Photoreduction of the Folate Cofactor in Members of the Photolyase Family*

Cryptochromes and DNA photolyases are related flavoproteins with flavin adenine dinucleotide as the common cofactor. Whereas photolyases repair DNA lesions caused by UV radiation, cryptochromes generally lack repair activity but act as UV-A/blue light photoreceptors. Two distinct electron transfer (ET) pathways have been identified in DNA photolyases. One pathway uses within its catalytic cycle, light-driven electron transfer from FADH−* to the DNA lesion and electron back-transfer to semireduced FADHo after photoproduct cleavage. This cyclic ET pathway seems to be unique for the photolyase subfamily. The second ET pathway mediates photoreduction of semireduced or fully oxidized FAD via a triad of aromatic residues that is conserved in photolyases and cryptochromes. The 5,10-methenyltetrahydrofolate (5,10-methenylTHF) antenna cofactor in members of the photolyase family is bleached upon light excitation. This process has been described as photodecomposition of 5,10-methenylTHF. We show that photobleaching of 5,10-methenylTHF in Arabidopsis cry3, a member of the cryptochrome DASH family, with repair activity for cyclobutane pyrimidine dimer lesions in single-stranded DNA and in Escherichia coli photolyase results from reduction of 5,10-methenylTHF to 5,10-methyleneTHF that requires the intact tryptophan triad. Thus, a third ET pathway exists in members of the photolyase family that remained undiscovered so far.

DNA photolyases and cryptochromes (cry) 2 form a large family of related flavoproteins with DNA repair activity and photoreceptor function, respectively. Members of this protein family were identified in all kingdoms of life and can be grouped in at least nine subclades (1). DNA photolyases repair cytotoxic and mutagenic DNA lesions that are formed during exposure of DNA to UV-B. These DNA lesions are cyclobutane pyrimidine dimers (CPDs) or pyrimidine-pyrimidone  photoproducts. According to their substrate specificity, DNA photolyases are designated as CPD photolyases or  photolyases (2). The repair of both types of DNA lesions by photolyase requires the catalytic fully reduced and anionic flavin cofactor FADH Ϫ that, when photoexcited, injects an electron directly into the DNA lesion (1) as shown in Fig. 1A (electron transfer pathway 1). During extraction from the cell and purification under aerobic conditions the flavin cofactor is usually oxidized to the semireduced and eventually to the fully oxidized form. Reduction of these flavin species to FADH Ϫ in vitro can be achieved by illumination of the enzyme in the presence of reducing agents such as dithiothreitol or ␤-mercaptoethanol. This process is named photoactivation (1). Photoactivation in vitro requires photoexcitation of the flavin and a triad of redox-active residues in the protein moiety that is highly conserved in DNA photolyases (3,4) as shown in Fig. 1A (electron transfer pathway 2). These residues are generally tryptophans that allow transport of an electron from the protein surface to the U-shaped flavin cofactor, which is buried within the C-terminal ␣-helical domain (5)(6)(7)(8)(9). Whether the same mechanism is used by photolyase to photoreduce FAD in vivo is a matter of debate (10). Photoreduction of the flavin cofactor was also observed in cryptochrome blue/UV-A photoreceptors. However, instead of fully reduced flavin, semireduced flavin species (either anionic flavin semiquinone radical or neutral semiquinone radical) accumulate. This form of the photoreceptor is considered as the signaling state (11)(12)(13)(14).
A recently discovered subclade of the DNA photolyase/cryptochrome family are DASH cryptochromes, which have members in plants, bacteria, and aquatic animals (6,(15)(16)(17). Because DASH cryptochromes were found to lack repair activity for CPDs in double-stranded DNA, they were considered as cryptochrome-type photoreceptors (6,16). However, it was recently shown that DASH cryptochromes repair CPDs in singlestranded DNA (18) and loop structures of double-stranded DNA (19) and, thus, belong to the CPD photolyase group. In contrast to conventional CPD photolyases, DASH cryptochromes are unable to flip the CPD lesion out of the DNA duplex (7).
Besides the flavin cofactor that is essential for enzymatic activity, DNA photolyases and most likely all cryptochromes contain a second chromophore (1). Like the catalytic flavin, the second chromophore is non-covalently attached to the protein moiety. The majority of DNA photolyases and, as far as studied, the cryptochromes including the DASH-type like cry3 from Arabidopsis thaliana contain polyglutamated 5,10-methenyltetrahydrofolate (5,10-methenylTHF) as the second chromophore (1,12,17,20,21) (see Fig. 1B for folate structures). Several organisms like the cyanobacterium Anacystis nidulans (Synechococcus elongatus) produce deazariboflavins (7,8didemethyl-8-hydroxy-5-deazariboflavin) and utilize them as second cofactor (22). In photolyases of thermophilic bacteria and Archaea of the genus Sulfolobus, FMN and FAD, respectively, were found as second cofactors (23,24). The sole function of the second cofactors demonstrated at present is transfer of excitation energy to the catalytic flavin cofactor via a Förstertype mechanism. The crystal structures of DNA photolyases and DASH cryptochromes revealed that the second chromophores are located in a cleft between the N-terminal ␣/␤ domain and the C-terminal ␣ domain (7)(8)(9). The centroid distances between the catalytic FAD and the second chomophore are in the range of 15-18 Å. The close distances and the angles between the transition dipole moments of the two cofactors are favorable for efficient energy transfer. Indeed, energy transfer efficiencies are about 70% for Escherichia coli photolyase (25), close to 100% for A. nidulans photolyase (26), and between 78% (dark-adapted) and 87% (light-adapted) for Arabidopsis cry3 (27). Although the second cofactors are not essential for catalysis (28,29), they increase the efficiency of repair and possibly of photoactivation by having higher extinction coefficients than FADH Ϫ in the near UV and blue region (30). The spectral overlap between 5,10-methenylTHF emission and the absorption of the different flavin redox states is on the order Illumination in vitro of photolyase that contains fully oxidized or semireduced flavin results in light-induced absorbance changes. The decrease in absorption in the 450 -470-nm region reflects a decrease in the amount of fully oxidized FAD concomitant with transient increase in absorption above 500 nm, which indicates the formation of a neutral semiquinone radical. Excitation of the 5,10-methenylTHF antenna chromophore at its absorption peak at 380 nm causes a likewise photoreduction of the catalytic FAD (1,27,28,30,31). However, irreversible bleaching of the 380-nm peak is observed under high irradiance UV-A or camera flash illumination (28,30). This irreversible bleaching goes along with release of the folate cofactor from the protein moiety (30) and was named photodecomposition of 5,10-methenylTHF (28). However, the identity of the formed folate species remained unknown (30). In our previous spectroscopic characterization of Arabidopsis cry3, a similar bleaching of the 380-nm peak was observed (27).
