Published online 24 April 2021 Nucleic Acids Research, 2021, Vol. 49, No. 9 4919–4933 doi: 10.1093/nar/gkab285 A unified computational view of DNA duplex, triplex, quadruplex and their donor–acceptor interactions Gyuri Park1, Byunghwa Kang1, Soyeon V. Park1, Donghwa Lee1,2,3,* and Seung Soo Oh 1,3,4,* 1Department of Materials Science and Engineering, Pohang University of Science and Technology (POSTECH), Pohang 37673, South Korea, 2Division of Advanced Materials Science, Pohang University of Science and Technology (POSTECH), Pohang 37673, South Korea, 3Institute for Convergence Research and Education in Advanced Technology (I-CREATE), Yonsei University, Incheon 21983, South Korea and 4School of Interdisciplinary Bioscience and Bioengineering, Pohang University of Science and Technology (POSTECH), Pohang 37673, South Korea Received October 15, 2020; Revised April 07, 2021; Editorial Decision April 07, 2021; Accepted April 14, 2021 ABSTRACT INTRODUCTION DNA can assume various structures as a result Deoxyribonucleic acid (DNA) is a unique material as com- of interactions at atomic and molecular levels posed of nitrogenous bases (adenine (A), thymine (T), (e.g., hydrogen bonds, ␲–␲ stacking interactions, guanine (G) or cytosine (C)), sugar rings, and phosphate and electrostatic potentials), so understanding of groups. DNA is programmable, so it can be rationally the consequences of these interactions could guide designed into molecular structures ranging from simple Watson–Crick base-pairing primers to DNA origami-based development of ways to produce elaborate pro- complex 3D constructs (1,2). Even spatial and temporal grammable DNA for applications in bio- and nan- control of DNA nanostructures is achievable; sophisticated otechnology. We conducted advanced ab initio cal- DNA molecular machines can perform a series of nanome- culations to investigate nucleobase model struc- chanical motions in a controllable manner, so this abil- tures by componentizing their donor-acceptor inter- ity provides unprecedented applications in bio- and nan- actions. By unifying computational conditions, we otechnology (3–6). Such active use of DNA requires funda- compared the independent interactions of DNA du- mental understanding of DNA folding and its stabilization. plexes, triplexes, and quadruplexes, which led us In particular, DNA hybridization programming extensively to evaluate a stability trend among Watson–Crick exploits understanding of various donor-acceptor interac- ␲ ␲ and Hoogsteen base pairing, stacking, and even tions of DNA, including hydrogen bonds, – stacking in- ion binding. For a realistic solution-like environ- teractions, and electrostatic potentials (7). Hydrogen bonds and ␲–␲ stacking are among the most ment, the influence of water molecules was care- important intra- and inter-molecular interactions in DNA fully considered, and the potassium-ion preference (8–11). In general, nucleobases create specific hydrogen of G-quadruplex was first analyzed at an ab initio bonds between a purine (A or G) and a pyrimidine (C and level by considering both base-base and ion-water T), which yield planar Watson–Crick base pairs (A–T; G– interactions. We devised new structure factors in- C). Between adjacent base pairs, ␲–␲ stacking interactions cluding hydrogen bond length, glycosidic vector an- result in sequential stacking of base pairs. Therefore, hy- gle, and twist angle, which were highly effective for drogen bonds and ␲–␲ stacking drive the formation of a comparison between computationally-predicted and duplex as a basic structure of DNA. However, the nucle- experimentally-determined structures; we clarified obases can be influenced by different interactions due to pH the function of phosphate backbone during nucle- and metal ions, so triplexes and quadruplexes are sometimes obase ordering. The simulated tendency of net inter- produced (12,13). For instance, C is protonated at slightly acidic pH to become a hydrogen-bond acceptor C+, which action energies agreed well with that of real world, binds to the guanine of a G–C pair as a bond donor to and this agreement validates the potential of ab ini- yield a C+•G–C triad in a DNA triplex (14). The C+ can tio study to guide programming of complicated DNA be also paired with a non-protonated form of C, and the re- constructs. sulting C–C+ pairs involve in formation of C-quadruplex, i.e., i-motif (15). Moreover, guanines can interact strongly *To whom correspondence should be addressed. Tel: +82 54 279 2144; Email: [email protected] Correspondence may also be addressed to Donghwa Lee. Tel: +82 54 279 2160; Email: [email protected] C The Author(s) 2021. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 4920 Nucleic Acids Research, 2021, Vol. 49, No. 9 with metal ions, and monovalent cations can influence a structures by ab initio calculations and experimentally- guanine-rich DNA strand to form a G-quadruplex (16,17). determined structures by X-ray crystallography and nuclear DNA should build thermodynamically-favored structures, magnetic resonance (NMR) spectroscopy, and those anal- which is a basic principle behind the spatiotemporal control yses were highly useful to identify how phosphate back- of DNA nanoconstructs by modifying their environmental bones influence nucleobase interactions. Owing to the uni- conditions. fied computational view, various DNA constructs could Rational creation of advanced DNA structures requires be compared to each other to provide useful information, in-depth understanding of DNA interactions under a vari- such as a stability trend among Watson–Crick base pairing, ety of conditions; ab initio calculations can provide valu- Hoogsteen base pairing, stacking, and metal-ion binding. able insights in the behavior of proposed designs (18). This method provides reasonable values of dipole moments, charge distributions, and vibrational frequencies, so it can MATERIALS AND METHODS be useful to describe non-covalent molecular interactions (19,20). Importantly, all donor-acceptor interactions can For our ab initio calculations, we used the Gaussian 09 be easily itemized, so that their independent functions in (G09) program. To effectively investigate nucleobase inter- structure stabilization can be readily analyzed, and the rele- actions, DNA backbones including sugar rings and phos- vant net energy is computed precisely. Therefore, the ab ini- phate groups were replaced with hydrogen atoms. Based on tio simulation has been exploited to interpret nucleic acid the ideal structure of B-DNA, an initial model of two-layer interactions. However, most previous studies have consid- structure was constructed with a parallel arrangement of base-pair layers, maintaining 3.4-A˚ stacking distance and ered only interactions in gas phase (21–24), whereas actual ◦ DNA structures are in aqueous solutions. Some researchers 36 angle under its center of mass. Previous calculation have conducted phase-dependent simulation (25–27), but and experimental determination reports have validated the feasibility of this model (24,29). The double-layer model the calculation works used different computational condi- = tions, making itemized DNA interactions not comparable was guided to initially have a fixed dihedral angle (Opt to each other. Furthermore, the difference between compu- ModRed), minimizing possible non-planarity caused by the tational and experimental results could not be well inter- absence of a backbone, and its optimization process was then followed. Every structure optimization and relaxation preted due to lack of effective analytical factors (26,28); in / // / some calculations, the Watson–Crick base pair was less sta- was conducted under M05–2X 6–31G(d,p) M05–2X 6– ble than a mismatch (MM) pair, i.e., a non-Watson–Crick 31G(d,p) condition, and the basis set superposition error base pair (26), and stacking of bases did not yield planarity (BSSE) calculations were further corrected by the counter- (28). Therefore, there is a strong need for more realistic cal- poise (CP) method (30). The water solvent effect was real- culations of itemized DNA interactions under the unified ized by using a conductor-like screening model (COSMO) conditions and analyses, making the computational results (31). comparable with experimental ones. Shapes of DNA structures were determined by consider- In this work, we conducted advanced ab initio calcu- ing the numbers of bases and layers. The number of bases af- fects the hydrogen bonding among the nucleobases, whereas lations to simulate DNA duplexes, triplexes, and quadru- ␲ ␲ plexes by using identical computation conditions and ratio- the number of layers represents – stacking interactions nally itemized donor-acceptor interactions (Scheme 1a). To among the base-pair layers. Metal ions were added to con- provide the fundamental information for designing and ma- struct G-tetrad or G-quadruplex structures in which the nipulating complicated DNA structures along with predic- cations interact electrochemically with guanines. All interactions are expressible in simple formulae. The tions of their potential stability and transient motions, we systematically analyzed stability orders of the itemized in- binding energy Ebind is determined as the