Ab-Initio Molecular Dynamics Simulation of the Electrolysis of Nucleobases

Electrolysis is potentially a valuable tool for cleansing waste water. One might even hope that it is possible to synthesize valuable products in this way. The question is how the reaction conditions can be chosen to obtain desired compounds. In the present study we use Car–Parrinello molecular dynamics to simulate the reaction of nucleobases under electrolytic conditions. We use our own scheme (F. Hofbauer, I. Frank, Chem. Eur. J., 18, 277, 2012) for simulating the conditions after the electron transfer in a self-consistent field calculation. This scheme was employed previously to the electrolysis of pure water and of polluted solutions. On the picosecond timescale, we find a strongly different reaction behavior for each of the four nucleobases contained in DNA.


Introduction
Two billion people have no access to clean toilets. Having clean toilets is not possible without clean water. In certain situations water is rare in contrast to electric current. This applies to arid regions on Earth but also to the situation in the international space station (ISS). It is necessary to have closed circuits which involve fast and efficient cleansing of waste water. Cleansing waste water is possible by various means. After filtration and biochemical processes it still contains bioorganic compounds. Nanofiltration with pores of about 2 nm can change that to some degree. Compounds like amino acids and nucleobases have a size of about 1 nm, hence even with nanofiltration they are not completely removed. The same is true for compounds like urea and uric acid which we investigated previously [1]. Electrolysis might help to remove such undesired compounds from a solution. It is even possible that valuable products are formed by electrolysis such as molecular hydrogen. In the present paper we investigate the fate of nucleobases. The experimental interest in the electrochemistry of nucleobases is old [2,3], but there are also more recent publications which typically describe more complex systems [4,5]. For simulating electrolytic reactions, we use our established scheme for simulating the situation near an electrode (Figure 1, see also [1,6,7]).
In this scheme we consider the situation after mass transport of positively/negatively charged species towards the negatively/positively charged electrodes and discharging of these species near the electrode. We simulate anodic and cathodic reactions separately which is justified by the fact that, seen at the nanometer scale, the two electrodes are far away from each other. After the mass transport, the chemical situation is clearly different for anodic and cathodic space. Near the anode OH − ions are discharged. Hence we simulate the situation of an aqueous liquid containing a pollutant by removing 4 H + ions and 4 electrons e − from 4 water molecules in the mixture. Close to the cathode, H + or H 3 O + is discharged. Here we remove 4 OH − ions and add four electrons. Alternatively, one could have added four H + ions and four electrons. The numbers are chosen in a way to observe reactions soon without too high concentrations of OH . or H . radicals; see also the Methods section. The situation before applying a potential is stable. The solution contains a small amount of OH − and H 3 O + molecules (in a much smaller concentration than shown here). (b) If a potential is applied there is a mass transport of the OH − ions to the anode and of the H 3 O + ions to the cathode. The same accounts for other charged species which may be formed during the reaction. (c) Close to the electrode surfaces, the ions are discharged. Due to electronic tunneling, a direct contact to the electrode is not necessarily needed. As a result, we have a pollutant in contact with radicals and with water. We take this as the starting point of our simulations.
Our procedure is different from other approaches [8][9][10]. It resembles the approach we are using when simulating photoreactions [11]: there we equilibrate the system in the ground state, then switch instantaneously to the excited state and let it go. This means that we have no reliable way to describe phenomena like the excitation itself on the attosecond timescale. We rather see what is properly called photochemistry, that is, the breaking and formation of chemical bonds. In addition, in the case of electrochemistry we instanteneously change the composition. We do not make an attempt to simulate the electron transfer itself, which is difficult in a reproducible way if using a self-consistent field (SCF) approach. Furthermore, we apply no external field. This is justified by the fact that we compute very small systems (≈1 nm 3 ), and at this scale the potential difference is negligible. We focus on the chemistry triggered by the instantaneous electron transfer. For the simulations we use the Car-Parrinello molecular dynamics code [12,13] with the BLYP functional [14]; for details see the Methods section.
Starting in the vast field of the rich nucleobase electrochemistry, we want to keep things as simple and elementary as possible. Hence, in the present study, we simulate the electrochemistry of the nucleobases in water. Further steps will be the addition of an electrolyte and maybe also of an electrode and/or a heterogeneous catalyst. Ideally the electrode does not undergo chemical reactions, but there might be to some degree also surface catalysis which is different from a purely electronic catalysis.

