A Novel Photoreaction Mechanism for the Circadian Blue Light Photoreceptor Drosophila Cryptochrome*

Cryptochromes are flavoproteins that are evolutionary related to the DNA photolyases but lack DNA repair activity. Drosophila cryptochrome (dCRY) is a blue light photoreceptor that is involved in the synchronization of the circadian clock with the environmental light-dark cycle. Until now, spectroscopic and structural studies on this and other animal cryptochromes have largely been hampered by difficulties in their recombinant expression. We have therefore established an expression and purification scheme that enables us to purify mg amounts of monomeric dCRY from Sf21 insect cell cultures. Using UV-visible spectroscopy, mass spectrometry, and reversed phase high pressure liquid chromatography, we show that insect cell-purified dCRY contains flavin adenine dinucleotide in its oxidized state (FADox) and residual amounts of methenyltetrahydrofolate. Upon blue light irradiation, dCRY undergoes a reversible absorption change, which is assigned to the conversion of FADox to the red anionic \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{FAD}^{{\bar{{\cdot}}}}\) \end{document} radical. Our findings lead us to propose a novel photoreaction mechanism for dCRY, in which FADox corresponds to the ground state, whereas the \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{FAD}^{{\bar{{\cdot}}}}\) \end{document} radical represents the light-activated state that mediates resetting of the Drosophila circadian clock.

Cryptochromes (CRYs) 4 constitute a family of flavoproteins that use flavin adenine dinucleotide (FAD) and sometimes an additional pterin derivative (methenyltetrahydrofolate, MTHF) as noncovalently bound cofactors and blue light absorbing chromophores (1). CRYs share moderate sequence, but significant structural homology with DNA photolyases, which repair UV-damaged DNA in a blue light dependent manner (2). Cyclobutane pyrimidine dimer (CPD) photolyases repair UV light-induced pyrimidine-dimer (PyrϽϾPyr) DNA lesions by intermolecular redox reactions between the catalytically active fully reduced flavin chromophore (FADH Ϫ ) and the PyrϽϾPyr substrate (2). A similar reaction mechanism is presumed for the (6-4)-photolyases, which repair pyrimidinepyrimidone  photoproducts (Pyr (6-4) Pyr), a second class of UV light-induced DNA lesions (3,4). Cryptochromes do not exhibit any DNA repair activities, despite significant sequence homology of plant cryptochromes to CPD-photolyases and animal cryptochromes to (6-4)-photolyases (5,6). Blue light absorption, phosphorylation, and effector interactions of plant cryptochromes control fundamental biological processes such as de-etiolation and flowering onset (7). Action spectra (8) and spectroscopic studies (9, 10) on Arabidopsis thaliana cryptochrome 1 (AtCRY1) suggest that its native and functionally active chromophore is oxidized FAD (FAD ox ), in contrast to photolyases, where the active chromophore is the two electronreduced FADH Ϫ . However, in both AtCRY1 (11,12) and photolyases (2), blue light activation leads to the formation of a neutral blue FADH ⅐ radical. Whereas in AtCRY1 FAD ox is photoreduced to FADH ⅐ upon blue light illumination, the FADH ⅐ radical in photolyases is produced after a blue lightactivated electron transfer from FADH Ϫ to the PyrϽϾPyr substrate.
Animal cryptochromes, either as photoreceptors or integral components of biological clocks, play crucial roles in the generation of 24-h (circadian) rhythms or their synchronization with the environmental light-dark cycle (6,13). Drosophila cryptochrome (dCRY) is a blue light photoreceptor mediating light synchronization of the circadian clock (14,15). The Drosophila clock is operated by a transcriptional and translational feedback loop in which the clock proteins Period and Timeless (dTIM) inhibit their own transcription by negatively regulating the transcription factors dClock and dCycle (16). dCRY resets the Drosophila clock by sequestering dTIM from the feedback loop through light-dependent dCRY-dTIM interactions (17), which trigger the proteasomal degradation of dTIM (18) and dCRY (19). The dCRY molecule comprises a chromophorebinding photolyase homology region (PHR) (514 amino acids) and a unique C-terminal extension (28 amino acids) referred to as "tail" region ( Fig. 1). Binding to its effector dTIM and light signaling to the clock is mediated by the PHR domain of dCRY. The tail region prevents the dCRY-dTIM interaction and stabilizes dCRY in the dark (20,21). A light-induced conformational change, possibly governed by intramolecular redox reactions, appears to be required to displace the tail region and to allow for dCRY-dTIM interactions and subsequent dTIM and dCRY degradation.