Here we show that a third electron transfer pathway exists in photolyase and DASH cryptochome, where the 5,10-methenylTHF cofactor is photoreduced to 5,10-methylene-THF. Thus, bleaching at 380 nm does not simply reflect destruction but is a specific chemical conversion of the second chromophore.
Expression and Purification of cry3-Expression and purification of proteins were done under red light conditions as described (21). However, the size-exclusion chromatography step was omitted, and the final NaCl concentration was adjusted to 200 mM during the concentrating step on Amicon ultra centrifugal filter concentrator with a 30-kDa cutoff (Millipore).
Spectroscopic Studies-UV-visible absorption spectra of purified cry3 wild type and mutant proteins were recorded using a 2-channel UV-2401 PC spectrophotometer (Shimadzu). For direct comparison of cry3 wild type and mutants, absorption cross-section spectra were calculated (8). Photoreduction of fully oxidized FAD by blue light and UV-A illumination and photobleaching of 5,10-methenylTHF by UV-A treatment in the protein samples were followed by recording difference spectra during illumination at 15°C as already described (8). Interference filters (Schott) were used for monochromatic radiation with blue light (450 Ϯ 6 nm; 50 mol m Ϫ2 s Ϫ1 ) or UV-A (386 Ϯ 5 nm; 112 mol m Ϫ2 s Ϫ1 ). Kinetics were plotted for the 450-nm absorption peak of fully oxidized FAD and for the 384-nm absorption peak of 5,10-methenylTHF. The data were fitted by mono-exponential decay curves. Excitation and emission spectra of cofactors released from cry3 wild type and mutants after trichloroacetic acid precipitation of the protein (27) and after trichloroacetic acid treatment of cofactor standards, respectively, were recorded in 50 mM sodium phos-phate buffer, pH 7.0, containing 200 mM NaCl, 10 mM ␤-mercaptoethanol, and 10% (v/v) glycerol using an RF-5301 PC spectrofluorophotometer (Shimadzu) in a three-window glass fluorescence cuvette (Hellma GmbH & Co. KG).
Cofactor Determination by HPLC and ESI-MS-The cofactor composition and the ratios of cofactor to protein were determined by HPLC reverse phase chromatography analysis as already described (27). Cofactors were separated and eluted from the column using a continuous linear gradient of methanol (60 -80%), and separation was monitored by absorption at 360 nm (5,10-methenylTHF and oxidized flavins), at 450 nm (only oxidized flavins), and at 295 nm (oxidized flavins and all folates except 5,10-methenylTHF). For comparison, 5 nmol of each standard were processed analogously. For ESI-MS analysis cofactor standards (5,10-methenylTHF and 5,10-methylene-THF, both monoglutamated) were separated by liquid chromatography, and mass spectra were recorded online by ESI-MS as described below. A 36 M E. coli photolyase protein solution was treated with sodium borohydride (final concentration 50 mM) to reduce the polyglutamated 5,10-methenylTHF and release the formed 5-methylTHF (29). The enzyme completely depleted of folate by chromatography on a Sephadex PD-10 (GE-Healthcare) desalting column was reconstituted with monoglutamated 5,10-methenylTHF by incubating 10 nmol of the enzyme with a 10-fold molar excess of 5,10-methenylTHF at 10°C for 75 min. The incorporation of 5,10-methenylTHF into E. coli photolyase was monitored spectroscopically and seen by the increase in absorbance at 380 nm. After removal of unbound 5,10-methenylTHF by repeated concentration and dilution of the protein solution in 500 l of Vivaspin ultrafiltration spin columns (Sartorius Stedim Biotech), the reconstituted protein solution (43 M) was used for UV-A (386 Ϯ 5 nm, 120 mol m Ϫ2 s Ϫ1 ) treatment followed by recording the absorption spectra. For liquid chromatography-MS measurements to 100 l of the E. coli photolyase protein solutions (concentration, 48 M for reconstituted dark sample and 43 M for reconstituted UV-A irradiated sample), 1 ml acetonitrile was added. After centrifugation for 15 min (18,000 ϫ g), the supernatants containing the extracted cofactors were dried by lyophilization. The pellet was redissolved in 50 l of H 2 O and subjected to liquid chromatography-MS analysis. For chromatography, a 125/2 Nucleodur C18ec column (Macherey-Nagel) was utilized with an Agilent 1100 chromatography system (Agilent). The following gradient was applied at a flow rate of 0.2 ml/min and a column temperature of 25°C with buffer A (H 2 O, 0.05% formic acid) and buffer B (acetonitrile, 0.045% formic acid), holding B for 5 min at 2% followed by a linear increase of B to 10% within an additional 5 min and to 30% within 15 min. Finally, buffer B was increased to 95% B within 2 min. Mass spectra were recorded online by ESI-MS using a LTQ-FT mass spectrometer (ThermoFinnigan). High resolution mass spectra were obtained using the ICR detection cell, and MSMS data were generated in the linear ion trap.
