of gel electrophoresis experiments comparing the directional nature of the bifunctional adduct after the monofunctional adduct was formed from single stranded DNA. 67 The following three DNA duplexes were used and vary in the length of the interstrand cross-link that can be formed in both 5'!5' and the 3'!3' directions: 5 -d(tctcctattcgcttatctctc) Figure 6.2 in Chapter 6 depicts the naming convention used when referring to the direction of bifunctional adduct formation. 13

35 d(gagagataagcgaataggaga)-3 (VB12), 5 -d(tctccttcttgttcttcctcc) d(ggattaagaacaagaaggaga)-3 (VB14) and 5 d(ctctctctattgttatctcttct)-3 5 -d(agaagagataactatagagagag)-3 (VB16). Increasing the length of the interstrand cross-link has the effect of changing the direction of the bifunctional adduct formed. The VB12 duplex forms almost exclusively 3'!3' adducts while the VB16 duplex forms adducts in the 5'!5' direction 70% of the time. 1.4 [ 1 H, 15 N] NMR Studies of Platinum Anticancer Drugs The first report on the use of [ 1 H, 15 N] HSQC NMR spectroscopy to study the chemistry of cisplatin was published in The [ 1 H, 15 N] HSQC NMR technique has three distinct advantages to the study of platinum based anticancer agents. First, the ability to detect 15 N labelled complexes at concentration down to 5 µm, a level that is relevant to physiological environment in which the anticancer agents will be employed. The second advantage is that the HSQC technique allows for the monitoring of reactions of 15 N labelled platinum am(m)ine complexes with biological molecules such as proteins or DNA. The success of this technique relies on knowing the one-bond 15 N- 1 H coupling constant of the 15 NH 3 spin system. In the case of cisplatin the 1 J( 15 N- 1 H) coupling constant of 73 Hz is optimized for the HSQC experiment. Finally, the HSQC technique also allows for the discrimination between substituted species and thus the determination of rate information for each step of the reaction. Over the past decade [ 1 H, 15 N] NMR spectroscopy has been applied to a wide range of studies of platinum anticancer drugs and these applications have recently been reviewed. 75 In particular, the [ 1 H, 15 N] HSQC NMR technique has been very 14

36 successfully applied to the problem of understanding the aquation of platinum based anticancer drugs. Cisplatin 23 aquation has been extensively studied along with a 76, 77 number of mononuclear analogues. Also of relevance to the work described in this thesis are [ 1 H, 15 N] HSQC NMR studies of the kinetics of the reaction of 15 N-cisplatin with a number of oligonucleotides. The decamer 5 -d(acatggtaca) was initially investigated. 78 Subsequently the duplex 5 -d(atacatggtacata)-3 5 d(tatgtaccatgtat)-3 (dsgg) was studied with 15 N-cisplatin 79 and cis- [Pt( 15 NH 3 ) 2 (H 2 O) 2 ] 2+80 This dsgg duplex has been extensively studied within the Berners-Price group and beyond. The solution structure has been elucidated by Parkinson et al. 81 Prior to the start of my candidature [ 1 H, 15 N] HSQC NMR spectroscopy had been successfully applied in a number of studies of di- and trinuclear platinum complexes in a collaboration between the Berners-Price and Farrell groups. The addition of an 15 N label to both the 15 NH 3 or 15 NH 2 groups of multinuclear platinum complexes allows for the easy discrimination of individual components of the reactions of these platinum anticancer agents as the 15 NH 3 and 15 NH 2 regions of the [ 1 H, 15 N] HSQC NMR spectra can be monitored independently. As discussed above (Section 1.3.1) the aquation profile and pk a of the dinuclear platinum complex 1,1/t,t were determined by [ 1 H, 15 N] NMR spectroscopy 65 The aquation profile of 1,0,1/t,t,t has been analysed as well as the interactions that occur when the complex is in the presence of a phosphate buffer. 64, My contribution to this published work was made prior to the start of my candidature and is not included in this thesis. 15

37 The DNA binding interactions of the dinuclear 1,1/t,t complex have been thoroughly studied by [ 1 H, 15 N] HSQC NMR spectroscopy. The first study involved a complete kinetic analysis of the stepwise formation of a 1,4- interstrand cross-link by monitoring the reaction of 15 N-1,1/t,t with the DNA oligonucleotide 5 -{d(atatgtacatat) 2 } (14XL). 71 Competition between formation of intra- or inter-strand cross-links was investigated in a study 82 of 15 N-1,1/t,t with the dsgg duplex (5 d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 ). The sequence offers three possible bifunctional modes for the 1,1/t,t complex: a 1,2-intrastrand adduct and two 1,4-interstrand adducts that differ in the direction of the cross-link (5-5 versus 3-3. The final product was identified as almost exclusively the 5-5 1,4-G(8)G(18) interstrand cross-link which exists as two conformational forms, which differ in the orientation of the bound G(18) residue. This study was followed by a [ 1 H, 15 N] NMR study of the reaction of 15 N-1,1/t,t with the single strand 5 -d(atacatggtacata) , 82, 83 The results showed that 1,1/t,t reacts much faster with the single strand than the duplex as a consequence of a greater slowing of the aquation of the complex in the presence of the duplex. 1.5 Molecular Modelling of Platinum-DNA Adducts Molecular modelling of biological macromolecules has been done for a number of years now. Molecular mechanics/dynamics software packages have long been the first line of attack when it comes to the investigation of these large macromolecules. Computer simulations require significant computer resources in both technology and CPU time. The level of simulation can be currently be assigned into three categories; molecular mechanics/dynamics, semi-empirical and ab initio or DFT. Progressing from one level My contribution to this published work in terms of molecular modelling was made prior to the start of my candidature. Details are included in Appendix A1 Figure A

38 to the next increases the level of theory or complexity with which the simulation is treated, additionally the computer resources required to run the simulation also increase. An inverse relationship exists with the number of parameters required to describe each system. While the molecular mechanics simulation may require a very large number of parameters to run, the ab initio simulation requires very few, only the selection of a basis set. Most molecular mechanics software packages are limited to organic compounds, that is to say carbon, hydrogen, nitrogen, sulphur and phosphorous atoms are recognized. Obviously the modelling of cisplatin or any other platinum based complex will require significant alteration to the generic parameters. The size of the molecular system and the relative ease with which parameters can be added was the principal reasons for using the Amber molecular mechanics/dynamics software package. The parameters used that allowed the modelling of platinum contain complexes along side of biological molecules were first developed by Yao et al. 84 These parameters although limited to cisplatin were sufficient for John Cox at Virginia Commonwealth University to use as the basis for the parameterization of 1,0,1/t,t,t, 1,1/t,t and a variety of other multinuclear platinum complexes. The problem of assigning partial atomic charges has not been adequately resolved in molecular mechanics simulations. This is especially the case when considering 3 rd row transitions metals such as platinum with 4f electrons. The charge distribution within these molecules is not necessarily a simple redistribution of the formal charge on the molecule. Advances in semi-empirical and more recently density functional theory 17

39 calculations have allowed for a more comprehensive and hopefully accurate assessment of the partial atomic charges. 1.6 Aims The aims of the work described in this thesis were to use a combination of [ 1 H, 15 N] NMR spectroscopy and molecular modeling to study aspects of the solution chemistry and DNA binding interactions of the clinical candidate 1,0,1/t,t,t and the related dinuclear compounds 1,1/t,t and 1,1/c,c Aquation Studies The previous [ 1 H, 15 N] NMR aquation study of 1,1/t,t 65 was extended in this work to study the aquation profile of 1,1/t,t in phosphate and acetate buffers. These simple buffer solutions are proposed as simple models to investigate possible transport mechanisms through membranes. During the course of the project the complex 1,1/c,c was introduced as a back-up clinical candidate and so the aquation studies were expanded to involve this drug as well. These studies are described in Chapter Studies of the Preassociation Step in the Formation of 1,4- and 1,6- Interstrand Cross-Links by 1,0,1/t,t,t At the start of my candidature [ 1 H, 15 N] NMR experiments had been performed (but not fully analysed) to follow the stepwise formation of 1,4- and 1,6-interstrand cross-links by 1,0,1/t,t,t. The experiments followed the reaction of 15 N-1,0,1/t,t,t with the DNA oligonucleotides 5 -{d(atatgtacatat) 2 } (14XL) and 5-18

40 {d(tatgtatacata) 2 } (16XL). These two sequences demonstrated markedly different characteristics in binding 1,0,1/t,t,t. In the case of 14XL the NMR data provided evidence for two distinct conformers of the interstrand cross-link, whereas there was evidence for only one conformer of the 1,6-interstrand crosslink. Does the preassociation of 1,0,1/t,t,t play a role in dictating the nature of the bifunctional interstrand cross-link? My role in this study was to use a combination of 1 H NMR and molecular modelling to study the initial electrostatic preassociation of the 1,0,1/t,t,t complex with the 14XL and 16XL oligonucleotide sequences and to separate the final products of the reactions by HPLC for characterisation by ESI-MS spectroscopy. These studies are described in Chapter A Kinetic Study of the Competitive Formation of 5'-5' vs 3'-3' Crosslinks by 1,0,1/t,t [ 1 H, 15 N] NMR was used to follow the binding of 15 N-1,0,1/t,t,t to the 14-mer duplex (5 -d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 ) (dsgg). This sequence offers the opportunity to investigate the potential formation of 1,4- and 1,5- interstrand cross-links in both 3'!3' and 5'!5' directions by 1,0,1/t,t,t. The experiment was carried out under identical conditions to the previous reaction of this sequence with 1,1/t,t 71 which formed exclusively the 1,4-GG crosslink in the 5'!5' direction in accordance with the findings of molecular biology experiments. The results of this study are described in Chapter 5. 19

41 1.6.4 Modelling Study of the Preassociation of 1,0,1/t,t,t with DNA The role of the central linker in preassociating with DNA is significant and during the course of this project it was determined that a molecular modelling study would be useful in testing the hypothesis that the direction of cross link formation is dictated by this preassociation. To this end the dsgg sequence and the three nucleotide sequences employed by Viktor Brabec in molecular biology experiments were modelled with the central linker of 1,0,1/t,t,t docked into the minor groove. The models of the dsgg sequence are included in Chapter 5 and the modelling study of the Brabec sequences (VB12, VB14 and VB16) are described in Chapter 6. 20

42 1.7 References 1. Peyrone, M., Ann. Chem. Pharm. 1845, Rosenberg, B.; VanCamp, L.; Krigas, T., Nature 1965, 205, Rosenberg, B.; VanCamp, L.; Trosko, J. E.; Mansour, V. H., Nature 1969, 222, Siafaca, K., In Oncology Trends Products Markets - Part I, Future Oncology, New Medicine Inc.: Lake Forest, CA, 1999; pp American Pharmaceutical Partners, I., American Pharmaceutical Partners Launches Carboplatin for Injection, USPAmerican Pharmaceutical Partners Launches Carboplatin for Injection, USP. In Trimmer, E. E.; Essigmann, J. M., Cisplatin. In Essays in Biochemistry, Vol 34, 1999, 1999; Vol. 34, pp Uchida, N.; Kasi, H.; Takeda, Y.; Maekawa, R.; Sugita, K.; Yoshioka, T., Anticancer Res. 1998, 18, Misset, J. L., Br. J. Cancer 1998, 77, Reedijk, J., Chem. Commun. 1996, Perez, R. P., Eur. J. Cancer 1998, 34, Lebwohl, D.; Canetta, R., Eur. J. Cancer 1998, 34, Hambley, T. W., Coord. Chem. Rev. 1997, 166, Levi, F.; Metzger, G.; Massari, C.; Milano, G., Clin. Pharmacokinet. 2000, 38, Bancroft, D. P.; Lepre, C. A.; Lippard, S. J., J. Am. Chem. Soc. 1990, 112, Hindmarsh, K.; House, D. A.; Turnbull, M. M., Inorg. Chim. Acta 1997, 257, Koubek, E.; House, D. A., Inorg. Chim. Acta 1992, 191, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1989, 166, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1989, 161, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1990, 173, Miller, S. E.; Gerard, K. J.; House, D. A., Inorg. Chim. Acta 1991, 190, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1991, 187, Miller, S. E.; Wen, H.; House, D. A.; Robinson, W. T., Inorg. Chim. Acta 1991, 184, Berners-Price, S. J.; Frenkiel, T. A.; Frey, U.; Ranford, J. D.; Sadler, P. J., J. Chem. Soc., Chem. Commun. 1992, Berners-Price, S. J.; Appleton, T. G., In Platinum-based drugs in cancer therapy, Kelland, L. R.; Farrell, N., Eds. Humana Press: Totowa, N.J., 2000; pp xii, Davies, M. S.; Berners-Price, S. J.; Hambley, T. W., Inorg. Chem. 2000, 39, Jennerwein, M.; Andrews, P. A., Drug Metab. Dispos. 1995, 23, Gately, D. P.; Howell, S. B., Br. J. Cancer 1993, 67,

43 28. Gale, G. R.; Morris, C. R.; Atkins, L. M.; Smith, A. B., Cancer Res. 1973, 33, Binks, S. P.; Dobrota, M., Biochem. Pharmacol. 1990, 40, Hromas, R. A.; North, J. A.; Burns, C. P., Cancer Lett. 1987, 36, Mann, S. C.; Andrews, P. A.; Howell, S. B., Cancer Chemother. Pharmacol. 1990, 25, Andrews, P. A.; Velury, S.; Mann, S. C.; Howell, S. B., Cancer Res. 1988, 48, Andrews, P. A.; Velury, S.; Mann, S. C.; Howell, S. B., In Platinum and other metal coordination compounds in cancer chemotherapy, Nicolini, M., Ed. Martinus Nijhoff: Boston, 1988; pp Safaei, R.; Katano, K.; Samimi, G.; Naerdemann, W.; Stevenson, J. L.; Rochdi, M.; Howell, S. B., Cancer Chemother. Pharmacol. 2004, 53, Holzer, A. K.; Samimi, G.; Katano, K.; Naerdemann, W.; Lin, X.; Safaei, R.; Howell, S. B., Mol. Pharmacol. 2004, 66, Holzer, A. K.; Katano, K.; Klomp, L. W. J.; Howell, S. B., Clinical Cancer Research 2004, 10, Safaei, R.; Holzer, A. K.; Katano, K.; Samimi, G.; Howell, S. B., J. Inorg. Biochem. 2004, 98, Appleton, T. G.; Connor, J. W.; Hall, J. R.; Prenzler, P. D., Inorg. Chem. 1989, 28, Berners-Price, S. J.; Kuchel, P. W., J. Inorg. Biochem. 1990, 38, Reedijk, J.; Teuben, J. M., Platinum-sulfur interactions involved in antitumor drugs, rescue agents, and biomolecules. In Cisplatin : chemistry and biochemistry of a leading anticancer drug, Lippert, B., Ed. Helvetica Chimica Acta, Wiley-VCH: Zürich, Weinheim, Chichester, 1999; pp xii, 563 p. 41. Ishikawa, T.; Bao, J.-J.; Yamane, Y.; Akimaru, K.; Frindrich, K.; Wright, C. D.; Kuo, M. T., J. Biol. Chem. 1996, 271, Dorr, R. T., A review of the modulation of cisplatin toxicities by chemoprotectants. In Platinum and other metal coordination compounds in cancer chemotherapy 2, Pinedo, H. M.; Schornagel, J. H., Eds. Plenum Press: New York, 1996; pp ix, 357 p. 43. Barnham, K. J.; Djuran, M. I.; del Socorro, P.; Sadler, P. J., J. Chem. Soc., Chem. Commun. 1994, Djuran, M. I.; Lempers, E. L. M.; Reedijk, J., Inorg. Chem. 1991, 30, Davies, M. S.; Berners-Price, S. J.; Hambley, T. W., J. Am. Chem. Soc. 1998, 120, Davies, M. S.; Berners-Price, S. J.; Hambley, T. W., J. Inorg. Biochem. 2000, 79, Rice, J. A.; Crothers, D. M.; Pinto, A. L.; Lippard, S. J., Proc. Natl. Acad. Sci. U. S. A. 1988, 85,

44 48. Bellon, S. F.; Lippard, S. J., Biophys. Chem. 1990, 35, Takahara, P. M.; Rosenzweig, A. C.; Frederick, C. A.; Lippard, S. J., Nature 1995, 377, Takahara, P. M.; Frederick, C. A.; Lippard, S. J., J. Am. Chem. Soc. 1996, 118, Chow, C. S.; Whitehead, J. P.; Lippard, S. J., Biochemistry 1994, 33, Jamieson, E. R.; Jacobson, M. P.; Barnes, C. M.; Chow, C. S.; Lippard, S. J., J. Biol. Chem. 1999, 274, Farrell, N., In Platinum-based drugs in cancer therapy: Cancer drug discovery and development ; 7., Kelland, L. R.; Farrell, N., Eds. Humana Press: Totowa, N.J., 2000; pp Farrell, N.; Qu, Y.; Bierbach, U.; Vaksecchi, M.; Menta, E., In Cisplatin : chemistry and biochemistry of a leading anticancer drug, Lippert, B., Ed. Verlag Helvetica Chimica Acta, Wiley-VCH: Zürich, Weinheim, New York, 1999; pp Manzotti, C.; Pratesi, G.; Menta, E.; Domenico, R. D.; Cavalletti, E.; Fiebig, H. H.; Kelland, L. R.; Farrell, N.; Polizzi, D.; Supino, R.; Pezzoni, G.; Zunino, F., Clinical Cancer Research 2000, 6, Hegmans, A.; Qu, Y.; Kelland, L. R.; Roberts, J. D.; Farrell, N., Inorg. Chem. 2001, 40, McGregor, T. D.; Hegmans, A.; Kasparkova, J.; Neplechova, K.; Novakova, O.; Penazova, H.; Vrana, O.; Brabec, V.; Farrell, N., J. Biol. Inorg. Chem. 2002, 7, Kraker, A. J.; Hoeschele, J. D.; Elliott, W. L.; Showalter, H. D. H.; Sercel, A. D.; Farrell, N. P., J. Med. Chem. 1992, 35, Perego, P.; Gatti, L.; Caserini, C.; Supino, R.; Colangelo, D.; Leone, R.; Spinelli, S.; Farrell, N.; Zunino, F., J. Inorg. Biochem. 1999, 77, Roberts, J. D.; Peroutka, J.; Beggiolin, G.; Manzotti, C.; Piazzoni, L.; Farrell, N., J. Inorg. Biochem. 1999, 77, Farrell, N.; Qu, Y.; Hacker, M. P., J. Med. Chem. 1990, 33, Pratesi, G.; Perego, P.; Polizzi, D.; Righetti, S. C.; Supino, R.; Caserini, C.; Manzotti, C.; Giuliani, F. C.; Pezzoni, G.; Tognella, S.; Spinelli, S.; Farrell, N.; Zunino, F., Br. J. Cancer 1999, 80, Farrell, N., Met. Ions Biol. Syst. 2004, 42, Davies, M. S.; Thomas, D. S.; Hegmans, A.; Berners-Price, S. J.; Farrell, N., Inorg. Chem. 2002, 41, Davies, M. S.; Cox, J. W.; Berners-Price, S. J.; Barklage, W.; Qu, Y.; Farrell, N., Inorg. Chem. 2000, 39, Hofr, C.; Farrell, N.; Brabec, V., Nucleic Acids Res. 2001, 29,

45 67. Kasparkova, J.; Zehnulova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2002, 277, Kloster, M.; Kostrhunova, H.; Zaludova, R.; Malina, J.; Kasparkova, J.; Brabec, V.; Farrell, N., Biochemistry 2004, 43, Kasparkova, J.; Vrana, O.; Farrell, N.; Brabec, V., J. Inorg. Biochem. 2004, 98, Hofr, C.; Brabec, V., Biopolymers 2005, 77, Cox, J. W.; Berners-Price, S. J.; Davies, M. S.; Qu, Y.; Farrell, N., J. Am. Chem. Soc. 2001, 123, Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, Kasparkova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2000, 275, Zehnulova, J.; Kasparkova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2001, 276, Berners-Price, S. J.; Ronconi, L.; Sadler, P. J., Prog. Nucl. Magn. Reson. Spectrosc. In Press, Corrected Proof. 76. Barton, S. J.; Barnham, K. J.; Habtemariam, A.; Sue, R. E.; Sadler, P. J., Inorg. Chim. Acta 1998, 273, Guo, Z.; Chen, Y.; Zang, E.; Sadler, P. J., J. Chem. Soc., Dalton Trans. 1997, Barnham, K. J.; Berners-Price, S. J.; Frenkiel, T. A.; Frey, U.; Sadler, P. J., Angew. Chem., Int. Ed. Engl. 1995, 34, Berners-Price, S. J.; Barnham, K. J.; Frey, U.; Sadler, P. J., Chem.--Eur. J. 1996, 2, Reeder, F.; Guo, Z.; Murdoch, P. D. S.; Corazza, A.; Hambley, T. W.; Berners-Price, S. J.; Chottard, J.-C.; Sadler, P. J., Eur. J. Biochem. 1997, 249, Parkinson, J. A.; Chen, Y.; Del Socorro Murdoch, P.; Guo, Z.; Berners-Price, S. J.; Brown, T.; Sadler, P. J., Chem.--Eur. J. 2000, 6, Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.--Eur. J. 2003, 9, Davies, M. S.; Berners-Price, S. J.; Cox, J. W.; Farrell, N., Chem. Commun. 2003, Yao, S.; Plastaras, J. P.; Marzilli, L. G., Inorg. Chem. 1994, 33,

46 2 Experimental Methods and Conditions Here I will discuss, in general, the theory that underpins the NMR experiments that were used in collecting data for this thesis. It should be noted that details specific to a given experiment will be found in the relevant materials and methods section of the appropriate chapter. Additional sections of this chapter present experimental work that provided the foundation for the work described in Chapters 3, 4, 5 and 6. This includes: i) the development of molecular modelling parameters and the determination of partial atomic charges required for accurate modelling of the electrostatic interactions of platinum complexes with DNA and ii) an introductory modelling experiment where the relationship between the platinum square plane and the plane of a guanine residue is investigated. 2.1 Nuclear Magnetic Resonance Techniques [ 1 H, 15 N] HSQC spectroscopy. The use of the inverse technique of [ 1 H, 15 N] HSQC NMR in the study of platinum anticancer agents is well developed. 1-3 The procedure has a number of advantages over direct 15 N methods. Primarily this advantage comes from the increased sensitivity to a theoretical maximum of 306 (i.e. ( γh / γn ) 5/2 ) compared to the directly detected 15 N. A requirement of this method is that a coupling constant be known for the N-H bonds of interest. This results in a spectrum where only 2D cross peaks for the atoms of interest are seen. Solutions must be prepared in water as to avoid deuterium exchange and tertiary amines are not detected by this method as they lack the requisite N-H coupling. The large coupling of 72 Hz for am(m)ines coordinated to platinum is ideal for this technique. As can be seen in Figure 2.1 the addition of GARP decoupling during acquisition simplifies the spectra by giving a 25

47 single peak rather than a doublet for a given species. As the nitrogen is attached to a platinum in all of the cases studied here 195 Pt satellites are expected. 195 Pt is a spin ½ nuclei with a natural abundance of 33.8%. Both the 1 J ( 15 N- 195 Pt) and 2 J ( 1 H- 195 Pt) satellites should be observed but in many cases these are broadened beyond detection by chemical shift anisotropy. 1 H Gradient Gradient N GARP INEPT 15 N Evolution Time INEPT Acquisition & 15 N Decoupling Figure 2.1 The [ 1 H- 15 N] HSQC pulse sequence. 4 The obvious disadvantage of this procedure is that the platinum complexes need to be 15 N labelled, which sometimes is not a trivial task. NH 2 NH δ 1 H Figure 2.2 Proton spectrum of a DNA oligonucleotide incubated with 15 N-1,0,1/t,t,t showing the regions where Pt-NH 2 and -NH 3 resonances are found. These regions are shown in the sample [ 1 H, 15 N] HSQC NMR spectrum shown in Figure

48 Figure 2.2 clearly demonstrates the difficulty of obtaining information from the 1 H spectra of Pt-DNA adducts. The two highlighted regions correspond to the regions where Pt-NH 2 and Pt-NH 3 resonances are found. In addition to crowding in these regions the proximity to the water peak increases the difficulty of analysis. The [ 1 H, 15 H] HSQC NMR experiment eliminates the 1 H 2 O signal, although some magnetization does appear at the water chemical shift in most HSQC spectra. The chemical shift range observed in the 15 N dimension for Pt(II) am(m)ine complexes is quite large, covering from -90 to -40 ppm. Figure 2.3 shows the regions where the 1 H, 15 N cross-peaks are observed for various platinum complexes. -90 H 3 N-Pt-O (cisplatin) -80 δ 15 N -70 Pt-O,N,Cl (cis) H 3 N H 3 N-Pt-N,Cl (cisplatin) H 2 N-Pt-O,N,Cl (trans) δ 1 H H 3 N-Pt-S (cisplatin) Figure 2.3 Regions of interest in the [ 1 H, 15 N] HSQC NMR spectra of cisplatin and multinuclear platinum drugs. The boxed areas roughly indicate the regions where 1 H, 15 N cross peaks appear. (trans) refers to an NH 2 group trans to the site of ligand substitution and (cis) refers to an NH 3 group cis to the site of ligand substitution in multinuclear platinum complexes. The (cisplatin) refers to the NH 3 group trans to a substituted chloride ligand in cisplatin type complexes. 27

49 The scheme shown in Figure 2.4 briefly describes the pathways that are available to 1,0,1/t,t,t when dissolved in solution with DNA present. The numbering associated with each species are used as labels subsequently in the text. Figure 2.4 Scheme depicting the pathway followed by 1,0,1/t,t,t from dissolution to the formation of a bifunctional adduct. Figure 2.5 shows a [ 1 H, 15 H] HSQC NMR spectrum during the reaction of 1,0,1/t,t,t with a 12mer DNA duplex. The 1 H, 15 N peaks seen in this spectrum correspond to the unreacted dichloro complex (1) a very small amount of aquachloro (2) and monofunctional (3) intermediates together with bifunctional adducts (5). Peaks for the NH 2 /NH 3 groups of the central {PtN 4 } linker are visible. Even though the spectra are very complicated they can be analysed to follow the stepwise formation of cross-links on the DNA and obtain rate constants for each step

50 Figure 2.5 A representative [ 1 H, 15 N] HSQC NMR spectrum obtained during the reaction between a 12mer duplex DNA and the multinuclear platinum complex 15 N-1,0,1/t,t,t. showing the Pt-NH 3 and -NH 2 regions. 7 1 H Watergate. The pulse sequence used for water suppression is the p3919 Watergate sequence as depicted in Figure 2.6. The gradient strength was also optimized beyond the standard 20% recommended to a value of 50%. Individual experiments required that the offset value (O1) be calibrated to the centre of the water peak within 2.5 Hz. Triple axis gradient shimming was applied on the Bruker spectrometers while z-axis gradient shimming was used on the Varian instrumentation. 1 H p16:1 p27*0.692 p27*1.462 p0*0.231 p1 p27*0.231 p27*1.462 p27*0.692 p16:1 Figure 2.6 Watergate pulse sequence p3919 from the Bruker Avance 600 MHz spectrometer. 2D 1 H NOESY spectroscopy. The acquisition of NOESY spectra was undertaken on both Bruker and Varian instruments. In both cases the pulse sequences employed 29

51 involved some degree of water suppression. On the Bruker Avance 600 a Watergate NOESY pulse sequence, seen in Figure 2.7, was employed. 8, 9 p1 p1 Grad p16:1 Grad p16:1 p1 p27*0.692 p27*1.462 p0*0.231 p27*0.231 p27*1.462 p27* H Figure 2.7 Pulse sequence used in obtaining NOESY spectra on the Bruker Avance 600 MHz spectrometer. 2.2 Molecular Modelling Background Introduction Three levels of theory can currently be applied to the computer modelling of molecular systems: molecular mechanics uses classic physical equations to calculate the potential energy of a system, semi-empirical programs use a limited implementation of the Schrödinger equations to solve the state of the molecule, while ab initio programs attempt to solve the wave function of a system completely. Generally as the complexity of the system being studied increases the level of the theory applied to that system decreases because of demands on computer resources. Consequently the study of biological macromolecules is usually restricted to the molecular mechanics or dynamics methods. The investigation of multinuclear platinum based complexes and their interactions with DNA required the development of parameters to allow the molecular mechanics 30

52 program Amber to model these interactions. A wide array of sources can be used in the development of parameters for molecular mechanics; experimental data from x-ray structures or extended X-ray absorption fine structure (EXAFS) can provide the basis for these parameters. The similarity between chemical structures is also a useful tool when developing parameters for molecular modelling. By comparing the atomic arrangements in an already parameterized molecule to the structural motifs in the molecule under investigation, variants of the original parameters can be developed for the new molecule being investigated. The process is iterative, so a number of testing cycles must be achieved before an appropriate set of parameters is obtained. Of special interest in this thesis is the determination of partial charges of substituted multinuclear complexes. For this the program MOPAC 2002 was used, the versions of this semi-empirical program released since 2002 have been able to perform calculations on molecules that contain platinum atoms. The following is a brief review of the equations and theory behind the two modelling programs used during the course of this project; HyperChem 5.11, Amber 7 and MOPAC Amber Bond Stretching. This term is associated with deformation of a bond from its standard equilibrium length. For small displacements from equilibrium, a harmonic function is often used: E 2 bond Κ r ( r r0 ) bonds = (2.1) 31

53 A larger value for the stretch force constant K r leads to a greater tendency for the bond to remain at its equilibrium distance r 0. Higher powers of r r 0, giving cubic, quartic, or higher terms are also common. Bond Angle Bending. This term determines the energy required to deform an angle from its equilibrium value. For small displacements from equilibrium, a harmonic function is often used: E 2 bond angle Κ θ ( θ θ 0 ) angles = (2.2) A larger value for the bending force constant K θ leads to a greater tendency for the angle to remain at its equilibrium value θ 0. Dihedrals. The term dihedral within the Amber program can be used in place of torsion angle. This term is associated with the tendency of dihedral angles to have a certain n-fold symmetry and to have minimum energy for the cis-, gauche-, or transconformation: Vn = [ 1+ cos(nφ φ )] (2.3) 2 Edihedral 0 dihedral The period of the interaction is 360/n. The phase angle φ 0 shifts the curve to the left or right. For n=1 and φ 0 =0, the curve represents the situation where the energy is a minimum for the trans-conformation with a barrier of V n to the highest energy cisconformation. A phase angle of φ 0 =180 represents the opposite situation with a minimum at the cis-conformation and a maximum at the trans-conformation. By including sums of terms of the above kind, dihedral angle interactions of arbitrary complexity can be described. 32

54 Improper Dihedrals. This energy term is exactly the same as the above dihedral term except that the atoms involved are not related in a linear sequence. The typical example is the trigonal planar arrangement of atoms. One might ask why this motif is singled out when they could be included in the dihedral energy term. The answer is that during the parameterization phase a minimum number of improper dihedrals are assigned, that is only when necessary, while every unique sequence of 4 atoms requires a dihedral set of parameters. As with the simple dihedral case generic terms can be introduced depending upon the situation. van der Waals. This term describes the repulsive forces keeping two non-bonded atoms apart at close range and the attractive force drawing them together at long range. A ij Bij E vanderwaals = 6 (2.4) 12 ij vdw Rij Rij The above potential is referred to as a Lennard-Jones or 6 12 potential and is summed over all non-bonded pairs of atoms ij. The first positive term is the short range repulsion and the second negative term is the long range attraction. The parameters of the interaction are A ij and B ij. There are six ways of specifying the constants A ij and B ij ; three by single atom type and three by pairs of atom types. Single atom type means that there are constants for individual atom types, i, that are summed by a combining rule that results in a parameters for a specific pair, ij, of atom types. Pairs of atom types means that parameter files contain explicit parameters for a pair of atom types ij and that no combining of single atom type parameters is necessary. If in the first case, there were N parameters, the second case would require N(N-1)/2 parameters for an equivalent set. If present, the constants specified by pairs of atoms types are used to override the values generated from single atom types. 33

