Relaxation mechanisms of UV-photoexcited DNA

Relaxation mechanisms of UV-photoexcited DNA and RNA nucleobases Mario Barbattia,b,1, Adélia J. A. Aquinoa, Jaroslaw J. Szymczaka, Dana Nachtigallovác, Pavel Hobzac, and Hans Lischkaa,c,1 a Institute for Theoretical Chemistry, University of Vienna, Waehringerstrasse 17, A 1090 Vienna, Austria; bMax-Planck-Institut für Kohlenforschung, Kaiser-Wilhelm-Platz 1, D 45470 Mülheim an der Ruhr, Germany; and cInstitute of Organic Chemistry and Biochemistry, Academy of Sciences of the Czech Republic, Flemingovo nam. 2, CZ-16610 Prague 6, Czech Republic.

Edited* by Josef Michl, University of Colorado at Boulder, Boulder, CO, and approved October 15, 2010 (received for review October 6, 2010)

O

wing to the importance of mutagenic and carcinogenic effects caused by UV radiation on DNA, the UV-induced photochemistry and photophysics of individual bases (1–7), base pairs (8–10), and nuclei acid strands (11–13) have been intensively studied. In each of these levels, it has been found that the genetic code may count on molecular mechanisms to get rid of the energy excess minimizing deleterious effects. Isolated nucleobases (Scheme 1), for instance, are all photostable by returning to the ground state in an ultrafast time scale of a few picoseconds (1, 2, 14–20), which reduces the probability of undergoing photochemical transformations induced by reactive excited states. The ultrafast deactivation of the five natural nucleobases contrasts with the long lifetime of analogous bases not found in DNA and RNA (2, 21), suggesting that it could have been a source of evolutionary pressure in early biotic ages. Ultrafast deactivation occurs through internal conversion mechanisms, where the molecule transfers the photoenergy stored in the electronic system to nuclear vibrational degrees of freedom. This means, the deactivation takes place without photoemission in contrast to what occurs for fluorescent and phosphorescent species. The energy transfer proceeds more intensely near nuclear geometries with state degeneracies also known as conical intersections, which motivates the search for mechanisms explaining how the molecular geometry is deformed from the ground state (or Franck–Condon) structure initially photoexcited www.pnas.org/cgi/doi/10.1073/pnas.1014982107

into the strongly distorted geometries where conical intersections usually occur. Even though the nucleobases share the property of photostability, their structures, electronic properties, and spectra show considerable variation. Therefore, the question arises whether there is a common pattern in the deactivation dynamics of the nucleobases or does one encounter only a collection of processes that finally lead to the same result of photostability. Advanced techniques of theoretical chemistry are very well suited for performing such comparative studies even though the severe difficulties have to be recognized where several electronic states and their nonadiabatic interactions have to be considered at the same time. Together with ultrafast time-resolved spectroscopy (1, 2, 22), these methods have provided a high degree of detail in the characterization of internal conversion processes, suggesting ring puckering as one major structural theme for the photodeactivation of nucleobases (3, 23–26). Based on the available knowledge of computed structures and connecting reaction paths, a unifying description of the photostability of all nucleobases has been proposed (5, 23, 27). However, considering the fact that up to now for all nucleobases about 20 energetically accessible minimum energy conical intersection structures, which still describe the intersection seam only in a rudimentary way, are known, the goal of providing a theoretical explanation for the photodynamics based on description of reaction paths alone seems to be almost Author contributions: H.L. designed research; M.B., A.J.A.A., J.J.S., D.N., and H.L. performed research; M.B., A.J.A.A., J.J.S., D.N., P.H., and H.L. analyzed data; and M.B., A.J.A.A., J.J.S., D.N., P.H., and H.L. wrote the paper. The authors declare no conflict of interest. *This Direct Submission article had a prearranged editor. 1

To whom correspondence may be addressed. E-mail: [email protected] or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/ doi:10.1073/pnas.1014982107/-/DCSupplemental.

PNAS ∣ December 14, 2010 ∣ vol. 107 ∣ no. 50 ∣ 21453–21458

BIOPHYSICS AND COMPUTATIONAL BIOLOGY

photodynamical simulation ∣ photostability ∣ ultrafast photodeactivation ∣ nonadiabatic interactions ∣ ab initio multireference methods

Scheme 1.

