Recognition and repair of UV lesions in loop structures of duplex DNA by DASH-type cryptochrome

Abstract

DNA photolyases and cryptochromes (cry) form a family of flavoproteins that use light energy in the blue/UV-A region for the repair of UV-induced DNA lesions or for signaling, respectively. Very recently, it was shown that members of the DASH cryptochrome subclade repair specifically cyclobutane pyrimidine dimers (CPDs) in UV-damaged single-stranded DNA. Here, we report the crystal structure of Arabidopsis cryptochrome 3 with an in-situ-repaired CPD substrate in single-stranded DNA. The structure shows a binding mode similar to that of conventional DNA photolyases. Furthermore, CPD lesions in double-stranded DNA are bound and repaired with similar efficiency as in single-stranded DNA if the CPD lesion is present in a loop structure. Together, these data reveal that DASH cryptochromes catalyze light-driven DNA repair like conventional photolyases but lack an efficient flipping mechanism for interaction with CPD lesions within duplex DNA.

Cryptochromes (cry) and DNA photolyases form a unique family of flavoproteins, with members present in all kingdoms of life (1). This family is divided into several subclades according to sequence similarity and function. DNA photolyases are enzymes that repair cytotoxic and mutagenic DNA lesions induced by UV-B, specifically cis-syn cyclobutane pyrimidine dimers (CPDs) or 6-4 pyrimidine-pyrimidone lesions (6-4 photoproduct) by using light energy in the UV-A/blue region (2). The catalytic cofactor of DNA photolyase is flavin adenine dinucleotide (FAD) in its fully reduced form (FADH⁻), and this is present in a U-shaped conformation, as shown in several DNA photolyase structures (3–7). Catalysis involves electron transfer from the excited catalytic cofactor to the UV-B photoproduct, splitting the cyclobutane or oxetane rings, and electron back-transfer to the semireduced FADH° (1). Excitation of FAD is accomplished either by direct photon absorption or by Förster-type energy transfer from an antenna cofactor (1). Despite considerable sequence and structural similarity with DNA photolyases and common cofactor compositions, cryptochromes generally lost repair activity but gained photoreceptor function operating in the same waveband region as DNA photolyases (1). In plants, cryptochromes trigger several developmental processes, such as deetiolation and photoperiodic flower induction, and they entrain the circadian clock (8). In animals such as Drosophila, cryptochromes also function as photoreceptors for light input to the clock (9) and in magnetoreception (10), whereas mammalian cryptochromes are central components of the circadian clock without proven photoreceptor function (11). The more recently discovered subclade of the cryptochrome/photolyase family, named cry DASH, includes members in plants, cyanobacteria, eubacteria, and vertebrates (12–15). It has been suggested that they represent photoreceptors because they lacked repair activity for CPDs in double-stranded DNA (dsDNA), despite their DNA-binding activity (12, 13). Positive support for photoreceptor function of cry DASH was provided by DNA microarray analysis of a Synechocystis mutant lacking the corresponding gene. The analysis showed increased levels of several transcripts compared with wild type, and thus suggested a role as a transcriptional repressor (12). The crystal structures determined from Synechocystis cry DASH (12) and Arabidopsis thaliana (A. t.) cry3 (16, 17) showed an overall protein fold similar to class I CPD DNA photolyases (3–5) and Arabidopsis cry1 (18), with an N-terminal α/β-domain and a C-terminal α-domain that binds the FAD cofactor in the U-shaped conformation. Similar to photolyases, photoreduction (photoactivation) of FAD to fully reduced FADH⁻ (that is, a potentially catalytically active form) has been demonstrated for cry DASH (16, 19). The surface features around the FAD-binding pocket of cry3 and Synechocystis cry DASH comprise several basic residues and are thus more similar to class I DNA photolyases than to Arabidopsis cry1 (12, 16–18). These residues have been shown to be essential for DNA binding in DNA photolyase (6). This property of DASH cryptochromes is consistent with their previously described DNA-binding activity (12, 13). Very recently, it was shown that DASH cryptochromes repair CPDs specifically in single-stranded DNA (ssDNA) (20), thus prompting a reclassification of DASH cryptochromes as ssDNA-specific photolyases. It was hypothesized that the ssDNA repair activity is a relict of an evolutionary period when lateral gene transfer in the form of ssDNA viruses and free ssDNA played an important role.

