Are Dinucleoside Monophosphates Relevant Models for the Study of

Jun 9, 2014 - Oxidatively generated tandem lesions such as G[8–5m]T pose a potent ... use of GpT in water as a model of the corresponding reaction i...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/crt

Are Dinucleoside Monophosphates Relevant Models for the Study of DNA Intrastrand Cross-Link Lesions? The Example of G[8−5m]T Julian Garrec*,†,‡ and Elise Dumont*,§ †

CNRS, Théorie-Modélisation-Simulation, SRSMC, Vandoeuvre-lès-Nancy F-54506, France Thérorie-Modélisation-Simulation, SRSMC, Université de Lorraine, Vandoeuvre-lès-Nancy F-54506, France § Institut de Chimie de Lyon, CNRS, Ecole normale supérieure de Lyon, Université de Lyon, 46 allée d’Italie, 69364 Lyon, Cedex 07, France ‡

S Supporting Information *

ABSTRACT: Oxidatively generated tandem lesions such as G[8−5m]T pose a potent threat to genome integrity. Direct experimental studies of the kinetics and thermodynamics of a specific lesion within DNA are very challenging, mostly due to the variety of products that can be formed in oxidative conditions. Dinucleoside monophosphates (DM) involving only the reactive nucleobases in water represent appealing alternative models on which most physical chemistry and structural techniques can be applied. However, it is not yet clear how relevant these models are. Here, we present QM/ MM MD simulations of the cyclization step involved in the formation of G[8−5m]T from the guanine−thymine (GpT) DM in water, with the aim of comparing our results to our previous investigation of the same reaction in DNA (Garrec, J., Patel, C., Rothlisberger, U., and Dumont, E. (2012) J. Am. Chem. Soc. 134, 2111−2119). We show that, despite the different levels of preorganization of the two systems, the corresponding reactions share many energetic and structural characteristics. The main difference lies in the angle between the G and T bases, which is slightly higher in the transition state (TS) and product of the reaction in water than in the reaction in DNA. This effect is due to the Watson−Crick H-bonds, which are absent in the {GpT +water} system and restrain the relative positioning of the reactive nucleobases in DNA. However, since the lesion is accommodated easily in the DNA macromolecule, the induced energetic penalty is relatively small. The high similarity between the two reactions strongly supports the use of GpT in water as a model of the corresponding reaction in DNA.



INTRODUCTION Biomolecules are frequently exposed to deleterious agents in vivo, such as reactive oxygen species (ROS),1,2 which cause important damage to biomolecules, in particular DNA. Among the lesions that can be formed, oxidatively generated tandem lesions (OTL), in which two neighboring DNA bases are covalently tethered together, have attracted a lot of interest because they can severely jeopardize genome integrity.1,3−9 One of the first OTLs that has been put forward involves a guanine and a thymine belonging to the same DNA strand,10 and is induced by the abstraction of a hydrogen atom from the thymine methyl group by a free radical such as the hydroxyl radical (HO•) (Figure 1). This guanine−thymine lesion was detected both in isolated DNA and in cultured human cells exposed to γ-rays. The C8 of guanine is covalently bonded to the C5 methyl (C5m) atom of the neighboring thymine, hence the G[8−5m]T abbreviation. This lesion is a substrate for nucleotide excision repair in vivo.11 However, as two contiguous nucleotides are damaged, the repair efficiency is significantly lower than that for single-nucleobase lesions such as 8-oxo-7,8dihydro-2′-deoxyguanosine.12,13 Because of its high mutagenicity, a deeper understanding of the formation of G[8−5m]T, and more generally of all OTLs, © 2014 American Chemical Society

is sought. G[8−5m]T occurs less frequently than photolesions or ROS-induced single-nucleobase defects. A yield of ∼0.05 lesions per 109 nucleosides per Gy has been reported,14 a value 3 orders of magnitude smaller than the one reported for 5formyl-2′-deoxyuridine, as one example of a simple, singlenucleotide lesion. However, it is even more difficult to obtain kinetic data that would rationalize the formation of these bulky adducts. Indeed, the rarity of tandem lesions makes them highly challenging to detect within DNA.15 To circumvent this difficulty, short DNA oligonucleotides containing a photolabile synthetic precursor of the first reaction intermediate, a T−CH2• radical (second structure in Figure 1), has been synthesized.16 This enabled researchers to force the system to initiate the radical reaction from this state and thus to follow the desired reaction pathway. However, to the best of our knowledge, this strategy has never been employed to derive the rate and equilibrium constants of the formation of G[8−5m]T. An appealing alternative consists of conducting auxiliary experiments on isolated dinucleoside monophosphates (DM) in solution.3 With such small systems (compared to doubleReceived: December 12, 2013 Published: June 9, 2014 1133

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

Figure 1. Reaction mechanism of the formation of G[8−5m]T. The boxed step (2) is the focus of the present study and of ref 25. The {2deoxyribose-phosphate-2-deoxyribose} backbone linking the thymine and guanine nucleobases is schematized by an -S-P-S- letter sequence.

