Internal conversion to the electronic ground stateoccurs via two distinct pathways for pyrimidinebases in aqueous solutionPatrick M. Hare, Carlos E. Crespo-Hernandez, and Bern Kohler†

Department of Chemistry, Ohio State University, 100 West 18th Avenue, Columbus, OH 43210

Edited by F. Fleming Crim, University of Wisconsin, Madison, WI, and approved October 30, 2006 (received for review September 21, 2006)

The femtosecond transient absorption technique was used tostudy the relaxation of excited electronic states created by absorp-tion of 267-nm light in all of the naturally occurring pyrimidine DNAand RNA bases in aqueous solution. The results reveal a surprisingbifurcation of the initial excited-state population in <1 ps to twononradiative decay channels within the manifold of singlet states.The first is the subpicosecond internal conversion channel firstcharacterized in 2000. The second channel involves passagethrough a dark intermediate state assigned to a lowest-energy1n�* state. Approximately 10–50% of all photoexcited pyrimidinebases decay via the 1n�* state, which has a lifetime of 10–150 ps.Three- to 6-fold-longer lifetimes are seen for pyrimidine nucleo-tides and nucleosides than for the corresponding free bases,revealing an unprecedented effect of ribosyl substitution on elec-tronic energy relaxation. A small fraction of the 1n�* population isproposed to undergo intersystem crossing to the lowest tripletstate in competition with vibrational cooling, explaining the highertriplet yields observed for pyrimidine versus purine bases at roomtemperature. Some simple correlations exist between yields of the1n�* state and yields of some pyrimidine photoproducts, but morework is needed before the photochemical consequences of thisstate can be definitively determined. These findings lead to adramatically different picture of electronic energy relaxation insingle pyrimidine bases with important ramifications for under-standing DNA photostability.

conical intersection � DNA photophysics � excited-state dynamics

There has been interest for many years in excited electronic statesof DNA and RNA bases because of their role in the UV

photodamage of nucleic acids (1). These excited states sometimesdecay to deleterious photoproducts, which can cause mutations andinterfere with the normal cellular processing of DNA. It has beenargued that damage is greatly inhibited by their intrinsic photo-physical properties. In particular, the ultrafast decay of excitedelectronic states to the electronic ground state (S0) by internalconversion (IC) has been proposed to account for the high pho-tostability of single bases (1) and base pairs (2). In essence,electronic energy is eliminated by rapid nonradiative decay fasterthan the time required for a photochemical reaction.

Microscopic understanding of IC begins with a description of thepathway through nuclear coordinate space traversed by a photo-excited molecule as it makes its way from the initial excited stateback to S0 (Fig. 1A). Such a pathway is akin to the ‘‘reactioncoordinate’’ describing the evolution of reactants to products in a(photo)chemical reaction (Fig. 1B). An important difference be-tween photochemical and photophysical nonradiative decay con-cerns the initial and final points of the reaction coordinate: In aphotochemical reaction, these are distinct points (Fig. 1B) corre-sponding to the different nuclear geometries of reactants andproducts. In photophysical nonradiative decay to S0 the decaypathway ends where it began (Fig. 1A). Photon absorption createsan excited molecule that initially has the same geometry as in S0, inaccordance with the Franck–Condon principle. The decay pathwayends when the molecule has reached thermal equilibrium once

again near the minimum of S0. In this process the original photonenergy has been degraded into excess vibrational energy of thesurrounding solvent molecules. By avoiding undesirable photo-chemical destinations, this road to nowhere is responsible for thehigh photostability of the molecular building blocks of DNA.

Why is nuclear motion necessary if the final geometry is the sameas the starting one? Simply stated, nonradiative decay occurs atmuch greater rates than radiative decay when nuclear motions takethe molecule to regions of the nuclear configuration space wherepotential-energy surfaces intersect. In polyatomic molecules thesurfaces meet at conical intersections (CIs) (3, 4). At a CI, couplingbetween electronic and nuclear degrees of freedom enables non-adiabatic transitions from one surface to another, frequently withina single vibrational period (4). Once of uncertain importance, CIsare now routinely implicated whenever ultrafast photoprocessesincluding rapid IC are observed (5, 6).

Determination of the nuclear motions responsible for nonra-diative decay presently requires concerted experimental andtheoretical efforts. Conventional femtosecond pump–probetechniques excel at measuring lifetimes and rates but cannotdetermine molecular structures. Time-resolved electron diffrac-tion is a promising all-in-one method for studying nonradiativedecay (7) but is not yet in widespread use. Presently, femtosec-ond measurements like the ones reported here are combinedwith quantum chemical calculations to assemble a comprehen-sive picture of excited-state dynamics.

