Dynamic determination of the functional state inphotolyase and the implication for cryptochromeZheyun Liu a,b, Meng Zhanga,b,c, Xunmin Guo a,b, Chuang Tan a,b,d, Jiang Li a,b, Lijuan Wang a,b, Aziz Sancare,1,and Dongping Zhong a,b,c,d,f,1

Departments of aPhysics and bChemistry and Biochemistry, and Programs of cBiophysics, dChemical Physics, and fBiochemistry, The Ohio State University,Columbus, OH 43210; and eDepartment of Biochemistry and Biophysics, University of North Carolina School of Medicine, Chapel Hill, NC 27599

Contributed by Aziz Sancar, June 28, 2013 (sent for review May 26, 2013)

The flavin adenine dinucleotide cofactor has an unusual bentconfiguration in photolyase and cryptochrome, and such a foldedstructuremay have a functional role in initial photochemistry. Usingfemtosecond spectroscopy, we report here our systematic charac-terization of cyclic intramolecular electron transfer (ET) dynamicsbetween the flavin and adenine moieties of flavin adenine di-nucleotide in four redox forms of the oxidized, neutral, and anionicsemiquinone, and anionic hydroquinone states. By comparing wild-type and mutant enzymes, we have determined that the excitedneutral oxidized and semiquinone states absorb an electron fromthe adenine moiety in 19 and 135 ps, whereas the excited anionicsemiquinone and hydroquinone states donate an electron to theadenine moiety in 12 ps and 2 ns, respectively. All back ET dynamicsoccur ultrafast within 100 ps. These four ET dynamics dictate that onlythe anionic hydroquinone flavin can be the functional state in photo-lyase due to the slower ET dynamics (2 ns) with the adenine moietyand a faster ET dynamics (250 ps) with the substrate, whereas theintervening adenine moiety mediates electron tunneling for repair ofdamaged DNA. Assuming ET as the universal mechanism for photo-lyase and cryptochrome, these results imply anionic flavin as themoreattractive form of the cofactor in the active state in cryptochrome toinduce charge relocation to causeanelectrostatic variation in theactivesite and then lead to a local conformation change to initiate signaling.

flavin functional state | intracofactor electron transfer | adenineelectron acceptor | adenine electron donor | femtosecond dynamics

The photolyase–cryptochrome superfamily is a class of fla-voproteins that use flavin adenine dinucleotide (FAD) as the

cofactor. Photolyase repairs damaged DNA (1–5), and crypto-chrome controls a variety of biological functions such as regulatingplant growth, synchronizing circadian rhythms, and sensing di-rection as a magnetoreceptor (6–10). Strikingly, the FAD cofactorin the superfamily adopts a unique bent U-shape configurationwith a close distance between its lumiflavin (Lf) and adenine (Ade)moieties (Fig. 1A). The cofactor could exist in four different redoxforms (Fig. 1B): oxidized (FAD), anionic semiquinone (FAD–),neutral semiquinone (FADH•), and anionic hydroquinone (FADH–).In photolyase, the active state in vivo is FADH–. We have recentlyshowed that the intervening Ade moiety mediates electron tun-neling from the Lf moiety to substrate in DNA repair (5). Becausethe photolyase substrate, the pyrimidine dimer, could be eitheran oxidant (electron acceptor) or a reductant (electron donor),a fundamental mechanistic question is why photolyase adoptsFADH– as the active state rather than the other three redoxforms, and if an anionic flavin is required to donate an electron,why not FAD–, which could be easily reduced from FAD?In cryptochrome, the active state of the flavin cofactor in vivo is

currently under debate. Two models of cofactor photochemistryhave been proposed (11–14). One is called the photoreductionmodel (11–13), which posits that the oxidized FAD is photo-reducedmainly by a conserved tryptophan triad to neutral FADH•

(signaling state) in plant or FAD– in insect, then triggering struc-tural rearrangement to initiate signaling. The other model (14, 15)hypothesizes that cryptochrome uses a mechanism similar to that

of photolyase by donating an electron from its anionic form (FAD–

in insect or FADH– in plant) to a putative substrate that inducesa local electrostatic variation to cause conformation changes forsignaling. Both models require electron transfer (ET) at the activesite to induce electrostatic changes for signaling. Similar to thepyrimidine dimer, the Ade moiety near the Lf ring could also bean oxidant or a reductant. Thus, it is necessary to know the role ofthe Ade moiety in initial photochemistry of FAD in cryptochrometo understand the mechanism of cryptochrome signaling.Here, we use Escherichia coli photolyase as a model system to