CHEMISTRY

A comprehensive effort in photodynamical ab initio simulations of the ultrafast deactivation pathways for all five nucleobases adenine, guanine, cytosine, thymine, and uracil is reported. These simulations are based on a complete nonadiabatic surface-hopping approach using extended multiconfigurational wave functions. Even though all five nucleobases share the basic internal conversion mechanisms, the calculations show a distinct grouping into purine and pyrimidine bases as concerns the complexity of the photodynamics. The purine bases adenine and guanine represent the most simple photodeactivation mechanism with the dynamics leading along a diabatic ππ* path directly and without barrier to the conical intersection seam with the ground state. In the case of the pyrimidine bases, the dynamics starts off in much flatter regions of the ππ* energy surface due to coupling of several states. This fact prohibits a clear formation of a single reaction path. Thus, the photodynamics of the pyrimidine bases is much richer and includes also nπ* states with varying importance, depending on the actual nucleobase considered. Trapping in local minima may occur and, therefore, the deactivation time to the ground state is also much longer in these cases. Implications of these findings are discussed (i) for identifying structural possibilities where singlet/ triplet transitions can occur because of sufficient retention time during the singlet dynamics and (ii) concerning the flexibility of finding other deactivation pathways in substituted pyrimidines serving as candidates for alternative nucleobases.

impossible. Dynamics simulations are required to add a new quality in terms of a realistic description of these complex photophysical systems and to provide extended tools for their assessment and understanding. To this date, photodynamical simulations have been performed at the ab initio level for selected nucleobases under varying conditions (6, 22, 28–30). The objective of the present work is the report of a comprehensive dynamics study on all five DNA/RNA bases using nonadiabatic trajectory dynamics simulations based on high-level ab initio methods for guanine, cytosine, and uracil combined with a reanalysis of the results of our recent dynamics simulations of adenine (6) and thymine (30). The aim is to show how the almost overwhelming flood of information provided by these calculations can be reduced in order to arrive at a general picture taking into account the details of the dynamics simulations but also keeping track of the general principles governing the photostability of DNA bases. The present results can also be used to gauge semiempirical work (31) resulting in efficient procedures for performing much faster photodynamical simulations also on significantly larger molecular systems. Characterization of Conical Intersections The large number of distinct conical intersections occurring for each nucleobase has already been mentioned above (3–6, 23, 32, 33). Before proceeding to the discussions, it will be useful to summarize main findings and to introduce a symbolic notation to allow reference to different conical intersections in a more pictorial way. Typical geometrical distortions producing conical intersections are illustrated in Fig. 1 A–D with characteristic examples shown in Fig. 1 E–G. In this figure, the thick lines represent the molecular ring plane and the atoms moved out of plane are connected with thin lines. Dotted lines indicate dissociated bonds. One main way to reach a conical intersection between the first excited state and the ground state is by twisting a double bond by about 90° forming a biradical state in which the energy gap between ground and excited states is reduced and using an additional degree of freedom like a wagging mode to tune the intersection (34). When this procedure is performed in a heterocycle, it gives rise to puckered structures. In the lowest energy conical intersection of adenine, for instance, the C2 H group is moved out of plane in order to become twisted in relation to the N1 atom (in plane) as shown in Fig. 1E. This kind of conical intersection describes a crossing between a ππ* state and the closed shell (cs) state.

Fig. 1. (A–D) Symbolic representation of conical intersections. Examples of conical intersections between the first excited state and the ground state for (E) adenine, (F) thymine, and (G) cytosine. A symbolic representation is shown at Right. 21454 ∣

www.pnas.org/cgi/doi/10.1073/pnas.1014982107

Several other types of conical intersections are found in the nucleobases as well. In the one given in Fig. 1B, the oxygen atom is moved out of plane as also shown in Fig. 1F for thymine, leading to a nO π ∕cs crossing. Fig. 1 C and D correspond to conical intersections formed by NH stretching and ring opening structures, respectively (35). In these latter two cases, the crossing is between πσ* and the cs states. The pyrimidine bases also possess conical intersections with quasiplanar conformation, like the one shown in Fig. 1G for cytosine (the dashed line indicates a puckering distortion in a background plane). An overview of the main types of conical intersections occurring for all five nucleobases and their relative stabilities is presented in Fig. 2. The energies given correspond to reference complete active space perturbation theory to second order (CASPT2) and multireference configuration interaction (MRCI) values taken from refs. 6, 23, and 33 and calculations of the present work. The figure indicates the energetic accessibility with respect to the first UV absorption band. Dynamics Simulations: Time Constants and Deactivation Mechanisms.

The deactivation pathways and the time scales at which individual stages are traversed are intimately connected and provide major insight into the entire photodeactivation process. In addition to this detailed view, global time constants have been determined as they can be directly compared to those obtained from fitting decay curves determined from pump–probe experiments. The outline of the subsequent sections is as follows: the presentation starts with a comparison of computed and experimental time constants in order to set the stage for the dynamics simulations and to present the achievable accuracy range. Next, the deactivation mechanisms are summarized and individual cases are characterized. Finally, the question of a unifying reaction scheme is discussed and an assessment of the impact of photodynamical simulations is given. Global time constants: an overview. A summary of experimental time constants for nucleobases and some relevant derivatives in gas phase are given in Table 1. A comprehensive version of Table 1 is given in Table S1. Short decay constants (∼100 fs) are very sensitive to the way the raw data are deconvoluted, being influenced by instrumental response, coherent processes, and the fast change in the ionization potential during the motion of the wave packet out of the Franck–Condon region. Because our dynamics simulations do not take into account any of these factors, we shall concentrate on the intermediary time constants (0.5–5 ps), which are more likely caused by deactivation to the ground state. Among the nucleobases, most of the experiments in gas phase have been performed for adenine, which provides detailed information on the dependence of the excited-state lifetime with the pump wavelength (1, 2, 14–20). The intermediary time constant

Fig. 2.