A growing number of computational studies describe CIsthought to be responsible for nonradiative decay by excited nucleo-bases (8–23). The nuclear coordinate space that must be exploredis potentially vast: intramolecular dynamics in uracil (Ura), thenucleobase with the smallest number of atoms, take place in a30-dimensional space. Furthermore, a large number of CIs havebeen located between the several low-energy excited states of thebases (21, 23). Not surprisingly, these issues and the inherentchallenge of calculating excited-state potential-energy surfaces formolecules of this size have led to some disagreement about theactual decay pathways. A point of contention is whether the maindecay channel proceeds via a 1n�* state. Some authors have arguedfor an intermediate 1n�* state (2, 12, 24–29), whereas others haveproposed that nonradiative decay occurs directly from the 1��*state to S0 (10, 15, 21, 22). Finally, a growing number of compu-tational studies (see, for example, ref. 21) have identified both kinds

Author contributions: B.K. designed research; P.M.H. and C.E.C.-H. performed research;P.M.H. and C.E.C.-H. analyzed data; and P.M.H., C.E.C.-H., and B.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission.

Abbreviations: Ura, uracil; Thy, thymine; Cyt, cytosine; DMU, 1,3-dimethyluracil; dTMP,thymidine 5�-monophosphate; 1CHU, 1-cyclohexyluracil; ISC, intersystem crossing; IC, in-ternal conversion; CI, conical intersection; S0, electronic ground state.

†To whom correspondence should be addressed. E-mail: kohler@chemistry.ohio-state.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0608055104/DC1.

© 2006 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0608055104 PNAS � January 9, 2007 � vol. 104 � no. 2 � 435–440

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0

of channels, but there is no experimental evidence that IC actuallyoccurs via more than one channel in solution. Here we presentfemtosecond transient absorption results showing decay via twodistinct channels for the hydrated pyrimidine bases Ura, thymine(Thy), and cytosine (Cyt). The first channel is direct IC from the1��* state to S0. The second channel, which accounts for 10–50%of all deexcitation events, involves decay via a comparativelylong-lived intermediate electronic state, which we assign to thelowest-energy 1n�* state. These results reveal a more complexpicture for pyrimidine base photophysics and DNA photostability.

ResultsFig. 2 compares transient absorption signals from several 5� mono-nucleotides in aqueous solution at 250 and 340 nm after excitationat 267 nm. These probe wavelengths provide a representative viewof the dynamics of the various electronic states, which absorbbroadly in this spectral region. At 250 nm, both pump and probewavelengths fall within the strong, lowest-energy �3�* transitionof each nucleotide, leading to negative induced absorbance (�A)signals, which allow the repopulation of S0 to be observed. Signalswere recorded at a probe wavelength of 340 nm because a numberof intermediates potentially absorb here, including vibrationally hotS0 molecules (�360 nm; refs. 30 and 31), pyrimidine triplet states

(300–550 nm; ref. 32), and a dark singlet state recently detected in1-cyclohexyluracil (1CHU) (300–450 nm; ref. 31).

For each of the pyrimidine mononucleotides, cytidine 5�-monophosphate (CMP), thymidine 5�-monophosphate (dTMP),and uridine 5�-monophosphate (UMP), the signal at 340 nm ispositive and approximately mirrors the 250-nm bleach signal atdelay times greater than �10 ps. Parameters from a global fit to thesignals at 250 and 340 nm using two exponentials plus an offset aresummarized in Table 1. At both probe wavelengths a positive spikeis present at short delay times when pump and probe pulses overlap.This feature arises from coherent two-photon absorption by solventmolecules and was excluded from the fits. No long-lived decaycomponents are observed from the purine ribonucleotide adeno-sine 5�-monophosphate (AMP) at the same two wavelengths (Fig.2). The slow-decay component (�2 in Table 1) is seen in allpyrimidine bases and base derivatives studied but is not present insignals recorded at 570 nm (Fig. 3). It is known that transientabsorption signals at visible wavelengths from pyrimidine andpurine nucleobases decay on a subpicosecond time scale (1, 6, 30).

Among the unmodified pyrimidine bases, �2 increases in thefollowing order: Cyt � Ura � Thy. The same order is observed forthe 5� mononucleotides of these compounds, but �2 is three to sixtimes longer (Fig. 4 and Table 1). The same lifetimes are observed,within experimental uncertainty, for the 5� mononucleotides andnucleosides (data not shown) of Cyt, Thy, and Ura. Lifetimes ofboth 1,3-dimethyluracil (DMU) and 1CHU are more similar to thefree base Ura than to the 5� mononucleotide UMP (Table 1).

DiscussionSignal Assignments. The pump pulse at 267 nm abruptly depletes theground-state population by exciting molecules to the 1��* state.This results in the negative absorbance changes (�A � 0) seenimmediately after time 0 at 250 nm (Fig. 2), a wavelength absorbed

Fig. 1. Nonradiative photophysical (A) and photochemical (B) decay path-ways on hypothetical potential-energy surfaces. The surfaces denote theenergies of the ground electronic state (S0) and the lowest excited electronicstate (S1) as a function of two representative coordinates, q1 and q2, chosenfrom a nuclear configuration space of much higher dimensionality. The decaypaths are shown by dashed lines on the surfaces and by solid lines projectedonto the q1, q2 plane. The photophysical pathway in A moves from theFranck–Condon region reached by photon absorption (black arrow) to a CIwith S0 and finally back to the starting point at the minimum of S0. Photo-chemical decay in B takes the excited-state wavepacket from a reactantgeometry on S0 (point R) to the geometry of a new chemical product (P).

