Crystal structure of a prokaryotic (6-4) photolyase withan Fe-S cluster and a 6,7-dimethyl-8-ribityllumazineantenna chromophoreFan Zhanga,1, Patrick Scheererb,1, Inga Oberpichlera, Tilman Lampartera,2, and Norbert Kraußc,2

aBotanical Institute, Karlsruhe Institute of Technology, D-76131 Karlsruhe, Germany; bInstitute of Medical Physics and Biophysics (CC2), AG Protein X-RayCrystallography, Charité–University Medicine Berlin, D-10117 Berlin, Germany; and cSchool of Biological and Chemical Sciences, Queen Mary Universityof London, London E1 4NS, United Kingdom

Edited by Aziz Sancar, University of North Carolina, Chapel Hill, NC, and approved March 18, 2013 (received for review February 5, 2013)

The (6-4) photolyases use blue light to reverse UV-induced (6-4)photoproducts in DNA. This (6-4) photorepair was thought to berestricted to eukaryotes. Here we report a prokaryotic (6-4) photo-lyase, PhrB from Agrobacterium tumefaciens, and propose that (6-4)photolyases are broadly distributed in prokaryotes. The crystal struc-ture of photolyase related protein B (PhrB) at 1.45 Å resolution sug-gests a DNA binding mode different from that of the eukaryoticcounterparts. A His-His-X-X-Arg motif is located within the proposedDNA lesion contact site of PhrB. This motif is structurally conservedin eukaryotic (6-4) photolyases for which the second His is essentialfor the (6-4) photolyase function. The PhrB structure contains6,7-dimethyl-8-ribityllumazine as an antenna chromophore anda [4Fe-4S] cluster bound to the catalytic domain. A significant partof the Fe-S fold strikingly resembles that of the large subunit ofeukaryotic and archaeal primases, suggesting that the PhrB-likephotolyases branched at the base of the evolution of the crypto-chrome/photolyase family. Our study presents a unique prokaryotic(6-4) photolyase andproposes that the prokaryotic (6-4) photolyasesare the ancestors of the cryptochrome/photolyase family.

DNA repair | UV light | energy transfer

The cryptochrome/photolyase family (CPF) encompasses struc-turally highly conserved flavoproteins with divergent functions.

Whereas cryptochromes regulate growth and development of plantsand circadian rhythm of plants and animals, photolyases repairUV-induced DNA lesions thereby preventing growth delay, mu-tagenesis, cell death, and cancer (1–3). Based on their substratespecificities, two classes of photolyases can be distinguished, cyclopyrimidine dimer (CPD) photolyases and (6-4) photolyases. CPDphotolyases have been identified in all three domains of life and aregenerally accepted as the common ancestors of the CPF. By con-trast, (6-4) photolyases have only been found in eukaryotes and aretherefore presumed to have formed later during evolution (1, 4, 5).Agrobacterium tumefaciens is a soil bacterium famous for its

gene transfer mechanism (6). The genome of A. tumefaciens bearstwo genes encoding the photolyase-related proteins A and B(PhrA and PhrB) (7). PhrA is a prototypical CPD class III pho-tolyase, whereas PhrB belongs to a group of photolyase-like pro-teins that has recently been denominated as iron-sulfur bacterialcryptochromes and photolyases (FeS-BCP) and is distantly relatedto other CPFs. Mutant studies have shown that both PhrA andPhrB are required for the full photoreactivation in A. tumefaciensand that the roles of PhrA and PhrB with respect to in vivo DNArepair are not redundant (7). We therefore reasoned that PhrBcould function as a (6-4) photolyase.Wedescribe here a functional study of recombinant PhrBwith an

HPLC-based in vitro DNA repair assay and confirmed it to be a(6-4) photolyase, which contains a 6,7-dimethyl-8-ribityllumazine(DMRL) chromophore and an iron-sulfur cluster. We also deter-mined the crystal structure of PhrB at 1.45 Å resolution. Significantdifferences and essential common features were identified in acomparative analysis of the crystal structures of (6-4) photolyasesincluding PhrB and its eukaryotic counterparts. Because PhrB

homologs of the FeS-BCP type are widely distributed in pro-karyotes, structural considerations lead us to suggest that (6-4)photorepair plays a significant role in bacteria and archaea.The structural homology between PhrB and the large subunitof eukaryotic and archaeal primases (PriL), sheds unique lighton the evolution of the CPF.

Results and DiscussionFunction of PhrB. Previous DNA repair assays with PhrA and PhrBbased on restriction digest and electrophoresis showed that PhrAbut not PhrB repairs CPD photoproducts in vitro. A spectral assayin which the specific absorption of the (6-4) photoproduct at 325nm was followed photometrically gave inconclusive results due tospectral overlap with the PhrB chromophores. In the present work,we used HPLC-purified single stranded (ss) or double stranded(ds) DNA oligonucleotides with a thymine-thymine (TT) (6-4)lesion (8) as substrate and monitored DNA repair by HPLC,combined with continuous recording of UV-visible (UV-vis)spectra. As shown in Fig. 1 A and B and Fig. S1, PhrB incubationwith ssDNA and dsDNA TT (6-4) photoproducts resulted ina characteristic mobility change on the HPLC column and a lossof the 325-nm absorption peak. The mobility of the product wasalways identical with that of control DNA without the lesion. Thisrepair of ssDNA and dsDNA (6-4) photoproducts was blue-lightand incubation-time dependent, as expected for a typical photo-lyase (9).We have thus shown that PhrB is a (6-4) photolyase. Thisfinding strongly suggests that prokaryotes are also capable ofthe photolyase-catalyzed repair of (6-4) lesions.Considering that PhrB homologs share the same function, the

(6-4) photorepair should be broadly distributed among bacteria. Ina Uniprot-KB BLAST survey, we found 462 bacterial and archaealPhrB homologs (FeS-BCPs), which form a group separate fromother CPFs (Table S1). In the catalytic region between amino acids276 and 474 that encompasses the FADbinding and putativeDNAbinding regions, residues are highly conserved. Therefore, wepropose that most if not all FeS-BCP members are (6-4)

