Formation mechanism and structure of a guanine-uracil DNA intrastrand cross-link

Spatial and Temporal Resolution of the Oxygen-Independent Photoinduced DNA Interstrand Cross-Linking by a Nitroimidazole Derivative

DNA damage is ubiquitous in nature and is at the basis of emergent treatments such as photodynamic therapy, which is based on the activation of highly oxidative reactive oxygen species by photosensitizing O₂. However, hypoxia observed in solid tumors imposes the necessity to devise oxygen-independent modes of action able to induce DNA damage under a low oxygen concentration. The complexity of these DNA damage mechanisms in realistic environments grows exponentially when taking into account light absorption and subsequent excited-state population, photochemical and (photo)-redox reactions, the multiple species involved in different electronic states, noncovalent interactions, multiple reaction steps, and the large number of DNA reactive sites. This work tackles all the intricate reactivity of a photosensitizer based on a nitroimidazole derivative reacting toward DNA in solution under UV light exposition. This is performed through a combination of ground- and excited-state quantum chemistry, classical molecular dynamics, and hybrid QM/MM simulations to rationalize in detail the formation of DNA interstrand cross-links (ICLs) exerted by the noncanonical noncovalent photosensitizer. Unprecedented spatial and temporal resolution of these phenomena is achieved, revealing that the ICL is sequence-specific and that the fastest reactions take place at AT, GC, and GT steps involving either the opposite nucleobases or adjacent Watson-Crick base pairs. The N7 and O6 positions of guanine, the N7 and N3 sites of adenine, the N4 position of cytosine, and the O2 atom of thymine are deemed as the most nucleophile sites and are positively identified to participate in the ICL productions. This work provides a multiscale computational protocol to study DNA reactivity with noncovalent photosensitizers, and contributes to the understanding of therapies based on photoinduced DNA damage at molecular and electronic levels. In addition, we believe the depth understanding of these processes should assist the design of new photosensitizers considering their molecular size, electronic properties, and the observed regioselectivity toward nucleic acids.

Theoretical studies on the purine radical induced purine-purine type intrastrand cross-links

At the density functional theory (DFT) level, addition reactions between the guanine-8-yl radical and its 3'/5' neighboring purine deoxynucleosides forming the purine-purine type intrastrand cross-links were studied. It is found that addition of the guanine-8-yl radical to the C8 site of its 5' neighboring deoxyguanosine or deoxyadenosine is a two-step reaction consisting of a structurally relatively unfavourable conformational transformation step, while the corresponding 3' C8 addition is straightforward and kinetically more efficient. The 3' C8 preference of the guanine-8-yl radical additions indicates the existence of an obvious sequence effect, which is completely opposite to that observed in the formation of pyrimidine radicals induced DNA intrastrand cross-links. The detrimental effects from steric hindrance and stabilizing weak interactions make these addition reactions markedly suppressed in double stranded DNA. This work broadens our knowledge about the possible types of DNA intrastrand cross-links.

DNA intrastrand cross-links induced by the purine-type deoxyguanosine-8-yl radical: a DFT study

Currently, all known DNA intrastrand cross-links are found to be induced by pyrimidine-type radicals; however, whether or not purine-type radicals are able to cause DNA intrastrand cross-links remains unclear. In the present study, probable additions of the highly reactive deoxyguanosine-8-yl radical to its 3'/5' neighboring pyrimidine nucleotides in four model compounds, 5'-G˙T-3', 5'-TG˙-3', 5'-G˙C-3', and 5'-CG˙-3', were studied using density functional theory (DFT) methods. In single-stranded DNA, the deoxyguanosine-8-yl radical is preferred to efficiently attack the C5 site of its 3' neighboring deoxythymidine or deoxycytidine, forming the G[8-5]T or G[8-5]C intrastrand cross-link rather than the C6 site forming the G[8-6]T or G[8-6]C intrastrand cross-link. The four corresponding sequence isomers, namely T[5-8]G, T[6-8]G, C[5-8]G, and C[6-8]G, formed by additions of deoxyguanosine-8-yl radical to its 5' neighboring pyrimidine nucleotides are predicted to be formed inefficiently. In double-stranded DNA, considering the detrimental effects of stabilizing weak interactions on related structural adjustments required in each addition reaction path, relatively lower reaction yields are suggested for the G[8-5]T and G[8-5]C intrastrand cross-links, while the formation of the other six intrastrand cross-links becomes quite difficult. All calculations definitely demonstrate that, in addition to pyrimidine-type radicals, the purine-type deoxyguanosine-8-yl radical is able to attack its 3'/5' neighboring pyrimidine nucleotides forming several DNA intrastrand cross-links.