55 Electrostatic. This term describes the classical non-bonded electrostatic interactions of charge distributions. qiq j E electrostatic = (2.5) ijelectrostatic εrij The above potential describes the monopole-monopole interactions of atomic charges q i and q j a distance R ij apart. Normally these charge interactions are computed only for non-bonded atoms and once again the 1 4 interactions might be treated differently from the more normal non-bonded interactions (1 5 relationship or more). The dielectric constant ε used in the calculation is sometimes scaled or made distance-dependent in the absence of explicit solvent where it is set to unity. Newer methods of implicit solvent calculation have been developed to mimic solvated conditions. Specifically, the Generalized Born approach to implicit solvent calculations is very useful when studying 10, 11 large systems or when computer resources are limited MOPAC 2002 Semi-empirical calculation methods differ from mechanics methods in that only the individual atoms within a molecule need to be parameterized. Another way to think of this is that in mechanics methods, all nitrogens are not created equal. Eliminating the need to parameterize large numbers of chemically distinct atoms is one reason why semi-empirical methods are attractive. The difficulty in performing semi-empirical calculations on complexes containing a transition metal has been that the functional used by the programs performing the calculations did not understand these elements of the size of platinum. Only recently has platinum been included in the AM1 basis set. The commercially available MOPAC 2002 includes this new AM1 basis set. The principle reason for using MOPAC 2002 was to determine the partial atomic charges in 34

56 the platinum complexes being studied. This distribution of charge is crucial when considering the electrostatic interactions between complex and DNA. Unfortunately the transition metals have not been included in the explicit water calculation method used in MOPAC The charges determined for all the complex at this point are gas phase only General Modelling Procedures Several distinct stages must be passed during the process of modelling any molecular system. coordinates must be obtained force fields need to be defined charges must be determined Even once these building blocks have been assembled the most important factor has yet to be addressed. What chemical/physical questions are to be answered by modelling these systems? The primary question to be investigated here is the nature of the preassociation of 1,0,1/t,t,t with DNA. Several oligonucleotides will be employed whose sequences are familiar from kinetics work. Occasional asides into smaller questions of structure or the parameterization of new drugs will also be considered in the chapter. 2.3 Parameterisation Molecular modelling parameters used in calculations performed in this thesis are based on those derived by John Cox during his PhD candidature at Virginia Commonwealth University, 12 which were in turn based on the work of Yao et al. 13 These parameters are 35

57 specific to the multinuclear platinum based drugs used in these studies. In several instances the variation between the ammine nitrogens may seem arbitrary, however the reason for defining multiple atom types for the apparently very similar atoms is for ease of assignment. The program HyperChem uses a rule based scheme to determine the atom type for a given atom. In the case of 1,0,1/t,t,t these rules were modified to result in the atom type assignments seen in Figure 2.8. However, in the case of 1,1/c,c it was decided that a manual assignment of the atom types was more efficient as the rule based system would not be able to distinguished between the nitrogens of the cis and trans complexes. In addition to that this route was chosen since Amber, the program with which most of the modelling was performed does not possess this rule based assignment scheme. The Amber force field, like most molecular mechanics force fields, divides the computational problem into components. The equations used to calculate the potential energy of a given system contain variables for the chemical bond, the dihedral angle and two types of torsion angles (proper and improper). Table 2.1 through Table 2.3 show those parameters derived for use with multinuclear platinum drugs. Only the parameters that were either missing from the original Amber force field, specifically the parm99.dat file distributed with Amber 7 or the equivalent set of files distributed with HyperChem version 5.11, or those that needed modification are included in the following Tables. 36

58 H3 H3 H3 H3 H3 H3 N3 H3 HC H3 N3 Y PT NM CT NM PT Y N3 H3 HC 6 H3 N3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 N3 H3 HC H3 N3 H3 HC H3 N3 Y PT NM CT NM PT NM CT NM PT Y N3 H3 HC 6 H3 N3 H3 HC 6 H3 N3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 NR H3 HC H3 NR H3 H3 H3 NI PT Y NL CT H3 HC 6 NL H3 PT Y NI H3 H3 Figure 2.8 Atom types assigned to 1,1/t,t and 1,0,1/t,t,t and 1,1/c,c where Y denotes a labile ligand, not an atom type. Table 2.1 Bond stretch parameters for the trans and cis multinuclear platinum complexes. Complex Bond (atom types) Force Constant (kcal/mol) Equilibrium Bond Length (Å) PT-N PT-NM PT-Cl trans PT-NB a N3-H H3-NM CT-NM PT-NR cis PT-NI PT-NL a NB is the atom type for the N7 nitrogen in guanine. 37

59 Table 2.2 Dihedral angle parameters for the trans and cis multinuclear platinum complexes. Complex Bond Angle Force Constant (kcal/mol) Equilibrium Angle (degrees) NM-PT-Cl NM-PT-N Cl-PT-N NM-PT-NB NB-PT-N N3-PT-N NM-PT-NM H3-NM-H trans PT-NM-CT PT-NM-H CT-NM-H PT-N3-H H3-N3-H PT-NB-CB PT-NB-CK CT-CT-NM HC-CT-NM cis NL-PT-NI NL-PT-Cl NL-PT-NB NL-PT-NR NR-PT-NB NR-PT-Cl NR-PT-NI NI-PT-Cl NI-PT-NB H3-NR-H H3-NR-PT H3-NI-PT H3-NI-H H3-NL-H H3-NL-CT H3-NL-PT CB-NB-PT CK-NB-PT CT-CT-NL HC-CT-NL The torsion and improper torsion angles fundamentally describe the same type of arrangement of atoms. Figure 2.9 demonstrates the difference between these to descriptors of atomic planes. 38

60 A B A B C D D C Figure 2.9 Proper and improper torsion angles. Table 2.3 Torsion angle parameters for the trans multinuclear platinum complexes. Complex Proper Torsion Angles Multiplicity Force Constant a Phase Periodicity X -CT-NM-X CT-CT-NM-H CT-CT-NM-PT trans HC-CT-NM-PT X -PT-N3-X X -PT-NM-X X -PT-NB-X X -N3-PT-NB cis trans X-PT-NI-X X-PT-NR-X X-PT-NL-X X-PT-NB-X X-PT-N3-X X-N3-PT-NB Improper Torsion Angles N3-N3-PT-NB N3-N3-PT-NM N3-N3-PT-Cl cis X-NL-PT-X X-NB-PT-X X-NI-PT-X X-NR-PT-X CK-NB-CB-PT PT-N3-NB-N PT-N3-CL-N PT-NB-N3-O PT-NB-N3-NB PT-NR-NI-N PT-NR-NI-NB PT-NI-NB-NR PT-NI-L-NR a The value for the force constant is treated differently depending on the multiplicity. The value in the table is that which is actually entered into the program. For explicit torsion angle the nascent force constant is divided by the multiplicity. X refers to any atom type at the given position. Even though HyperChem uses the same parameters and force field as Amber 7 the syntax used in defining the improper torsion angles is slightly different. As seen in 39

61 Figure 2.9 the central atom of the improper torsion angle, B, must be in the third position in of the list of atoms used to define the torsion angle. The end result of this is that improper torsion angles defined in HyperChem may not be valid in Amber 7, while the reverse is not true. All nomenclature used in Table 2.3 is suitable to Amber 7. Hydrogen bonding interactions were explicitly modelled in HyperChem while the newer version of the Amber 7 force field employed in the Amber 7 suite of programs no longer uses this form of the algorithm. The change in parameterization reflects the trend to calculations using explicit solvent where all non-bonding interactions are contained in a single potential term. This change is in line with the trend toward explicit solvent modelling in current simulations. In addition to this new explicit solvent model more accurate methods of implicit solvent calculations are now also available. Further parameterization was required for the detailed study of the aquation/substitution profiles of these drugs. Until now the only labile ligand to be modelled or parameterized was the chloro group. However, this is not particularly realistic since all of the complexes being studied contain two labile ligands that will be at least aquate and depending upon the solution may in fact interact with buffers that are present. To this end the aqua, phosphate (both terminal and bridging) and acetate groups were parameterized. Table 2.4 contains additional parameters required for these ligands. The method by which these parameters were obtained is described in the manual for Amber 7, the salient portion of which is quoted below. The general principle is to use analogy as much as possible. The amount of effort that should be expended is related to the scientific question being asked. To accurately calculate thermodynamic interactions with water or a macromolecule, one needs the best parameters that can be obtained. If only qualitatively reasonable geometries are needed, less work may be required

62 The paper by Bayly et al. also contains background information on the procedure used to determine new force field parameters. 15 As already indicated the initial parameters developed by John Cox were for the dichloro substituted complexes of 1,1/t,t and 1,0,1/t,t,t. Parameters have now been developed to include aqua, phosphato and acetato ligands in both the trans and cis arrangements. Figure 2.10 and Table 2.4 show the new parameters added to the Amber force field. H3 H3 H3 H3 H3 H3 H3 H3 H3 O3 O3 O3 P OS N3 H3 PT NM N3 H3 H3 H3 H3 HC HC HC CT C O OS N3 PT N3 H3 H3 H3 H3 NM H3 HW HW OW N3 PT N3 H3 H3 H3 H3 NM H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 NR H3 H3 NR H3 H3 NR H3 H3 H3 NI PT OS P O3 O3 O3 NL H3 H3 H3 O NI C PT OS CT NL H3 H3 H3 NI PT NL OW H3 HW HW HC HC HC Figure 2.10 Atom types developed for phosphato, acetato and aqua ligands of both cis and trans platinum. 41

63 Table 2.4 Additional parameters required for the aqua, acetato and phosphato ligands. Bond (atom types) Force Constant (kcal/mol) Equilibrium Bond Length (Å) Ligand PT-OS OAc, PO 4 PT-OW H 2 O C-OS OAc O3-P PO 4 Bond Angle Force Constant Equilibrium Angle Ligand (kcal/mol) (degrees) NM-PT-OS/O OAc, PO 4 N3-PT-OS/O OAc, PO 4 CT-C-OS OAc O-C-OS OAc C-OS-PT OAC NM-PT-OW H 2 O N3-PT-OW H 2 O HW-OW-PT H 2 O P-OS-PT PO 4 O3-P-O3/O PO 4 O3-P-OS PO 4 Proper Torsion Angles Multiplicity Force Constant a Phase Periodicity Ligand N3-PT-OW-HW H 2 O NM-PT-OW-HW H 2 O NM-PT-OS-C OAc N3-PT-OS-C OAc PT-OS-C-O OAc PT-OS-C-CT OAc NM-PT-OS-P PO 4 N3-PT-OS-P PO 4 O3-P-OS-PT PO 4 a The value for the force constant is treated differently depending on the multiplicity. The value in this Table is that which is actually entered into the program. For explicit torsion angle the nascent force constant is divided by the multiplicity. Many of the parameters that are concerned with the binding of oxygen to platinum are in common whether it is for the phosphato or acetato ligand. The Pt-O bond length in the aqua species is most certainly going to be different from that in the phosphato species. Refinement of these parameters will need to be addressed in future work. 42

64 Figure 2.11 shows two atoms types assigned to two structures unique to the 1,1/c,c dinuclear complex. Parameterization of these bridged species was performed in the same manner as described for the extended structures above. H3 H3 H3 H3 H3 H3 H3 NR H3 H3 NR H3 H3 H3 O2 NI P PT OS NL H3 H3 H3 NI PT OS NL H3 O2 H3 OS H3 H3 H3 H3 H3 NI PT NR NL H3 H3 H3 NI PT NR NL H3 H3 H3 H3 H3 H3 H3 Figure 2.11 Atom types assigned to the bridging µ-phosphate and µ-oh species of 1,1/c,c. The following is a brief discussion of any issues peculiar to a given ligand s parameter development. The Aqua Ligand. The parameterization of any atom that has one or more labile hydrogens possibly bound to it is a difficult task. The implied result in the case of aqua or hydroxo ligand is ph. However, ph in a molecular dynamic (MD) simulation has no meaning unless the explicit solvent used in creating the periodic boundary conditions includes some level of free hydronium or hydroxide molecules. This method would be intriguing but as yet has not been performed. It was therefore decided to use the HW and OW atom types used for free water in the periodic boundary conditions to simulate a bound water ligand. This was a compromise between a reasonable representation of the hydrogen bonding characteristics of water and the more realistic characteristics of the OS atom type for oxygen bound to electron withdrawing group such as phosphorus or platinum. Recent work by David A. Case at the Scripps Research Institute 43

65 demonstrates a methodology that allows modelling of systems at specified ph. 17 Although this method currently is limited to the Generalized Born implicit solvate calculation method. The Phosphato Ligand. Several difficulties were encountered with this ligand. The published parameters are quite sufficient however they model a fully deprotonated species and this is not the ideal state. 16 This aside, the geometry of the phosphate group is perfectly acceptable. Using the deprotonated ligand will overemphasize any H- bonding interactions between phosphate oxygens and adjacent amines or amines at other platinum centres. However, for the purposes of this work, understanding the possible interactions was of paramount importance as will be seen in the case of 1,1/c,c in a phosphate buffer system. The Acetato Ligand. The bulk of the values used in the acetato ligand were derived from existing Amber parameters for the amino acids which contain side chain acid groups, such as glutamic and aspartic acid. 2.4 Charge Determination (MOPAC 2002) Introduction Before any of the platinum complexes could be used for simulation reasonable charges needed to be calculated. MOPAC 2002 was chosen for this task since its semi-empirical parameterization was able to perform operations on molecules containing platinum. A simply geometry optimization was performed and the partial atomic charges were extracted for use in subsequent molecular dynamics simulations. 44

66 Determination of the partial atomic charge of atoms within a given molecule is crucial when calculations using these molecules are to investigate surface or electrostatic interactions. A major topic of this thesis that is described in Chapters 4, 5 and 6 is to investigate the preassociation of the central platinum or linker of 1,0,1/t,t,t which occurs by an electrostatic interaction within the minor groove of DNA. 5 A variety of methods exist for the calculation of partial charges. Semi-empirical, ab initio and DFT methods all of which employ various models for the calculation of partial atomic charges. The di and trinuclear platinum drugs studied here present a difficulty not encountered in the investigation of most organic drugs. Modelling transition metals and their complexes is still a complex problem. Most semi-empirical programs and many basis sets for ab initio calculations can not cope with the heavier transition metals. Density functional theory (DFT) is a valuable tool and is applicable to the determination of partial atomic charges of platinum complexes. MOPAC 2002, a semi-empirical modelling package which has been parameterized to make calculations on platinum among other transition metals. The calculation of partial charges in DNA oligonucleotides is reliably handled within the Amber suite of programs. The methodology employed was a restrained electrostatic potential fit (RESP) and was based on a the previous method of 15, 18 electrostatic potential fitting (ESP). The 1994 and 1999 force field charges were based on the 6-31G* basis set Method Models of 1,0,1/t,t,t, 1,1/t,t and 1,1/c,c were generated in HyperChem 5.11, while Vega, version was used to generate the input file for MOPAC A consistent set of A comprehensive study using DFT has recently been completed by Ph.D. student Joseph Moniodis in the Berners-Price lab. 45

67 input parameters were applied for the determination of charge within MOPAC The following table details the standard set of charges used. Calculations were performed on the supercomputer at APAC with the results being combined for application locally. MOPAC 2002, a semi-empirical calculation program, was used to determine the partial atomic charges of each platinum containing molecule investigated. Table 2.5 Standard input parameters used when running MOPAC 2002 to optimize the geometry and determine the partial atomic charges of the platinum complexes studied. Keyword Description AM1 RHF a Use the AM1 Hamiltonian Restricted Hartree-Fock CHARGE=number charge assigned to GNORM=0.01 Exit calculation when gradient drops below value specified BONDS GEO-OK Prints final bond order matrix Override interatomic distance check VECTORS DENSITY Print final eigenvectors Print final density matrix a Allows both open and closed shell systems Results The dinuclear complex 1,1/t,t demonstrated a high degree of symmetry in the partial atomic charges determined by MOPAC The atomic charges are paired about the longitudinal axis and the mirror plane normal to that axis within the molecule. 46

68 H H N (0.2806) (0.2433) ( ) (0.2801) H H (0.2241) H (0.1335) (0.1091) H H (0.1007) (0.1006) (0.1091) H H H (0.1333) H H (0.2806) (0.2241) H N (0.2420) ( ) H (0.2359) Cl ( ) (0.6661) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Pt N C C C C C C N Pt (0.6662) Cl ( ) H N H (0.2427) ( ) (0.2352) H (0.2256) H (0.1274) H (0.1091) (0.1002) (0.1001) (0.1092) (0.1274) H H H H H H (0.2258) (0.2803) N ( ) H (0.2343) H (0.2337) Figure 2.12 Semi-empirical partial charges determined for 1,1/t,t using MOPAC H (0.2425) Figure 2.13 The 1,1/t,t charge distribution represented by colour with blue corresponding to the most negative values and red to the most positive values. The wire mesh is based on the van der Waals radii of the atoms in the complex. The cis variant of the dinuclear complex (1,1/c,c) displays a similar distribution of charges with respect to the atomic charges across the mirror plane normal to the longitudinal axis. However, the charges determined for the two hydrogens on each of the carbon atoms within the linker chain are noticeably different. This difference in the charge on the hydrogens is more positive for 1,1/t,t, specifically the increase for the hydrogen bound to the two central carbons is and for the hydrogens 47

69 attached to the outside carbons of the linker. It is likely that this alteration in the charge distribution can be rationalized by consideration of the closer proximity of the chloro ligand to the linker hydrogens in the 1,1/c,c structure. H (0.2363) Cl ( ) H (0.2316) N ( ) H (0.2433) H (0.2766) H (0.2550) (0.1688) (0.1108) (0.1038) (0.0987) (0.1032) (0.1293) (0.2184) H H H H H H H H (0.2407) H (0.2446) ( ) (0.6715) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (0.6714) ( ) N Pt N C C C C C C N Pt N H (0.2443) H H (0.2404) (0.2306) N H ( ) (0.2373) H H (0.2179) (0.1297) (0.1031) (0.0990) (0.1035) (0.1113) (0.1683) (0.2551) (0.2767) (0.2430) H H H H H H H Cl ( ) H Figure 2.14 Semi-empirical partial charges determined for 1,1/c,c using MOPAC As seen in the 1,1/t,t representation of the charge distribution the 1,1/c,c structure has the largest localization of positive charge at the platinum centres surrounded by the negatively charged nitrogens of the ammines. A slightly larger charge gradient exists between the chloro ligands and the platinum atoms of 1,1/c,c than was seen in the 1,1/t,t model. 48

70 Figure ,1/c,c charge distribution represented by colour with blue corresponding to the most negative values and red to the most positive values. The wire mesh is based on the van der Waals radii of the atoms in the complex. As was seen in with the dinuclear complex, 1,1/t,t, the symmetry of 1,0,1/t,t,t results in the pairing of partial atomic charges both along the longitudinal axis and across the mirror plane normal to that axis. The most notable difference is seen in the charge that resides on the central platinum atom of the trinuclear 1,0,1/t,t,t. That platinum carries a charge of which is greater than the charge on either terminal platinum atom. The charges on the terminal platinum atoms are 0.661(1/2) which falls within 0.01 of the charges determined for the platinum atoms in either of the dinuclear complexes. 49

71 (0.2320) H (0.2614) (0.2305) H H (0.2846) H ( ) N (0.2440) H (0.2645) H N H H N H ( ) (0.2577) (0.2845) ( ) (0.2443) (0.2237) H (0.1316) H (0.1096) H (0.1306) (0.2465) (0.2469) (0.1361) (0.1170) (0.1031) (0.1034) H H H H H H H (0.1173) H (0.1021) H (0.1035) H (0.1091) H (0.1287) H (0.2244) H ( ) Pt N C C C C C C N Pt Cl ( ) (0.6612) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (0.7647) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (0.6611) N C C C C C Pt N C Cl H H H H H H H H H H H (0.1093) (0.1046) (0.1029) (0.1177) H H H H (0.1357) H H H ( ) (0.2241) (0.2322) (0.1171) (0.1031) (0.1036) (0.1101) (0.1315) (0.1347) (0.2470) (0.2581) (0.2456) (0.1383) (0.2220) (0.2847) N H H N H ( ) (0.2605) (0.2331) ( ) N (0.2442) H (0.2437) (0.2663) H H (0.2847) H Figure 2.16 Semi-empirical partial charges determined using MOPAC 2002 for the complex 1,0,1/t,t,t. Figure ,0,1/t,t,t charge distribution with the scale on the left displaying the negative charges represented in blue and the positive charges in red. 50

72 2.5 Molecular Dynamics Protocols Introduction Sequence Generation. Two routes are open to the molecular modeller to obtain the coordinates required for a simulation. Either a theoretical model can be used or coordinates derived from some experimental method. In the case of biological molecules such as DNA these experimentally determined coordinates come from X-ray diffraction crystal structures or NMR solution structures. Sequence Equilibration. A 200 ps molecular dynamics simulations was used to equilibrate the DNA sequences used. A large number of system parameters are monitored throughout the simulation. Potential and kinetic energy terms along with values describing the geometric conditions of the experiment such as density, volume are recorded over the length of the simulation When these values have stabilized the simulation is considered to be at equilibrium at in a suitable state to initiate an experiment. Docking. The docking of the drugs into the minor groove of a given sequence was accomplished using the NMR restraints capacity of the Sander program. Specific atoms in both the minor groove of the DNA sequence and in the central linker region of 1,0,1/t,t,t were selected. These atoms were assigned a separation of 5 Å and a force constant for this pseudo bond was input. Refer to Appendix A2.2.2 and A for the actually sander input file. The positioning of the central linker took place over a 20 ps MD simulation. 51

73 Production MD. As before Sander was used for the molecular dynamics calculations. Here a self submitting script was used to perform the extended calculations. This was required for a number of reasons the best two of which are data security and increased priority on the queuing system employed by APAC. An example script is available in Appendix A The length of the individual MD slices varied depending upon the size of the system being studied and how many water molecules were contained in the periodic box. It was found that 4 hours of wall time running on 4 cpus provided the best economy in terms of access, efficient use of computer resources and the allocation awarded to the Berners-Price research group in the APAC Merit Allocation Scheme. Analysis. A variety of programs were used to analyse the structural and energetic properties of these DNA/drug adducts. In all relevant cases the values being quoted are comparisons between a bound and a free state. Full analysis of the equilibrated DNA sequence was also performed in order to assess the impact on the structure of the modelling process. Where applicable relative values will be reported. The programs ptraj and carnal, both supplied with Amber, are tools for assessing the relative motion of the atoms in a system. Using both of these programs the root mean square deviation (RMSD) of the structure was determined over the course of the simulation relative to its starting position. The script used can be found in Appendix A In the case of DNA a root mean standard deviation of distance of less than 5 Å is considered stable. 14 This would be outside the acceptable limits for a globular protein due to their generally compact nature. 52

74 Sander generates a large amount of energy/physical property data during the MD simulation. A perl script originally create by David A. Case of the Scripps Research Institute and modified by myself, extracts these values for the course of a simulation. The program 3DNA created by Lu performs a complete structural analysis of DNA in several forms. 19 Parameters such as tilt, twist, roll among others that describe the duplex under simulation can be extracted from the structures saved during the course of the experiment Methods The coordinates used for the in silico oligonucleotides were derived from two sources. In the case of the 5 -d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 (dsgg) duplex a computer generated duplex based on database values defining the B- DNA three dimensional structure was created and an average coordinates set based on the NMR solution structure determined by Chen et al. was obtained, with thanks to Professor Peter Sadler at Edinburgh University. 20 Both sets of coordinates were subjected to a 200 ps equilibration procedure. A detailed comparison between these two DNA sequences was performed in order to determine the validity of using computer generated coordinates for subsequent sequences that did not have NMR solution structure data. Additional information regarding the comparison of these two DNA structures can be found in Appendix A2. Three such additional DNA sequences were studied; VB12, VB14 and VB16. These sequences were computer generated using coordinates provided in the Amber suite of programs. As with the dsgg sequence 200 ps of equilibration were performed. A detailed example of the equilibration protocol input files can be found in the Appendix A2. 53

75 14XL 5 -{d(atatgtacatat) 2 } 16XL 5 -{d(tatgtatacata) 2 } dsgg 5 -d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 VB12 5 -d(tctcctattcgcttatctctc)-3 5 -d(gagagataagcgaataggaga)-3 VB14 5 -d(tctccttcttgttcttcctcc)-3 5 -d(ggattaagaacaagaaggaga)-3 VB16 5 -d(ctctctctattgttatctcttct)-3 5 -d(agaagagataactatagagagag)-3 Figure 2.18 DNA nucleotide sequences used to generate the duplex structures studied in silico. The parm99.dat parameters were employed within the Amber 7 suite. This set of parameters is optimized for DNA and with the addition of the parameters derived for the multinuclear platinum complexes molecular mechanics/dynamics modelling of platinum-dna adducts is possible. The platinum specific parameter files can be found in Appendix A Calculations were performed on a variety of computers. To construct the molecular systems a dual PIII 866 MHz desktop computer was used with 512 MB of RAM running Linux RedHat 7.2 with MPICH version message passing interface installed for the purposes of handling the sharing of tasks between processors. Computer resources at the Western Australian Interactive Virtual Environment Centre (IVEC) and APAC super computer facilities were both used extensively for equilibration and production dynamics simulations. Again MPICH libraries were installed on these computers to allow for multiple CPU calculations. In general optimum scaling for Amber was achieved at 4 cpus, as beyond this level the efficiency started to drop off. An additional consideration was the time allocated on the super computer facilities which was based on time used per cpu, thus calculations employing more cpus at lower efficiency actually cost more in terms of the grant time. 54

76 All molecular manipulations were performed using the xleap program contained within the Amber suite of programs and all energy calculations were executed with the sander sub-program. The following is the protocol used in the equilibration of duplex DNA structures. The initial coordinates were verified for accuracy and protons were added to the structure. A short 1000 step minimization was performed on the added hydrogen atoms while keeping all other atoms in the system fixed with a large harmonic restraint. Sodium atoms were added to give an overall neutral charge to the system. As with the protons the sodium atom positions were minimized (1000 steps) while keeping all of the other heavy atoms in the system fixed. The restraints were removed and the entire system was subjected to a short minimization (1000 steps). At this point the periodic boundary conditions were applied to the system using the standard TIP3 model for water and rectangular box conditions. The TIP3 model of water is a rigid 3 point water molecule used for explicit solvent calculations. 21 This model of water limits much of the nonbonding interactions that would be calculated if water molecules were fully applied to a system. The outer edges of the box were 10 Å away from the outer most edge of a box defined by the solute molecules. This resulted on average in to water molecules being added to the system. The following table summarizes the minimization and molecular dynamics steps involved in equilibrating these systems. A comparison was made between the 3D structure of the dsgg DNA duplex generated by the nucgen database for nucleic acids, a sub-program of Amber 7, and the structure obtained using the coordinates from the duplex solution structure solved by NMR. The relative similarity in structure between the two dsgg sequences and the overall low 55

77 RMS difference between both suggests that the Amber 7 coordinates and force field perform to a reasonable standard when generating B form DNA oligonucleotides. Restrained Minimization - 1 DNA Fixed kcal/mol 10 psec Restrained Dynamics - 1 DNA Fixed kcal/mol constant volume only Restrained Minimization - 2 DNA Fixed kcal/mol Restrained Minimization - X DNA Fixed X = kcal/mol kcal/mol kcal/mol 6-50 kcal/mol 7-25 kcal/mol 8-10 kcal/mol 9-1 kcal/mol 10 psec Restrained Dynamics - 2 DNA Fixed kcal/mol constant volume/pressure Unrestrained Minimization psec Unrestrained Dynamics - 3 to 22 constant volume/pressure Figure 2.19 Method followed for the equilibration of oligonucleotide sequences for subsequent use in Amber MD simulations. 56

78 Docking. The drug of interest was introduced into the DNA system at a distance of at least 10 Å. Identical procedures were followed as for the equilibration of the DNA sequence alone. After the equilibration phase was completed the drug molecule was docked into predetermined site on the DNA using the NMR restraints capabilities of Amber. The net effect of this process was to move the drug from its original 10 Å position to one of 3 to 4 Å distance from the DNA helix. Production MD. Initially a 2 ns production MD run was performed on every system studied. These preliminary results allowed for assessment of the system without committing to larger amount of computer time Analysis The analysis of an MD simulation is accomplished using a variety of accompanying and 3 rd party programs. All of the scripts used to execute these programs can be found in Appendix A2.3 and on the back cover CD. Structural changes. A primary measure of the stability of a simulation is the measure of the positional difference of atoms of interest at given time intervals during the course of the simulation. In the Amber package this can be accomplished using two utilities; carnal and ptraj. As a consistency check both utility programs were used and the results compared. Energetic and Simulation Parameters. A wide variety of parameters can be recorded during an execution of the Sander MD portion of Amber. A post run processing script 57

79 written by David Case of the Scripps Research Institute was used to extract the following parameters: total energy, potential energy, kinetic energy, density, volume, temperature and pressure. A copy of the two files required to run this script can be found on the accompanying compact disc within the scripts directory. The execute file is called run_mdout and the processing script is title process_mdout.perl. DNA structure parameters. A 3 rd Party program called 3DNA version 1.5 written by Xiang-Jun Lu 19 was used to analyse the structural parameters of the DNA helix during the progress of a MD simulation. Compilation of the data obtained from this program was performed by a PHP script written by Julian Tonti-Filippini and myself. 3DNA provides a vast amount of information ranging from local base pair parameters such as tilt and roll to global helix information like the kink and helix form whether it is A form, B form or other. Of the myriad parameters available to describe the DNA helix the most useful are the pseudorotation angle and Chi. These parameters provide information about the local and global changes in the DNA structure. The pseudorotation angle describes, in one term, the pucker of the deoxyribose. C 2' - endo C 3' - endo C 5' C 5' Base O 3' Base O 3' Figure 2.20 Structures for (a) the C2'-endo and (b) the C3'-endo conformations of ribose

80 It has been shown that the sugar pucker is indicative of the helix form of the DNA. The pseudorotation angle has several regions of interest. If the value was between 108 and 180 a C1 -exo to C2 -endo configuration for the sugar was observed. Within that region the S Type conformation is observed at angles of B-DNA has a characteristic value of or a C2 -endo configuration. Chi (χ), the torsion angle which is defined by the four atoms shown in the Figure 2.21 describes the relationship between the sugar and the nucleobase. In general it is confined to two configurations, the syn or anti. As with the pseudorotation angle, χ bears a relationship to the form of the DNA. For angles between -90 to -180 or 90 to 180 the anti configuration is observed. While angles of -60 to 90 results in a syn configuration. The high anti configuration occurs with χ values between -90 and Normally B-DNA has a χ value of ~-95 meaning an anti configuration. The binding of platinum to the N7 position of guanine may result in a change from syn to high anti. The A-DNA structure has anti values for its χ value while Z-DNA alternates between the syn and anti configurations. O O N NH HN N N * * N NH 2 H 2 N N N * * HO H O * H * HO H O * H * H OH H H H OH H H anti Figure 2.21 The anti and syn configurations of the DNA nucleoside guanosine. * indicates atoms that define Chi (χ). syn 59