Author contributions: F.Z., P.S., I.O., T.L., and N.K. planned the project; I.O. optimized thePhrB purification procedure; F.Z. and P.S. conducted the PhrB crystallization screening andoptimization; F.Z. and I.O. performed the HPLC-based antenna chromophore analyses;F.Z. performed the DNA repair assays; P.S. cryocooled the crystals; P.S. and N.K. collectedthe X-ray diffraction data; P.S. performed data processing and solved the PhrB structure;P.S. and N.K. refined the PhrB structure; F.Z. and T.L. performed the alignments; T.L.performed phylogenetic studies; F.Z., P.S., I.O., T.L., and N.K. analyzed data; and T.L.and N.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID code 4DJA).1F.Z. and P.S. contributed equally to this work.2To whom correspondence may be addressed. E-mail: tilman.lamparter@kit.edu orn.krauss@qmul.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1302377110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1302377110 PNAS | April 30, 2013 | vol. 110 | no. 18 | 7217–7222

BIOCH

EMISTR

Y

photolyases, which is further supported by detailed analysis ofsequences and in context with our crystal structure.Another FeS-BCP member that has been investigated is

cryptochrome B (CryB) of Rhodobacter sphaeroides. CryB hasbeen identified as a photoreceptor that regulates photosynthesisgenes (10). Tests for in vitro DNA repair activity were negative(11), but this could result from inappropriate in vitro conditions.Several features are in support of a repair activity of CryB. It isrequired for the full in vivo photorepair in R. sphaeroides (12),as is PhrB in A. tumefaciens (7). Surface charge and DNA bindingproperties of CryB (10, 11) are also compatible with DNArepair activity.

Crystal Structure of PhrB. We determined the crystal structure ofPhrB at 1.45 Å resolution, the best resolution for a CPF proteinstructure published so far. The presence of the [4Fe-4S] cluster wasconfirmed by a strong anomalous scattering signal in the X-ray

diffraction data collected at the iron absorption edge (Table S2).The refined model (Fig. 1C) closely matches (rmsd= 1.3 Å for 502Cα positions) the structure of CryB (11). Superpositions with othermembers of the CPF yielded rmsds between 2.8 and 3.5 Å (13–22).Remarkably, the two largest rmsds of 3.2 Å and 3.5 Å wereobtained with Drosophila melanogaster and Arabidopsis thaliana(6-4) photolyases [Drome (6-4) PL andArath (6-4) PL], respectively.In the phylogenetic analysis, the prokaryotic and eukaryotic (6-4)photolyases appear as the most distantly related proteins withinCPF (Fig. S2A). Besides the FAD chromophore, which assumesa prototypical U-shaped conformation, another organic cofactorwas identified. Its electron density does not match with knownantenna chromophores of other photolyases, but could be fitted toDMRL (Fig. 2A). We confirmed its identity by HPLC analysis. Inthe elution profiles, themajor peak was found at the same positionas that of synthetic DMRL and the UV-Vis spectral properties ofboth compounds are identical (Fig. S3). The same DMRL

Fig. 1. Function and crystal structure of PhrB. (A) HPLC analysis of PhrB (6-4) photoproduct repair with single stranded DNA. The “standard” is a mixture ofdamaged DNA with a TT (6-4) lesion and undamaged DNA (black dotted line). In the other samples, damaged DNA was incubated with PhrB for 60 and120 min under blue light (BL, purple and magenta lines, respectively) or 60 min in darkness (orange line). Thereafter, DNA was separated from the protein andsubjected to HPLC. (B) HPLC analysis of PhrB (6-4) photoproduct repair with double stranded DNA. The standard is a mixture of damaged DNA with a TT (6-4)lesion and undamaged DNA (black dashed line). In the other samples, damaged DNA was incubated with PhrB for 5 and 15 min under blue light (BL, blue andred lines, respectively) or 15 min in darkness (green line). Thereafter, DNA was separated from the protein and subjected to HPLC. (C) Overall structure andcofactor arrangement with distances of PhrB. The ribbon representation shows the α/β-domain (N-terminal antenna binding domain; blue) and the α-helicaldomain (catalytic domain; green) connected by a long interdomain linker (orange). The cofactors DMRL, FAD, and the [4Fe-4S] cluster are illustrated in ball-and-sticks mode and as enlarged view (Lower).

Fig. 2. DMRL binding site. (A) DMRL chromophore with σA-weighted 2Fo-Fc electron density map contoured at 1.0σ (blue mesh) and residues in potentialhydrogen bonding or van der Waals contacts (Ile51 is omitted for clarity). Hydrogen bonds between DMRL and surrounding amino acids shown by blackdotted lines, van der Waals contacts shown in red. (B) Antenna chromophores of photolyases superimposed onto the PhrB structure. Overall view and close-upview of antenna chromophores. PhrB is illustrated in ribbon representation and α-helices of PhrB are shown as cylinders (green). FAD and the DMRL antenna ofPhrB are drawn as ball and sticks, the corresponding antenna chromophores are shown as sticks. Homologous binding pockets are used by 8-HDF in Anacystisnidulans CPD photolyase (PDB code 1TEZ), 8-HDF in Thermus thermophilus CPD photolyase (PDB code 2J07), 8-HDF in Drosophila melanogaster (6-4) pho-tolyase (soaked chromophore, PDB code 3CVV), or the isoalloxazine ring of FAD in Sulfolobus tokodaii CPD photolyase (PDB code 2E0I). A different bindingpocket is used by methenyltetrahydrofolate in Escherichia coli CPD photolyase (PDB code 1DNP) or MTHF in Arabidopsis thaliana Cry3 (PDB code 2VTB).

7218 | www.pnas.org/cgi/doi/10.1073/pnas.1302377110 Zhang et al.

cofactor was also found at a homologous binding site in CryB (11).Another feature in common with CryB is the strict structuralconservation of aromatic residues that define the intraproteinelectron transfer route for the FAD photoreduction (11).

