Photoinduced charge separation and DNA self-repair depend on sequence directionality and stacking pattern

Charge separation is one of the most common consequences of the absorption of UV light by DNA. Recently, it has been shown that this process can enable efficient self-repair of cyclobutane pyrimidine dimers (CPDs) in specific short DNA oligomers such as the GAT 
<svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="13.200000pt" height="16.000000pt" viewBox="0 0 13.200000 16.000000" preserveAspectRatio="xMidYMid meet"><metadata>
Created by potrace 1.16, written by Peter Selinger 2001-2019
</metadata><g transform="translate(1.000000,15.000000) scale(0.017500,-0.017500)" fill="currentColor" stroke="none"><path d="M0 440 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z M0 280 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z"/></g></svg>
 T sequence. The mechanism was characterized as sequential electron transfer through the nucleobase stack which is controlled by the redox potentials of nucleobases and their sequence. Here, we demonstrate that the inverse sequence TTAG promotes self-repair with higher quantum yields (0.58 ± 0.23%) than GATT (0.44 ± 0.18%) in a comparative study involving UV-irradiation experiments. After extended exposure to UV irradiation, a photostationary equilibrium between self-repair and damage formation is reached at 33 ± 13% for GATT and at 40 ± 16% for TTAG, which corresponds to the maximum total yield of self-repair. Molecular dynamics and quantum mechanics/molecular mechanics (QM/MM) simulations allowed us to assign this disparity to better stacking overlap between the G and A bases, which lowers the energies of the key A−˙G+˙ charge transfer state in the dominant conformers of the TTAG tetramer. These conformational differences also hinder alternative photorelaxation pathways of the TTAG tetranucleotide, which otherwise compete with the sequential electron transfer mechanism responsible for CPD self-repair. Overall, we demonstrate that photoinduced electron transfer is strongly dependent on conformation and the availability of alternative photodeactivation mechanisms. This knowledge can be used in the identification and prediction of canonical and modified DNA sequences exhibiting efficient electron transfer. It also further contributes to our understanding of DNA self-repair and its potential role in the photochemical selection of the most photostable sequences on the early Earth.