Enzymatic Characterization of Folate Species in cry3 and E. coli Photolyase-5,10-MethyleneTHF dehydrogenase/5,10-methenylTHF cyclohydrolase was purified from E. coli cells (BL21 (DE3)) following established procedures (32). In brief, the enzyme was purified from total soluble protein extracts by protamine sulfate (2%) and ammonium sulfate (35, 55, and 70%) precipitation in 50 mM Tris-HCl, pH 7.5, 50 mM ␤-mercaptoethanol followed by heparin and anion exchange chromatography in the same buffer system with 1 M KCl for elution. The specificity of the purified enzyme for the two substrates, NADP ϩ and 5,10-methyleneTHF, was tested and is shown in supplemental Fig. S1B. The release of the cofactors from cry3 or E. coli photolyase by trichloroacetic acid precipitation, heat treatment, or incubation of the sample in 20% acetonitrile (as used for reverse phase chromatography analysis) inhibited the enzyme assay. Thus, the cofactors were released from the protein moiety by incubation in imidazole (33), resulting in intact cofactor (supplemental Fig. S1A). Thereto the protein sample was incubated at 4°C for 5 h in the dark in 50 mM sodium phosphate buffer, pH 7.5, 200 mM NaCl, 10 mM ␤-mercaptoethanol, and 10% (v/v) glycerol containing 500 mM imidazole. After removal of precipitated proteins by centrifugation (20,000 ϫ g, 4°C, 10 min) the released cofactors were separated from the protein by a concentrating step on an Amicon ultra centrifugal filter concentrator with 30-kDa cut-off (Millipore) and tested in the enzyme assay. Cofactor standard solutions were treated and analyzed the same way, but the concentrating step was omitted (supplemental Fig. S1C). The conversion of 5,10-methyleneTHF to 5,10-methenylTHF by the dehydrogenase activity was traced spectroscopically by the increase in 340-nm absorbance caused by the formation of NADPH (supplemental Fig. S1D) (34). Unless otherwise stated, the standard enzyme assay mixture contained 0.2 mM NADP ϩ , 50 mM sodium phosphate, pH 7.5, 200 mM NaCl, 500 mM imidazole, 10 mM ␤-mercaptoethanol, and 10% (v/v) glycerol in a final volume of 130 l. The mixture was prepared without NADP ϩ ; after the addition of the enzyme solution, the mixture was incubated for 5 min at 20°C. Afterward, NADP ϩ was added to the mixture to start the reaction. For spectroscopic measurements the reference contained the same enzyme assay mixture of released cofactors or cofactor standards like the measured sample except that no NADP ϩ was added.
Structural Analysis of Photobleached cry3⅐T5 Cocrystals-cry3 was cocrystallized with a thymine dimer-comprising oligonucleotide (T-TϽϾT-T-T) as previously described (19). Bleaching of cry3⅐T5 cocrystals and recording of fluorescence and UV-visible spectra was achieved at the Cryobench, European Synchrotron Radiation Facility. X-ray data were collected for photobleached cry3 cocrystals at beamline ID14-1, European Synchrotron Radiation Facility and processed and scaled by XDS and XSCALE (35,36). Refinement of the cry3 structure utilized COOT and REFMAC5 (37) (supplemental Table 1).