systematically study the dynamics of the excited cofactor infour different redox forms. Using site-directed mutagenesis, wereplaced all neighboring potential electron donor or acceptoramino acids to leave FAD in an environment conducive to for-mation of one of the four redox states. Strikingly, we observedthat, in all four redox states, the excited Lf proceeds to intra-molecular ET reactions with the Ade moiety. With femtosecondresolution, we followed the entire cyclic ET dynamics and de-termined all reaction times of wild-type and mutant forms of theenzyme to reveal the molecular origin of the active state of flavinin photolyase. With the semiclassical Marcus ET theory, wefurther evaluated the driving force and reorganization energyof every ET step in the photoinduced redox cycle to understandthe key factors that control these ET dynamics. These observa-tions may imply a possible active state among the four redoxforms in cryptochrome.

Results and DiscussionPhotoreduction-Like ET from Adenine to Neutral Oxidized (Lf) andSemiquinoid (LfH•) Lumiflavins. As reported in the preceding pa-per (16), we have shown that the excited FAD in photolyase isreadily quenched by the surrounding tryptophan residues, mainlyW382 with a minor contribution from W384, and that the ETdynamics from W382 to FAD* occurs ultrafast in 0.8 ps. ByreplacingW382 andW384 to a redox inert phenylalanine (W382F/W384F) using site-directed mutagenesis, we abolished all possibleET between FAD* and the neighboring aromatic residues andobserved a dominant decay of FAD* in 19 ps (an average time ofa stretched exponential decay with τ = 18 ps and β = 0.92) as shownin Fig. 2A (kFET

−1) with a probing wavelength at 800 nm. Theobserved stretched behavior reflects a heterogeneous quenchingdynamics, resulting from the coupling of ET with the active-sitesolvation on the similar timescales (17). The dynamics in 19 psreflects the intramolecular ET from the Ade to Lf moieties toform a charge-separated pair of Ade++Lf–. Tuning the probewavelengths to shorter than 700 nm to search for the maximum

Author contributions: D.Z. designed research; Z.L., M.Z., X.G., C.T., J.L., L.W., and D.Z.performed research; Z.L. and D.Z. analyzed data; and Z.L., A.S., and D.Z. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: dongping@mps.ohio-state.edu oraziz_sancar@med.unc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1311077110/-/DCSupplemental.

12972–12977 | PNAS | August 6, 2013 | vol. 110 | no. 32 www.pnas.org/cgi/doi/10.1073/pnas.1311077110

Dow

nloa

ded

by g

uest

on

June

19,

202

0

contribution of the putative Ade+ intermediate, we show twotypical transients in Fig. 2 B and C probed at 630 and 580 nm,respectively. We observed the formation of Ade+ in 19 ps anddecay in 100 ps (see all data analyses thereafter in SI Text). Thedecay dynamics reflects the charge recombination process (kBET

−1)and leads to the completion of the redox cycle. As discussed in thepreceding paper (16), such ET dynamics between the Lf and Ademoieties is favorable by negative free-energy changes.Similarly, we prepared the W382F mutant in the semiquinone

state (FADH•) to eliminate the dominant electron donor ofW382. Without this tryptophan in proximity, we observed adominant decay of FADH•* in 85 ps (τ = 82 ps and β = 0.93)probed at 800 nm (Fig. 3A), which is similar to the previouslyreported 80 ps (18) that was attributed to the intrinsic lifetime ofFADH•*. In fact, the lifetime of the excited FMNH•* in fla-vodoxin is about 230 ps (19), which is nearly three times longerthan that of FADH•* observed here. Using the reductionpotentials of +1.90 V vs. normal hydrogen electrode (NHE) foradenine (20) and of +0.02 V vs. NHE in photolyase for neutralsemiquinoid LfH• (21), with the S1←S0 transition of FADH• at650 nm (1.91 eV) we find that the ET reaction from Ade toLfH•* has a favorable, negative free-energy change of −0.03 eV.

Thus, beside the intrinsic lifetime, the excited LfH• is likely to bequenched by intramolecular ET with Ade to form a charge-separated pair of Ade++LfH–. Taking 230 ps as the lifetime ofLfH•* without ET, we derive a forward ET dynamics with Ade in135 ps, contributing to an overall decay of FADH•* in 85 ps.To probe the intermediate Ade+, we tuned the probe wave-

lengths to the shorter wavelengths to detect the maximum in-termediate contribution. The best probing wavelength would be theone at which the absorption coefficients of the excited and groundstates are equal, resulting in cancellation of the positive LfH•*signal by the negative partial LfH• formation signal, leading to thedominant rise and decay signal of Ade+. Fig. 3B shows the typicalsignal probed at 555 nm. We observed negative signals due to theinitial bleaching of FADH•. We can regroup all three signalsof LfH•*, Ade+, and LfH• into two dynamic types of transients(SI Text): one represents the summation of two parts (LfH•*and LfH•) with an excited-state decay time of 100 ps and its am-plitude is proportional to the difference of absorption coefficientsbetween the two parts. Because LfH• has a larger absorption co-efficient (eLfH•*< eLfH•), the signalflips and shows as a negative rise(Fig. 3B). The second-type transient reflects the summation of twoparts (Ade+ and LfH•) with a dynamic pattern of Ade+ in a rise and