Figure 2. Example snapshots of our QM/MM simulation of step (2) of the formation of G[8−5m]T in water. The snapshots show (A) the reactants in a nonreactive, disordered conformation (Rdisordered), (B) the π-stacked reactants in the reactive conformation in the same narrow solvent cage (Rcage), and (C) G[8−5m]T⌉• (Pcage). In all snapshots, the nucleobases, the {2-deoxyribose-phosphate-2-deoxyribose} backbone, and the solvent are represented with balls-and-sticks, tan sticks, and van der Waals spheres, respectively. Atoms that are represented with balls-and-sticks (colored by chemical element) are treated quantum-mechanically (QM), while all the others are modeled classically (MM). The green arrows in panels A and B indicate qualitatively the orientation of p orbitals of the C5m atom of the thymine and the C8 atom of the guanine.

ically, we resort to the implementation of Rothlisberger and coworkers22,23 who integrated the QM/MM strategy within the Car−Parrinello molecular dynamics (CPMD)24 framework. In a previous work,25 we used this approach to model the formation of G[8−5m]T placed in the center of a solvated DNA dodecamer. This initial study provided electronic, structural, and energetic details about the reaction, in particular the cyclization step during which the two reactive bases are brought close together (step (2) in Figure 1). The latter is the most critical step in terms of structural reorganization.25,26 We showed that this reaction is a rather facile process with a free energy barrier of 9.7 kcal/mol and that G[8−5m]T is accommodated easily within B-DNA, i.e., with no major structural distortion of the macromolecule. Here, we present complementary simulations in which we have applied the same protocol to the analogous reaction from the guanine−thymine DM (GpT hereafter) in water (Figure 2). This provides a direct microscopic view of the relative effect of various environments (solvated DNA and bulk water, which are represented explicitly in our model) on the formation of G[8−5m]T. Our calculations actually show that several key characteristics of this reaction from GpT in water can be extrapolated to solvated DNA.

stranded DNA), organic synthesis and kinetics measurements can be performed in a systematic manner. One can sketch selectivity rules and even establish a firm reactivity order among radical nucleobases.17 In addition, synthesized DMs can be sitespecifically incorporated into oligonucleotides.18 However, it is not clear to what extent the results obtained for the reactivity of a DM are transferable to the corresponding moiety in a B-DNA macromolecular environment. So far, little is known about the difference of behavior between these two systems at the atomic scale. The present study aims at addressing this issue by means of molecular modeling including free energy calculations for the two corresponding reactions. To do so, we employ a relevant microscopic model that incorporates the two crucial ingredients that are necessary to provide a realistic description of the reactive fragments embedded in their environment, namely, (i) the explicit representation of the whole solute−solvent system19 and (ii) the dynamics of the system at room temperature.20,21 The first ingredient is included by means of a mixed quantum mechanics/molecular mechanics (QM/MM) approach19 in which the reactive fragments are treated quantum mechanically (QM), which enables the description of bond formation and breaking, while the rest of the system is represented by a cheaper (in terms of computational cost) molecular mechanics (MM) potential. The second ingredient is accounted for through molecular dynamics (MD) simulations. More specif1134

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

Figure 3. Organization of the system during step (2) in (A) water and (B) in DNA. For the GpT DM in water (A), the reactants (R) exhibit a large conformational space, due to free rotations around backbone dihedrals. Only one specific conformation, with the proper positioning of the two reactive nucleobases in a π-stacked conformation in the same solvent cage (represented by a thick dashed circle; see also Figure 2-B), can give rise to the formation of G[8−5m]T⌉•. In B-DNA (B), the system is initially highly preorganized, with the two reactive bases in a π-stacked conformation. The Watson−Crick H-bond network (represented by thin dashed lines), and thus the high organization of the system, is preserved throughout the whole reaction, as demonstrated in our previous investigation of the formation of G[8−5m]T in DNA.25 Note that these sketches exagerate the πstacked character of the cross-linked T and G nucleobases in both PDNA and Pcage. See Figure 8b in ref 25 and Figure 2C for representative snapshots of PDNA and Pcage, respectively.



with CPMD-QM/MM simulations because we are mostly interested with the reactive conformation of the reactant, Rcage (Figure 3A and Figure 2B). We will discuss this issue further in the Results and Discussion section. Rcage corresponds to a small portion of the full conformational landscape of the reactant and can be modeled with QM/MM simulations. However, it is not trivial to guess its structure a priori. Pcage, on the other hand, is a much more rigid entity because of the additional chemical bond linking the two bases. It cannot intrinsically exhibit major conformational transitions so that its structure can be unambiguously guessed from the equivalent G[8−5m]T⌉• radical moiety in DNA that we generated in silico in ref 25. Thus, it was much more sound to start our simulations from Pcage and then to bring the system backward along step (2) (using the thermodynamic integration procedure described bellow) up to Rcage. First, we extracted the coordinates of the G[8−5m]T⌉• moiety (i.e., the cross-linked nucleobases together with their {2-deoxyribose-phosphate-2-deoxyribose} bridge) from a selected snapshot of the corresponding groups in damaged DNA.25 Then, we immersed this solute in a 41.5 × 41.4 × 41.5 Å3 simulation box containing 2161 water molecules. Finally, the

