Report

Crystal Structure of a Photolyase Bound to a CPD-Like DNA Lesion After in Situ Repair

See allHide authors and affiliations

Science  03 Dec 2004:
Vol. 306, Issue 5702, pp. 1789-1793
DOI: 10.1126/science.1101598

Abstract

DNA photolyases use light energy to repair DNA that comprises ultraviolet-induced lesions such as the cis-syn cyclobutane pyrimidine dimers (CPDs). Here we report the crystal structure of a DNA photolyase bound to duplex DNA that is bent by 50° and comprises a synthetic CPD lesion. This CPD lesion is flipped into the active site and split there into two thymines by synchrotron radiation at 100 K. Although photolyases catalyze blue light–driven CPD cleavage only above 200 K, this structure apparently mimics a structural substate during light-driven DNA repair in which back-flipping of the thymines into duplex DNA has not yet taken place.

Life under the sun is endangered by ultraviolet (UV) radiation that causes the formation of genotoxic photoproducts in DNA (1). Major UV-induced lesions include cis-syn cyclobutane pyrimidine dimers (CPDs) formed by a [2+2] cycloaddition of two adjacent pyrimidine bases, usually thymine. The importance of efficient repair systems for UV lesions is highlighted by hereditary diseases such as xeroderma pigmentosum. In prokaryotes, plants, and many animals, DNA photolyases (EC no. 4.1.99.3) are mainly responsible for the repair of CPD lesions by catalyzing the cleavage of the cyclobutane ring, using blue or near-UV light [absorbance (λ) of 360 to 500 nm] as the energy source (2, 3).

Despite three crystal structures of DNA photolyases (46) and the high affinity of photolyases toward DNA strands with cis-syn CPD lesions [dissociation constant KD ∼10–9 M (2)], it has proved difficult to obtain structural information on a substrate complex with CPD-comprising DNA. Thus, questions remain concerning the mechanism of CPD lesion recognition and binding. A long-standing hypothesis, supported by biochemical data (7, 8), computer modeling (6, 9, 10), and nuclear magnetic resonance (NMR) spectroscopy (11), states that photolyases flip the damaged dinucleotide out of the DNA double helix into their active site. After substrate binding, photon absorption by an antenna pigment (deazaflavin or methenyltetrahydrofolate) triggers transfer of the excitation energy to the catalytic flavin adenine dinucleotide, reduced state (FADH) cofactor (Fig. 1A). The excited cofactor then transmits an electron to the CPD lesion to induce splitting of the cyclobutane ring. The resulting radical anion then transfers back the excess electron to the FADH cofactor (semiquinone), closing the catalytic cycle within 0.5 to 2 ns after initial photon absorption (2).

Fig. 1.

(A) Mechanism of blue light–mediated repair of CPD lesions by DNA photolyases. Asterisks indicate an excited state of the flavin cofactor, hv indicates a blue-light or near-UV photon. (B) Ribbon model of complex A with the CPD-DNA in sticks representation and the 2Fobs-Fcalc electron density (contouring level, 1σ) that defines the duplex DNA (blue labels, CPD strand; red labels, counterstrand). The thymine dimer is highlighted in blue; the protruding α6 helix (red) contacts DNA along the minor groove. Apart from the adenine moiety of FAD (flavin adenine dinucleotide, purple), the cofactors are colored in yellow. Nomenclature and definition of secondary structure elements are given in (4).

To elucidate the recognition mode of CPD lesions, we crystallized a complex between the Anacystis nidulans DNA photolyase (4, 12) and a 14-nucleotide oligomer DNA duplex with a CPD analog in the central position (13). The synthetic CPD analog has the same cis-syn stereochemistry as natural CPD lesions and is efficiently photo-reactivated by DNA photolyases (14); however, it contains a formacetal linkage instead of the intradimer phosphate. The latter was highly useful in preparing the CPD lesion analog in quantities sufficient for crystallization studies. After data collection at a synchrotron beamline, structure solution, and refinement to 1.8Å resolution (Rfactor/Rfree: 0.206/0.226), the crystals, which comprise four photolyase/DNA complexes per asymmetric unit, revealed the DNA photolyase in two different states (fig. S1A): In two complexes, duplex DNA is bound to the enzyme (complexes A and B), whereas the other two complexes (C and D) show only short stretches of single-stranded DNA (13). Unless otherwise stated, the following structural analysis concerns complex A. Despite the extensive structural distortion of the duplex DNA upon binding to the DNA photolyase, the protein itself undergoes only minor changes, with a root mean square deviation (RMSD) of 0.583Å for 473 Cα-positions as compared to the uncomplexed enzyme. Marked differences are found only along the protein-DNA interface, which buries about 1216Å2 of molecular surface, rather little compared to other DNA repair enzymes (Fig. 1B).