To elucidate the cry DASH recognition mode for CPD lesions, we crystallized the complex of Arabidopsis cry3 with a synthetic oligonucleotide consisting of 5 thymines with a synthetic CPD analog between T2 and T3. The synthetic CPD analog has the same cis-syn stereochemistry as natural CPD lesions, but contains a formacetal linkage instead of the intradimer phosphate (6). This oligonucleotide was predicted to cover at least the minimal recognition motif of CPD photolyases, NpT<>TpNpN (21, 22). The X-ray crystal structure of this complex was solved by molecular replacement using the available cry3 structure (16) as a search model and refined with 6 cry3·T5 complexes per asymmetric unit at 2.0-Å resolution (R factor to R free: 0.185:0.222). In these cocrystals, cry3 forms homodimers, where a noncrystallographic twofold axis runs perpendicularly through a plane spanned by the methenyltetrahydrofolate (MTHF) and FAD chromophores. This dimeric organization, which involves salt bridges between the MTHF antenna chromophore and the side chains of D189, D192, and K338 of the other cry3 monomer, was observed before in the DNA-free crystal form of cry3 (16) but not in solution.

Results and Discussion

Overall Structure of cry3·T5 Complexes.

The structure of the cry3·T5 complexes shows fully defined electron density for the CPD oligonucleotide in 4 of 6 complexes [Fig. 1 and supporting information (SI) Fig. S1]. In 2 other complexes, the 5′-thymine is disordered. In the structure the chemically synthesized CPD lesion is found to be broken along the C5C5 and C6C6 bonds, thus presenting 2 intact thymine bases inside the active site. Such a repair of CPD lesions in synchrotron radiation-exposed crystals has been described before for the complex between the CPD photolyase from Anacystis nidulans (A. n.) and a CPD-comprising duplex DNA (6), but not for other crystal structures comprising CPD lesions (23–25). The hallmark of cry3 as a conventional DNA photolyase is the improved stacking of the 5′ and 3′ thymines by decreasing the tilt angle between the base planes from 56° (23) to 16° (6) after cleavage and a rotational offset of −30° perpendicular to the base planes of the thymine dinucleotide; the synthetic lesion thus mimics the conformations of CPD lesions as observed in duplex DNA. Compared with uncomplexed cry3 (16), the binding of the CPD oligonucleotide triggers only local structural changes in cry3. Apart from the C terminus, where H496 forms π-stacking interactions with the thymine base of T5, a flexible ridge region (G437–S449) that links helices α15 and α16 moves by up to 4.6 Å upon DNA binding (Fig. 2A). Similar movements of the corresponding region have been observed in the complex of the A. n. photolyase with DNA, indicating local, DNA-dependent reordering of the protein surface. In cry3, the side chain of R446 is close to the intralesion formacetal linkage (≈4.6 Å) that replaces in the synthetic CPD lesion the intralesion phosphate. Accordingly, this highly conserved residue might stabilize the base pair flip out either by forming a direct salt bridge with the intralesion phosphate or by compensating its charge only via long-range electrostatic interactions.

Fig. 1.

Ribbon model of complex A from A. t. cry3 and the repaired CPD damage in sticks representation with SIGMAA-weighted F_obs − F_calc omit electron density (light blue, contoured at 2 σ) defining the oligonucleotide. The catalytic domain (dark red) of A. t. cry3 contains the catalytic FAD cofactor (yellow) and the antenna chromophore MTHF (orange) in the contact region to the antenna domain (blue). Nomenclature and definition of secondary structure elements are given in Brudler et al. (12).

Fig. 2.

Interactions between A. t. cry3 and the DNA backbone. (A) Structural movements in the active site upon substrate binding to A. t. cry3. The loop region α15–α16 moves up to 4.6 Å and allows several stabilizing interactions between the DNA and the side chains of E444 and R446 as well as between the main chain of D445 and the 2 thymines T1 and T4. The ribbon models show A. t. cry3 with bound DNA (red) and the free protein (pdb code: 2J4D; green). The T5 oligonucleotide (yellow) is shown in stick presentation together with its molecular surface. The formacetal linkage within the synthetic CPD lesion that was split into 2 thymine residues by X-ray radiation is marked as P⁰. Notably, the side chain of R446 is close enough to form a salt bridge with the intralesion phosphate moiety. (B) Comparison of active sites between cry3 (purple) and the class I photolyase from A. n. (cream) complexed to CPD lesion comprising DNA (PDB code 1TEZ). Note that K452 in A. t. cry3 and K414 in A. n. photolyase are noncorresponding residues but occupy very close spaces.