To date, no spectroscopic data on the blue light responses of dCRY or any other animal cryptochrome are available. To study the photochemistry of dCRY as a circadian blue light photoreceptor, we have established an expression and purification scheme that enables us to purify mg amounts of monomeric dCRY from Sf21 insect cell cultures. The chromophore content of the purified material is analyzed by UV-visible absorption and fluorescence spectroscopy, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS), and reversed phase high pressure liquid chromatography (RP-HPLC) analysis. We show that insect cell purified dCRY contains FAD in its oxidized state (FAD ox ) and residual amounts of MTHF. Intriguingly, dCRY undergoes a reversible absorption change upon blue light irradiation, which is assigned to the conversion of FAD ox to an anionic FAD . radical. These findings lead us to propose a novel photoreaction mechanism for dCRY, which is expected to have crucial implications for the resetting mechanism of the Drosophila circadian clock.

Recombinant Expression and Purification of dCRY
Virus Generation-Full-length Drosophila cryptochrome (dCRY 1-542) was heterologously expressed in Sf21 insect cells using the Bac-to-Bac System (Invitrogen). A dCRY fusion protein comprising an N-terminal His 6 tag followed by a tobacco etch virus protease cleavage site for removal of the hexahistidine moiety was subcloned into the pFastBac-HTb vector (Invitrogen) using the restriction sites NotI and XhoI. Tag removal yielded a recombinant dCRY protein with a 19-residue extension (GAMGSGIERPTSTSSLVAA) at the N terminus corresponding to a molecular mass of 64,155 Da. Recombinant bacmid DNA was generated after a transposition step in the Escherichia coli DH10Bac strain and isolated according to the manufacturer's protocols. For recombinant virus generation, 1 ϫ 10 6 Sf21 cells were transfected using Cellfectin reagent (Invitrogen). The parental virus was isolated 3 days after infection. After two more amplification steps, the P3 high titer stock with usually Ն1 ϫ 10 8 plaque-forming units/ml was obtained and used for subsequent large scale expressions.
Cell Culture and Recombinant Protein Expression-Sf21 cells were grown as suspension cultures at 27°C in TC100 complete medium supplemented with 10% (v/v) fetal calf serum (Invitrogen) and 1% (v/v) Pleuronic-F68 (Invitrogen). For expression in a 1 liter of culture, 2.5 ϫ 10 9 cells were infected with the third passage virus at an optimized multiplicity of infection and incu-bated at 60 rpm in Fernbach flasks. Optimal expression kinetics were established in pilot studies. The cells were centrifuged at 500 ϫ g, and the pellet was washed with ice-cold phosphatebuffered saline, resuspended in 30 ml of buffer A (50 mM NaH 2 PO 4 , pH 8.0, at 4°C, 5% (v/v) glycerol, 2 mM ␤-mercaptoethanol (␤-ME)), and snap-frozen in liquid nitrogen prior to storage at Ϫ20°C.
Purification of dCRY-For purification of dCRY, pellets from 5 liters of expression cultures were thawed on ice and homogeneously resuspended in buffer A supplemented with Complete EDTA-free protease inhibitor tabs (Roche Applied Science) and 1 mM phenylmethylsulfonyl fluoride. The cells were lysed using a Branson 450 sonifier equipped with a microtip, and the lysate was spun for 60 min at 100,000 ϫ g. The supernatant was loaded onto a DEAE-Sepharose column (Amersham Biosciences) preequilibrated with buffer A. The column was processed by applying a linear gradient ranging from 0 to 100% buffer B (buffer A ϩ 500 mM NaCl). dCRY-containing fractions were pooled, adjusted to 300 mM NaCl, and loaded on a nickelnitrilotriacetic acid-agarose column (Qiagen). dCRY was eluted in buffer C (50 mM Tris-HCl, pH 8.0, at 4°C, 300 mM NaCl, 5% glycerol (v/v), 2 mM ␤-ME) applying a linear gradient of 25-300 mM imidazol. Affinity purified dCRY was concentrated to 10 mg/ml according to Bradford assay (Bio-Rad) using an Amicon Ultra-15 filter device (Millipore, Bedford, MA) with a 30-kDa molecular mass cut-off (MWCO). The concentrated material was subjected to gel filtration on a HiLoad S200 16/60 column (GE Healthcare) using a buffer system containing 25 mM Tris, pH 8.0, at 4°C, 150 mM NaCl, 5% (v/v) glycerol and 2 mM ␤-ME. For analytical gel filtration runs, a HiLoad S200 10/30 (GE Healthcare) column was used with a running buffer containing 25 mM Tris, pH 8.0, at 4°C, 300 mM NaCl, 5% (v/v) glycerol, and 2 mM ␤-ME. Fractions containing highly purified and monomeric dCRY were pooled, concentrated to typically 13 mg/ml (200 M) and snap-frozen in liquid nitrogen. The samples were stored at Ϫ80°C until measured.

