Structural and Evolutionary Aspects of Antenna Chromophore Usage by Class II Photolyases*

Background: Photolyases are light-driven DNA repair enzymes harboring a catalytic FAD cofactor and usually an antenna chromophore. Results: 8-Hydroxydeazaflavin is the cognate antenna of the Methanosarcina mazei photolyase, an archaeal representative of the clade of otherwise metazoan class II photolyases. Conclusion: Phylogenetically, photolyases lost 8-hydroxydeazaflavin as antenna only in higher plants. Significance: 8-Hydroxydeazaflavin occurs as cofactor within major parts of the metazoan phylome. Light-harvesting and resonance energy transfer to the catalytic FAD cofactor are key roles for the antenna chromophores of light-driven DNA photolyases, which remove UV-induced DNA lesions. So far, five chemically diverse chromophores have been described for several photolyases and related cryptochromes, but no correlation between phylogeny and used antenna has been found. Despite a common protein topology, structural analysis of the distantly related class II photolyase from the archaeon Methanosarcina mazei (MmCPDII) as well as plantal orthologues indicated several differences in terms of DNA and FAD binding and electron transfer pathways. For MmCPDII we identify 8-hydroxydeazaflavin (8-HDF) as cognate antenna by in vitro and in vivo reconstitution, whereas the higher plant class II photolyase from Arabidopsis thaliana fails to bind any of the known chromophores. According to the 1.9 Å structure of the MmCPDII·8-HDF complex, its antenna binding site differs from other members of the photolyase-cryptochrome superfamily by an antenna loop that changes its conformation by 12 Å upon 8-HDF binding. Additionally, so-called N- and C-motifs contribute as conserved elements to the binding of deprotonated 8-HDF and allow predicting 8-HDF binding for most of the class II photolyases in the whole phylome. The 8-HDF antenna is used throughout the viridiplantae ranging from green microalgae to bryophyta and pteridophyta, i.e. mosses and ferns, but interestingly not in higher plants. Overall, we suggest that 8-hydroxydeazaflavin is a crucial factor for the survival of most higher eukaryotes which depend on class II photolyases to struggle with the genotoxic effects of solar UV exposure.

Light-harvesting and resonance energy transfer to the catalytic FAD cofactor are key roles for the antenna chromophores of light-driven DNA photolyases, which remove UV-induced DNA lesions. So far, five chemically diverse chromophores have been described for several photolyases and related cryptochromes, but no correlation between phylogeny and used antenna has been found. Despite a common protein topology, structural analysis of the distantly related class II photolyase from the archaeon Methanosarcina mazei (MmCPDII) as well as plantal orthologues indicated several differences in terms of DNA and FAD binding and electron transfer pathways. For MmCPDII we identify 8-hydroxydeazaflavin (8-HDF) as cognate antenna by in vitro and in vivo reconstitution, whereas the higher plant class II photolyase from Arabidopsis thaliana fails to bind any of the known chromophores. According to the 1.9 Å structure of the MmCPDII⅐8-HDF complex, its antenna binding site differs from other members of the photolyase-cryptochrome superfamily by an antenna loop that changes its conformation by 12 Å upon 8-HDF binding. Additionally, so-called Nand C-motifs contribute as conserved elements to the binding of deprotonated 8-HDF and allow predicting 8-HDF binding for most of the class II photolyases in the whole phylome. The 8-HDF antenna is used throughout the viridiplantae ranging from green microalgae to bryophyta and pteridophyta, i.e. mosses and ferns, but interestingly not in higher plants. Overall, we suggest that 8-hydroxydeazaflavin is a crucial factor for the survival of most higher eukaryotes which depend on class II photolyases to struggle with the genotoxic effects of solar UV exposure.
Besides photosynthetic reaction centers and protochlorophyllide reductases DNA photolyases belong to a few lightdriven enzymes, which convert light energy into breakage of UV-B-induced bonds between neighboring pyrimidines (1). As highly efficient DNA repair enzymes, photolyases harvest blue/ near-UV light photons for channeling excitation energy into the cleavage of non-oxidative photoproduct lesions in DNA. These lesions are caused by solar exposure of non-shielded DNA to UV-B ( ϳ 280 -315 nm) and UV-C ( ϳ 100 -280 nm) and consist for 10 -20% of the pyrimidine (6-4) 3 pyrimidone photoproduct, and for 80 -90% of the cyclobutane pyrimidine dimer (CPD) lesion in its cis-syn isoform (2). Given the predominance and genotoxicity of the latter, most organisms confronted with sunlight, including archaea, bacteria, and eukaryotes, utilize CPD photolyases to maintain their genomic integrity. Only a few evolutionary later evolved organisms such as the placental mammals afford the loss of photolyase orthologues by relying on alternative DNA repair pathways for the maintenance of their genomic integrity (3,4).