Binding Mode of the T<>T Lesion.

The binding mode of the thymine pair within the active site of cry3 is very similar to that of the A. n. CPD photolyase: The C4 carbonyl groups of the 5′ and 3′ thymines form hydrogen bonds with the adenine N6 amino group of the FADH⁻ cofactor (Fig. 2B). Additional, conserved interactions include the hydrogen bonds between the 5′ thymine and E325 (A. n. photolyase: E283) and the 3′ thymine and N391 (A. n. photolyase: N349; Fig. 2B). In comparison with the A. n. CPD photolyase, the active site of the cry3·T5 complex not only harbors 6 structurally conserved water molecules, but is also considerably more polar (Fig. 3). One of the tryptophans of the L-shaped hydrophobic wedge of class I CPD photolyases (A. n. photolyase: W286, W392) is replaced by a tyrosine, Y434. Upon CPD lesion binding, the side chain of this residue swivels by ≈70° into the active site to contact the cyclobutane moiety of the lesion via its aromatic system. Its hydroxyl group makes a hydrogen bond with the P⁻¹ phosphate of the T5 oligonucleotide (Fig. 2B). As a consequence of the replacement of an indole by a phenyl group, 3 additional water molecules invade the active site and form a polar shell around the 3′-thymine ring (Fig. 3C). Two water molecules that are halfway between the 3′-thymine and the isoalloxazine ring in the A. n. CPD photolyase, and thus cross the path of direct electron transfer between the CPD lesion and the FADH⁻ cofactor (26), are replaced by the carboxamide group of N431 (Fig. 3B). The hydrophilic side chain of Q395 replaces a methionine (A. n. photolyase: M353) that was previously considered to be a unique discriminant between CPD photolyases and cryptochromes (27) because it contacts the ring system of the 3′ thymine from the side (Fig. 2B).

Fig. 3.

Isoelectric surface potential of A. t. cry3 bound to the single-stranded pentameric DNA containing a CPD analog. (A) Top and side (Inset) views. (B and C) Hydration and electrostatics of the active site in the substrate-bound state of cry3. The black arrows in C indicate the water molecules intruded into the active site because of replacement of a tryptophan conserved in class I CPD photolyases by Y434.

The P⁻¹, P⁺¹, and P⁺² phosphates of the T5 oligonucleotide are engaged in polar interactions with residues, which mostly are also found in class I CPD photolyases (Fig. 2A). Accordingly, DNA binding, with its concomitant flip of the CPD lesion into the active site of cry3, appears comparable to canonical photolyases. Interestingly, the flip of a CPD lesion out of duplex DNA is already energetically favored by about 9 kcal/mol compared with undamaged DNA, and is only slightly sequence-dependent (28, 29). This raises the question of why Arabidopsis cry3 and other DASH-type cryptochromes can repair CPDs in ssDNA but not in duplex DNA. One notion might be steric interference between cry3 and the counterstrand of the CPD-comprising DNA strand (20). However, a model of duplex DNA with a flipped CPD lesion that was derived on the base of the A. n. CPD photolyase complex (Fig. S2) indicates only an absence of stabilizing interactions between helix α5 in cry3 (α6 in A. n. CPD photolyase) and the phosphoribose backbone of the counterstrand. Alternatively, it was hypothesized from structure comparison with class I photolyases that the inability of cry DASH to bind dsDNA is caused by an overall decrease in the hydrophobicity of the substrate-binding pocket, resulting in lower affinity for the CPD lesion (17) and rendering the enzyme incapable of expending 5.3–7.5 kcal/mol for the required flip of the CPD lesion (29). Decreased hydrophobicity of the active site and the intrusion of water are indeed observed in our structure (Fig. 3 B and C). This could disfavor the competition of the protein with the residual stacking and hydrogen-bonding interactions of the CPD lesion within the duplex DNA. Because DASH cryptochromes are widely distributed in living organisms ranging from cyanobacteria to vertebrates, we reason that the specificity for ssDNA may have biological relevance aside from a suggested evolutionary role in transient repair of ssDNA during lateral gene transfer (20). For example, it is known that duplex DNA becomes strongly distorted and partly wound up during essential processes, such as replication, transcription, or lesion repair. The repair activity of DASH cryptochromes, therefore, might be permitted only during these processes.

