Efficient UV-induced charge separation and recombination in an 8-oxoguanine-containing dinucleotide

Significance

UV photons are absorbed strongly by DNA, but rarely cause permanent photodamage. Single nucleobases are protected by ultrafast nonradiative decay, but excited states in single- and double-stranded DNA decay very differently. An intensely debated question is whether a UV photon can move an electron from one nucleobase to another along a single strand. This study demonstrates that UV absorption efficiently transfers an electron from an oxidatively damaged guanine (8-oxo-G) to adenine in a dinucleotide mimic of the flavin cofactor FADH₂, yielding radicals that decay in 60 ps. It is proposed that the photoredox activity of 8-oxo-G, which may have repaired cyclobutane pyrimidine dimers in the RNA world, reflects the importance of ultrafast charge separation between stacked nucleobases by UV radiation.

Abstract

During the early evolution of life, 8-oxo-7,8-dihydro-2′-deoxyguanosine (O) may have functioned as a proto-flavin capable of repairing cyclobutane pyrimidine dimers in DNA or RNA by photoinduced electron transfer using longer wavelength UVB radiation. To investigate the ability of O to act as an excited-state electron donor, a dinucleotide mimic of the FADH₂ cofactor containing O at the 5′-end and 2′-deoxyadenosine at the 3′-end was studied by femtosecond transient absorption spectroscopy in aqueous solution. Following excitation with a UV pulse, a broadband mid-IR pulse probed vibrational modes of ground-state and electronically excited molecules in the double-bond stretching region. Global analysis of time- and frequency-resolved transient absorption data coupled with ab initio quantum mechanical calculations reveal vibrational marker bands of nucleobase radical ions formed by electron transfer from O to 2′-deoxyadenosine. The quantum yield of charge separation is 0.4 at 265 nm, but decreases to 0.1 at 295 nm. Charge recombination occurs in 60 ps before the O radical cation can lose a deuteron to water. Kinetic and thermodynamic considerations strongly suggest that all nucleobases can undergo ultrafast charge separation when π-stacked in DNA or RNA. Interbase charge transfer is proposed to be a major decay pathway for UV excited states of nucleic acids of great importance for photostability as well as photoredox activity.

The RNA world hypothesis suggests that ancient life originated from RNA-based oligomers due to their ability to both store genetic information and catalyze reactions in a manner similar to protein-based enzymes (1). In proteins, the 20 canonical amino acids are not versatile enough in their redox activity for many purposes, and special redox cofactors, such as the dinucleotides FADH₂ and NAD(P)H, are often recruited to facilitate a desired transformation. Similarly, facile oxidation or reduction reactions involving the canonical nucleobases could put the integrity of the genome at risk. Recent work has shown that 8-oxo-7,8-dihydroguanine (8-oxo-G), an oxidatively damaged form of guanine (G), is a redox-active base capable of photoinduced reversal of thymine dimers in DNA oligomers (2, 3). In particular, continuous irradiation of substrates containing a thymine dimer and a nearby 8-oxo-G using a UVB lamp decreases the amount of dimers over time (2, 3), but direct evidence of photoinduced electron transfer (ET) has been lacking.

The ability of 8-oxo-G to act as an excited-state electron donor is plausible, but by no means assured in light of conflicting indications. On the one hand, 8-oxo-G is easier to oxidize in its electronic ground state by ∼30 kJ⋅mol^–1 compared with G (4), the most easily oxidized of the canonical bases, but this advantage may be lost for excited-state oxidation due to the lower energy of the first excited singlet state of 8-oxo-G. On the other hand, the 2′-deoxynucleoside of 8-oxo-G (8-oxo-dGuo, or O) has an excited-state lifetime of just 0.9 ± 0.1 ps at physiological pH (5)—a value that is similar to the subpicosecond lifetimes of the undamaged bases (6). Rapid nonradiative decay could thus frustrate ET despite favorable thermodynamics. Time-resolved spectroscopy can resolve this puzzle by detecting any short-lived radicals produced by photoinduced ET. Here, we use femtosecond time-resolved infrared (TRIR) spectroscopy to definitively show that UV excitation of the dinucleotide d(OA) (Fig. 1) transfers an electron from O to 2′-deoxyadenosine (A) on a subpicosecond timescale to form a contact radical ion pair (exciplex) that can be unambiguously identified by comparison with density functional theory (DFT) calculations.