Fig. 1. (A) Configuration of the FAD cofactor with four critical residues(N378, E363, W382, and W384 in green) in E. coli photolyase. The lumiflavin(Lf) (orange) and adenine (Ade) (cyan) moieties adopt an unusual bentconfiguration to ensure intramolecular ET within the cofactor. The N and Eresidues mutated to stabilize the FAD– state and the two W residues mu-tated to leave FAD and FADH• in a redox-inert environment are indicated.(B) The four redox states of FAD and their corresponding absorption spectra.

Fig. 2. Femtosecond-resolved intramolecular ET dynamics between theexcited oxidized Lf and Ade moieties. (A–C) Normalized transient-absorp-tion signals of the W382F/W384F mutant in the oxidized state probed at 800,630, and 580 nm, respectively, with the decomposed dynamics of the re-actant (Lf*) and intermediate (Ade+). Inset shows the derived intramolecularET mechanism between the oxidized Lf* and Ade moieties.

Liu et al. PNAS | August 6, 2013 | vol. 110 | no. 32 | 12973

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

19,

202

0

decay behavior and similarly the signal flips due to the larger ab-sorption coefficient of FADH•.Kinetically, we observed an apparent rise in 20 ps and a decay in

85 ps. Fig. 3C shows that, when the transient is probed at 530 nm,the ground-state LfH• recovery in 85 ps dominates the signal.Thus, the observed dynamics in 20 ps reflects the back ET processand the signal manifests as apparent reverse kinetics, leading toless accumulation of the intermediate state. Here, the charge re-combination in 20 ps is much faster than the charge separation in135 ps with a driving force of −1.88 eV in the Marcus invertedregion. In summary, although the neutral FAD* and FADH•*states can draw an electron from a strong reductant and the dimersubstrate can be repaired by a strong oxidant (22) by donating anelectron to induce cationic dimer splitting, the ultrafast cyclic ETdynamics with theAdemoiety in themutants reported here or withthe neighboring tryptophans in the wild type (23, 24) exclude thesetwo neutral redox states as the functional state in photolyase.

ET from Anionic Semiquinoid Lumiflavin (Lf–) to Adenine. In photo-lyase, FAD– cannot be stabilized and is readily converted toFADH• through proton transfer from the neighboring residuesor trapped water molecules in the active site. However, in type 1insect cryptochromes, the flavin cofactor can stay in FAD– invitro under anaerobic condition and this anionic semiquinone wasalso proposed to be the active state in vivo (14, 15). By examiningthe sequence alignment and X-ray structures (25, 26) of these twoproteins, the key difference is one residue near the N5 atom of theLf moiety, N378 in E. coli photolyase and C416 in Drosophilacryptochrome. Through structured water molecules, the N378 isconnected to a surface-exposed E363 in the photolyase but C416is connected to the hydrophobic L401 in the cryptochrome. Thus,we prepared a double-position photolyase mutant E363L/N378Cto mimic the critical position near the N5 atom in the crypto-chrome. With a higher pH 9 and in the presence of the thyminedimer substrate at the active site to push water molecules out ofthe pocket to reduce local proton donors, we were able to suc-cessfully stabilize FAD– in the mutant for more than several hoursunder anaerobic condition. Fig. 4 shows the absorption transientsof excited FAD– probed at three wavelengths. At 650 nm (Fig.4A), the transient shows a decay dynamics in 12 ps (τ = 12 ps andβ = 0.97) without any fast component or long plateau. We also didnot observe any measurable thymine dimer repair and thus ex-clude ET from FAD–* to the dimer substrate (SI Text).The radical Lf–* probably has a lifetime in hundreds of pico-