Binding to and Repair of a Single T<>T Dimer in Distorted dsDNA by cry3.

To test whether cry3 is capable of binding and repairing CPD lesions in distorted duplex DNA, we used cry3 that was catalytically active against ssDNA (see Fig. 5A), similar to the cry DASH ortholog from Vibrio cholerae (20). In the binding assay with cry3, we used dsDNA substrates that harbor T<>T lesions in loop structures of various size. Loops were created by annealing complementary oligonucleotides that either allowed complete base pairing or left an increasing number of bases within and around the T<>T dimer unpaired (Fig. 4A, probes 1–8). Binding of these probes was studied by EMSAs performed under red light to avoid concomitant photorepair (Fig. 4B). As controls, the same dsDNAs but without the central T<>T dimer were used (Fig. 4C). Quantitative analysis of binding to these probes (Fig. 4D) revealed significant substrate binding if at least one of the thymines in the T<>T lesion formed no hydrogen bonds to the complementary strand (Fig. 4, probes 2 and 3). The weaker binding of probe 2 compared with probe 3 is apparently caused by the fact that the 5′ thymine in the T<>T lesion makes only one hydrogen bond to the complementary adenine, whereas the 3′ thymine makes canonical two (23). Accordingly, a further increase in binding is observed when both thymines of the T<>T dimer lack hydrogen bonding to the complementary strand (Fig. 4, probe 4). Binding efficiency is close to saturation when the T<>T dimer is positioned in a loop structure of 4 bases (Fig. 4, probe 5) and increases only slightly upon further increase in the loop size (Fig. 4, probes 6–8). An additional slight increase in binding is observed for the complete single-stranded control (Fig. 4, probe 9). This reflects some background binding of cry3 to single-stranded undamaged DNA (a shifted band for probe 9 indicated with an arrow in Fig. 4C and open bar for the same probe in Fig. 4D), as already reported (13). The second higher shifted band seen for the single-stranded probe with a T<>T dimer (indicated with a thin arrow in Fig. 4B) is presumably caused by simultaneous binding of cry3 to the central T<>T lesion and additional unspecific binding to another region of the same DNA molecule.

Fig. 5.

Repair of T<>T dimers by cry3. (A) Single-stranded oligo(dT)₁₈. The reaction mixture contained 8.9 μM T<>T dimers in 10 μM oligo(dT)₁₈ and either 10 nM cry3 (squares) or the same aliquot of buffer (circles). Mixtures were treated with photoreactivating light (open symbols) or kept in the dark (filled symbols) at 10 °C. Data represent the mean value of 2 independent experiments, with error bars indicating standard errors. The rate of cry3-catalyzed photorepair was calculated by using the slope of its linear part between 20 and 60 min (solid line) obtained by the least-square method. (B) Quantitative data of repair of a single T<>T dimer in loop structure probes. Data for probes 1–6 (shown in Fig. 4A) are based on the gels shown in Fig. S3.

Fig. 4.

Binding of cry3 to DNA probes containing a single T<>T dimer in the central position. (A) Sequences and structures of probes. The T<>T dimer is positioned within the VspI recognition site (boxed in probe 1). Probe 1 forms a perfect duplex. In probes 2 and 3, the 5′ and 3′ thymines, respectively, of the T<>T dimer are not hydrogen bonded to the complementary strand. In probe 3, only one hydrogen bond is formed between the 5′ thymine of the T<>T dimer and the complementary adenine (23). In probes 4–8, the T<>T lesion is positioned in the center of loop structures with 2–10 base pairs. Hydrogen bonds between complementary bases are shown as dashed lines. The upper strand (50 nt) was labeled at the 5′ position with IRDye700 (MWG Biotech AG) (marked with asterisk). (B and C) EMSA showing cry3 binding to probes with (B) or without (C) the central T<>T dimer. Probes shown in A and the single-stranded control (probe 9) were mixed with cry3 (+) or with the same aliquot of buffer (−). Arrows indicate the positions of shifted bands. Representative gels from 2 independent experiments are shown. (D) Quantitative binding data. Mean values and standard errors of the 2 independent experiments are shown.