Fig. 1.

UV-visible (A) and FTIR spectra (B) for d(OA) at neutral pH. The spectra of monomeric 8-oxo-dGuo (red dashed curves) and AMP (green dotted curves) are shown for comparison. The excitation wavelengths used in the pump-probe experiments are indicated in A by arrows.

The dinucleotide d(OA) was chosen as a crude mimic of FADH₂ in which 8-oxo-G replaces the dihydroflavin moiety of the cofactor that serves as the electron source for photoinitiated ET to a cyclobutane pyrimidine dimer (CPD) (7). Combining O with A offers the further advantage that the steady-state UV-visible and IR absorption spectra have several nonoverlapping transitions that arise from just one of the two chromophores (Fig. 1 and Fig. S1). These spectral characteristics permit selective excitation of O and selective detection of the localization site of an excited state via mid-IR probing of either O or A vibrations. A vibrational spectrum with its comparatively narrow resonances is frequently easier to assign than overlapping electronic absorption spectra. Consequently, TRIR spectroscopy can often differentiate between charge transfer (CT) states and other excited states that may have similar electronic absorption spectra (8, 9).

Besides its interest as a mimic of a redox cofactor, the d(OA) dinucleotide also provides valuable insights into the role of CT states in DNA. As shown below, the CT state of d(OA) decays in ∼60 ps by charge recombination (CR). This decay is much longer than the excited-state lifetimes of either A or O as monomers. Similarly long-lived excited states are formed in high yields whenever DNA bases are stacked with one another both in single- and double-stranded forms (10–14). The identity of these long-lived states has been one of the most debated issues in the photophysics and photochemistry of nucleic acids during the past decade. The results from this study suggest that a primary decay channel for excited states of stacked nucleobases in DNA, whether modified or not, is ultrafast interbase ET.

Results

Experimental Results.

TRIR spectra of d(OA) following excitation at 265 nm are shown in Fig. 2A. All measurements were carried out in D₂O solution at pH 7.3 because deuteration improves the IR transparency of water in the double-bond stretching region of interest. The difference spectra consist of negative signals due to ground-state bleaching (GSB), and positive signals that arise from vibrational modes of transient species. All signals disappear completely by 300 ps after the pump pulse, leaving only a featureless negative offset, which originates from altered absorption by D₂O at the elevated temperature produced by the pump pulse (15).

Fig. 2.

TRIR spectra at the indicated pump-probe delay times from a 5-mM solution of the d(OA) dinucleotide (A and C) and 5 mM 8-oxo-dGuo (O) + 5 mM AMP (A) mixture (B and D) following 265-nm (A and B) and 295-nm (C and D) excitation. Red arrows point to positive signals assigned to vibrational marker bands of 8-oxo-dGuo^•+ (see text). The inverted and scaled steady-state FTIR spectrum for each sample is shown by the dot-dashed gray line. Vibrational mode assignments are included for convenience.

Immediately after 265-nm excitation, three strong bleaches are observed at the frequencies of the strongest peaks in the steady-state FTIR absorption spectrum of d(OA). Each band is labeled in Fig. 2 by its dominant vibrational character as determined by normal mode calculations. The relative amplitudes of the negative ΔA signals are in excellent agreement with the inverted FTIR spectrum, which is shown in Fig. 2A by the dot-dashed line. Strong positive bands in the transient absorption difference spectra are seen at 1,602 and 1,684 cm⁻¹ (marked with red arrows). Both positive and negative signals decay biexponentially and can be globally fit with the same time constants (4 ± 1 ps and 60 ± 20 ps; Table 1). TRIR spectra for an equimolar mixture of 8-oxo-dGuo and AMP excited at 265 nm (Fig. 2B) display a strikingly different pattern of initial bleach amplitudes compared with when O and A are covalently linked in the d(OA) dinucleotide (Fig. 2A). In particular, the ratios of the bleach signal of the A band at 1,623 cm⁻¹ to the O bands at 1,562 and 1,662 cm⁻¹ are four times greater in the monomer mixture than in d(OA). Also, only short 2- to 3-ps decays are observed for the O and A monomers (Fig. S2 A and B), indicating that the 60-ps time constant is unique to the dinucleotide.