Introduction
Photoinduced charge separation is one of the main phenomena occurring in nucleic acids during the exposure to ultraviolet (UV) light. 1 Despite its far reaching implications for biochemistry, biology and even materials sciences, [2][3][4][5] the key experimental proof of photoinduced charge separation in native unmodied DNA was delivered less than 20 years ago.This experiment involving transient absorption spectroscopy demonstrated that UV excitation of polydeoxyadenosine could populate long-lived electronic states, which were assigned to charge transfer (CT) between the neighbouring stacked adenine nucleobases. 6While numerous examples of CT states have been corroborated in different modied and native forms of DNA since then, 1,[7][8][9][10][11][12][13][14] the key factors controlling the efficiency of this process still remain obscure.Consequently, prediction of DNA sequences capable of performing efficient UV-induced charge transfer is still a challenge.
Among different processes that can be triggered by charge separation in DNA, self-repair of cyclobutane pyrimidine dimers (CPDs) attracted substantial attention, recently. 15,168][19][20][21] Formation of this cyclobutane ring affects the structure of the sugar-phosphate backbone and precludes biochemical activity such as DNA replication and transcription. 21,229][30] The most prominent examples of DNA self-repair were demonstrated for the damaged GAT]T sequence ("]" representing the CPD) and for 2,6-diaminopurine (D) and 8-oxoguanine (O) nucleobases located in the vicinity of CPDs. 31,324][35] In other words, the yields of nonenzymatic DNA self-repair are a manifestation of how efficiently photoinduced charge separation can occur in specic DNA sequences and whether lifetimes of the CT states are sufficiently long to invoke a photochemical reaction.
It is worth emphasizing that highly efficient self-repair of CPDs greatly improves the photostability of specic sequences and was also suggested as a possible selection factor for primordial RNA and DNA oligomers from the rich pool of random sequences. 1,15,36,379][40][41][42][43][44][45][46] This resulted in the consideration of the above mentioned D and O nucleobases as potential components of rst informational polymers owing to their improved electron-donating and CPD-repairing properties when compared to canonical nucleobases. 31,32,47In particular, DNA trinucleotides containing a D nucleobase and a T]T dimer were shown to repair the CPD with yields reaching up to 92% when irradiated at 280 nm and, thus, D could protect DNA oligomers from photodamage under prebiotic conditions. 31s shown by Bucher et al. 1 the direction of charge transfer in stacked DNA is controlled by the oxidation potentials of the nucleobases.However, the efficiency with which excited CT states are populated seems to depend on much more subtle aspects related to the local environment of the sequence of stacked bases. 2 This could be either governed by the spatial overlap of the neighbouring stacked bases as well as by the lifetimes of the different electronic states involved in photoinduced charge separation. 31,48More specically, as suggested for DT]T and T]TD trimers, a possible factor which could enable funneling the excited-state population to a CT state is the inacessibility of efficient direct photorelaxation channels of the locally excited (LE) state on D via an S 1 /S 0 conical intersection. 31ere, we present a surprising example of the damaged T]TAG oligomer (5 ′ -end denoted rst), which exhibits self-repair quantum yields which are higher by ∼30% than in the case of the equivalent GAT]T oligomer with opposite sequence direction when exposed to UV irradiation at 285 nm.Since this difference cannot be explained by the ordering and oxidation potentials of nucleobases, we provide a mechanistic rationale for this process based on molecular dynamics simulations, quantum mechanics/molecular mechanics (QM/MM) calculations and explorations of excited-state potential energy surfaces using the algebraic diagrammatic construction to the second order method [ADC (2)]. 49,50We show that in contrast to GAT]T, direct photorelaxation of the LE and CT intermediate states of the photoexcited T]TAG tetramer is substantially hindered, which enhances sequential electron transfer between stacked bases and enables efficient and selective CPD self-repair.

Results and discussion
Experimental observation of higher self-repair quantum yields in T]TAG when compared to GAT]T In our previous works, we demonstrated the self-repair of the short single-stranded DNA sequence, GAT]T, via photoexcitation at 290 nm. 33Aer absorption of a UV photon, the GAT]T molecule sequentially populates minima on the potential energy surface of the lowest excited electronic state (S 1 minima), which can be associated with the G + cA − cT]T and G + cAT]T − c charge transfer states (Fig. 1 le). 14,35These states can be accessed right aer the photoexcitation of the G base, which is responsible for the majority of the absorption of the oligomer at wavelengths longer than 280 nm. 14,30Here, we performed comparative UV irradiation experiments of the DNA oligonucleotides T] TAG and GAT]T (Fig. 1).Exposure of the sequences to longwavelength irradiation from an LED, centered around 285 nm (average power at the sample position 0.36 mW), allowed us to excite the Guanine (predominantly) and Adenine (partially) bases and minimized photoreversal from direct absorption of the T]T dimer (Fig. 2A and S7 †). 51,52As presented in Fig. 2B, continuous UV-irradiation of an aqueous T]TAG solution (pH 6.9, ∼30 mM) in our system lead to an absorbance increase of several 10 mOD at 266 nm within 10 minutes.The irradiation induced difference spectra clearly correspond to typical absorption pattern exhibited by thymine bases in DNA, which is strongly indicative of the self-repair process (Fig. 2A).Complementary HPLC analysis corroborated this nding (Fig. 2C).The chromatogram of the damaged sequence T]TAG as starting material (bottom) shows a peak at 8.4 min.Upon exposure to 285 nm light, the undamaged sequence TTAG gradually recovers, visible as an increase in absorption at 9.9 min.Aer 30 min of irradiation, the irradiated sample was spiked with the undamaged sequence TTAG for reference (top chromatogram).The emergence of one single peak at 9.9 min conrms the self-repair to TTAG.The corresponding chromatograms for the sequence GAT]T are shown in Fig. S1 in the ESI.† The increase in absorbance at 266 nm due to the recovery of undamaged TTAG from T]TAG (black arrow in Fig. 1B) can be plotted as a function of the photon dose absorbed by the molecules (Fig. 3).At low irradiation doses, the increase in absorbance at 266 nm is linear.The data in this range can be tted with a linear trendline (red).The slope of the initial absorbance increase is linearly proportional to the quantum yield of the CPD self-repair (see ESI † for details).The quantum yield of CPD self-repair was found to be 0.44 ± 0.18% for GAT] T (blue) and 0.58 ± 0.23% for T]TAG (black), respectively.At higher absorbed doses, the slope of both plots decreases.This is largely the result of approaching the photostationary state of equilibrium, but can be also attributed to the formation of secondary products, which lower the yield of the net reaction.Aer absorption of high irradiation doses, a photostationary equilibrium between net damage formation and self-repair can be reached (Fig. 4).In case of the sequence GAT]T, the equilibrium is reached aer the absorption of 3.5 J at a level of 33 ± 13% self-repair.This result is higher than the previously reported 25% at 290 nm irradiation as well as the corresponding quantum yields. 33,53The differences can be attributed to the 15 nm broad LED light source, in comparison to the previously used narrowband (3 nm broad) 290 nm excitation, as self-repair quantum yields and direct photoreversal may be higher at lower wavelengths.In case of T]TAG, the equilibrium is reached only aer absorption of 2.5 J at a higher level of 40 ± 16%.The error bars were estimated to 40% of the provided values according to the previous work by some of us were an analogous experimental setup was used. 54These results indicate that the ratio of the self-repair vs. the net rate of CPD formation is higher in case of T]TAG when compared to GAT]T.The photostationary equilibrium indicates the maximum total yield of the repair in UV-rich environments, under persistent exposure to irradiation.