Outside the site of lesion recognition, the DNA adopts a B-type conformation (Figs. 1B and 2A) and makes numerous interactions with the protein via its phosphodeoxyribose backbone, as expected for a sequence-independent DNA repair enzyme. The overall orientation of the DNA strand that contains the CPD is consistent with models derived from both biochemical (8) and NMR spectroscopic data (11). The thymine dimer is specifically recognized in the active site by being completely flipped out of the duplex DNA (Fig. 2A). The complementary adenines stack with their outside neighbors but do not stack on top of each other because of a large intrahelical bend at the CPD site. Previous studies showed that a single CPD lesion induces bending of regular B-DNA by 20 to 30° (15, 16). In the complex with DNA photolyase, the DNA bending was increased to about 50° (Fig. 2B). The photolyase/DNA complex differs significantly from a complex with the DNA-repair enzyme T4 endonuclease V (17), which also recognizes CPD lesions. Instead of the thymine dimer, the endonuclease flips the adenine opposite the 5′-T of the CPD lesion into its active site, thus causing the DNA to bend in the opposite direction from that in the photolyase/DNA complex (Fig. 2A).

Fig. 2.

(A) CPD-comprising DNA duplexes bound to DNA photolyase, to T4 endonuclease V (17), or in the uncomplexed state (16). Accession numbers are in parentheses. (B) The overall bend of modeled duplex B-DNA with an internal CPD lesion (13) increased from about 22° (gray) to 50° on binding to DNA photolyase (yellow). The inset shows the expected structural changes around the thymine dimer in the modeled states before and after base flipping. R232 might assist in the recognition of CPD-comprising DNA before base flipping by forming a salt bridge with the P0 phosphate. The flip-out of the thymine dimer into the active site pocket by ∼13 Å is accompanied by large structural changes in the CPD-comprising DNA strand.

The disruption of the two base pairs at the CPD lesion site and partial unwinding of the duplex DNA created a large hole of about 10 by 10Å. This was partly occupied by a nonregular ridge comprising residues G397 to F406 (18) of the photolyase. Here, the only specific interactions between the photolyase and DNA involved van-der-Waals contacts between P402 and the adenines opposite the CPD lesion and a hydrogen bond between the amide of L403 and the phosphate group between the two adenines (Fig. 3). The conformation of this ridge differs between the complexed and uncomplexed photolyase molecules (fig. S1). In the complex, the ridge G397-F406 is displaced by about 4.0Å with the biggest movement found at its tip for D399 (10Å) and in a large swiveling motion of the side chain of R404.

Fig. 3.

Schematic diagram showing the interactions between duplex DNA and the enzyme. Nucleotides not defined by electron density are shown faded. Dashed arrows indicate interactions with the protein backbone; solid arrows, with side chains. Numbering of the phosphate groups starts from the intradimer formacetal group (0). Interactions between the enzyme and the complementary strand may stabilize the bending of duplex DNA, because there are no major differences in the affinities toward CPD-comprising single- and double-stranded DNA (2).

Salt bridges and hydrogen bonds are extensively formed between the photolyase and the phosphates P–1, P+1, P+2, and P+3 (Fig. 3). Although the synthetic CPD lesion comprises a formacetal group instead of the intradimer phosphate P0, major interactions with the missing P0 phosphate are unlikely, because there are no residues close to the flipped thymine dimer that could interact with P0. This observation agrees with biochemical footprinting data (19, 20). Like several other DNA binding proteins, DNA photolyases use the dipole moment of helices for electrostatic stabilization of the protein-DNA complex. For example, the N terminus of helix α6 is directed toward the minor groove around the P–1 phosphate (Fig. 1B). An analogous interaction may be postulated between the P+2 phosphate and helix α18, which moves by ∼0.6Å upon DNA binding.

Negative Fobs-Fcalc difference electron density at the postulated cyclobutane ring revealed that the C5-C5 and C6-C6 bonds of the synthesized CPD lesion were broken. Thus, the active site harbors a repaired thymine dinucleotide (Fig. 4, B and C). To exclude adventitious DNA repair during crystallization or crystal handling, single crystals were analyzed by capillary electrophoresis (CE) for the presence of the synthetic CPD lesion. We found that the cyclobutane ring was still intact after 12 months of crystallization, showing that neither crystal growth nor harvest, both performed under strict red light conditions, induce repair (fig. S2). Likewise, the CE analysis of crystals only briefly exposed (∼2 s) to synchrotron irradiation showed intact CPD lesions. However, for crystals exposed for a complete data collection run (overall exposure time, 600 to 1000 s), we observed mainly decomposition products of the DNA strands in the CE runs. Unlike blue light–catalyzed cycloreversions of CPD lesions by photolyases, which proceed only above 200 K (21), the observed cleavage of the CPD lesion by prolonged synchrotron exposure occurred at the data collection temperature of 100 K. Although the repair of CPDs by radiolytically produced electrons has been reported even at 77 K (22), it has not been observed in other structures of DNA strands containing CPD lesions, either uncomplexed (16) or complexed to T4 endonuclease V (17) or an archaeal Y-type polymerase (23). The vulnerability of the CPD lesion in our crystals could be caused either by the unusual flip of the CPD lesion out of the DNA helix or by true catalysis in the active site that mimics blue light–mediated CPD cleavage. The latter is supported by the observation that a second photochemical reaction of photolyases, the reduction of the catalytic flavin cofactor from an inactive, oxidized state to FADH, is likewise triggered in the A. nidulans enzyme by high brilliance synchrotron radiation (24). In our structure, flavin reduction by radiation-generated electrons was sustained by the 9° “butterfly bend” along the N5-N10 axis of the isoalloxazine ring (Fig. 4A). CPD photolyases have high quantum yields for photochemical dimer splitting (∼0.9); thus, it is likely that the geometry of the active site allows particularly efficient unidirectional electron transfer and dimer splitting.