seconds as observed in insect cryptochrome (15), also similar to thelifetime of the radical LfH•* (19). Thus, the observed dynamics in12 ps must result from an intramolecular ET from Lf–* to Ade toform the Lf+Ade– pair. Such an ET reaction also has a favorabledriving force (ΔG0 = −0.28 eV) with the reduction potentials ofAde/Ade– and Lf/Lf– to be −2.5 and −0.3 V vs. NHE (20, 27),respectively. The observed initial ultrafast decay dynamics ofFAD–* in insect cryptochromes in several to tens of picoseconds, inaddition to the long lifetime component in hundreds of pico-seconds, could be from an intramolecular ET with Ade as well asthe ultrafast deactivation by a butterfly bending motion througha conical intersection (15, 19) due to the large plasticity of cryp-tochrome (28). However, photolyase is relatively rigid, and thusthe ET dynamics here shows a single exponential decay witha more defined configuration.Similarly, we tuned the probe wavelengths to the blue side to

probe the intermediate states of Lf and Ade– and minimize thetotal contribution of the excited-state decay components. Around350 nm, we detected a significant intermediate signal with a rise in2 ps and a decay in 12 ps. The signal flips to the negative ab-sorption due to the larger ground-state Lf– absorption. Strikingly,at 348 nm (Fig. 4C), we observed a positive component with theexcited-state dynamic behavior (eLf–* > eLf–) and a flippednegative component with a rise and decay dynamic profile (eLf+eAde– < eLf–). Clearly, the observed 2 ps dynamics reflects theback ET dynamics and the intermediate signal with a slow for-mation and a fast decay appears as apparent reverse kineticsagain. This observation is significant and explains why we did notobserve any noticeable thymine dimer repair due to the ultrafastback ET to close redox cycle and thus prevent further electrontunneling to damaged DNA to induce dimer splitting. Thus, inwild-type photolyase, the ultrafast cyclic ET dynamics determinesthat FAD– cannot be the functional state even though it can donateone electron. The ultrafast back ET dynamics with the interveningAdemoiety completely eliminates further electron tunneling to thedimer substrate. Also, this observation explains why photolyase usesfully reduced FADH– as the catalytic cofactor rather than FAD–

even though FAD– can be readily reduced from the oxidized FAD.

ET from Anionic Hydroquinoid Lumiflavin (LfH–) to Adenine. Pre-viously, we reported the total lifetime of 1.3 ns for FADH–* (2).Because the free-energy change ΔG0 for ET from fully reduced

Fig. 3. Femtosecond-resolved intramolecular ET dynamics between theexcited neutral semiquinoid Lf and Ade moieties. (A–C) Normalized tran-sient-absorption signals of the W382F mutant in the neutral semiquinoidstate probed at 800, 555, and 530 nm, respectively, with the decomposeddynamics of two groups: one represents the excited-state (LfH•*) dynamicbehavior with the amplitude proportional to the difference of absorptioncoefficients between LfH•* and LfH•; the other gives the intermediate (Ade+)dynamic behavior with the amplitude proportional to the difference ofabsorption coefficients between Ade+ and LfH•. Inset shows the derivedintramolecular ET mechanism between the neutral LfH•* and Ade moieties.For the weak signal probed at 555 nm, a long component (∼20%) was re-moved for clarity and this component could be from the product(s) resultingfrom the excited state due to the short lifetime of 230 ps.

12974 | www.pnas.org/cgi/doi/10.1073/pnas.1311077110 Liu et al.

Dow

nloa

ded

by g

uest

on

June

19,

202

0

LfH–* to adenine is about +0.04 eV (5, 21), the ET dynamicscould occur on a long timescale. We observed that the fluores-cence and absorption transients all show the excited-state decaydynamics in 1.3 ns (Fig. 5A, τ = 1.2 ns and β = 0.90). Similarly, weneeded to tune the probe wavelengths to maximize the in-termediate absorption and minimize the contributions of excited-state dynamic behaviors. According to our previous studies (4, 5),at around 270 nm both the excited and ground states have similarabsorption coefficients. Fig. 5 B and C show the transients probedaround 270 nm, revealing that the intermediate LfH• signal ispositive (eLfH•+eAde– > eLfH–+eAde) and dominant. Similarly, weobserved an apparent reverse kinetics with a rise in 25 ps anda decay in 1.3 ns. With the N378C mutant, we reported the life-time of FADH–* as 3.6 ns (4) and taking this value as the lifetimewithout ET with the Ade moiety, we obtain the forward ET timeas 2 ns. Thus, the rise dynamics in 25 ps reflects the back ET andthis process is ultrafast, much faster than the forward ET. Thisobservation is significant and indicated that the ET from the co-factor to the dimer substrate in 250 ps does not follow the hopping

mechanism with two tunneling steps from the cofactor to adenineand then to dimer substrate. Due to the favorable driving force,the electron directly tunnels from the cofactor to dimer substrateand on the tunneling pathway the intervening Ade moiety medi-ates the ET dynamics to speed up the ET reaction in the first stepof repair (5).