Structural differences between the T]TAG and GAT]T tetramers
We further performed classical molecular dynamics and quantum chemical simulations, to provide mechanistic rationale explaining this substantial difference in self-repairing activity between the T]TAG and GAT]T tetramers.Since we previously performed such calculations for the GAT]T tetranucleotide, here, we applied an analogous computational protocol to the T]TAG tetramer. 35The overall lengths of the trajectories of our force-eld based molecular dynamics simulations for T]TAG amounted to 10 ms per tetranucleotide.We selected two of the most representative stacked conformers, which have the highest contribution to the overall conformational space of T]TAG (see the ESI † for more details).The AGanti conformer of T]TAG discussed in the main article is shown in Fig. 5.We used their averaged MD structures for subsequent QM/MM simulations in order to rst optimize the ground-state geometry of the tetramer using density functional theory with dispersion correction (PBEh-3c functional) and next calculate their photophysical and photochemical properties.
Notably, as shown in Fig. 5, the highest populated stacked conformers of T]TAG and GAT]T differ by the relative spatial orientation of the nucleobases and the stacking patterns.The positioning of the G and A bases at the 3 ′ -end of the T]TAG tetramer results in the two nucleobases predominantly populating the anti orientations with respect to the sugar scaffold.This entails a very good overlap of the six-membered counterparts of the purine rings and partial overlap of the vemembered subunits for most of the populated conformers.While the T]TAG conformer presented in Fig. 5 was populated in 9% of the simulation time, we also observed this AG-anti arrangement for conformers with partially unstacked or overhanging T]T dimer.Overall, the AG-anti arrangement was present in >50% of sampled conformations of T]TAG (see Fig. S4 in the ESI † to this article).In contrast, the G base of the GAT]T tetramer prefers to form a hydrogen bond between its N3 atom and the free 5 ′ -OH group of the sugar, which is accompanied by the syn orientation of the nucleobase with respect to the sugar ring. 35As observed previously, the neighbouring adenine of GAT]T also prefers the syn orientation to maintain better stacking and the resulting GA-syn stacked conformer was determined to be dominant (populated in over >30% of conformations). 35It is worth noting though that the G and A bases of this conformer are somewhat displaced with respect to one another and less favorably stacked, having only the veand six-membered counterparts of G and A bases stacked.As previously indicated for long-range electron transport in DNA, the degree of stacking and conformational arrangement could strongly affect the rate and yield of CT along the stack. 2 Therefore, these structural differences are the rst indication that disparities between the photochemistry of the two tetranucleotides should be expected.