Table 1.

Global fit parameters for the d(OA) GSB signals

λpump	A1	τ1, ps	A2	τ2, ps
265 nm*	0.6 ± 0.1	4 ± 1	0.4 ± 0.1	60 ± 20
295 nm†	0.9 ± 0.4	3 ± 1	0.10 ± 0.05	60 ± 30

Normalized traces were fit to $A_1{e}^{-t/{\tau }_1}+A_2{e}^{-t/{\tau }_2}$ after subtracting the long-time signal due to solvent heating.

^*Fit to bleach signals for O (ν_C6=O) at 1,662 cm⁻¹, O (ν_amino) at 1,562 cm⁻¹, and A (ν_rings) at 1,623 cm⁻¹ (Fig. S2 A and B).

^†Fit to bleach signals for O (ν_C6=O) at 1,662 cm⁻¹ and O (ν_amino) at 1,562 cm⁻¹ (Fig. S2 C and D).

TRIR spectra recorded with excitation at 295 nm, a wavelength that is expected to excite O and not A (Fig. 1), are shown for d(OA) and the equimolar mixture in Fig. 2 C and D, respectively. A slow time constant (60 ± 30 ps) is again observed for d(OA) that is absent in the monomer mixture (Fig. S2 C and D), but now the prominent positive features seen at 1,602 and 1,684 cm⁻¹ following excitation at 265 nm are attenuated, and the relative amplitude of this component is four times weaker (10 ± 5% vs. 40 ± 10%) for 295- vs. 265-nm excitation. Comparison of Fig. 2 A and C clearly shows that the pattern of bleached bands seen at the earliest delay times differs dramatically for 265- vs. 295-nm excitation. The 1,623-cm⁻¹ ground-state mode of A, which is strongly bleached upon excitation at 265 nm, is bleached only weakly in the TRIR spectra at 295 nm (blue arrow in Fig. 2C). There is no bleaching of this band when the equimolar mixture of O and A is excited at 295 nm (Fig. 2D), confirming that this wavelength selectively excites O.

Computational Results.

Vibrational spectra for ground-state O and A were computed for the 2′-deoxynucleoside pentahydrates (8-oxo-dGuo·5D₂O and dAdo·5D₂O, structures shown in Fig. S1 J and K) using the harmonic approximation at the polarizable continuum model (PCM)/PBE0/6–31+G(d,p) level of theory (Fig. 3A). Assignments for 8-oxo-dGuo vibrations are based on the dominant character and match ones in ref. 16. Inclusion of 2′-deoxyribose and explicit D₂O molecules dramatically improves the agreement with experiment for the ring in-plane and C=O stretching vibrations, respectively (17). Absolute integrated absorption coefficients from the calculations were scaled by a constant factor that gave the best agreement with the experimental cross-sections for the O and A fundamentals. The experimental and theoretical frequencies are compared in Table 2.

Fig. 3.

Calculated vibrational spectra for monomeric species by PCM/PBE0/6–31+G(d,p) and the harmonic approximation. (A) Ground-state 8-oxo-dGuo·5D₂O (O) and dAdo·5D₂O (A) spectra; (B) ground-state 8-oxo-dGuo^•+·5D₂O (O^•+) and dAdo^•−·5D₂O (A^•−) spectra; (C) difference spectrum obtained from the above monomeric species. Molecular structures and mode assignments are shown for convenience (R, 2′-deoxyribose; explicit D₂O not shown). All transition frequencies are broadened by a Gaussian function with 20 cm⁻¹ FWHM.

Table 2.