UV-visible Absorption and Fluorescence Spectroscopy
UV-visible absorption spectra of purified dCRY were recorded using an Uvikon 933 spectrophotometer (Kontron, Neufahrn, Germany) at room temperature with cuvettes of 1-cm path length. The dCRY sample was measured at a concentration of 13 mg/ml (200 M). Flavin fluorescence spectra were recorded on a RF-1501 Fluorimeter (Shimadzu, Duisburg, Germany) with a bandwidth of 10 nm using a 450-nm excitation wavelength for the emission spectra and a 530-nm emission wavelength for the excitation spectra. The detection of the folate chromophore was performed using a FluoroMax II Fluorimeter (Spex Industries, Edison, NJ) by recording emission spectra at an excitation wavelength of 380 nm and excitation spectra at an emission wavelength of 460 nm.

Analysis of dCRY-bound Chromophores by MALDI-TOF Mass Spectrometry
Purified dCRY was mixed in a 1:1 ratio (v/v) with a saturated ␣-cyano-4-hydroxy-cinnamicacid (CHCA; Sigma) MALDI matrix solution dissolved in 0.1% (v/v) trifluoroacetic acid and 50% (v/v) acetonitrile. 1 l of this protein-matrix mixture was spotted on a MALDI steel target plate (100-well plate; Applied Biosystems) and allowed to air dry. The measurements were carried out using a Voyager-DE TM MALDI-TOF-MS PRO Bio-Spectrometry TM work station (PerSpective Biosystems) with delayed extraction and a nitrogen laser (337.1 nm) in the linear mode. External mass calibration was performed using the c1 calibration mixture ( After calibration, mass analysis was performed with the Voyager TM 5.1 Software Data Explorer TM (Applied Biosystems).

RP-HPLC Analysis of dCRY-bound Chromophores
30-l samples of purified dCRY at various concentrations were heat-denatured for 3 min at 97°C and spun at 13 krpm for 5 min at 4°C. The supernatant containing the released chromophores was subjected to RP-HPLC analysis using the protocol described in Ref. 22 with a Waters (Toronto, Ontario, Canada) Chromatography System and a C18 reversed phase column (Hypersil-ODS, Bischoff Chromatography, Leonberg, Germany). Flavin chromophores and MTHF were followed by absorption at 370 nm. To quantify the chromophore loading, a 150 M dCRY solution was denatured and subjected to RP-HPLC analysis. The percentage of FAD loading was determined by comparing the integrated peak areas with an FAD calibration curve (50,150, and 200 M FAD). MTHF degradation products, which do not absorb at wavelengths Ͼ300 nm, were checked for by diode array detection.

Blue Light Illumination and Dark Recovery
Series of UV-visible absorption spectra were recorded using a Shimadzu UV 2401 spectrometer with a spectral bandwidth of 2 nm. The temperature of the sample was adjusted to 10°C by a circulating water bath. A light-emitting diode (Luxeon Star; Lumileds) with an emission maximum at 445 nm (14 nm full width at half-maximum) provided 19 milliwatts/cm 2 of blue light at the sample. dCRY was photoreduced by blue light illumination for 100 s. The spectrum of 100% product formation was calculated by subtracting the spectrum of the dark form from that obtained after 100 s of illumination. The limit of subtraction was reached when an untypical shoulder appeared at 500 nm in the product spectrum (23).
Time traces of dCRY dark recovery were recorded at 450 nm after a 10-s illumination with blue light. Fitting a single expo-nential curve to these experimental data did not yield satisfying results, showing that one time constant was not sufficient to describe the decay. Therefore, a biexponential function was used that resulted in an almost perfect match of the curves. The two obtained time constants describe the time until all but 1/e Ϸ 37% of the molecules recovered to the dark form.
dCRY protein was diluted in 25 mM Tris, pH 8.0, 150 mM NaCl, 5% glycerol (v/v), and 2 mM ␤-ME. Anaerobic conditions were produced by directing a stream of argon onto the solution for 40 min. ␤-ME was removed by passage through a Sephadex G 25 column (PD-10; Amersham Biosciences). Lyophilized glucose oxidase from Aspergillus niger (Sigma-Aldrich G 2133) was dissolved in 110 mM sodium phosphate, pH 9.2, 90 mM NaCl, and 9 mM EDTA. Possible traces of free chromophore were removed by ultrafiltration using a filter with a 10-kDa cut-off (Vivaspin 500; Vivasciences, Hannover, Germany). Glucose oxidase was rendered anaerobic and photoreduced with blue light (445-nm light-emitting diode, 19 milliwatts/cm 2 ) for a time period of 2 h.