Unusual Bent Configuration, Intrinsic ET, and Unique Functional State.With various mutations, we have found that the intramolecularET between the flavin and the Ade moiety always occurs with thebent configuration in all four different redox states of photolyaseand cryptochrome. The bent flavin structure in the active site isunusual among all flavoproteins. In other flavoproteins, the flavincofactor mostly is in an open, stretched configuration, and if any,the ET dynamics would be longer than the lifetime due to the longseparation distance. We have found that the Ademoiety mediatesthe initial ET dynamics in repair of damaged DNA using thisunusual bent structure (5, 29). Currently, it is not known whetherthe bent structure has a functional role in cryptochrome. If theactive state is FAD– in type 1 insect cryptochromes or FADH– in

Fig. 4. Femtosecond-resolved intramolecular ET dynamics between theexcited anionic semiquinoid Lf and Ade moieties. (A–C) Normalized tran-sient-absorption signals of the E363L/N378C mutant in the anionic semi-quinoid state probed at 650, 350, and 348 nm, respectively, with thedecomposed dynamics of two groups: one exhibits the excited-state (Lf–*)dynamic behavior with the amplitude proportional to the difference ofabsorption coefficients between Lf–* and Lf–; the other has the intermediate(Lf or Ade–) dynamic behavior with the amplitude proportional to the dif-ference of absorption coefficients between (Lf+Ade–) and Lf–. Inset showsthe derived intramolecular ET mechanism between the anionic Lf–* and Ademoieties.

Fig. 5. Femtosecond-resolved intramolecular ET dynamics between theexcited anionic hydroquinoid Lf and Ade moieties. (A–C) Normalized tran-sient-absorption signals in the anionic hydroquinoid state probed at 800,270, and 269 nm with the decomposed dynamics of two groups: one rep-resents the excited-state (LfH–*) dynamic behavior with the amplitude pro-portional to the difference of absorption coefficients between LfH–* andLfH–; the other reflects the intermediate (LfH• or Ade–) dynamic behaviorwith the amplitude proportional to the difference of absorption coefficientsbetween (LfH•+Ade–) and (LfH–+Ade). Inset shows the derived intramolecularET mechanism between the anionic LfH–* and Ade moieties.

Liu et al. PNAS | August 6, 2013 | vol. 110 | no. 32 | 12975

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

19,

202

0

plant cryptochrome, then the intramolecular ET dynamics withthe Ade moiety could be significant due to the charge relocationto cause an electrostatic change, even though the back ET couldbe ultrafast, and such a sudden variation could induce local con-formation changes to form the initial signaling state. Conversely,if the active state is FAD, the ET dynamics in the wild type ofcryptochrome is ultrafast at about 1 ps with the neighboringtryptophan(s) and the charge recombination is in tens of pico-seconds (15). Such ultrafast change in electrostatics could besimilar to the variation induced by the intramolecular ET of FAD–*

or FADH–*. Thus, the unusual bent configuration assures an“intrinsic” intramolecular ET within the cofactor to induce a largeelectrostatic variation for local conformation changes in crypto-chrome, which may imply its functional role.We believe the findings reported here explain why the active

state of flavin in photolyase is FADH–. With the unusual bentconfiguration, the intrinsic ET dynamics determines the onlychoice of the active state to be FADH–, not FAD–, due to themuch slower intramolecular ET dynamics within the cofactor inthe former (2 ns) than in the latter (12 ps), although both anionicredox states could donate one electron to the dimer substrate.With the neutral redox states of FAD and FADH•, the ET dy-namics are ultrafast with the neighboring aromatic tryptophan(s)even though the dimer substrate could donate one electron tothe neutral cofactor, but the ET dynamics is not favorable, beingmuch slower than those with the tryptophans or the Ade moiety.Thus, the only active state for photolyase is anionic hydroqui-none FADH– with an unusual, bent configuration due to theunique dynamics of the slower intramolecular ET (2 ns) in thecofactor and the faster intermolecular ET (250 ps) with the di-mer substrate (4). These intrinsic intramolecular cyclic ET dy-namics in the four redox states are summarized in Fig. 6A.