Fig. 4.

Synchrotron-induced structural changes inside the DNA/DNA photolyase complex and recognition of the CPD lesion. (A) The bent isoalloxazine moiety of the catalytic flavin is shown with 2Fobs-Fcalc electron density contoured at 1σ. For comparison, the flavin cofactor of noncomplexed DNA photolyase (accession no. 1QNF) is shown as light red wireframe. (B) With the intact CPD lesion (pink) being modeled in the active site, Fobs-Fcalc difference electron density around the C5 and C6 atoms of the cyclobutane ring (red, –3σ; blue, +3σ) indicated the cleavage of the CPD group. For comparison, the cleaved CPD lesion (blue) is shown with its 2Fobs-Fcalc density (gray, 1σ). (C) Hydrogen bonds with the thymine dimer are shown as blue dashed bonds, others in orange. (D) Diagram of the interactions between the CPD lesion and active site residues; distances are given in Å. Possible electron transfer routes are indicated either via the adenine moiety of FADH (purple) or between the isoalloxazine and the 3′-T (green).

As might be expected for the structure of a cryotrapped reaction product, the thymine dinucleotide differs structurally only slightly from a hypothetical CPD lesion within the active site (Fig. 4B). The pyrimidine rings of the 5′-T and 3′-T improve their stacking by decreasing the tilt angle between the base planes from 56° (16) to 16° after cleavage. The rotational offset of –26° perpendicular to the base planes of the thymine dinucleotide mimics quite well the CB+ pucker of the cyclobutane ring that was found in the crystal structure of a CPD lesion within duplex DNA [C6*-C5*-C5-C6 dihedral: –24° (16)] and in theoretical calculations (25, 26). The interactions that we observed between the thymine dinucleotide and the active site are likely preserved before cleavage of a bound CPD lesion. An L-shaped wedge comprising the conserved tryptophans W286 and W392 shielded the cyclobutane ring during the reaction course by making van-der-Waals interactions with the ring plane of the 5′-T and the edge of the thymine dinucleotide. The C4 carbonyl groups of the 5′-T and the 3′-T formed hydrogen bonds with the adenine N6 amino group of the FADH cofactor (Fig. 4, C and D). In addition, the two thymines formed hydrogen bonds via their C4-carbonyl and N3-imide groups to the side chains of E283 and N349, although these residues are apparently free to flip their side chain for an alternative hydrogen-bonding pattern with cytosine-comprising CPD lesions. Nevertheless, the hydrogen bonds between the 3′-T and N349 and between the 5′-T and E283 might be important for catalysis. Protonated E283 could stabilize the radical anion of the CPD formed after electron transfer from FADH (Fig. 1A). A mutation of this residue to alanine in the yeast photolyase did not affect substrate binding, but diminished the quantum yield for the CPD cleavage reaction by 60% (8).

A still unresolved issue of the photolyase mechanism is the nature of the electron transfer pathway used in blue light–driven electron transfer from FADH to the CPD lesion. Compared to other flavoproteins, the FADH cofactor of photolyases adopts a unique U-shaped conformation and might hence facilitate indirect electron transfer through its adenine moiety toward the CPD lesion. Quantum chemical calculations corroborated such an electron transfer pathway from the electron-donating π-system of the isoalloxazine via the 1′-CH and 2′-OH groups of the ribityl group and the adenine moiety (27, 28). Our structure firmly supports this pathway, because the adenine ring bridges the electron donating isoalloxazine ring and the thymine dinucleotide via two hydrogen bonds (Fig. 4, C and D). Nevertheless, a direct electron transfer pathway cannot be excluded, because the C4-carbonyl of the 3′-T almost contacts the C8-methyl group of the isoalloxazine (4.3Å), and the center of the electron-donating isoalloxazine ring system of FADH is only 7Å away from the 3′-T. Further experimental and theoretical data are needed to clarify which of the two pathways is operational.

Supporting Online Material

www.sciencemag.org/cgi/content/full/306/5702/1789/DC1

Materials and Methods

Figs. S1 and S2

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article