Main types of S0 ∕S1 conical intersections for all five nucleobases. Barbatti et al.

Table 1. Experimental and simulated (present work) time constants for deactivation of nucleobases in gas phase Base

λpump , nm

τ1 , ps

τ2 , ps

Ade

250 267 277 267 267 267 267 250 267 239 250 267 239 236 250 267 314

1.7 63

0.53 ½0.65 ½0.74

1.5 1.1 (2.4) ½> 1.8 12 24

1.7* >2.3* >1.8* >1.5* 3.08

77 55 13

L, C, and H stand for low-, center-, and high-energy window in the initial conditions, whose technical definitions are given in SI Text. *S2 → S1 decay time. Barbatti et al.

In the case of gas phase uracil, experiments at 250 nm, near the first band center, show an intermediary (0.5 ps) and a long (2.4 ps) time constant. At 267 nm, only the long time constant is reported. The simulations have been performed starting the trajectories in two initial energy windows, a low-energy window, which should be compared to experiments pumped at 267 nm, and a high-energy window, which should be compared to experiments pumped at 250 nm. Most of trajectories of the low-energy window decay with a time constant much longer than the maximum time of our simulations. The fitting of the S2 occupation shows that the S1 state is populated in 1.8 ps. This establishes the lower limit for the long time constant. A minor fraction of trajectories in the low-energy window decays to the ground state with the intermediary time constant of 0.74 ps. In the high-energy window, uracil trajectories are split into those that deactivate to the ground state in 0.65 ps and those with a longer time constant for which the lower limit is 1.5 ps. For gas phase cytosine, experiments at 250-nm pump wavelength show an intermediary time constant of 0.8 ps and a long time constant of 3.2 ps (Table 1). At 267 nm, near the center of the first absorption band, the intermediary time constant has not been reported, whereas different measurements give values of 1.9 ps (2) and 3.2 ps (16) for the long time constant. Most of trajectories deactivated to the ground state with a time constant of 0.53 ps, whereas a minor fraction deactivated with time constant 3.08 ps (Table 2). These time values are in good agreement with the intermediary and long time constants measured at 250 nm. Although we do not intend to discuss the short time constants here, it is worth noting that 13% of trajectories deactivated within only 0.01 ps, a result that was also observed in multiple spawning simulations (28). The results presented successfully document a series of observations in time-resolved spectroscopy, namely, (i) the general trends of the lifetime (increasing from guanine to thymine) are predicted and related to the different deactivation mechanisms, (ii) the dependence of time constants on the excitation energy (larger lifetimes at low excitation energies) is well reproduced, (iii) the dominance of a single time constant in the decay of the purine bases, and (iv) the more complex deactivation of the pyrimidine bases with multiple time constants. Deactivation mechanisms. Based on the analysis of the dynamics simulations, the most important reaction pathways connecting the Franck–Condon region and the most relevant types of conical intersections are characterized in Fig. 3. Additionally, other energetically feasible alternative reaction paths are indicated. Table 3 summarizes the fraction of trajectories actually following PNAS ∣ December 14, 2010 ∣

vol. 107 ∣

no. 50 ∣

21455

CHEMISTRY

for adenine near the absorption band origin (277 nm) is about 2 ps. It steadily decreases to 0.75 ps when the pump wavelength is shortened to 250 nm (Table 1). We have reanalyzed our previous simulations data for adenine (6) to take into account this dependence on the pump wavelength. Not only the same trend of shorter lifetimes for shorter wavelengths is observed (Table 2), but also good agreement with the experimental values is obtained (Table 1). For all energy windows, all trajectories deactivate with a single time constant indicating, together with the structural analysis discussed below, a single homogeneous mechanism for the internal conversion (Table 2). Guanine deactivation has been measured at 267-nm pump wavelength, which corresponds approximately to the center of the first absorption band. Experimental investigations have reported somewhat different time constants of 0.36 ps (2) and 0.8 ps (16) depending on the fitting procedure. Our simulations have been performed starting the dynamics in the S1 state in a 0.5-eV window around the center of the first UV absorption band. The trajectories deactivated with a time constant of 0.28 ps, supporting the experimental result of ref. 2. As in adenine, the occurrence of a single time constant and the structural analysis indicate the dominance of a single deactivation mechanism. Experiments performed for thymine at 250- and 267-nm pump energy show that this molecule decays with relatively long time constants of a few picoseconds. As first reported in ref. 22, the reason for this behavior is in part due to a very slow S2 → S1 decay. Our computed trajectories, separated into two groups according to their initial energy, establish the lower limit for the excited-state lifetime, >2.3 ps in the high-energy window and >1.7 ps in the low-energy window.