The catalytic mechanism of photolyases generally employs the photo-induced injection of an electron from a fully reduced and typically U-shaped flavin adenine dinucleotide cofactor (FADH Ϫ ) onto the DNA lesion. This electron transfer triggers cleavage of CPD or (6-4) photoproducts inside duplex DNA (1,5) within less than a nanosecond and achieves quantum yields close to one for CPD repair (6,7). Spectroscopic and structural studies (8 -10) showed that class I, class II, and (6-4) photolyases catalyze transient electron transfer to these DNA lesions by binding the lesions next to the FADH Ϫ cofactor in the catalytic ␣-helical C-terminal domain. The N-terminal domain of these photolyases adopts a Rossmann fold, which can accommodate additional prosthetic groups, so called antenna chromophores. Compared with other members of the photolyase-cryptochrome family, class II photolyases are evolutionary distant with pairwise sequence identities of Ͻ16%. Nevertheless, structures of class II photolyases from a methanogenic archaeon (8) as well as from a higher plant (11) indicate a similar bilobal architecture as class I and (6-4) photolyases and cryptochromes. Although class II photolyases catalyze like class I CPD photolyases and DASH cryptochromes light-driven DNA-repair and photoreduction of the catalytic FAD cofactor, they show marked differences for their active site structures, the mode of DNA binding, the electron transfer pathways required for photoreduction as well as for the discrimination between intact and UV lesion-containing DNA.
A shortcoming of the catalytic FADH Ϫ cofactor is its poor absorption in the blue to near-UV range with extinction coefficients between 2800 M Ϫ1 cm Ϫ1 (400 nm) and 5600 M Ϫ1 cm Ϫ1 (360 nm). For efficient light-driven repair in the blue/near-UV region, photolyases hence rely on a second auxiliary chromophore with large absorption cross-sections to broaden up their limited spectral properties. So far, five different classes of antenna chromophores have been identified from members of the photolyase-cryptochrome superfamily (Fig. 1). These antenna chromophores comprise aromatic moieties absorbing in the range between 380 and 420 nm and include (i) 5,10methenyltetrahydrofolate (MTHF) as found in many microbial class I photolyases (12) and DASH cryptochromes (13); (ii) 8-hydroxydeazaflavin (8-HDF) from (6-4) photolyases (14,15) and several class I photolyases (16); (iii and iv) flavin mononucleotide (FMN) and FAD in the class I photolyases from Thermus thermophilus (17) and Sulfolobus tokodaii (18), respectively; and finally (v) 6,7-dimethyl-8-ribityl-lumazin in a novel class of proteo-/cyanobacterial cryptochromes (19). For light harvesting the antenna chromophores absorb a photon and transfer its energy via a Förster mechanism to the catalytic active FADH Ϫ . All nucleotide-like chromophores (FAD, FMN, and 8-HDF) are bound within the N-terminal ␣/␤ domain of photolyases with its Rossmann-like fold in a distance of 17-18 Å to the catalytic FAD cofactor. In contrast to the deeply buried nucleotide-like antenna, the pterin derivative MTHF is bound close to the protein surface along the interface of the N-and C-terminal domains (20).
Like MTHF, the riboflavin derivatives FAD, FMN, as well as the riboflavin biosynthesis intermediate 6,7-dimethyl-8-ribityllumazin are commonly found in archaea, bacteria, and eukaryotes, whereas the occurrence of deazaflavin cofactors such as 8-HDF is more limited. For example, deazaflavins are signature molecules of methanogenic archaea, where 8-HDF in its oligoglutamylated and protonated F420 form plays an important role as a low potential hydride carrier (21). Nevertheless, deazaflavins are also found in several other bacterial clades such as the actinobacteria (22) as well as in unicellular green algae. The biological role of 8-HDF in photolyases was demonstrated in the green algae Chlamydomonas reinhardtii. Here, inactivation of the gene for the deazaflavin synthase PHR1 caused a loss of the photorepair of UV lesions as catalyzed in C. reinhardtii by the class II photolyase PHR2 (23). Interestingly, the (6-4) and class II CPD photolyases from the insect Drosophila melanogaster utilize 8-HDF as antenna chromophore as well (15), although this organism misses like other animals a genomic copy of the deazaflavin synthase gene. Instead, the bacterial symbiont Wolbachia was suggested to supply 8-HDF as a vitamin essential for photolyase function (14).
