Electron-driven proton transfer relieves excited-state antiaromaticity in photoexcited DNA base pairs†

The Watson–Crick A·T and G·C base pairs are not only electronically complementary, but also photochemically complementary. Upon UV irradiation, DNA base pairs undergo efficient excited-state deactivation through electron driven proton transfer (EDPT), also known as proton-coupled electron transfer (PCET), at a rate too fast for other reactions to take place. Why this process occurs so efficiently is typically reasoned based on the oxidation and reduction potentials of the bases in their electronic ground states. Here, we show that the occurrence of EDPT can be traced to a reversal in the aromatic/antiaromatic character of the base upon photoexcitation. The Watson–Crick A·T and G·C base pairs are aromatic in the ground state, but the purines become highly antiaromatic and reactive in the first 1ππ* state, and transferring an electron and a proton to the pyrimidine relieves this excited-state antiaromaticity. Even though proton transfer proceeds along the coordinate of breaking a N–H σ-bond, the chromophore is the π-system of the base, and EDPT is driven by the strive to alleviate antiaromaticity in the π-system of the photoexcited base. The presence and absence of alternative excited-state EDPT routes in base pairs also can be explained by sudden changes in their aromatic and antiaromatic character upon photoexcitation.


References for
Based on the equilibrium geometries of base pairs, dissected nucleus-independent chemical shifts, NICS(1) zz , were computed for each base, by splitting the optimized structures of base pairs into individual bases and considering each fragment separately (with the purported charge and electronic state). NICS(1) zz values were computed at 1 Å above the five and six membered ring centers and include only contributions from the "out-of-plane" (zz) tensor component perpendicular to the ring plane. All NICS(1) zz data were computed at CASSCF(10,10)/6-311+G(d,p) using an active space of ten orbitals (five occupied and five unoccupied π-orbitals, with ten electrons), employing the Dalton2016 program. Choice of active space for NICS(1) zz calculations was selected by comparing computed carbon shieldings at the CASSCF/6-311+G(d,p) level with different active space sizes (Tables S2, S3, S4 and S5). Computed NICS(1) zz at the PW91/IGLOIII level for separate bases in base pair structures were benchmarked against results for the corresponding base pairs and showed excellent agreement (Table S1). Multicenter indices (MCI) values were computed for base pairs structures at CASSCF(8,8)/6-311+G(d,p) using an active space of eight orbitals (four occupied π orbitals and four unoccupied π-orbitals, with eight electrons) employing ESI-3D.

Computed carbon shieldings at CASSCF/6-311+G(d,p), with different active spaces.
Based on ground state equilibrium structures of the WC A•T and G•C base pairs, carbon shieldings for the base fragments were computed to identify an appropriate active space for NICS calculations at the CASSCF level. Base fragments were considered so that a larger active space (up to 12,12) could be examined with the 6-311+G(d,p) basis set. Computed carbon shielding values converged at an active space size of 10 π-orbitals (five occupied and five unoccupied π-orbitals) and 10 electrons (10,10).    The base pair 1 H chemical shifts were calculated using the following equation.
is the calculated shielding of the base pairs, is the calculated shielding for benzene, and is the experimental H NMR shift (7.3 ppm) of benzene.   to the GS states). Note also changes in the pyrimidine fragments; increased bond length alternation in the LE state, and bond length equalization in the CT state.

Multicenter indices (MCI) values
Multicenter indices (MCI) values were calculated for the C s geometries of base pairs, in the GS, LE, and CT states, at the CASSCF/6-311+G(d,p) level with an active space of eight π-orbitals (four occupied and four unoccupied π-orbitals, with eight electrons). MCI values were computed for the complete base pair structures (instead of base fragments), and show a loss of aromaticity upon excitation, with gain in aromaticity upon proton-coupled electron transfer. The MCI quantifies the extent of delocalized cyclic bonding-larger values reflect a more aromatic ring. For example, the MCI for benzene is 0.0435 at the CASSCF(6,6)/6-311++G(d,p) level.

Computed MCI values for Watson-Crick A•T and G•C structures at C 1 symmetry
Since minima geometries of the LE and CT states of WC A•T and WC G•C are nonplanar, we also considered the effects of aromaticity gain and loss based on the C 1 structures. Minima geometries, energies, and MCI values of the GS, LE, and CT states were computed at the (TD-)B97XD/6-311+G(d,p) level and the trends agree with that observed for the planar models. Upon excitation to the LE state (A*•T and G*•C), both base pairs show a loss of aromatic character in the six membered ring of the purine.
Crossing to the CT state ([AT]* and [GC]*) regains a large part of the aromatic character ( Figure S4).
Note the near planar geometry of the purines in base pairs at the CT state, indicating re-aromatization ( Figure S5).

Alternative proton-coupled electron transfer pathway in the Watson-Crick G•C base pair.
We also considered an alternative PCET pathway in the WC G•C pair, involving excitation on the purine, and proton transfer from the exocyclic NH 2 group. Compared to transfer of the pyridinium proton, the resulting CT structure is higher in energy (3.0 eV relative to ground state G•C, compared to 2.6 eV for the pyridinium proton transfer), and computed NICS (1)

Homolytic N-H -bond cleavage of adenine and guanine.
Computed energy profile for the homolytic N-H -bond cleavage of adenine and guanine at CASPT2(10,10)/6-311+G(d,p), in the ground state (GS) and 1 ππ* excited state (LE). Note that, for both adenine and guanine, the GS and LE curves for -bond cleavage parallel each other and are highly endothermic. We show in Figure S10 that the power to photo-oxidize a nucleobase (as evaluated by the triplet state IPs) and to reduce a nucleobase (as evaluated by the ground state EAs), can be related to, respectively, the excited-state antiaromatic character and the ground state aromatic character of the nucleobase considered.
As shown in Figure S10

DNA duplex model study
Excited-state proton-coupled electron transfer in a GC:CG DNA duplex was considered, following the deactivation pathway proposed by Kohler and coworkers in J. Am. Chem. Soc. 137, 7059-7062 (2015) (see Figure S7). The studied GC:CG duplex structure was modified from a PDB file (1BNA) and computed using an ONIOM scheme. All four nucleobase structures were optimized at ωB97XD/6-31+G(d). The sugar phosphate backbone (see Figure S11, in grey) was optimized using a semiempirical method, PM6, with fixed atomic positions for all heavy atoms. Relative energies and multicenter index (MCI) analyses were computed at ωB97XD/6-31+G(d) for the resulting ONIOM optimized geometries. Figure S12. Optimized structure of guanine 1 in the LE state and CT state. Note that 1 is non-planar in the LE state, but becomes planar upon electron transfer to cytosine 3.
Upon local excitation on guanine 1 (LE state), the energy of the CT and PCET states lower as an electron is transferred from guanine 1 to cytosine 3 through base stacking, and as a proton is transferred from guanine 4 to cytosine 3 (see relative energies in Table S10). In the LE state, guanine 1 (green color in LE state) loses aromaticity (note lower MCI value, Table S11) and planarity ( Figure S12). Transferring an electron to cytosine 3 (red color in CT state) regains some aromatic character restores planarity (see Table   S11 and Figure S11). Cytosine 3 loses a lesser degree of aromaticity) upon accepting an electron (cf. MCI value for 1 in the LE state.   Figure 3A).  Figure 4A).  Figure 2B).  Figure 3B).  Figure 4B).  Figure S6).