Simulated data are bold in brackets. Time constants in parentheses correspond to monoexponential fitting.

Table 3. Characterization of the deactivation mechanism (%) for the paths shown in Fig. 3 Ade* Yield (%) Time (ps) Gua Yield (%) Time (ps) Thy† Yield (%) Time (ps) Ura† Yield (%) Time (ps) Cyt Yield (%) Time (ps)

A1a 100 (80) 0.63 G1a 100 0.28 P1a 100 2.3 P1a 60 1.5 C1a 87 0.01

A2a 0 (20) —

P1b NA NA P1b NA NA C1b 58 0.5

P1c NA NA P1c NA NA C1c 16 —

P1d NA NA P1d NA NA C1d 12 0.5

P2a 0 — P2a 31 0.65 C2 13 0.01

NA, not available. *Energy window C. (Values for 4-aminopyrimidine are given in parenthesis.) High-energy window.

†

Fig. 3. Schematic description of the dynamics of the five nucleobases based on dynamics simulations results. The curves connecting the nπ* minimum to the ππ* state (path C1c in Cyt and path P1c in Thy/Ura) do not cross the Franck–Condon region as it might appear in this one-dimensional projection of the energy surfaces.

these paths in the dynamics simulation and collects the average times required for passing them. It is important to note that all five bases have a barrierless pathway directly connecting the ππ* state (dashed curve) to the ground state (solid curve) through a carbon atom puckering structure (right side of each graph in Fig. 3). This feature has been described before, and it has been proposed that this pathway constitutes a main mechanism for deactivation of the nucleobases (5, 23, 27). Note that in Fig. 3 the curve labeled ππ* is to be considered as a general representation for this type of state. ππ* states of different character dominate at the left and right sides of each graph. At the right side, the ππ* state has ionic character close to the Franck–Condon region and biradical character close to the intersection (36). The purine bases adenine and guanine. Fig. 3 shows a clear distinction in the complexity of the dynamics between the purine and the pyrimidine nucleobases. The dynamics simulations performed for adenine and the adenine model 4-aminopyrimidine show a predominance of the pathway along the ππ* state with C2 -atom puckering (path A1 in Fig. 3). In the adenine dynamics including only one n orbital in the active space, this pathway is the unique one, whereas the calculations for 4-aminopyrimidine including two n orbitals in the active space show a small but nonnegligible fraction of about 18% of trajectories deactivating through path A2. Note that neither in the A1 path nor in the A2 path the nπ* state plays any role, which means that the dynamics simulations results do not support the assignment of the intermediary time constant of adenine to mechanisms involving the nπ* state as proposed in refs. 14 and 18. For guanine the deactivation mechanism is rather homogeneous, as is revealed by the small variation of geometry structures at the hopping time. This means that guanine is always driven to the conical intersection by directly following the ππ* state along the puckering deformation of the C2 atom (Fig. 3, Path G1). The straightforward deactivation mechanism to the ground state observed in the dynamics simulations for guanine and adenine leads, in agreement with experimental results, to the shortest lifetimes of all nucleobases. Computed values (Table 3) are 0.28 ps for guanine and 0.77 ps for adenine. These relatively short timings indicate a direct mechanism without possibilities for re21456 ∣