Experimental and calculated [PCM/PBE0/6–31+G(d,p)] vibrational frequencies in inverse centimeters for the ground states of 8-oxo-dGuo·5D₂O (O), 8-oxo-dGuo^•+·5D₂O (O^•+), dAdo·5D₂O (A), and dAdo^•−·5D₂O (A^•−)

Species	${\mathrm{\nu }}_{\mathrm{amino}}$*	${\mathrm{\nu }}_{\mathrm{C}4\mathrm{C}5}$†	${\mathrm{\nu }}_{\mathrm{rings}}$‡	${\mathrm{\nu }}_{\mathrm{C}6\mathrm{=O}}$§	${\mathrm{\nu }}_{\mathrm{C}8\mathrm{=O}}$¶
O (expt.)	1,562	1,601	―	1,662	1,704
O (calc.)	1,566	1,609	―	1,655	1,717
O•+ (expt.)	1,602||	―	―	1,684||	―
O•+ (calc.)	1,602	1,473	1,649	1,677	1,763
A (expt.)	―	―	1,623	―	―
A (calc.)	―	―	1,625	―	―
A•− (expt.)	―	―	―	―	―
A•− (calc.)	―	―	1,565	―	―

calc., calculated; expt., experimental.

^*Vibrations localized on the C₂–ND₂ group.

^†Pyrimidine ring in-plane vibration with dominant C₄ = C₅ stretching.

^‡Ring in-plane vibration.

^§C₆ = O stretch.

^¶C₈ = O stretch.

^||Band maximum from the TRIR spectrum.

PCM/time-dependent (TD)-M052X calculations on the entire d(OA) dinucleotide show that a state with well-defined CT character lies close in energy to the locally excited states of O and A (Table S1). Geometry optimization of this state leads to a minimum with strong CT character and a small interbase separation (see computational details in SI Text). The vibrational spectra of O^•+ (8-oxo-dGuo^•+·5D₂O) and A^•− (dAdo^•−·5D₂O), two possible species created by UV excitation, are shown in Fig. 3B. Vibrational modes of the radical cation O^•+ are generally blue-shifted compared with its unoxidized form, whereas the single vibrational band of A^•− in our window of interest is red-shifted (Table 2). Fig. 3C displays the difference spectrum obtained by subtracting the spectra in Fig. 3 A and B. Fig. S3 displays similar difference spectra calculated for the excited state of the entire dinucleotide.

Kinetic Modeling.

An excellent global fit to the time- and frequency-dependent TRIR spectra for d(OA) (Fig. 2) was obtained using two exponential time constants and a long time offset to account for heating of the solvent. Target analysis (18) was used to identify a kinetic model that best accounts for the observed signals. A model with two parallel decay channels gave the most reasonable predictions (Figs. S4 and S5). In this model, initial excited states of d(OA) created by the femtosecond pump pulse branch faster than our experimental time resolution to two states that decay monoexponentially with the lifetimes determined by global fitting.

The difference spectrum of the long-lived state [i.e., its species-associated difference spectrum (SADS)] (18) determined by global analysis of the TRIR spectra recorded with a pump wavelength of 265 nm is shown by the dashed curve in Fig. 4A. The SADS derived from the d(OA) TRIR spectra excited at 295 nm is identical in appearance (dotted curve in Fig. 4A). The excellent match of the two SADS indicates that the same transient species is present at the longest delay times for 265- and 295-nm excitation. Furthermore, this common spectrum is in excellent agreement with the calculated difference spectrum of the O and A radical ions shown in Fig. 3C, which has been reproduced in Fig. 4A. The SADS of the short-lived state along with other details are presented in Figs. S4–S6.

Fig. 4.

Comparison of experimental and calculated vibrational difference spectra of d(OA). (A) Long-lived SADS for 265-nm (dashed) and 295-nm (dotted, shown after scaling) excitation obtained from global fitting compared with the theoretical difference spectrum from Fig. 3C (solid) calculated at the PCM/PBE0/6–31+G(d,p) level with explicit D₂O molecules; (B) the difference spectra calculated with 8-oxo-dGuo(–D_N7)^• (gray) at the same level of theory; (C) comparison of the experimental difference spectrum of 8-oxo-dGuo^•+ (dot-dashed) obtained by two-photon ionization of 8-oxo-dGuo and the theoretical spectrum (solid).

Discussion

The remarkable agreement between the experimental difference spectrum of the 60-ps state and the calculated difference spectrum obtained from the redox pairs O^•+/O and A/A^•− (Fig. 4A) indicates that UV excitation of d(OA) transfers an electron from O to A, mimicking the first step of photolyase-catalyzed repair of CPDs. Ultrafast spectroscopic measurements on the photolyase system suggest that ET from the excited state of the isoalloxazine moiety of FADH⁻ to the 5′ thymine residue of the dimer occurs via the adenine base situated between the isoalloxazine and the thymine (7). In the present studies, interbase charge separation is supported by a calculation performed on the entire dinucleotide, which predicts that the CT state is the global minimum among all excited states, and that O^•+ and A^•− interact weakly (see computational details in SI Text). The latter observation validates the computational approach in which the radical ion pair is treated as two separate monomers.