Molecular Modeling of dCRY
Structural models of dCRY were created by the program Modeller (24) (27) and modified, when necessary, to avoid the disruption of secondary structures. Model structures were superimposed on template structures using the combinatorial extension method (28). Both ligands, FAD and MTHF, were placed in the corresponding binding pockets, yielding only few unfavorable contacts between the ligands and adjacent protein side chains. These unfavorable contacts were removed by a conformational search of the clashing residues finding the conformations with minimal interaction energy between them, the ligands, and the rest of the protein. The models were refined by a short minimization (150 steps) of complex energies with fixed positions of the ligands. Refinement was performed with the program CHARMm (29) using the default parameters for energy calculations.

Expression and Purification of dCRY from Sf21 Insect Cells-
Full-length Drosophila cryptochrome (dCRY ) fused to an N-terminal hexahistidine tag was recombinantly expressed and purified from Spodoptera frugiperda (Sf21) insect cells using a baculovirus expression vector system (Fig. 2). After cell lysis and centrifugation, the cleared lysate from typically 5 liters of insect cell expression culture was initially purified on a DEAE-Sepharose anion exchange column ( Fig. 2A). This purification step enhanced dCRY stability by removing (next to other contaminating proteins) proteases that otherwise led to proteolytic degradation of dCRY. dCRY-containing fractions, which were colored yellowish because of the presence of chromophore, were pooled and subjected to nickel-nitrilotriacetic acid affinity (Fig. 2B) and size exclusion chromatography (Fig. 2C). The described purification scheme resulted in overall yields of 3-5 mg of highly purified photoreceptor/ liter of cell culture (Fig. 2, C and D). The identity of the purified protein was confirmed by mass spectrometry (data not shown). Comparison with standard proteins suggests that purified dCRY behaves as a monomer under analytical gel filtration conditions (Fig. 2E).
Analysis of Chromophore Content-The yellow color of the purified dCRY protein (Fig. 2D) suggested that the isolated photoreceptor contains FAD and possibly MTHF, as present in other spectroscopically characterized cryptochrome/photolyase family members (2,5). We have analyzed the chromophore content of dCRY by absorption and fluorescence spectroscopy as well as MALDI-TOF mass spectrometry and RP-HPLC. Absorption spectra of freshly purified dCRY show maxima at 475, 450, 428, 374, and 360 nm, consistent with the presence of oxidized FAD (Fig. 3; see also Fig. 7A, dark form) (2). Because in photolyases MTHF strongly absorbs with a maximum at 380 -410 nm (30), the absorption spectrum of dCRY does not speak to the presence of significant MTHF amounts. The chromophore content was further analyzed by fluorescence spectroscopy. To detect the presence of the flavin moiety, we recorded excitation spectra for emission at 530 nm and emission spectra for excitation at 450 nm. Consistent with the presence of FAD, we observed two excitation maxima at 370 and 450 nm and a strong emission signal at 530 nm (Fig. 4A). However, the absence of any fine structure and the similarity of the dCRY fluorescence profile with that of free FAD suggests that the fluorescence signal might arise from small amounts of free FAD ox . It therefore appears that the dCRY-bound FAD ox evidenced by the fine structure of the absorption spectra (   because of a very short fluorescence life time. Based on the known cryptochrome and photolyase structures (25,31), FAD is expected to bind to dCRY in a U-shaped conformation. The U-shape might allow for a photoinduced electron transfer between the adenine and the isoalloxazine ring, thereby causing a short fluorescence life time as has been determined for cryptochrome 3 from A. thaliana (32). Additionally, electron transfer from adjacent amino acids to FAD is thought to contribute to fluorescence life time quenching (32). For detection of MTHF, emission spectra were recorded for excitation at 380 nm, and excitation spectra were recorded for emission at 460 nm. The emission spectra showed a strong peak at 520 nm and a small peak at 460 nm (Fig. 4B). Whereas the 520 nm peak again documents the presence of FAD, the peak at 460 nm is indicative of residual amounts of MTHF, as previously reported for Vibrio cholerae CRY2 (33). However, this peak was not observed in some other dCRY preparations, and moreover, no significant signal was obtained in excitation spectra for 460-nm emission. We conclude that at best a small proportion of the insect cell-purified dCRY contains MTHF.