laxation or trapping in a local minimum. On the other hand, they are much larger than the initial relaxation times in the order of 100 fs (adenine) and 150 fs (guanine) as obtained from biexponential fitting of pump–probe decay curves, ruling out the possibility of interpreting these decay times as those corresponding to deactivation to the ground state (23). The pyrimidine bases thymine, uracil and cytosine. The most prominent example where the differences between pyrimidine and purine bases can be seen is thymine. It quickly relaxes into a shallow minimum of the second excited state of ππ* character and remains there for a few picoseconds (Fig. 3, path P1). This behavior was first documented by ab initio multiple spawning dynamics (22). This minimum is characterized by interaction of two ππ* states of different character, and the long stay of thymine in S2 shapes its entire dynamics. The direct pathway along the ππ* state analogous to that accessed by the purine bases (Fig. 3, path P2) is never activated in thymine. After relaxation to the first excited state, thymine reaches the nπ* region of this state and moves toward the minimum (Fig. 3, path P1a). [A different result, however, was recently obtained by starting the dynamics at the transition state of S2 rather than in the Franck–Condon region. In this case, thymine moved toward the ππ* region (29).] From the nπ* minimum, thymine may deactivate through the nπ ∕cs conical intersection (path P1b) or cross the barrier with the ππ* state to deactivate at ππ ∕cs conical intersections (paths P1c and P1d). Our simulations limited to 3 ps do not allow a decisive prediction which pathway is the dominant one, although the high energy of the nπ ∕cs conical intersection (see Fig. 2) most certainly indicates dominance of the ππ ∕cs intersection through the P1c pathway. Even though the uracil dynamics is characterized by reaction pathways similar to those observed for thymine, a splitting into three mechanisms is observed here in contrast to thymine. The most accessed mechanism is analogous to thymine, with trapping in the minimum (Fig. 3, path P1) of the second excited state followed by relaxation to the nπ* minimum (path P1a). Starting the dynamics simulations in a high- and in a low-energy window of the absorption spectrum leads to 60% (Table 3) and 38%, respectively, of trajectories following this pathway. This mechanism gives the main contribution to the trajectories decaying with the long time constant τ3 (Table 2). The dynamics simulations limited to 2.5 ps do not allow the prediction whether these trajectories that relax to the nπ* minimum deactivate afterward through the nπ ∕cs crossing (path P1b) or cross the barrier to the ππ* state (path P1c). Barbatti et al.

Barbatti et al.

Outlook The dynamics simulations discussed in the preceding

sections show that there is no unique photodynamical deactivation mechanism for all five bases. Based on the theme of ring puckering, a pool of reaction pathways exists that is significantly narrower for the purine bases as compared to the pyrimidine bases. Trapping in certain regions of the energy surface within a time range up to several picoseconds is observed for the pyrmidine bases. These findings lead to interesting consequences and conclusions beyond the specific range of the photodynamics of singlet states and the five nucleobases discussed so far. Triplet states of nucleobases have attracted strong interest in recent years because of their importance as precursor for photo damage, especially pyrimidine dimerization (37). Because singlet–triplet transitions are much slower and cannot compete with subpicosecond processes, it is crucial to identify those regions of phase space where the dynamics of the singlet state remains for a longer period of time. Investigations on the mechanism for triplet formation and photodimerization are only starting to be elucidated. The study of internal conversion of the pyrimidine bases in aqueous solution (38) postulating the nπ* state to be the doorway to the triplet state fits very well into the present picture showing the possibility of trapping in local energy minima as opposed to the straightforward dynamics of the purine bases. Thus, the absence of a long-lived decay for the purine ribonucleotide adenosine 5′-monophosphate under the same experimental conditions is very well in line with the present analysis. Alternative nucleobases forming unnatural base pairs are gaining increasing attention because of the aim of incorporating them into RNA or DNA. The pair xanthine and 2,6-diaminopyrimidine (2-6-DAPy; for numbering, see adenine in Scheme 1) is such an example (39). Thus, the assessment of the photophysical properties of substituted pyrimidines and purines is of great interest in this context. It is well-known that amino substitution at the C2 position of the purine ring producing 2-aminopurine (2-APu) leads to a large increase in the lifetime (Table 1) as compared to adenine (6-APu). This fact stresses the role of puckering at C2 observed in the simulations. However, what will happen if the imidazole ring is removed while the blocking NH2 substitution is retained? The afore-mentioned 2,6-DAPy is such a case. Will it be long-lived? The answer can be given from previous experience with the dynamics of 6-APy (40) and recent static (41) and photodynamical (42) calculations for 2,6-DAPy, which show that either keeping or not the blocking C2 substitution on the pyrimidine ring, the removal of the imidazole ring will allow fast deactivation by puckering at the C4 position. These last cases should show only some representative examples demonstrating the promising perspectives of the dynamics simulations. Increasing experience with this technique will allow an even better understanding and predicting potential of the complex photophysical and photochemical behavior of nucleobases and related compounds. Methods Nonadiabatic trajectories dynamics simulations were performed for all five nucleobases using the fewest switches surface-hopping method (43). Details about the method are discussed in the SI Text, sections II and III. Energies, energy gradients, and nonadiabatic coupling vectors were computed at ab initio MRCI with singles excitations (MR-CIS) level for adenine and guanine and at the CASSCF level for cytosine, thymine, and uracil after careful selecPNAS ∣ December 14, 2010 ∣