His365-His366-X-X-Arg369 Motif. In a structural alignment (23) ofPhrB and two eukaryotic (6-4) photolyases (15, 22), we haveidentified 34 identical amino acids, 10 of which are conserved to99–100% in all FeS-BCPs. Three of these amino acids cluster to-gether in the His365-His366-X-X-Arg369 motif. These aminoacids are located at the bottom of the DNA lesion’s entrancechannel. His365 and Arg369 are conserved in (6-4) photolyasesand in many CPD homologs. His366 is conserved in FeS-BCPs, ineukaryotic (6-4) photolyases, and in animal cryptochromes; CPDphotolyases have an Asn at the homologous position. The His366homolog of eukaryotic (6-4) photolyases is essential for DNA re-pair; it serves as a donor/acceptor in catalytic proton transfer,which accompanies the light-driven electron transfer (22, 24).His366 in PhrB is stabilized by van derWaals contacts with Leu370and Met410 (Fig. 3A), whereas the homologous His in eukaryotic(6-4) photolyases is stabilized by a hydrogen bonding network (15,22). The orientation of His366 in PhrB, however, matches with thehomologous His residue of Drome (6-4) PL (Fig. 3A), suggestingfunctional identity despite the different environments of theseresidues. Gln306, conserved in FeS-BCPs and eukaryotic (6-4)photolyases, is likely to form hydrogen bonds to the DNA lesion(Fig. 3A and Table S3). This residue has a conserved Glu as itscounterpart in CPD photolyases. Overall, already the high degree

of conservation within FeS-BCPs of amino acid residues thatare probably in contact with the DNA lesion in PhrB (Fig. 3Aand Table S3) supports the idea that all FeS-BCPs are (6-4)photolyases.

(6-4) Photoproduct Binding. According to a DNA–protein in-teraction model that was obtained using the Hex-Server (http://hexserver.loria.fr) (25), eight amino acids of PhrB that are allconserved within FeS-BCPs interact with the DNA lesion (TableS3). One of these, Arg183, is part of the loop region connecting α7and α8; this loop is longer in FeS-BCPs than in other CPFs. ThreeDNA-interacting amino acids of this loop, Ala180 to Asn182, arenot visible in the electron density map (Fig. 4 A and B). Theseamino acids might be stabilized by the DNA to acquire a moreordered conformation. Arg183 could change its orientation to in-teract via its positively charged guanidinium group by forming a saltbridge with the phosphate within the DNA lesion, thereby stabi-lizing the flip out of the lesion (Fig. 4A andB). In this respect, PhrBdiffers from prototypical CPFs, in which another Arg residue[Arg421 in Drome (6-4) PL] within the region corresponding to theα17–α18 connecting loop in PhrB stabilizes the flip out of the lesion(Fig. 4 C and D) (17, 20, 26). This Arg is missing in FeS-BCPs. Wetherefore suggest that FeS-BCPs bind UV-damaged DNA ina mode significantly different from prototypical photolyases.

C-Terminal Extension of PhrB. Prototypical photolyases have a com-mon set of helices corresponding to α1–α19 in PhrB, and the lastcommon helix α19 forms a part of the DNA binding groove. Arath(6-4) PL has an additional α-helix (15), referred to as C-terminalextension. This helix forms an antiparallel two-helix bundle withα19 and (assuming a prototypical binding mode) faces away frombound DNA. The C terminus of PhrB is extended by two addi-tional helices α20 and α21 (Fig. S4), of which α21 displays se-quence similarity with the C-terminal helix α20 of Arath (6-4) PL.Sequence alignments show that α20 and α21 are present in allFeS-BCPs. According to a PhrB–DNA interaction model, whichis compatible with a prototypical photolyase–DNA binding mode,helix α20 of PhrB interacts with the DNA, whereas α21 faces awayfrom the DNA. The surface region of α20 displays a positivecharge (Fig. S5) that is consistent with its predicted role in DNAbinding. Because the C-terminal extension is absent in Drome (6-4)PL and most other photolyases, we suppose that it is dispensablefor repair activity and consider a regulatory role for this region.

Chromophore Binding. In the neutral radical (FADH•) and theanionic fully reduced (FADH−) states of the FAD cofactor inprototypical photolyases, the N5 atom of the isoalloxazine moietyis protonated and hydrogen bonded to an Asn residue (17, 27).Apparently, this stabilizing function is taken over by a conservedwater molecule in PhrB that is held in place by additional hy-drogen bonds with the side chain of Arg369 and the backbonecarbonyl group of Tyr391 (Fig. 3B). The binding sites for thedicyclic DMRL in PhrB and for the tricyclic antenna chromophores8-hydroxy-7,8-didemethyl-5-deazariboflavin, FAD, or FMN inother photolyases (14, 21, 28) are in homologous positions (Fig.2B). His43 of PhrB prevents binding of tricyclic cofactors by sterichindrance, and is in van der Waals contact with the two methylgroups of DMRL (Fig. 2A). This residue is 67% conserved in FeS-BCPs, and is replaced by bulky Tyr and Phe residues in other FeS-BCPs. We therefore predict that none of these proteins bindsa tricyclic antenna chromophore.

[4Fe-4S] Cluster of PhrB. The [4Fe-4S] cluster of PhrB is co-ordinated by Cys350, Cys438, Cys441, and Cys453. Structuralhomology between the carboxyl-terminal domain of PriL (PriL-CTD) of eukaryotic and archaeal primases that also contain[4Fe-4S] clusters and CPF members suggests that both groups ofproteins have a common ancestor (29). Structural similarity canbe seen between PhrB and PriL for a stretch from amino acids348–434 of PhrB, which comprises helices α13–α19 (Fig. 5A; the

Fig. 3. Catalytic center. (A) The active site of PhrB. Sections of Drome (6-4) PLin complexwith photodamagedDNA (PDB code 3CVU, DNA inmagenta, aminoacids in purple, and FAD in blue) and of PhrB (amino acids in green and FAD inyellow) are shown after superposition of the FAD molecules. Amino acid resi-dues inDrome (6-4) PL, which interactwith theDNA lesion and the homologousresidues in PhrB, are shown. Potential hydrogen bonds are represented bydashed lines. (B) WaterW731 of PhrB (green) forms a potential hydrogen bondwith theN5atomof FAD (Catoms in yellow), replacing the roleofAsn378of theclass I CPD photolyase from E. coli (PDB code 1DNP), drawn in purple. Its PhrBhomolog, Glu403, faces away from FAD and interacts with His388.