COMPUTATIONAL METHODS

The computational methodology used in the present study is the same as that in our previous investigations of the formation of G[8−5m]T,25 G[8−5]C,26 and G[8−5m]mC26 in B-DNA. It follows the work performed by others on similar DNA-based systems.27,28 Our approach differs substantially from the static approach mainly used by others in the field.29,30 In our scheme, the most decisive feature is to provide a realistic description of a biomolecular reaction and in turn to compute and compare the corresponding free energy profiles. We provide more details about the corresponding theoretical aspects in Supporting Information. Starting Structure. Our simulations were started from the product of the cyclization step (noted G[8−5m]T⌉• or Pcage herein). An example snapshot of G[8−5m]T⌉• is provided in Figure 2C. The reason for this choice is that, before cyclization, the reactant is very flexible due to free rotation around single bonds of the {2-deoxyribosephosphate-2-deoxyribose} backbone (this is depicted schematically in Figure 3A). As a result, the conformational landscape of the reactant state is large, and one typically needs nanosecond simulations to sample it.31−33 This time scale is much longer than what we can afford 1135

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

Figure 4. Evolution of key order parameters along step (2b)w. (A) Distance between the geometric center of the T and G bases (RG−T) as a function of the reaction coordinate d. (B) Angle between the planes of the T and G bases (α(G, T)̂ as a function of d. (C) α(G, T)̂ as a function of RG−T. The two regimes discussed in the text are represented by dashed lines, which are obtained from least-squares fits of the data. The color scale, from blue to red, corresponds to specific values of d (or, equivalently, to TI windows) and hence reflects the position of the system along the reaction path. The ranges of d corresponding to the regions of Pcage and the TS are defined qualitatively using the free energy profile (Figure 5). The reactant state is subdivided in reactive (Rcage) and disordered (Rdisordered) substates, as explained in the text. (D) Example snapshot showing the definition of key geometric parameters. d and RG−T are represented with a green dashed line and a yellow stick, respectively. α(G, T)̂ was calculated as the angle between the planes represented in magenta. α(G, T)̂ itself in not represented, for the sake of clarity. reasonable description of the π-stacking interaction between the DNA bases treated at the QM level.40 We used the QM/MM coupling scheme developed by Rothlisberger and co-workers,22,23 which couples the CPMD24 and GROMOS41 software. The equations of motion were integrated following the Car−Parrinello method.24 The system was first equilibrated for 27 ps. Then, the system was forced to move from Pcage to Rcage in a stepwise fashion, using a series of constrained CPMD simulations (or thermodynamic integration (TI) windows42) in which the distance d between the C8 atom of the guanine and the C5m atom of the thymine was fixed to a value ranging from 1.6 to 3.7 Å. We used 22 TI windows of 3 ps each. Once the reactant region was reached, geometric criteria were used to distinguish reactive (i.e., π-stacked, Figure 2B) from disordered (i.e., non-π-stacked, Figure 2A) reactant conformations (see the Results and Discussion section for more details). This analysis enabled us to define the Rcage state as the ensemble of conformations for which d is slightly

system was neutralized by adding one potassium ion. The total number of atoms was 6547. A snapshot of the resulting system is provided in Figure S1 (Supporting Information). Equilibration at the Classical Level. The system was equilibrated using classical MD simulations in the NPT ensemble (T = 300 K, P = 1 bar). The calculations were performed using the AMBER1134 suite of programs. The simulations parameters, in particular the AMBER99 force field,35 were kept the same as those in our previous investigation of DNA lesions.25,26 QM/MM Simulations and Free Energy Calculations. The last configuration of our classical simulation was used as a starting structure for QM/MM simulations. The QM/MM partition was the same as that in our study of G[8−5m]T in DNA,25 i.e., the QM part included the two reactive nucleobases. This corresponds to atoms that are represented with balls-and-sticks in Figure 2. This QM region was described using the same DFT/BLYP-DCACP36−39 level of theory as in refs 25 and 26. The DCACP correction is important to provide a 1136