Fig. 2. Transient absorption traces of dTMP, UMP, CMP, and AMP in pH 7buffer pumped at 267 nm and probed at 340 nm (blue squares) and 250 nm(circles). Signals have been scaled to match the short time signal at 340 nm.Solid lines are from fits to the data. Fit parameters can be found in Table 1.

436 � www.pnas.org�cgi�doi�10.1073�pnas.0608055104 Hare et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0

strongly by ground-state molecules. Steady-state absorption spectraare shown for all molecules studied in supporting information (SI)Fig. 6. The recovery of this signal to �A � 0 measures the timeneeded for the excited population to return to S0, as discussedelsewhere (33). The pronounced biphasic recovery at 250 nmindicates unambiguously that S0 is repopulated on two distinct timescales defined by �1 and �2. In the following discussion we refer tothese as the fast and slow channels for IC.

Assignment of the fast channel to direct IC from the initial 1��*state to S0 is supported by signals at several probe wavelengths. At570 nm, where the 1��* states of the nucleobases show excited-stateabsorption (30), subpicosecond decays are observed (Fig. 3). Non-radiative decay to S0 creates vibrationally hot ground-state mole-cules. As a result, the S0 absorption spectrum is transiently shiftedto longer wavelengths until vibrational cooling can reestablishthermal equilibrium with the surrounding solvent molecules (6, 30,31, 33). The hot S0 absorption spectrum changes most rapidly atlong wavelengths. Thus, relatively rapid decays of �1 ps (�1 in Table1) are observed at 340 nm. There may also be a contribution at thisprobe wavelength from excited-state absorption by the short-lived1��* population. The fast component (�1 in Table 1) of the bleach

signal at 250 nm of between 2 and 4 ps indicates that vibrationalcooling is complete on this time scale (6, 30, 31, 33). The observedvalues for �1 are in good agreement with previous observations forAMP and dTMP in aqueous solution (33).

The slow channel is the central discovery communicated in thiswork and will be the focus of subsequent discussion. It is observedin all pyrimidine bases in aqueous solution but is conspicuously

Fig. 4. Normalized transient absorption signals for the bases (circles) andnucleotides (squares) of Ura, Cyt, and Thy pumped at 267 nm and probed at 250nm. Solid lines are fits to the data. Fit parameters can be found in Table 1.

Table 1. Fit parameters at indicated probe wavelengths for the pyrimidines studied inaqueous solution

Nucleobase �, nm �1, ps A1, % �2, ps A2, % A3, %

CMP 250 3.7 � 0.8 �56 � 34 34 � 3 �41 � 19 �7 � 4340 0.7 � 0.3 30 — 50 20

Cyt 250 2.9 � 0.7 �87 � 9 12 � 3 �9 � 7 �4 � 4340 0.7 � 0.1 77 — 18 6

TMP 250 2.2 � 0.1 �83 � 3 127 � 15 �14 � 4 �2 � 3340 0.41 � 0.09 61 — 18 21

Thy 250 2.8 � 0.4 �87 � 13 30 � 13 �11 � 13 �1.3 � 1340 0.72 � 0.07 77.4 — 8 14

UMP 250 2.3 � 0.2 �51 � 6 147 � 7 �42 � 6 �7 � 1340 0.45 � 0.05 66 — 23 11

Ura 250 1.9 � 0.1 �68 � 5 24 � 2 �28 � 5 �3 � 4340 0.55 � 0.06 68 — 22 9

DMU 250 3 � 1 �68 � 10 9.9 � 0.8 �28 � 14 �3 � 3340 0.57 � 0.03 76 — 17 6

1CHU* 250 2.4 � 0.3 �62 � 5 26 � 2 �32 � 5 �3 � 2340 0.9 � 0.3 75 — 24 0.90

Each parameter marked with a dash was globally linked to the row above in fitting. Uncertainty intervals are95% confidence estimates.*Data are from ref. 31.

Fig. 3. Solvent-corrected transient absorption traces of CMP, dTMP, UMP,and AMP in pH 7 buffer pumped at 267 nm and probed at 570 nm. Structuresof the bases are shown with R representing a ribose-phosphate group. Solidlines are from global fits to the data.