Zhang et al. PNAS | April 30, 2013 | vol. 110 | no. 18 | 7219

BIOCH

EMISTR

Y

rmsd for PhrB and Saccharomyces cerevisiae PriL [Protein DataBank (PDB) code 3LGB] is 3.1 Å). The structurally most con-served core region stretches from amino acids 348 to 398. Theiron-sulfur clusters are found at almost the same positions in thesuperimposed structures, but are tilted against each other (Fig.5B). This is consistent with the fact that only the first cysteines inthe core regions of both structures are strictly conserved. Theside chains of the remaining three Cys residues are located onlyat roughly similar positions, and the sequential order of corre-sponding cysteines differs (PriL: 336/417/434/474; PhrB: 350/454/438/441). We therefore assume that outside the core region the

protein fold diverged and the coordinating Cys residues wereexchanged during evolution.Besides primases and FeS-BCPs, several other DNA-interacting

proteins contain Fe-S clusters (30–32). Fe-S clusters are regardedas ancient features. Their typical function is electron transport, andthey are also often required for protein folding and stability (30).The Fe-S cluster of human PriL is specifically required for initiationof primer synthesis, but the molecular details of its function are asyet unknown (33). For the base excision repair Mutator Y (MutY,a DNA glycosylase) and Endonuclease III (EndoIII), a chargetransfer between DNA and the Fe-S cluster has been proposed tobe involved in detecting DNA lesions (34). The present study does

Fig. 4. Potential role of the PhrB α7–α8 loop region in binding of the (6-4) DNA lesion in comparison with eukaryotic Drome (6-4) PL. PhrB and Drome (6-4) PLare illustrated in ribbon representation and α-helices are shown as cylinders in green and brown, respectively. The α7–α8 loop is drawn in black; the (6-4) lesionstabilizing loop of Drome (6-4) PL is drawn in yellow. This loop in Drome (6-4) PL is located between helices that correspond to α17 and α18 in PhrB andstabilizes the flip out of the (6-4) photoproduct by inserting a pair of amino acid side chains (Gln418 and Arg421) into the resulting gap in the double strandedDNA oligomer. (A) The DNA fragment with the (6-4) lesion (shown as ball and sticks) from the Drome (6-4) PL–DNA complex (PDB code 3CVU) superimposedonto the PhrB structure. (B) A close-up view of A. Two side chain conformations shown for Arg183 (as sticks), the original one as seen in the crystal structure(yellow) and a second one (brown) being a rotamer that is in close contact (<4 Å) with the central phosphate as indicated by the dotted lines. (C) The Drome(6-4) PL complex with photodamaged DNA oligomer (PDB code 3CVU). DNA is drawn as sticks and the (6-4) lesion highlighted as ball and sticks. Only a portionof the DNA oligomer is shown for clarity. (D) A close-up view of C. Arg421 forms a salt bridge with the DNA (6-4) lesion.

Fig. 5. Structural comparison between PhrB and the primase S. cerevisiae large subunit PriL. Overall structural comparison of PhrB and the PriL carboxyl-terminal domain (PriL-CTD) are drawn in green and blue, respectively; coordinating Cys residues and Fe-S clusters in orange/yellow and black/red, respectively.Numbers of Cys residues and α-helices refer to PhrB. (A) Superposition of PhrB and PriL–CTD (Sacce PriL, PDB Code 3LGB) based on Cα atoms, yielding an rmsdof 3.1 Å for 112 Cα atoms. The homologous regions are drawn in ribbon representation; the rest of the protein is transparent. (B) Close-up view of [4Fe-4S]clusters, coordinating Cys residues and relevant portions of the protein backbones.

7220 | www.pnas.org/cgi/doi/10.1073/pnas.1302377110 Zhang et al.

not yet provide insight into the function of the Fe-S cluster of PhrBapart from it being an integral component of the protein fold, butknowledge of the high-resolution structure of PhrB might help tosolve this task in the future.

Evolutionary Scenario of CPF. To gain a deeper insight into evolu-tionary relationships between primases and CPF, we performedphylogenetic studies based on structurally aligned sequences,using parsimony and distance-based algorithms. Irrespective ofwhether the entire homologous region (348–434) or the moreconserved core region (348–398) were used as input, primases andPhrB were placed in one monophyletic group (Fig. S2B). Weconclude that the primases, FeS-BCPs, and other CPFs divergedearly in the evolution and that the first common ancestor of CPFwas a (6-4) photolyase with an Fe-S cluster. The Fe-S cluster waslost in the early evolution of non–FeS-BCP CPF members. Inseveral subsequent independent steps, antenna chromophoreswere replaced, (6-4) repair changed to CPD repair, and DNArepair activity was replaced by the signal transduction that rep-resents a transition from photolyases to cryptochromes (Fig. 6).Photolyases are regarded as ancient DNA repair enzymes al-

ready present before the formation of the oxygen-rich atmosphere(35), and our results imply that the (6-4) photorepair occurredearly in the evolution of photolyases. However, CPD repair mighthave also evolved so early. Most essential features such as DNAbinding, the cavity for the DNA lesion, photoreduction, an FADchromophore close to that cavity, and the overall protein fold areidentical in CPD and (6-4) photolyases. The combination of CPDand (6-4) photorepair provides a major evolutionary advantage.Our survey suggests that a large number of prokaryotes livingtoday still rely on the combined photorepair of CPD and (6-4)photoproducts: 80% of those prokaryotes that contain an FeS-BCPmember (Table S1) also contain a CPD photolyase homolog.