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

between the planes of the bases, α(G, T),̂ along the reaction coordinate d (Figure 4). These order parameters reflect the stacking or nonstacking between the two bases in the reactant state and hence enable us to discriminate reactive conformations from others which are much less ordered (e.g., Figure 2B vs A). Figure 4 shows that when the system progresses backward from the products to the transition state and then to the reactants up to d ≈ 2.8 Å (green in Figure 4), RG−T and α(G, T)̂ both decrease. This is due to the fact that the C8(G)C5m(T) bond in Pcage (Figure 2C) constrains the system and prevents the T and G bases from adopting a parallel relative positioning. The progressive breaking of this bond along the Pcage → TS→ ··· process favors the π-stacking between the bases. Then, as d continues to increase, RG−T slightly increases (orange in Figure 4A), but the two bases remain close to a parallel conformation (Figure 4B). In this portion of the reaction path, π-stacking is optimal (within the intrinsic constraints of the system, due to the {2-deoxyribosephosphate-2-deoxyribose} backbone and the water cage). As a consequence, the p orbitals of the C8(G) and C5m(T) atoms are aligned, enabling a significant overlap between them, and placing the system in a reactive conformation. The snapshot of Figure 2B is extracted from this region. Finally, for d ⩾ 3.5 Å, both RG−T and α(G, T)̂ increase in a steep manner, and the magnitude of their fluctuations is much higher than that in other portions of the reaction path. This is characteristic of two nucleobases that are no longer in a π-stacked, reactive conformation. Figure 2A provides an example snapshot for d ≈ 3.7 Å. This transition region at d ≈ 3.5 Å is also clear from Figure 4C, in which two regimes can be identified. From this analysis, we can define Rcage as the ensemble of conformations for which d is slightly below or equal to 3.4 Å. This qualitative definition, based on geometrical criteria, can be completed by energetic considerations, based on the free energy profile that can be recovered from the TI procedure, as we shall see in the last section of this article. We should also note that a very similar average value of d is found for the reactant state in DNA.25 As a final remark, note that Florian et al.31 proposed a more sophisticated order parameter, noted ξ, which accounts for both RG−T and α(G, T)̂ at the same time. This choice is more relevant to the study of the stacking/destacking of the DM but is not mandatory here because we are pursuing a different goal. We provide the plot of ξ as a function of d in Figure S2 (Supporting Information), which shows that it basically leads to the same conclusions. Here, we stick to using RG−T and α(G, T)̂ because they provide a simpler and more intuitive description of Rcage in terms of positioning of the system on our reaction coordinate d. Preorganization in the GpT DM and in DNA. In order to establish a relevant comparison between the formation of G[8−5m]T in DNA and in water, it is crucial to stress that the reactants in these two environments differ in their preorganization. Figure 3 sketches the relative positioning of the G and T nucleobases at specific points of the reaction coordinate, as well as key preorganization factors. In GpT DM in water (Figure 3A), the reactive nucleobases are attached by the {2deoxyribose-phosphate-2-deoxyribose} bridge. This provides a certain degree of preorganization to the system because the reactants do not need to diffuse for long distances to come close together. However, as stressed in the previous section, the intrinsic flexibility of the DM makes most of the reactant

inferior or equal to 3.4 Å. Following this definition, we computed the free energy profile of step (2) in the reaction coordinate interval 1.6 Å ≤ d ≤ 3.4 Å, using the TI42 method. The statistical error on the free energy profile was computed as in our previous studies of DNA lesions25,26 and was found to be of the same magnitude, i.e., ≈1 kcal/mol. The possible limitations that are inherent to the QM(DFT)/MM method that we use have been discussed in ref 26. We would like to stress that the present study aims at comparing two analogous series of simulations performed to model step (2) in distinct environments, i.e., DNA and water. Although both series of calculations are associated with the above-mentionned limitations, the fact that they are obtained with the same computational protocol enables an insightful comparison.



RESULTS AND DISCUSSION Reactive Substate of the Reactants in Water and in DNA. Figure 2 shows representative snapshots extracted from our simulations of step (2). The formation of a covalent bond between the C5m atom of the thymine and the C8 atom of the guanine requires the overlap of their p orbitals (whose orientation is represented qualitatively by green arrows in Figure 2). This overlap is possible only if the two nucleobases are nearly π-stacked. As can be seen from Figure 2A and B, the reactant state exhibits both reactive and nonreactive conformations. Indeed, DMs are very flexible molecules due to free rotations around single bonds of the {2-deoxyribose-phosphate2-deoxyribose} backbone. This gives rise to many conformational substates for G[8−5m]T, only one of which corresponding to the nucleobases in the appropriate relative positioning in the same narrow solvent cage (Rcage, Figure 2B). Hence, step (2) for the reaction in water can be subdivided in the two steps sketched in Figure 3A, namely, the initial conformational transition toward a reactive conformation (Rdisordered → Rcage, step (2a)w) and the subsequent cyclization reaction (Rcage → Pcage, step (2b)w). Step (2a)w is similar to the stacking/destacking process of GpT studied previously by other authors.31−33 Our system differs from GpT in that the thymine moiety is a radical species with a missing H atom on the C5 methyl group. Although this difference in chemical structure affects base−base interactions, it should be stressed that the free energy difference between the stacked and unstacked states is governed by several factors, in particular solvent reorganization and barriers of rotation around backbone dihedrals, which have little to do with the radical or nonradical character of the thymine. Hence, we expect that the free energy difference associated with step (2a)w has the same order of magnitude as the value of −0.7 kcal/mol obtained for GpT in ref 31 using classical MD simulations. This speaks for a flat energy landscape in the reactants state, in which the system can pass from one conformation (e.g., an extended one like that in Figure 2A) to another (e.g., Rcage, Figure 2B) with almost no energetic penalty. From now, we will focus on step (2b)w and the comparison with its counterpart in DNA in order to provide a complete picture of the whole cyclization process in both environments. This relies on a proper definition of Rcage in terms of positioning of the system on the reaction coordinate. To do so, we started our simulations from Pcage and then forced the system to progress backward toward the reactants by increasing in a stepwise fashion the distance d between the C5m atom of the thymine and the C8 atom of the guanine (following the TI prodecure described in the Computational Methods section). Meanwhile, we monitored the distance between the geometric center of the T and G bases, RG−T, as well as the angle defined 1137