Hare et al. PNAS � January 9, 2007 � vol. 104 � no. 2 � 437

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0

absent in purine bases like AMP (Fig. 2) and guanosine 5�-monophosphate (data not shown). The slow lifetime (�2 in Table 1)is assigned to relaxation of an intermediate electronic state thatabsorbs at 340 nm (Fig. 2) but not at visible wavelengths (Fig. 3).The approximate mirror symmetry of the 250- and 340-nm tran-sients indicates that population in the intermediate state relaxesdirectly to S0. The electronic state absorbing at 340 nm cannot bethe 1��* state because transients at visible probe wavelengths (Fig.3) and previous transient absorption and fluorescence-upconver-sion measurements (1) indicate that the latter state has a lifetimeof �1 ps. The long-lived population at 340 nm is in an intermediateelectronic state along the decay pathway from the 1��* state to S0.Assignment of the intermediate state to a minor tautomer or aspecies produced by photoionization was considered and dismissed(see SI Discussion). Because emission from pyrimidine nucleobasesis not observed on the �2 time scale (15, 34, 35), the intermediatestate is a ‘‘dark’’ state with a negligible radiative transition rate toS0. The nature of this dark state is discussed below in more detail.

Residual negative and positive offsets seen at 250 and 340 nm,respectively (A3 in Table 1), are assigned to absorption by long-livedtriplet states (31). These states have absorption maxima between350 and 450 nm but do not absorb beyond 550 nm (36), consistentwith the weak offset seen at 340 but not at 570 nm (Figs. 2 and 3).Triplet yields for pyrimidine bases and their nucleosides andnucleotides are �2% in aqueous solution (37, 38). The amplitudeA3 has a large uncertainty because of the weak signals but isconsistent with triplet state formation by no more than a fewpercent of excited molecules. This assignment is further supportedby our prior study of 1CHU in a variety of solvents, including lesspolar ones where intersystem crossing (ISC) yields are large enoughto be accurately measured (31). Some of the long-time bleach signalat 250 nm may also arise from photohydrates. These photoadditionproducts with water are formed in CMP and UMP with quantumyields of no more than 2% (38).

Nonradiative Decay Yields. For the pyrimidine bases, at least 98% ofall excited molecules decay nonradiatively to S0 in aqueous solutionat room temperature (38). Until now this decay was thought tooccur solely via ultrafast IC to S0 (the fast channel) (1). Theadditional slow channel reported here shows that the 98% IC yieldmust contain contributions from two parallel decay pathways.

The fraction of excited molecules that decay via the fast and slowchannels was evaluated from the relative amplitudes A1, A2, and A3(Table 1) of the bleach recovery signals at 250 nm (Figs. 2 and 4).The yield for the fast direct IC is A1, whereas A2 is the yield ofmolecules that decay via IC from the dark state to S0. Theseestimates are accurate provided that S0 is the only state withsignificant absorption at this wavelength (31, 33). There is nostimulated emission at 250 nm because this wavelength lies abovethe electronic origin of the emissive 1��* state of each compound.Hence, all states other than S0 can contribute only positive �Achanges. The overall negative sign of the signals at 250 nm indicatesthat S0 is the dominant absorber at this wavelength. If the inter-mediate electronic state were to absorb at 250 nm, then the relativeamplitude of the �2 decay (A2) would decrease. A2 is thus a lowerbound for the true yield (33).

From the A2 values in Table 1, the dark-state yield is 42 � 6%in UMP, 41 � 19% in CMP, and 14 � 4% in dTMP. Because ISCto the triplet state is proposed to occur from the dark state (seebelow), the yield of molecules entering the slow channel (1 � A1)is likely somewhat higher. The free bases have somewhat smalleryields of nonradiative decay via the dark state compared with theirnucleotides (Table 1 and Fig. 4). Interestingly, a larger yield is notsimply due to substitution at N1 of the pyrimidines because similaryields are found for Ura (28 � 5%), 1CHU (32 � 5%), and DMU(28 � 14%). Collectively, these observations show that decay via theslow channel accounts for between 10% and 50% of all deexcitationevents.

Assignment of the Dark State. Next, we propose an assignment forthe dark state populated in the slow-decay channel and discuss itsdynamics. The main criteria used in making our assignment are theaccessibility and stability of this state. The dark state must beenergetically accessible from the 1��* state in an aqueous envi-ronment because only this state is significantly excited. In addition,the dark state requires a stable minimum, which is separated fromCIs with lower energy states by sufficiently high barriers to explainthe many-picosecond lifetimes. Ab initio calculations provide es-sential guidance. Numerous calculations have shown that there areseveral low-lying singlet excited states classified by their 1��* and1n�* character.

The number of candidate singlet states is limited (see SI Discus-sion for why triplet states can be ruled out). A singlet biradical statereachable from the 1��* state was proposed by Zgierski andcolleagues (13, 14) to be a key intermediate in the nonradiativedecay of Cyt and Ura. This state is reached by a barrierless pathwayinvolving torsional motion about the C5,C6 double bond reminis-cent of dynamics seen in the 1��* state of ethene (39). Zgierski andcolleagues (13, 14) suggest that a state switch occurs from the 1��*state to the biradical state, but calculations by others (12, 20, 22) findan essentially identical pathway, which does not appear to departfrom the 1��* surface. This fact and the apparent absence of astable minimum before the CI with S0 lead us to reject assignmentto a biradical state.