Here, we investigate the antenna chromophore of the class II CPD photolyase from Methanosarcina mazei (MmCPDII) and the spectroscopic and structural characterization of its complex with the 8-HDF antenna. Together with phylogenetic data the structure of the MmCPDII⅐8-HDF complex and its differences from higher plant photolyases allow us to predict the occurrence of 8-HDF in the whole metazoan branch of life.
Generation and Preparation of MmCPDII Mutants-Photolyase mutants MmCPDII-S26L and MmCPDII-S26F were obtained from pET28a-MmCPDII by site-directed mutagenesis (QuikChange II; Agilent); resulting plasmids were verified by dideoxy-sequencing (Qiagen). Co-expression and purification of the Ser 26 mutants were performed as described for the wild type photolyase.
In Vitro Reconstitution of ApoMmCPDII with Antenna Chromophores-ApoMmCPDII (53 M) was mixed with the equimolar amount of either FAD, FMN, MTHF, F420, or chemically synthesized 8-HDF. Afterward, all mixtures as well as negative controls were dialyzed at 4°C against 1.5 ml of buffer I (10 mM Tris-HCl, 100 mM NaCl, pH 8.0) in a pre-greased VDX TM plate (Hampton Research) using 10-l microdialysis rods (Hampton Research) and a 3.5-kDa cutoff SnakeSkin dialysis membranes (Pierce). After 24 h, reconstitution was analyzed by UV/visible spectroscopy using a NanoDrop 1000 spectrophotometer (Thermo Scientific). Likewise, the 8-HDF reference spectrum was recorded with the NanoDrop 1000 spectrophotometer and normalized to the MmCPDII concentration used before.
Structure Determination of ApoMmCPDII Crystals Soaked with 8-HDF-MmCPDII crystals grown in 0.5 M lithium sulfate and 7.5% PEG 8000 were soaked for 30 min in crystallization buffer supplemented with 100 M synthetic 8-HDF. During soaking changes in the integrity of MmCPDII crystals were observed and are most likely reflected by the lower resolution limit of 2.7 Å compared with apoMmCPDII crystals (1.5 Å). Before flash-freezing in liquid nitrogen crystals were soaked in the crystallization solution supplemented with 30% glycerol as cryoprotectant. Diffraction data were collected from a single crystal at 100 K at beamline ID23-2, European Synchrotron Radiation Facility (ESRF). Data processing was carried out using XDS and XSCALE (24) and further refinement using COOT (25) and REFMAC5 (26) at 2.7 Å resolution led to R-factors of R work ϭ 20.6% and R free ϭ 27.0%. Data processing and refinement statistics are summarized in Table l.
Crystallization and Structure Determination of the MmCPDII⅐8-HDF Complex-An initial 96-well format crystallization screening was performed with a Cartesian robot system and commercially available crystallization screens (Qiagen) using the sitting drop vapor diffusion method at 18°C. First crystals were obtained in a crystallization condition containing 0.1 M lithium sulfate, 0.1 M trisodium citrate, pH 5.6, and 12% PEG 6000 within 48 h. Using the hanging drop vapor diffusion method and this condition, crystals suitable for data collection were grown in a 24-well plate at a protein concentration of 11.8  JULY 11, 2014 • VOLUME 289 • NUMBER 28 mg/ml. For cryoprotection, glycerol was added to a final concentration of 33%, and MmCPDII⅐8-HDF crystals were flashfrozen in liquid nitrogen. X-ray data were collected from a single crystal at 100 K at beamline ID23-1, ESRF. MmCPDII⅐8-HDF crystallized in the orthorhombic space group P2 1 2 1 2 1 with unit cell parameters a ϭ 78.97 Å, b ϭ 114.7 Å, c ϭ 141.4 Å. Diffraction data were processed using XDS and XSCALE (24), and initial phases were obtained by molecular replacement using PHASER (27) and the structure of MmCPDII in a tetragonal crystal form (2XRY) as search model. The correct solution with data up to 2.5 Å had two molecules per asymmetric unit with z scores of 25.7 for rotation and 45.9 for translation function for the first molecule and z scores of 26.2 and 79.7 for the second molecule, respectively. After initial automated model building from experimental phases as done by ARP/wARP (28), further refinement using COOT (25) and REFMAC5 (26) at 1.9 Å resolution led to R-factors of R work ϭ 14.9% and R free ϭ 17.8%. Data processing and refinement statistics are summarized in Table 1. The nomenclature for the secondary structure of MmCPDII is taken from Kiontke et al. (8). Structural analysis and figures were done with PyMOL 1.5 (29).