Photophysical properties of T]TAG
According to the vertical excitation energies computed for the T]TAG tetramer (see the ADC(2)/MM results Table 1), the charge transfer states can be found in the higher energy range of the UV absorption spectrum, whereas the lower energy range of the spectrum is dominated by local pp* excitations of the purine bases and np* excitations of the T bases constituting the T]T dimer.The lowest-energy CT state in the Frank-Condon region involves an electron transferred from G to A and can be identied as the S 6 state with the p G p * A molecular orbital character.The analogous p G p * A CT state of the GA-syn stacked conformer of GAT]T has the vertical excitation energy higher by 0.2 eV and was identied as the S 10 state.This demonstrates that the stacking overlap between the G and A bases can strongly affect the ordering of the electronic states and the energies of CT excitations.
We previously identied another CT excitation in the Franck-Condon region of the GAT]T tetramer, namely the p A p * TT CT state associated with an electron transferred between the A base and the T]T dimer.In fact, this excitation is the lowest energy CT state found for the ground-state geometry of the GA-anti conformer of GAT]T, and was identied as the S 10 state with the excitation energy of 5.81 eV.In the case of the GAsyn conformer of the GAT]T tetranucleotide, the p A p * TT CT state was found to be the S 7 state with the vertical excitation energy of 5.32 eV.Nevertheless, we did not nd this electronic state among the ten lowest vertical excitations of the T]TAG tetranucleotide.Furthermore, low self-repair yields of canonical trinucleotides containing CPDs was ascribed to the limited accessibility of the p A p * TT CT state. 30,31Therefore, as proposed for the GAT]T tetranucleotide, we postulate that the initial photoinduced charge separation event in T]TAG is initiated with the population of the p G p * A state outside of the Franck-Condon region.The details of this mechanism are discussed in the following section.