We assign the dark state with absorption at 340 nm and lifetime�2 to a lowest-energy 1n�* state. Calculations on Cyt and Ura haveshown that the 1n�* state is the lowest-energy excited singlet stateat its energy-minimized geometry (10, 22, 28). The existence of astable 1n�* minimum on the excited-state potential-energy surfaceis suggestive of conditions necessary for the slow relaxation seenhere. Lifetimes of more than a few picoseconds are rarely observedin the condensed phase for excited states that are not the lowest-energy excited state at their minimum energy geometry (Kasha’srule). However, the stability of this state is difficult to predict,because we know of no calculations of energetic barriers separatingthe minimum of the lowest 1n�* state of each pyrimidine from a CIwith, for example, S0. Studies of the energy-minimized 1n�* stateof 9H-adenine have indicated that barriers of just 0.1 eV separateits minimum from CIs with other states, including S0 (18, 19, 23).Thus, even if the 1n�* state were accessible from the 1��* state of9H-adenine, this low barrier suggests that it would decay to S0 onan ultrafast time scale, explaining the absence of slow channel decayin AMP (Fig. 2). Marian (23) has suggested that the small barrierin 9H-adenine would vanish in solution because of blue shifting ofthe 1n�* state.

According to calculations, both Ura (12, 15, 40) and Thy (15, 40)have a lowest-energy 1n�* state in the gas phase at the equilibriumground-state geometry. On the other hand, the 1n�* state ofisolated Cyt lies slightly above the 1��* state (22). When aqueoussolvation effects are included, the lowest 1n�* state shifts to beessentially isoenergetic with the 1��* state of Ura and moves abovethe 1��* state for Thy (15). However, the energetic state orderingat the ground-state geometry (the vertical state ordering) is notpredictive of the order found at other geometries as demonstratedin many recent calculations (21, 22, 28). Thus, a 1n�* state that liesat higher energy in the Franck–Condon region can become acces-sible at another geometry reached by nuclear motions in the 1��*state. Fig. 5 is a qualitative illustration of the two nonradiative decaypathways after bifurcation of the initial excited-state population. Itwill be important to investigate in the future how solvation affectsthese state crossings, which to our knowledge have only beenstudied computationally in the gas phase. Finally, we note that thereis experimental precedent for a 1n�* state with a picosecondlifetime in a closely related compound in the condensed phase (41).Chachisvillis and Zewail (41) measured a lifetime of 9–23 ps,depending on solvent, for the 1n�* state of pyridine, pyrimidine,and triazine in femtosecond transient absorption experiments.

438 � www.pnas.org�cgi�doi�10.1073�pnas.0608055104 Hare et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0

Sequential decay via a 1n�* state (1��*3 1n�*3 S0) has beenfrequently discussed as the main deactivation channel for isolatednucleobases in the gas phase (2, 27, 29, 42–44), although there isconsiderable disagreement about the precise kinetics. Reportedlifetimes from photoelectron measurements for the 1n�* state areon the order of several hundred femtoseconds to several picosec-onds for all of the pyrimidines (29, 42, 43). In contrast to thesereports, Kong and coworkers (45, 46) detected a state with amany-nanosecond lifetime upon electronic excitation of isolatedThy and Ura. They assigned this state to a 1n�* state, which theysuggested would be undetectable in solution (46). Our results showthat this state is clearly not quenched by hydration. Additionally, theobservation that branching yields for 1CHU are essentially the samein polar and nonpolar solvents (31) suggests that both fast and slowchannels could be observable in the isolated molecules.

Mechanism of Dark-State Decay. The nuclear motions that lead todecay of the dark state cannot be determined from our transientabsorption measurements. Nonetheless, the experimental resultssuggest some general aspects of the decay mechanism, which wenow discuss. The lack of emission on the picosecond time scale forany pyrimidine base (15, 35, 47) indicates that population in theintermediate electronic state never returns to the fluorescentregion of the 1��* state. We propose that the 10- to 150-ps lifetimeof the dark 1n�* state is due to a sizeable barrier that restricts accessfrom the minimum of this state to a CI with S0 (Fig. 5).

One of the most striking results from this study is the 3- to6-fold-longer lifetime of the dark state in the pyrimidine nucleotidescompared with their free bases (Fig. 4). In contrast, the 1��*lifetimes of free bases and their mononucleotides change by nomore than 50% (1). These observations suggest strongly thatdifferent nuclear coordinates are responsible for nonradiative decayby these excited states. The effect on the dark-state lifetime is the

specific result of ribosyl substitution because two other N1-substituted derivatives, 1CHU and DMU, have shorter lifetimesthat match the free base Ura (Table 1).