Antenna Chromophore Usage by Class II Photolyases
UV/Visible and Fluorescence Spectroscopy-UV/visible spectra were recorded using a V-660 spectrometer (JASCO) at 8°C. Fluorescence spectroscopy of MmCPDII in the oxidized state was carried out at 4°C on a FP-6500 spectrofluorometer (Jasco). Samples of apoMmCPDII and MmCPDII⅐8-HDF with concentrations of 10 M or 5.8 M in buffer I were excited at a wavelength of 420 nm with band widths of 3 nm for excitation and 5 nm for emission, respectively. Reference fluorescence emission spectra were recorded with a high power LED (Roithner Lasertechnik), Maya 2000 Pro spectrometer (Ocean Optics) and SpectraSuite software (Ocean Optics). To generate the reduced state, samples of 16 M MmCPDII in 10 mM DTT were deoxygenized before being photoreduced on ice by illumination at 450°nm for 10 min using a 450°nm high power LED (Roithner Lasertechnik, 0.24 W/cm 2 ). Fluorescence emission/ excitation spectra were recorded at 4°C on a Fluorolog-3 (Horiba) using slit widths of 2 nm.

RESULTS AND DISCUSSION
In Vitro Reconstitution of MmCPDII with Synthetic 8-HDF-When purified via the E. coli expression system, recombinant class II photolyase from M. mazei (MmCPDII) possesses the catalytic FAD cofactor but lacks any bound and host-derived antenna chromophore as shown before by UV/visible spectroscopy and structural analysis (8). To identify the cognate lightharvesting prosthetic group of MmCPDII, we performed in vitro reconstitution of this so called apoMmCPDII with known antenna chromophores such as FAD, FMN, MTHF, 8-HDF, and F420. Only for synthetic 8-HDF, the UV/visible spectroscopic analysis reveals a characteristically modified absorbance spectrum of the M. mazei class II photolyase. The absorbance spectrum of oxidized apoMmCPDII with peaks at 362, 377, 421, 444, and 469 nm is dominated in the reconstituted MmCP-DII⅐8-HDF complex by a significant peak at 435 nm (Fig. 1). Furthermore, incorporation of deprotonated 8-HDF ( max 420 nm) into MmCPDII causes a 15-nm red shift of the chromophore absorbance. This red shift of bound 8-HDF complies well with data reported for D. melanogaster (6-4) photolyase (14) and A. nidulans class I photolyase (AnCPDI), the latter either purified directly from A. nidulans cells (16) or in vitro reconstituted with 8-HDF using recombinant enzyme (30,31).