Adenine
We start with adenine (compound 1, Figure 2). The observed reaction mechanisms have some properties in common for the anodic and the cathodic reactions. In both cases we observe no ring opening. The hydrogen atom connected to the six-membered ring is never attacked. The same is true for the hydrogen atom connected to the five-membered ring. We observe several reaction products, see Figure 2. The compounds 1a and 1b are formed by the simple abstraction of hydrogen atoms from adenine. OH radicals are the active moieties. By abstraction of hydrogen, they are transformed into stable water molecules. An alternative reaction is the addition of an OH radical, leading to compound 1c. All these reactions of OH radicals which occur at the anode are strongly determined by the initial geometry. A more interesting reaction product is compound 1d, which is formed at the cathode in three of eight simulations. It represents a less stable isomer of adenine (BLYP: 7.3 kcal/mol, B3LYP: 8.4 kcal/mol). The mechanism is described in Figures 3 and 4. Figure 3 demonstrates how this reaction is connected with the formation of molecular hydrogen. It is possible with little motion of the nuclei. Figure 4 shows the change of single coordinates. The H 2 formation is the driving force for the formation of a less stable isomer.

Cytosine
Cytosine (2, Figure 5) showed almost no reactivity at the cathodic side, while interesting products were formed under anodic conditions. Again, the rings were not broken. In addition, like for adenine, some parts of the structure are relatively stable, in particular the CO double bond and the CC double bond. The latter is attacked only in one case which leads to formation of compound 2c. Compounds 2a and 2b are formed by simple hydrogen abstraction, compound 2d by OH addition. Compound 2e was formed at the cathodic side by addition of nascent hydrogen. Its formation is strongly dependent on the initial geometry; normally cytosine is stable under cathodic conditions. The most interesting product is compound 2f. After anodic hydrogen abstraction, dimerization is possible. This result was observed in two out of eight simulations. Nevertheless, as it was observed immediately after discharging, its formation was strongly dependent on the initial geometry (Figures 6 and 7). The abstraction of two hydrogen atoms by two OH radicals is complete after 0.05 ps. The formation of the N-N bond occurs about 0.1 ps later (Figures 7 and 8). A movie of the reaction, showing also the spin densities, is available in the Supplementary Material. Remarkably, during the reaction, almost no spin densities are observed, hence almost all electrons succeed in being paired during this reaction. This is in contrast to what we observed for other compounds [1,6].

Guanine
Electrolysis of guanine 3 leads to a variety of interesting products (Figure 9). 3a is formed near the anode by simple hydrogen abstraction of the OH radical. It is also observed, however, near the cathode by hydrogen abstraction of nascent hydrogen. The products 3b-3e stem from several reaction steps, which, in the anodic space, always consist of hydrogen abstraction and OH addition. 3g is the result of two isomerization reactions at the cathode. Finally, we observe a ring-opening reaction leading to 3f. Figures 10 and 11 illustrate the reaction mechanism for this ring opening. It is started by two abstraction reactions in which OH radicals take two hydrogen atoms from the ring system. This unstable situation is stabilized about 0.1 ps later by the addition of a third OH radical which again leads to the opening of the ring about 0.3 ps after the start of the reaction.

Thymine
Electrolysis of thymine 4 led to several products without a clear pattern ( Figure 12). Again, the single reaction steps are simple hydrogen abstractions or OH additions. Except in the case of 4f, where the methyl group is oxidized, the resulting molecules are unstable radicals and are rather intermediates than products. Figure 12. Reaction products of the electrolysis of thymine 4. All these products were observed near the anode. 4c was observed near both electrodes.