Energetics of ET in Photolyase Analyzed by Marcus Theory. The in-trinsic intramolecular ET dynamics in the unusual bent cofactorconfiguration with four different redox states all follow a singleexponential decay with a slightly stretched behavior (β = 0.90–0.97) due to the compact juxtaposition of the flavin and Ademoieties in FAD. Thus, these ET dynamics are weakly coupledwith local protein relaxations. With the cyclic forward and backET rates, we can use the semiempirical Marcus ET theory (30) as

treated in the preceding paper (16) and evaluate the drivingforces (ΔG0) and reorganization energies (λ) for the ET reac-tions of the four redox states. Because no significant conforma-tion variation in the active site for different redox states isobserved (31), we assume that all ET reactions have the similarelectronic coupling constant of J = 12 meV as reported for theoxidized state (16). With assumption that the reorganizationenergy of the back ET is larger than that of the forward ET, wesolved the driving force and reorganization energy of each ETstep and the results are shown in Fig. 6B with a 2D contourplot. The driving forces of all forward ET fall in the regionbetween +0.04 and −0.28 eV, whereas the corresponding backET is in the range from −1.88 to −2.52 eV. The reorganizationenergy of the forward ET varies from 0.88 to 1.10 eV, whereasthe back ET acquires a larger value from 1.11 to 1.64 eV. Thesevalues are consistent with our previous findings about the re-organization energy of flavin-involved ET in photolyase (5),which is mainly contributed by the distortion of the flavin co-factor during ET (close to 1 eV). All forward ET steps fall in theMarcus normal region due to their small driving forces and all ofthe back ET processes are in the Marcus inverted region. Notethat the back ET dynamics of the anionic cofactors (2 and 4 inFig. 6B) have noticeably larger reorganization energies thanthose with the neutral flavins probably because different high-frequency vibrational energy is involved in different back ETs.Overall, the ET dynamics are controlled by both free-energychange and reorganization energy as shown in Fig. 6B. Theactive site of photolyase modulates both factors to control theET dynamics of charge separation and recombination orcharge relocations in every redox state.

ConclusionWe reported here our direct observation of intramolecular ETbetween the Lf and Ade moieties with an unusual bent configu-ration of the flavin cofactor in photolyase in four different redoxstates using femtosecond spectroscopy and site-direct mutagen-esis. Upon blue-light excitation, the neutral oxidized and semi-quinone lumiflavins can be photoreduced by accepting an electronfrom the Ade moiety (or neighboring aromatic tryptophans),while the anionic semiquinone and hydroquinone lumiflavins canreduce the Ade moiety by donating an electron. After the initial

Fig. 6. Summary of the molecular mechanisms and dynamics of cyclic intramolecular ET between the Lf and Ade moieties of photolyase in the four differentredox states and their dependence on driving forces and reorganization energies. (A) Reaction times and mechanisms of the cyclic ET between the Lf and Ademoieties in all four redox states. (B) Two-dimensional contour plot of the ET times relative to free energy (ΔG0) and reorganization energy (λ) for all electrontunneling steps. All forward ET reactions are in the Marcus normal region (−ΔG0 < λ), whereas all back ET steps are in the Marcus inverted region (−ΔG0 > λ).

12976 | www.pnas.org/cgi/doi/10.1073/pnas.1311077110 Liu et al.

Dow

nloa

ded

by g

uest

on

June

19,

202

0

charge separation or relocation, all back ET dynamics occur ul-trafast in less than 100 ps to close the photoinduced redox cycle.Strikingly, in contrast to the oxidized state, all other three back ETdynamics are much faster than their forward ET processes,leading to less accumulation of the intermediate state. To capturethe intermediate states, it is necessary to find an appropriateprobing wavelength to cancel out the contributions from both theexcited state (positive signal) and ground state (negative signal),leaving the weak intermediate signal dominant.The intramolecular ET dynamics in the four redox states with

the bent cofactor configuration reveal the molecular origin of theactive state in photolyase and imply a universal ET model forboth photolyase and cryptochrome. To repair damaged DNA inphotolyase, the ET must be from the anionic flavin cofactor andthe intramolecular ET dynamics unambiguously reveal that onlythe FADH– as the active state rather than FAD– due to the in-trinsically slower ET (2 ns) in the former and faster ET (12 ps) inthe latter, allowing a feasible, relatively fast, ET (250 ps) to thedamaged-DNA substrate from FADH– with the intervening Ademoiety in the middle to mediate such initial electron tunnelingfor repair. In cryptochrome, either neutral FAD and FADH• oranionic FAD– and FADH– can proceed to an ET dynamics uponblue-light excitation. For the former, the ET with the neigh-boring aromatic tryptophans occur in 1 and 45 ps or with the Ademoiety in 19 and 135 ps, and for the latter, the ET with the Ademoiety occur in 12 ps and 2 ns, respectively. All back ET dy-namics occur within 100 ps. Such ET dynamics induce an elec-trostatic variation in the active site, leading to local conformationchanges to form the initial signaling state. A unified ET mech-anism for both photolyase and cryptochrome would imply that ananionic redox form is more attractive as a functional state incryptochrome. Further studies are needed, however, to understandthe signaling mechanism(s) of photosensory cryptochrome.