In Vivo Assembly of the MmCPDII⅐8-HDF Holoenzyme-To confirm these results and to rule out that incorporation of synthetic 8-HDF during in vitro reconstitution and crystal soaking is enforced by the chosen experimental conditions, an in vivo reconstitution system of MmCPDII with endogenously produced 8-HDF was established. For this purpose, a cofactor plasmid encoding the S. coelicolor FO synthase (ScFbiC), a bifunctional, S-adenosylmethionine-dependent enzyme (26), was constructed that enables the biosynthesis of 8-HDF in E. coli from 4-hydroxypyruvate and 5-amino-6-ribityl-amino-2,4(1H,3H)-pyrimidinedione, an intermediate of the riboflavin synthesis pathway ( Fig. 2A). After nickel-nitrilotriacetic acid affinity chromatography, the MmCPDII⅐8-HDF complex was separated from ScFbiC by size exclusion chromatography (Fig. 2B). Like in vitro reconstituted MmCPDII, the UV/visible spectrum of the MmCPDII⅐8-HDF holoenzyme in the oxidized state is hallmarked by the distinct absorption maximum at 435 nm (Fig.  2C, black line) indicating that 8-HDF is bound to the photolyase. Moreover, the fluorescence emission spectrum of the complex exhibits a strong fluorescence peak at em 467 nm (Fig.  2C, thick red line) and differs significantly from apoMmCPDII with the latter exhibiting only broader emission at em of approximately 528 nm for the oxidized FAD cofactor (Fig. 2C, dotted red line). In accordance with data from AnCPDI (16,31) and Dm   (14), the fluorescence maximum of MmCPDIIbound 8-hydroxydeazaflavin resembles the maximum of deprotonated 8-HDF in solution with its em of 470 nm (Fig.  3A). In contrast, the fully reduced FADH Ϫ state exhibits an emission peak for the 8-HDF antenna that is blue-shifted by 12 nm to 455 nm, thus suggesting some interaction between the redox state of the FAD cofactor and the structure of the antenna binding site (Fig. 2D). Furthermore, excitation spectra recorded at em ϭ 533 nm, the emission maximum of the reduced FADH Ϫ chromophore, demonstrate efficient Förster energy transfer between the antenna and the flavin cofactor, as the resulting spectra are clearly dominated by the 8-HDF antenna with its max of 435 nm (Fig. 2D).
To clear up the question, whether Methanosarcinales are generally capable of synthesizing 8-HDF as cognate antenna of their photolyases, we performed a protein-protein BLAST search with Methanocaldococcus jannaschii CofG and CofH, which have been used before to produce artificial 8-HDF in E. coli (32). Indeed, M. mazei harbors like other methanosarcinal species, e.g. Methanosarcina acetivorans or Methanosarcina barkeri, one CofG and two CofH orthologues (SwissProt entries COFG_ METMA, COFH1_METMA, COFH2_METMA).

Validation of the Heterologous in Vivo 8-HDF Reconstitution
System-To rule out that our in vivo reconstitution system generates noncognate antenna-photolyase hybrids, class I and (6-4) photolyases were co-expressed with the ScFbiC FO synthase. AnCPDI was chosen as reliable positive control for successful incorporation of the 8-HDF antenna chromophore, because AnCPDI that is directly purified from A. nidulans cells harbors 8-HDF as cognate antenna (16). This cyanobacterium carries genomic copies of the 8-HDF synthase genes cofG and cofH (UniProt entries Q5N4D1, Q5N3T7), on which biosynthesis of the 8-HDF chromophore depends. For comparison, the (6-4) photolyase from D. melanogaster as well as the class I CPD photolyase from T. thermophilus (TtCPDI) were used, as the corresponding organisms lacking CofG/CofH orthologues. Here, 8-HDF has been successfully incorporated by either in vitro reconstitution of Ds(6-4) (14) or crystal soaking of TtCPDI (33).
After purification, all putative recombinant photolyase complexes were analyzed by UV/visible spectroscopy to analyze for in vivo reconstitution with endogenously produced 8-HDF. In case of AnCPDI and Dm(6-4) the UV/visible spectra exhibit prominent absorbance maxima at approximately 440 -450 nm (Fig. 3B). Whereas the absorbance maximum of the AnCPDI⅐8-HDF complex at max 438 nm corresponds closely to spectra already published (16,30,31), the maximum at max 448 nm for the in vivo reconstituted Dm(6-4)⅐8-HDF complex is redshifted by 8 nm compared with the in vitro reconstituted photolyase with a max of 440 nm (14). Interestingly, in vivo reconstitution with 8-HDF failed for TtCPDI, although the FO synthase was solubly expressed (Fig. 3, A-C). Accordingly, the co-crystal structure of TtCPDI with synthetic 8-HDF, which was generated by soaking TtCPDI crystals with millimolar concentrations of 8-HDF (33), corresponds to a noncognate complex of this photolyase. However, the lack of endogenous 8-HDF biosynthesis pathways is not per se an indicator for 8-HDF playing no role as a cognate antenna chromophore of the respective photolyase. The fruit fly D. melanogaster lacks the genes required for 8-HDF biosynthesis, but its (6-4) and CPD photolyases are nevertheless capable of accommodating 8-HDF as antenna chromophore (15). This insect species may derive 8-HDF from a bacterial, hitherto unknown commensal as suggested before (14).