Energetics
The energy changes connected with the most important pathways are listed in Table 1, as computed with Hartree-Fock, post-Hartree-Fock and DFT methods [14][15][16][17]. The reactions observed after the electron transfer are mostly exothermic. An exception is the isomerization of 1 to form 1d, which is just catalyzed by the electrical current. Hartree-Fock, as usual [18], deviates quite a bit, while the hybrid and double hybrid approaches nicely converge to reasonable results.

Conclusions
What do we learn from that? It is obvious that all four bases have a different reactivity. The reactivity of the bases near the cathode is lower than that near the anode. Generally, ring opening reactions are rare. We observe primarily radical reactions at the anode, namely additions and abstractions. In both electrode spaces, direct radical recombination reactions are observed leading to HOOH and H 2 , respectively, whereby the latter reaction is dominating at the cathode and is responsible for the low reactivity with the pollutant under these conditions.
The most characteristic reaction in the case of adenine is the cathodic isomerization reaction leading to compound 1d. This is not in agreement with experiment, where a reduction is described [2]. In view of our limited statistics this discrepancy may be accidental. In the case of cytosine, the anodic dimerization leading to 2f is the most interesting reaction. At the cathode we observe the formation of 2e, which is in agreement with the first step of the reaction observed in experiment [3]. In the case of guanine the ring opening reaction leading to 3f is a remarkable feature. Finally, in the electrolysis of thymine, no characteristic pattern was observed. A cathodic reaction different from the evolution of molecular hydrogen was observed in a single case only.
In view of the surprisingly many different reaction pathways observed, it has to be stated that the investigation of 128 nucleobases in 64 reaction runs is merely a starting point. In contrast to amino acids, where the formation of carbondioxide is predominant, we observe a wide variety of products. This accounts to all four nucleobases. Our simulations give just a crude idea, which must be checked experimentally. The number of runs is limited not only by CPU time but also by the time needed for the evaluation. It is not practical to evaluate several hundred reactive molecular dynamics simulations if the reaction pathway is complex and totally unknown.
We observed little which would answer the original question how one might form harmless or even precious compounds by electrolysis of contaminated water-an exception is of course the cathodic formation of molecular hydrogen. We can contribute, however, a little to the general knowledge about DNA bases: there is nothing which could be identified as a special property or reactivity which would make the four nucleobases the obvious building blocks of the genetic code. On the other hand, these different chemical reactivities might be useful for DNA sequencing using nanodevices [19,20].

Methods
Car-Parrinello molecular dynamics simulations [12,13,21] have been performed in the NVE ensemble using the Becke-Lee-Yang-Parr (BLYP) functional in connection with the Grimme dispersion correction [22]. The time step was chosen as 2 a.u. (0.048 fs) and the fictitious electron mass as 200 a.u. Troullier-Martins pseudopotentials as optimized for the BLYP functional were employed for describing the core electrons [23,24]. The plane-wave cutoff which determines the size of the basis set was set to 70.0 Rydberg. The simulation cell size was 20 × 20 × 20 a.u. 3 (10.6 × 10.6 × 10.6 Angstrom 3 ). Solutions with a density of roughly 1 g/cm 3 were generated and equilibrated for 0.96 ps using the TEMPCONTROL option with the parameters 300 and 100. After equilibration of stable, neutral closed-shell systems, reactive species were generated by removing four protons and four electrons, leading to OH . radicals which simulates the anodic conditions. Alternatively, to simulate cathodic conditions, OH was removed from the solution leading to nascent hydrogen. For the reactive simulations, which took 0.48 ps, the spin-unrestricted version of Kohn-Sham theory was employed [25]. In the production runs, the temperature was not controlled. A total of eight simulations was performed for all four nucleobases, both for the anodic and the cathodic reactions. In every simulation cell, two nucleobases were surrounded by water molecules, hence we have to evaluate the fate of 128 nucleobases.