RESULTS
Irreversible photobleaching of the 5,10-methenylTHF antenna chromophore has been described before for E. coli DNA photolyase and was named photodecomposition (28,29). However, the identity of the formed product(s) remained elusive (30). In cry3 we have observed the same bleaching of the 5,10-methenylTHF cofactor when the protein was illuminated with UV-A (27). Two tyrosine residues (Tyr-429, Tyr-423) are positioned between the two cofactors in cry3 (8). The distances of the phenyl rings of Tyr-423 and Tyr-429 from the centroid of the FAD isoalloaxine ring are 8.45 and 9.80 Å, respectively. Tyr-423 is within H-bonding distance of the N1 and N8 nitrogen atoms of 5,10-methenylTHF. Trp-409, the central tryptophan of the tryptophan triad, is 7.21 Å away from Tyr-429, and the distance from Tyr-429 to Tyr-423 is 6.44 Å (see Fig. 1A for a schematic illustration of steric order of the residues in cry3). We, therefore, reasoned that the light-induced redox changes of FAD could be coupled to changes in side-chain orientation, protonation, and/or redox state of these two aromatic residues as well as of 5,10-methenylTHF itself (8). To test for a role of the tryptophan triad on photobleaching of 5,10-methenylTHF, we performed site-directed mutagenesis of Trp-356 (at the protein surface) to phenylalanine, which is considered to be redox inactive in the process of FAD photoreduction within DNA photolyases (4,38). Absorption cross-section spectra indicate that the W356F mutant contains both cofactors, 5,10-methenylTHF and FAD, close to stoichiometric amounts (supplemental Fig.  S2A). The strong decrease of absorption in the region between 450 and 470 nm (caused by photoreduction of FAD ox ) seen for wild type cry3 during blue light irradiation ( Fig. 2A and C) completely lacks the W356F mutant (Fig. 2, B and C), whose spectra after illumination (dashed curve) and after the same period in the dark (dotted curve) overlap well. In general these results demonstrate an essential role of the tryptophan triad for FAD photoreduction in cry3 in vitro as expected. Illumination with UV-A (386 nm) causes a strong and irreversible decrease in the 380-nm peak in wild type cry3 (Fig. 2, D and F) as described before (27). The 380-nm peak originates mostly from 5,10-methenylTHF (27). The bathochromic shift of the absorption maximum of 5,10-methenylTHF in solution (360 nm) and associated to the protein was found before in E. coli photolyase and cry3 (20,27). In contrast to wild type, the absorption peak of fully oxidized FAD increases in the W356F mutant upon UV-A treatment, and the final decrease in 380-nm absorption is very small (Fig. 2, E and F). This increase in the amount of FAD ox in the W356F mutant under UV-A irradiation can be explained by electron donation to 5,10-methenylTHF by residual FADH Ϫ but a lack of photoreduction of the flavin caused by the interrupted tryptophan triad (see below). It should be noted that the increase of the 450-nm absorption peak of FAD ox is accompanied with an increase in absorption at 380 nm (⑀ 443 ϭ 11,140 M Ϫ1 cm Ϫ1 , ⑀ 380 ϭ 8,850 M Ϫ1 cm Ϫ1 ; values for FAD ox in cry3) and shown for cry3 in supplemental Fig. S2B. The extinction coefficients of FADH . and FADH Ϫ at 380 nm are in the range of 4.49 ϫ 10 3 M Ϫ1 cm Ϫ1 . Had no photobleaching of 5,10-methenyl-THF occurred in the W356F mutant under UV-A irradiation, one would have expected some increase in the 380-nm absorption (caused by the formation of FAD ox ) instead of the detectable slight decrease. Thus, the minor change seen in 380-nm absorbance (Fig. 2E) between the original spectrum (solid curve) and the spectrum after UV-A irradiation (dashed curve) signifies less change in the 5,10-methenylTHF content than actually occurred. However, the contribution of 5,10-methenylTHF (⑀ 380 ϭ 25,400 M Ϫ1 cm Ϫ1 ) in 380-nm absorbance is at least 2.8 higher that that of FAD. Taking into account the contribution of the various FAD redox states to absorbance at 380 nm, it is evident that photobleaching of 5,10-methenylTHF under UV-A irradiation is strongly reduced in the W356F mutant compared with wild type. The amount of 5,10-methenyl-THF after UV-A irradiation is reduced by 64% for wild type and by 20% for the W356F mutant (Fig. 2F). Similar results were obtained for another mutant in the tryptophan triad (W432F; data not shown). Thus, an intact tryptophan triad is required for efficient photobleaching of 5,10-methenylTHF in cry3.
Photobleaching of 5,10-methenylTHF could result from stepwise reduction to 5-methylTHF via 5,10-methyleneTHF as observed before with chemical reduction using sodium borohydride (30,39,40) (see Fig. 1B for folate structures). Because the photodecomposition products of 5,10-methenylTHF have not been identified unambiguously in either E. coli photolyase or any other member of the photolyase/cryptochrome family, we further analyzed the folate cofactors of cry3 and of E. coli DNA photolyase before and after photobleaching.
We performed reverse phase chromatography (RPC) of the folates released from cry3 under acidic conditions. To assign the cry3 cofactors after RPC, the standards RF, FAD, 5,10-methenylTHF, 5,10-methyleneTHF, and THF were used (Fig.  3C). The retention times of the cofactors released from the cry3 dark sample are very similar to those of standard FAD and 5,10-methenylTHF (Fig. 3A). In contrast, the UV-A-irradiated sample showed two novel peaks with prominent absorption at 295 nm besides the 5,10-methenylTHF peak. One peak eluted faster (peak 1) and the other one slower (peak 2) than 5,10-methenyl-THF (Fig. 3B). Peak 1 can be assigned to THF, whereas peak 2 most likely corresponds to 5,10-methyleneTHF. The shift in retention times between the standard folates (THF, 5,10-methenylTHF, 5,10-methyleneTHF) and the folates released from UV-A-irradiated cry3 (peak 1 and peak 2) is presumably caused by the fact that the standard folates were monoglutamated in contrast to the folates of cry3 that are polyglutamated (data not shown).