The free base and its nucleotide have very similar electronicstructure because excitation is localized on the aromatic basemoiety. The absorption spectra of nucleosides and nucleotides ofpyrimidine bases are slightly red-shifted compared with the freebases, but similar red shifts are observed for 1CHU and DMU (SIFig. 6), which do not show increased lifetimes compared with Ura(Table 1). Additionally, preliminary experiments in which theexcitation wavelength was varied from 260 to 285 nm show nosignificant change in either dark-state yield or lifetime for 1CHU.These findings indicate that the amount of excess energy in theFranck–Condon excited state does not control the 1n�* lifetime.

We propose that the rate of excess vibrational energy transfer tothe solvent is responsible for the lifetime differences between thefree bases and their nucleotides. It has been shown previously thatsolute-solvent hydrogen bonds strongly enhance vibrational coolingin nucleobase monomers (30). The additional hydrogen bondingsites of the ribosyl group thus allow the nucleotides to dissipateexcess vibrational energy to the solvent more rapidly than in the freebases, provided that vibrational energy flow from the base to thesugar is unrestricted. As a result of their greater energy content, thefree bases undergo more rapid barrier crossing to S0 and thereforehave shorter lifetimes than the nucleotides.

Mediation of ISC by the 1n�* State. Although a minor decay channelin pyrimidines, ISC to the triplet state is important because of therole that these long-lived states may play as precursors to variousDNA photoproducts (38). We propose that the 3��* state ispopulated in pyrimidine bases via the intermediate 1n�* state (31),a transition favored over direct 1��* to 3��* ISC by classicalpropensity rules (48). ISC is proposed to occur efficiently only whenthe 1n�* state has sufficient excess energy. Once molecules in the1n�* state have cooled, ISC is no longer possible and the onlyaccessible decay channel is IC to S0. In support of this model, nogrowth in transient signals was observed in 1CHU at wavelengthswhere the triplet state absorbs (31). In support of the notion thatISC occurs along the 1n�* decay pathway, we note that the purinebases, which lack observable 1n�* population, have much lowertriplet yields in room temperature aqueous solution (38).

Photochemical Possibilities. Although the photoproducts formed inDNA by UV light have been known for decades in some cases, itis still unknown what excited states lead to reaction. Because thiswork shows that decay of the initial 1��* population creates largenumbers of relatively long-lived 1n�* states, it is important toconsider whether the latter states are responsible for any of thepyrimidine photoproducts known to form from singlet precursors.This includes the pyrimidine photohydrates, which cannot beformed via triplet sensitization, and the pyrimidine (6–4) pyrimi-done photoproducts (38).

Pyrimidine photohydrates are formed in monomeric bases and inDNA when a water molecule adds across the C5,C6 double bond.Quantum yields of hydration are similar for CMP and UMP butvanishingly small for dTMP and other 5-substituted pyrimidines onaccount of steric hindrance (49). Likewise, our results indicatehigher 1n�* yields for CMP and UMP compared with dTMP.Additionally, increased photohydrate yields are observed in nucle-otides relative to the free bases (38) in agreement with theexperimental 1n�* yields. However, there is one trend that wecannot explain: the photohydrate yield of DMU is reported to beone order of magnitude higher than Ura (49), yet similar 1n�* yieldsare seen for both.

The first step in the reaction to form a pyrimidine (6–4)pyrimidone is formation of either an oxetane or azetidine inter-mediate (38). This is a [2�2] photocycloaddition reaction betweenthe carbonyl group of Thy or the imino group of Cyt and the C5,C6

Fig. 5. Schematic of proposed nonradiative decay mechanism for the pyrimi-dines shown in two arbitrary coordinates as a function of energy. After excitationto the 1��* state (green), population moves toward the minimum of this state(1��*min), encountering a CI with the 1n�* state (blue) along the way. Uponleaving this CI it can either go toward 1��*min leading to fluorescence (bluearrow)or the 1��*/S0 CI (rightpath),or it cangotowardtheminimumofthe 1n�*state(leftpath).Afterthe 1��*/S0 CI, themoleculerelaxesbacktotheS0 minimum(gray). To reach S0 from the 1n�* state, a significant barrier must be overcome.Not shown are the 1n�* to 3��* crossing and any photochemical paths that mayoccur from any of the three states depicted.

Hare et al. PNAS � January 9, 2007 � vol. 104 � no. 2 � 439

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0

double bond of a neighboring pyrimidine. It has thus been sug-gested that a 1n�* state is the reactive excited state (50). Higheryields are found for doublets where the 3� base is Cyt and not Thy(51). This matches the higher yield we see for the 1n�* state of Cytcompared with Thy.