Materials and MethodsPhotolyase Mutants and Their Redox States. The preparation of His-tag fusedE. coli photolyase E109A mutant (EcPL) has been described before (32). Basedon this template, we mutated two tryptophans of W382 and W384 near theflavin and prepared this double mutant in the oxidized form (FAD). For theanionic semiquinone (FAD–), wemutated two critical positions in EcPL of E363Land N378C. After purification, we obtained the mutant protein in completelyoxidized state. Before ultrafast experiments, the mutant enzyme of a concen-tration of 100 μM was exchanged into a basic reaction buffer at pH 9 with50 mM Tris·HCl, 300 mM NaCl, 1 mM EDTA, 20 mM DTT, 1 mM oligo-(dT)15containing cyclobutane thymine dimers, and 50% (vol/vol) glycerol. Afterpurgewith argon and irradiationwithUV light at 365nm (UVP; 8W), theflavincofactor is stabilized at the FAD– state under anaerobic conditions. The neutralsemiquinone (FADH•) EcPL was prepared by mutation of W382F in EcPL andthe anionic hydroquinone (FADH–) EcPL was stabilized under anaerobic con-ditions after purge with argon and subsequent photoreduction.

Femtosecond Absorption Spectroscopy. All of the femtosecond-resolvedmeasurements were carried out using the transient-absorption method. Theexperimental layout has been detailed previously (24). Enzyme preparationswith oxidized (FAD) and anionic semiquinone (FAD–) flavin were excited at480 nm. For enzyme with neutral semiquinone (FADH•), the pump wave-length was set at 640 nm. For the anionic hydroquinone (FADH–) form of theenzyme, we used 400 nm as the excitation wavelength. The probe wave-lengths were tuned to cover a wide range of wavelengths from 800 to260 nm. The instrument time resolution is about 250 fs and all of theexperiments were done at themagic angle (54.7°). Samples were kept stirringduring irradiation to avoid heating and photobleaching. Experiments withthe neutral FAD and FADH• states were carried out under aerobic conditions,whereas those with the anionic FAD– and FADH– states were executed underanaerobic conditions. All experiments were performed in quartz cuvetteswith a 5-mm optical length except that the FADH– experiments probed at 270and 269 nm were carried out in quartz cuvettes with a 1-mm optical length.

ACKNOWLEDGMENTS. This work is supported in part by National Institutesof Health Grants GM074813 and GM31082, the Camille Dreyfus Teacher–Scholar (to D.Z.), the American Heart Association fellowship (to Z.L.), andThe Ohio State University Pelotonia fellowship (to C.T. and J.L.).

1. Sancar A (2003) Structure and function of DNA photolyase and cryptochrome blue-

light photoreceptors. Chem Rev 103(6):2203–2237.2. Kao Y-T, Saxena C, Wang L, Sancar A, Zhong D (2005) Direct observation of thymine

dimer repair in DNA by photolyase. Proc Natl Acad Sci USA 102(45):16128–16132.3. Li J, et al. (2010) Dynamics and mechanism of repair of ultraviolet-induced (6-4)

photoproduct by photolyase. Nature 466(7308):887–890.4. Liu Z, et al. (2011) Dynamics and mechanism of cyclobutane pyrimidine dimer repair

by DNA photolyase. Proc Natl Acad Sci USA 108(36):14831–14836.5. Liu Z, et al. (2012) Electron tunneling pathways and role of adenine in repair of cy-

clobutane pyrimidine dimer by DNA photolyase. J Am Chem Soc 134(19):8104–8114.6. Liu HT, et al. (2008) Photoexcited CRY2 interacts with CIB1 to regulate transcription

and floral initiation in Arabidopsis. Science 322(5907):1535–1539.7. Liu B, Liu HT, Zhong D, Lin CT (2010) Searching for a photocycle of the cryptochrome

photoreceptors. Curr Opin Plant Biol 13(5):578–586.8. Sancar A (2004) Regulation of the mammalian circadian clock by cryptochrome. J Biol

Chem 279(33):34079–34082.9. Ozturk N, Selby CP, Annayev Y, Zhong D, Sancar A (2011) Reaction mechanism of