In general, the analysis of folates by RPC has the problem of pH-dependent non-enzymatic interconversions of some folate species (41,42). 5-MethylTHF is relatively stable over a broad range of pH values (42). In contrast, 5,10-methenylTHF is mostly stable under acidic conditions (as used in our experiments), whereas it is converted to 5-formylTHF or 10-formylTHF at neutral and alkaline pH (42). 10-FormylTHF as well as 10-formiminoTHF are cyclized back to 5,10-methenyl-THF under acidic conditions (41). 5,10-MethyleneTHF is unstable under acidic conditions and converted to THF and formaldehyde (41). However, this conversion of 5,10-methylene- THF can be diminished by the addition of ␤-mercaptoethanol (39). Indeed, the presence of ␤-mercaptoethanol causes a reduction of the novel folate peak 1 and an increase of peak 2 (Fig. 3D) similar to stabilization of the 5,10-methyleneTHF standard by ␤-mercaptoethanol (Fig. 3E). Thus, RPC analysis suggests, as the fluorescence excitation and emission spectra shown in supplemental Fig. S3, that cry3 converts 5,10-methenyl-THF to 5,10-methyleneTHF under UV-A irradiation.
The molecular mass of 5,10-methyleneTHF is 1 Da higher than of 5,10-methenylTHF (see Fig. 1B). Thus, we used ESI-MS analysis to check for the expected mass change of the folate cofactor after UV-A treatment. MALDI-TOF MS analysis has shown that the folate cofactor in cry3 is polyglutamated with two to seven glutamate residues (data not shown). This polyglutamation strongly complicates the mass determination by ESI-MS. Thus, sodium borohydride treatment of the enzyme was performed to reduce and release the folate and reconstitute it with the monoglutamated 5,10-methenylTHF as described before for E. coli photolyase (29,30). However, sodium borohydride treatment of cry3 led only to partial release of the folate (data not shown), in contrast to E. coli photolyase that was treated exactly the same (Fig. 4A). Folate-depleted and reconstituted E. coli photolyase showed the 380-nm peak (Fig. 4A) typical for protein-bound 5,10-methenylTHF. Therefore, we  Because of protonation of 5,10-methyleneTHF during ESI-MS performed in positive ion mode, a mass difference of 2 Da compared with 5,10-methenylTHF is seen in ESI mass spectra. The folate released from E. coli photolyase that was not exposed to UV-A-treated sample showed a peak at 456.1626 Da (Fig. 4, C and E), matching exactly the mass of monoglutamated 5,10-methenylTHF, whereas the UV-A-treated sample had a prominent peak at 458.1779 Da (Fig. 4, D and F) that corresponds exactly to the mass of the 5,10-methyleneTHF standard besides residual detection of 5,10-methenylTHF in the same sample (Fig. 4D). Gas phase fragmentation spectra of the ESI-MS analyzed folates released from E. coli photolyase (Fig. 4, E and F, lower panels) correspond to gas phase fragmentation spectra of standard monoglutamated 5,10-methenylTHF and 5,10-methyleneTHF (data not shown). Therefore, the data from the mass analysis are in strong favor of UV-A-driven formation of 5,10-methyleneTHF formation in E. coli photolyase.
When cocrystals of cry3 and a T5 oligonucleotide containing a synthetic CPD-like lesion (19) were exposed to UV-A at 180 K, a decrease in absorption at 380 nm together with the accumulation of semiquinoid and probably also fully reduced FAD was observed (Fig.  5A). In agreement with FAD photoreduction, the fluorescence emission signal of fully oxidized FAD at 520 nm progressively decreased upon 355-nm laser excitation (Fig.  5B). The initial strong fluorescence emission of 5,10-methenylTHF at 440 -460 nm seen for cry3 in solution is decreased in the crystal. This suggests enhancement of energy transfer from 5,10-methenylTHF to FAD in the crystal that could be caused by the crystalline nature of the probe, the bound substrate, or the low temperature during spectroscopy. Together these data indicate that essentially the same photochemistry takes place in solution as in cry3 crystals; however, with better energy transfer from 5,10-methenylTHF to FAD in the crystal. X-ray diffraction analysis of a single UV-A-treated cry3 cocrystal revealed that the methine bond between N5 of the pteridine moiety and N10 of the para-aminobenzoate moiety remained intact (Fig. 5C, supplemental Table 1). This data set precludes the formation of 5-methylTHF (or any other ring opening process) as a product of 5,10-methenylTHF photobleaching in cry3 and supports our above statement that 5,10-methyleneTHF could have been formed.
To exclude the possibility that the lack of NADPH formation seen for the dark cry3 sample resulted from the presence of compounds that inhibit the 5,10-methyleneTHF dehydrogenase activity, we added 5,10-methyleneTHF to both reaction mixtures. A similar rate of NADPH formation was seen for the cry3 samples kept in darkness or irradiated with UV-A (Fig. 6B) when the reaction mixture were supplemented with 5,10-methyleneTHF. This result excludes that compounds in the dark sample inhibited the enzymatic reaction. In line with the above-described formation of NADPH seen specifically in the reaction assay with cofactors from UV-A-treated cry3 are the difference spectra shown in Fig. 6, C and D. Only the assay mixture that contains the cofactors of UV-A-treated cry3 (Fig.  6C) shows a decrease at 260 nm (peak absorbance of NADP ϩ and NADPH) concomitant with an increase in absorbance at 340 nm (peak absorbance of NADPH but no absorbance of NADP ϩ ). These spectral changes reflect the conversion of NADP ϩ to NADPH that is not seen for the dark control (Fig.  6D). The decrease at around 295 nm in Fig. 6C reflects the decrease of 5,10-methyleneTHF (⑀ 294 nm 25,000 M Ϫ1 cm Ϫ1 ).  Fig. S1) followed by concentration of the cofactors. A, cofactors released from UV-A-irradiated cry3 (dashed curve) or dark sample (solid curve) were incubated with 5,10-methyleneTHF dehydrogenase. After 5 min of preincubation, NADP ϩ was added to a final concentration of 0.2 mM, and changes in absorbance at 340 nm were monitored continuously to trace the formation of NADPH. For control, the same reaction was performed with standard 0.1 mM 5,10-methenylTHF (short dashed curve) and 0.1 mM 5,10-methyleneTHF (dotted curve). B, control reactions where 5,10-methyleneTHF was added to a final concentration of 0.09 mM to the reaction mixture of cofactors from the dark control (solid curve) and the UV-A-treated cry3 sample (dashed curve). C and D, difference spectra of reaction mixtures taken before (ϪNADP ϩ ) and after (ϩNADP ϩ ) the addition of NADP ϩ . In C spectra are shown for the UV-A-treated sample and in D for the dark control.