vol. 107 ∣

no. 50 ∣

21457


What Can We Learn from the Dynamics Simulations? The dynamics simulations provide a detailed view of the ultrafast deactivation paths responsible for the photostability of the nucleobases. The discussion presented above shows that the analysis of reaction paths alone is not sufficient for obtaining a comprehensive picture of the actual processes occurring in the photodynamics. The purine bases adenine and guanine represent the most simple photodeactivation mechanism with the dynamics leading along a diabatic ππ* path directly and without barrier to the conical intersection with the ground state. Even though the reaction paths are direct, their real motions are described by coupled puckering modes of the heteroaromatic rings that need certain coordination until sufficiently large out-of-plane distortions are produced to achieve the crossing of the S1 and S0 states. The minimum time for this process is about 280 fs in the present calculations as found for guanine deactivation. Respective timings of 440 to 770 fs (depending on the initial pump energy) are significantly larger for adenine. Note that the nπ* state is not involved in the dynamics of the purine bases. The photodynamics of the pyrimidine bases is much richer and also takes much longer (0.53 to several ps) as compared to the purine bases. According to our analysis, a number of factors are responsible for this situation. One of them is the occurrence of several ππ* states of different character with the π orbital localized in different parts of the molecule. This fact leads to a flatter region on the energy hypersurface of the ππ* state. The extreme case is given by thymine for which a shallow energy minimum in the S2 (ππ*) state is found. This initial flatness of the energy surface prohibits any direct ππ* reaction mechanism as

observed for the purine bases even though a barrierless pathway is available. The remaining dynamics depends on the details of energy barriers and energy surfaces. Accordingly, different detailed behavior of the pyrimidine bases is observed. Because the later stages of the dynamics significantly depend on theses subtleties, it is not possible to claim complete quantitative agreement with experiment. However, the enhanced complexity of the photodynamics of the pyrimidine bases is certainly correctly reproduced in the present simulations.

CHEMISTRY

In the second mechanism, about 31% of uracil trajectories (both for the high- and low-energy windows) follow the ππ* reaction pathway (Fig. 3, path P2). Although most of the structures at hopping time have an out-of-plane distortion of the C6 atom in common so as to make a twist around the C5 C6 bond, they can have very different conformations. This implies that there is no homogeneous mechanism, and uracil has many different ways to proceed to the conical intersection. The remaining trajectories follow a third mechanism that leads to deactivation at a N3 -C4 ring opening conical intersection (not shown in Fig. 3). The trajectories deactivating through these latter two mechanisms mainly contribute to a single intermediary time constant (highenergy window: 0.74 ps; low-energy window: 0.65 ps) even though the structural changes associated with these two mechanisms are quite different. In the case of cytosine, the direct ππ* pathway is not activated at all. Instead, a more complex sequence of events takes place. In 87% of cases (see Table 3) cytosine very quickly relaxes from the ππ* to the nπ* state via the quasiplanar conical intersection (path C1a). From the nπ* minimum, it can deactivate either again through the quasiplanar conical intersection (path C1b), or it can overcome the barrier separating the nπ* and ππ* states (C1c) and switch to the ground state at a ππ ∕cs conical intersection (paths C1d). Both mechanisms were observed in the dynamics simulations. About 58% of trajectories passed through the nπ ∕cs conical intersection (path C1b), whereas another fraction of 16% deactivated through ππ* pathways of type C1d. For comparison, multiple spawning ab initio dynamics based on a very small active space (28) predicted the same relaxation to the S1 minimum, but the escape proceeds in a somewhat different way favoring the C1d path. It is notable that the deactivation times in both kinds of conical intersections (C1b and C1d) are about the same, giving rise to a single intermediary time constant of 0.53 ps. A minor fraction of 13% of trajectories deactivated within only 10 fs by following path C2. The maximum simulation time limited to 1.2 ps did not allow identifying the pathway followed by the remaining fraction of 13% of trajectories with 3.08-ps time constant.

tion of methods aimed at balancing accuracy and computational efficiency as much as possible. As usual, these methods lead to an overshooting of the ionic ππ* state at the Franck–Condon region. Other features of the potential energy surfaces like minima, conical intersections, and reaction pathways are adequately described. This excess of energy in the Franck–Condon region speeds up the dynamics leading to overall shorter time constants than in the experiments, but the correct activation of the several reaction pathways does not seem to be affected by it, as has been discussed in refs. 6 and 30. All simulations were started with the molecules excited in the first ππ* band. The dynamics data for adenine have been reported before (6). They are composed of 60 trajectories. For thymine, the set of 35 trajectories reported in ref. 30 was augmented by a new set of 65 trajectories. For guanine, cytosine, and uracil, respectively, 60, 105, and 89 trajectories were computed. Trajectories were propagated for 0.6 ps for adenine and gua-

nine, 1.2 ps for cytosine and uracil, and 3 ps for thymine. Calculations were performed with the Newton-X program (44) interfaced with the Columbus program (45).