Formation of pyrimidine-pyrimidine type DNA intrastrand cross-links: a theoretical verification

Pyrimidine-type radicals have been demonstrated to be able to attack their 3' or 5' neighboring purine nucleotides forming diverse DNA intrastrand cross-links, but whether or not these radicals can attack their surrounding pyrimidine nucleotides forming pyrimidine-pyrimidine type DNA intrastrand cross-links remains unclear. To resolve this question, probable additions of the uracil-5-methyl (˙U_CH2) radical to the C₅[double bond, length as m-dash]C₆ double bond of its 3'/5' neighboring pyrimidine nucleotides in the four models, 5'-T(˙U_CH2)-3', 5'-C(˙U_CH2)-3', 5'-(˙U_CH2)T-3', and 5'-(˙U_CH2)C-3', are explored in the present work employing density functional theory (DFT) methods. The C₆ site of its 5' neighboring thymidine is the preferred target for ˙U_CH2 radical addition, while additions of the ˙U_CH2 radical to the C₆ and C₅ sites of its 5' neighboring deoxycytidine are found to be competitive reactions. The ˙U_CH2 radical can react with both the C₆ and C₅ sites of its 3' neighboring pyrimidine nucleotides, but the efficiencies of these reactions are predicted to be much lower than those of the corresponding addition reactions to its 5' neighboring pyrimidine nucleotides, indicating the existence of an obvious sequence effect. All the addition products could be finally transformed into closed-shell intrastrand cross-links, the molecular masses of which are found to be exactly the same as certain MS values determined in a recent study of an X-irradiated deoxygenated aqueous solution of calf thymus DNA. The present study thus not only definitely corroborates the fact that the reactive ˙U_CH2 radical can attack its 3'/5' neighboring pyrimidine nucleotides forming several pyrimidine-pyrimidine type DNA intrastrand cross-links, but also provides a plausible explanation for the identities of these structurally unknown intrastrand cross-links.

Mechanism studies of addition reactions between the pyrimidine type radicals and their 3′/5′ neighboring deoxyguanosines

To clarify the biologically significant sequence effect existing in the formation of the pyrimidine-type radicals induced DNA intrastrand cross-links, addition mechanisms between the uridine-5-methyl (˙U_CH2), 6-hydroxy-5,6-dihydrothymidine-5-yl (˙T_6OH), and 6-hydroxy-5,6-dihydrocytidine-5-yl (˙C_6OH) radicals and their 3′/5′ neighboring deoxyguanosines (dG) are explored in the present study employing the model 5′-G(˙U_CH2)-3′, 5′-(˙U_CH2)G-3′, 5′-G(˙T_6OH)-3′, 5′-(˙T_6OH)G-3′, 5′-G(˙C_6OH)-3′, and 5′-(˙C_6OH)G-3′ sequences. It is found that the 5′ G/C₈ additions of the three radicals are all simple direct one-step reactions inducing only relatively small structural changes, while a conformational adjustment involving orientation transitions of both nucleobase moieties and twisting of the DNA backbone is indispensable for each 3′ G/C₈ addition. Furthermore, markedly positive reaction free energy requirements are estimated for these conformational transformations making the 3′ G/C₈ additions of the three radicals thermodynamically much more unfavorable than the corresponding 5′ G/C₈ additions. Such essential conformational adjustments along the 3′ G/C₈ addition paths that structurally greatly influence the local DNA structures and thermodynamically substantially reduce the addition efficiencies may be the reasons responsible for the differences in the formation yields and biological consequences of the pyrimidine-type radicals induced DNA intrastrand cross-link lesions.

For each radical, the 5′ G/C₈ addition is a simple direct one-step reaction, while a structurally significant and thermodynamically markedly unfavorable conformational adjustment is indispensable for the 3′ G/C₈ addition.

Structural basis for substrate discrimination byE. colirepair enzyme, AlkB

Positive charge on methylated nucleotides is a prime criterion for substrate recognition byE. coliAlkB.

Interstrand cross-linking implies contrasting structural consequences for DNA: insights from molecular dynamics

Oxidatively-generated interstrand cross-links rank among the most deleterious DNA lesions. They originate from abasic sites, whose aldehyde group can form a covalent adduct after condensation with the exocyclic amino group of purines, sometimes with remarkably high yields. We use explicit solvent molecular dynamics simulations to unravel the structures and mechanical properties of two DNA sequences containing an interstrand cross-link. Our simulations palliate the absence of experimental structural and stiffness information for such DNA lesions and provide an unprecedented insight into the DNA embedding of lesions that represent a major challenge for DNA replication, transcription and gene regulation by preventing strand separation. Our results based on quantum chemical calculations also suggest that the embedding of the ICL within the duplex can tune the reaction profile, and hence can be responsible for the high difference in yields of formation.