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

The activation barrier of steps (2b)w (ΔG⧧w) and (2)DNA (ΔG⧧DNA) are of 8.5 and 9.7 kcal/mol, respectively. Consistently, the position of the transition state along the reaction coordinate is slightly shifted toward the reactant in the former reaction (d⧧w ≈ 2.15 Å) with respect to the latter (d⧧DNA ≈ 2.05 Å). The difference between the two cyclization profiles is more pronounced in the product region, with free energy differences for the whole steps of ΔrGw = 2.4 kcal/mol and ΔrGDNA = 6.4 kcal/mol. These differences in the free energy profiles are best explained by the steric constraints that G[8−5m]T ⌉ • experiences within DNA. Figure 6 shows that in the product and TS regions (d ≤ 2.1 Å), RG−T and α(G, T)̂ are on average lower in DNA than in water. This is illustrated in Figure 6C, in which representative snapshots of the products in the two environments are superimposed. Note that the plots in both Figure 6A and B exhibit some significant fluctuations despite averaging other TI windows. This is due to the limited sampling that is inherent to any QM(DFT)/MM simulations of condensed phase systems26 (see Computational Methods). However, the amplitude of these fluctuations are lower than the difference between the curves obtained for the DM and DNA in the product region. This is a good indication that our sampling is enough to provide an instructive qualitative comparison between the reactions we are studying. In the Rcage region, RG−T and α(G, T)̂ exhibit similar values in both environments, which is consistent with the fact that the relative positioning of the reactive nucleobases in the reactant state is very similar in the two environments.31 The small increase in structural differences between the reactions in water and in DNA, when the systems proceeds from the reactants through the TS to the products, suggests a picture in which the constraint of the DNA framework onto the reactive bases subsystem increases along the reaction coordinate. This is consistent with the fact that the difference in transitions state free energy (ΔG⧧DNA − ΔG⧧w ≈ 1.2 kcal/mol) is lower than the difference in reaction free energy (ΔrGDNA − ΔrGw ≈ 4.0 kcal/ mol). Although G[8−5m]T⌉• is very well accommodated inside DNA,25 the Watson−Crick H-bonds between the reactive G and T bases and their C-A partners tighten α(G, T).̂ This results in a less optimal relative orientation of G and T⌉•, which comes with a small energetic penalty. It is worth stressing that we are dealing with a free radical reaction with small variation of the dipole moment of the reacting groups along the reaction path. This means that the energetics of the chemical step is not affected by the difference in electrostatic environment when considering the reactions in water and inside the macromolecule, as in enzymatic reactions with closed-shell substrates.43,44 Here, the preorganization of DNA is even slightly anticatalytic. Nevertheless, steps (2b)w and (2)DNA are overall very similar both from the structural and energetic points of view. This means that the formation of G[8−5m]T from GpT in water represents an interesting reference to the corresponding reaction in DNA. We expect that this finding can be transposed to other intrastrand basepairs, provided the DNA environment accommodates the resulting lesion easily, as is the case for G[8−5m]T and the peroxyl-based lesion.45,46

conformations nonreactive. In contrast in DNA (Figure 3B), the reactants (RDNA) are locked in a reactive conformation, ready to undergo the cyclization process RDNA → PDNA (step (2)DNA). This is due to the overall B-DNA structure and in particular to (i) the Watson−Crick H-bonds network between the reactive G and T⌉• nucleobases and their C-A partners, and (ii) the π-stacking interactions with adjacent nucleobases. As we have shown in our previous investigation of the formation of G[8−5m]T in DNA,25 this local structural organization is maintained throughout the whole reaction path (see also the illustrative snapshots provided in Figure 8 of ref 25). This means that B-DNA is a highly preorganized system with respect to the formation of G[8−5m]T. From a structural point of view, this has some similarities with what is observed with enzymatic reactions.43 However, we will show in the next section that this preorganization has no catalytic effect here. For the time being, the main difference between the formation of G[8−5m]T in solvated GpT and in DNA lies in the dynamics of these reactions: step (2a)w has no counterpart in DNA. Comparison of the Reaction Paths in Water and in DNA. Figure 5 compares the free energy profiles computed for

Figure 5. Free energy profiles for the cyclization step during the formation of G[8−5m]T. Reaction of GpT⌉• in water (step (2b)w in Figure 3A) (solid line). Corresponding reaction in DNA (step (2)DNA in Figure 3B) (dashed line). The profiles are shifted along the y axis so that the free energy of both Rcage and RDNA is zero (see text).

the cyclization of GpT⌉• in water (step (2b)w in Figure 3A) and the corresponding reaction in DNA (step (2)DNA in Figure 3B). It is important to stress that the relative positioning of the two profiles along the y-axis depends on the origin that we have chosen for free energies. Because (i) the preliminary step (2a)w for the reaction in water comes with almost no free energy change31 and (ii) no such preliminary step exists in DNA, it is convenient to shift the free energy profiles so that G(Rcage) = G(Pcage) = 0 kcal/mol. This, of course, relies on our definition of Rcage in terms of the value of the reaction coordinate (d = 3.4 Å). A different definition would shift the free energy profile of (2b)w up or down. However, the curvature of the free energy profile in the reactants region is small so that this difference in free energy would be negligible compared to the relevant quantities that we want to discuss here, e.g., TS free energies.