The data shown above demonstrate that cry3 photoreduces 5,10-methenylTHF to 5,10-methyleneTHF, whereas 5,10-methyleneTHF dehydrogenase catalyzes the opposite reaction. The equilibrium of the reaction catalyzed by 5,10-methylene-THF dehydrogenase can be shifted toward 5,10-methenylTHF and NADPH by removing 5,10-methenylTHF from the equilibrium under high NADP ϩ /NADPH ratios. We reasoned that cry3 lacking its folate cofactor could deplete the reaction mixture of 5,10-methenylTHF by incorporating it in its folate binding pocket. To test this hypothesis, we performed coupled enzyme tests with UV-A-treated cry3 in the presence of 5,10-methyleneTHF dehydrogenase and NADP ϩ . Indeed, we observed an increase in the 380-nm peak (Fig. 7A) that is completely missing when 5,10-methyleneTHF dehydrogenase was omitted (Fig. 7B). The recovery of the 380-nm peak depends specifically on both the presence of 5,10-methyleneTHF dehydrogenase and NADP ϩ as shown in Fig. 7C. The dark reversion of the flavin cofactor to fully oxidized FAD did not depend on the presence of 5,10-methyleneTHF dehydrogenase and/or NADP ϩ , as expected. Because E. coli 5,10-methyleneTHF dehydrogenase is a bifunctional enzyme that has also cyclohydrolase activity (32,34), some of the formed 5,10-methenylTHF is probably also converted to 10-formylTHF (see Equation 2). However, the cyclohydrolase activity is much lower than the dehydrogenase activity (32), thus removing only a small fraction of 5,10-methenylTHF from the equilibrium between cry3 and 5,10-methyleneTHF dehydrogenase. Furthermore, cry3 is able to bind 10-formylTHF and cyclize it within its binding site to 5,10-methenylTHF (data not shown), as already reported for E. coli DNA photolyase (30).
As outlined above, irreversible photobleaching of the 5,10-methenylTHF cofactor has been described before for E. coli photolyase (28 -30). When comparing UV-A-driven photobleaching of 5,10-methenylTHF in cry3 and E. coli photolyase, we observed that this process is even faster in E. coli photolyase than in cry3 (Fig. 8A). Likewise the photoreduction of FAD under blue light irradiation is faster in E. coli photolyase than in cry3 (Fig. 8B). As shown in Fig. 4 by mass spectrometry, E. coli photolyase forms 5,10-methyleneTHF when treated with UV-A. This result is confirmed by the data shown in Fig. 8C. UV-A-pretreated E. coli photolyase builds up the 380-nm peak when incubated with 5,10-methyleneTHF dehydrogenase in the presence of NADP ϩ . In the absence of 5,10-methyleneTHF dehydrogenase and NADP ϩ , this increase in 380-nm absorbance is not detectable (Fig. 8D). When UV-A treatment of E. coli photolyase or cry3 is done in the presence of 5,10-methyleneTHF dehydrogenase and NADP ϩ , the decrease in 380-nm absorbance is strongly reduced (Fig. 8E) or completely lacking (Fig. 8F), respectively. The continuous increase in absorbance at 340 nm reflects the formation of NADPH. The enzymatic coupling of UV-A-driven 5,10-methyleneTHF formation catalyzed by cry3 or E. coli photolyase and the reverse reaction catalyzed by 5,10-methyleneTHF dehydrogenase is schematically shown in Fig. 8G.