DNA Distortion Caused by Uracil-Containing Intrastrand Cross-Links

Four uracil-containing intrastrand cross-links have been detected in human cells upon UV irradiation of 5-bromouracil-containing DNA, namely 5'-G[8-5]U-3', 5'-U[5-8]G-3', 5'-A[8-5]U-3', and 5'-A[2-5]U-3'. These lesions feature unique composition and connectivity compared with other intrastrand cross-links reported in the literature. For the first time, structural information obtained using molecular dynamics (MD) simulations reveal that all four lesions distort the DNA helix, which can involve an extrahelical location of the cross-link, changes in the helical interactions of the complementary nucleotides, or disruption of hydrogen bonding in the flanking base pairs up to two positions from the cross-linked site; however, the degree of distortion varies between the cross-links, being affected by the sequence, nucleobase-nucleobase connectivity, and the purine involved. Most importantly, the relative distortion of the damaged DNA provides the first structural explanation for the observed abundances of the four uracil-containing cross-links. Furthermore, the highly distorted conformations suggest that these lesions will likely have severe implications for DNA replication and repair processes in cells.

Insights into the structure of intrastrand cross-link DNA lesion-containing oligonucleotides: G[8-5m]T and G[8-5]C from molecular dynamics simulations

Oxidatively generated complex DNA lesions occur more rarely than single-nucleotide defects, yet they play an important role in carcinogenesis and aging diseases because they have proved to be more mutagenic than simple lesions. Whereas their formation pathways are rather well understood, the field suffers from the absence of structural data that are crucial for interpreting the lack of repair. No experimental structures are available for oligonucleotides featuring such a lesion. Hence, the detailed structural basis of such damaged duplexes has remained elusive. We propose the use of explicit solvent molecular dynamics simulations to build up damaged oligonucleotides containing two intrastrand cross-link defects, namely, the guanine-thymine and guanine-cytosine defects. Each of these lesions, G[8-5m]T and G[8-5]C, is placed in the middle of a dodecameric sequence, which undergoes an important structural rearrangement that we monitor and analyze. In both duplexes, the structural evolution is dictated by the more favorable stacking of guanine G6, which aims to restore π-stacking with the 3' purine nucleobase. Subsequently, transient formation of hydrogen bonds with a strand shifting is observed. Our simulations are combined with density functional theory to rationalize the structural evolution. We report converging computational evidence that the G[8-5m]T- and G[8-5]C-containing structures evolve toward "abasic-like" duplexes, with a stabilization of the interstrand pairing noncovalent interactions. Meanwhile, both lesions restore B-helicity within tens of nanoseconds. The identification of plausible structures characterizes the last hydrogen abstraction step toward the formation of such defects as a non-innocent chemical reaction.

Glycosidic bond cleavage in deoxynucleotides: effects of solvent and the DNA phosphate backbone in the computational model

Density functional theory (B3LYP) was employed to examine the hydrolysis of the canonical 2'-deoxynucleotides in varied environments (gas phase or water) using different computational models for the sugar residue (methyl or phosphate group at C5') and nucleophile (water activated through full or partial proton abstraction). Regardless of the degree of nucleophile activation, our results show that key geometrical parameters along the reaction pathway are notably altered upon direct inclusion of solvent effects in the optimization routine, which leads to significant changes in the reaction energetics and better agreement with experiment. Therefore, despite the wide use of gas-phase calculations in the literature, small model computational work, as well as large-scale enzyme models, that strive to understand nucleotide deglycosylation must adequately describe the environment. Alternatively, although inclusion of the phosphate group at C5' also affects the geometries of important stationary points, the effects cancel to yield unchanged deglycosylation barriers, and therefore smaller computational models can be used to estimate the energy associated with nucleotide deglycosylation, with the 5' phosphate group included if full (geometric) details of the reaction are desired. Hydrogen-bonding interactions with the nucleobase can significantly reduce the barrier to deglycosylation, which supports suggestions that discrete hydrogen-bonding interactions with active-site amino acid residues can play a significant role in enzyme-catalyzed nucleobase excision. Taken together with previous studies, the present work provides vital clues about the components that must be included in future studies of the deglycosylation of isolated noncanonical nucleotides, as well as the corresponding enzyme-catalyzed reactions.