Sequential electron transfer in the T]TAG tetranucleotide
According to the calculations of vertical excitation energies, irradiation of the T]TAG tetranucleotide at 285 nm primarily results in the population of the S 1 (p G p * G ) state of the G base.We argue that subsequent electron transfer events will occur on the hypersurface of the lowest excited singlet state involving changes of molecular orbital character until the key S 1 /S 0 state crossing is reached.Such a mechanism leading to partial splitting of the cyclobutane ring is presented in Fig. 6 and involves three intermediate S 1 minima.Therefore, the presented self-repair mechanism of T]TAG is analogous to the sequential electron transfer (SET) process described for the GAT]T tetranucleotide. 35However, we identied the key differences in the SET mechanism for the GAT]T and T]TAG tetranucleotides which could further explain the higher selfrepair yields found for T]TAG.
Initial vibrational relaxation of the p G p * G state results in modest puckering of the aromatic ring of guanine and reaching the vicinity of the rst S 1 minimum that participates in the SET mechanism.This minimum is denoted as G* in Fig. 6 and lies 4.27 eV above the ground-state structure of the AG-anti conformer of T]TAG.Similarly as in the case of GAT]T, ring puckering is most pronounced at the C4 and C5 atoms and greater out of plane distortion of the C4 and N3 atoms and rotation about the C4]N3 bond leads to the S 1 (p G p * G )/S 0 minimum-energy crossing point (MECP) lying ∼0.75 eV above the G* S 1 minimum (see Fig. 6 and 7).Considerably sloped topography of this state crossing and the associated energy Table 1 Vertical excitations energies (in eV) of the T]TAG tetranucleotides obtained at the ADC(2)/TZVP level of theory, assuming the ground-state minimum-energy geometry optimized with the PBEh-3c method.The results for the GAT]T tetranucleotide were taken from ref. 33  barrier hinder the direct photorelaxation of the G base.We anticipate that this will increase the importance of the competitive forward electron transfer process from the G base to the neighbouring A base.It is worth noting that the analogous S 1 (p G p * G )/S 0 MECP of the of GAT]T tetranucleotide was reported to lie only 0.15 eV and 0.50 eV above the S 1 (G*) minima for the GA-syn and GA-anti conformers, respectively.This indicates that direct photorelaxation of the G base should be more efficient in the GAT]T tetranucleotide, whereas UV-excited T] TAG tetramer should more easily undergo charge separation.
Subsequent excited-state electron transfer from the G base may allow to reach the vicinity of the A − cG + c S 1 minimum.This entails structural changes in the ring-puckering pattern of the G base, with most pronounced pyramidalization of the C2 atom and additional pyramidalization of the C6 atom of A. Consequently, the two pyramidalized C atoms create the main contact between these purine bases in the A − cG + c CT minimum.This S 1 minimum lies merely 0.12 eV below the G* minimum of T] TAG AG-anti conformer, which implies a weaker driving force for e − transfer than in the case of the GAT]T tetramer (DE = −0.9eV for the GA-anti conformer).While driving force is an important component of the electron rate within the Marcus model, the efficiency of photoinduced CT between stacked nucleobases is also dependent on the excited-state lifetime of the donor state.Given that, the direct photorelaxation of the G base is hindered in major conformers of T]TAG, we expect the G to A electron transfer process to be an important contributor to the photochemistry of this tetranucleotide.
Similarly as in the case of the p G p * G state, the direct photorelaxation from the A − cG + c minimum is hindered owing to very high energy of the S 1 (p G p * A )/S 0 state crossing, which lies 1.18 eV above the corresponding S 1 minimum.The energy of this state crossing is even higher than the vertical excitation energy of lowest optically bright state of the G base.This state crossing involves formation of a (transient) covalent bond between the C2 atom of G and C6 atom of A. Consequently, the second electron transfer event from the radical anion of the A base to the T]T dimer should be the preferred event occurring aer the population of the p G p * A of the T]TAG tetranucleotide.In contrast, direct photorelaxation of GAT]T from its p G p * A state is again more efficient than for T]TAG, since the analogous S 1 (p G p * A )/S 0 state crossing lies 0.42 eV above and 0.25 eV below the G + cA − c minimum located for the GA-syn and GA-anti conformers, respectively.
The above interpretation is further supported by the recent investigation of the photodynamics of the GA and AG dinucleotides with transient absorption spectroscopy, which showed that the yield of excited-state interbase charge transfer is higher by ∼75% for GA than for AG.Similarly the GA dinucleotide was reported to exhibit a longer lifetime of the CT state (170 ± 10 ps) than the AG dinucleotide (112 ± 12 ps). 55While in the case of these dinucleotides the trend is clearly opposite than for the damaged GAT]T and T]TAG tetranucleotides, dinucleotides are characterized by very different conformational spaces than longer oligomers owing to both purine bases being located at the termini of the mini-strands.Consequently, the results of Petropoulos et al. 55 are consistent with the picture emerging from our QM/MM calculations, that is, that the population of DNA excited CT states is strongly affected by the interbase stacking pattern.
The second excited-state e − transfer allows the tetranucleotide to reach the T]T − cAG + c electronic conguration that enables direct photoreversal of the CPD.This CT event is associated with much stronger driving force for electron transfer as the T]T − cAG + c S 1 minimum lies 1.22 eV below the A − cG + c minimum.Similarly as in the case of the GAT]T tetranucleotide as well as damaged trimers containing 2,6-diaminopurine, a very modest barrier (<0.1 eV) separates the nal S 1 CT minimum of T]TAG from the S 1 (p G p * TT )/S 0 state crossing responsible for CPD repair (see the PES presented in Fig. S10 in the ESI †).Beyond this barrier, the C5-C5 bond breaking process can occur spontaneously and the S 1 (p G p * TT )/S 0 conical intersection has a peaked topography (see Fig. S10 in the ESI †).This geometry of the T]T dimer in this MECP is characterized by the C5/C5 distance (between constituent T bases) equal to 2.55 Å, which means that this covalent bond of the CPD is completely broken at the point of conical intersection.The remaining C6-C6 bond of the CPD maintains its length of 1.56 Å at the S 1 /S 0 state crossing, but can be subsequently broken in the hot electronic ground state, which completes the CPD self-repair process. 35,56,57In other words, CPD reversal is a stepwise process and as reported previously for enzymatic repair of thymine dimers, the barrier associated with C6-C6 bond breaking does not exceed 0.15 eV (3.1 kcal mol −1 ). 56,57e emphasize, that reaching this S 1 (p G p * TT )/S 0 state crossing does not ensure CPD self-repair, since the C5-C5 bond may still be reformed aer non-radiative transition to the S 0 state.Therefore, the experimentally measured self-repair quantum yield cannot be directly associated with the quantum yield of the SET process in T]TAG.However, the peaked topography of this state crossing indicates that C5-C5 bond rupture involves relatively high momentum, which should generally drive the C5 atoms of the T bases towards greater separation, which can be subsequently followed by C6-C6 bond breaking. 31,35