To provide additional evidence for the presence of MTHF, we performed MALDI-TOF MS on the purified dCRY protein (Fig. 5)   cence signal obtained with our purified dCRY is therefore likely to be correlated with the weak fluorescence of its folic acid degradation product (34). Note that none of the observed masses correspond to polyglutamate species of MTHF or its degradation products folic acid and THF. Furthermore, no riboflavin (theoretical mass 377.37; [M ϩ H] ϩ ) is detected. However, based on our MALDI data, we cannot exclude the presence of FMN, whose theoretical mass of 459.3 Da [M ϩ H] ϩ is close to that of MTHF. The large peak at 380.04 Da is assigned to the CHCA matrix dimer. The remaining peaks could not be assigned.
To further analyze the nature and stoichiometry of the cofactors that are noncovalently bound to dCRY, we have heat-denatured the purified protein and subjected the chromophore containing supernatant obtained after a brief spinning step to RP-HPLC analysis (Fig. 6). The RP-HPLC analysis shows that dCRY contains FAD, but no riboflavin or FMN, providing further evidence that the MALDI peak at 458.94 Da (Fig. 5) is indeed due to the presence of MTHF. Comparison of the integrated peak areas of the FAD released from dCRY ( Fig. 6A) with an FAD calibration curve (not shown), revealed that 30 -35% of the purified dCRY are loaded with FAD. The HPLC elution profiles did not reveal MTHF or other folate-type chromophores, suggesting that these compounds are at best present in low amounts.
Blue Light Induced Conversion of Oxidized FAD to an Anionic Red FAD Radical-dCRY incorporating oxidized FAD was illuminated using a light-emitting diode blue light source with an emission maximum at 445 nm and an intensity of 19 milliwatts/cm 2 . Blue light irradiation resulted in a changed absorption spectrum with newly emerging maxima at 472, 403, and 367 nm and a weak but distinct absorption between 550 and 700 nm. Concomitantly, the FAD ox absorption bands at 475, 450, and 428 nm were reduced (Fig. 7A). Comparison with literature (23) and with spectra of photoreduced glucose oxidase (Fig. 7B) identifies the blue light-induced species as an anionic red FAD radical (FAD . ) (Fig. 7C). In contrast to CPD photolyases (2) and AtCRY1 (10,11), no neutral blue FADH ⅐ radical is observed. The neutral FADH ⅐ radical would be expected to display significant absorption maxima at or near 580 and 625 nm, as described e.g. for E. coli DNA photolyase (35). The weak and very broad absorption of blue light illuminated dCRY between 550 and 700 nm is not due to residual amounts of the neutral FADH ⅐ radical but rather represents a typical feature of anionic FAD . radicals (23). It is also present in the spectrum of the glucose oxidase radical, which exclusively forms the anionic form at the applied pH of 9.2 (23). Of note, with the exception of glucose oxidase, all other flavoproteins examined so far fall in two classes, forming either neutral or anionic flavin radicals upon photoreduction, irrespective of pH (23). This was also shown for dCRY by lowering the pH of the buffer from pH 8 to pH 7 and 6. Under these conditions, the light-induced difference spectra of dCRY strongly resemble those obtained at pH 8 (Fig. 7D), displaying only a slight increase in absorbance between 600 and 650 nm (data not shown). The presence of isosbestic points at 495, 414, 345, and 317 nm (Fig. 7A) shows that the reaction of dCRY occurs exclusively between the two species FAD ox and FAD . , i.e. no fully reduced FAD is formed in the reaction cycle. Because the proposed light activation mechanism of dCRY involves an electron transfer to the oxidized FAD (Fig. 7C), the reaction efficiency was analyzed in the presence and absence of ␤-ME as an external electron donor using aerobic as well as anaerobic conditions. Superposition of the absorption difference spectra shows that under all tested reaction conditions the same photoprod- uct is formed (Fig. 7D). The large positive peaks at 403 and 367 nm and the smaller peaks at 655, 605, and 510 nm are assigned to the generation of the anionic red FAD radical, whereas the negative peaks at 475, 450, and 428 nm are due to the disappearance of oxidized FAD. Under anaerobic conditions and in the presence of ␤-ME, the conversion of oxidized FAD to the ani-  (FAD ox ). B, reference spectra for oxidized and anionic radical states of flavin in glucose oxidase; comparison with dCRY. C, reaction scheme for the conversion of oxidized FAD to the anionic FAD radical upon