1. Ullrich S, Schultz T, Zgierski MZ, Stolow A (2004) Electronic relaxation dynamics in DNA and RNA bases studied by time-resolved photoelectron spectroscopy. Phys Chem Chem Phys 6:2796–2801. 2. Canuel C, et al. (2005) Excited states dynamics of DNA and RNA bases: Characterization of a stepwise deactivation pathway in the gas phase. J Chem Phys 122:074316. 3. Marian CM (2005) A new pathway for the rapid decay of electronically excited adenine. J Chem Phys 122:104314. 4. Perun S, Sobolewski AL, Domcke W (2005) Ab initio studies on the radiationless decay mechanisms of the lowest excited singlet states of 9 H-adenine. J Am Chem Soc 127:6257–6265. 5. Merchan M, et al. (2006) Unified model for the ultrafast decay of pyrimidine nucleobases. J Phys Chem B 110:26471–26476. 6. Barbatti M, Lischka H (2008) Nonadiabatic deactivation of 9 H-adenine: A comprehensive picture based on mixed quantum-classical dynamics. J Am Chem Soc 130(21): 6831–6839. 7. Lan Z, Fabiano E, Thiel W (2009) Photoinduced nonadiabatic dynamics of pyrimidine nucleobases: On-the-fly surface-hopping study with semiempirical methods. J Phys Chem B 113(11):3548–3555. 8. Nir E, Plutzer C, Kleinermanns K, de Vries M (2002) Properties of isolated DNA bases, base pairs and nucleosides examined by laser spectroscopy. Eur Phys J D 20:317–329. 9. Sobolewski AL, Domcke W, Hättig C (2005) Tautomeric selectivity of the excited-state lifetime of guanine/cytosine base pairs: The role of electron-driven proton-transfer processes. Proc Natl Acad Sci USA 102:17903–17906. 10. Groenhof G, et al. (2007) Ultrafast deactivation of an excited cytosine-guanine base pair in DNA. J Am Chem Soc 129:6812–6819. 11. Crespo-Hernandez CE, Cohen B, Kohler B (2005) Base stacking controls excited-state dynamics in A-T DNA. Nature 436:1141–1144. 12. Santoro F, Barone V, Improta R (2008) Absorption spectrum of A-T DNA unraveled by quantum mechanical calculations in solution on the (dA)(2) (dT)(2) tetramer. ChemPhysChem 9:2531–2537. 13. Lange AW, Herbert JM (2009) Both intra- and interstrand charge-transfer excited states in aqueous B-DNA are present at energies comparable to, or just above, the 1 ππ excitonic bright states. J Am Chem Soc 131:3913–3922. 14. Ullrich S, Schultz T, Zgierski MZ, Stolow A (2004) Direct observation of electronic relaxation dynamics in adenine via time-resolved photoelectron spectroscopy. J Am Chem Soc 126:2262–2263. 15. Kang H, Jung B, Kim SK (2003) Mechanism for ultrafast internal conversion of adenine. J Chem Phys 118:6717–6719. 16. Kang H, Lee KT, Jung B, Ko YJ, Kim SK (2002) Intrinsic lifetimes of the excited state of DNA and RNA bases. J Am Chem Soc 124:12958–12959. 17. Chin CH, et al. (2007) Theoretical investigations of spectroscopy and excited state dynamics of adenine. Chem Phys Lett 445:361–369. 18. Satzger H, et al. (2006) Primary processes underlying the photostability of isolated DNA bases: Adenine. Proc Natl Acad Sci USA 103:10196–10201. 19. Satzger H, Townsend D, Stolow A (2006) Reassignment of the low lying cationic states in gas phase adenine and 9-methyl adenine. Chem Phys Lett 430:144–148. 20. Bisgaard CZ, Satzger H, Ullrich S, Stolow A (2009) Excited-state dynamics of isolated DNA bases: A case study of adenine. ChemPhysChem 10:101–110. 21. Blancafort L, Cohen B, Hare PM, Kohler B, Robb MA (2005) Singlet excited-state dynamics of 5-fluorocytosine and cytosine: An experimental and computational study. J Phys Chem A 109:4431–4436. 22. Hudock HR, et al. (2007) Ab initio molecular dynamics and time-resolved photoelectron spectroscopy of electronically excited uracil and thymine. J Phys Chem A 111:8500–8508. 23. Serrano-Andres L, Merchan M, Borin AC (2008) A three-state model for the photophysics of guanine. J Am Chem Soc 130:2473–2484. 24. Blancafort L (2006) Energetics of cytosine singlet excited-state decay paths—A difficult case for CASSCF and CASPT2. Photochem Photobiol 83:603–610.