The excellent agreement between the ground-state FTIR spectrum of d(OA) (Fig. 1B) and the calculated spectrum (Fig. 3A) validates the level of theory and the inclusion of explicit D₂O molecules and the ribose group. To confirm that this computational approach is suitable for modeling vibrational spectra of radical ions, the vibrational difference spectrum of O^•+ was recorded by two-photon ionization of an aqueous solution of the O monomer (dot-dashed curve in Fig. 4C). The calculated difference spectrum (solid curve in Fig. 4C) is in excellent agreement with the experiment. The positive features in the experimental spectrum at 1,602 and 1,684 cm⁻¹ are assigned to ν_amino and ν_C6=O modes of O^•+, respectively. These modes have a similar spacing, but are red-shifted by ∼20 cm⁻¹ compared with the vibrational marker bands of the guanine radical cation in D₂O at ∼1,620 and ∼1,700 cm⁻¹ (9, 19).

The kinetic steps in the parallel model used to analyze the TRIR data are assigned in Fig. 5. According to our model, the CT state is populated faster than the experimental time resolution (<0.5 ps) in competition with the appearance of vibrationally hot nucleobases produced by ultrafast internal conversion. The latter species undergo vibrational cooling on a timescale of several picoseconds. For excitation at 265 nm, the pattern of bleached bands at early times is consistent with the creation of an exciton that is delocalized over both O and A (see discussion of short-time dynamics in SI Text for more details). The rapid appearance of the CT state is supported by the absence of any signal growth at frequencies where O^•+ and A^•− absorb. Once populated, the CT state decays monoexponentially with a lifetime of ∼60 ps. Because there is no spectral signature of a further state, the CT state is proposed to decay to the ground state of d(OA) by CR.

Fig. 5.

Kinetic schemes for excited-state decay of d(OA) in aqueous solution. The relative population of each decay channel from the fitting analysis is also indicated (Table 1). The initial steps in gray occur faster than the instrument response time.

ET from O to A is considerably more efficient for 265- vs. 295-nm excitation as seen from the relative amplitudes of the long-lived signal in the kinetic traces in Fig. S2 A and C. The normalized GSB signals at several different frequencies are identical (Fig. S2 B and D), indicating that the bleach signals are free of excited-state and/or hot ground-state absorption. In this case, the amplitude of the slow decay component (A₂ in Table 1) is an estimate of the fraction of all excited states that decay to the CT state. Our calculations indicate that the shorter wavelength populates predominantly the bright, locally excited L_a state of A, which couples strongly with the CT state (see computational details in SI Text), which may explain the fourfold larger CT yield upon 265-nm excitation. Strong coupling between the L_a state of A and the CT state also suggests that the CT state is populated extremely rapidly.

Nucleobase radical ions have enhanced acid/base properties compared with the parent compounds (20), but proton transfer occurs too slowly to compete with CR in d(OA). Notably, the experimental difference spectrum (Fig. 4A) disagrees with the spectrum calculated assuming that O is present as a neutral radical (Fig. 4B, gray curve). Reported deprotonation rates for G^•+ at neutral pH range from 3.5 × 10⁵ s⁻¹ (21) to 1.8 × 10⁷ s⁻¹ (22). O^•+ (pK_a = 6.6) is a weaker acid than G^•+ (pK_a = 3.9) (23), suggesting that it would lose a proton at a still slower rate. A^•− is protonated by a water molecule at a rate of 1.4 × 10⁸ s⁻¹ (20). These reactions are too slow to compete with CR in 60 ps. Rapid CR and the prevention of follow-on reactions that would permanently alter O support the finding that O in an oligodeoxynucleotide is able to catalyze the reversal of a thymine dimer for at least five turnover cycles with no detectable degradation of the 8-oxo-G base (2).

On the Generality of Photoinduced CT in Nucleobase Stacks.