Materials and MethodsProtein Purification and Crystallization. Recombinant His-tagged PhrB proteinwas expressed in Escherichia coli and purified via Ni-affinity chromatographyand size-exclusion chromatography as described (7) andused for crystallizationat concentrations of 4–6 mg/mL under dim light conditions or in darkness.Crystallization screens were carried out by the sparse matrix method (36) ina sitting-drop vapor diffusion approach testingmore than 1,000 crystallizationconditions at 277 K and 298 K in 96-well Medical Research Council (MRC)plates. Promising conditions were systematically screened further by changing

protein concentration, pH, and the concentration of precipitation agents.After optimization, crystals were grown by hanging drop vapor diffusion at298 K in 24-well Linbro plates. Each hanging drop was prepared on a silicon-ized glass plate bymixing 5 μL protein solutionwith 5 μL reservoir solution. Theprotein solution contained 12.5 mM Tris (hydroxymethyl) amino methane,1.25 mM EDTA, 2.5% (vol/vol) glycerol, 75 mM sodium chloride, pH 7.8. Thereservoir solution contained 2–6% (wt/vol) polyethylene glycol 400, 100 mM2-(N-morpholino) ethanesulfonic acid buffer, pH 6.0. Yellow rectangular PhrBcrystals appeared within 3 d and grew further for 7 d. PhrB crystals were flashfrozen in liquid nitrogen after cryosoakingwith 4M trimethylamineN-oxide fora fewminutes (37). Fully grown crystals had dimensions of∼0.9 × 0.3 × 0.3mm3.

Structure Analysis. Diffraction data collection was performed at 100 K usingsynchrotron X-ray sources at the European Synchrotron Radiation Facility(Grenoble, France) and the Berlin Radiation Synchroton Source II (BESSY II)(Berlin, Germany). Best diffraction data for optimized anomalous dispersionand highest resolution were collected at beamline BL 14.2 (38) at BESSY II/Helmholtz Zentrum Berlin für Materialien und Energie with an MAR-225CCDdetector at λ = 1.7409 Å and λ = 0.91841 Å, respectively. All images wereindexed, integrated, and scaled using the XDS program package (39) and theCCP4 program SCALA (40, 41). The single-wavelength anomalous dispersion(SAD) dataset at λ = 1.7409 Å (peak) was collected to 1.95 Å resolution. Initialexperimental phaseswere determined to 1.95 Å resolution by the SADmethodbased on this dataset by using SHELXC/D (42) and a beta version of SHELXE (43)to solve the iron-atom substructure. Substructure solution with SHELXD wassuccessful for space group P21212, yielding four iron sites (with occupanciesgreater than 0.62). With this solution a mean figure of merit of 0.640 wasobtained after initial phasing, polyalanine-chain tracing, and density modifi-cationwith SHELXE (43) that increased to 0.770 after densitymodificationwiththe program DM (44). The electron density mapwas readily interpretable, andthe phases were input for initial automated model building using ARP/wARP7.0.2 (45). Subsequent steps of crystallographic refinement, consisting of tor-sion angle molecular dynamics, simulated annealing using a slow-coolingprotocol and amaximum likelihood target function, energyminimization, andB-factor refinement by the program CNS (46) were carried out using a high-resolution native dataset in the resolution range of 19.95–1.45 Å. The [4Fe-4S]cluster, the FAD, and 6,7-dimethyl-8-ribityllumazine were clearly visible in theelectrondensity of a σA−weighted 2Fo− Fcmapalready after thefirst roundofrefinement, as well as in the σA − weighted simulated annealing omitteddensitymaps that were calculated in thefinal stage of refinement. Restrained,individual B-factors were refined and the crystal structure was finalized usingthe CCP4 program REFMAC5, another program of the CCP4 suite (40), andPHENIX (47). The final model has agreement factors Rfree and Rcryst of 18.0%and 13.8%. Table S2 summarizes the statistics for crystallographic data col-lection and structural refinement. Manual rebuilding of the PhrB model andelectron density interpretation was performed after each refinement cycleusing the program COOT (48) and WHAT_CHECK (49). Potential hydrogenbonds and van der Waals contacts were analyzed using the programs HBPLUS(50) and LIGPLOT (51). All crystal structure superpositions of backbone α carbontraces were performed using the DALI server (23). The PhrB structure was an-alyzed for salt bridges using the WHAT IF server (52). Electrostatic surfacepotentials were calculated through solution of the Poisson–Boltzmann equa-tion using the programAPBS (53). All molecular graphics representationswerecreated using PyMOL (54).

Analytical HPLC. An Agilent 1200 Series HPLC system managed by ChemStation software, equipped with a diode array detector (DAD) (Agilent) anda Nucleosil 100–5 analytical column (250 × 4.6 mm; Macherey and Nagel) wasused for analysis of organic cofactors and the single stranded DNA (6-4)photoproduct repair assay, whereas equipped with a Nucleodur 100–3 col-umn (250 × 4.0 mm; Macherey and Nagel) for the double stranded DNA (6-4)photoproduct repair assay.

A protocol established for Rhodobacter CryB (10) was used for analyses oforganic cofactors: Purified PhrB was mixed with 7.2% trichloroacetic acid(TCA) followed by 1 -h shaking at 100 rpm on ice, and then mixed with 5%(vol/vol) of 0.1% formic acid (HCO2H) in acetonitrile (ACN) and 0.35 mMNaOH (final concentrations). The mixture was centrifuged at 13,000 × g for10 min, and the supernatant was analyzed by HPLC. Synthetic DMRL (agenerous gift from A. Bacher, Technical University Munich, Germany) andFAD were used as references. For the mobile phase, we used solution A(0.1% HCO2H in ACN) and solution B (0.1% HCO2H in H2O). After injection of10-μL samples, separations were achieved at 298 K and a flow rate of 1 mL/min. A 5-min isocratical run with 95% (vol/vol) of solution B was followed bya gradient from 95% to 25% solution B within 20 min. Elution was moni-tored at 264 nm and UV-Vis spectra were automatically recorded and stored.