Structures of the MmCPDII⅐8-HDF Holoenzyme-To characterize 8-HDF antenna binding by MmCPDII we derived cocrystal structures either by crystallization of in vivo reconstituted MmCPDII⅐8-HDF holoenzyme at 1.9 Å resolution or by soaking apoMmCPDII crystals with 8-HDF and subsequent structural analysis at 2.7 Å resolution. Interestingly, the holoenzyme crystallized in a novel, orthorhombic crystal form with two molecules per asymmetric symmetry unit (Table 1). Apart from differences in the linker region between the N-and C-terminal domain (Val 186 -Glu 231 ) and the antenna binding loop there are almost no deviations between the MmCPDII⅐8-HDF and the apoMmCPDII structures. Furthermore, the root mean square deviations for complexes A and B of the MmCPDII⅐8-HDF holoenzyme are low with 0.20 and 0.28 Å (330, 353 C ␣ atoms), respectively, relative to the apoMmCPDII structure. In both reconstitutions, the 8-HDF chromophore is buried deeply within the N-terminal domain in a distance of 16.7 Å to the catalytic FAD cofactor, which is well in range for efficient Förster energy transfer (Fig. 4, A and B). According to difference electron density the in vivo reconstituted MmCPDII⅐8-HDF holoenzyme shows quantitative incorporation of the 8-HDF chromophore, whereas soaking achieved under our conditions only 70% occupancy (Fig. 4, C and D). The inhomogeneity of the latter is reflected by the alternative conformations taken up by parts of the antenna loop (Leu 57 -Ala 64 ) that links ␤-strand ␤2 with helix ␣2. Here, the short helical turn (Glu 60 -Glu 63 ) that is found in 8-HDF-free MmCPDII structures (PDB codes 2XRY and 2XRZ (8)) undergoes a conformational change by closing the entrance to the 8-HDF binding site. The largest change within the antenna loop is made by Phe 61 -Leu 62 , which swivel their packed side chains by a distance of over 12 Å to form --interaction between Phe 61 and the middle ring of the deazaflavin (Fig. 4B).
The in vivo reconstituted MmCPDII⅐8-HDF holoenzyme shows even larger conformational changes compared with the apo-structures of MmCPDII. The rearrangement of the  antenna loop involves the region between Asp 59 and Ala 64 and seals the 8-HDF binding site from bulk solvent access. Furthermore, major parts of the linker between the N-and C-terminal domains of this photolyase (Val 186 -Glu 231 ) either adopt a conformation different from in 8-HDF lacking MmCPDII structures or are disordered as in complex B (Glu 189 -Leu 217 ). In complex A, helix ␣7 (Leu 200 -Glu 214 ), which is structurally also conserved in the class II photolyase from rice (Oryza sativa, PDB code 3UMV, Asp 221 -Glu 232 (11)), breaks down into three shorter helical segments (␣7A, Leu 200 -Glu 205 ; ␣7B, Val 208 -Leu 211 ; ␣7C, Glu 214 -Lys 219 ). The newly formed ␣7C segment hereby stabilizes the closed conformation of the antenna loop by several, newly formed van der Waal interactions. As a result Asp 222 of the linker that is exposed before as part of a loop becomes buried and forms a salt bridge with Arg 67 . Overall, these structural adaptations of the second shell around the antenna chromophore binding site most likely impede reopening and hence release of bound 8-HDF.