The main finding of this study is that absorption of a UVB or UVC photon by the d(OA) dinucleotide transfers an electron from O to A, producing a radical ion pair; this supports the role of O as a flavin mimic, but this study also adds to evidence that photoinduced ET occurs generally and with relatively high quantum yield between any pair of stacked nucleobases (11, 24–26). The latter conclusion rests on strong thermodynamic and kinetic similarities between CT states in other dinucleotides and in d(OA).

Although O is easily oxidized, the Gibbs energy change for photoinduced ET ($\mathrm{\Delta }{G}_{\text{ET}}$) depends not only on the oxidation half-reaction, but also on the reduction potential of the acceptor, the energy of the excited state, and the electrostatic work of bringing free ions to the short separation at which ET takes place (27). The work term appropriate for charge separation favors forward ET. However, even if this term is neglected, as is commonly done (28), combining the large singlet-state energies of the canonical DNA bases (29) with experimental reduction potentials (30) predicts that photoinduced ET is thermodynamically allowed for five of the six possible heterodimers made from A, G, C, and T. In fact, the use of the relaxed singlet excited state energy likely underestimates the energy available for an ET reaction that proceeds on an ultrafast timescale before the thermalization of excess vibronic energy (31). If the available excited-state energy is equal to the energy of the 265-nm photon (4.68 eV), then photoinduced ET is predicted to be exergonic for all dinucleotides made from A, G, C, and T, including the four homodimers, with $\mathrm{\Delta }{G}_{\text{ET}}$ ranging between −0.27 and −1.05 eV (Table S2). Importantly, $\mathrm{\Delta }{G}_{\text{ET}}$ for ET from O to A induced by a 295-nm photon (−0.58 eV; Table S2) falls in the middle of this range and is less favorable than ET from G to A, C, or T upon 265-nm excitation. These considerations suggest that photoinduced ET between pairs of the canonical DNA bases is thermodynamically every bit as likely as between O and the others.

In addition to these thermodynamic similarities, there are strong kinetic parallels between the excited-state dynamics of d(OA) and model compounds made of natural bases. The 60-ps lifetime of the radical ion pair state of d(OA) is in the range of lifetimes previously reported for the long-lived excited states detected in DNA/RNA dinucleotides (11, 24–26) and oligonucleotides (10, 26, 32). Earlier, Takaya et al. (11) proposed that the long-lived excited states are exciplex states with substantial CT character that decay via CR on timescales of 10–100 ps, depending on the thermodynamic driving force. These long-lived excited states are only observed when π−π stacking brings two or more bases into van der Waals contact (10, 11, 32). This minimal separation enables ET to proceed on a femtosecond timescale (33), allowing charge separation to compete with the subpicosecond dynamics responsible for the ultrashort excited-state lifetimes of single bases.

In further support of the similarities between d(OA) and compounds containing natural bases, we note that the CD spectrum of d(OA) (Fig. S1A) is nearly identical to that of d(GA) (34), indicating that both have similar electronic structure and distributions of conformers. Dinucleotides are relatively flexible and adopt a range of conformations in aqueous solution, which may differ from the prevalent ones in double-stranded DNA. However, the dynamics of the long-lived exciplex states have been shown to be identical for different initial stacking conformations (14). Finally, although d(OA) is single-stranded, long-lived excited states have been detected in duplex DNAs made from all four bases (12, 35). The similar dynamics observed in double- and in single-stranded forms (10, 13) suggests that base pairing perturbs, but does not fundamentally alter, the charge separation that we propose occurs generally between π-stacked nucleobases.

The striking excitation wavelength dependence of the CT state yield seen in our experiments suggests that the energetics and coupling between the excitonic and CT states at the FC region play an important role in the excited-state dynamics. Similar to d(OA), calculations on canonical base stacks predict that CT states lie in close proximity to the bright excitonic states in the FC region (36, 37). Furthermore, dynamical calculations predict that even a small amount of vibronic coupling between these states can populate the CT state, independent of their relative energy (38). The higher quantum yield observed at shorter excitation wavelength is also consistent with subpicosecond charge separation that occurs before vibronic relaxation.