Fig. 6. Two possible evolutionary pathways from the ancestor of primase PriLand CPFs to present PriL and CPF group members. Proposed major functionalchanges are indicated by colored dots. The Lower tree has fewer functionalchanges. In this preferred scenario, thefirst photolyasewas a (6-4) photolyaseswith an iron-sulfur cluster. Eu(6-4)PL: eukaryotic (6-4) photolyases; Cry: cryp-tochromes; CPD PL: CPD photolyases; CryDASH: DASH-type cryptochromes.

Zhang et al. PNAS | April 30, 2013 | vol. 110 | no. 18 | 7221

BIOCH

EMISTR

Y

For the DNA (6-4) photoproduct repair assay, HPLC was conducted as de-scribed (55). The single stranded or double stranded DNA probe comprisingthe (AGGT(6-4)TGGC or GCGGT(6-4)TGGCG paired with TCGCCAACCGCT,100 μmol/L) was dissolved in the repair buffer (50 mM Tris·HCl, 100 mMNaCl, 5 mM DTT, 1 mM EDTA, 5% glycerol, pH 7.5) and incubated with PhrB(300 μmol/L in the repair buffer). The reaction mixture was irradiated witha 405-nm light emitting diode (GB333UV1C/L1, Farnell Diodes) at an intensityof 40 μmol·m−2·s−1. Controls in which the TT (6-4) lesion comprising probesincubated with PhrB in darkness were performed in parallel. The reaction wasstopped by heating to 368 K for 10 min and the sample was centrifuged at13,000 × g for 10 min. The supernatant was subjected to HPLC at a tempera-ture of 298 K for single strand or 333 K for double strand DNA. For compar-isons, untreated DNA probes with or without TT (6-4) lesion were processedanalogously. The mobile phase consisted of solution C [0.1 M triethylammo-nium acetate (TEAA) in H2O] and solution D (0.1 M TEAA in H2O/ACN 20/80).The gradient was 0–17% solution D in 60 min with a flow rate of 0.5 mL/min.Elution was monitored at 260 nm and 325 nm. Based on the retention timeand UV-Vis spectral properties, the TT (6-4) lesion-comprising strand was

easily distinguished from the repaired and complementary strand as well asfrom the standard undamaged DNA strand (Fig. S1).

ACKNOWLEDGMENTS. We thank Richard Pokorny for critical reading ofthe manuscript; Korbinian Heil and Thomas Carell for the gift of (6-4)photoproducts and for advice in photorepair assays; Markus Fischer andAdelbert Bacher for synthetic DMRL; Sybille Wörner and Ernst Heene fortechnical assistance; David von Stetten and Antoine Royant of the ID 29S-Cryobench [European Synchrotron Radiation Facility (ESRF), Grenoble,France] and the scientific staff of the ESRF at beamlines ID 14-1, ID 14-4, ID23-1, and ID 23-2; and Uwe Müller, Manfred Weiss, and the scientific staff ofthe BESSY-MX/Helmholtz Zentrum Berlin für Materialien und Energie atbeamlines BL 14.1 and BL 14.2 operated by the Joint Berlin MX-Laboratoryat the BESSY II electron storage ring (Berlin-Adlershof, Germany) where thedata were collected for continuous support. This work was supported by theDeutsche Forschungsgemeinschaft Cluster of Excellence “Unifying Conceptsin Catalysis” (to P.S.; Research Field D3), Sonderforschungsbereich 1078/B6 (toP.S.), Deutsche Forschungsgemeinschaft 799/8-1 (to T.L.), the China Scholar-ship Council (F.Z.), and Karlsruhe House of Young Scientists (F.Z.). P.S.acknowledges K. P. Hofmann and his advanced investigator European Re-search Council Grant (ERC-2009/249910TUDOR) for support.

1. Sancar A (2003) Structure and function of DNA photolyase and cryptochrome blue-light photoreceptors. Chem Rev 103(6):2203–2237.

2. Cashmore AR (2003) Cryptochromes: Enabling plants and animals to determine cir-cadian time. Cell 114(5):537–543.

3. Chaves I, et al. (2011) The cryptochromes: Blue light photoreceptors in plants andanimals. Annu Rev Plant Biol 62:335–364.

4. Essen LO (2006) Photolyases and cryptochromes: Common mechanisms of DNA repairand light-driven signaling? Curr Opin Struct Biol 16(1):51–59.

5. Müller M, Carell T (2009) Structural biology of DNA photolyases and cryptochromes.Curr Opin Struct Biol 19(3):277–285.

6. Gelvin SB (2008) Agrobacterium-mediated DNA transfer, and then some. Nat Bio-technol 26(9):998–1000.

7. Oberpichler I, et al. (2011) A photolyase-like protein from Agrobacterium tumefacienswith an iron-sulfur cluster. PLoS ONE 6(10):e26775.

8. Glas AF, Schneider S, Maul MJ, Hennecke U, Carell T (2009) Crystal structure of theT(6-4)C lesion in complex with a (6-4) DNA photolyase and repair of UV-induced(6-4) and Dewar photolesions. Chemistry 15(40):10387–10396.

9. Sancar A (2004) Photolyase and cryptochrome blue-light photoreceptors. Adv ProteinChem 69:73–100.

10. Hendrischk AK, et al. (2009) A cryptochrome-like protein is involved in the regulationof photosynthesis genes in Rhodobacter sphaeroides. Mol Microbiol 74(4):990–1003.

11. Geisselbrecht Y, et al. (2012) CryB from Rhodobacter sphaeroides: A unique class ofcryptochromes with new cofactors. EMBO Rep 13(3):223–229.

12. Hendrischk AK, Braatsch S, Glaeser J, Klug G (2007) The phrA gene of Rhodobactersphaeroides encodes a photolyase and is regulated by singlet oxygen and peroxide ina sigma(E)-dependent manner. Microbiology 153(Pt 6):1842–1851.

13. Brudler R, et al. (2003) Identification of a new cryptochrome class. Structure, function,and evolution. Mol Cell 11(1):59–67.

14. Fujihashi M, et al. (2007) Crystal structure of archaeal photolyase from Sulfolobustokodaii with two FAD molecules: Implication of a novel light-harvesting cofactor. JMol Biol 365(4):903–910.