DISCUSSION
The mechanisms by which DNA photolyases repair CPDs and the (6-4) photoproduct are well established. Biochemical and structural analyses have proven that the only cofactor in photolyase essential for catalysis is FAD (1,44). For both CPD photolyase and (6-4) photolyase, the flavin cofactor is needed in its fully reduced and anionic form (FADH Ϫ ) to allow single electron transfer from its light-activated form (FADH Ϫ *) to the UV lesion (25,45,46). Cocrystal structures of DNA photolyases with bound substrate show distances between FAD and the UV lesion close enough for efficient electron transfer ( Fig. 1) (47,48). The absorption of FADH Ϫ peaks in the UV-B and UV-A region at around 350 nm. The radiation of sunlight that reaches the surface of the earth is, however, pretty low in this waveband region (ϳ1.5% of solar energy). It seems that all photolyases can enhance photorepair by carrying a second chromophore with extinction coefficients in this waveband region that are much higher than that of FADH Ϫ . These antenna chromophores transfer excitation energy to the catalytic flavin cofactor because there is spectral overlap between flavin absorption and antenna fluorescence (31). The nature of the second chromophore that is used by photolyase is specific for each species. Organisms that can synthesize deazariboflavins such as the cyanobacterium A. nidulans use 8-hydroxy-5-deazariboflavin (22), thermophilic bacteria use FMN (23), and Archaea use FAD (24). However, photolyases of most organisms contain 5,10-methenylTHF as antenna pigment (1,2,44). The energy transfer efficiencies between the antenna chromophore and the catalytic flavin are very high and were determined to be in the range between 70% for E. coli (5,10-methenylTHF antenna) and close to 100% for A. nidulans photolyase (8-hydroxy-5-deazariboflavin antenna). In cry3 this energy transfer efficiency was calculated to be between 78 and 87% (27). Thus, the role of 5,10-methenylTHF as antenna chromophore in photolyase is well established. Former studies on E. coli photolyase have shown that its 5,10-methenylTHF cofactor shows bleaching under high light and camera flash irradiation that is irreversible, in contrast to the photoreduction of the catalytic flavin cofactor that is accompanied by a decrease in absorption in the 450 -470-nm range (28,29). This bleaching was considered as photodecomposition of 5,10-methenylTHF as a result of ring opening between N5 of the pteridine and N10 of the p-aminobenzoate moiety concomitant with the release of formaldehyde that is eventually oxidized to formic acid (Ref. 30; see Fig. 1B for folate structures). We have observed the same photobleaching of 5,10-methenylTHF in cry3 (Fig. 2) and confirmed that the light conditions used for these studies were such that the 5,10-methenylTHF chromophore in E. coli photolyase was also bleached and at even higher rates than in cry3 (Fig. 8A). The conserved tryptophan triad in cry3 is not only required for the photoreduction of FAD but also for efficient photobleaching of 5,10-methenylTHF, as shown in Fig. 2. These observations are in line with previous studies on E. coli photolyase showing essentially no photobleaching of the folate cofactor in the W306F mutant and in the absence of reducing agents (38).
Our spectroscopic, biochemical, mass spectroscopic, and structural analyses together show unequivocally that cry3 and E. coli photolyase form 5,10-methyleneTHF during UV-A irradiation. These data are in conflict with the previously published ones for E. coli photolyase that suggested the release of a C1 unit (formaldehyde or formic acid) in the process of photodecomposition (30). A simple explanation for this discrepancy could be the fact that in the previous study the folate was released from the protein under acidic conditions. This might have caused rapid conversion of 5,10-methyleneTHF to THF and formaldehyde as 5,10-methyleneTHF is known to be unstable under acidic conditions (49). Indeed, we have also detected THF after RPC separation of the chromophores released from UV-A-irradiated cry3 (Fig. 3). However, in the presence of ␤-mercaptoethanol as stabilizer, we identified 5,10-methylene-THF using the same RPC separation technique (Fig. 3).
The conversion of 5,10-methenylTHF to 5,10-methylene-THF requires the transfer of two electrons and one proton. As outlined above, we found that the tryptophan triad, which is conserved essentially in all members of the photolyase/crypto- Half-lives of the 382-nm absorbing 5,10-methenylTHF were calculated as described before (8). The protein samples did not contain the same ratios of flavin redox states after purification. Therefore, the protein samples were pre-illuminated with blue light (450 Ϯ 6 nm; 95 mol m Ϫ2 s Ϫ1 , 15°C, 60 min) to reach the full content of fully reduced FAD before UV-A treatment. UV-A illumination led to a decrease in 382 nm absorption with a half-life t1 ⁄2 of 63 Ϯ 1 min for cry3 and t1 ⁄2 of 7 Ϯ 0.2 min for E. coli photolyase. B, photoreduction of oxidized FAD by blue light illumination. The protein samples were stored at 4°C in darkness to gain oxidized FAD in the samples before blue light illumination. Absorbance changes of cry3 (closed squares) and chrome family, is required for efficient photoreduction of 5,10-methenylTHF. This result favors the previously proposed model (38) that fully reduced FADH Ϫ is required and is possibly the obligatory electron donor for photoreduction of 5,10-methenylTHF. In line with such a model are studies on E. coli photolyase that contained only the 5,10-methenylTHF cofactor. This study showed that photodecomposition of 5,10-methenylTHF was 5-fold reduced when flavin was missing (31). The residual photobleaching of 5,10-methenylTHF in the flavin-free E. coli photolyase indicates an alternative pathway for electron transfer to 5,10-methenylTHF during photoreduction that is independent of FADH Ϫ . We exclude, however, that the two tyrosines (Tyr-423, Tyr-429) that are sandwiched in cry3 between the two chromophores could be involved in such an alternative electron pathway for two reasons; first, these tyrosines are not conserved in E. coli photolyase, although photobleaching in E. coli photolyase is faster than in cry3 (Fig. 8A). Second, mutation of both tyrosines to phenylalanine in cry3 led to the loss of 5,10-methenylTHF binding during purification, but the enzyme could be reconstituted with 5,10-methenyl-THF. The reconstituted enzyme showed photobleaching very similar to the wild type (data not shown).