81 The base-pair and base-step parameters observed in DNA duplex structures can be useful in understanding the DNA helix. Of all the parameters monitored by 3DNA the following are of particular interest. Buckle usually falls in the range of ± 1.0 for the B- DNA duplex and variations outside of this range are considered to be significant. The opening parameter for B form DNA is -3.8 while the propeller twist is often negative for right handed DNA but can cover the range of ± 1.4. Tilt is a sequence dependent parameter that generally falls between 0 and 5 and twist is usually 36 in B-DNA. Large changes to the roll parameter can suggest disruption of the base stacking within the DNA duplex. 23 Only when exceptional changes are observed do these parameters play a significant role in the analysis. Generally they are less likely to change substantially when electrostatic interactions are being monitored. The program 3DNA functions by converting the final coordinates of a simulation into a PDB file. This PDB file provides the basis for the analysis. To simplify the analysis process only the final structure of any given MD slice was analysed. As noted previously, the MD simulation was divided into segments that would complete within 4 hours of wall time on the APAC supercomputer in Canberra. A large number of datasets were created for each simulation since on average a 4 hour interval was equivalent to 20 ps of simulation with the minimum acceptable simulation running to over 2000 ps. This allowed for additional structural and energetic information to be extracted from molecular dynamics files at virtually any point within the simulation. The data obtained from 3DNA are usually presented in the form of an X-Y plot, with time on the X axis and the particular parameter being described on the Y axis. This being the case what is presented is a time dependent variation of the given parameter. Some consideration must be paid to the possibility that the change observed is in fact a function of Amber and the MD simulation. For this reason the analysis was performed not only on production 60

82 simulations but also on the sequence and pre-docking equilibration MD runs. This provides a basis for comparison, a baseline in effect, for 3DNA and the systems being studied. Of the myriad parameters available to describe the DNA helix the most useful are P and χ. These parameters provide information about both local and global changes in the DNA structure. P describes, in one term, the pucker of the sugar. It has been shown that the sugar pucker is indicative of the helix form of the DNA. As with P, χ, which is the torsion angle describing the relationship between the sugar and the nucleobase bears out a relationship to the form of the DNA helix. Other common base pair or base step parameters such as buckle, opening, tilt, twist, propeller twist are considered, although their usefulness is limited. Only when exceptional changes are observed do these parameters play a significant role. Generally they are less likely to change substantially when electrostatic interactions are being monitored. Figure 2.22 Pseudorotation wheel, used to describe the relationship between the pseudorotation angle and sugar pucker.24 61

83 The program 3DNA functions by converting the final coordinates of a simulation into a PDB file which was subsequently used as the basis for the analysis. To simplify the analysis process only the final structure of any given MD slice was analysed. As noted previously the MD simulation was divided into blocks that would complete within 4 hours on the APAC super computer. The large number of files resulting from this approach to molecular dynamics provided very nice sampling of the data. On average the sampling took place at 20 ps intervals over the course of a 2 ns simulation. In addition this approach provides an excellent failsafe mechanism in case of a machine failure. The simulation can be restarted from any completed 20 ps time interval. Further sampling is possible if exceptional circumstances present themselves. The 3DNA analysis was performed on all simulations that were run. In the case of equilibration runs this analysis was intended to provide a baseline to compare to. It was crucial to be sure that a change seen in the 3DNA analysis was caused by the close contact between DNA and the platinum complexes under study Visualization MPEG movies have been made of all the 2 ns production molecular dynamics simulations and these can be found on the accompanying compact disc. The program VMD 25 was used to load the trajectory files generated by Amber. Plugin programs available from the VMD website ( were then used to generate movies from the simulation. The static images seen in this thesis were generated in a two step process, first SwissPDBViewer 3.7 SP5 26 was used to create an appropriate view of the system being observed from a PDB file. This was then saved as a POV-Ray 3.5 scene. POV-Ray 27 is 62

84 a ray tracing program freely available from POV-Ray ( POV-Ray allows for high quality 3D computer generate images to be created from minimally 3D representations such as PDB files. 2.6 Rotation freedom at N7 of guanine Introduction The orientation of the platinum plane at the N7 position of guanine is most likely dictated by geometric and steric considerations. In an attempt to further understand the binding of 1,1/t,t and 1,0,1/t,t,t to DNA the angular dependence of a Pt(NH 3 ) 3 group binding to both the 5' and 3' N7 of the dsgg sequence has been investigated. These models proved useful in the interpretation of the NMR spectra of the reaction of 1,1/t,t with the dsgg duplex where evidence for two different conformations of the 5'!5' 14XL interstrand adduct was found Materials and Methods The 14 base pair DNA duplex was generated using the database within HyperChem The dsgg sequence was subjected to extensive minimization using a modified version of the Amber force field included with HyperChem. These modifications, described earlier, allow for the inclusion of a variety of platinum based anticancer drugs into the model systems. A Pt(NH 3 ) 3 group was covalently linked to the N7 of G(7) or G(8) where the angle between the plane of the Pt(NH 3 ) 3 and the guanine was systematically varied until a representative sampling of the entire range of binding 63

85 approaches was modelled. Each system was then minimized to a gradient of kcal/mol. Modelling was performed on the dsgg sequence using HyperChem 5.11 and a modified version of the Amber 94 parameter set. 13 A Dell 500 MHz PC with 128 MB of RAM was used to perform the following calculations. The angle between the plane formed by the platinum centre and that of the guanine nucleobase was varied systematically over two regions as defined in Figure A Pt(NH 3 ) 3 group was used as a model for the multinuclear drugs being studied in order to simplify the system. After each variation in the torsion angle between drug and base planes was made the system was minimized to an energy gradient of kcal/mol. Both the G(7) and G(8) N7 sites were studied using this protocol. An implicit solvent method was employed to mimic a water environment. A distant dependent dielectric constant of 4 was used to make this possible. Both potential binding sites in the central region of the dsgg sequence were modelled Results and Discussion It appears that two distinct conformations exist in both the G(7) and G(8) environments. The results are not all that surprising in that an initial angle of 0 to 90 results in a minimized angle of 45 or 30 for G8 or G7 respectively. Similarly with an initial angle of 90 to 180 the resultant angle is 145 or 130 for G(8) and G(7) respectively. What is of interest is that the two regions do not yield the same results, that is G(8) and G(7) are different at least in terms of Pt(NH 3 ) 3 binding with the HyperChem force field. The minimized energies obtained were consistent with the angles obtained. However, the difference in energy between one set of torsion angles and another is not really 64

86 significant. Also the energy difference between G(7) and G(8) conformers with similar torsion angles is insignificant. Figure 2.23 An example of the initial angle between the square plane of the platinum and the plane of the guanine Initial Angle (degrees) Initial Angle (degrees) Final Angle (degrees) Final Angle (degrees) Figure 2.24 Final torsion angle dependency upon initial torsion angle for the G(7) or 5' system and the G(8) or 3' system. 65

87 Figure 2.25 The two orientations of the platinum plane shown in the above figure both demonstrate an angular dependence upon the initial torsion angle but are also energetically indistinguishable Conclusions It has been clearly shown that two regions exist at approximately 40 off the plane of the guanine nucleotide base. There appears to be little or no difference between the 5' and 3' situation with regard to the orientation of a PtN 3 group around the bond connecting the platinum to the N7 position in guanine. With the generically defined torsions that John Cox 12 used this torsional angle about the Pt-N7 bond would have been represented as an energy maximum at 0. 66

88 2.7 References 1. Barnham, K. J.; Berners-Price, S. J.; Guo, Z.; Murdoch, S. d. P.; Sadler, P. J. In 7th International Symposium on Platinum and Other Metal Coordination Compounds in Cancer Therapy, Plenum: New York, Berners-Price, S. J.; Sadler, P. J., Coord. Chem. Rev. 1996, 151, Chen, Y.; Guo, Z.; Sadler, P. J., In Cisplatin : chemistry and biochemistry of a leading anticancer drug, Lippert, B., Ed. Verlag Helvetica Chimica Acta, Wiley-VCH: Zürich, Weinheim, New York, 1999; pp Palmer, A. G., III; Cavanagh, J.; Wright, P. E.; Rance, M., J. Magn. Reson. 1991, 93, Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.--Eur. J. 2003, 9, Cox, J. W.; Berners-Price, S. J.; Davies, M. S.; Qu, Y.; Farrell, N., J. Am. Chem. Soc. 2001, 123, Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, Piotto, M.; Saudek, V.; Sklenar, V., J. Biomol. NMR 1992, 2, Sklenar, V.; Piotto, M.; Leppik, R.; Saudek, V., J. Magn. Reson., A 1993, 102, Tsui, V.; Case, D. A., Biopolymers 2001, 56, Tsui, V.; Case, D. A., J. Am. Chem. Soc. 2000, 122, Cox, J. W. Interactions between multinuclear platinum complexes and DNA with a focus on ligand, linker groups and kinetics: A molecular modelling [ 1 H, 15 N] NMR study. Virginia Commonwealth University, Richmond, Yao, S.; Plastaras, J. P.; Marzilli, L. G., Inorg. Chem. 1994, 33, Case, D. A.; Pearlman, D. A.; Caldwell, J. W.; Cheatham III, T. E.; Wang, J.; Ross, W. S.; Simmerling, C.; Darden, T.; Merz, K. M.; Stanton, R. V.; Chen, A.; Vincent, J. J.; Crowley, M.; Tsui, V.; Gohlke, H.; Rader, R. V.; Chen, P. J.; Massova, I.; Kollman, P. A., Amber Users' Manual. University of California: San Fransisco: Bayly, C. I.; Cieplak, P.; Cornell, W.; Kollman, P. A., J. Phys. Chem. 1993, 97, Meagher, K. L.; Redman, L. T.; Carlson, H. A., J. Comput. Chem. 2003, 24, Mongan, J.; Case, D. A.; McCammon, J. A., J. Comput. Chem. 2004, 25, Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Kollmann, P. A., J. Am. Chem. Soc. 1993, 115, Lu, X.-J.; Shakked, Z.; Olson, W. K., J. Mol. Biol. 2000, 300, Chen, Y.; Parkinson, J. A.; Del Socorro Murdoch, P.; Guo, Z.; Berners-Price, S. J.; Brown, T.; Sadler, P. J., Chem.--Eur. J. 2000, 6,

89 21. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L., J. Chem. Phys. 1983, 79, Blackburn, G. M.; Gait, M. J., Nucleic acids in chemistry and biology. 2nd ed.; Oxford University Press: Oxford [England] ; New York, 1996; p xv, 528 p. 23. Dornberger, U.; Flemming, J.; Fritzsche, H., J. Mol. Biol. 1998, 284, Altona, C.; Sundaralingam, M., J. Am. Chem. Soc. 1972, 94, Humphrey, W.; Dalke, A.; Schulten, K., J. Mol. Graph. 1996, 14, Guex, N.; Peitsch, M. C., Electrophoresis 1997, Cason, C. POV-Ray(tm) rendering engine for Windows, 3.1g;

90 3 Aquation of 1,1/t,t and 1,1/c,c in Various Buffered Environments 3.1 Introduction The aquation of cisplatin has been extensively studied providing an excellent knowledge base for developing an understanding of how platinum based complexes interact within the biological milieu Specifically, the profile of platinum species that will be able to interact with DNA when the buffer environment is made up of potential donor ligands needs to be fully understood, in that it directly impacts on the efficacy of platinum complexes as antitumour drugs. The complicated environment of the blood plasma and the completely different intracellular environment provide a host of potential ligands. Phosphate in the blood plasma along with glutathione are two obvious ligands, however other species such as the phospholipids that make up part of the cellular membrane could potentially interact with platinum based complexes and may play a role in their transport across cellular membranes. [ 1 H, 15 N] HSQC NMR spectroscopy has been used to study the aquation of cisplatin, 13, 14 [PtCl(dien)] + 15, 16, [PtCl 2 (NH 3 )(2-picoline)], 17, 18 19, 20 and cis-[ptcl 2 (NH 3 )(cyclohexylamine)]. The aquation of 1,1/t,t in perchlorate solution has been previously studied by [ 1 H, 15 H] HSQC NMR, but was not used to follow the kinetics of the reaction. 21 More recently the kinetic profile of 1,0,1/t,t,t was examined in the presence of perchlorate and phosphate solutions using [ 1 H, 15 H] HSQC NMR. 22 Preliminary experiments on 1,1/t,t (not included here) were carried out in conjunction with Dr. Murray Davies. Experiments on the 1,1/c,c complex were performed in conjunction with Dr. Junyong Zhang. All modelling studies are the sole work of the author. The work in this Chapter is included in the following publication. 1 69

91 In this chapter [ 1 H, 15 N] HSQC NMR is used to investigate the aquation/ligation profiles of 1,1/t,t and 1,1/c,c. Three aqueous systems are employed: perchlorate, phosphate and acetate. The study allows for comparison with previously published data and also expands the understanding of platinum-ligand interactions in solution. The following scheme (Figure 3.1) shows the complete set of reaction pathways possible when 1,1/t,t and 1,1/c,c are dissolved in aqueous buffer environments. H 3 N NH 2 (CH 2 )n H 2 N NH 3 2+ H 3 N NH 2 (CH 2 )n H 2 N 2+ NH 3 Pt Pt Pt Pt H 3 N Y Y NH 3 Y NH 3 H 3 N Y 1,1/c,c (n=6) Y/Y 1,1/t,t (n=6) Cl/Cl (1) Cl/HO (4) H + K a1 k -1 k 1 K 1 Cl - H 2 O Cl/H 2 O (2) L H 2 O K 3 k 3 k -3 Cl/L (7) HO/H 2 O (5) H + K a2 k -2 k 2 K 2 k -4 Cl - H 2 O Cl- H 2 O/H 2 O L H 2 O K 5 k 5 k -5 H 2 O H 2 O/L (3) (8) k 4 K 4 H 3 O + H 3 O + k 6 k -6 1,1/c,c µ-po 4 (10) H + K a3 k -7 H 3 O + k 7 K 7 H 3 O + H 2 O L 1,1/c,c µ-oh HO/HO L/L (6) (3') (9) Model A K 1 K 3 k 1 k 3 Pt-Cl Pt-H 2 O Pt-L k (I) -1 k (II) -3 (III) Model B Figure 3.1 A general scheme (Model A) representing the possible pathways available to 1,1/t,t and 1,1/c,c in perchlorate, phosphate or acetate solution (L=OAc or PO 4 ). Bridging species of hydroxide (3 ) and phosphate (10) are only generated from 1,1/c,c. Model B depicts a simplified scenario employed in the kinetic modelling in cases where the equivalent {PtN 3 Y} end groups in different species were indistinguishable by NMR. 70

92 Molecular models of 1,1/c,c ligand substituted species were created and used to substantiate the assignment of species that appeared in the [ 1 H, 15 H] HSQC NMR spectra. 3.2 Materials and Methods Both the nitrate salt of 15 N-[{trans-PtCl(NH 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ 1,1/t,t and 15 N- [{cis-ptcl(nh 3 ) 2 } 2 (µ-nh 2 (CH 2 ) 6 NH 2 )] 2+ 1,1/c,c were provided by Professor Nick Farrell. NMR Spectroscopy. A detailed explanation of the pulse sequences and the experimental method associated with them can be found in Chapter 2. The following is a brief description of the specific conditions for the NMR experiments in this Chapter. A Bruker 600 MHz spectrometer ( 1 H, MHz, 15 N 60.8 MHz) equipped with a 5 mm triple resonance probe with triple axis gradients and dedicated 1 H, 13 C, 15 N channels as well as 2 H for locking was employed in all experiments. Shigemi NMR tubes were used for these experiments in order to reduce the volume needed and a minimal amount of D 2 O (5% of total volume) was added to avoid signal loss due to deuterium exchange. The 1 H chemical shifts were referenced by the addition of 1,4-dioxane (δ 3.767) to the solution. The 15 N chemical shifts were calibrated externally against 15 NH 4 Cl (1.0 M in 1.0 M HCl in 5% D 2 O/95% H 2 O) at δ ( 15 N) H NMR experiments. The Watergate pulse sequence was used to suppress the 1 H 2 O signal in the 1 23, 24 H NMR spectra. The 1 H spectra, were either 32 or 64 scans with 32 K or 64 K points, the spectral width of 12 khz corresponds to a sweep width of 20 ppm. A relaxation delay of 2.5 seconds was employed. 71

93 [ 1 H, 15 N] HSQC NMR experiments. In the case of the 2D [ 1 H, 15 N] HSQC experiments the standard Bruker phase sensitive HSQC pulse sequence was employed, setting the 1 J ( 15 N, 1 H) coupling constant to 72 Hz. 25 Decoupling of the 15 N signals was accomplished with the GARP-1 sequence during the acquisition segment of the pulse sequence. The [ 1 H, 15 N] HSQC spectra consisted of 4 transients of 48 or 96 increments in t 1 yielding experiments of approximately 7 minutes or 14 minutes in length, with an acquisition time of seconds and spectral widths of 6 khz in f 2 ( 1 H, ppm) and 2.1 khz (for 1,1/t,t) or 5.5 khz (for 1,1/c,c) in f 1 ( 15 N, ppm). Zero-filling to the next power of 2 in both f 2 and f 1 dimension was the only processing measure applied to the data. ph measurements. In order to determine the pk a of 1,1/c,c in solution the ph was measured using a Corning ph Micro NMR Combo Electrode (model ) and an Ecoscan ph meter. The Corning electrode was designed to be inserted into a 5 mm NMR tube and measure ph from a small volume. The electrode was a standard type that used KCl as a supporting electrolyte. For this reason over the course of the experiment chloride leached from the electrode into the solution. This fact is covered in more detail in the discussion of the pk a results. For general ph measurements during the kinetic experiments a handheld Shindengen ph Boy-P2 (su19a) ph meter was employed. To avoid leaching of chloride aliquots of 5 µl of the solution were placed on the electrode surface and the ph was recorded (the aliquots were not returned to the sample). Both meters were calibrated using ph buffers at ph 6.9 and 4.0. Adjustments in ph were made using 0.1 M and 0.01 M HClO 4, or 0.1 M and 0.01 M NaOH. Data Analysis. The kinetic analyses of the aquation or ligation reactions were undertaken by measuring the peak volumes in the [ 1 H, 15 N] HSQC NMR spectra using 72

94 the Bruker XWINNMR software and calculating relative concentrations of {Pt( 15 NH 3 ) 2 } at each time point, using the technique described by Davies et al. for the aquation of 1,0,1/t,t,t. 22 Briefly, for a given reaction, peak volumes were determined using an identical vertical scale and threshold value. The HSQC peak corresponding to the aquachloro species (2) overlaps with the peak for the diaqua peak (3) (when considering the {PtN 3 H 2 O} group). This problem of overlap also applies to the chloro species, the HSQC peak for the {PtN 3 Cl}group of the dichloro (1) and aquachloro (2) complexes are indistinguishable. The practical consequence of this is the inability to calculate the concentration of individual species such as the diaqua and aquachloro directly from the peak volumes. As will be shown, the exception to this case is the aquation of 1,1/c,c in phosphate, where distinct peaks are seen and attributed to additional species described by Model A. The first stage in processing the peak volumes was to account for the overlapping species. Every asymmetrical complex gives rise to two peaks in the HSQC spectra; it was assumed that these pairs of peaks will have the same intensity and/or volumes. Thus the 1 H, 15 N peak resulting from the {PtN 3 (H 2 O)} group of the aquachloro complex is subtracted from the peak attributed to the dichloro complex. The result is a peak volume that represents the amount of dichloro species since the contribution from the chloro end of the aquachloro species has been removed. However, the peak attributed to the aquachloro species also contains a contribution from the diaqua species. This must be taken into account otherwise the calculated aquachloro species will be underestimated and the dichloro species overestimated. In order to perform the calculation it is assumed that the two platinum coordination units (separated by an alkanediamine chain of six carbon atoms) do not interact with each other and can be treated independently. 21 This assumption made for previous studies of 1,1/t,t 21 and 1,0,1/t,t,t 22 was based on calculation performed for 73

95 dicarboxylates by Rajasekaran 26 and has been substantiated recently by van Eldik and co-workers 27 in a study of the interactions between Pt(II) centres in [Pt 2 (N,N,N`,N`tetrakis(2-pyridylmethyl)diamine)-(H 2 O) 2 ] 4+ complexes. When these platinum units act independently of each other their chemical environments are the same, thus resulting in coincident chemical shifts in the HSQC spectra. The practical outcome of the assumption is that the equilibrium constant for the first aquation step (K 1 ) is assumed to be equal to that of the second step (K 2 ). In order to obtain individual peak volumes for all species (dichloro 1, aquachloro 2, and diaqua 3) a Visual Basic program was written to iteratively determine the concentration of all three species while maintaining the K 1 = K 2 condition. The calculation was initiated with 10% of the peak volume being assigned to diaqua. Convergence was achieved when K 1 -K 2 < Aquation of 1,1/t,t and 1,1/c,c in Perchlorate Solution Sample Preparation. Solutions of 1.02 mm in 15 N-1,1/t,t (0.37 mg, 0.50 µmol, 485 µl) and 0.94 mm in 15 N-1,1/c,c (0.35 mg, 0.45 µmol, 480 µl) were prepared in 15 mm sodium perchlorate buffer, with 5% D 2 O for locking purposes. The initial ph values were 5.4 and 6.0 for 1,1/t,t and 1,1/c,c, respectively, and in both cases the ph dropped by units over the course of the reaction. A series of 1 H and [ 1 H, 15 N] HSQC NMR spectra were recorded at 298 K until the system reached equilibrium. During the early stages of the reaction HSQC spectra were recorded every 7 minutes and later 14 minute intervals were used. Finally, as equilibrium was approached a 46 minute interval was inserted in between the recording of HSQC spectra. 74

96 Aquation of 1,1/t,t. A number of impurities have been identified in the 1,1/t,t solution and are labelled i in Figure 3.2a. The chemical shifts of these impurities do not change during the course of the reaction and do not interfere or overlap with any of the peaks related to 1,1/t,t. The three impurity peaks amount to less than 10% of the total material observed. A more complete analysis of these impurity species is described in reference. 21 Initially only the 1 H/ 15 N NMR peaks at δ 3.89/-64.6 in the NH 3 region and the peak in the NH 2 region (δ 5.05/-47.0) were observed and assigned to the starting material 15 N-1,1/t,t. Figure 3.2a depicts a typical spectrum obtained near equilibrium. The peaks attributed to the starting material decreased in intensity over the course of the experiment while a second pair of peaks assigned to aquated species of 15 N-1,1/t,t appeared at δ 4.11/-62.1 and δ 5.08/ These peaks grew in intensity until equilibrium was attained after 5 hours. NH 2 region NH 3 region trans-nh 3 region cis-nh 3 region -68 i * a b,3 3' u b δ( 15 N) b, δ( 1 H) i i 2b, ,2a * 3.8 δ ( 15 N) * i * 1,2a i 1,2a * 2b,3,4 * δ ( 1 H) Figure 3.2 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t (5.2 hours) and (b) 15 N-1,1/c,c (4.5 hours) in 15 mm sodium perchlorate solution (95% H 2 O/5% D 2 O) at 298 K. Peaks labelled * are 195 Pt satellites, those labelled are artefacts and i are impurities in the sample at 298 K. The peak assignments are shown in Figure 3.3; labels a and b refer to {PtN 3 Cl} and {PtN 3 H 2 O} ends, respectively and u is an unidentified peak. The relative concentrations of the three species 1, 2 and 3 at each time point were calculated as described above, and the rate constants were obtained using the kinetic model shown in Figure 3.3. The Scientist equation file is included in Appendix A

97 Cl/Cl (1) k 1 k -1 Cl/H 2 O (2) k 2 k -2 H 2 O/H 2 O (3) k 3 k -3 µ-oh (3') Figure 3.3 The scheme represents the model used in the kinetic fit of the aquation of 1,1/t,t and 1,1/c,c in perchlorate solution. The equilibrium involving the hydroxide bridged species (3 ) is only observed for 1,1/c,c. The time dependence of the concentrations of species and the kinetic fit derived from the model (Figure 3.3) are shown in Figure 3.4a and the rate constants are shown in Table a 1.0 b Concentration (mm) Concentration (mm) Time (hours ) Time (hours) Figure 3.4 Plots of the time dependence of species in the aquation of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c at 298 K in 15 mm NaClO 4 (95% H 2 O/5% D 2 O) according to the model shown in Figure 3.3. Key: dichloro (1), aquachloro (2), * diaqua (3),! µ-oh (3 ) and chloride. For 15 N-1,1/c,c the concentrations are based on the peaks in the trans-nh 3 region. Examination of the rate and equilibrium constants (Table 3.1) shows good agreement with previous studies of the aquation of 15 N-1,1/t,t in 0.1 M NaClO 4, where the equilibrium constants (pk 1 = pk 2 ) were estimated to be 3.9 ± 0.2 (at 25 C) from the [ 1 H, 15 N] NMR spectrum at equilibrium. 21 In the latter study no significant difference was found for the equilibrium constant at 25 C and 37 C and the aquation (k 1, k 2 ) and anation (k -1, k -2 ) rate constants were derived from a chloride release experiment at 37 C, determined using a chloride selective electrode. 76

98 Comparison of these data with the rate constants obtained here at 25 C show that the aquation rate constant increases with temperature and the anation rate constant increases in proportion, such that equilibrium is reached more rapidly at 37 C (1.9 hours), but the position of the equilibrium remains the same. Similar temperature dependent behaviour was observed in the case of 1,0,1/t,t,t. 22 Table 3.1 Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm NaClO 4 in comparison to 1,0,1/t,t,t. a Parameter 1,1/c,c b 1,1/t,t 1,0,1/t,t,t 22 k 1 (10-5 s -1 ) 2.26 ± ± ± 0.2 k -1 (M -1 s -1 ) 0.50 ± ± ± k 2 (10-5 s -1 ) 5.3 ± ± ± 1.5 k -2 (M -1 s -1 ) 1.6 ± ± ± 0.05 k 3 ( s -1 ) ± 0.73 k -3 ( s -1 ) 0.24 ± 1.74 pk ± ± 0.01 c 3.35 ± 0.04 c pk ± 0.19 pk a 1,1/t,t and 1,1/c,c are 1.0 mm, 1,0,1/t,t,t is 3.39 mm. b The fitting was based on concentrations of 1, 2 and 3 derived from the peaks in the trans-nh 3 region. c For 1,1/t,t and 1,0,1/t,t,t, pk 1 = pk 2 (see text). Aquation of 1,1/c,c. For the aquation of 15 N-1,1/c,c in 15 mm perchlorate solution at 298 K a major difference compared to the 15 N-1,1/t,t reaction is the appearance of a 1 H/ 15 N NMR peak in the trans-nh 3 region which has different 1 H/ 15 N shifts to that of the aquachloro species (2). This peak is visible after about 2 hours as a shoulder and then becomes a clear peak at δ 4.30/-85.0 (Figure 3.2b). This is due to the changing ph of the solution during the course of the experiment causing a slight change in the 1 H/ 15 N shift of the peak for the aquated species (2) (see Figure 3.12). A possible assignment for this new peak is the diaqua species (3). The molecular model (see below and Figure 3.13) shows the possibility of hydrogen bonds between the two aqua ligands which could result in a slight difference in 1 H chemical shifts compared to the aquachloro species (2). However, the chemical shift of the new peak does not change with ph and 77

99 a more likely assignment is the hydroxo-bridged species (3 ) (see model in Figure 3.14). Hence the assumption made in the analysis of the 1,1/t,t reaction that the two positively charged {PtN 3 Y} groups act independently of one another is still applicable here, since only one peak is observed for the aquachloro (2) and diaqua (3) species. The absolute amount of the µ-oh species (3 ) in solution is at no time greater than 2%. For consistency, and in order to allow comparison between aquation rate constants for the 1,1/t,t and 1,1/c,c complexes, data were analysed with the same kinetic model (Figure 3.3), but with inclusion of the µ-hydroxo species which was assumed to form from the diaqua species (3). The time dependence for the concentrations of species and the kinetic fit derived from this model are shown in Figure 3.4b and the rate and equilibrium constants are given in Table 3.1. Comparison of Figure 3.4a and b shows that equilibrium is attained after a similar time (5-6 hours) for the two complexes with slightly more of the 1,1/c,c (78.4%) than the 1,1/t,t (71.6%) remaining at equilibrium. It is interesting to observe that the rate constant for the first aquation step more than doubled for 1,1/t,t compared to 1,1/c,c, whereas the anation rate constant for the first step of each reaction is virtually the same, resulting in an equilibrium constant which favours the aquated species of 1,1/t,t. Previous studies have shown that for 1,1/t,t and 1,0,1/t,t,t the dichloro form is strongly favoured over the aquated forms under these conditions and the equilibrium for 1,0,1/t,t,t lies further toward the aquated species. Examination of the data in Table 3.1 shows that overall 1,1/c,c is the least hydrolysed of the three complexes under similar conditions. The assignment of the hydroxo-bridged species was confirmed by Dr. Junyong Zhang using ESI-MS. 1 78

100 3.4 Aquation of 1,1/t,t and 1,1/c,c in Acetate Solution Sample Preparation. Solutions of 0.93 mm in 15 N-1,1/t,t (0.31 mg, 0.40 µmol, 430 µl) and 0.95 mm in 15 N-1,1/c,c (0.33 mg, 0.43 µmol, 450 µl) were prepared in 15 mm sodium acetate buffer with 5% D 2 O for locking purposes. The initial ph of the solutions were ph 5.3 and ph 6.0 for 15 N-1,1/t,t and 15 N-1,1/c,c, respectively and this did not change over the course of the reaction. Proton and [ 1 H, 15 N] HSQC NMR spectra were obtained at 298 K until the system reached equilibrium. A 46 minute interval was inserted in between the recording of later HSQC spectra. 1,1/t,t in Acetate. Two additional peaks (δ 3.95/-63.7 and δ 4.95/-61.8) in the NH 3 and NH 2 regions, respectively are observed in the [ 1 H, 15 N] HSQC NMR spectra for the reaction of 1,1/t,t in 15 mm acetate (Figure 3.5a) which were not seen in the reaction of 1,1/t,t in perchlorate. These peaks are assigned to acetato species. NH 2 region NH 3 region trans-nh 3 region cis-nh 3 region a -90 b -68 i * -85 III II δ( 15 N) II III i i II III I δ ( 15 N) I i II III I δ( 1 H) * δ ( 1 H) Figure 3.5 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c at 298 K in 15 mm sodium acetate solution (95% H 2 O/5% D 2 O) after 12.3 hours. Peaks labelled * are 195 Pt satellites, those labelled are artefacts and i are impurities in the sample. The peak assignments are shown in Figure 3.6 (Model B). 79

101 Equilibrium conditions were attained very slowly (~64 hours after mixing, Figure 3.7a). Due to the number of overlapped peaks it is not possible to ascertain the concentration of all species at each time point in this reaction. For example the peak at δ 3.88/-64.4 contains contributions from the dichloro species (1), the aquachloro species (2) and the mixed chloroacetato species (7) while the peak at δ 3.95/-63.7 theoretically contains contributions from each of the acetato species chloroacetato (7), aquaacetato (8) and diacetato (9). To analyse the data a simplified kinetic model (Figure 3.6) that treats each complex as a monomeric species, {PtN 3 Y}, was employed. K 1 K 3 k 1 Pt-Cl Pt-H 2 O Pt-OAc k (I) -1 k (II) -3 (III) k 3 Figure 3.6 Scheme representing the aquation and ligand substitution (Model B) of 1,1/t,t and 1,1/c,c in acetate buffer. The observed 1 H/ 15 N peaks account for the total concentration of chloro, aqua and acetato species. The time dependence of the concentrations of these species and the kinetic fit derived from the model shown in Figure 3.6 (Model B) are shown in Figure 3.7a. The rate constants are listed in Table 3.2. At equilibrium acetato species account for 53% of the total platinum species and chloro species have decreased to 43%. Table 3.2 Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm sodium acetate (at 298 K) according to the model shown as Figure 3.6. Compound a k 1 (10-5 s -1 ) k- 1 (M -1 s -1 ) k 3 (M -1 s -1 ) k -3 (10-5 s -1 ) pk 1 pk 3 1,1/c,c b 1.56 ± ± ± ± ± ± ,1/t,t 1.83 ± ± ± ± ± ± 0.02 a The fitting was based on concentrations of the total chloro (I), aqua (II) and acetato (III) species. b For the reaction of 1,1/c,c in acetate concentrations of species were based on the peaks in the trans-nh 3 region. 80