CONCLUSIONS We have simulated the cyclization of GpT⌉• in water, enabling a direct comparison with our previous simulations of the same reaction occurring in DNA.25 We have highlighted the 1138

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

Figure 6. Structural differences between the reaction paths in DNA (dashed line) and in water (solid line). (A) Average RG−T (see Figure 4D for definition) as a function of d. (B) Average α(G, T)̂ (see Figure 4D for definition) as a function of d. The averages are computed over each TI window, hence the notations ⟨RG−T⟩d and ⟨α(G, T)⟩̂ d. (C) Superimposition of example snapshots of the products in the GpT DM in water (red) and in DNA (green). The reactive bases are represented as balls and sticks.

cyclization process leading to the formation of G[8−5m]T⌉• in water and in DNA are similar, both in terms of energetics and structures. A subtle difference is found between the geometries of the products of the two reactions. The B-DNA framework, and in particular the Watson−Crick hydrogen-bonds involving the reactive subsystem, slightly tightens the angle between the reactive bases by about 10°. This results in an energetic penalty of about 4.0 kcal/mol. This effect is less pronounced in the TS, with an estimated difference of activation free energy of 1.7 kcal/mol. Overall, the two reactions share many common traits, with similar structural and energetic characteristics. Our results suggest that further experimental studies of the GpT DM reactivity in water could improve our understanding of the formation of G[8−5m]T. The extrapolation of our conclusions to other lesions, and thus to other corresponding DMs, depends on the ability of DNA to accommodate these lesions. For those reactions in which DNA does not significantly affect

difference in preorganization between the two systems, stressing that this mainly affects the reaction dynamics of the corresponding reactions. In DNA, the preorganization is high, with the reactive nucleobases locked in a reactive conformation, i.e., with a relative positioning that is favorable to the attack of the methylene radical of the H-abstracted thymine onto the C8 of the nearby guanine. In the GpT DM in water, the reactants have much more degrees of freedom because of the absence of the DNA framework, although the {2-deoxyribose-phosphate2-deoxyribose} bridge still maintains a certain level of preorganization. Using geometric criteria, the reactant state has been subdivided in reactive and disordered substates termed Rcage and Rdisordered, respectively. This analysis has enabled us to show that the structure of the reactive bases is very similar in GpT and in DNA, in particular in terms of separating distance and parallelism between the bases. We have used Rcage and its counterpart in DNA, RDNA, as references for energies. Our free energy calculations have shown that the reaction paths of the 1139