Likewise it is yet not clear whether photoreduction of 5,10-methenylTHF is initiated by photoexcitation of the antenna chromophore, of FADH Ϫ , or of both chromophores as the absorption peak of 5,10-methenylTHF in cry3 (and other photolyases) is at 380 nm, where FADH Ϫ also absorbs. The midpoint redox potential of the 5,10-methenylTHF/5,10-methylene-THF couple is Ϫ300 mV (50). The estimated redox potential for the FAD/FAD . pair is Ϫ153 mV for Arabidopsis cry1 and Ϫ48 mV for the FADH . /FADH Ϫ pair in E. coli photolyase (51). Upon photoexcitation, FADH Ϫ * represents an extremely powerful reductant with a redox potential that is estimated as low as Ϫ2.8 V (52, 53) and sufficient to inject an electron via electron transfer pathway 1 onto CPD lesions (redox potential about Ϫ2.2 V (52, 54)) ( Fig. 1). This again supports the notion that FADH Ϫ * is the most likely electron donor for 5,10-methenylTHF reduction along pathway 3 (Fig. 1). However, the reduction of 5,10-methenylTHF to 5,10-methyleneTHF requires a second electron. The origin of this second electron remains elusive and could be stripped from the protein matrix. Consistent with the model that FADH Ϫ donates an electron to 5,10-methenylTHF is our observation that the bleaching of 5,10-methenylTHF in cry3 is decelerated in the presence of TϽϾT containing singlestranded DNA, the substrate of cry3 (data not shown). In this situation a competition for electron transfer exists where FADH Ϫ can donate the electron either to the CPD lesion or to 5,10-methenylTHF. Reduced photobleaching of 5,10-methenyl-THF in the presence of CPD-containing substrate was also found for E. coli photolyase (31). Together our results demonstrate that cry3 and other photolyases are able to photoreduce 5,10-methenylTHF using the photoexcited FADH Ϫ * cofactor.
Folate (vitamin B), in particular the THF species 10-formylTHF, 5,10-methenylTHF, 5,10-methyleneTHF, and 5-methylTHF, have a crucial role in the C1 metabolism. They are needed for the formation of formylmethionyl-tRNA, purines, thymidylate, and pantothenate, respectively (55,56). 5-MethylTHF is the direct precursor of methionine, which can be converted to S-adenosyl methionine (57), the methyl group donor for synthesis of polyamines and of the phytohormone ethylene; S-adenosyl methionine is also required for the methylation of proteins and DNA. Furthermore, the biosynthesis of chlorophylls, lignins, and other compounds depends on S-adenosyl methionine (57). It is intriguing to speculate that cry3 could play a role in the folate pathway by catalyzing the formation of 5,10-methyleneTHF in a light-dependent step. Indeed, it is known that the synthesis of folates is up-regulated in dividing tissues such as meristems (58) and in developing seedlings that synthesize chlorophylls (59).
Light-induced expression of early enzymes in THF synthesis such as dihydropterin pyrophosphokinase-dihydropterate synthase was described before (58). We have detected a significant up-regulation of the cry3 transcript and protein levels in Arabidopsis seedlings that are controlled by phytochrome A (data not shown). The requirement of light for both enhanced cry3 expression and catalytic activity would fit with a model where cry3 plays a role in seedling development in the transition phase between dark and light.
Considering the redundant pathways by which 5,10-methyleneTHF can be formed in plants, it is likely that the blocking of one pathway (e.g. by mutation) can be compensated at least in part by the others. There are three established pathways for the formation of 5,10-methyleneTHF in plants; that is, via the C1-THF synthase pathway, by glycine decarboxylase, or by serine hydroxymethyltransferase (56). In plants, the C1-THF synthase pathway depends on a bifunctional enzyme with 5,10-methenylTHF cyclohydrolase (EC 3.5.4.9) and 5,10-methylene-THF dehydrogenase (EC 1.5.1.5) activity (60) and a 10-formylTHF synthetase (EC 6.3.4.3) (60). C1-THF synthase isoenzymes were found in the cytosol, chloroplasts, and mitochondria (61). The mitochondrial glycine decarboxylase multienzyme system (EC 1.4.4.7, 1.8.1.4, 2.1.2.10) catalyzes the irreversible conversion of two glycines into one serine, an important step in photorespiration (62). Serine hydroxymethyltransferase (EC 2.1.2.1) catalyzes the reversible conversion of glycine into serine (62). Isoenzymes of serine hydroxymethyltransferase are also present in the cytosol, chloroplasts, and mitochondria. The data we present here show that cry3 possibly constitutes a fourth pathway in 5,10-methylene-THF formation. The unique feature of cry3 is, however, its dependence on light energy for the reduction of 5,10-methenyl-THF. The extent to which cry3 contributes to the formation of 5,10-methyleneTHF in vivo remains unclear. Based on previous in vitro spectroscopic studies, the quantum efficiency for the photobleaching of 5,10-methenylTHF in cry3 was found to be in the range of 3.7 ϫ 10 Ϫ5 (27) and, thus, is quite low. Therefore, it is possible that cry3 contributes to 5,10-methyleneTHF formation only under very high irradiance conditions. Furthermore, under standard conditions the midpoint redox potential of the NADP ϩ /NADPH couple is Ϫ320 mV and of the 5,10-methenylTHF/5,10-methyleneTHF couple is Ϫ300 mV (50). Thus, there is an equilibrium between NADPH formation and 5,10-methyleneTHF consumption that depends on the NADP ϩ /NADPH ratio.
Another aspect of our studies is the discovery that the lightdriven formation of 5,10-methyleneTHF by cry3 and E. coli photolyase can be coupled with the formation of NADPH in the presence of 5,10-methyleneTHF dehydrogenase (Fig. 8G). Because cry3 is located in chloroplasts and mitochondria (16), where 5,10-methyleneTHF dehydrogenase activity is present, these compartments might be able to form NADPH in a lightdependent way but independent of photosynthesis.