Conclusions
In summary, we demonstrated that the directionality of DNA sequences can substantially inuence the efficiency of photoinduced electron transfer through the base stack.We present this based on the example of photoinduced self-repair of cyclobutane pyrimidine dimers in the GAT]T and T]TAG sequences, which is controlled by the sequential electron transfer mechanism. 35In particular, the T]TAG sequence is characterized by higher quantum yields of CPD self-repair when compared to the GAT]T tetramer (0.58% vs. 0.44%).Overall, the T]TAG tetranucleotide can repair up to 40% of the formed photodimers before reaching photostationary equilibrium when irradiated at 285 nm.In comparison, we managed to achieve only up to 33% of self-repair for the GAT]T sequences under equivalent conditions.
We ascribe this phenomenon to the differences in the conformational ensembles between the two tetranucleotides.Firstly, for the major conformers of the T]TAG tetramer (AGanti), we observe much higher degree of stacking between the G and A bases than in the case of the GAT]T tetranucleotide.Better stacking overlap lowers the energy of p G p * A CT state in the Franck-Condon region.This electronic CT state is responsible for the rst electron transfer event in the CPD self-repair process.Secondly, the S 1 /S 0 minimum-energy crossing points (MECPs) responsible for the direct photorelaxation of the intermediate p G p * G LE and p G p * A CT states are practically energetically inaccessible in the T]TAG tetramer.As a result, excited-state electron transfer process is a much more competitive and favorable photorelaxation mechanism for T] TAG than in the case of the GAT]T tetranucleotide.Even though many more photorelaxation mechanisms are usually available in DNA strands, involving e.g.locally-excited np* or ps* states, our calculations demonstrate that the well-stacked conformation of the T]TAG tetramer can effectively restrain some of these channels and promote interbase electron transfer.
We anticipate that these alternative photorelaxation mechanisms are the main reason for the modest self-repair quantum yields resulting from our measurements.Here, we were able to identify two other photorelaxation pathways of the intermediate p G p * G LE and p G p * A CT electronic states with our static excitedstate PES explorations.However, in a dynamic picture higher electronic states can easily interchange their order with the S 1 state and lead to other S 1 /S 0 state crossings, which are more challenging to grasp with the static QM/MM approach.These state crossings can potentially enable back-electron transfer which was observed for polyadenosine sequences. 58Furthermore, as discussed above, reaching the key S 1 (p G p * TT )/S 0 state crossing which entails C5-C5 bond breaking may still be followed by CPD reformation.Nevertheless, the quantum yields are sufficiently high to enable the accumulation of high quantities of repaired material during continuous irradiation (∼33% for GAT]T and ∼40% for T]TAG).
Overall, we show that photoinduced electron transfer in DNA and the associated CPD self-repair process are strongly dependent on conformation and the availability of alternative (direct) photorelaxation channels of the intermediate states.This demonstrates that the efficiency of electron transfer cannot be simply predicted based on sequence.However, prior computational exploration of the conformational spaces of DNA sequences and their associated photochemical properties (including the energetics of CT states and S 1 /S 0 state crossings) can offer valuable predictive capacity for the identication of DNA oligomers that can undergo efficient charge separation upon UV absorption.