Drosophila cryptochrome. Proc Natl Acad Sci USA 108(2):516–521.10. Gegear RJ, Foley LE, Casselman A, Reppert SM (2010) Animal cryptochromes me-

diate magnetoreception by an unconventional photochemical mechanism. Nature

463(7282):804–807.11. Bouly JP, et al. (2007) Cryptochrome blue light photoreceptors are activated through

interconversion of flavin redox states. J Biol Chem 282(13):9383–9391.12. Lin CT, et al. (1995) Association of flavin adenine dinucleotide with the Arabidopsis

blue light receptor CRY1. Science 269(5226):968–970.13. Malhotra K, Kim S-T, Batschauer A, Dawut L, Sancar A (1995) Putative blue-light

photoreceptors from Arabidopsis thaliana and Sinapis alba with a high degree of

sequence homology to DNA photolyase contain the two photolyase cofactors but lack

DNA repair activity. Biochemistry 34(20):6892–6899.14. Song SH, et al. (2007) Formation and function of flavin anion radical in cryptochrome

1 blue-light photoreceptor of monarch butterfly. J Biol Chem 282(24):17608–17612.15. Kao Y-T, et al. (2008) Ultrafast dynamics and anionic active states of the flavin co-

factor in cryptochrome and photolyase. J Am Chem Soc 130(24):7695–7701.16. Liu Z, et al. (2013) Determining complete electron flow in the cofactor photoreduc-

tion of oxidized photolyase. Proc Natl Acad Sci USA 110:12966–12971.17. Chang C-W, et al. (2010) Ultrafast solvation dynamics at binding and active sites of

photolyases. Proc Natl Acad Sci USA 107(7):2914–2919.

18. Byrdin M, Eker APM, Vos MH, Brettel K (2003) Dissection of the triple tryptophanelectron transfer chain in Escherichia coli DNA photolyase: Trp382 is the primarydonor in photoactivation. Proc Natl Acad Sci USA 100(15):8676–8681.

19. Kao Y-T, et al. (2008) Ultrafast dynamics of flavins in five redox states. J Am ChemSoc 130(39):13132–13139.

20. Seidel CAM, Schulz A, Sauer MHM (1996) Nucleobase-specific quenching of fluores-cent dyes. 1. Nucleobase one-electron redox potentials and their correlation withstatic and dynamic quenching efficiencies. J Phys Chem 100(13):5541–5553.

21. Gindt YM, Schelvis JPM, Thoren KL, Huang TH (2005) Substrate binding modulates thereduction potential of DNA photolyase. J Am Chem Soc 127(30):10472–10473.

22. Vicic DA, et al. (2000) Oxidative repair of a thymine dimer in DNA from a distance bya covalently linked organic intercalator. J Am Chem Soc 122(36):8603–8611.

23. Byrdin M, et al. (2010) Quantum yield measurements of short-lived photoactivationintermediates in DNA photolyase: Toward a detailed understanding of the tripletryptophan electron transfer chain. J Phys Chem A 114(9):3207–3214.

24. Saxena C, Sancar A, Zhong D (2004) Femtosecond dynamics of DNA photolyase: En-ergy transfer of antenna initiation and electron transfer of cofactor reduction. J PhysChem B 108(46):18026–18033.

25. Park HW, Kim ST, Sancar A, Deisenhofer J (1995) Crystal structure of DNA photolyasefrom Escherichia coli. Science 268(5219):1866–1872.

26. Zoltowski BD, et al. (2011) Structure of full-length Drosophila cryptochrome. Nature480(7377):396–399.

27. Balland V, Byrdin M, Eker APM, Ahmad M, Brettel K (2009) What makes the differ-ence between a cryptochrome and DNA photolyase? A spectroelectrochemical com-parison of the flavin redox transitions. J Am Chem Soc 131(2):426–427.

28. Partch CL, Clarkson MW, Ozgür S, Lee AL, Sancar A (2005) Role of structural plasticityin signal transduction by the cryptochrome blue-light photoreceptor. Biochemistry44(10):3795–3805.

29. Antony J, Medvedev DM, Stuchebrukhov AA (2000) Theoretical study of electrontransfer between the photolyase catalytic cofactor FADH− and DNA thymine dimer. JAm Chem Soc 122(6):1057–1065.

30. Page CC, Moser CC, Chen XX, Dutton PL (1999) Natural engineering principles ofelectron tunnelling in biological oxidation-reduction. Nature 402(6757):47–52.

31. Maul MJ, et al. (2008) Crystal structure and mechanism of a DNA (6-4) photolyase.Angew Chem Int Ed Engl 47(52):10076–10080.

32. Li J, Uchida T, Todo T, Kitagawa T (2006) Similarities and differences between cy-clobutane pyrimidine dimer photolyase and (6-4) photolyase as revealed by reso-nance Raman spectroscopy: Electron transfer from the FAD cofactor to ultraviolet-damaged DNA. J Biol Chem 281(35):25551–25559.

Liu et al. PNAS | August 6, 2013 | vol. 110 | no. 32 | 12977

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

June

19,

202

0