102 2.0 a 2.0 b Concentration (mm) Concentration (mm) Time (hours ) Time (hours ) Figure 3.7 Plots of the time dependence of species in the aquation of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c at 298 K in 15 mm sodium acetate according to the model shown in Figure 3.6 (Model B). Key: chloro (I), aqua (II),! acetato (III) and chloride. For 1,1/c,c the concentrations are based on the peaks in the trans-nh 3 region. 1,1/c,c in Acetate. When the aquation of 1,1/c,c is carried out in 15 mm acetate the [ 1 H, 15 N] HSQC spectra also exhibit the peaks for the chloro and aqua species as well as new peaks at δ 4.20/-83.6 (trans-nh 3 ) and δ 3.83/-65.3 (cis-nh 3 ) assignable to O- bound acetato species (Figure 3.5b). These peaks first appear ~50 minutes into the reaction and their chemical shifts do not change for the remainder of the reaction. No new peak is visible in the Pt-NH 2 region, as the peak for the acetato species overlap with that of the aqua species (δ 4.57/-42.8) at this ph. As for the reaction with 1,1/t,t, 1 H/ 15 N peaks for the different dinuclear species can not be distinguished and the data were analysed by the simplified kinetic model (Figure 3.6 Model B) in which each complex is treated as a monomeric species {PtN 3 Y}. The plot shown in Figure 3.7 shows that equilibrium is reached more slowly (~104 hours after mixing) than for the reaction with 1,1/t,t and there is a greater percentage of acetato species present. At equilibrium acetato and chloro species account for 64% and 32% of the total platinum species, respectively. The rate and equilibrium constants derived from the model shown in Figure 3.6 (Model B) are shown in Table 3.2. The values can not be compared with those determined 81

103 using Model A (Table 3.1) but allow for the comparison of reactions carried out in different buffer conditions and between 1,1/t,t and 1,1/c,c. A comparison of pk 1 and pk 3 in Table 3.2 indicates that for both dinuclear complexes there is a similar preference for chloride over acetate. 3.5 Aquation of 1,1/t,t and 1,1/c,c in Phosphate Solution Sample Preparation. Solutions of 0.93 mm in 15 N-1,1/t,t (0.31 mg, 0.40 µmol, 430 µl) and 0.94 mm in 15 N-1,1/c,c (0.30 mg, 0.39 µmol, 410 µl) were prepared in 15 mm sodium phosphate buffer with 5% D 2 O for locking purposes. The initial ph values of the solutions were for 5.3 and 5.9 for 15 N-1,1/t,t and 15 N-1,1/c,c, respectively. Over the course of the reaction the ph dropped by about 0.7 ph units in each case. Proton and [ 1 H, 15 N] HSQC NMR spectra were obtained at 298 K until the system reached equilibrium. δ( 15 N) II 5.2 NH 2 region 5.0 III 4.8 δ( 1 H) NH 3 region i i i II III * I * 3.8 a δ ( 15 N) a 3' 2b,3 * trans-nh 3 region cis-nh 3 region * * * * * 7b,8b 1,2a,7a i 2b,3,3',8a i 4.0 δ ( 1 H) b 1,2a,7a 7b,8b, Figure 3.8 [ 1 H, 15 N] HSQC NMR spectra of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c in 15 mm sodium phosphate solution (95% H 2 O/5% D 2 O) after 12.3 hours Peaks labelled * are 195 Pt satellites, those labelled are artefacts and i are impurities in the sample. The peak assignments are shown in Figure 3.9 for 1,1/t,t and in Figure 3.1 for 1,1/c,c; labels a and b refer to {PtN 3 Cl} and {PtN 3 H 2 O} ends of the dinuclear complexes, respectively. 82

104 1,1/t,t in Phosphate. The [ 1 H, 15 N] HSQC spectra acquired for the reaction of 15 N- 1,1/t,t in 15 mm phosphate show peaks assigned to the phosphato species at δ 3.98/ and δ 4.94/-63.9 in the NH 3 and NH 2 regions, respectively. These peaks first appear ~22 minutes after the commencement of the reaction and their chemical shifts are unchanged throughout the course of the reaction. The 15 N chemical shift of the phosphato species is consistent with oxygen binding and is similar to that of the aqua species. The reaction attained equilibrium after 12 hours, which is substantially faster than that observed for the process with acetate buffer yet, slower than the perchlorate system. The time dependent plot of the concentration of the different species is shown in Figure 3.10a, along with the kinetic fit derived using the simplified model shown in Figure 3.9 (Model B). It is evident that the equilibrium in phosphate strongly favours the dichloro species; 71% of the total platinum remains coordinated to chloride, 22% to phosphate and 7% to aqua ligands. K 1 K 3 k 1 Pt-Cl Pt-H 2 O Pt-PO 4 k (I) -1 k (II) -3 (III) k 3 Figure 3.9 Simplified scheme (Model B) used to model phosphate ligand substitution of both 1,1/t,t and 1,1/c,c. The reaction of 1,0,1/t,t,t with phosphate 22 reached equilibrium in a time similar to that of 1,1/t,t. However, a greater percentage (50%) of phosphato species was present at equilibrium. Table 3.3 shows the rate and equilibrium constants derived from the simplified Model B in comparison to the reaction of 1,0,1/t,t,t under similar conditions. 22 Examination of the data in Tables 3.2 and 3.3 shows that the rate constants (k 3 ) for binding of acetate and phosphate to 1,1/t,t are very similar, but the 83

105 2.0 a 2.0 b Concentration (mm) Concentration (mm) Time (hours ) Time (hours ) Figure 3.10 Plots of the time dependence of species in the aquation of (a) 15 N-1,1/t,t and (b) 15 N-1,1/c,c at 298 K, in 15 mm sodium phosphate solution according to the model shown in Figure 3.9 (Model B). Key: chloro (I), aqua (II),! phosphate (III) and chloride. For 1,1/c,c the concentrations are based on the peaks in the cis-nh 3 region. anation rate constant (k -3 ) is close to 10 fold higher in the case of phosphate, and this value is comparable to that found in the reaction of 1,0,1/t,t,t. A comparison of pk 1 and pk 3 in Table 3.3 demonstrates the clear preference for chloride over phosphate for both 1,1/t,t and 1,0,1/t,t,t. Table 3.3 Rate and equilibrium constants for the aquation of 1,1/c,c and 1,1/t,t in 15 mm sodium phosphate at 298 K, according to the model shown in Figure 3.9. Compound a k 1 (10-5 s -1 ) k- 1 (M -1 s -1 ) k 3 (M -1 s -1 ) k -3 (10-5 s -1 ) pk 1 pk 3 1,1/c,c b 2.18 ± ± ± ± ± ± ,1/t,t 2.49 ± ± ± ± ± ± ,0,1/t,t,t c 3.4 ± ± ± ± ± ± 0.01 a The fitting was based on concentrations of the total chloro (I), aqua (II) and phosphato (III) species. b For the reaction of 1,1/c,c in phosphate the peaks in the cis-nh 3 region were used for the determination of concentrations. c Data obtained from Davies et al. 22 were refitted according the model shown in Figure ,1/c,c in Phosphate. When the aquation of 1,1/c,c is carried out in 15 mm phosphate the [ 1 H, 15 N] HSQC spectra show several new peaks consistent with the binding of phosphate (Figure 3.8b) and the individual species (as shown in Figure 3.1) can be identified and monitored by observing the peaks in the trans-nh 3 region. The first 1 H/ 15 N peaks assignable to a phosphate bound species appear ~1.2 hours into the reaction at δ 4.15/-85.7 (trans-nh 3 ), 3.84/-62.5 (cis-nh 3 ) and 4.65/-41.4 (NH 2 ) and are 84

106 assigned to the {PtN 3 PO 4 } group of the phosphatochloro species (7). The partner peaks, for the {PtN 3 Cl} group, are all concealed by the peaks for 1. One new peak assignable to the aquaphosphato species (8) (δ 4.38/-85.7) is first visible in the trans-nh 3 region after about 3 hours, the partner is assumed to be concealed by the peak for 7. The existence of a distinct peak for the aquated group in 8 suggests there may be an interaction between the coordinated aqua and phosphato groups which influences the 1 H chemical shift of the trans-nh 3 ligand. The molecular model of 8 (see below and Figure 3.15) supports the existence of such a H-bond. The cis-nh 3 and NH 2 groups will be less sensitive to this interaction so that in these regions the peaks for the aqua group of 8 are coincident with those of the aquated species (2 and 3). Another new peak (δ 4.22/-87.1) appears at this time as a shoulder to peak 7b in the trans-nh 3 region and grows in intensity until equilibrium is attained. The chemical shift is insensitive to ph and is assigned to the macrochelate phosphate bridged species (10) (see model in Figure 3.16). To compare the reaction with that of 1,1/t,t the reaction was analysed by the simplified kinetic model shown in Figure 3.9 with the peaks in the cis-nh 3 region used to obtain the concentrations of the total chloro (I), aqua (II) and phosphato (III) species at each time point. The time dependent plot and kinetic fit derived from Model B is shown in Figure 3.10b. Equilibrium is reached very slowly (>80 hours), a marked contrast to the reaction of 1,1/t,t with phosphate which reaches equilibrium in 12 hours. The derived rate and equilibrium constants are listed in Table 3.3. Comparison of these data with those in Table 3.2 show that for 1,1/c,c the rate constant for binding of phosphate and acetate (k 3 ) are quite similar, as are the anation rate constants (k -3 ). A comparison of the pk 1 and pk 3 values indicates a slight preference for chloride over phosphate which is The assignment of species 10 has been confirmed by 31 P NMR and ESI-MS experiments performed by Dr. Junyong Zhang. 1 85

107 Time (hours) Figure 3.11 Plots of the time dependence of species in the aquation of 15 N-1,1/c,c at 298 K in 15 mm phosphate (95% H 2 O/5% D 2 O) according to Model A in Figure 3.1. Key: dichloro (1), aquachloro (2), * diaqua (3),! chlorophosphato (7), aquaphosphato (8), + µ-phosphate (10) and chloride. The concentrations are based on the 1 H, 15 N peaks in trans-nh 3 region where every species can be distinguished. similar to the preference of chloride over acetate. The major difference for the reaction of 1,1/c,c with phosphate compared to 1,1/t,t and 1,0,1/t,t,t is that the bound phosphate is much less labile. The anation rate constant k -3 is about 12-fold lower. This difference means that equilibrium is attained much more slowly and leads to a greater proportion of phosphate bound species. The difference can be explained by the fact that only 1,1/c,c is able to form the macrochelate phosphate-bridged species (10). Although the formation of the species is reversible, the lability of the bound phosphate is considerably reduced in comparison to the 1,1/t,t compounds. The reaction of 1,1/c,c with phosphate was analysed further by obtaining the concentration of every species at each time point from the 1 H/ 15 N peaks in the trans- NH 3 region to give the time dependent plot shown in Figure The major species are the chlorophophato species 7 (58% of the total Pt) and the µ-po 4 species 10 (25%) and only a small amount of the intermediate aquaphosphato species 8. The kinetic fit of 86

108 the data, performed by Dr. Junyong Zhang, derived from Figure 3.1 (Model A) and the rate and equilibrium constants for the complete reaction can be found in the full publication. 1 Table 3.4 Rate and equilibrium constants derived from the kinetic fit of the reaction 1,1/c,c and 1,0,1/t,t,t in phosphate buffer using Model A (Figure 3.1). Parameter 1,1/c,c 1,0,1/t,t,t a k 1 (10-5 s -1 ) 3.84 ± ± 0.1 k -1 (M -1 s -1 ) ± ± k 2 (10-5 s -1 ) 4.43 ± 0.48 k -2 (M -1 s -1 ) ± k 3 (M -1 s -1 ) ± ± k -3 (10-5 s -1 ) 0.46 ± ± 0.12 k 4 (10-5 s -1 ) 1.02 ± ± 0.10 k -4 (M -1 s -1 ) ± ± 0.03 k 5 (M -1 s -1 ) ± k -5 (10-5 s -1 ) 1.56 ± 1.93 k 7 (10-5 s -1 ) 7.33 ± 0.69 k -7 (10-5 s -1 ) 1.18 ± 0.16 k 8 (s -1 ) 0.68 ± 7 k -8 (s -1 ) 481 ± 4000 pk ± ± 0.01 pk ± 0.19 pk ± ± 0.01 pk ± ± 0.3 pk pk ± 0.10 a Data from reference. 22 In this model the aquation of 2 to 3 was not considered. 87

109 3.6 pk a determination of 15 N-[{cis-Pt(H 2 O)(NH 3 ) 2 } 2 (µ- NH 2 (CH 2 ) 6 NH 2 )] 2+ AgNO 3 (34.0 mg, 0.2 mmol) was dissolved in H 2 O (1.0 ml) and a 6.8 µl (1.8 equiv.) aliquot was added to 15 N-1,1/c,c (0.53 mg, 0.7 µmol) in a solution containing 35 µl of 2.0 M NaClO 4, 35 µl D 2 O, 620 µl H 2 O and 10 µl 1,4-dioxane (10 mm, as reference). The solution was incubated overnight at 37 C and then centrifuged to remove the AgCl precipitate. The final concentration of the 1,1/c,c diaqua complex was 0.98 mm in NaClO 4 (100 mm in 5% D 2 O/95% H 2 O). Examination of the [ 1 H, 15 N] NMR spectrum indicated small amounts (< 4%) of the chloro forms 1 or 2 of 1,1/c,c, however their presence did not compromise the experiment. Adjustments in ph were carried out by the addition of NaOH (0.1 or 0.01 M in 5% D 2 O/95% H 2 O) and HClO 4 (0.1 or 0.01 M in 5% D 2 O/95% H 2 O), respectively. The [ 1 H, 15 N] HSQC spectra were recorded in the ph range 2.20 to As indicated earlier the ph probe used was able to be inserted into 5 mm NMR tube however the probe did leach potassium chloride. The consequence of this is that the equilibrium that was rapidly skewed to the diaqua (3) form of the complexes slowly returned to favouring the dichloro species (1). The ph titration data were analysed using the following equation: δ = (δ A [H + ] + δ B K a )/([H + ] + K a ) (3.1) Where K a is the acid dissociation constant for one Pt-OH 2 group of the diaqua complex (3) and δ A and δ B are the chemical shifts of the diaqua and dihydroxo complexes, respectively. The program KaleidaGraph (Synergy Software, Reading, PA) was used for fitting, as described previously for 1,1/t,t 21 and 1,0,1/t,t,t. 22 Figure 3.12 shows the changes in chemical shift of the 1 H and 15 N resonances of 1,1/c,c diaqua complex with ph. 88

110 -78 a 4.8 b δ 15 N δ 1 H ph ph Figure 3.12 Plots showing the change in (a) 15 N and (b) 1 H chemical shifts with ph for Pt-NH 3 (cis- ( ), trans ( )) and Pt-NH 2 ( ) groups in the 1,1/c,c diaqua complex (3) in 100 mm ClO 4 at 298 K. Using equation 3.1 the three 1 H titrations curves were fitted and an average pk a value of 6.01 ± 0.03 was calculated. The value is 0.4 ph units greater than the pk a of 5.62 determined for both 1,1/t,t and 1,0,1/t,t,t. This higher pk a results in a larger percentage of the 1,1/c,c complex being in the active diaqua form at physiological ph. The pk a of 1,1/c,c is still significantly below that of cisplatin at (6.41). 14 It is notable that only one point of inflexion is detected in the titration curves indicating (as for 1,1/t,t and 1,0,1/t,t,t) that K a2 K a3, although both the NMR and molecular modelling experiments (see below) suggest that in the 1,1/c,c case the two positively charged {PtN 3 O} groups can interact. 3.7 Molecular Modelling Molecular modelling studies of 1,1/c,c substituted complexes was performed to provide further evidence for assignment of species observed in the HSQC NMR spectra. Specifically the aquaphosphato (8), diaqua (3), µ-hydroxo (3 ) and µ-phosphato (10) species were modelled. MOPAC is a semi-empirical molecular modelling program and calculations were performed using the AM1 method in a restricted Hartree-Frock 89

111 formalism in advance of molecular dynamics calculations. These initial calculations served two purposes; first to provide a reasonable starting structure for subsequent molecular dynamics experiments and second the calculation provided a set of partial atomic charges for the molecules being studied. The MOPAC 2002 version was required since it has been parameterized to allow calculations on molecules containing platinum. The molecular dynamics calculations were performed using HyperChem 5.11 employing the amber94 force field modified by Yao et al. 28 and by John Cox in his Ph.D. dissertation. 29 A simulated annealing routine was employed to isolate a good structure from the high energy molecular dynamics calculation. The equilibration phase of the simulated annealing was performed at 300 K over 200 ps while the cooling phase took 100 ps with a final temperature of 100 K. The entire molecular dynamics process was conducted using periodic boundary conditions and explicit solvent. The structures isolated from this process were subjected to minimization until the energy gradient was less than kcal/mol. Images were rendered from Swiss PDB Viewer 3.7 SP5 30 using POV-Ray Diaqua complex (3) and µ-hydroxo bridged species (3 ). The model of the diaqua complex (Figure 3.13) was constructed to establish whether hydrogen bonding interactions between the two aqua groups could account for the distinct peak observed in the trans-nh 3 region (δ 4.30/-85.0) of the [ 1 H, 15 N] HSQC spectra of 1,1/c,c solutions in perchlorate and phosphate buffers. 90

112 Figure 3.13 The molecular model of the diaqua species (3) of 1,1/c,c suggesting possible hydrogen bonding between the two aqua ligands indicated by the dashed line. The distance between the proton of an aqua ligand and the oxygen of its opposing aqua ligand increased from a minimum of 3.5 Å at the start of the simulated annealing process to maximum of 5 Å in the minimized structure. The strength of the attraction between the two aqua ligands appeared to be insufficient to maintain the strained hydrogen bonded structure. The 1 H/ 15 N peak assigned to the diaqua species (3) was later identified as the bridging hydroxo species (3 ) by ESI-MS. 1 The model of this species is shown in Figure

113 Figure 3.14 The bridging hydroxo species (3 ) observed in the 1,1/c,c aquation reaction in perchlorate solution. Aquaphosphato species (8). The [ 1 H, 15 N] HSQC spectra of the reaction of 1,1/c,c in phosphate buffer showed a distinct peak in the trans-nh 3 region near to the peak for the {PtN 3 OH 2 } group which had not previously been attributed to any known species in either the perchlorate or acetate reaction systems. Figure 3.15 is the minimized, restraint free, model of the aquaphosphato substituted 1,1/c,c complex. Hydrogen bonding between an oxygen atom of the phosphate group and the hydrogen atom of the aqua ligand is likely, with a distance of 1.89 Å, accounting for the difference in 1 H shift of the {PtN 3 (H 2 O)} group compared to the other aquated species (2 or 3). 92

114 Figure 3.15 The aquaphosphato species (8) of 1,1/c,c demonstrating the possible hydrogen bonding interaction between the aqua and phosphato ligands. The phosphate is di-protonated in this case however it is possible that the mono-protonated speicies would exist at higher ph. The model of the macrochelate phosphato species (10) was constructed to substantiate assignment of the distinct 1 H/ 15 N peak in the trans-nh 3 region close to those of other {PtN 3 PO 4 } groups. 93

115 Figure 3.16 The bridged 1,1/c,c phosphate species (10). The potential energy of the µ-po 4 structure seen in Figure 3.16 reflects its likelihood to occur. Prior to running the simulated annealing the energy of the system was 39.0 kcal/mol which compares favourably to the diaqua system (Figure 3.13) whose energy prior to the molecular dynamics run was 45.6 kcal/mol. The potential energy of these two systems, µ-po 4 and diaqua, are not strictly comparable since the chemical structures are different. However, as an indication of the stability of the macrochelate the energy comparison is of interest since the bridged species is in fact of lower energy suggesting a more stable structure than the extended diaqua species. As can be seen from both Figure 3.13 and Figure 3.15 hydrogen bonding involving the aqua ligand attached to 1,1/c,c acts as a stabilizing factor while in contrast in 1,1/t,t the potential hydrogen bond donors and acceptors are pointing away from each other in the 94

116 theoretically most stable structure where none of the tetrahedral centres in the six carbon chain have eclipsed hydrogen atoms. The 1,1/c,c can adopt two possible orientations, one where the labile ligands are facing each other, or a geometry where one of the platinum square planes is reflected resulting in a loss of interaction between the labile sites. Since all of the bonds in 1,1/c,c are free to rotate the situation where the labile ligands can interact in a hydrogen bonding capacity could result in a favourable energy interaction. It does need to be stated that the alternate conformation would result in the ammine being accessible for hydrogen bonding. Based on an assessment of the geometry of the phosphato/aqua species it is not difficult to see how the three atoms involved in the putative bridge O-P-O are in closer proximity to one another and the possibility of hydrogen bonding stabilizing a transition between the phosphato/aqua and the µ-po 4 species is viable. 3.8 Conclusions The aquation profile of any metal based drug is crucial for a full understanding of its ultimate fate within the body. Knowing if the complex is in fact a pro-drug or that the ligands make up nothing more than a carrier for the metal itself, can be the critical difference between success in the clinic or not. In the case of cisplatin, its aquation has been extensively studied and much is known about the rates and mechanisms by which aquation occurs. Multinuclear platinum complexes provide an interesting opportunity in the study of aquation kinetics. The dinuclear complexes 1,1/t,t and 1,1/c,c differ only in the arrangement of ligands on the platinum atoms. The aquation profiles of 1,1/t,t and 1,1/c,c are quite similar although significantly less aquation occurs for 1,1/c,c. A more significant difference between these two complexes is the pk a values of the aquated species. The pk a of 1,1/c,c is 0.4 ph units higher than that of 1,1/t,t or 95

117 1,0,1/t,t,t, resulting in more of the aquated (i.e. active) species available at physiological ph. The role that ligand substitution may play in affecting transport across cellular membranes is supported by the fact that both 1,1/t,t and 1,1/c,c experience reversible binding with the phosphate and acetate ligands. Hence a phospholipid shuttling mechanism could exist across membranes involving reversible binding to phospholipids. However, 1,1/t,t and 1,1/c,c do not interact with phosphate in the same manner. A phosphate ligand is a more labile when reacting with 1,1/t,t than is acetate, while the reverse is true for 1,1/c,c. In addition the reaction of phosphate with 1,1/t,t is similar to that of 1,0,1/t,t,t. Formation of the bridged phosphato species for the 1,1/c,c complex may be the reason for this. A further unique aspect of the 1,1/c,c reaction with phosphate is its slow approach to equilibrium and again this is attributed to the phosphate-bridged species acting as a reservoir during the reaction. Finally, the bridged macrochelate species formed in the reactions of 1,1/c,c appear to be a unique feature in comparison to the 1,1/t,t. The molecular modelling of these systems demonstrated that the cis geometry places the two labile ligands directed towards each other in a position which is prone to the formation of macrochelate ring structures. 96

118 3.9 References 1. Zhang, J.; Thomas, D. S.; Davies, M. S.; Berners-Price, S. J.; Farrell, N., J. Biol. Inorg. Chem. 2005, 10, Bancroft, D. P.; Lepre, C. A.; Lippard, S. J., J. Am. Chem. Soc. 1990, 112, Johnson, N. P.; Hoeschele, J. D.; Rahn, R. O., Chem.-Biol. Interact. 1980, 30, Reishus, J. W.; Martin, D. S., Jr., J. Am. Chem. Soc. 1961, 83, Hindmarsh, K.; House, D. A.; Turnbull, M. M., Inorg. Chim. Acta 1997, 257, Koubek, E.; House, D. A., Inorg. Chim. Acta 1992, 191, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1989, 161, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1989, 166, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1990, 173, Miller, S. E.; House, D. A., Inorg. Chim. Acta 1991, 187, Miller, S. E.; Wen, H.; House, D. A.; Robinson, W. T., Inorg. Chim. Acta 1991, 184, Miller, S. E.; Gerard, K. J.; House, D. A., Inorg. Chim. Acta 1991, 190, Davies, M. S.; Berners-Price, S. J.; Hambley, T. W., Inorg. Chem. 2000, 39, Berners-Price, S. J.; Frenkiel, T. A.; Frey, U.; Ranford, J. D.; Sadler, P. J., J. Chem. Soc., Chem. Commun. 1992, Guo, Z.; Chen, Y.; Zang, E.; Sadler, P. J., J. Chem. Soc., Dalton Trans. 1997, Murdoch, P. D. S.; Guo, Z.; Parkinson, J. A.; Sadler, P. J., J. Biol. Inorg. Chem. 1999, 4, Chen, Y.; Guo, Z.; Parsons, S.; Sadler, P. J., Chem.--Eur. J. 1998, 4, Chen, Y.; Parkinson, J. A.; Guo, Z.; Brown, T.; Sadler, P. J., Angew. Chem., Int. Ed. 1999, 38, Barton, S. J.; Barnham, K. J.; Habtemariam, A.; Sue, R. E.; Sadler, P. J., Inorg. Chim. Acta 1998, 273, Barton, S. J.; Barnham, K. J.; Frey, U.; Habtemariam, A.; Sue, R. E.; Sadler, P. J., Aust. J. Chem. 1999, 52, Davies, M. S.; Cox, J. W.; Berners-Price, S. J.; Barklage, W.; Qu, Y.; Farrell, N., Inorg. Chem. 2000, 39, Davies, M. S.; Thomas, D. S.; Hegmans, A.; Berners-Price, S. J.; Farrell, N., Inorg. Chem. 2002, 41, Sklenar, V.; Piotto, M.; Leppik, R.; Saudek, V., J. Magn. Reson., A 1993, 102, Piotto, M.; Saudek, V.; Sklenar, V., J. Biomol. NMR 1992, 2, Palmer, A. G., III; Cavanagh, J.; Wright, P. E.; Rance, M., J. Magn. Reson. 1991, 93,

119 26. Rajasekaran, E.; Jayaram, B.; Honig, B., J. Am. Chem. Soc. 1994, 116, Hofmann, A.; van Eldik, R., J. Chem. Soc., Dalton Trans. 2003, Yao, S.; Plastaras, J. P.; Marzilli, L. G., Inorg. Chem. 1994, 33, Cox, J. W. Interactions between multinuclear platinum complexes and DNA with a focus on ligand, linker groups and kinetics: A molecular modelling [ 1 H, 15 N] NMR study. Virginia Commonwealth University, Richmond, Guex, N.; Peitsch, M. C., Electrophoresis 1997, Cason, C. POV-Ray(tm) rendering engine for Windows, 3.1g;

120 4 Studies of 1,4- or 1,6-Interstrand Cross-links Formation by 1,0,1/t,t,t 4.1 Introduction Previous studies by Berners-Price and Farrell followed the stepwise formation of a 1,4- interstrand cross-link by reaction of 1,1/t,t with the duplex 5 -{d(atatgtacatat) 2 } (14XL) by using [ 1 H, 15 N] NMR. 1 This approach allowed the rate constants of this reaction to be determined for each step in the pathway. It was shown that the 1,4-GG interstrand cross-links were the major product formed. H 3 N H 2 N X H 2 N Pt Pt Y NH 3 H 3 N NH 3 Y' n+ 5' DNA 5' 1,1/t,t or 1,0,1/t,t,t Y/Y' G 5 G 4 Cl/Cl (1) (k H ) Cl/Cl [DNA] k H' G 5 ' G 4 ' Cl/H 2 O (2) G/G (5) k CH Cl/H 2 O [DNA] k MF G/Cl (3) 5' 14XL 5' 16XL Figure 4.1 A generic scheme showing the pathway for the binding of multinuclear platinum complexes to DNA. The generalized structure for the platinum complexes demonstrates how the linker (X) may be changed to alter the length of the molecule or its chemical properties. For 1,0,1/t,t,t X = Pt(NH 3 )(NH 2 C 6 H 12 ). Y/Y = Cl and are also the site of covalent binding to the guanine N7. 14XL and 16XL represent the two DNA duplexes used in the study. Prior to the commencement of my candidature other members of the Berners-Price and Farrell groups had performed (but not analysed) [ 1 H, 15 N] HSQC NMR experiments to 99

121 follow the stepwise formation of 1,4- and 1,6-interstrand cross-links by reaction of 15 N- 1,0,1/t,t,t with the sequences 14XL and 5 -{d(tatgtatacata) 2 } (16XL). The objective was to compare the kinetics of formation of 1,4-interstrand cross-links by 1,0,1/t,t,t and 1,1/t,t and to compare the kinetics of formation of 1,4-and 1,6-interstrand cross-links by 1,0,1/t,t,t. My role in this study was (i) to purify the bifunctional adducts by HPLC for characterization by ESI-MS spectrometry and (ii) to characterize the first step of the reaction, the preassociation between 1,0,1/t,t,t and DNA. The latter involved assigning the NOESY spectra of the 14XL and 16XL sequences before and after addition of 1,0,1/t,t,t, and the construction of molecular models to aid in the interpretation of the NMR data. The complete study, including the results of the work described here have been published in the paper by Hegmans et al. 2 The figures from this paper showing 1 H and [ 1 H, 15 N] NMR spectra over the course of these reactions are provided in Appendix A4 and are discussed in the context of the results of my work in Section Materials and Methods Chemicals. HPLC purified DNA oligonucleotides 14XL and 16XL were purchased from OSWEL. The 15 N-1,0,1/t,t,t as a nitrate salt was kindly provided by Professor Nick Farrell. 100

122 Sample Preparation. Reaction of 14XL with 1,0,1/t,t,t. A stock solution of the 14XL duplex was prepared in 500 µl of 5% D 2 O in H 2 O. The concentration of the DNA (14XL) duplex was determined using UV/Vis spectroscopy to be 15.6 mm. This value was obtained using a molar absorption coefficient of ε 260 = M -1 cm -1 based on the method derived by Kallansrud and Ward. 3 The NMR sample was prepared by combining 76 µl of the stock solution of 14XL, 30 µl sodium phosphate buffer (200 mm, ph 5.3) and 2 µl of TSP solution (sodium-3-trimethylsilyl-d 4 -propionate, 13.3 mm) in 272 µl 5% D 2 O/95% H 2 O and placing in a plastic vial for subsequent annealing. The duplex was melted in a water bath ( K) and then allowed to cool slowly to room temperature. A 20 µl aliquot of a freshly prepared solution of fully 15 N-labelled 1,0,1/t,t,t (1.50 mg, 1.20 µmol) in 37.5 µl 5% D 2 O/95% H 2 O was added to the duplex solution resulting in a total volume of 400 µl, with final concentrations of duplex (2.96 mm), phosphate buffer (15 mm) and 1,0,1/t,t,t (1.60 mm). After vortexing, the sample was transferred into a Shigemi D 2 O matched small volume 5 mm NMR tube. The reaction was carried out at 298 K and was followed by 1 H and [ 1 H, 15 N] NMR over a total time of 45 hours. 2 The final ph of the solution was determined to be 5.9. Reaction of 16XL with 1,0,1/t,t,t. A stock solution of the 16XL duplex was prepared in 300 µl of 5% D 2 O in H 2 O. The duplex concentration was estimated spectrophotometrically to be 13.2 mm based on the absorption coefficient of ε 260 = M -1 cm -1 derived using the method of Kallansrud and Ward µl of the DNA stock solution, 33 µl sodium phosphate buffer (200 mm, ph 5.3) and 2 µl of The NMR samples were prepared by Dr. Alexander Hegmans and Dr. Murray Davies and the NMR data were acquired prior to the commencement of my candidature. 101