The Antenna Binding Site of MmCPDII-Although the 8-HDF chromophore locates in the same region of the N-terminal domain of MmCPDII as in the cognate 8-HDF complexes of class I and (6-4) photolyases (Fig. 5), there are several peculiar differences for its binding site. First, the re-side of the aromatic ring system of 8-HDF, i.e. in Fig. 5 the backside, makes extensive van der Waal interactions with the bulky side chains of Phe 273 and Phe 55 as well as with Met 72 via its middle ring. In other photolyases, the latter residue is exchanged by a leucine, whereas the walling by phenylalanines is conserved. More important, the si-side of the chromophore aromatic system is covered by Thr 58 and Phe 61 from the nonregular antenna loop, the indole of Trp 126 as well as the side chain of Arg 411 . In other photolyases, the antenna loop is replaced by a short helical segment, which points only with a leucine toward the middle ring of the 8-HDF chromophore. Second, Trp 126 replaces in class II photolyases an otherwise highly conserved arginine residue (AnCPDI, Arg 109 ; Ds , Arg 118 ), which forms in class I and (6-4) photolyases an H-bond to the C2-carbonyl of the deazaflavin moiety. In MmCPDII, this role is taken by Ser 26 , which forms hydrogen-bonds to the C2-carbonyl as well as the 3Ј-hydroxy group of the ribityl moiety (Fig. 5). Finally, two basic residues surround the phenolic ring of the 8-hydroxydeazaflavin to stabilize its deprotonated, anionic state. Only one of these residues, His 272 , is similar in class I and 6-4 photolyases (AnCPDI, Lys 248 ; Ds , Lys 266 ), where it forms a salt bridge to the 8-oxy group of 8-HDF. The conformation of this histidine is stabilized by a salt bridge to Asp 412 from the C-terminal domain of MmCPDII (Fig. 5). Interestingly, its preceding residue, Arg 411 , makes the second salt bridge to the 8-oxy group. This residue has no counterpart in class I and (6-4) photolyases and is replaced in the latter by an arginine from the N-terminal domain (AnCPDI, Arg 51 ; Ds(6-4), Arg 60 ).
Given other structural differences, which are crucial for function, e.g. the distinct electron-transfer pathways for photoreduction or the binding sites of the catalytic FAD chromophore (8), these observations corroborate the notion of a large evolutionary gap between class II photolyases and other members of the photolyase-cryptochrome superfamily. For example, class I photolyases that lack 8-HDF as cognate antenna such as TtCPDI, miss all of the ascribed motifs of the 8-HDF binding site (Fig. 5).
Conservation of the 8-HDF Antenna Chromophore within Class II Photolyases-Apart from the structural differences described above, a comparison of the MmCPDII⅐8-HDF complex and the recently published O. sativa class II CPD photolyase (OsCPDII) shows no other major deviations. Nevertheless, heterologously expressed O. sativa photolyase lacks any kind of an additional antenna chromophore as shown previously by UV/visible spectroscopy (34) as well as by current structural characterization (11). Likewise, the heterologously overexpressed class II photolyases from Arabidopsis thaliana, whose sequence identity to OsCPDII exceeds 65%, fails to incorporate under our in vivo conditions any 8-HDF chromophore (Fig. 3, B and C). A superimposition of the OsCPDII structure with the MmCPDII⅐8-HDF complex (root mean square deviation 0.60 Å for 338 C␣ atoms) shows that although most amino acids lining the antenna binding pocket are either conserved (OsCPDII: Trp 43 , Leu 76 , Ala 85 , His 289 , Phe 290 , Arg 429 ) or have at least similar biochemical properties (see Fig.  6C). However, there are several peculiar differences correlating with the loss of 8-HDF binding. First, the "lower" part of the pocket, where the aromatic ring moiety of 8-HDF is accommodated, lacks in OsCPDII the aromatic residues from the antenna loop that are responsible forstacking from the si-side (MmCPDII, Phe 61 ; OsCPDII, Leu 82 ) as well as the edge-to-face interaction with the Cys 6 of the deazaflavin (MmCPDII, Tyr 69 ; OsCPDII, Leu 90 (35)). Instead, the binding pocket is partly filled in OsCPDII by the bulky side chain of Phe 77 (MmCPDII: Thr 58 ) from the antenna loop. Second, in OsCPDII a leucine (Leu 45 ) replaces the polar residue Ser 26 of MmCPDII, which is crucial for the recognition of 8-HDF via the H-bonds to the C2-carbonyl and 3Ј-hydroxy group. In contrast, the basic residues His 272 and Arg 411 of the C-terminal catalytic domain, which stabilize the deprotonation of the 8-hydroxy group, are preserved in the higher plant photolyases like AtCPDII or OsCPDII and are hence only weak indicators for antenna chromophore binding.