15. Hitomi K, et al. (2009) Functional motifs in the (6-4) photolyase crystal structure makea comparative framework for DNA repair photolyases and clock cryptochromes. ProcNatl Acad Sci USA 106(17):6962–6967.

16. Huang YH, et al. (2006) Crystal structure of cryptochrome 3 from Arabidopsis thalianaand its implications for photolyase activity. Proc Natl Acad Sci USA 103(47):17701–17706.

17. Kiontke S, et al. (2011) Crystal structures of an archaeal class II DNA photolyase and itscomplex with UV-damaged duplex DNA. EMBO J 30(21):4437–4449.

18. Komori H, et al. (2001) Crystal structure of thermostable DNA photolyase: Pyrimidine-dimer recognition mechanism. Proc Natl Acad Sci USA 98(24):13560–13565.

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

20. Pokorny R, et al. (2008) Recognition and repair of UV lesions in loop structures of duplexDNA by DASH-type cryptochrome. Proc Natl Acad Sci USA 105(52):21023–21027.

21. Tamada T, et al. (1997) Crystal structure of DNA photolyase from Anacystis nidulans.Nat Struct Biol 4(11):887–891.

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

23. Holm L, Rosenström P (2010) Dali server: Conservation mapping in 3D. Nucleic AcidsRes 38(Web Server issue):W545-9.

24. Li J, et al. (2010) Dynamics and mechanism of repair of ultraviolet-induced (6-4)photoproduct by photolyase. Nature 466(7308):887–890.

25. Macindoe G, Mavridis L, Venkatraman V, Devignes MD, Ritchie DW (2010) HexServer:An FFT-based protein docking server powered by graphics processors. Nucleic AcidsRes 38(Web Server issue):W445–W449.

26. Mees A, et al. (2004) Crystal structure of a photolyase bound to a CPD-like DNA lesionafter in situ repair. Science 306(5702):1789–1793.

27. Xu L, et al. (2008) Active site of Escherichia coli DNA photolyase: Asn378 is crucial bothfor stabilizing the neutral flavin radical cofactor and for DNA repair. Biochemistry47(33):8736–8743.

28. Klar T, et al. (2006) Natural and non-natural antenna chromophores in the DNAphotolyase from Thermus thermophilus. ChemBioChem 7(11):1798–1806.

29. Sauguet L, Klinge S, Perera RL, Maman JD, Pellegrini L (2010) Shared active site ar-chitecture between the large subunit of eukaryotic primase and DNA photolyase.PLoS ONE 5(4):e10083.

30. White MF, Dillingham MS (2012) Iron-sulphur clusters in nucleic acid processing en-zymes. Curr Opin Struct Biol 22(1):94–100.

31. Wu Y, Brosh RM, Jr. (2012) DNA helicase and helicase-nuclease enzymes with a con-served iron-sulfur cluster. Nucleic Acids Res 40(10):4247–4260.

32. Netz DJ, et al. (2012) Eukaryotic DNA polymerases require an iron-sulfur cluster forthe formation of active complexes. Nat Chem Biol 8(1):125–132.

33. Klinge S, Hirst J, Maman JD, Krude T, Pellegrini L (2007) An iron-sulfur domain of theeukaryotic primase is essential for RNA primer synthesis. Nat Struct Mol Biol 14(9):875–877.

34. Boal AK, et al. (2009) Redox signaling between DNA repair proteins for efficient le-sion detection. Proc Natl Acad Sci USA 106(36):15237–15242.

35. Eisen JA, Hanawalt PC (1999) A phylogenomic study of DNA repair genes, proteins,and processes. Mutat Res 435(3):171–213.

36. Jancarik J, Kim SH (1991) Sparse matrix sampling: A screening method for crystalli-zation of proteins. J Appl Cryst 24:409–411.

37. Mueller-Dieckmann C, Kauffmann B, Weiss MS (2011) Trimethylamine N-oxide as aversatile cryoprotective agent in macromolecular crystallography. J Appl Cryst 44:433–436.

38. Müller U, et al. (2012) Facilities for macromolecular crystallography at the Helmholtz-Zentrum Berlin. J Synch Rad 19:442–449.

39. Kabsch W (2010) XDS. Acta Crystallogr D Biol Crystallogr 66(Pt 2):125–132.40. Bailey S; Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Pro-

grams for protein crystallography. Acta Crystallogr D Biol Crystallogr 50(Pt 5):760–763.41. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D Biol Crys-

tallogr 62(Pt 1):72–82.42. Sheldrick GM (2008) A short history of SHELX. Acta Crystallogr A 64(Pt 1):112–122.43. Sheldrick GM (2010) Experimental phasing with SHELXC/D/E: Combining chain tracing

with density modification. Acta Crystallogr D Biol Crystallogr 66(Pt 4):479–485.44. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.

Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.45. Langer G, Cohen SX, Lamzin VS, Perrakis A (2008) Automated macromolecular model

building for X-ray crystallography using ARP/wARP version 7. Nat Protoc 3(7):1171–1179.46. Brünger AT, et al. (1998) Crystallography & NMR system: A new software suite for

macromolecular structure determination. Acta Crystallogr D Biol Crystallogr 54(Pt 5):905–921.

47. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221.

48. Emsley P, Cowtan K (2004) Coot: model-building tools for molecular graphics. ActaCrystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2126–2132.

49. Hooft RWW, Vriend G, Sander C, Abola EE (1996) Errors in protein structures. Nature381(6580):272–272.

50. McDonald IK, Thornton JM (1995) The application of hydrogen bonding analysis inX-ray crystallography to help orientate asparagine, glutamine and histidine sidechains. Protein Eng 8(3):217–224.

51. Wallace AC, Laskowski RA, Thornton JM (1995) LIGPLOT: A program to generateschematic diagrams of protein-ligand interactions. Protein Eng 8(2):127–134.

52. Rodriguez R, Chinea G, Lopez N, Pons T, Vriend G (1998) Homology modeling, modeland software evaluation: Three related resources. Bioinformatics 14(6):523–528.

53. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA (2001) Electrostatics of nano-systems: Application to microtubules and the ribosome. Proc Natl Acad Sci USA 98(18):10037–10041.

54. DeLano WL (2002) The PyMOL Molecular Graphics System (DeLano Scientific, SanCarlos, CA).

55. Glas AF, et al. (2010) DNA (6-4) photolyases reduce Dewar isomers for isomerizationinto (6-4) lesions. J Am Chem Soc 132(10):3254–3255.

7222 | www.pnas.org/cgi/doi/10.1073/pnas.1302377110 Zhang et al.

Supporting InformationZhang et al. 10.1073/pnas.1302377110

Fig. S1. UV-visible (UV-Vis) spectra of HPLC fractions as recorded by the HPLC-coupled diode array detector (DAD) detector. (A) Spectra from the repair assaywith single stranded DNA under blue light for 120 min (Fig. 1A). (B) Spectra from the repair assay with double stranded DNA under blue light for 15 min (Fig. 1B).

Fig. S2. Phylogenetic trees based on structural alignments. Both trees were constructed with the PROTPARS program of the PHYLIP program package (1).Bootstrap values (in percent) are printed along the branches if <95%. (A) Analysis based on a structural alignment of full-length cryptochrome/photolyasefamilies (CPFs). The weak relationship between photolyase related protein B (PhrB) and eukaryotic (6-4) photolyases is also obtained with maximum likelihoodand distance-based algorithms. (B) Analysis based on a structural alignment of the protein fold common to primase large subunit (PriL) and CPFs, core region(amino acid residues 348–398 of PhrB). PhrB and PriL are placed in one monophyletic group, a result that is also obtained with distance-based algorithms, andwith the same algorithms and a longer homologous stretch (amino acids 333–465). Color code: red, (6-4) photolyases; cyan, CPD class I photolyases; green, CPDclass II photolyases; black, plant cryptochromes and Cry-DASH proteins; blue, primase PriL subunits. AgrtuPhrB: (6-4) photolyase of Agrobacterium tumefaciens,Protein Data Bank (PDB) code 4DJA; Anani CPD I: CPD class I photolyase of Anacystis nidulans, PDB code 1OWL; ArathCry1: cryptochrome 1 of Arabidopsisthaliana, PDB code 1U3C; ArathCry3: cryptochrome 3 (Cry-DASH) of Arabidopsis thaliana, PDB code 2J4D; Arath (6-4) PL: (6-4) photolyase of Arabidopsisthaliana, PDB code 3FY4; Drome (6-4) PL, (6-4) photolyase of Drosophila melanogaster, PDB code 3CVW; Escco CPD I: CPD class I photolyase of Escherichia coli,PDB code 1DNP; HomsaPriL: DNA primase large subunit of Homo sapiens, PDB code 3Q36; Metma CPD II: CPD class II photolyase of Methanosarcina mazei, PDBcode 2XRZ; SaccePriL: DNA primase large subunit of Saccharomyces cerevisiae, PDB code 3LGB; Sulto CPD I: CPD class I photolyase of PDB code Sulfolobustokodaii, PDB code 2E0I; SynspDASH: Cry-DASH of Synechocystis strain PCC 6083, PDB code 1NP7; Theth CPD I: Thermus thermophilus CPD class I photolyase,PDB code 1IQR.

1. Felsenstein J (2005) PHYLIP (Phylogeny Inference Package), Version 3.6. Available at http://evolution.genetics.washington.edu/phylip.html.

Zhang et al. www.pnas.org/cgi/content/short/1302377110 1 of 3

Fig. S3. The antenna chromophore of PhrB is 6,7-dimethyl-8-ribityllumazine (DMRL). All spectra are normalized to A264nm. (A) HPLC profiles, released PhrBcofactors (black line) and mixture of authentic FAD and DMRL (red line). Peaks 1 and 8 refer to DMRL; peak 9 could relate to a DMRL-like derivative. (B) UV-Visspectra of DMRL (-like) fractions (HPLC peaks 1, 8, and 9). (C) UV-Vis spectra of FAD (-like) fractions (HPLC peaks 4, 5, 7, 11, 12, and 13). (D) UV-Vis spectra of UV-absorbing compounds of unknown identity (HPLC peaks 2, 3, 6, and 10).

Fig. S4. Comparison of C-terminal helices in superimposed (6-4) photolyases. Drome (6-4) PL and Arath (6-4) PL are illustrated in ribbon representation in blueand orange, respectively; PhrB is drawn in ribbon representation and transparent surface in green. Helix α19 of PhrB is common to all three photolyases, Drome(6-4) PL ends with this helix. Arath (6-4) PL has one additional helix and PhrB has two additional helices, α20 and α21.

Zhang et al. www.pnas.org/cgi/content/short/1302377110 2 of 3

Fig. S5. Electrostatic surface representation of PhrB. Electrostatic surface potentials were calculated using the program APBS (1) with the nonlinear Poisson–Boltzmann equation and contoured at ±5kT/e. Negatively and positively charged surface areas are colored in red and blue, respectively. The cofactors DMRL,FAD, and Fe-S cluster are illustrated in ball-and-sticks representation. The DNA binding cavity is characterized by its positive charge; the helix α20 of theC-terminal extension with its positive surface charge appears to delimit this cavity.

1. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA (2001) Electrostatics of nanosystems: Application to microtubules and the ribosome. Proc Natl Acad Sci USA 98(18):10037–10041.

Other Supporting Information Files

Table S1 (DOCX)Table S2 (DOCX)Table S3 (DOCX)

1. Lovell SC, et al. (2003) Structure validation by Calpha geometry: Phi, psi and Cbeta deviation. Proteins 50(3):437–450.2. Macindoe G, Mavridis L, Venkatraman V, Devignes MD, Ritchie DW (2010) HexServer: An FFT-based protein docking server powered by graphics processors. Nucleic Acids Res 38(Web

Server issue):W445–W449.

Zhang et al. www.pnas.org/cgi/content/short/1302377110 3 of 3