blue light irradiation of dCRY and its dark recovery. D, normalized absorbance difference spectra ("after" minus "before" illumination). Difference spectra obtained under anaerobic and aerobic conditions, each in the presence and absence of ␤-ME, are superimposed. Under all of the tested reaction conditions, FAD ox is converted to the anionic FAD radical FAD . . E, dark recovery of FAD ox monitored as increase of absorbance at 450 nm over time. The samples were irradiated with 19 milliwatts/cm 2 445-nm blue light for 10 s. Whereas the rate of FAD radical generation depends on the presence/absence of ␤-ME (compare black and green lines with, respectively, red and blue lines), the recovery speed of dCRY depends on the presence of oxygen (compare black and red lines with, respectively, green and blue lines). Time constants for dark recovery are several hours for anaerobic conditions (0 (green) or 2 mM (blue) ␤-ME). Under aerobic conditions (0 (black) or 2 mM (red) ␤-ME), ϳ50% of the molecules recover with time constants of 4 -5 min, the other 50% with 20 -30 min time constants, and Ͻ1% of the molecules do not return. onic FAD radical is completed to at least 85% after ϳ100 s of illumination (Fig. 7A). The reaction does not fully revert even after several hours in darkness (Fig. 7E). Removal of ␤-ME under anaerobic conditions leads to a slower conversion of oxidized FAD to the radical. As observed in presence of ␤-ME, the anionic FAD . radical is very stable (Fig. 7E). Interestingly, FAD ox is converted to the anionic FAD radical even in the presence of oxygen and without ␤-ME (Fig. 7, D and E), clearly distinguishing the process from the common procedure of flavoprotein photoreduction (23). The addition of ␤-ME again leads to a faster conversion of FAD ox to FAD . (Fig. 7E), suggesting that ␤-ME facilitates the reaction in absence or presence of oxygen. In contrast to the forward reaction, the back reaction rate does not depend on the presence of ␤-ME but heavily depends on the presence of oxygen (Fig. 7E). Under aerobic conditions, ϳ50% of the molecules revert to FAD ox with a time constant of 4 -5 min, the other half revert with a time constant of 20 -30 min, and Ͻ1% of the molecules never revert. This contrasts with the time scales of several hours observed under anaerobic conditions. Taken together, our model for blue light activation of dCRY (Fig. 7C) is supported by the observation that ␤-ME as external electron donor enhances the forward reaction (reduction of FAD ox to FAD . via electron uptake), whereas oxygen enhances the back reaction (dark recovery of FAD ox ).

DISCUSSION
To shed light into the hitherto unknown photochemistry of the circadian blue light photoreceptor dCRY, we have established the expression and purification of full-length dCRY from Sf21 insect cells, which are closely related to the native Drosophila cells (Fig. 2). Purified dCRY behaves as a monomer in analytical gel filtration (Fig. 2E), suggesting that it might function as a monomer in vivo. Of note, A. thaliana CRY1 (AtCRY1) is suggested to form constitutive homodimers in plants, which are mediated by the PHR domain (36). Our freshly purified dCRY contains oxidized FAD and residual amounts of the postulated light-harvesting chromophore MTHF (Figs. 3-6). In earlier reports, maltose-binding protein fused dCRY expressed in E. coli was shown to contain oxidized FAD, whereas MTHF binding could not be established (37,38). Note, however, that the reported absorption spectra obtained from E. coli expressed maltose-binding protein fused dCRY were almost featureless with a small maximum at 410 -420 nm. In contrast, our insect cell-purified material provided high quality absorption spectra with the typical fine structure of protein-bound oxidized FAD (Fig. 3), allowing us to conduct UV-visible spectroscopic studies on the blue light responses of dCRY. Interestingly, blue light irradiation of our insect cell-purified dCRY leads to the conversion of FAD ox into a red anionic FAD radical (FAD . ) (Fig. 7), a radical species that has hitherto not been described for cryptochrome/photolyase family members. In contrast to the CPD photolyases (2) or the more closely related (6-4)-photolyases (3,4), neither fully reduced FADH Ϫ nor the neutral blue FADH ⅐ radical (nor other byproducts) are observed, implying a different activation mechanism for dCRY. Notably, the anionic FAD . radical of dCRY is also formed and stable under oxidizing conditions and in the absence of an external electron donor (Fig. 7).