123 TSP solution were combined with 281 µl 5% D 2 O/95% H 2 O. 20 µl of the solution were removed for ph and UV absorption measurements. The duplex was annealed and allowed to cool to room temperature, after which a 40 µl aliquot of a freshly prepared solution of 15 N-1,0,1/t,t,t (1.14 mg, 0.91 µmol) in 46 µl 5% D 2 O was added to the duplex to reach a total volume of 420 µl, with final concentrations of duplex (2.62 mm), phosphate buffer (15 mm) and 15 N-1,0,1/t,t,t (1.9 mm). The reaction was followed by 1 H and [ 1 H, 15 N] NMR at 298 K until completion after approximately 47 hours. 2 The final ph was determined to be 6.2. Instrumentation. The NMR spectra were recorded on a Varian UNITY-INOVA-600 MHz spectrometer ( 1 H, MHz; 15 N, MHz). A detailed explanation of the acquisition parameters and setup can be found in Chapter 2. Two dimensional NOESY spectra were acquired with 4032 points over a sweep width of 12 khz and zero filled to the next power of two for each of 384 t1 increment (STATES) using a mixing time of 150 ms and a relaxation delay of 2 seconds. Water suppression was performed using the Watergate pulse train incorporated into the NOESY pulse sequence. The ph of the solutions was measured on a Shindengen ph Boy-P2 (su19a) ph meter and calibrated against ph buffers of ph 6.9 and 4.0. The volume of 5.0 µl of the solution was placed on the electrode surface and the ph recorded. Due to chloride leaching from the ph electrode the solution used for measurement was not returned to the bulk sample. Adjustments in ph were made using 0.04 M, 0.2 M, and 1.0 M HClO 4 in 5% D 2 O in H 2 O, or 0.04 M, 0.2 M, and 1.0 M NaOH in 5% D 2 O in H 2 O. NOESY Analysis. The 1 H NMR resonances of the duplexes 14XL and 16XL prior to platination were assigned from the NOESY spectra using established assignment 102

124 strategies for right-handed B-form DNA duplexes. 4 The program Sparky was used to record the sequential assignment and generate the figures. 5 HPLC and ESI Mass Spectrometry. The final products from the reactions of 1,0,1/t,t,t with 14XL and 16XL were analysed by HPLC. For the reaction with the 16XL sequence the 15 N-labelled sample from the NMR experiment was used, whereas for the 14XL sequence the reaction was repeated under identical conditions using unlabelled 1,0,1/t,t,t. The solutions were stored at 277 K for several months, then diluted with H 2 O by a factor of 100 prior to injection onto the HPLC column. A 10 µl or 90 µl aliquot was loaded onto either a Phenomenex Aqua analytical (5 µm, 125 Å particle size, mm) or semi-preparative (5 µm, 125 Å particle size, mm) column and eluted using a Waters 600 gradient pump system with a Waters 486 UV detector. Computer control was maintained with the Millennium 32 software suite provided by Waters. Elutions of the DNA and DNA adducts from the reverse phase HPLC column were performed with acetonitrile in 50 mm ammonium acetate buffer (ph 5.4) at flow rates of 1 (analytical) or 6 (preparative) ml min -1. The gradient profile is included in Table 4.1. The fractions were collected into clean bottles and then extensively dialyzed for 24 hours against cold water (277 K) to remove any salts. The samples were lyophilized and then stored frozen prior to shipment to the University of Arizona where electrospray mass spectrometry was performed. The purified fractions were dissolved in 1 ml of H 2 O/MeOH (1:5) with 5% NH 4 OH. Data were acquired in negative ion mode using an electrospray ionization ion source. 103

125 Table 4.1 The gradient profile used for the elution of DNA adducts in the 14XL and 16XL experiments with 1,0,1/t,t,t. Analytical Time Flow Rate % 50 mm Ammonium % Acetonitrile (min) (ml/min) Acetate Preparative Molecular Modelling. Computer models depicting the electrostatic preassociation of 1,0,1/t,t,t in the minor groove of 14XL and 16XL were generated using HyperChem version Computations were performed using a derivative of the AMBER parameter set and an all-atom force field developed by Yao et al., 6, 7 as described previously. 8 1,0,1/t,t,t was manually docked in the minor groove of 14XL and 16XL with the positioning of the linker based on the observed proton chemical shift changes in the first 1 H NMR spectra obtained immediately after addition of 1,0,1/t,t,t to each sequence (see below). Small adjustments in the orientation of the -CH 2 - chains were made in order to direct the terminal platinum atoms towards the major groove. The system was relaxed between each of these iterations with 100 steps of molecular mechanics using a steepest descent algorithm. The final images were created in SwissPDBViewer v and rendered using POVRay v3.11g Reaction of duplex 14XL with 15 N-1,0,1/t,t,t Analysis of the preassociation step. A NOESY spectrum was obtained before the reaction of 14XL with 1,0,1,/t,t,t. Figure 4.2 depicts the NOESY sequential assignment walk for the 14XL DNA duplex in the H6/H8 aromatic, H1 sugar proton region. 104

126 C8H1' T10H1' T6H1' T2H1' T4H1' G5H1' T12H1' A7H1' A9H1' A1H1' A11H1' A3H1' 6.4 δ 1 H (ppm) Figure 4.2 Contour plot of the 2D NOESY NMR spectrum (mixing time = 150 ms, 5% D 2 O in H 2 O., 298 K) of the 14XL duplex showing the H6/H8 aromatic ( ppm) to sugar ring H1 ( ppm) connectivities. The sequential NOESY walk is shown and the assigned protons are labelled on the axis of the plot. The assignments were used to identify proton resonances in the 1D 1 H spectra for comparison to the first 1 H experiment performed after the addition of 1,0,1/t,t,t. Figure 4.3 shows a comparison of the aromatic regions of the 1 H NMR spectra of the 14XL DNA duplex before and after addition of 1,0,1/t,t,t. Significant changes in chemical shift are observed for the A7 H2 ( δ 0.05) and the T2 and T6 or T10 H1' ( δ 0.03) protons. All of these protons reside in the minor groove of the DNA and with the exception of T10 H1' are located between the two guanine N7 binding sites of the duplex. The aromatic A7 H2 and A7 H8 protons exhibit a downfield chemical shift change after the addition of the platinum complex, which is consistent with a decrease in electron density caused by the close proximity of the positively charged platinum atoms. Significant shifts were also observed for several of the imino protons. A 105

127 downfield shift ( δ 0.013) for the G(5)-C(8') base pair along with a shift ( δ 0.012) for T(4)-A(9') were observed. The base pair A(3)-T(10') was found to have an upfield shift ( δ ). A3 H8 (8.345 ppm) A3 H8 (8.358 ppm) A7 H8 (8.250 ppm) A7 H8 (8.239 ppm) C8 NH 2 (8.051 ppm) C8 NH 2 (8.067 ppm) G5 H8 (7.792 ppm) G5 H8 (7.800 ppm) A7 H2 (7.380 ppm) A7 H2 (7.335 ppm) b a ppm Figure 4.3 The aromatic region of the 1 H spectrum of the 14XL DNA duplex (a) before and (b) immediately after addition of 1,0,1/t,t,t. Peaks which underwent a notable change in chemical shift are annotated on the figure. Characterization of the bifunctional adduct. The HPLC elution profile of the final reaction mixture from the reaction of 1,0,1/t,t,t with the 14XL duplex (Figure 4.4), shows peaks for two major adducts with similar retention times (28.5 and 29 minutes). The two fractions were separated and the ESI mass spectra of both isolated conformers showed peaks attributable to a cross-linked adduct of 1,0,1/t,t,t and the duplex (calculated molecular weight amu). For both fractions peaks from Na + adducts reduced the intensity of the drug-dna adduct peaks but the 4-, 5-, 6- and 7- charge 106

128 states are clearly observed. The mass spectra for both fractions are included in Appendix A4 Figures A4.6 and A free DNA Absorbance bifunctional adducts Time (min) Figure 4.4 HPLC chromatogram (preparative column) of the products from the reaction of 1,0,1/t,t,t with the 14XL duplex (detection wavelength 254 nm). Construction of Molecular Models. Molecular models of the preassociated state were created and are shown in Figure 4.5. Based on the analysis of the initial proton NMR spectrum obtained after addition of 1,0,1/t,t,t to the 14XL solution the central platinum of 1,0,1/t,t,t was placed within the minor groove between the two guanine residues. The HPLC data show the existence of two cross-linked adducts. Similarly, analysis of the [ 1 H, 15 N] HSQC and 1 H NMR spectra from the reaction (Appendix A4) 2 indicate formation of two conformers of the 1,4-interstrand cross-links. Figure 4.5 shows two possible routes around the backbone of the DNA that will allow for the linker to remain within the minor groove and still approach the guanine N7 position in the major groove. 107

129 The difference between these two pathways is the number of intervening phosphate groups. Figure 4.5a depicts the preassociated state with one intervening phosphate group belonging to the C8 nucleotide. While Figure 4.5b has two intervening phosphate groups C8 and A7. Figure 4.5 Two possible routes to the 1,4GG interstrand bifunctional adduct of 1,0,1/t,t,t with 14XL. (a) depicts the possible pathway with one intervening backbone phosphate while (b) show the alternative pathway where two backbone phosphates separate the free ends of 1,0,1/t,t,t as it wraps around the DNA. The prime denotes the complementary strand of the duplex. 4.4 Reaction of duplex 16XL with 15 N-1,0,1/t,t,t Analysis of the preassociation step. A NOESY spectrum was obtained for the 16XL duplex prior to reaction with 1,0,1/t,t,t for the purpose of determining the position within the DNA duplex 3D structure where the 1,0,1/t,t,t complex preassociated. Figure 4.6 depicts the NOESY sequential assignment walk for the 16XL DNA duplex in the H6/H8 aromatic, H1 sugar proton region. 108

130 C9H1' T7H1' T5H1' T3H1' T1H1' T11H1' G4H1' A8H1' A10H1' A6H1' A12H1' A2H1' δ 1 H (ppm) Figure 4.6 Contour plot of the 2D NOESY NMR spectrum (mixing time = 150 ms, 5% D 2 O in H 2 O., 298 K) of the 16XL duplex showing the H6/H8 aromatic ( ppm) to sugar ring H1 ( ppm) connectivities. The sequential NOESY walk is shown and the assigned protons are labelled on the axis of the plot. The assignments were used to identify proton resonances in the 1D 1 H spectra for comparison to the first 1 H experiment performed after the addition of 1,0,1/t,t,t. Significant changes in the proton chemical shifts of 16XL sequence are observed after the addition of the 1,0,1/t,t,t. Figure 4.7 show the changes in the aromatic region of the 1 H spectrum. As seen in the reaction with the 14XL sequence, protons found within the minor groove of the DNA are significantly affected during the preassociation phase. Specifically, the T7 H1 ( δ 0.05), A8 H1 ( δ -0.02) and A8 H2 ( δ 0.09) protons, and possibly the A6 H2 proton are all significantly shifted. Also the T7 H6 ( δ 0.005) and C9 H6 ( δ 0.09) protons located in the major groove were found to be shifted. 109

131 A6 H8 (8.312 ppm) A6 H8 (8.267 ppm) C9 NH2 (8.095 ppm) C9 NH2 (8.081 ppm) A2 H2 / G4 H8 (7.833 ppm) A2 H2 (7.839 ppm) G4 H8 (7.815 ppm) C9 H6 (7.345 ppm) C9 H6 (7.259 ppm) T7 H6 / T11 H6 (7.174 ppm) (7.149 ppm) T7 H6 / T11 H6 (7.144 ppm) b a ppm Figure MHz 1 H NMR spectra of the 16XL duplex showing the aromatic region (a) before and (b) immediately after the addition of 1,0,1/t,t,t. Characterization of the Bifunctional Adduct. The HPLC elution profile in Figure 4.8 shows only one major adduct (retention time 28 minutes). The ESI mass spectrum of the isolated adduct is consistent with a cross-linked adduct of 15 N-1,0,1/t,t,t and the duplex (calculated molecular weight amu) as the major species, with peaks observed attributable to the 4-, 5-, 6-, 7- and 8- charge states of the double stranded adduct. The mass spectrum is included in Appendix A4 Figure A

132 bifunctional adduct Absorbance free DNA Time (min) Figure 4.8 Preparative scale separation of the 16XL-1,0,1/t,t,t product mixture. The peak at ~24.5 minutes corresponds to free/unreacted DNA while the peak at ~28 minutes is due to the major (1,6-GG) cross-linked adduct (detection wavelength 254 nm). Construction of Molecular Models. The changes in proton chemical shift found from the 1 H spectra were used to localize a possible docking site for 1,0,1/t,t,t in the minor groove of the 16XL sequence. In the molecular modelling program HyperChem a standard DNA duplex was generated with the nucleotide sequence corresponding to the 16XL duplex. The central platinum atom of 1,0,1/t,t,t was placed within 5 Å of the A8 H2, T7 H6 and C9 H6 aromatic protons which reside in the minor groove. The G4 H8 proton is presumed to be associated with portions of the platinum complex further away from the central platinum. As can be seen in Figure 4.9 the geometry of the possible bifunctional 1,6-interstrand cross-link is such that the central linker is unable to remain anchored within the minor groove of the duplex once the two terminal platinum atoms are coordinated to the N7 guanine atoms in the major groove in order to form the 1,6- interstrand adduct. 111

133 Figure 4.9 Molecular model depicting the 16XL DNA duplex with 1,0,1/t,t,t electrostatically preassociated in the minor groove of the DNA. The two free ends of the platinum complex are reaching around the backbone in an attempt to make the monofunctional adduct by binding to guanine N7 in the major groove. 4.5 Discussion The combined results from the full 1 H and [ 1 H, 15 N] NMR study of which this work forms a part show that two conformers of the interstrand cross-link were formed in the reaction between 1,0,1/t,t,t and the 14XL duplex while only one conformer was observed in the reaction between 1,0,1/t,t,t and the 16XL DNA duplex. The [ 1 H, 15 N] HSQC NMR spectra for the reaction with the 16XL sequence show only one set of 1 H, 15 N peaks corresponding to the final cross-linked adduct while the spectra obtained in the reaction with the 14XL duplex have two distinct sets of 1 H/ 15 N cross-peaks assigned to the bifunctional adduct of 1,0,1/t,t,t with the 14XL duplex (see Appendix A4 Figure A4.1). The aromatic region of the 1 H NMR spectra (Appendix A4 Figure 112

134 A4.2) show two signals assignable to H8 protons of platinated guanine residues for the 14XL duplex, whereas only a single guanine H8 resonance was observed for the 1,4- GG cross-link formed by 1,1/t,t with the same sequence 11 and one major platinated guanine H8 signal is observed in the case of the 16XL duplex. Similarly, a comparison of the imino region (11 14 ppm) of the proton spectra for both reactions again shows two distinct sets of peaks for the final product from the 14XL reaction while only one set is observed for the 16XL reaction. Finally, the resonances associated with the methylene protons of 1,0,1/t,t,t clearly demonstrate the existence of two conformers of the bifunctional adduct in the 14XL reaction and only one in the 16XL reaction (see Appendix A4 Figure A4.1). HPLC purification of the different products from the two reactions was performed as part of this work and two major adducts were observed in the case of the reaction with the 14XL sequence while only one significant adduct was in evidence for the reaction with the 16XL sequence. The most intense peaks observed by UV/Vis spectroscopy in the reaction with 14XL and with 16XL were collected and subjected to ESI-MS analysis which confirmed the assignment in each case as bifunctional cross-linked adducts of 1,0,1/t,t,t with either 14XL or 16XL. The two conformers of the 1,4-interstrand cross-link are not interconvertible as demonstrated by temperature dependent NMR experiments described in the paper by Hegmans et al. 2 In the work described here molecular modelling was used to help rationalize the behaviour of 1,0,1/t,t,t with the two DNA duplexes. In particular, to aid in the understanding of why two conformers of the 1,4-interstrand cross-link form for 1,0,1/t,t,t whereas a single conformer was observed on reaction of 1,1/t,t with the same sequence. The placement of the linker within the minor groove of each sequence was 113

135 based on evidence found in the first proton spectrum acquired after addition of 1,0,1/t,t,t to the DNA. The assignment of the NOESY spectra allowed for the identification of the specific protons whose resonances were affected by the electrostatic interaction of 1,0,1/t,t,t. Once identified the affected protons were used as anchors to dock the platinum of the central linker into the minor groove of the DNA. This form of electrostatic interactions in the minor groove of DNA are not unknown and examples of 12, 13 this are seen in the complexes studied by Grant Collins. The molecular models presented here provide a mechanism by which the central linker of 1,0,1/t,t,t can preassociate in the minor groove of DNA with the arms reaching around the backbone of the DNA to enable coordination of the terminal platinum atoms to the N7 position of guanine in the major groove. For the 14XL sequence two models were generated that allowed for the central linker to remain within the minor groove yet also provided a pathway by which two separate (and non-interconvertible) bifunctional adducts could be formed from the 14XL duplex requiring only small distortions in the DNA structure. The model allows speculation that the 1,4-interstrand cross-link formed by 1,0,1/t,t,t involves simultaneous binding in both the minor and major grooves. This proposal is supported by the partial NMR solution structure obtained by Qu et al. of the 1,4-GG interstrand cross-link formed by 1,0,1/t,t,t and the 8mer duplex 5 -{d(atgtacat) 2 }. 14 The structure indicates binding to the N7 positions of the two guanines in the major groove and also has NOE contacts between the central linker and protons within the minor groove. A model of the 1,0,1/t,t,t-16XL preassociated system was created that also placed the central linker within the minor groove yet required the removal of the linker in order to 114

136 make a monofunctional or bifunctional adduct involving coordination of guanine N7 in the major groove. The proposed mechanism of formation of the 1,4- and 1,6-interstrand cross-links derived from these models is supported by several pieces of evidence from the NMR analyses of the stepwise formation of the cross-links. 2 As discussed above the comparison of 1 H NMR spectra obtained before and immediately after the addition of 1,0,1/t,t,t to both the 14XL and 16XL DNA duplexes indicate that the linker is situated within the minor groove of the DNA during the preassociation stage. An examination of the 1 H NMR spectra in the region spanning 1 to 2 ppm has shown that protons assigned to the linker methylene groups of 1,0,1/t,t,t have very different time profiles in the reactions of 1,0,1/t,t,t with 14XL and 16XL (see Appendix A4 Figure A4.4). 2 Over the course of the reaction of 1,0,1/t,t,t with 14XL the CH 2 (2/5) proton environment does not change. The chemical shift during the preassociation, monofunctional and bifunctional binding stages is very similar. On the other hand the changes seen in the reaction of 1,0,1/t,t,t with 16XL where the same CH 2 (2/5) protons have distinct environments during the preassociation and monofunctional adduct stages demonstrated by very different 1 H chemical shifts. This suggests that for the reaction of 1,0,1/t,t,t with the 14XL duplex the central platinum or linker of 1,0,1/t,t,t remains locked in the minor groove of the duplex for the duration of the reaction. Whereas the linker region of 1,0,1/t,t,t must exit the minor groove in order to form the bifunctional adduct of 1,0,1/t,t,t with the 16XL duplex. 115

137 4.6 Conclusions The role of preassociation in any DNA complex interaction must be considered, especially in the case of a positively charged complex such as 1,0,1/t,t,t. Both cisplatin and 1,1/t,t have been shown to preassociate with the backbone of DNA. These electrostatic interactions are critical in the overall scheme or pathway that platinum complexes take en route to the formation of coordinated adducts. However, the nature and outcome of the preassociation of 1,0,1/t,t,t in the minor groove of DNA provides a new avenue to investigate in terms of platinum-dna interactions. The complete study of which this work forms a part revealed very little difference in terms of the kinetics of formation of 1,4-GG cross-links by 1,1/t,t and 1,0,1/t,t,t. 2 In addition the kinetics of formation of 1,4-and 1,6-interstrand cross-links by 1,0,1/t,t,t are also similar. However, the nature of the adduct formed appears to be dictated by preassociation of the central linker of 1,0,1/t,t,t in the minor groove of the DNA. In the case of the 1,4-interstrand cross-links formed with 1,0,1/t,t,t simultaneous binding in both the minor and major groove yields two conformers of the bifunctional adduct. This unique DNA binding mode proposed for 1,0,1/t,t,t could account for its much greater potency compared to the dinuclear 1,1/t,t. 116

138 4.7 References 1. Cox, J. W.; Berners-Price, S. J.; Davies, M. S.; Qu, Y.; Farrell, N., J. Am. Chem. Soc. 2001, 123, Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, Kallansrud, G.; Ward, B., Anal. Biochem. 1996, 236, Wijmenga, S. S.; Mooren, M. M. W.; Hilbers, C. W., NMR of macromolecules: a practical approach. In Practical approach series., Roberts, G. C. K., Ed. IRL Press at Oxford University Press: Oxford ; New York, 1993; pp Goddard, T. D.; Kneller, D. G. SPARKY 3, 3.110; University of California, San Francisco: San Francisco, Yao, S.; Plastaras, J. P.; Marzilli, L. G., Inorg. Chem. 1994, 33, HYPERCHEM Computational Chemistry, 5.11; Hypercube, Inc.: 419 Phillip St., Waterloo, Ont., Canada N2L 3X2, Brabec, V.; Kašparková, J.; Vrána, O.; Nováková, O.; Cox, J. W.; Qu, Y.; Farrell, N., Biochemistry 1999, 38, Guex, N.; Peitsch, M. C., Electrophoresis 1997, Cason, C. POV-Ray(tm) rendering engine for Windows, 3.1g; Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.--Eur. J. 2003, 9, Wheate, N. J.; Collins, J. G., J. Inorg. Biochem. 2000, 78, Wheate, N. J.; Cutts, S. M.; Phillips, D. R.; Aldrich-Wright, J. R.; Collins J. G., J. Inorg. Biochem. 2001, 84, Qu, Y.; Scarsdale, N. J.; Tran, M.-C.; Farrell, N. P., J. Biol. Inorg. Chem. 2003, 8,

139 5 Binding of 1,0,1/t,t,t to a DNA Duplex Containing Multiple Guanine Binding Sites: A Competition Study 5.1 Introduction The DNA duplex 5 -d(atacatggtacata)-3 5 -d(tatgtaccatgtat)-3 (dsgg) duplex has been studied extensively using a variety of NMR techniques. Structural determination based on NOESY NMR spectroscopy resulting in the elucidation of the three dimensional NMR solution structure of the dsgg duplex and the structure of the cisplatin adduct formed by binding to the two central guanine residues is described in the work by Parkinson et al. 1 Additional studies 2, 3 describing the reaction of the dsgg duplex with cis-[ptcl 2 (NH 3 )(2-picoline)] and [PtCl 2 (dien)] + have also been conducted. ssgg 5'-d(A 1 T 2 A 3 C 4 A 5 T 6 G 7 G 8 T 9 A 10 C 11 A 12 T 13 A 14 )-3' sscc 3'-d(T 28 A 27 T 26 G 25 T 24 A 23 C 22 C 21 A 20 T 19 G 18 T 17 A 16 T 15 )-5' Figure 5.1 The dsgg duplex and the nucleotide sequence of the complementary single stranded oligonucleotides ssgg and sscc. [ 1 H, 15 N] HSQC NMR spectroscopy has also been employed to investigate the reaction 1, 4-6 of cisplatin with the dsgg duplex. Cisplatin monofunctional binding to the N7 of both G(8) 3' and G(7) 5' guanines was observed with the rate constant favouring the formation of the 3' monofunctional adduct. 5, 6 The formation of the 1,2-GpG intrastrand adduct from the 3' monofunctional adduct in the reaction with the duplex was more rapid than the closure from the 5' monofunctional adduct. 5 In a similar experiment with the single stranded oligonucleotide (ssgg), the rate constant for the chelation step was the same regardless of whether the monofunctional adduct was bound to the 3' or 5' guanine. 118

140 [ 1 H, 15 N] HSQC NMR was used to follow the reaction of the dinuclear 1,1/t,t with the dsgg duplex. 7 A completely different reaction profile was observed compared to the cisplatin reaction. The monofunctional adduct is formed primarily at the 3' base G(8) of the GG dinucleotide pair. The major product was found to be the 5'!5' 1,4 interstrand cross-link between G(8) and G(18). This result is consistent with molecular biology experiments that found 1,1/t,t formed interstrand cross-links exclusively in the 5'!5' direction The flexibility of the am(m)ine linker of 1,1/t,t was considered as a possible source for the variation observed in the bifunctional adduct formed from 1,1/t,t, and the dsgg duplex. Molecular models were built that allowed the visualization of two conformations of the G(8)-G(18) interstrand cross-link. The linker s flexibility and the orientation of the platinum square plane at the site of bifunctional adduct closure resulted in two distinct conformations of the 1,4 interstrand cross-link. These two conformations were assigned by an analysis of the [ 1 H, 15 N] HSQC NMR and NOESY spectra. 7 It has been shown in Chapter 4 that the central platinum of 1,0,1/t,t,t appears to be critical in determining the nature of the bifunctional adduct formed. The preassociation of the charged central linker within the minor groove allowed for the formation of two conformations of the of the 1,4 interstrand cross-link. On the other hand in the case of the 1,6 interstrand cross-link only a single conformation was evident and the minor groove preassociation was not preserved. 12 Insight into the role that the charged central platinum linker plays in the interaction of 1,0,1/t,t,t with DNA is vital if the mechanism of action of 1,0,1/t,t,t is to be understood. 119

141 DNA footprinting analyses performed by Brabec et al. 13 have shown that for 1,0,1/t,t,t interstrand cross-links are formed in both the 3'!3' and 5'!5' directions. The proportion of each directional isomer depends on the length of the cross-link. As the number of base pairs between the two guanine bases increases a preference for the 5'!5' interstrand cross-link over the 3'!3' interstrand cross-link is observed. Since only 5'!5' cross-links are formed for 1,1/t,t it is possible that preassociation of the central linker of 1,0,1/t,t,t in the minor groove could account for the formation of directional isomers. To explore this possibility this chapter describes [ 1 H, 15 N] HSQC NMR experiments following the reaction of 1,0,1/t,t,t with the dsgg duplex. Experiments were carried out under identical conditions to previous studies of 1,1/t,t. 7 Figure 5.2 shows the possible pathways for the reaction given the possibility of both 1,4 and 1,5 interstrand cross-links forming in both the 3'!3' and 5'!5' directions. Y/Y Cl/Cl (1) k H H 2 O k 3' Cl/H 2 O (2) + k 5' 5' G(7)G(8) 3' G(25) G(18) dsgg 3' 5' G(8)/Cl G(7)/Cl (3) (4) k 3'C k 5'C k 3'C k 5'C G(8)/G(18) 1,4 IXL 5'-5' G(8)/G(25) 1,5 IXL 3'-3' G(7)/G(18) 1,5 IXL 5'-5' Bifunctional Adducts (5) G(7)/G(25) 1,4 IXL 3'-3' Figure 5.2 Scheme depicting the possible adducts formed from the reaction of 1,0,1/t,t,t with dsgg duplex. (IXL = interstrand cross-link). 120

142 5.2 Experimental Materials and Methods Chemicals. The 15 N-1,0,1/t,t,t was provided by Professor Nick Farrell and synthesized as described recently. 14 The sodium salts of the HPLC purified oligonucleotide 5 d(atacatggtacata)-3 (ssgg) and 5 -d(tatgtaccatgtat)-3 (sscc) were purchased from OSWEL. Instrumentation. NMR spectra were recorded on a Varian UNITY-INOVA-600 spectrometer ( 1 H, MHz, 15 N, MHz). The 1 H{ 15 N} NMR spectra were recorded using a 5 mm triple resonance probehead equipped with pulsed field gradients. The 1 H spectra were acquired with water suppression using the WATERGATE sequence. 15 Both the 1D 15 N- edited 1 H NMR spectra and 2D [ 1 H, 15 N] HSQC NMR spectra (optimized for 1 J( 15 N, 1 H) = 72 Hz) were recorded using the sequence of Stonehouse et al H chemical shifts were referenced to TSP (sodium 3-trimethylsilyl-[D4]-propionate) and 15 N chemical shifts were calibrated externally against 15 NH 4 Cl (1.0 M in 1.0 M HCl in 5% D 2 O/95% H 2 O). The 15 N signals were decoupled by irradiating with the GARP-1 sequence at a field strength of 1 khz during the acquisition time. Typically for 1D 1 H spectra, 64 or 128 transients were acquired using a spectral width of 12 khz, a relaxation delay of 1.5 seconds, and a line broadening of 1.0 Hz. For kinetics studies involving [ 1 H, 15 N] HSQC NMR spectra, between 4 and 12 transients 121

143 were collected for increments of t 1 (allowing spectra to be recorded on a suitable timescale for the observed reaction), with an acquisition time of seconds, spectral widths of 4 khz in t 2 ( 1 H) and khz in t 1 ( 15 N). 2D spectra were completed in minutes. The 2D HSQC spectra were processed using Gaussian weighting functions in both dimensions, and zero-filling by 2 in the f 1 dimension. The ph of the solutions was measured on a Shindengen ph Boy-P2 (su19a) ph meter and calibrated against ph buffers at ph 6.9 and 4.0. A volume of 5 µl of the solution was placed on the electrode surface and the ph recorded. These aliquots were not returned to the bulk solution in order to eliminate the leaching of Cl -. Adjustments in ph were made using 1.0 M HNO 3 in 5% D 2 O/95% H 2 O, or 1.0 M NaOH in 15% D 2 O/95% H 2 O. Sample Preparation. Stock solutions of each of ssgg and sscc were prepared in 500 µl 5% D 2 O/95% H 2 O. The concentrations were estimated spectrophotometrically to be mm and mm respectively using the absorption coefficient calculated using the method described by Kallansrud and Ward (ε 260 = M -1 cm -1 and M -1 cm -1 for ssgg and sscc respectively). 17 All samples (including buffers, acids etc.) were prepared so that there was a 95% H 2 O/5% D 2 O concentration. Duplicate reactions were prepared in the following manner µl of stock solution of ssgg (12.53 mm) and 53.9 ml of sscc (15.41 mm) were combined with 30 µl sodium phosphate buffer (200 mm, ph 5.3) and 2 µl of TSP solution (13.3 mm) in µl 5% D 2 O/95% H 2 O. The samples was heated in a hot water bath (60-70 C) and subsequently allowed to cool to room temperature. A 20 µl aliquot of a freshly 122

144 prepared solution of 15 N-1,0,1/t,t,t (1.12 mg, 0.90 µmol) in 28.0 µl of 5% D 2 O/95% H 2 O was added to the duplex resulting in a total volume of 400 µl, and a concentration of duplex (2.0 mm), phosphate buffer (15 mm) and 1,0,1/t,t,t (1.6 mm), respectively. The reaction was followed at 298 K by 1 H and [ 1 H, 15 N] NMR over a time period of 46 hours. The final ph of the solution was found to be 6.2. Kinetic runs were carried out at 298 K, and covered the time periods 0 46 hours and 0 19 hours, to maximize the use of NMR time. Samples were maintained at 298 K when not in the spectrometer. Kinetic data were derived from both experiments in which spectra with identical reaction times were collected to confirm that the rates of reaction were identical. Data Analysis. An analysis of the kinetic profile of the reaction between 1,0,1/t,t,t and dsgg was accomplished by measuring the peak volumes from the NH 3 region of the [ 1 H, 15 N] HSQC spectra as previously described. 5, 7, 12 The VNMR software package was used to integrate the peaks and provide the volumes over the course of the experiment. All species depicted in Figure 5.2 result in two non-equivalent peaks in the NH 3 region, except for 15 N-1,0,1/t,t,t due to its symmetry. However, a number of peaks overlap, for example the chloro end of the aquachloro species is coincident with the chloro end of the dichloro species. All species, other than 15 N-1,0,1/t,t,t, gave rise to two NH 3 peaks for the non-equivalent {PtN 3 Y} groups. In some cases, overlap between peaks was significant (as with the peak for the non-aquated {PtN 3 Cl} group of the aquachloro species (2) is coincident with the peak for (1) 18 ), but in these instances, it was only one of the pair of the peaks that was overlapped. Thus, reliable intensities were obtained by doubling the volume of the second (discrete) peak. Comparison of the time dependent changes of the observed 123