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

thymine tandem lesion G[8,5-Me]T by UvrABC nuclease. Chem. Res. Toxicol. 18, 1339−1346. (9) Gu, C., Zhang, Q., Yang, Z., Wang, Y., Zhou, Y., and Wang, Y. (2006) Recognition and Incision of oxidative intrastrand cross-link lesions by UvrABC nuclease. Biochemistry 45, 10739−10746. (10) Box, H. C., Budzinski, E. E., Dawidzik, J. B., Gobey, J. S., and Freund, H. G. (1997) Free radical-induced tandem base damage in DNA oligomers. Free Radical Biol. Med. 23, 1021−1030. (11) Jiang, Y., Wang, Y., and Wang, Y. (2009) In vitro replication and repair studies of tandem lesions containing neighboring thymidine glycol and 8-Oxo-7,8-dihydro-2′-deoxyguanosine. Chem. Res. Toxicol. 22, 574−583. (12) Imoto, S., Bransfield, L. A., Croteau, D. L., Van Houten, B., and Greenberg, M. M. (2008) DNA tandem lesion repair by strand displacement synthesis and nucleotide excision repair. Biochemistry 47, 4306−4316. (13) Wang, J., Cao, H., You, C., Yuan, B., Bahde, R., Gupta, S., Nishigori, C., Niedernhofer, L. J., Brooks, P. J., and Wang, Y. (2012) Endogenous formation and repair of oxidatively induced G[8−5m]T intrastrand cross-link lesion. Nucleic Acids Res. 40, 7368−7374. (14) Hong, H., Cao, H., and Wang, Y. (2007) Formation and genotoxicity of a guanine−cytosine intrastrand cross-link lesion in vivo. Nucleic Acids Res. 35, 7118−7127. (15) Cadet, J., Douki, T., Ravanat, J.-L., and Wagner, J. (2012) Measurement of oxidatively generated base damage to nucleic acids in cells: facts and artifacts. Bioanal. Rev., 1−20. (16) Romieu, A., Bellon, S., Gasparutto, D., and Cadet, J. (2000) Synthesis and UV photolysis of oligodeoxynucleotides that contain 5(phenylthiomethyl)-2′-deoxyuridine: a specific photolabile precursor of 5-(2′-deoxyuridilyl) methyl radical. Org. Lett. 2, 1085−1088. (17) Bellon, S., Ravanat, J.-L., Gasparutto, D., and Cadet, J. (2002) Cross-linked thymine-purine base tandem lesions: synthesis, characterization, and measurement in γ−irradiated isolated DNA. Chem. Res. Toxicol. 15, 598−606. (18) Bellon, S., Gasparutto, D., Saint-Pierre, C., and Cadet, J. (2006) Guanine−thymine intrastrand cross-linked lesion containing oligonucleotides: from chemical synthesis to in vitro enzymatic replication. Org. Biomol. Chem. 4, 3831−3837. (19) Warshel, A., and Levitt, M. (1976) Theoretical studies of enzymatic reactions: dielectric, electrostatic and steric stabilization of the carbonium ion in the reaction of lysozyme. J. Mol. Biol. 103, 227− 249. (20) Warshel, A. (1991) Computer Modeling of Chemical Reactions in Enzymes and Solutions, Wiley-Interscience, New York. (21) Klahn, M., Braun-Sand, S., Rosta, E., and Warshel, A. (2005) On possible pitfalls in ab initio quantum mechanics/molecular mechanics minimization approaches for studies of enzymatic reactions. J. Phys. Chem. B 109, 15645−15650. (22) Laio, A., VandeVondele, J., and Rothlisberger, U. (2002) A Hamiltonian electrostatic coupling scheme for hybrid Car−Parrinello molecular dynamics simulations. J. Chem. Phys. 116, 6941. (23) Laio, A., Gervasio, F. L., VandeVondele, J., Sulpizi, M., and Rothlisberger, U. (2004) A variational definition of electrostatic potential derived charges. J. Phys. Chem. B 108, 7963−7968. (24) Car, R., and Parrinello, M. (1985) Unified approach for molecular dynamics and density-functional theory. Phys. Rev. Lett. 55, 2471−2474. (25) Garrec, J., Patel, C., Rothlisberger, U., and Dumont, E. (2012) Insights into intrastrand cross-link lesions of DNA from QM/MM molecular dynamics simulations. J. Am. Chem. Soc. 134, 2111−2119. (26) Patel, C., Garrec, J., Dupont, C., and Dumont, E. (2013) What singles out the G[8−5]C intrastrand DNA cross-link? mechanistic and structural insights from quantum mechanics/molecular mechanics simulations. Biochemistry 52, 425−431. (27) Gossens, C., Tavernelli, I., and Rothlisberger, U. (2008) DNA structural distortions induced by ruthenium- arene anticancer compounds. J. Am. Chem. Soc. 130, 10921−10928. (28) Spiegel, K., Magistrato, A., Maurer, P., Ruggerone, P., Rothlisberger, U., Carloni, P., Reedijk, J., and Klein, M. L. (2008)

the structure of the product, as in the present study, the DM in water represents a very interesting model that can help to shed some light on several aspects of OTLs chemistry.



ASSOCIATED CONTENT

* Supporting Information S

Additional discussion and bibliographic references on the microscopic QM/MM-MD approach; starting structure of our simulations; and evolution of the stacking coordinate ξ of Florian and co-workers31 along the reaction coordinate d. This material is available free of charge via the Internet at http:// pubs.acs.org.



AUTHOR INFORMATION

Corresponding Authors

*(J.G.) Phone: +33 0 4 72 72 88 46. Fax: +33 0 4 72 72 88 60. E-mail: [email protected]. *(E.D.) E-mail: [email protected]. Funding

This work was performed within the framework of the LABEX PRIMES (ANR-11-LABEX-0063) of Université de Lyon, within the program “Investissements d’Avenir” (ANR-11IDEX-0007) operated by the French National Research Agency (ANR). Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS We thank the PSMN at ENS-Lyon for computational resources. ABBREVIATIONS ROS, reactive oxygen species; OTL, oxidatively generated tandem lesion; GpT, guanine−thymine; DM, dinucleoside monophosphate; Rcage, reactants in water cage; Pcage, products in water cage; RDNA, reactants in DNA; PDNA, products in DNA; QM/MM, quantum mechanics, molecular mechanics; CPMD, Car−Parrinello molecular dynamics; TI, thermodynamic integration



REFERENCES

(1) Cooke, M. S., Evans, M. D., Dizdaroglu, M., and Lunec, J. (2003) Oxidative DNA damage: mechanisms, mutation, and disease. FASEB J. 17, 1195−1214. (2) Wiseman, H., and Halliwell, B. (1996) Damage to DNA by reactive oxygen and nitrogen species: role in inflammatory disease and progression to cancer. Biochem. J. 313, 17−29. (3) Wang, Y. (2008) Bulky DNA lesions induced by reactive oxygen species. Chem. Res. Toxicol. 21, 276−281. (4) Cadet, J., Douki, T., Gasparutto, D., and Ravanat, J.-L. (2003) Oxidative damage to DNA: formation, measurement and biochemical features. Mutat. Res. 531, 5−23. (5) Zeng, Y., and Wang, Y. (2006) Sequence-dependent formation of intrastrand crosslink products from the UVB irradiation of duplex DNA containing a 5-bromo-2′-deoxyuridine or 5-bromo-2′-deoxycytidine. Nucleic Acids Res. 34, 6521−6529. (6) Zeng, Y., and Wang, Y. (2007) UVB-Induced formation of intrastrand cross-link products of DNA in MCF-7 cells treated with 5bromo-2′-deoxyuridine. Biochemistry 46, 8189−8195. (7) Lin, G., Zhang, J., Zeng, Y., Luo, H., and Wang, Y. (2010) Conformation-dependent formation of the G[8−5]U intrastrand Cross-link in 5-bromouracil-containing G-quadruplex DNA induced by UVA irradiation. Biochemistry 49, 2346−2350. (8) Yang, Z., Collis, L. C., Basu, A. K., and Zou, Y. (2006) Recognition and incision of γ-radiation-induced cross-linked guanine1140