Moreover, it is also formed at pH 6 and pH 7, i.e. well below the pK a of 8.3 quoted for the equilibrium between anionic and neutral FAD radicals (39). It is therefore conceivable that the FAD . radical corresponds to the light-activated signaling state of dCRY, whereas oxidized FAD corresponds to the ground or dark state. This interpretation of our UV-visible absorption spectra is supported by a number of in vivo action spectra for light effects on circadian rhythms; when Drosophila flies are irradiated with 400 -700-nm monochromatic light of 1-milliwatt/cm 2 intensity and 50-nm band widths for 10 min at circadian times ZT15 (early night with maximal phase delay) or ZT21 (late night with maximal phase advance), maximum phase shifts of the circadian clock are observed for 400 -500-nm light sources, with almost no responses above 600 nm (40). Likewise, phase delays and advances of the circadian rhythm of adult eclosion of Drosophila flies show maximum responses between 420 and 480 nm with a sharp decline above 550 nm (41). At a molecular level, phase shifting of the Drosophila clock is based on light-dependent interactions of dCRY with dTIM followed by degradation of both proteins (17)(18)(19). In agreement with the above mentioned physiological responses, 10 min of illumination with monochromatic 1-milliwatt/cm 2 light sources results in most efficient dTIM (40) and dCRY (21) degradation using 450 -500-nm blue light, with little effect above 600 nm. By comparison, hypocotyl growth inhibition in A. thaliana, which is mediated by AtCRY1, displays the maximum response between 390 and 480 nm (8). Therefore, hypocotyl growth inhibition has been suggested to be linked to the activity of oxidized FAD (9,10). Despite the similar action spectra and the FAD ox ground states proposed for both photoreceptors, the activation mechanism of dCRY is clearly different from that of AtCRY1; FAD ox is converted into the red anionic FAD . radical in dCRY as opposed to the blue neutral FADH ⅐ radical observed in AtCRY1. It is still possible that a secondary photoreaction takes place in dCRY starting from the anionic radical state, which would explain the residual physiological response to light above 500 nm. To prove this hypothesis, further experimental validation with higher time resolution is required. We can, however, exclude from our experiments that the two electron-reduced state of flavin (FADH Ϫ ) is the starting point of the reaction as in the case of DNA photolyases. The dCRY action spectra are clearly different from those of the DNA photolyases, where the maximal DNA repair activity is observed in the range of 320 to 440 nm because of the strong MTHF absorption (42).
Interestingly, a 10-min white light pulse of 5-milliwatt/cm 2 intensity, followed by 50 min of incubation in the dark, leads to a complete degradation of dTIM, but not dCRY in Schneider 2 cultured Drosophila cells and in flies (21). Although both proteins degrade in the light with similar half-lives of ϳ25 min, dCRY appears to rapidly revert to a stable form in darkness and therefore requires continuous light exposure for complete degradation. We propose that the blue light-induced conversion of FAD ox to the anionic FAD . radical leads to conformational changes within the dCRY molecule, likely affecting the C-terminal tail region, that trigger dCRY-dTIM interactions. dTIM is subsequently degraded by a light-independent mechanism that involves tyrosine phosphorylation and ubiquitination (18).
In its light-activated state, dCRY is also degraded, possibly because of the displacement of the tail region, which stabilizes dCRY in the dark in flies and Schneider 2 cells (21). Reversion to the oxidized FAD state in darkness would bring the tail back into its dCRY-protecting ("closed") conformation. Under oxidizing conditions, our UV-visible spectroscopic measurements on purified dCRY have revealed two dCRY populations, which revert to the FAD ox dark form with time constants of 4 -5 min and 20 -30 min, respectively (Fig. 7E). At this stage we can only speculate about the nature of these two dCRY populations. Possible explanations include different phosphorylation stages of dCRY or different conformations of the C-tail regions. However, both time constants appear long compared with the dark recovery kinetics of dCRY in Schneider 2 cells and flies (21), suggesting that additional factors speed up the dark reversion in vivo. The much slower intrinsic back reaction rates of purified dCRY compared with DNA photolyases (2) might be crucial to its biological function, namely resetting the clock by sequestering cellular dTIM from the circadian feedback loop through light-dependent dCRY-dTIM interactions and subsequent proteasomal degradation of both proteins.