145 peaks in the Pt-NH 2 region of the [ 1 H, 15 N] spectra was used to confirm the peak assignments. The appropriate differential equations were integrated numerically, and rate constants determined by a nonlinear optimization procedure using the program SCIENTIST (Version 2.0, MicroMath, Inc.). The errors represent one standard deviation. In all cases the data were fit using appropriate first and second-order rate equations. Example SCIENTIST input files are available in Appendix 5.1. Molecular Modelling. Computer models depicting the electrostatic preassociation of 1,0,1/t,t,t in the minor groove of dsgg were generated based on the solution structure of the DNA duplex published in by Parkinson et al. 1 Computations were performed using a derivative of the AMBER parameter set and an all-atom force field developed by Yao et al., 19, 20 as described previously. 21 1,0,1/t,t,t was manually docked in the minor groove of at two sites of dsgg with the positioning of the linker based on the observed proton chemical shift changes in the first 1 H NMR spectra obtained immediately after addition of 1,0,1/t,t,t (see below). Small adjustments in the orientation of the -CH 2 - chains were made in order to direct the terminal platinum atoms towards the major groove. The system was relaxed between each of these iterations with 100 steps of molecular mechanics using a steepest descent algorithm. The final images were created in SwissPDBViewer v and rendered using POVRay v3.11g Results and Discussion Preassociation. Figure 5.3 shows a comparison of the imino regions of the 1 H spectra of the dsgg duplex before and immediately after the addition of 1,0,1/t,t,t. There is a 124

146 notable shift in three of the guanine imino resonances G(7), G(18) and G(25) with peaks of equivalent intensity remaining at the original chemical shift. The G(8) imino resonance is unaltered. A possible interpretation is that the solution contains a mixture of two preassociated states (in an approximately 1:1 ratio) in which the complex (1,0,1/t,t,t) is aligned in different directions along the DNA (i.e. 5'!5' and 3'!3'). G8 G7 G18 G25 b a ppm δ 1 H Figure 5.3 A comparison of the imino region of the 1 H spectra of the dsgg duplex (a) before and (b) immediately after the addition of 1,0,1/t,t,t. Modelling of Preassociated States. Recently published work of which Chapter 4 forms a part has shown that preassociation of the central linker of 1,0,1/t,t,t in the minor groove of DNA can dictate the nature of the cross-links formed. 12 Therefore models were constructed to visualize the two possible preassociated states. Molecular dynamics (MD) simulations were performed on systems constructed using the NMR solution structure of dsgg as a starting point. 1 The central linker of 1,0,1/t,t,t was docked using Amber 7 to regions in the minor groove between G(7) and G(25) or 125

147 G(8) and G(18), labelled as 3'!3' and 5'!5' orientations, respectively and 2 ns MD simulations were performed. A significant indicator of structural deviation within a MD simulation can be found in the RMSD of distance. This statistic provides an indication of the overall mobility of the system. Figure 5.4 contains the RMSD distance plots for both the 5'!5' and the 3'!3' orientations. The atoms selected for the RMSD analysis were the heavy atoms of the DNA only. This avoids the influence of the highly mobile solvent molecules and the hydrogens atoms of the DNA, additionally the 1,0,1/t,t,t molecule was excluded from this analysis. For both orientations of 1,0,1/t,t,t within the minor groove it is seen that the RMSD varies about an average of approximately 3.5 Å. It is significant that the observed fluctuation of the RMSD does not show dramatic variation over the time frame of the MD simulation. This suggests that the movement of the platinum complex in and out of the minor groove does not substantially alter the 3D structure of the DNA as a whole. 5.0 a 5.0 b RMSD Distance (Å) RMSD Distance (Å) Time (ps) Time (ps) Figure 5.4 The RMSD plot depicting the movement of the DNA duplex phosphate backbone for the 2 ns production MD simulation of (a) the G(8)-G(18) (5'!5') configuration and (b) the G(7)-G(25) (3'!3') configuration. 126

148 Another indication that the 1,0,1/t,t,t complex can reside within the minor groove of DNA without significant structural distortion can be found in an analysis of the energy profile obtained during the MD simulation. Figure 5.5 contains the plots of the total energy of the two systems studied. The energy profiles do not contain any indication of major distortions resulting in a fluctuation in the total energy of the systems. The total energy is a sum of both the kinetic and potential energies of the system during the MD simulation x10 5 a x10 5 b x x10 5 Total Energy (kcal/mol) x x x10 5 Total Energy (kcal/mol) x x x x x x x Time (ps) Time (ps) Figure 5.5 The total energy plots from the 2 ns preassociation production MD run of (a) the G(8)-G(18) (5'!5') configuration and (b) the G(7)-G(25) (3'!3') configuration. Dynamic behaviour of the 3'!3' preassociated state. The central linker moved into the minor groove and appeared to be anchored near the phosphate backbone. The complex adopts two distinct conformations during the simulation. The first being the linear structure where the complexes lie along the minor groove. The second structure has the arms of the complex extending across the minor groove and backbone of both strands of the DNA while the central platinum remains at least above the minor groove. Dynamic behaviour of the 5'!5' preassociated state. A very similar pattern is observed in the 5'!5' simulation where both motifs exist, that of the linear complex lying in the minor groove and the second system with the complex normal to the line of 127

149 the minor groove. Interestingly the insertion of the central platinum linker into the minor groove in the 5'!5' system seems to have been more successful. The insertion is much more strongly held. Figure 5.6 contains two representative states taken from the MD simulations demonstrating the minor groove preassociation. Both the 5'!5' and the 3'!3' preassociated states demonstrated high levels of stability while located within the minor groove. Figure 5.6 The 1,0,1/t,t,t complex aligned in the minor groove of dsgg in both the (a) 3'!3' orientation, (960 ps) and (b) the 5'!5' orientation (1200 ps). The guanine residues are indicated in yellow. [ 1 H, 15 N] NMR experiments. The reaction of 1,0,1/t,t,t with dsgg was followed by [ 1 H, 15 N] HSQC NMR over a period of 46 hours. Two regions in the [ 1 H, 15 N] HSQC spectra are observed. The NH 2 and NH 3 regions are distinct from each other with 15 NH 2 128

150 peaks between δ 15 N -40 to -50 ppm and 15 NH 3 peaks in the range of δ 15 N -60 to -70 ppm. The proximity of the NH 2 region to the 1 H 2 O resonance at ~4.8 ppm in the 1 H dimension is unfortunate as resonances due to the substituted species tend to be masked by the residual water peak. In addition, the integration of these peaks close to the water resonance is unreliable as their intensity is diminished by the water suppression. Initially peaks are observed for the dichloro species (1) at δ 3.90/-64.3 in the NH 3 region and δ 5.05/-47.1 in the NH 2 region for the terminal {PtN 3 Cl} groups and at δ 4.28/-63.3 for the NH 3 groups of the central linker. A small peak at δ 4.18/-62.2 is observed in the first spectrum recorded after 18 minutes and is assigned to the terminal {PtN 3 O} group of the aquachloro species (2). Its partner peak is assumed to be coincident with the peak for the dichloro species (1). As the reaction progresses, peaks assignable to at least two monofunctionally bound species are observed. A peak assigned to the {PtN 3 Cl} group (unbound end) of the combined monofunctional adducts (labelled 3b/4b) is seen at δ 3.92/-64.5 adjacent to the peak corresponding to the dichloro starting material (1). The deshielding by δ 0.02 ppm is similar to that observed for the monofunctional bound adducts of 1,1/t,t with this duplex 7 and is indicative of an electrostatic interaction with the DNA. Two peaks appear (each with similar time dependent behaviour as the peak corresponding to 3b and 4b) and are assigned to the Pt-NH 3 groups bound to guanine N7 in two different monofunctional adducts. The 1 H/ 15 N chemical shifts for the first peak (3a) appear at δ 4.19/-59.7 and the second (4a) at δ 4.31/ The two peaks have very similar chemical shifts to peaks assigned to 3' and 5' monofunctional adducts in the reaction of 1,1/t,t with the dsgg duplex 7 at δ 4.14/-59.9 and δ 4.30/ Based on these assignments 3a is assigned to the 3' G(8) monofunctional adduct and 4a to the 5' G(7) 129

151 monofunctional adduct. Interestingly, the peak assigned to the 3' monofunctional species (3a) develops an asymmetry in the early stages of the reaction indicative of two distinct environments which could arise from adducts aligned in different directions along the DNA. A similar asymmetry was not observed for the reaction with 1,1/t,t. The chemical environment of the N7 guanine bound end of each monofunctional species is not substantially different to the environment of the bifunctional adduct that is formed from it. For this reason the 1 H/ 15 N peaks of the {PtN 3 N 7 G} group of the monofunctional (3 or 4) and bifunctional species (5) are overlapped. Of principle importance is the fact that the monofunctional adduct concentration can be derived from the integration of the peak (3b/4b) corresponding to the unbound end of the monofunctional adduct which is free from overlap. At the end of the reaction there are two major peaks (δ 4.32/-60.7 and δ 4.18/-59.5) in the PtNH 3 (end) region assignable to bifunctional adducts. Three peaks observed at δ 4.54/-60.1, δ 4.49/ 59.6 and δ 4.44/ 59.8 that account for less than 7% of the total product profile have been designated as minor products (6) and excluded from the kinetic analysis to follow. Kinetic Analysis. Since the [ 1 H, 15 N] HSQC peaks representing the N 7 -G bound end of the monofunctional species (3a, 4a) are coincident with the peaks for the bifunctional adduct (5) it was not possible to determine the concentration of the individual monofunctional adducts. Their total concentration was derived from the discrete peak (3b, 4b) for the unbound {PtN 3 Cl} end. 130

152 15 NH 2 15 NH 3 (a) 2.05 h 3,4 1 5 H 2 O Linker 4a 2a 3b,4b 1 3a (b) 6.68 h 5 Linker 3b,4b 3,4 1 2a 1,2b 4a 3a (c) h 5 Linker 3b,4b 3,4 1 4a,(5) 1,2b 3a,(5) (d) h Linker Minor Products 5 Figure 5.7 2D [ 1 H, 15 N] HSQC NMR (600 MHz) spectra at 298 K of the reaction of 15 N-1,0,1/t,t,t during the reaction with the dsgg duplex after (a) 2.05 (b) 6.68 (c) and (d) hours. Peaks are assigned to the NH 3 and NH 2 groups of structures 1-5 as shown in Figure 5.2. Linker refers to the central PtN 4 unit of 1,0,1/t,t,t while the other identified peaks are attributed to the terminal NH 2 or NH 3 groups. a and b refer to non-equivalent terminal {PtN 3 Y} groups in non-symmetrical species, where Y=Cl (a) and H 2 O or N 7 -guanine (b). represents artefacts associated with very intense cross peaks. 131

153 k H Cl/Cl Cl/H 2 O k MF G/Cl k BF G/G (1) (2) (3,4) (5) Figure 5.8 Kinetic scheme used to model the rate constants for the reaction of 1,0,1/t,t,t with the dsgg duplex. The reaction pathway for the kinetic model used is shown in the scheme depicted in Figure 5.8. The rate constants obtained are listed in Table 5.1 and the computer best fits for the rate constants are shown in Figure 5.9 Table 5.1 Rate constants determined for the reaction of 1,0,1/t,t,t with the dsgg duplex according to the kinetic model shown in Figure 5.8. k H (10-5 s -1 ) k MF (M -1 s -1 ) k BF (10-5 s -1 ) 1,0,1/t,t,t 3.43 ± ± ± 0.2 1,1/t,t ± ± ± 0.04 The rate constant for combined monofunctional adduct formation is ca. 2-fold higher for 1,1/t,t compared to 1,0,1/t,t,t. A possible explanation is that for 1,0,1/t,t,t strong preassociation in the minor groove alters the approach to the N7 of guanine in the major groove. A similar lowering of the rate constant for monofunctional binding was observed for 1,0,1/t,t,t binding to the 1,4-GG duplex in comparison to 1,1/t,t. 12 It is surprising that, the rate constant for conversion of the combined monofunctional species to the combined bifunctional adducts is two fold higher than that for the formation of the 5'!5' 1,4GG interstrand cross-link in the reaction of 1,1/t,t with duplex dsgg. 7 In all other kinetic studies of interstrand cross-linking by 1,1/t,t and 1,0,1/t,t,t performed to date 7, 12 the rate constant for bifunctional adduct formation has been found 132

154 Time (hours) Figure 5.9 Kinetic fit of the combined model for 1,0,1/t,t,t forming bifunctional adducts with dsgg. Key combined bifunctional adducts (5), + free dsgg, " 1,0,1/t,t,t, # Cl/H 2 O (2),! combined monofunctional adducts (3,4). to be similar in magnitude to the rate constant for the aquation consistent with aquation of the monofunctional adduct being the rate determining step. Characterisation of Products. Due to time constraints only limited analysis of the products from the reaction was possible. A NOESY spectrum of the final product was obtained but could not be analysed due to the complexity which was exacerbated by the presence of unreacted dsgg in the sample. A sample was prepared using similar conditions to the kinetic experiment however the ratio of DNA:drug was increased to 2:1.8 from the previous ratio of 2:1.6 in order to obtain larger quantities of product to improve the quality of the NOESY spectra. However, the NOESY spectrum obtained was still too complicated to obtain any structural information on the nature of the final adducts. Figure 5.10 shows the imino region of the 1 H NMR spectra of the final products of the reaction of 1,0,1/t,t,t with the dsgg duplex in comparison to that obtained from the identical reaction with 1,1/t,t. 7 In the latter case analysis of the NOESY spectrum showed the presence of one major adduct, the 5'!5' 1,4 interstrand cross-link between 133

155 G(8) and G(18) in two conformational forms. 7 The greater complexity of the resonances observed in the imino region indicates that more than one adduct is formed in the reaction with 1,0,1/t,t,t. AT GC a * * * * b G25 (b) G8 (b) G7 (b) G18 (b) * * * * ppm δ 1 H Figure 5.10 Comparison of the imino region of the 1 H spectra at equilibrium containing a mixture of final products and reactants from the reaction of (a) 1,0,1/t,t,t and (b) 1,1/t,t with dsgg (the latter taken from reference 7 ). * Imino resonances from GC base pairs in unreacted dsgg duplex. 5.4 Conclusions The reaction of 1,0,1/t,t,t with the dsgg duplex is more complicated than the reaction of 1,1/t,t with the same sequence. Analysis of the initial 1 H NMR spectrum from the reaction, combined with molecular modelling provides evidence for two different 134

156 preassociated states in which the 1,0,1/t,t,t is aligned in different directions along the DNA. The asymmetry of the 1 H/ 15 N peaks for the major monofunctional adduct in the [ 1 H, 15 N] HSQC spectrum is also consistent with two different orientations of the monofunctional adduct which may be coordinated to the N7 of the G(8) residue. The 1 H NMR spectrum of the final product indicates more than one major adduct is formed, as opposed to only the 5'!5' G(8)-G(18) 1,4 interstrand cross-link formed by 1,1/t,t. It is possible that interstrand cross-links are formed in both 3'!3' and 5'!5' directions consistent within the results of DNA footprinting experiments performed by Brabec and co-workers where a directional dependency based on separation of interstrand guanine sites was observed. 13 Further work is needed to explore this possibility, in particular to characterize the products of the reaction of 1,0,1/t,t,t with the dsgg duplex. The use of HPLC purification to isolate the bifunctional adducts of the reaction followed by analysis of their NOESY spectrum ultimately leading to a 3D structure of each bifunctional adduct would be the next logical step to undertake. 135

157 5.5 References 1. Parkinson, J. A.; Chen, Y.; Del Socorro Murdoch, P.; Guo, Z.; Berners-Price, S. J.; Brown, T.; Sadler, P. J., Chem.--Eur. J. 2000, 6, Chen, Y.; Parkinson, J. A.; Guo, Z.; Brown, T.; Sadler, P. J., Angew. Chem., Int. Ed. 1999, 38, Murdoch, P. D. S.; Guo, Z.; Parkinson, J. A.; Sadler, P. J., J. Biol. Inorg. Chem. 1999, 4, Barnham, K. J.; Berners-Price, S. J.; Frenkiel, T. A.; Frey, U.; Sadler, P. J., Angew. Chem., Int. Ed. Engl. 1995, 34, Berners-Price, S. J.; Barnham, K. J.; Frey, U.; Sadler, P. J., Chem.--Eur. J. 1996, 2, Reeder, F.; Guo, Z.; Murdoch, P. D. S.; Corazza, A.; Hambley, T. W.; Berners-Price, S. J.; Chottard, J.-C.; Sadler, P. J., Eur. J. Biochem. 1997, 249, Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.--Eur. J. 2003, 9, Zou, Y.; Van Houten, B.; Farrell, N., Biochemistry 1994, 33, Zaludová, R.; Zakovská, A.; Kašpárková, J.; Balcarová, Z.; Kleinwächter, V.; Vrána, O.; Farrell, N.; Brabec, V., Eur. J. Biochem. 1997, 246, Kašpárková, J.; Nováková, O.; Vrána, O.; Farrell, N.; Brabec, V., Biochemistry 1999, 38, Kasparkova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2000, 275, Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, Kasparkova, J.; Zehnulova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2002, 277, Qu, Y.; Harris, A.; Hegmans, A.; Petz, A.; Kabolizadeh, P.; Penazova, H.; Farrell, N., J. Inorg. Biochem. 2004, 98, Piotto, M.; Saudek, V.; Sklenar, V., J. Biomol. NMR 1992, 2, Stonehouse, J.; Shaw, G. L.; Keeler, J., J. Biomol. NMR 1994, 4, Kallansrud, G.; Ward, B., Anal. Biochem. 1996, 236, Davies, M. S.; Cox, J. W.; Berners-Price, S. J.; Barklage, W.; Qu, Y.; Farrell, N., Inorg. Chem. 2000, 39, HYPERCHEM Computational Chemistry, 5.11; Hypercube, Inc.: 419 Phillip St., Waterloo, Ont., Canada N2L 3X2, Yao, S.; Plastaras, J. P.; Marzilli, L. G., Inorg. Chem. 1994, 33, Brabec, V.; Kašparková, J.; Vrána, O.; Nováková, O.; Cox, J. W.; Qu, Y.; Farrell, N., Biochemistry 1999, 38,

158 22. Guex, N.; Peitsch, M. C., Electrophoresis 1997, Cason, C. POV-Ray(tm) rendering engine for Windows, 3.1g;

159 6 Modelling Studies of 3'-3' vs. 5'-5' Interstrand Crosslinking 6.1 Introduction Preassociation, or the electrostatic interaction between DNA and 1,0,1/t,t,t, has been encountered as a major focus of investigation in Chapters 4 and 5. This preassociation appears to play an important role in determining the direction and nature of the cross-link formed between duplex DNA and 1,0,1/t,t,t. Viktor Brabec and co-workers have performed a series of experiments using gel electrophoresis to study the relationship between 5'!5' and 3'!3' bifunctional adduct formation by 1,0,1/t,t,t in oligonucleotide sequences where both interstrand cross-links are possible. 1 Figure 6.1 shows the nucleotide sequences used in the study by Brabec et al. 1 The abbreviations VB12, VB14 and VB16 will be used to refer to the sequences throughout this chapter. VB12 5 -d(tctcctattcgcttatctctc)-3 5 -d(gagagataagcgaataggaga)-3 VB14 5 -d(tctccttcttgttcttcctcc)-3 5 -d(ggattaagaacaagaaggaga)-3 VB16 5 -d(ctctctctattgttatctcttct)-3 5 -d(agaagagataactatagagagag)-3 Figure 6.1 The three sequences used in the study by Viktor Brabec. 1 The guanine residues highlighted in bold are those involved in the interstrand cross-linking. In this study Brabec et al. demonstrated that the VB12 sequence almost exclusively formed the 3'!3' cross-link. This is an unexpected result since this type of adduct is virtually unheard of in DNA cross-linking agents. The VB14 sequence had an equal preference for cross-links in either the 5'!5' or the 3'!3' direction (Figure 6.2), whereas the VB16 sequence demonstrated a mark preference (70%) for the formation of a 5'!5' adduct. 138

160 1 5`-d( T C T C C T T C 3'-3' 11 5'-5' T T G T T C T T C C T 21 C C)-3` 3`-d(A G A G G A A G A A C A A G A 29 A G G A G G)-5` 22 Figure 6.2 A DNA sequence depicting the possible 5'!5' and 3'!3' interstrand cross-links. The sequence is that used by Viktor Brabec 1 and referred to here as the VB14 sequence. In Chapter 4, the variation in product profile observed for the reaction of 1,0,1/t,t,t with the 14XL and the 16XL sequences suggested differing mechanisms of binding. Analysis of the initial 1 H spectra led, in part, to the preassociation hypothesis; that being, the process by which the central platinum of the trinuclear complex inserts into the minor groove, resulting in two non-interconvertible routes to a bifunctional adduct in the case of the 14XL sequence, whereas only one such route is possible for the 16XL sequence. Preassociation with the dsgg sequence was studied in Chapter 5 where the complex was found to form more than one bifunctional adduct. These combined results led to the conclusion that the unfavourable 3'!3' form of the bifunctional adduct was in fact possible in the case of binding of 1,0,1/t,t,t to the dsgg duplex because the central platinum preassociated in the minor groove thus providing an anchor which allowed the formation of bifunctional adducts in both directions. In this chapter the dependency of the number of intervening base pairs between potential cross-links with relation to the preference to form 5'!5' or 3'!3' bifunctional adducts is investigated using molecular modelling techniques. The Brabec gel electrophoresis experiments demonstrated a trend, that of increasing separation between the two guanines favouring 5'!5' interstrand cross-link formation. These experiments used single stranded DNA during the monofunctional binding step followed by annealing of the complementary strand and subsequent formation of the bifunctional 139

161 adduct. The 3D structure of single stranded DNA is not well defined and molecular mechanics/dynamics modelling of these structure is even less well understood. In addition our primary interest is to understand the role the central linker of 1,0,1/t,t,t plays in dictating the direction and nature of interstrand cross-linking. For these reasons the modelling protocol described in Chapter 2 and implemented in Chapter 5 was also applied to the three Brabec sequences: VB12, VB14 and VB16. However, to allow a more direct comparison to the results presented by Brabec et al. 1 a single monofunctional adduct was also generated and subjected to a molecular dynamics simulation where the central linker was not resident within the minor groove. 6.2 Materials and Methods Sequence Generation The initial coordinates of the three DNA duplexes: VB12, VB14 and VB16 were generated using the standard B form DNA parameters supplied by the Amber program. A full description of the molecular dynamics methodology can be found in Chapter 2. Unless otherwise stated the methods employed are those detailed in Chapter 2. The models were generated using the nukit and nucgen functions of Amber 7. Hydrogen atoms were added and relaxed while the heavy atoms were initially restrained. The structure was fully minimized using a distance dependent dielectric constant to mimic water. Sodium counter ions were added using xleap from the Amber 7 suite. Again minimization was performed first on the newly added sodium atoms and then on the entire system. Addition of the 1,0,1/t,t,t into the model system was made using xleap. 140

162 6.2.2 Docking The docking of 1,0,1/t,t,t into the minor groove of a given sequence was accomplished using the NMR restraints capability within the Sander program. Specific atoms in both the minor groove of the DNA sequence and in the central linker region of 1,0,1/t,t,t were selected. These atoms were assigned a separation of 5 Å and a force constant for this pseudo bond was provided. The actual sander input file is provided in Appendix A The positioning of the central linker took place over a 20 ps molecular dynamics simulation. The rationale for the placement of the central linker in the minor groove was based on the work performed with the dsgg sequence. In that case the central linker was placed in the minor groove between the two guanine base pairs. As with the dsgg sequence this strategy resulted in two preassociated systems for each DNA sequence. The docking of 1,0,1/t,t,t into the minor groove of VB12, VB14 and VB16 occurred in two regions referred to as 5'!5' and 3'!3'. The final position of the drug is not subjected to any restraints, periodic boundary conditions were in force and a visual inspection at the completion of the docking phase was the last check before production dynamics were ready to proceed. The models seen in Figure 6.6 are the end result of this process Equilibration Molecular Dynamics Standard periodic boundary conditions were applied to the system. The resulting cubic system was submitted to molecular dynamics simulations. All equilibration, docking and production molecular dynamics simulations were performed either on the super computer at Canberra or on the IVEC computer in Western Australia. Appendix A2.2 contains full copies of the scripts used to perform these simulations. Equilibration of the system was performed over 200 picoseconds. A variety of energy and structural 141

163 data were considered to ensure that the system was at equilibrium. Examples of the scripts used to obtain or process output data derived from an molecular dynamics run can be found in Appendix A Production Molecular Dynamics As before Sander was used for the molecular dynamics calculations. Here a selfsubmitting script was used to perform the extended calculations. This was required for a number of reasons the most important of which are data security and increased priority on the queuing system employed at the Australian Partnership for Advanced Computing (APAC). An example script is available in Appendix A The length of the individual molecular dynamics slices varied depending upon the size of the system being studied and how many water molecules were contained in the periodic box. Four hours of wall or running time on 4 cpus provided the best economy in terms of access, efficient use of computer resources and the allocation given us by in the APAC Merit Allocation Scheme. Production dynamics was performed on these systems until at least 2 nanoseconds were recorded. In total six systems, employing the three DNA duplexes were studied and reported here Analysis A variety of programs were used to analyse the structural and energetic properties of these DNA/1,0,1/t,t,t preassociated states. In all relevant cases the values being quoted are comparisons between a bound and a free state. Full analysis of the equilibrated 142

164 DNA sequence was also performed in order to assess the impact on the structure of the modelling process. Where applicable relative values are reported. The programs ptraj and carnal, both supplied with Amber, are tools for assessing the relative motion of the atoms in a system. Using both of these programs the root mean square deviation of the DNA atom positions over the course of the simulation has been determined. Sander generates a large amount of data relating to energy and physical properties during the molecular dynamics simulation. A perl script originally created by David Case of the Scripps Research Institute and modified by myself extracts these values for the course of a simulation (refer to Appendix A2.3.3 and script directory on accompanying CD for complete details). The program 3DNA created by Xiang-Jun Lu performs a complete structural analysis of DNA in several forms. 2 From these data parameters describing the duplex being simulated are extracted such as tilt, twist, roll etc. To aid in the visualization of these systems movies were made of the molecular dynamics simulations using VMD. The frames of these movies are made up from the checkpoint or restart files saved approximately every 50 ps during the molecular dynamics simulations. These movies are included on the CD accompanying this thesis. 6.3 Results Sequence Generation and Equilibration The first criterion to be assessed when determining whether a system has reached equilibrium is the root mean square of the deviation (RMSD) of distance of the structure. The internal motion of the system relative to itself is an indication of heavy atoms within the system stability for a given set of conditions. It is generally accepted 143

165 that an RMSD of between 3 and 4 Å for a DNA duplex can be taken as stable. 3-7 This value would decrease to below 2 Å for a globular protein. Figure 6.3 depicts the change in RMSD for the duration of the equilibration simulation. Two significant factors should be noticed. First the structure appears to go through an initial phase where the RMSD is quite small although changing rapidly. This region corresponds to the time where positional restraints existed in the system or in the time immediately following their removal. The second fact of note is that the RMSD does not appear to be completely stable. This is not an absolute requirement since the system is going to be subjected to 2 nanoseconds of production molecular dynamics simulation, therefore suggestion of approach to equilibrium conditions is sufficient. 5 a 5 b 5 c RMSD Distance (Å) 3 2 RMSD Distance (Å) 3 RMSD Distance (Å) Time (ps) Time (ps) Time (ps) Figure 6.3 The RMSD of distance plots for the equilibration molecular dynamics simulations of (a) VB12 (b) VB14 and (c) VB16 DNA sequences determined using the ptraj trajectory analysis program in Amber VB12 Sequence Pre-Docking Equilibration. The pre-docking phase of the molecular dynamics protocol is used to ensure that the addition of the platinum complex 1,0,1/t,t,t is at a distance that does not destabilize or impact in any other way on the duplex DNA. This distance between the complex and the DNA should be sufficient to rule out any nonbonding interactions. Figure 6.4 depicts the stabilization of two systems where the complex, 1,0,1/t,t,t is ~10 Å away from the DNA. Both systems attain a reasonable 144

166 equilibrium within the 200 ps simulation. The variation observed has two major contributing factors, the first being the mobility of the complex and second the relative weakness of the DNA strand interactions. As opposed to the forces constraining the 3D structure of globular proteins the H-bonding forces holding DNA strands together are rather weak, especially in short oligonucleotide sequences such as the ones used in this study. 4 a 5 b 3 4 RMSD Distance (Å) 2 RMSD Distance (Å) Time (ps) Time (ps) Figure 6.4 Plots of the RMSD of distance for the pre-docking equilibration molecular dynamics simulations of VB12 with 1,0,1/t,t,t (a) in the 3'!3' direction and (b) in the 5'!5' direction derived from ptraj. As seen with the RMSD, the total energy of each pre-docking system reached an equilibrium state before the end of the equilibration phase (Figure 6.5). What is interesting to observe is that the total energy appeared to reach equilibrium almost immediately after the initiation of the constant pressure and volume molecular dynamics phase, which are finished after 30 ps. As will be seen the total energy of a molecular dynamics system is not a particularly useful indicator of an equilibrium state especially when taken as an average. However, the minima and maxima have proven to be useful aids in locating structures of interest. 145

167 a b Total Energy (kcal/mol) Total Energy (kcal/mol) Time (ps) Time (ps) Figure 6.5 Plot of the total energy for the pre-docking molecular dynamics equilibrations of VB12 oligonucleotide sequence in (a) the 3'!3' and (b) the 5'!5' preassociated orientations. Docking. The placement of the platinum of the central linker with the proper alignment in the minor groove of the DNA helix was verified visually after converting the docked structure into a PDB file. Figure 6.6 shows the initial structures used in the production dynamics. It can be seen clearly from Figure 6.6 that the central platinum of 1,0,1/t,t,t is not able to reside in between the two guanine residues. For the purpose of comparison the central linker region was positioned just outside the guanine region within the minor groove. The placement is not without justification as in Chapter 5 it was shown that the drug extends along and within the minor groove during the preassociation phase which preceded the formation of monofunctional and bifunctional adducts. 146

168 Figure 6.6 The complex 1,0,1/t,t,t docked in the minor groove of VB12 in (a) the 5'!5' orientation and (b) the 3'!3' orientation prior to the initiation of production molecular dynamics. The sections of the DNA duplex depicted as yellow space filling atoms are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. Production Molecular Dynamics. After the systems were equilibrated, production molecular dynamics was performed. These simulations typically ran for 1500 to 2000 ps. As was seen in the equilibration molecular dynamics runs both the RMSD and the total energy have been reported. Figure 6.7 shows the RMSD for both the 5'!5' and the 3'!3' systems. As opposed to the equilibration molecular dynamics results where the value tended to an equilibrium maximum, here the systems oscillate around that 147