Taken together, we can now predict the signature motifs, by which members of the class II photolyase family are capable of utilizing 8-HDF as a cognate antenna (Fig. 6, A and B). First, the C-motif harboring the basic residues His 272 and Arg 411 has to be intact for interacting with the deprotonated 8-oxy group of 8-HDF. Second, the N-motif (WMS) has to provide a polar residue for H-bonding interactions with 8-HDF as well as formation of the binding site wall. A replacement of this residue by a bulky aromate like in the S26F mutant of MmCPDII and predictably most plant eudicots is incompatible with 8-HDF antenna binding due to steric hindrance, whereas a smaller exchange by leucine is still tolerated for the in vivo incorporation of 8-HDF (Fig. 6, D and E). Third, a small nonbulky residue (MmCPDII: Thr 58 ) as well as the aromatic residues of the antenna loop are required for forming the binding site covering the si-side of the 8-HDF chromophore. Apart from the methanosarcinal photolyases, several bacterial class II photolyases fulfill these criteria such as those from Rubrobacter xylanophilus or members of the classes Chlorobiaceae and Desulfovibrionales. Interestingly, only a few of these microorganisms carry genes for a CofG orthologue for endogenous 8-HDF biosynthesis (Figs. 6A and 7), implying that many microbes may depend on the supply of 8-HDF from the environment for assembling antenna-bearing CPD photolyases. In the plant kingdom, green algae like C. reinhardtii or the Ostreococcus species own class II photolyases with cognate motifs for 8-HDF recognition together with genome-encoded deazaflavin synthases, e.g. PHR1 from C. reinhardtii (23). Interestingly, the ability to utilize 8-HDF antenna comprising class II photolyases for DNA repair is not restricted to these simple, still microbial model systems of higher plants, but is also maintained in land plants like mosses, lycophytes, and ferns. These embryophytes encode class II photolyases with intact 8-HDF binding sites as well as bifunctional CofG/CofHlike synthases for 8-HDF (Physcomitrella patens, A9SZ46; Selaginella moellendorffii, XP_002968233). Seed plants, i.e. angio-and gymnosperms, have apparently lost the ability to equip their class II photolyases with the deazaflavin antenna chromophore due to a loss of the corresponding 8-HDF synthases. Given the structural features at the antenna binding site of the monocot OsCPDII photolyase, where not only the region accommodating the pyrimidine ring is occupied by a leucine replacing Ser 26 of MmCPDII, but also affected by a two residue longer antenna loop (Fig. 7), one may argue that the photolyases from monocots and eudicots are capable of recognizing other antenna chromophores than the known ones. For example, colored secondary metabolites such as isoflavonoids, aurones, and anthocyanins have only evolved in gymnosperms and angiosperms (36) and may replace the 8-HDF antenna of class II photolyases, which are predicted to be still present in fern and moss photolyases.
Interestingly, class II photolyases from animals present a more perplexing view. Clearly, all known animal genomes indicate the absence of any biosynthesis pathway for deazaflavins. Nevertheless, both the (6-4) as well as the CPD photolyases from the insect D. melanogaster can be reconstituted with an 8-HDF antenna (14,15). Likewise, other diphtheria, e.g. the vectors for malaria and West Nile fever, Anopheles gambiae and Aedes aegyptii, encode class II photolyases compatible with 8-HDF antenna chromophores (Fig. 7). Interestingly, the earlier notion that endosymbiotic Wolbachia or Spiroplasma species may be the source of 8-HDF for D. melanogaster photolyases (14) is not supported by genomic data, which predict a lack of 8-HDF biosynthesis pathways in these bacteria. Accordingly, class II photolyases compatible with 8-HDF antennas are not a general trait in the insect class, because the class II photolyases from hemiptera and hymenoptera like the honey bee (Apis mellifera) show substitutions similar to those found in the higher plant photolyases. The same diversity also holds for the vertebrates. Here, many amphibians, reptilians, and even non-placental mammals encode 8-HDF utilizing class II photolyases (Fig. 7), whereas others like birds and bony fishes show substitutions in their photolyases genes, which ablate 8-HDF binding.

CONCLUSION
8-HDF is an apparently widespread antenna chromophore of photolyases not only among microbial organisms, but also in plants and animals. In the latter, it may fulfill a role as a vitamin as suggested before for the (6-4) photolyases from D. melanogaster (14). Suitable supplies for 8-HDF are many classes of gut microbes, e.g. methanosarcinales, or the environment. Given previous attempts to improve the UV resistance of mammals and plants by introducing photolyase transgenes into these organisms one can postulate that the provision of a suitable pathway for antenna chromophore biosynthesis may be a prerequisite as well.