Anionic FAD radicals are to the best of our knowledge unprecedented in the cryptochrome/photolyase family. However, this FAD oxidation state has been observed in several flavoproteins catalyzing redox reactions, e.g. in glucose oxidase (Fig.  7B) (23), oxylnitrilase (23), L-and D-amino acid oxidase (23,43), cholesterol oxidase (44), monomeric sarcosine oxidase (45), and choline oxidase (46). Compared with these enzymes, the 472-nm peak of the FAD . radical in dCRY is blue-shifted by 10 -15 nm (Fig. 7B and references cited above). Such a blue shift has also been observed for the neutral FADH ⅐ radical in AtCRY1 (10), when compared with CPD photolyases (47) and flavoprotein radicals (48), suggesting that it might be a con-served and distinguishing feature of the cryptochrome family. Based on the known crystal structure of the AtCRY1 PHR domain (31) and solution studies (39), this blue shift was tentatively explained by a relatively high polarity of the flavin environment. In contrast to AtCRY1, a negative charge developing near the N(1)-position of the FAD . isoalloxazine ring needs to be stabilized in the radical state of dCRY. For this purpose, all structurally characterized flavoprotein oxidases have either a positively charged residue or a helix dipole oriented toward the N(1)-C(2)ϭO region of the enzyme-bound flavin (49). In photolyases, the U-shaped conformation of the FAD chromophore may contribute to the stabilization of the negative charge of the two electron-reduced FADH Ϫ by the 3Ј hydroxyl group of the riboflavin moiety, which is in H-bonding distance to the nitrogen N(1) of the isoalloxazine ring (25). Assuming that dCRY, like all structurally characterized photolyases and cryptochromes, has FAD bound in a U-shaped conformation, we propose that the negative charge of the anionic FAD . radical may partially be stabilized in this way. Moreover, a structural model of dCRY suggests that Arg 271 , which is conserved in insect and mammalian CRYs (Fig. 1B), helps to stabilize the negative charge developing in the N(1)-C(2)ϭO region of FAD. In our dCRY model, the guanidinium group of Arg 271 is close to the O 2 of the isoalloxazine ring of FAD (Fig. 8). Of note, E. coli photolyase (25) and the DASH cryptochrome AtCRY3 (26), which form a neutral FAD radical (2,32), have an alanine in place of Arg dCRY 271 , whereas AtCRY1, which is also reported to form the neutral FAD radical (10), contains a histidine.
Because the formation of an anionic FAD radical requires an electron transfer to the oxidized FAD (Fig. 7C), an intramolecular electron transfer is to be expected in the blue light activation of dCRY. Tryptophane residues (Fig. 1B), which are involved in an electron transfer pathway for photoreduction in E. coli photolyase (50) and AtCRY1 (9), do not appear to act as electron donors in dCRY (51). Although the exchange of the secondary electron donors Trp dCRY 342 and Trp dCRY 397 to alanine interferes with dCRY-dependent transcriptional derepression through dTIM sequestration, no effect is observed when Trp dCRY 342 and Trp dCRY 397 are exchanged to the redox-inactive but structurally more similar phenylalanine. Furthermore, even the drastic exchange of the primary donor Trp dCRY 420 to alanine leads to an almost wild type-like behavior.
Kottke et al. (10) propose that in AtCRY1 an aspartic acid (Asp AtCRY1 396 ) opposite the N(5) position of the isoalloxazine ring (31) acts as a proton donor in the generation of a neutral FADH ⅐ radical from oxidized FAD. However, the formation of an anionic FAD radical from oxidized FAD in dCRY does not require a proton transfer (Fig. 7C). Consistently, dCRY has a cysteine residue (Cys 416 ) in place of the aforementioned Asp AtCRY1 396 ( Figs. 1B and 8), a fact that might be of significant importance for the different blue light responses of AtCRY1 and dCRY. The pK a of Cys 416 is expected to be more than four units higher than that of an aspartic acid at this position, thereby making proton transfer more unfavorable. The functional importance of Cys 416 is, however, suggested by its conservation within photosensitive CRY1-type insect cryptochromes such as the monarch butterfly (Danaus plexippus) CRY1 and the mosquito (Anopheles gambiae) CRY1 (Fig. 1B) (52).
In the absence of a three-dimensional structure of dCRY, we can only speculate about additional candidate residues for electron transfer, catalytic functions, or mechanisms of blue light signaling to the C-tail and effector proteins. However, our data suggest a primary photoreaction mechanism for dCRY that is clearly different from AtCRY1, class I-CPD, and (6-4)-photolyases. It will be interesting to see whether the dCRY photoreaction mechanism proposed herein is valid for other animal cryptochromes.