An open question is whether the 1n�* state detected here in basemonomers is present in base multimers, including DNA andpolynucleotides. Based on the findings described here, we are nowconfident that the long-lived singlet state seen in our earlierexperiments on (dT)18 (33) is the same state seen in dTMP.However, it has been argued that there is little stacking in (dT)n(52), and further work is needed to determine whether basestacking, which strongly influences the electronic structure of DNA(33), inhibits the formation or alters the decay rate of the 1n�* stateseen in the monomeric bases. This clearly has important conse-quences for pyrimidine (6–4) pyrimidone formation because thesephotoproducts are formed only in base-stacked doublets. In sum-mary, much more work is needed before any photoproducts can bedefinitively associated with 1n�* states.

ConclusionsWe have shown that the 1��* population created by UV excitationof pyrimidine bases in room temperature aqueous solution decaysvia two very different nonradiative decay channels shown schemat-ically in Fig. 5: ultrafast IC to S0 and decay to a dark state. The latterstate, which has an �100-fold-longer lifetime than the 1��* state,is tentatively assigned to a 1n�* state and is proposed to be agateway state to the triplet state. The photochemical consequencesof this state are poorly understood at present, but there are somesuggestive correlations with yields for photohydrates and pyrimi-dine (6–4) pyrimidone photoproducts. Future computational mod-els of excited states of pyrimidine bases must account for theunusual branching in the singlet manifold, which appears to alsotake place in the gas phase. Dynamical calculations will be essentialfor estimating branching yields and for testing the hypothesis thatvibrational excess energy in the 1n�* state is an important variablecontrolling its lifetime.

This work shows the power of the transient absorption methodfor following the evolution of dark excited states, which areundetectable in emission measurements. High yields of long-livedsinglet states have been observed in DNA oligomers (33),polynucleotides (53), and now in monomeric bases in aqueous

solution. This shows that it is an oversimplification to equate thephotostability of DNA or its building blocks with subpicosecond ICdynamics. Although ultrafast nonradiative decay is still operative inall of these systems, large numbers of singlet excitations are formedin each, which decay orders of magnitude more slowly than the1��* states. Much more experimental and theoretical work isneeded to understand the nature of these long-lived states and theirpotential for photochemical damage.

MethodsThe experimental setup has been described previously (30, 33).Briefly, signals were recorded by using a Ti:sapphire-based tran-sient absorption spectrometer. Pump pulses were provided by thethird harmonic of the laser fundamental. Probe pulses were ob-tained from a tunable optical parametric amplifier (Coherent,Santa Clara, CA). Pump pulse intensities were typically in the rangeof 0.2–2 GW�cm�2. Probe pulses were always attenuated to have atmost 1/10th of the intensity of the pump pulses. The transmittedprobe pulse was detected with a photomultiplier tube after spectralfiltering with a double grating monochromator. Pump and probepulse polarizations were set to magic angle to eliminate reorienta-tion signals. Transient absorption signals were recorded by using alock-in amplifier referenced to a chopper placed in the pump path.The instrument response function was �200 fs as determined by theFWHM of the coherent two-photon absorption seen in solvent-onlyscans. Approximately 0.4 ml of the solution under study was held ina 1.2-mm-path-length cell between CaF2 windows. The cell wasspun at several hundred rpm about its axis to avoid reexcitation ofthe pumped volume by successive laser pulses. Signals were care-fully monitored for photodegradation by UV absorption spectros-copy during the experiments, and solutions were replaced with freshones as needed. All chemicals were purchased from Sigma-Aldrich(St. Louis, MO) and used as received. Samples were prepared tohave an absorbance between 1 and 2 at the pump wavelength in pH7 phosphate buffer.

B.K. acknowledges helpful discussions with Prof. Todd J. Martinez andProf. Wolfgang Domcke. Measurements were performed in Ohio StateUniversity’s Center for Chemical and Biophysical Dynamics with equip-ment funded by the National Science Foundation and the Ohio Board ofRegents. This research was supported by National Institutes of HealthGrant R01 GM64563.

1. Crespo-Hernandez CE, Cohen B, Hare PM, Kohler B (2004) Chem Rev 104:1977–2019.2. Abo-Riziq A, Grace L, Nir E, Kabelac M, Hobza P, de Vries MS (2005) Proc Natl Acad

Sci USA 102:20–23.3. Robb MA, Bernardi F, Olivucci M (1995) Pure Appl Chem 67:783–789.4. Domcke W, Yarkony DR, Koppel H (2004) Conical Intersections: Electronic Structure,

Dynamics and Spectroscopy (World Scientific, Singapore).5. Bernardi F, Olivucci M, Robb MA (1996) Chem Soc Rev 25:321-328.6. Pecourt J-ML, Peon J, Kohler B (2000) J Am Chem Soc 122:9348–9349.7. Srinivasan R, Feenstra JS, Park ST, Xu SJ, Zewail AH (2005) Science 307:558–563.8. Ismail N, Blancafort L, Olivucci M, Kohler B, Robb MA (2002) J Am Chem Soc

124:6818–6819.9. Sobolewski AL, Domcke W (2002) Eur Phys J D 20:369–374.