169 maximum value. Interestingly, that equilibrium value is substantially lower than in the equilibrium case by up to 2 Å. The cause of this may be the formation of stabilizing hydrogen bonds while within the minor groove although no direct evidence is available to confirm this theory. The total energy as shown in Figure 6.8 for both systems is relatively constant over the course of the simulation. However, the high energy spikes do correlate in time to the areas of greater variation in RMSD. The higher values of RMSD distance correlate with a more flexible structure, while the lower values in Figure 6.7 correspond to more rigid structures. 6 a 6 b 5 5 RMSD Distance (Å) RMSD Distance (Å) Time (ps) Time (ps) Figure 6.7 Plots of the RMSD of distance from the production molecular dynamics simulations of the VB12 DNA duplex with 1,0,1/t,t,t bound in (a) the 3'!3' orientation and (b) the 5'!5' direction. The fluctuations of the RMSD suggest that the DNA is maintaining a duplex conformation with significant distortion occurring at times where the RMSD is > 5 Å. Data obtain by analysis of the molecular dynamics trajectory using ptraj. 148

170 The total energy plots shown in Figure 6.8 suggest a much more stable arrangement than might be suspected from the RMSD plots. Although in both cases a high-energy spike does correspond to an RMSD maximum at 800 ps in the 3'!3' case and at two time points, 600 ps and 1200 ps in the 5'!5' case a b Total Energy (kcal/mol) Total Energy (kcal/mol) Time (ps) Time (ps) Figure 6.8 Plots of the total energy sampled over the course of the production molecular dynamics simulation of the VB12 DNA sequence with 1,0,1/t,t,t preassociated in the minor groove of the duplex in (a) the 3'!3' and (b) 5'!5' orientations. The total energy of the system is essentially consistent except in a few areas where spikes occur. These spikes do not indicate structures that are of interest rather structures that were quickly excluded from the molecular dynamics simulation. The regions of moderate and more sustained changes in energy may suggest valid pathways into intermediate conformations of interest. Movies of the VB12-5'!5' and VB12-3'!3' trajectories were generated using VMD and are located on the accompanying CD. Observation of the VB12-5'!5' simulation reveals that the platinum of the central linker moves out of the minor groove and associates with the phosphate backbone while the terminal platinum stays in the minor groove. At the end of the simulation the 1,0,1/t,t,t was aligned completely within the minor groove. In the early stages of the VB12-3'!3' molecular dynamics simulation the central platinum of 1,0,1/t,t,t stays in or over the minor groove while the terminal arms extend over both of the phosphate backbones of the DNA duplex. The central platinum of 1,0,1/t,t,t stays roughly anchored in the location it occupied at the start of both of the simulations. 149

171 Figure 6.9 Structures obtained after the completion of the production molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB12 DNA duplex (a) in the 5'!5' orientation, where only half of the complex is preassociated within the minor groove and (b) in the 3'!3' orientation where only the central linker region of the complex remains positioned over the minor groove, in this case allowing the alkyl chains of the complex to reach around the backbone of the DNA duplex. While these structures may not be the most energetically significant or most likely to result in potential monofunctional or bifunctional adducts this figure does represent a large percentage of the conformations that occur during the molecular dynamics simulation. The sections of the DNA duplex depicted in yellow are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. The red nucleotide represents the 5' end of the DNA duplex VB14 Sequence Pre-Docking Equilibration. The pre-docking equilibration of the two VB14-1,0,1/t,t,t systems were performed and the RMSD (Figure 6.10) and the total energy (Figure 6.11) both comply with the expected limits. 150

172 3.5 a 4 b RMSD Distance (Å) RMSD Distance (Å) Time (ps) Time (ps) Figure 6.10 Plots of the RMSD of distance for the pre-docking equilibration molecular dynamics simulations of VB14 with 1,0,1/t,t,t (a) in the 3'!3' direction and (b) in the 5'!5' direction derived from ptraj a b Total Energy (kcal/mol) Total Energy (kcal/mol) Time (ps) Time (ps) Figure 6.11 Plot of the total energy for the pre-docking molecular dynamics equilibrations of VB14 oligonucleotide sequence in the (a) 3'!3' and (b) 5'!5' preassociated orientations. Docking. For the VB14 sequence both the 5'!5' and 3'!3' docking structures resulted in the central platinum of 1,0,1/t,t,t being positioned within the minor groove between the appropriate guanine residues. The docked structures of both VB14 systems can be seen in Figure

173 Figure 6.12 The equilibrated docked configurations of the VB14 duplex with 1,0,1/t,t,t in (a) the 3'!3' and (b) the 5'!5' orientations. Equilibration of the docked complex sometimes does not completely remove the impact of the docking procedure. The exaggerated V conformation seen in the 3'!3' figure is due to incomplete relaxation of the complex. This does not result in any problems as the complex completes the relaxation process during the initial stages of the production molecular dynamics simulation. The sections of the DNA duplex depicted in yellow space filling atoms are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. Production Molecular Dynamics. After the systems were equilibrated, production dynamics was performed. These simulations typically ran for 1800 to 2500 ps. As was seen in the equilibration molecular dynamics runs, both the RMSD and the total energy are reported. Figure 6.13 shows the RMSD for both the 5'!5' and the 3'!3' systems. As opposed to the equilibration molecular dynamics results where the value tended to an equilibrium maximum, here the systems oscillate around that maximum value. 152

174 The total energy as shown in Figure 6.14 for both systems is relatively constant over the course of the simulation. However, the high-energy spikes do correlate in time to the areas with extreme values of RMSD. The higher values of RMSD distance correlate with a more flexible structure. While the lower values in Figure 6.13 correspond to more rigid structures. At 800 and 1500 ps in the VB14-3'!3' molecular dynamics simulation two energy spikes where observed which correlated to minima in the RMSD plots (Figure 6.13). The VB14-5'!5' simulation displayed an energy maximum and high RMSD at 400 ps. 5 a 5 b 4 4 RMSD Distance (Å) 3 2 RMSD Distance (Å) Time (ps) Time (ps) Figure 6.13 Plots of the RMSD of distance from the production molecular dynamics simulations of the VB14 DNA duplex with 1,0,1/t,t,t bound in (a) the 3'!3' orientation and (b) the 5'!5' direction. The fluctuations of the RMSD suggest that the DNA is maintaining a duplex conformation with significant distortion occurring at times where the RMSD is > 5 Å. Data obtained by analysis of the molecular dynamics trajectory using ptraj. 153

175 a b Total Energy (kcal/mol) Total Energy (kcal/mol) Time (ps) Time (ps) Figure 6.14 Plots of the total energy sampled over the course of the production molecular dynamics simulation of the VB14 DNA sequence with 1,0,1/t,t,t preassociated in the minor groove of the duplex in the (a) 3'!3' and (b) 5'!5' orientations. The total energy of the system is essentially consistent except in a few areas where spikes occur. These spikes do not indicate structures that are of interest rather structures that were quickly excluded from the molecular dynamics simulation. The regions of moderate and more sustained changes in energy may suggest valid pathways into intermediate conformations of interest. During the production phase of the molecular dynamics simulations no restraints were placed on the atoms in the DNA-1,0,1/t,t,t system, thus allowing all atoms to move freely within the bounds of the Amber force field. The two movies, VB14 3'!3' and VB14 5'!5', included on the accompanying compact disc depict the movement of the complex within the minor groove. In general, 1,0,1/t,t,t adopts an elongated conformation. This is a similar structure to the one observed when the complex is subjected to the Amber force field in isolation. In addition the central platinum linker remain lodged within or over the minor groove of the DNA in almost all cases. The instances when the platinum complex left the minor groove were few, but interestingly the complex did find its way back into the minor groove with time. 154

176 Figure 6.15 Structures obtained after the completion of the production molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB14 DNA duplex (a) in the 5'!5' orientation, where only half of the complex is preassociated within the minor groove and (b) in the 3'!3' orientation where only the central linker region of the complex remains positioned over the minor groove, in this case allowing the alkyl chains of the complex to reach around the backbone of the DNA duplex. While these structures may not be the most energetically significant or most likely to result in potential monofunctional or bifunctional adducts this figure does represent a large percentage of the conformations that occur during the molecular dynamics simulation. The sections of the DNA duplex depicted in yellow are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. The red nucleotide represents the 5' end of the DNA duplex. Specifically the VB14-3'!3' movie demonstrates the alignment of 1,0,1/t,t,t along the phosphate backbone of the DNA duplex. While alignment within and along the minor groove of the duplex is observed during the VB14-5'!5' movie the most significant orientation found during the molecular dynamics simulation is the wrapping of the terminal arms of 1,0,1/t,t,t around the backbone of the duplex. This is the best example of this transition state of moving from central linker minor groove preassociation to simultaneous minor/major groove binding. 155

177 6.3.4 VB16 Sequence Pre-docking Equilibration. The pre-docking equilibration of the two VB16-1,0,1/t,t,t systems were performed and the RMSD (Figure 6.16) and the total energy (Figure 6.17) both fall with in the expected limits a 4.00 b RMSD Distance (Å) RMSD Distance (Å) Time (ps) Time (ps) Figure 6.16 Plots of the RMSD of distance for the pre-docking equilibration molecular dynamics simulations of VB16 with 1,0,1/t,t,t (a) in the 3'!3' direction and (b) in the 5'!5' direction derived from ptraj a b Total Energy (kcal/mol) Total Energy (kcal/mol) Time (ps) Time (ps) Figure 6.17 Plot of the total energy for the pre-docking molecular dynamics equilibrations of VB16 oligonucleotide sequence in (a) the 3'!3' and (b) the 5'!5' preassociated orientations. Docking. For the VB16 sequence both the 5'!5' and 3'!3' docking structures resulted in the central platinum of 1,0,1/t,t,t being positioned within the minor groove between 156

178 the appropriate guanine residues. Figure 6.18 contains representations of both the 5'!5' and 3'!3' docked systems. Figure 6.18 Docking position for (a) the 5'!5' and (b) the 3'!3' preassociated states of 1,0,1/t,t,t with the VB16 sequence. The sections of the DNA duplex depicted in yellow space filling atoms are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. Production Molecular Dynamics. After the systems were equilibrated, production dynamics was performed. These simulations typically ran for 2000 ps. As was seen in the equilibration molecular dynamics runs, both the RMSD and the total energy are reported. Figure 6.20 shows the RMSD for both the 5'!5' and the 3'!3' systems. As 157

179 opposed to the equilibration molecular dynamics results where the value tended to an equilibrium maximum, here the systems oscillate around that maximum value. The total energy as shown in Figure 6.19 for both systems is relatively constant over the course of the simulation, only the scale suggests a difference in the systems. Unlike the previous systems the VB16 production simulations do not demonstrate any correlation between the RMSD and total energy of the dynamics simulations. The RMSD of both the 5'!5 and the 3'!3' simulations, as seen in Figure 6.20, do have a higher absolute value of RMSD suggesting a more mobile system which would be consistent with the longer DNA duplex of VB16. a b Figure 6.19 Plots of the total energy sampled over the course of the production molecular dynamics simulation of the VB16 DNA sequence with 1,0,1/t,t,t preassociated in the minor groove of the duplex in (a) the 3'!3' and (b) the 5'!5' orientations. The total energy of the system is essentially consistent except in a few areas where spikes occur. These spikes do not indicate structures that are of interest rather structures that were quickly excluded from the molecular dynamics simulation. The regions of moderate and more sustained changes in energy may suggest valid pathways into intermediate conformations of interest. 158

180 5 a 6 b RMSD Distance (A) 3 2 RMSD Distance (A) Time (ps) Time (ps) Figure 6.20 Plots of the RMSD of distance from the production molecular dynamics simulations of the VB16 DNA duplex with 1,0,1/t,t,t bound in (a) the 3'!3' orientation and (b) the 5'!5' direction. The fluctuations of the RMSD suggest that the DNA is maintaining a duplex conformation with significant distortion occurring at times where the RMSD is > 5 Å. Data obtained by analysis of the molecular dynamics trajectory using ptraj. During the production phase of the molecular dynamics simulations no restraints were placed on the atoms in the DNA-1,0,1/t,t,t system, thus allowing all atoms to move freely within the bounds of the Amber force field. The two movies, VB16 3'!3' and VB16 5'!5', included on the accompanying CD depict the movement of the complex within the minor groove. In general, 1,0,1/t,t,t adopts an elongated conformation. This is a similar structure to the one observed when the complex is subjected to the Amber force field in isolation. In addition the central platinum linker remain lodged within or over the minor groove of the DNA in almost all cases. The instances when 1,0,1/t,t,t left the minor groove were few, but interestingly the complex did find its way back into the minor groove with time. 159

181 Figure 6.21 Structures obtained after the completion of the production molecular dynamics simulation of 1,0,1/t,t,t docked in the minor groove of the VB16 DNA duplex (a) in the 5'!5' orientation, where only half of the complex is preassociated within the minor groove and (b) in the 3'!3' orientation where only the central linker region of the complex remains positioned over the minor groove, in this case allowing the alkyl chains of the complex to reach around the backbone of the DNA duplex. While these structures may not be the most energetically significant or most likely to result in potential monofunctional or bifunctional adducts this figure does represent a large percentage of the conformations that occur during the molecular dynamics simulation. The sections of the DNA duplex depicted in yellow are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. The red nucleotide represents the 5' end of the DNA duplex. Early in the VB16-3'!3' movie produced from the molecular dynamics trajectory it was observed that 1,0,1/t,t,t lines up along the backbone of the DNA. Although the central platinum does move out of the minor groove ultimately the entire complex rolls deep into the minor groove causing a widening of the minor groove. Ultimately the central platinum remains in the minor groove as the terminal arms start to wrap around the phosphate backbones of the DNA duplex. Similarly the VB16-5'!5' movie shows in the early stages significant segments of time where the terminal arms of the 1,0,1/t,t,t wrap around the phosphate backbone of the DNA duplex. Later in the simulation the 160

182 1,0,1/t,t,t aligns along the minor groove. The significant observation here is the relatively static position of the central platinum Monofunctional Adduct Production Molecular Dynamics The study by Brabec and co-workers 1 which indicated a preference between the formation of 5'!5' and 3'!3' cross-links by 1,0,1/t,t,t depending on the number of intervening guanine base pairs was performed by first generating the monofunctional adducts on the single stranded oligonucleotides and then annealing to the appropriate complementary DNA strand. The modelling of single stranded DNA adducts is not a reasonable proposition since the Amber force field is designed to model the tertiary structure of duplex DNA. Single stranded DNA with few exceptions lacks any predictable secondary structure. A direct comparison to the gel electrophoresis experiments was therefore not possible. The closest system that could be modelled was one where 1,0,1/t,t,t is bound to a guanine N7 forming a monofunctional adduct of the VB14 duplex. Figure 6.22 is a representation of the starting structure used in the molecular dynamics simulation. The orientation of the 1,0,1/t,t,t with respect to the helix was approximately orthogonal to the global helical axis. It was desired that the initial configuration of the adduct would not impart any bias in terms of a preferred orientation adopted during the molecular dynamics simulation. 161

183 Figure 6.22 Initial conformation of the 5' guanine N7 bound monofunctional adduct of VB14-1,0,1/t,t,t. The initial orientation of the 1,0,1/t,t,t is approximately orthogonal to the axis of the DNA duplex in order to prevent the imparting of any bias based on the starting position of the molecular dynamics simulation. The sections of the DNA duplex depicted in yellow are the guanine residues. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a smaller gold atom. Over the course of the molecular dynamics simulation the previously observed variation of the RMSD was again seen in the experiment with the monofunctional system until 162

184 approximately 1600 ps (see Figure 6.23). At this point a large increase in the RMSD was observed reaching a maximum of 6 Angstroms. This large increase in mobility of the heavy or backbone atoms was very long lived, nearly 200 ps, and is suggestive of a larger scale change in the structure of the DNA duplex RMSD Distance (A) Time (ps) Figure 6.23 Plot of the RMSD of distance from the production molecular dynamics simulations of the monofunctionally bound 1,0,1/t,t,t adduct of VB14. Data obtain by analysis of the molecular dynamics trajectory using ptraj. The RMSD of the distance for the monofunctionally bound 1,0,1/t,t,t adduct of VB14. Figure 6.24 shows two structures taken from the time period where the RMSD was very large. The 2 ns simulation proved a useful introduction into the modelling of monofunctionally bound DNA adducts. As with all simulations performed on the preassociated systems the kinetic energy was provided in the form of 300 K of thermal 163

185 Figure 6.24 Two structures that demonstrate the dramatic changes observed around 1600 ps of the molecular dynamics simulation. In yellow is the central guanine residue of the VB14 duplex. The 1,0,1/t,t,t molecule is also shown in a space filling representation with the carbon atoms (white), hydrogen (light blue), nitrogen (dark blue), platinum (gold) and the terminal chloro ligand is displayed as a larger white atom. The red nucleotide represents the 5' end of the DNA duplex. energy. By comparing standard DNA parameters it was shown that no distortion was induced anywhere along the length of the duplex, even at the monofunctional binding site. Interestingly this region does not correlate to a high or low energy system identified in Figure Specifically the two systems at 700 and 1200 ps were also considered. 164

186 Total Energy (kcal/mol) Time (ps) Figure 6.25 The total energy plot for the production molecular dynamics simulation of 1,0,1/t,t,t bound monofunctionally to VB14. The most interesting observation from the 1,0,1/t,t,t-VB14 monofunctional molecular dynamics simulation was the apparent preference for the monofunctional adduct to align itself along the backbone of the DNA to which it was bound in the 5'!5' direction. This has to be considered as a preliminary finding as the experiment was only performed once and only with the VB14 DNA sequence. Does this alignment corroborate the findings of Brabec et al. 1 or is it simply the random motion of a tethered molecule at 300 K? For these results to be ultimately meaningful a large number of simulations would be required with a variety of original configurations of the monofunctional adduct. Examination of the movie generated from the production molecular dynamics simulation showed a distinct tendency to align in the 5'!5' direction along the phosphate backbone of the VB14 DNA duplex. This alignment occurs later in the simulation and generally is consistent with the uncharacteristically large change in the RMSD observed in Figure Interestingly, this significant physical alteration does 165

187 not correspond to any substantial change in the energy profile obtained from the simulation. 6.4 Discussion Three DNA sequences each containing two distinct preassociated adducts of 1,0,1/t,t,t have been considered in this work. The variation between the three sequences, VB12, VB14 and VB16 is found in the number of bases that separate the central guanine from the two distal guanines, in both the 3' and 5' directions. For the VB14 and VB16 sequences a direct analogy to the dsgg duplex can be made. The dsgg duplex has either three intervening bases when considering the longest possible cross-link (1,5-GG interstrand) or two intervening bases in the case of the shorter (1,4-GG interstrand) cross-link. The VB14 duplex is directly analogous to the shorter cross-link in the dsgg system in that is contains two bases between the cross-link sites. The VB16 duplex, while containing 4 intervening bases has a longer distance between guanine residues it is still useful for comparison as part of the series of DNA duplexes being studied. However, for the VB12 sequence no such intervening bases exists. It has been shown that as this separation between the guanine bases decreases the preference for 3'!3' binding increases. 1 Several observations were made from the analysis of the movies created from the molecular dynamics simulations. Three modes of interaction between 1,0,1/t,t,t and the DNA duplex were observed (i) backbone alignment (ii) minor groove insertion and (iii) wrap around association. No indication of a pattern was found from the visual analysis of the movies. Alignment along the phosphate backbone is easily understood, the negatively charged oxygens of the phosphate groups interact electrostatically with the 166

188 positively charged platinum centres. Insertion into the minor groove is less easily rationalized but appears in the molecular dynamics simulations to be more closely akin to being locked over the minor groove between the two phosphate backbones. If this is the case then the wrap around association is simply a progression from minor groove insertion to where the entire 1,0,1/t,t,t complex is aligned along and within the minor groove to an orientation where only the central platinum remains locked in the minor groove. The molecular dynamics simulations of both VB12 systems show that the central linker is retained within or over the minor groove of the duplex. This free movement of the central platinum of the 1,0,1/t,t,t is an indication that with the VB12 duplex, which lacks the intervening bases, the platinum complex is not anchored as strongly within the minor groove. The 1,0,1/t,t,t complex in the VB14 molecular dynamics simulations would appear to be firmly aligned along one of the phosphate backbones and would leave this position and adopt a wrap around motif. The terminal platinum atoms reach around the phosphate backbone into the major groove of the duplex. This was seen very clearly in the VB14-5'!5' simulation. Analysis of the atomic positional fluctuations during the production dynamics simulations of both the preassociated states of 1,0,1/t,t,t with the VB14 sequence demonstrates a similar response to the complex being positioned within the minor groove. 167

189 In the study by Brabec et al. 1 monofunctional adducts were prepared on single stranded DNA sequences containing a central guanine residue. Once monofunctional adduct formation was complete the complementary DNA strand was annealed and the bifunctional adducts subsequently formed were monitored by gel electrophoresis. As with the HPLC-based experiments of Chottard 8 where the formation of bifunctional DNA adducts of cisplatin were monitored by annealing the complementary strand of the DNA to the already formed monofunctional adduct, questions of the applicability to biological systems arise. Strictly speaking does the bifunctional adduct formed in this process truly represent what occurs when the platinum complex binds to duplex DNA within the nucleus? Probably not, although an equally strong argument can be made for the case where the formation of the monofunctional adduct occurs during the replication phase of the DNA where the platinum complex has access to single stranded DNA. In addition, no molecular modelling package exists that can simulate the annealing of DNA in any realistic fashion. This is only a minor concession to the methodology since the Brabec study analysed the subsequent bifunctional adduct formation after annealing had occurred. An interesting opportunity for future work presented itself from the results of the monofunctional simulation. If a preference exists between the formation of 5'!5' and 3'!3' bifunctional adducts the question can be asked, is it a kinetically or thermodynamically controlled process? It is hard to conceive of why this preference exists at all but if it is kinetically based a variation in the simulation temperature should statistically vary the ratio of 5'!5' to 3'!3' adducts indicated by the orientation adopted by the monofunctional adduct. However, if the process is thermodynamically 168

190 controlled a comparison of the free energies of 5'!5' and 3'!3' monofunctional adducts may be relevant. On comparison of the energy, RMSD profiles and trajectory movies (included on back cover CD) several interesting structures were identified in each molecular dynamics simulation. Analysis of the DNA parameters derived from 3DNA for these particular slices in the trajectories have given us a few interesting ideas about the preassociation process. Based on the analysis of the trajectories alone it can be seen that 1,0,1/t,t,t does move around within the minor groove and can, at the simulation temperature of 300 K, move out of the minor groove and reach across the backbone of the duplex. The localized distortion of the DNA caused by the insertion of the drug into the minor groove may be the cause for the directional preference. The Brabec experiments (where single stranded DNA was used to form the monofunctional adduct of 1,0,1/t,t,t then annealed to its complementary strand) differs as discussed earlier to the method used here. A directional preference was observed in the formation of the bifunctional adducts by Brabec where the preference for 5'!5' increased with an increase in separation of the interstrand guanines. The approach was to preassociate the drug with duplex and investigate any potential preference in directional binding based on the preassociated state. If any molecular basis is to be found in the preference for the directionality of bifunctional adduct formation is to be established using in silico techniques a large number of replicates will be required. 6.5 Conclusions Electrostatic association within the minor groove of duplex DNA is clearly demonstrated in the six molecular dynamics simulations presented. In all cases the 169

191 1,0,1/t,t,t complex remain aligned in, over or across the minor groove the duplex. No preference was observed in the alignment of the complex that would suggest a bias in the formation of 5'!5' or 3'!3' cross-links. However, as seen in Chapters 4 and 5 the preassociation of the central linker within the minor groove can act to anchor the platinum complex and thus allow the formation of a wide variety of interstrand crosslinks. The simplest explanation for the preference for the formation of 5'!5' interstrand cross-links observed by Brabec et al. is that geometrically the 3'!3' interstrand crosslink becomes increasingly more difficult to make as the separation between the two guanine residues increases. In all of the above cases it can be seen that the preassociated species in the minor groove is sufficiently energetic to move and align itself in a variety of orientations. The interactions that hold the complex within the minor groove are sufficient to maintain a general trend on keeping the complex more or less localized within the minor groove. A more detailed study of the electrostatic characteristics of the minor groove and any sequence dependencies that might exist would be a useful next step in understanding the nature of the interaction of trinuclear platinum complexes with duplex DNA. 170

192 6.6 References 1. Kasparkova, J.; Zehnulova, J.; Farrell, N.; Brabec, V., J. Biol. Chem. 2002, 277, Lu, X.-J.; Shakked, Z.; Olson, W. K., J. Mol. Biol. 2000, 300, Cheatham, T. E., III. Realistic simulation of nucleic acids in solution (DNA, RNA) Spector, T. I.; Cheatham, T. E., III; Kollman, P. A., J. Am. Chem. Soc. 1997, 119, Cieplak, P.; Cheatham, T. E., III; Kollman, P. A., J. Am. Chem. Soc. 1997, 119, Cheatham, T. E., III; Kollman, P. A., Book of Abstracts, 213th ACS National Meeting, San Francisco, April , COMP Cheatham, T. E., III; Kollman, P. A., J. Mol. Biol. 1996, 259, Reeder, F.; Guo, Z.; Murdoch, P. D.; Corazza, A.; Hambley, T. W.; Berners-Price, S. J.; Chottard, J. C.; Sadler, P. J., European journal of biochemistry / FEBS 1997, 249,

193 A1 Chapter 1 Appendix Figure A1.1 Monofunctional adduct of 1,1/t,t with dsgg sequence. The publication Berners-Price, S. J.; Davies, M. S.; Cox, J. W.; Thomas, D. S.; Farrell, N., Chem.--Eur. J. 2003, 9, and the associated supporting information can be found on the accompanying CD in the publications directory. 172

194 A2 Chapter 2 Appendix 2.1 Amber Parameter Sets frcmod.101ttt The frcmod file is used by Amber to provide additional parameters to the force field. Rather than modifying the basic parameter set which could result in errors being introduced into the calculations. The following three sections are the actual frcmod input files for 1,0,1/t,t,t and the two dinuclear complexes 1,1/t,t and 1,1/c,c. This is the additional/replacement parameter set for 101/ttt MASS PT platinum NM sp3 N bound to platinum BOND PT-N John Cox PT-NM John Cox PT-Cl John Cox N3-H John Cox H3-NM John Cox PT-NB John Cox CT-NM John Cox ANGLE H3-NM-H John Cox NM-PT-Cl John Cox NM-PT-N John Cox Cl-PT-N John Cox PT-NM-CT John Cox PT-N3-H John Cox PT-NM-H John Cox PT-NB-CT John Cox H3-N3-H John Cox NM-PT-NB John Cox NB-PT-N John Cox CT-CT-NM John Cox HC-CT-NM John Cox CT-NM-H John Cox N3-PT-N John Cox NM-PT-NM John Cox DIHE X -CT-NM-X John Cox CT-CT-NM-H CT-CT-NM-PT Marzilli 94 HC-CT-NM-PT Marzilli 94 X -PT-N3-X John Cox X -PT-NM-X John Cox X -PT-NB-X John Cox X -N3-PT-NB John Cox X -CH-NM-C John Cox X -NB-PT-X John Cox PT-N3-NB-N John Cox PT-N3-NM-N John Cox PT-N3-Cl-N John Cox 173

195 NONBON NM John Cox R is ~half of Marzilli 94 PT John Cox R is ~half of Marzilli frcmod.11tt This is the additional/replacement parameter set for 11/tt MASS PT platinum NM sp3 N bound to platinum BOND PT-N John Cox PT-NM John Cox PT-Cl John Cox N3-H John Cox H3-NM John Cox PT-NB John Cox CT-NM John Cox ANGLE H3-NM-H John Cox NM-PT-Cl John Cox NM-PT-N John Cox Cl-PT-N John Cox PT-NM-CT John Cox PT-N3-H John Cox PT-NM-H John Cox PT-NB-CT John Cox H3-N3-H John Cox NM-PT-NB John Cox NB-PT-N John Cox CT-CT-NM John Cox HC-CT-NM John Cox CT-NM-H John Cox N3-PT-N John Cox DIHE X -CT-NM-X John Cox CT-CT-NM-H CT-CT-NM-PT Marzilli 94 HC-CT-NM-PT Marzilli 94 X -PT-N3-X John Cox X -PT-NM-X John Cox X -PT-NB-X John Cox X -N3-PT-NB John Cox X -CH-NM-C John Cox X -NB-PT-X John Cox PT-N3-NB-N John Cox PT-N3-NM-N John Cox PT-N3-Cl-N John Cox NONBON NM John Cox R is ~half of Marzilli 94 PT John Cox R is ~half of Marzilli frcmod.11cc This is the additional/replacement parameter set for 11/cc 174

196 MASS PT platinum NM sp3 N bound to platinum NI sp3 N NR sp3 N NL sp3 N BOND PT-NI John Cox PT-NR John Cox PT-NL John Cox PT-Cl John Cox NR-H John Cox H3-NI John Cox H3-NL John Cox PT-NB John Cox CT-NL John Cox ANGLE H3-NI-H John Cox H3-NR-H John Cox H3-NL-H John Cox NR-PT-Cl John Cox NR-PT-NB John Cox NB-PT-NI John Cox Cl-PT-NL Jaroslav NL-PT-NB John Cox NI-PT-NL John Cox NI-PT-NR John Cox NL-PT-NR John Cox NI-PT-Cl John Cox PT-NR-H John Cox PT-NL-H John Cox PT-NI-H John Cox PT-NB-CT John Cox PT-NL-CT John Cox CT-CT-NL as in trans HC-CT-NL as in trans CT-NL-H as in trans DIHE X -CT-NR-X John Cox X -CT-NL-X John Cox X -CT-NI-X John Cox CT-CT-NL-H as in trans CT-CT-NL-PT as in trans Marzilli 94 HC-CT-NL-PT as in trans Marzilli 94 X -PT-NR-X John Cox X -PT-NL-X John Cox X -PT-NI-X John Cox X -PT-NB-X John Cox PT-NI-Cl-NR John Cox NONBON NI John Cox R is ~half of Marzilli 94 NR John Cox R is ~half of Marzilli 94 NL John Cox R is ~half of Marzilli 94 PT John Cox R is ~half of Marzilli Acquisition Scripts 175

197 Appendix is the file or script used to execute a molecular dynamics run on the APAC supercomputer. This was required since the computer system managed its allocation of time using a queuing system called PBS. The script specifies the number of CPUs used and automatically resubmits the file for execution after a fixed time. This is required for two reasons, logistical and safety. The queue system would not allow a job to run with the rime required for our simulations and safety because if the system goes down the simulation can be restarted just prior to the failure point Self submitting script for MD on APAC SC #!/bin/tcsh #PBS -P f63 #PBS -l walltime=08:00:00 #PBS -l vmem=1000mb #PBS -l ncpus=4 #PBS -l software=amber7 #PBS -l other=mpi #PBS -l other=rms #PBS -v NJOBS,NJOB,LJOB,RUN #PBS -wd #PBS -q normal # =========================================================================== # Self resubmitting PBS tcsh script: # # * Submits a follow on job before running the current job. The follow on # job will be in the "H"eld state until the current job completes # # * Assumes program being run is checkpointing at regular intervals and is # able to resume execution from a checkpoint # # * Does not assume the program will complete within the requested time. # # * Uses an environment variable (NJOBS) to limit the total number of # resubmissions in the sequence of jobs. # # * Allows the early termination of the sequence of jobs - just create/touch # the file STOP_SEQUENCE in the jobs working directory. This may be done # by the executable program when it has completed the "whole" job or by hand # if there is a problem # # * This script may be renamed anything (<15 characters) but if you use the -N # option to qsub you must edit the qsub line below to give the script name # explicitly # # * To use: # - place script in a sub-directory of /home to help with house keeping # - place all input files in a sub-directory of /short/f63/your_directory # with the same name as the script # - # - make appropriate changes to the PBS options above and to the # execution and file manipulation lines belo # - submit the job with the appropriate value of NJOBS, eg: # qsub -v NJOBS=5 <scriptname> 176