25. Perun S, Sobolewski AL, Domcke W (2006) Conical intersections in thymine. J Phys Chem A 110:13238–13244. 26. Matsika S (2004) Radiationless decay of excited states of uracil through conical intersections. J Phys Chem A 108:7584–7590. 27. Serrano-Andrés L, Merchán M, Borin AC (2006) A three-state model for the photophysics of adenine. Chem-Eur J 12:6559–6571. 28. Hudock HR, Martinez TJ (2008) Excited-state dynamics of cytosine reveal multiple intrinsic subpicosecond pathways. ChemPhysChem 9:2486–2490. 29. Asturiol D, Lasorne B, Robb MA, Blancafort Ls (2009) Photophysics of the π,π* and n,π* states of thymine: MS-CASPT2 minimum-energy paths and CASSCF on-the-fly dynamics. J Phys Chem A 113:10211–10218. 30. Szymczak JJ, et al. (2009) Photodynamics simulations of thymine: Relaxation into the first excited singlet state. J Phys Chem A 113:12686–12693. 31. Fabiano E, Thiel W (2008) Nonradiative deexcitation dynamics of 9 H-adenine: An OM2 surface hopping study. J Phys Chem A 112:6859–6863. 32. Matsika S (2005) Three-state conical intersections in nucleic acid bases. J Phys Chem A 109:7538–7545. 33. Zechmann G, Barbatti M (2008) Photophysics and deactivation pathways of thymine. J Phys Chem A 112:8273–8279. 34. Bonačić-Koutecký V, Koutecký J, Michl J (1987) Neutral and charged biradicals, zwitterions, funnels in S1, and proton translocation—Their role in photochemistry, photophysics, and vision. Angew Chem Int Edit 26:170–189. 35. Perun S, Sobolewski AL, Domcke W (2005) Photostability of 9 H-adenine: mechanisms of the radiationless deactivation of the lowest excited singlet states. Chem Phys 313:107–112. 36. Zgierski MZ, Patchkovskii S, Fujiwara T, Lim EC (2005) On the origin of the ultrafast internal conversion of electronically excited pyrimidine bases. J Phys Chem A 109:9384–9387. 37. Kwok WM, Ma C, Phillips DL (2008) A doorway state leads to photostability or triplet photodamage in thymine DNA. J Am Chem Soc 130:5131–5139. 38. Hare PM, Crespo-Hernandez CE, Kohler B (2007) Internal conversion to the electronic ground state occurs via two distinct pathways for pyrimidine bases in aqueous solution. Proc Natl Acad Sci USA 104:435–440. 39. Piccirilli JA, Krauch T, Moroney SE, Benner SA (1990) Enzymatic incorporation of a new base pair into DNA and RNA extends the genetic alphabet. Nature 343:33–37. 40. Barbatti M, Lischka H (2007) Can the nonadiabatic photodynamics of aminopyrimidine be a model for the ultrafast deactivation of adenine? J Phys Chem A 111:2852–2858. 41. Gengeliczki Z, et al. (2010) Effect of substituents on the excited-state dynamics of the modified DNA bases 2,4-diaminopyrimidine and 2,6-diaminopurine. Phys Chem Chem Phys 12:5375–5388. 42. Nachtigallova D, Barbatti M, Szymczak JJ, Hobza P, Lischka H (2010) The photodynamics of 2,4-diaminopyrimidine in comparison with 4-aminopyrimidine: The effect of amino-substitution. Chem Phys Lett 497:129–134. 43. Tully JC (1990) Molecular-dynamics with electronic-transitions. J Chem Phys 93:1061–1071. 44. Barbatti M, et al. (2007) The on-the-fly surface-hopping program system Newton-X: Application to ab initio simulation of the nonadiabatic photodynamics of benchmark systems. J Photoch Photobio A 190:228–240. 45. Lischka H, et al. (2001) High-level multireference methods in the quantum-chemistry program system COLUMBUS: Analytic MR-CISD and MR-AQCC gradients and MR-AQCC-LRT for excited states, GUGA spin-orbit CI and parallel CI density. Phys Chem Chem Phys 3:664–673. 46. He Y, Wu C, Kong W (2003) Decay pathways of thymine and methyl-substituted uracil and thymine in the gas phase. J Phys Chem A 107:5145–5148. 47. Nir E, Müller M, Grace LI, de Vries MS (2002) REMPI spectroscopy of cytosine. Chem Phys Lett 355:59–64.

21458 ∣


ACKNOWLEDGMENTS. The authors thank Vienna Scientific Cluster for technical support and computer time (Projects 70019 and 70151). This work has been supported by the Austrian Science Fund [Special Research Programs F16 (Advanced Light Sources) and F41 (ViCoM), and Project P18411-N19]. Support by the grant from the Ministry of Education of the Czech Republic (Center for Biomolecules and Complex Molecular Systems, LC512) and by the Praemium Academiae of the Academy of Sciences of the Czech Republic (to P.H. in 2007) is gratefully acknowledged. This work was part of the research project Z40550506 of the Institute of Organic Chemistry and Biochemistry of the Academy of Sciences of the Czech Republic.

Barbatti et al.