10. Merchan M, Serrano-Andres L (2003) J Am Chem Soc 125:8108–8109.11. Blancafort L, Robb MA (2004) J Phys Chem A 108:10609–10614.12. Matsika S (2004) J Phys Chem A 108:7584–7590.13. Zgierski MZ, Patchkovskii S, Lim EC (2005) J Chem Phys 123:081101.14. Zgierski MZ, Patchkovskii S, Fujiwara T, Lim EC (2005) J Phys Chem A 109:9384–9387.15. Gustavsson T, Banyasz A, Lazzarotto E, Markovitsi D, Scalmani G, Frisch MJ, Barone

V, Improta R (2006) J Am Chem Soc 128:607–619.16. Chen H, Li SH (2006) J Chem Phys 124:154315.17. Chen H, Li SH (2005) J Phys Chem A 109:8443–8446.18. Blancafort L (2006) J Am Chem Soc 128:210–219.19. Perun S, Sobolewski AL, Domcke W (2005) J Am Chem Soc 127:6257–6265.20. Sobolewski AL, Domcke W (2004) Phys Chem Chem Phys 6:2763–2771.21. Serrano-Andres L, Merchan M, Borin AC (2006) Proc Natl Acad Sci USA 103:8691–8696.22. Tomic K, Tatchen J, Marian CM (2005) J Phys Chem A 109:8410–8418.23. Marian CM (2005) J Chem Phys 122:104314.24. Kang H, Lee KT, Jung B, Ko YJ, Kim SK (2002) J Am Chem Soc 124:12958–12959.25. Kang H, Jung B, Kim SK (2003) J Chem Phys 118:6717–6719.26. Broo A (1998) J Phys Chem A 102:526–531.27. Canuel C, Mons M, Piuzzi F, Tardivel B, Dimicoli I, Elhanine M (2005) J Chem Phys 122:074316.

28. Marian CM, Schneider F, Kleinschmidt M, Tatchen J (2002) Eur Phys J D 20:357–367.29. Samoylova E, Lippert H, Ullrich S, Hertel IV, Radloff W, Schultz T (2005) J Am Chem

Soc 127:1782–1786.30. Pecourt J-ML, Peon J, Kohler B (2001) J Am Chem Soc 123:10370–10378.31. Hare PM, Crespo-Hernandez CE, Kohler B (2006) J Phys Chem B 110:18641–18650.32. Gut IG, Wood PD, Redmond RW (1996) J Am Chem Soc 118:2366–2373.33. Crespo-Hernandez CE, Cohen B, Kohler B (2005) Nature 436:1141–1144.34. Pal SK, Peon J, Zewail AH (2002) Chem Phys Lett 363:57–63.35. Onidas D, Markovitsi D, Marguet S, Sharonov A, Gustavsson T (2002) J Phys Chem B

106:11367–11374.36. Wood PD, Redmond RW (1996) J Am Chem Soc 118:4256–4263.37. Salet C, Bensasson R, Becker RS (1979) Photochem Photobiol 30:325–329.38. Cadet J, Vigny P (1990) in Bioorganic Photochemistry, ed Morrison H (Wiley, New York), Vol

1, pp 1–272.39. Ben-Nun M, Martinez TJ (2000) Chem Phys 259:237–248.40. Lorentzon J, Fulscher MP, Roos BO (1995) J Am Chem Soc 117:9265–9273.41. Chachisvillis M, Zewail AH (1999) J Phys Chem A 103:7408–7418.42. Ullrich S, Schultz T, Zgierski MZ, Stolow A (2004) J Am Chem Soc 126:2262–2263.43. Ullrich S, Schultz T, Zgierski MZ, Stolow A (2004) Phys Chem Chem Phys 6:2796–2801.44. Schneider M, Maksimenka R, Buback FJ, Kitsopoulos T, Lago LR, Fischer I (2006) Phys

Chem Chem Phys 8:3017–3021.45. He Y, Wu C, Kong W (2003) J Phys Chem A 107:5145–5148.46. He Y, Wu C, Kong W (2004) J Phys Chem A 108:943–949.47. Gustavsson T, Sharonov A, Markovitsi D (2002) Chem Phys Lett 351:195–200.48. El-Sayed MA (1962) J Chem Phys 36:573–574.49. Fisher GJ, Johns HE (1976) in Photochemistry and Photobiology of Nucleic Acids, ed

Wang SY (Academic, New York), Vol 1, pp 169–224.50. Patrick MH (1977) Photochem Photobiol 25:357–372.51. Lemaire DGE, Ruzsicska BP (1993) Photochem Photobiol 57:755–769.52. Mills JB, Vacano E, Hagerman PJ (1999) J Mol Biol 285:245–257.53. Crespo-Hernandez CE, Kohler B (2004) J Phys Chem B 108:11182–11188.

440 � www.pnas.org�cgi�doi�10.1073�pnas.0608055104 Hare et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

13,

202

0