The creation of a radical ion pair by UV light in d(OA) or in a general DNA sequence is reminiscent of the creation of a bound polaron pair in polymer blends (39, 40), establishing an unexpected link between DNA photophysics and photoprocesses in organic solar cells. The factors that determine whether such an ion pair can escape to form free charges may also govern whether nucleobase radicals live long enough to initiate damaging photoreactions. The high quantum yields of charge separation noted here for d(OA), and in other systems with modified bases (26), greatly exceed the highest DNA photoproduct yields, suggesting that CR, which occurs in d(OA) with a time constant of 60 ps, reverses transient damage before follow-on reactions can take place, including proton transfer with the solvent. However, reversal of a CPD following one-electron reduction occurs in less than 100 ps (7), suggesting that photoreactivation may be possible despite the generally short lifetimes of exciplexes or radical ion pairs in DNA.

We suggest that the efficient population of the CT state and its subsequent decay by CR result from the intimate positioning of acceptors next to donors by π−π stacking of the bases in a DNA strand; this may explain how a molecular architecture that produces high yields of CT states can nevertheless exhibit high photostability. Notably, this same paradigm raises questions for future investigations about whether the facile ET from O to a neighboring base on the same strand can repair a CPD that is located several bases away from O or even on the opposite strand. These studies will reveal whether the abundant radicals formed in nucleic acids by UV radiation can contribute to photoreactivation without increasing the risk of photodamage to the genome.

Conclusions

UV excitation of d(OA) creates a radical ion pair with a lifetime of 60 ps that is unambiguously assigned to interbase CT through the vibrational signature of this state in the double-bond stretching region. The observation of ET from O to a π-stacked neighbor, which is observed here, to our knowledge for the first time, supports the hypothesis that O can repair CPDs in a photolyase-like manner (2, 3). Strikingly, the quantum efficiency of reaching the CT state is four times higher for excitation with UVC than with UVB photons. This observation is important for developing models to explain how CT states, which generally have low oscillator strengths, are reached from the initial excitons. These results and ones from recent studies (24–26) suggest that interbase ET is a primary event whenever nucleic acids absorb UV light.

Methods

The UV-pump/broadband-mid-IR-probe experimental setup has been described in detail elsewhere (41). Broadband mid-IR probe pulses centered at 6,150 nm were generated with an optical parametric amplifier (OPA) (TOPAS-C+NDFG; Light Conversion). The probe beam was split into “signal” and “reference” portions to reduce pulse-to-pulse noise. Both probe beams passed through the sample and were detected by a dual-row 64-element HgCdTe array (Infrared Systems Development).

The 265- and 295-nm pump pulses were generated by a second OPA. The pump fluence at the sample was attenuated to between 1.7 and 2.0 mJ⋅cm⁻². The relative polarization of the pump and probe pulses was set to the magic angle (54.7°). Kinetics traces were fit to exponential functions in Igor Pro software, and the reported uncertainties are 2σ. Global fitting of the TRIR spectra was done using programs described in ref. 42.

A 5.0-mM d(OA) solution was prepared in 50 mM phosphate buffer and 100 mM NaCl in D₂O. D₂O solutions of 3′-AMP (Sigma-Aldrich) and 8-oxo-dGuo (Berry and Associates, Inc.) were prepared with the same buffer and salt concentrations. The synthesis (43) and characterization of d(OA) are described in SI Text and Figs. S7–S9. The sample solution under study (total volume of 2.0 mL) was recirculated through a flow cell (Harrick Scientific Products, Inc.) with a 100-μm path length.

Geometry optimizations and vibrational frequencies of monomeric molecules and radicals were calculated using DFT with the PBE0 functional (44) and the 6–31+G(d,p) basis set. All labile hydrogen atoms were substituted by deuterium atoms. The 2′-deoxyribose group and five explicit D₂O molecules were also included (see Dataset S1 for the Cartesian coordinates). A polarizable continuum model was used to model the bulk solvent effect (45). For calculations on the entire dinucleoside monophosphate, the ground-state geometry and vibrational frequencies were obtained using both PBE0 and M052X (46) functionals. The excited-state geometry and vibrational frequencies were obtained by their time-dependent extensions (47). Due to the cost of the computation, a smaller 6–31G(d) basis set was used, and explicit D₂O molecules were not included. All calculated vibrational frequencies were multiplied by 0.972.