198 # - the script will create a tar file with the name $PBS_JOBNAME.tar # in /short/f63/your_directory # # * If restarting a simulation at a mid-point set: # - NJOB = last completed run # - LJOB = (last - 1) completed run # # * Any references to /short/f63/your_directory in the script must be altered # to your particular directory on the short file system # # * To kill a job sequence, either touch the file STOP_SEQUENCE or qdel # the held job followed by the running job # # * To test, try "sleep 100" as your executable line # # * This script is written for the SC at apac. The project identifier # #PBS -P f63 must be changed to #PBS -P ivec0004 to be used on carlin # # =========================================================================== # Where do the echos go? set ECHO=/usr/bin/echo # # These variables are assumed to be set: # NJOBS is the total number of jobs in a sequence of jobs (defaults to 1) # NJOB is the number of the current job in the sequence (defaults to 1) # LJOB is the number of the previous job in the sequence (defualts to 0) # RUN is the working directory (defaults to /short/f63/dst/$pbs_jobname) # if (! $?NJOBS ) then $ECHO "\nnjobs (total number of jobs in sequence) is not - defaulting to 1\n" setenv NJOBS 1 endif if (! $?NJOB) then $ECHO "\nnjob (previous job number in sequence) is not set - defaulting to 1\n" setenv NJOB 1 endif if (! $?LJOB) then $ECHO "\nljob (previous job number in sequence) is not set - defaulting to 0\n" setenv LJOB 0 endif if (! $?RUN) then $ECHO "\nrun (working directory in short) is not set - defaulting to /short/f63/dst/pbs_jobname\n" setenv RUN "/short/f63/dst/$pbs_jobname" endif # # Quick termination of job sequence - look for a specific file # (the filename could be qsub -v argument) # if (-f STOP_SEQUENCE) then $ECHO "\nterminating sequence after $NJOB jobs\n" exit 0 endif 177

199 # # Increment the counters to get current job number njob++ setenv NJOB ljob++ setenv LJOB $ljob # # Are we in an incomplete job sequence - more jobs to run? # if ( $njob < $NJOBS ) then # # Now submit the next job # (Assumes -N option not used to change job name.) njob++ # WHY DOES THIS WORK? $ECHO "\nsubmitting job number $njob in sequence of $NJOBS jobs\n" qsub -W depend=afterany:$pbs_jobid $PBS_JOBNAME else $ECHO "\nrunning last job in sequence of $NJOBS jobs\n" endif # # File manipulation prior to job commencing, eg. clean up previous output # files, check for consistency of checkpoint files,... # if ($NJOB > 2) then # This should exclude 1st run therefore eo_files directory already exists gzip $RUN/md$LJOB.mdvel $RUN/md$LJOB.mdcrd mv $PBS_JOBNAME.* $RUN/eo_files # tar -rf /short/f63/dst/$pbs_jobname.tar $RUN/md$LJOB.mdvel.gz $RUN/md$LJOB.mdcrd.gz # rm -f $RUN/md$LJOB.md* endif # # Now run the job... # #[===================================================================] #[ # Stuff for MPI jobs - REMOVE ] #[ # ] #[ # set this in case an MPI process crashes ] #[ # ] setenv RMS_EXITTIMEOUT 60 #[ ] #[ # ] #[ # now run the job - output goes from each MPI process goes ] #[ # to the file log.jobnumber.procnumber ] #[ # ] #[ prun -s -o log.job$njob.proc%./a.out ] #[===================================================================] if ( $NJOB == 2 ) then # Inital Run mkdir $RUN/eo_files # tar -cf /short/f63/dst/$pbs_jobname.tar $RUN/mdcont.in 178

200 prun -n ${PBS_NCPUS} sander -O -i $RUN/mdcont.in -o $RUN/md$NJOB.out -c $RUN/md$LJOB.restrt -p $RUN/input.prmtop -r $RUN/md$NJOB.restrt -x $RUN/md$NJOB.mdcrd -v $RUN/md$NJOB.mdvel -ref $RUN/md$LJOB.restrt sleep 60 else # Continuation prun -n ${PBS_NCPUS} sander -O -i $RUN/mdcont.in -o $RUN/md$NJOB.out -c $RUN/md$LJOB.restrt -p $RUN/input.prmtop -r $RUN/md$NJOB.restrt -x $RUN/md$NJOB.mdcrd -v $RUN/md$NJOB.mdvel -ref $RUN/md$LJOB.restrt sleep 60 endif # # Not expected to reach this point in general but if we do, check that all # is OK. If the job command exited with an error, terminate the job # set errstat=$? if ($errstat!= 0) then sleep 5 # A brief nap so PBS kills us in normal termination $ECHO "\njob number $NJOB returned an error status $errstat - stopping job sequence.\n" touch STOP_SEQUENCE exit $errstat endif Docking Using the NMR Restraints Option in Sander The docking script specifies the conditions under which docking will take place. Note that the MD preferences hold all the DNA heavy atoms (non-hydrogen) fixed. &cntrl imin=0, ntc=2, tol= , ntf=2, cut=9.0, igb=0, ntpr=100, ntwr = 1000, ntwx = 1000, ntwv = 1000, nstlim=10000, dt=0.002, ntt=1, tempi=300.0, temp0=300.0, tautp=2.0, ntx=7, irest=1, ntb=2, ntp=1, scee=1.2, nscm=1000, ntr=1, nmropt=1, &end &wt type='end' &end DISANG=dock.rst.f Restrain the DNA heavy atoms I FIND * CT * * * C * * * CA * * * CB * * * CK * * * CQ * * * CM * * * OH * * * OS * * 179

201 SEARCH RES 1 42 END Restrain the DNA heavy atoms II FIND * O2 * * * O * * * NA * * * N* * * * N2 * * * NC * * * NB * * * P * * SEARCH RES 1 42 END END dock.rst.f Example input script defining atoms and force constants to be acted on during the docking process. &rst iat=-1,-1, atnam(1)='dumm',atnam(2)='dumm', nstep1=1,nstep2=5000, iresid=1,irstyp=0,ifvari=1,ninc=0,imult=0,ir6=0,ifntyp=0, r1= e+00,r2= ,r3= ,r4=99.000,rk2=1.0000,rk3=1.0000, r1a= e+00,r2a=5.000,r3a=5.000,r4a=99.000,rk2a=1.0000,rk3a=1.0000, igr1 = 29,13 grnam1(1)='h2','o2' igr2 = 83 grnam2(1)='pt12' &end &rst iat=-1,-1, atnam(1)='dumm',atnam(2)='dumm', nstep1=5001,nstep2=10000, iresid=1,irstyp=0,ifvari=1,ninc=0,imult=0,ir6=0,ifntyp=0, r1= e+00,r2=5.0000,r3=5.0000,r4=99.000,rk2=1.0000,rk3=1.0000, r1a= e+00,r2a=5.0000,r3a=5.0000,r4a=99.000,rk2a=1.0000,rk3a=0.010, igr1 = 29,13 grnam1(1)='h2','o2' igr2 = 83 grnam2(1)='pt12' &end &rst iat=0, &end 2.3 Processing Scripts Renaming Equilibration Output Files 180

202 Some of the early molecular dynamics simulations used a different naming convention. As a result this script was used to rename the files for processing with the newer processing scripts. #!/bin/csh -f # Created by Donald Thomas 11/03/04 # # This script renames old MD Equilibration files so that new processing scripts will # work with the files. # It renames md01 or min01 to md1 or min1 # You need to enter a value on the command line when you execute the script # # ex) renx 9 --> will rename md01 to md09 and min01 to min09 # set i=1 set j=1 # if (- d md01) then while ( $i <= $1 ) mv md0$i md$i cd md$i mv md0$i.out md$i.out mv md0$i.pdb md$i.pdb i = ( $i + 1) end # endif # if (-d min01) then while ( $j <= $1 ) mv min0$j min$j cd min$j mv min0$j.out min$j.out mv min0$j.pdb min$j.pdb j = ( $j + 1) end # endif Energy Output Analysis Script used to run the energy processing perl script. #processing script # # This script will process all amber.out files in a directory # # the command line is./run_mdout # # If you wish to use this script for equilibration or simulated annealing runs the file # prefix must be modified # # This script runs without modification on both the LC and SC # perl process_mdout.perl md*.out 181

203 process_mdout.perl Perl script that processes the energy output of Amber molecular dynamics simulations. #!usr/bin/perl if ($#ARGV < 0) { print " Incorrect usage...\n"; exit; } foreach $i ( 0..$#ARGV ) { $filein = $ARGV[$i]; $checkfile = $filein; $checkfile =~ s/\.z//; if ( $filein ne $checkfile ) { open(input, "zcat $filein ") die "Cannot open compressed $filein -- $!\n"; } else { open(input, $filein) die "Cannot open $filein -- $!\n"; } print "Processing sander output file ($filein)...\n"; &process_input; close(input); } print "Starting = sort by_number = sort by_number keys(%avg_time); foreach $i ( TEMP, TSOLUTE, TSOLVENT, PRES, EKCMT, ETOT, EKTOT, EPTOT, DENSITY, VOLUME ) { print "Outputing summary.$i\n"; open(output, "> summary.$i"); %outarray = eval "\%$i"; foreach $j ) { print OUTPUT "$j ", $outarray{$j}, "\n"; } close (OUTPUT); print "Outputing summary_avg.$i\n"; open(output, "> summary_avg.$i"); %outarray = eval "\%AVG_$i"; foreach $j ) { print OUTPUT "$j ", $outarray{$j}, "\n"; } close (OUTPUT); print "Outputing summary_rms.$i\n"; open(output, "> summary_rms.$i"); %outarray = eval "\%RMS_$i"; foreach $j ) { print OUTPUT "$j ", $outarray{$j}, "\n"; } close (OUTPUT); } sub by_number { 182

204 } if ($a < $b) { -1; } elsif ($a == $b) { 0; } elsif ($a > $b) { 1; } sub process_input { $status = 0; $debug = 0; while ( <INPUT> ) { $string = $_; print $_ if (! /NB-upda/ && $debug ); if (/A V E R A G E S/) { $averages = 1; ($averages_over) = /.*O V E R.*(\d*).*S T E P S/; } $rms = 1 if (/R M S/); if (/NSTEP/) { ($time, $temp, $pres) = /NSTEP =.*TIME.* =(.*\d*\.\d*).*temp.* =(.*\d*\.\d*).*press = (.*\d*\.\d*)/; if ( $debug ) { print $_; print "time is $time, temp is $temp, pres is $pres\n"; } $_ = <INPUT>; if (/Etot/) { ($etot, $ektot, $eptot) = /Etot.*=(.*\d*\.\d*).*EKtot.*=(.*\d*\.\d*).*EPtot.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "Etot is $etot, ektot is $ektot, eptot is $eptot\n"; } $_ = <INPUT>; } if (/BOND.*ANGLE.*DIHED/) { ($bond, $angle, $dihedral) = /BOND.*=(.*\d*\.\d*).*ANGLE.*=(.*\d*\.\d*).*DIHED.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "bond is $bond, angle is $angle, dihedral is $dihedral\n"; } $_ = <INPUT>; } if (/1-4 NB/) { ($nb14, $eel14, $nb) = /1-4 NB.*=(.*\d*\.\d*).*1-4 EEL.*=(.*\d*\.\d*).*VDWAALS.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "nb14 is $nb14, eel14 is $eel14, vdwaals is $nb\n"; } $_ = <INPUT>; } if (/EELEC/) { ($eel, $ehbond, $constraint) = 183

205 /EELEC.*=(.*\d*\.\d*).*EHBOND.*=(.*\d*\.\d*).*CONSTRAINT.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "eel is $eel, ehbond is $ehbond, constraint is $constraint\n"; } $_ = <INPUT>; # # check to see if EAMBER is in the mdout file (present when # NTR=1) # if ( /EAMBER/ ) { $_ = <INPUT>; } } if (/EKCMT/) { ($ekcmt, $virial, $volume) = /EKCMT.*=(.*\d*\.\d*).*VIRIAL.*=(.*\d*\.\d*).*VOLUME.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "Ekcmt is $ekcmt, virial is $virial, volume is $volume\n"; } $_ = <INPUT>; } if (/T_SOLUTE/) { ($tsolute, $tsolvent) = /T_SOLUTE =(.*\d*\.\d*).*t_solvent =(.*\d*\.\d*)/; if ( $debug ) { print $_; print "Temp solute is $tsolute, temp solvent is $tsolvent\n"; } $_ = <INPUT>; } # update arrays if (/Density/) { ($density) = /.*Density.*=(.*\d*\.\d*)/; if ( $debug ) { print $_; print "Density is $density\n"; } $_ = <INPUT>; } if ( $averages == 1 ) { $AVG_TIME{$time} = $time; $AVG_TEMP{$time} = $temp; $AVG_PRES{$time} = $pres; $AVG_ETOT{$time} = $etot; $AVG_EKTOT{$time} = $ektot; $AVG_EPTOT{$time} = $eptot; $AVG_BOND{$time} = $bond; $AVG_ANGLE{$time} = $angle; $AVG_DIHEDRAL{$time} = $dihedral; $AVG_NB14{$time} = $nb14; $AVG_EEL14{$time} = $eel14; $AVG_NB{$time} = $nb; $AVG_EEL{$time} = $eel; $AVG_EHBOND{$time} = $ehbond; $AVG_CONSTRAINT{$time} = $constraint; $AVG_EKCMT{$time} = $ekcmt; $AVG_VIRIAL{$time} = $virial; $AVG_VOLUME{$time} = $volume; 184

206 $AVG_TSOLUTE{$time} = $tsolute; $AVG_TSOLVENT{$time} = $tsolvent; $AVG_DENSITY{$time} = $density; $averages = 0; } elsif ( $rms == 1 ) { $RMS_TIME{$time} = $time; $RMS_TEMP{$time} = $temp; $RMS_PRES{$time} = $pres; $RMS_ETOT{$time} = $etot; $RMS_EKTOT{$time} = $ektot; $RMS_EPTOT{$time} = $eptot; $RMS_BOND{$time} = $bond; $RMS_ANGLE{$time} = $angle; $RMS_DIHEDRAL{$time} = $dihedral; $RMS_NB14{$time} = $nb14; $RMS_EEL14{$time} = $eel14; $RMS_NB{$time} = $nb; $RMS_EEL{$time} = $eel; $RMS_EHBOND{$time} = $ehbond; $RMS_CONSTRAINT{$time} = $constraint; $RMS_EKCMT{$time} = $ekcmt; $RMS_VIRIAL{$time} = $virial; $RMS_VOLUME{$time} = $volume; $RMS_TSOLUTE{$time} = $tsolute; $RMS_TSOLVENT{$time} = $tsolvent; $RMS_DENSITY{$time} = $density; $rms = 0; } else { $TIME{$time} = $time; $TEMP{$time} = $temp; $PRES{$time} = $pres; $ETOT{$time} = $etot; $EKTOT{$time} = $ektot; $EPTOT{$time} = $eptot; $BOND{$time} = $bond; $ANGLE{$time} = $angle; $DIHEDRAL{$time} = $dihedral; $NB14{$time} = $nb14; $EEL14{$time} = $eel14; $NB{$time} = $nb; $EEL{$time} = $eel; $EHBOND{$time} = $ehbond; $CONSTRAINT{$time} = $constraint; $EKCMT{$time} = $ekcmt; $VIRIAL{$time} = $virial; $VOLUME{$time} = $volume; $TSOLUTE{$time} = $tsolute; $TSOLVENT{$time} = $tsolvent; $DENSITY{$time} = $density; } } } } Geometry Output Analysis (Carnal) The carnal execution script was used only to confirm the rms output the ptraj rms analysis vida infra. 185

207 #!/bin/csh # # This script will process all amber trajectory files in a directory # # the command line is./run_carnal # # If you wish to use this script for equilibration or simulated annealing runs the file # prefixes in the ptraj.in file must be modified # # The files to be processed need to be specified in the carnal_rms_start.in file # # This script runs without modification on both the LC and SC # # carnal -O < carnal_rms_start.in carnal_rms_start.in The carnal input script that describes the input files and the atoms to be used in calculating the root mean square deviation of distance. FILES_IN PARM p1 input.prmtop; STREAM s1 md3.mdcrd.gz md4.mdcrd.gz md5.mdcrd.gz... md136.mdcrd.gz md137.mdcrd.gz; FILES_OUT TABLE tab1 rms_to_start; DECLARE GROUP gall (RES 1-28); RMS r1 FIT gall s1; OUTPUT TABLE tab1 r1; END Geometry Output Analysis (ptraj) The execution script for the analysis of the molecular dynamics trajectory. #!/bin/csh # # This script will process all amber trajectory files in a directory # # the command line is./run_ptraj # # If you wish to use this script for equilibration or simulated annealing runs the file # prefixes in the ptraj.in file must be modified # # The files to be processed need to be specified in the ptraj.in file # # This script runs without modification on both the LC and SC # 186

208 # ptraj input.prmtop ptraj.in ptraj.in Input file for the ptraj trajin md3.mdcrd.gz trajin md4.mdcrd.gz trajin md5.mdcrd.gz... trajin md135.mdcrd.gz trajin md136.mdcrd.gz trajin md137.mdcrd.gz trajout md.strip.traj nobox center :1-14 mass origin image origin center center :1-28 mass origin image origin center rms first mass out rms_to_first.data :1-24 average ptraj_avg.pdb :1-24 pdb strip :WAT X3DNA Structural analysis script Script used to process the DNA structural information from pdb files at the end of each MD segment. #!/bin/csh -f # script to automatically generate pdb files from equilibration runs and subject pdb files to X3DNA analysis # # This script is executed from the command line in linux as follows: #./run_x3dna $n where $n=the number of md files to process # # This script also runs on the Linux cluster if X3DNA is installed # # If you wish to use this script for equilibration or simulated annealing runs the file # prefix must be modified # set n=1 mkdir X3DNA while ( $n <= $1 ) mkdir X3DNA/md$n ambpdb -p input.prmtop -bres < md$n.restrt > X3DNA/md$n/md$n.pdb cd X3DNA/md$n find_pair -t md$n.pdb stdout analyze n = ( $n + 1 ) end 187

209 Compilation of X3DNA results Script used to compile the individual physical characteristics of DNA over each MD simulation. <?php // X3DNA post processing script // created by Donald Thomas and severely modified by Julian T-F // 24/02/2004 // // This script must be installed in the www directory and the linux // computer must be running php and apache // // If this doesn't make sense find someone who understand it and ask // them for help // unset($http_get_vars, $HTTP_POST_VARS); // Security for older versions of PHP. $out_file_names = array( "system_id" => 4, "RMSD_bases" => 6, "H-Bond_info" => 7, "coordinates" => 8, "local_parameters" => 9, "local_step_parameters" => 10, "local_helix_parameters" => 11, "lambda" => 13, "helix_form" => 14, "groove_widths" => 15, "torsions" => 17, "sugars" => 18, "P-P_distances" => 19, "helix_radii" => 20, "local_helix_axis" => 21 ); // Input $i = 0; $in_filesize = 0; do { $i++; $in_file = $directory. "/md". $i. "/md". $i. ".out"; $in_filesize = filesize($in_file); if ($in_filesize > 0) { $in_file_pointer = fopen($in_file, "r"); $in_contents = fread($in_file_pointer, $in_filesize); fclose($in_file_pointer); } $component_array = explode(str_repeat("*", 76), $in_contents); foreach ($out_file_names as $key => $value) { $out_file_contents[$key].= $component_array[$value]; } } while ($in_filesize > 0); 188

210 // Output foreach ($out_file_contents as $key => $value) { $output_file_pointer = fopen("../../../../../tmp/x3dna/". $key, "w"); fwrite($output_file_pointer, $value); }?> <html> <head> <meta http-equiv="content-type" content="text/html; charset=iso "> <title>x3dna Output file fixer</title> </head> <body> <br> This script is designed to process the data created by X3DNA<br><br> Please copy and paste the full path to the data directory into the box below<br> ex.) /usr/local/modelling/finished/docking/production/prod-g25/x3dna<br> DO NOT include the trailing "/"<br><br> Press "Click Me" to execute the script<br><br> <form action="<?php echo $PHP_SELF?>" method="post"> <input type="text" name="directory"> <br><br> <input type="submit" value="click Me"> </form> The warning message will tell you if the processing was completed successfully<br> md119.out should refer to the first non existent file in your X3DNA directory.<br> The script fails when it can't find the next md#.out file.<br> If the processing fails prior to your last md#.out file then you have a problem with<br> the input data. Fix it and try again.<br> <br> Also you may notice that the script executes of its own accord when it is reloaded. You<br> can ignore this and re-run the script explicitly. This is in fact a very good idea since<br> the previous directory settings may not be valid any longer.<br> <br> Best of Luck.<br> <br> Don Thomas<br> </body> </html> Methyl-fix Script used to convert the Methyl naming convention used in Amber when creating a PDB file from restart or trajectory files. #!/bin/csh # This script adjusts the naming of DNA from ambpdb -bres so that the T-methyl groups # recognized in Swiss PDB Viewer # Note that this script removes all TER cards from pdb file. sed "s/c7 T/C5M T/" $1.pdb > methyl.temp sed "s/1h7 / H51/" methyl.temp > methyl1.temp sed "s/2h7 / H52/" methyl1.temp > methyl2.temp sed "s/3h7 / H53/" methyl2.temp > methyl3.temp sed "/WAT/d" methyl3.temp > methyl4.temp sed "s/1h3 ttt/ H31 ttt/" methyl4.temp > methyl5.temp sed "s/2h3 ttt/ H32 ttt/" methyl5.temp > methyl6.temp sed "s/3h3 ttt/ H33 ttt/" methyl6.temp > methyl7.temp sed "s/1h4 ttt/ H41 ttt/" methyl7.temp > methyl8.temp sed "s/2h4 ttt/ H42 ttt/" methyl8.temp > methyl9.temp sed "s/3h4 ttt/ H43 ttt/" methyl9.temp > methyl10.temp 189

211 sed "s/1h5 ttt/ H51 ttt/" methyl10.temp > methyl11.temp sed "s/2h5 ttt/ H52 ttt/" methyl11.temp > methyl12.temp sed "s/3h5 ttt/ H53 ttt/" methyl12.temp > methyl13.temp sed "s/1h6 ttt/ H61 ttt/" methyl13.temp > methyl14.temp sed "s/2h6 ttt/ H62 ttt/" methyl14.temp > methyl15.temp sed "s/3h6 ttt/ H63 ttt/" methyl15.temp > methyl16.temp sed "/TER/d" methyl16.temp > $1.new.pdb rm methyl* A3 Chapter 3 Appendix 3.1 Scientist equation file for Model A fitting Scientist equation file used in the fitting of Model A in perchlorate solution for 1,0,1/t,t,t // MicroMath Scientist Model File // 1,0,1/t,t,t hydrolysis second order reversible - equilibrium first step // 15mM ClO 4 25 o C NMR IndVars: T DepVars: A, B, Cl Params: KAB, KBA A'=-KAB*A+KBA*B*Cl B'=KAB*A-KBA*B*Cl Cl'=KAB*A-KBA*B*Cl // A=101ttt, B=aqua/Cl, Cl=chloride T=0.0 A= B= Cl= *** 3.2 Scientist equation file for Model 2 fitting Scientist equation file used in the fitting of Model 2 in perchlorate solution for 1,0,1/t,t,t // MicroMath Scientist Model File // 1,0,1/t,t,t hydrolysis second order reversible - equilibrium first step // 15mM ClO 4 25 o C NMR IndVars: T The publication Zhang, J.; Thomas, D. S.; Davies, M. S.; Berners-Price, S. J.; Farrell, N., J. Biol. Inorg. Chem. 2005, 10, and the associated supporting information can be found on the accompanying CD in the publications directory. 190

212 DepVars: A, B, Cl, C Params: KAB, KBA, KBC, KCB A'=-KAB*A+KBA*B*Cl B'=KAB*A-KBA*B*Cl-KBC*B+KCB*C*Cl Cl'=KAB*A-KBA*B*Cl+KBC*B-KCB*C*Cl C'=KBC*B-KCB*C*Cl // A=101ttt, B=aqua/Cl, Cl=chloride C=diaqua T=0.0 A= B= Cl= C=0.0 *** 3.3 Scientist equation file for phosphate model fitting Scientist equation file used in the fitting in phosphate solution for 1,0,1/t,t,t // MicroMath Scientist Model File // 1,0,1/t,t,t hydrolysis second order reversible - equilibrium first step // 15mM PO 4 25 o C NMR IndVars: T DepVars: A, B, Cl, C, D, P Params: KAB, KBA, KBC, KCB, KCD, KDC A'=-KAB*A+KBA*B*Cl B'=KAB*A-KBA*B*Cl-KBC*B*P+KCB*C Cl'=KAB*A-KBA*B*Cl+KCD*C-KDC*D*Cl C'=KBC*B*P-KCB*C-KCD*C+KDC*D*Cl D'=KCD*C-KDC*D*Cl P'=-KBC*B*P+KCB*C // A=101ttt, B=aqua/Cl, Cl=chloride C=phosphato/Cl, D=phosphato/H2O, P=phosphate T=0.0 A= B= Cl= C=0.0 D=0.0 P=0.015 *** 191

213 A4 Chapter 4 Appendix 4.1 Selected Figures from Hegmans et al. The following sections of this appendix contain selected figures from the paper by Hegmans et al. Long Range 1,4 and 1,6-Interstrand Cross-Links Formed by a Trinuclear Platinum Complex. Minor Groove Preassociation Affects Kinetics and Mechanism of Cross-Link Formation as Well as Adduct Structure. Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, This work studies the stepwise formation of 1,4- and 1,6-interstrand cross-links by reaction of 15 N-1,0,1/t,t,t (1) with the duplexes: 5 -{d(atatgtacatat) 2 } 14XL - (I) 5 -{d(tatgtatacata) 2 } 16XL - (II) The labelling of intermediates and products (1-5) is shown in the following scheme. Scheme A4.1. Interstrand cross-linking of 1 with duplexes I (14XL) and II (16XL), according to Model 1 as defined in Hegmans et al. The publication Hegmans, A.; Berners-Price, S. J.; Davies, M. S.; Thomas, D. S.; Humphreys, A. S.; Farrell, N., J. Am. Chem. Soc. 2004, 126, and the associated supporting information can be found on the accompanying CD in the publications directory. 192

214 a b Figure A4.1 2D [ 1 H, 15 N] HSQC NMR (600 MHz) spectra at 298K of (a) duplex I and (b) duplex II after reaction with 15 N-1 for the times shown. Peaks are assigned to the Pt-NH 3 and Pt-NH 2 end and central linker groups in structures 1-5 in the pathway to formation of the 1,4- or 1,6 interstrand cross-linked adducts. For duplex I there are two conformers of the bifunctional adduct 5X and 5Y. The assignments A 1 /A 2 and B 1 /B 2 correspond to the different NH 3 environments for the end groups. Peaks for the individual conformers are not distinguishable in the Pt-NH 2 region but there are several different environments for the Pt-NH 2 groups as shown in the 45 h spectrum. In (a) peaks corresponding to the Pt- NH 2 groups of the central linker are mostly concealed by proximity to the H 2 O peak for all species; peaks labelled * in the NH 3 region at δ are assigned to other products. In (b) Peaks labelled m in the NH 3 region are assigned to other monofunctional adducts. 193

215 Figure A4.2 1 H NMR spectra (600 MHz) of the aromatic regions of (a) duplex I and (b) duplex II after reaction with 15 N-1 for between 0 and 47 h. In (a) the peaks are assigned to the H8 resonances of the G5 (and G5') bases coordinated to platinum in the monofunctional adduct (3) and the two conformers (X and Y) of the 1,4-interstrand cross-link (5). In (b) the peaks are assigned to the H8 resonances of the G4 (and G4') bases coordinated to platinum in the monofunctional adduct (3) and the 1,6-interstrand cross-link (5). The minor peak (at δ 8.73) is unassigned; it increases in intensity at a similar rate to the G H8 peak of 5 (δ 8.67) but curiously, there are no peaks in the [ 1 H, 15 N] NMR spectra which correlate with this minor species. Peaks labelled o are assigned to other products. 194

216 Figure A4.3 1 H NMR spectra (600 MHz) of the imino regions of (a) duplex I and (b) duplex II after reaction with 15 N-1 for between 0 and 45 h. In (a) assignments of resonances to base-pairs are indicated by numbers (5 = G(5)-C(8*) etc.), and letters indicate assignments to platinated adducts: a 3; b 5. In (b) the control spectrum shows the position of the imino protons before addition of 15 N-1. Pre-covalent binding association causes a selective shift for one of resonances of the AT base pairs (most likely T(5)A(8 ) or T(3)A(10 )). This peak reverts to its original position as the bound drug covalently attaches to the DNA to form the interstrand 1,6 interstrand cross-link (via the monofunctional adduct). Although the reactions are complete at 48 h intense signals for unplatinated duplex are present due to the excess of DNA used in the reactions. A greater excess of DNA was used in the reaction with duplex I. 195

217 Figure A4.4 Comparison of the 1 H spectra following the formation of (a) 1,4 - and (b) 1,6-interstrand cross-links by 15 N-1 showing the region of the CH 2 protons of the linker. The bottom spectrum shows the 1 H spectrum of 15 N-1 in 15 mm phosphate (ph 5.3) in the absence of DNA and the spectra at 0 h indicate the positions of the T-methyl protons in the two duplexes (I: 2.96 mm, II: 2.62 mm, in 15 mm phosphate, ph 5.3) before the addition of 15 N-1 (1.60 or 1.90 mm). The labelled peaks correspond to the CH 2 protons of the linker in the monofunctional (3) and bifunctional (1,4 or 1,6) adducts (5). On addition of 1 to I there is little change in the chemical shift of the multiplet at δ 1.68 corresponding to the CH 2 (2 and 5) groups, whereas for duplex II pre-covalent binding association leads to non-equivalence of these groups and two peaks are observed (δ 1.68 and 1.63). Formation of the 1,4 cross-link gives rise to a major pair of peaks in this region (δ 1.69,1.66) as well as three minor peaks (δ 1.74, 1.59 and 1.85) which is possibly consistent with the formation of major (5X) and minor (5Y) conformers observed in the [ 1 H, 15 N] NMR spectra. On the other hand on formation of the 1,6 interstrand cross-link only one pair of peaks (δ 1.78 and 1.75) is observed assignable to the CH 2 groups 2 and 5 and these peaks are more strongly deshielded than for the 1,4 interstrand cross-link. 196

218 Figure A4.5 Plots of the relative concentration of species observed during formation of (a, b) 1,4- and (c) 1,6- interstrand cross-links by reaction (at 298 K) of 15 N-1 with duplex I and II. The concentrations are based on the relative peak volumes of peaks in the Pt-NH 3 region. The curves are computer best fits for the rate constants shown in Table 4 of the paper. Labels: 1 open squares, 2 open circles, 3 triangles, 5 filled circles, other products diamonds. (b) shows the formation of the major (5X,filled circles ) and minor (5Y, filled squares) conformers of the 1,4-interstrand cross-link which are shown as combined product in (a). 197

219 4.2 Mass Spectrometry of isolated Pt-DNA adducts Figure A4.6 ESI-mass spectrum of the first 1,0,1/t,t,t-14XL adduct obtained from the HPLC purification of the product mixture after the completion of the reaction. The 4-, 5-, 6- and 7- charge states of the adduct are indicated. Figure A4.7 ESI-mass spectrum of the second 1,0,1/t,t,t-14XL adduct obtained from the HPLC purification of the product mixture after the completion of the reaction. The 4-, 5-, 6- and 7- charge states of the adduct are indicated. 198