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141

Chemical Research in Toxicology

Article

Parameterization of azole-bridged dinuclear platinum anticancer drugs via a QM/MM force matching procedure. J. Comput. Chem. 29, 38−49. (29) Xerri, B., Morell, C., Grand, A., Cadet, J., Cimino, P., and Barone, V. (2006) Radiation-induced formation of DNA intrastrand crosslinks between thymine and adenine bases: a theoretical approach. Org. Biomol. Chem. 4, 3986−3992. (30) Labet, V., Morell, C., Grand, A., Cadet, J., Cimino, P., and Barone, V. (2008) Formation of cross-linked adducts between guanine and thymine mediated by hydroxyl radical and one-electron oxidation: a theoretical study. Org. Biomol. Chem. 6, 3300−3305. (31) Jafilan, S., Klein, L., Hyun, C., and Florián, J. (2012) Intramolecular base stacking of dinucleoside monophosphate anions in aqueous solution. J. Phys. Chem. B 116, 3613−3618. (32) Norberg, J., and Nilsson, L. (1995) Stacking free energy profiles for all 16 natural ribodinucleoside monophosphates in aqueous solution. J. Am. Chem. Soc. 117, 10832−10840. (33) Norberg, J., and Nilsson, L. (1995) Potential of mean force calculations of the stacking-unstacking process in single-stranded deoxyribodinucleoside monophosphates. Biophys. J. 69, 2277−2285. (34) Case, D. et al. (2006) AMBER 9, University of California, San Francisco, CA. (35) Wang, J. M., Cieplak, P., and Kollman, P. A. (2000) How well does a restrained electrostatic potential (RESP) model perform in calculating conformational energies of organic and biological molecules? J. Comput. Chem. 21, 1049−1074. (36) Becke, A. D. (1988) Density-functional exchange-energy approximation with correct asymptotic behavior. Phys. Rev. A 38, 3098−3100. (37) Lee, C., Yang, W., and Parr, R. G. (1988) Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Phys. Rev. B 37, 785. (38) von Lilienfeld, O. A., Tavernelli, I., Rothlisberger, U., and Sebastiani, D. (2004) Optimization of effective atom centered potentials for London dispersion forces in density functional theory. Phys. Rev. Lett. 93, 153004.1−153004.4. (39) von Lilienfeld, O. A., Tavernelli, I., Rothlisberger, U., and Sebastiani, D. (2005) Variational optimization of effective atom centered potentials for molecular properties. J. Chem. Phys. 122, 014113.1−014113.6. (40) Lin, I.-C., and Rothlisberger, U. (2008) Describing weak interactions of biomolecules with dispersion-corrected density functional theory. Phys. Chem. Chem. Phys. 10, 2730−2734. (41) Scott, W. R. P., Hunenberger, P. H., Tironi, I. G., Mark, A. E., Billeter, S. R., Fennen, J., Torda, A. E., Huber, T., Kruger, P., and van Gunsteren, W. F. (1999) The GROMOS biomolecular simulation program package. J. Phys. Chem. A 103, 3596−3607. (42) Sprik, M., and Ciccotti, G. (1998) Free energy from constrained molecular dynamics. J. Chem. Phys. 109, 7737−7744. (43) Warshel, A., Sharma, P. K., Kato, M., Xiang, Y., Liu, H., and Olsson, M. H. M. (2006) Electrostatic basis for enzyme catalysis. Chem. Rev. 106, 3210−3235. (44) Garcia-Meseguer, R., Marti, S., Ruiz-Pernia, J. J., Moliner, V., and Tunon, I. (2013) Studying the role of protein dynamics in an SN2 enzyme reaction using free-energy surfaces and solvent coordinates. Nat. Chem. 5, 566−571. (45) Hong, I. S., Carter, K. N., Sato, K., and Greenberg, M. M. (2007) Characterization and mechanism of formation of tandem lesions in DNA by a nucleobase peroxyl radical. J. Am. Chem. Soc. 129, 4089−4098. (46) Dupont, C., Patel, C., Ravanat, J.-L., and Dumont, E. (2013) Addressing the competitive formation of tandem DNA lesions by a nucleobase peroxyl radical: a DFT-D screening. Org. Biomol. Chem. 11, 3038−3045.

1141

dx.doi.org/10.1021/tx4004616 | Chem. Res. Toxicol. 2014, 27, 1133−1141