Fig. 1 (
Fig. 1 (A) Molecular structures of the nucleobases in the DNA sequence T]TAG.(B) Schematic representation of the CPD self-repair mechanism promoted by photoinduced sequential electron transfer in the DNA sequences, GAT]T and T]TAG.

Fig. 2 (
Fig. 2 (A) Molar decadic absorption coefficients of the T]T CPD lesion (red) and the undamaged TT dinucleotide (black).The emission spectrum of the LED centered around 285 nm, which was used for irradiation, is shown as blue shade for comparison.(B) UV absorption difference spectra of a 30 mM solution of the sequence T]TAG after increasing times (1-10 min) of exposure to 285 nm irradiation with an average power of 0.36 mW.The recovery of the 266 nm absorption (arrow) is indicative of the self-repair to TTAG.(C) Analytical HPLC analysis of the sequence T]TAG (8.4 min) after different irradiation times.Upon irradiation a recovery of the undamaged sequence TTAG (9.9 min) can be observed.The chromatogram on top is spiked with undamaged TTAG for reference.

Fig. 3
Fig.3Absorbance difference spectra of T]TAG (black) and GAT]T (blue) at 266 nm as a function of absorbed light dose at 285 nm.The data at low doses can be fitted with a linear trend line (red).The slope of the trend line is linearly proportional to the quantum yield of the self-repair.

Fig. 4
Fig.4Concentration of the undamaged sequence GATT (top) and TTAG (bottom) divided by the initial concentration of the damaged sequence GAT]T (top) and T]TAG (bottom) as starting material as a function of absorbed photon dose at 285 nm.In case of the GAT]T sequence a photostationary equilibrium between repair and damage formation is reached after absorption of 3.5 J at a level of ∼33%.The sequence T]TAG reaches the equilibrium earlier after absorption of 2.5 J at a level ∼40%.The absorption coefficients at 266 nm were taken from Pan et al.30

Fig. 5
Fig. 5 Averaged structures of the major stacked conformers obtained from the classical MD simulations of T]TAG (left) and GAT]T (right) 35 tetranucleotides.The stacking of the G and A bases is presented at the bottom.

Fig. 6
Fig. 6 Energy diagram showing the sequential electron transfer mechanism initiated with the photoexcitation of the guanine base of the T]TAG tetranucleotide to the S 1 (p G p * G ) state in the Franck-Condon (FC) region (left-hand side of the diagram).The leftmost structure corresponds to the ground-state geometry of the AG-anti conformer of the T]TAG tetramer.The three middle energy levels are associated with the key S 1 minima.The partly transparent pathways leading to S 1 /S 0 MECPs demonstrate the possible competing direct photorelaxation mechanism from the G* and A − cG + c intermediate states.The rightmost structure is associated with the key S 1 /S 0 MECP responsible for C5-C5 bond breaking.The energies were obtained with the QM bases /MM setup at the ADC(2)/def2-SVP level of theory (see the Computational Methods section in the ESI † for more details).

Fig. 7
Fig. 7 Geometries of the G and A bases located for the intermediate excited-state (S 1 ) G* and A − cG + c minima (top) and the geometries of the corresponding S 1 /S 0 MECPs.The presented geometries and energies were obtained with the QM bases /MM setup at the ADC(2)/ def2-SVP level of theory (see the Computational Methods section in the ESI † for more details).
and obtained with the same approach