In general, the binding behavior of cry3 to DNA duplexes containing a T<>T dimer within a loop structure is well reflected in its specific repair activity for these substrates (compare Fig. 4D with Fig. 5B). Whereas essentially no repair is detectable for the complete duplex DNA even after 2 h of exposure to UV-A, there is quantitative repair activity seen for loop structure substrates that increases with the size of the loop harboring the T<>T lesion. For substrates that contain the T<>T lesions in the center of a loop with 4 or 6 unpaired bases' DNA, repair is almost completed within 5 min of incubation (Fig. 5B, probes 5 and 6).

Conclusions

Our structural and functional data provide clear evidence that cry3 and apparently other DASH cryptochromes have the ability to recognize and repair CPDs in dsDNA, because essentially all of the residues that make contact to the DNA lesion as well as those that contact the catalytic FAD cofactor are conserved (Fig. S4). However, for damage recognition and repair by cry3, the hydrogen bonding of the CPD lesion to the counterstrand has to be disturbed at least at one base in the lesion site. These data thus provide experimental confirmation of the previously proposed model (17) that DASH cryptochromes lost the ability to flip the CPD lesion out of the duplex but still retained DNA repair activity. Since disruption of hydrogen bonding in the DNA double helix resulting in even small loop structures is accompanied by replication, transcription and lesion repair, we propose that DASH cryptochromes have important functions during such essential cellular processes. This does not preclude that these proteins have additional but thus far unidentified roles unrelated to DNA repair. Indeed, an intriguing example that even conventional photolyases can exert dual roles as photolyases and photoreceptors has been recently found in the fungi Trichoderma atroviride and Aspergillus nidulans, where class I CPD photolyase autoregulates its own photoinduction (30) and triggers light-dependent sexual differentiation without losing its DNA repair capacity (31), respectively.

Materials and Methods

Oligonucleotides and Enzymes.

All oligonucleotides, including the IRDye700-labeled one, were purchased from MWG Biotech AG. Sequences are listed in the SI Text. T4 DNA ligase and VspI were from Fermentas. T4 pyrimidine dimer glycosylase (T4 endonuclease V) was from New England Biolabs.

Structure Determination of the cry3·T5 Complex.

The expression and purification of cry3 and its cocrystallization are described in the SI Text. A 2.0-Å data set was collected from a single cry3·T5 crystal cryocooled directly from mother liquor at beamline BW7A (EMBL/DESY). Data reduction was done with MOSFLM and SCALA (32). The structure was solved by molecular replacement using MOLREP (33) and a structure of A. t. cry3 (pdb code 2J4D) as a search model. Semiautomatic tracing and model rebuilding were done by ARP/wARP (34). Further refinement was carried out by using CNS 1.1 (35) followed by REFMAC5 (32) until the R factor to R free converged at 18.5:22.2 for data between 15.1 and 2.0 Å (Table S1). Throughout refinement, noncrystallographic symmetry restraints were applied to all 6 complexes, including the cofactors and the residues I4-V352, W356-G437, S449-L474, and M486-L494. The other regions, including loops and the N and C termini, showed local conformational variation, mostly due to crystal packing and intrinsic flexibility. To avoid bias, reference data for calculation of R free were selected shell-wise. Apparently, the repair of CPD lesions by synchrotron radiation proceeds fast in the crystals. Between the thymine base pair, no significant differences in electron densities were observed if the subsets of the X-ray data set were used for calculation, which differed in the overall X-ray dose absorbed by the cry3·T5 cocrystal.

Binding Assay.

Probe synthesis is described in the SI Text. A total of 20 fmol of the respective labeled probe (representing 20 fmol of labeled oligonucleotide with or without central T<>T dimer annealed with 2 pmol of the corresponding counterstrand) were mixed with 1 pmol of cry3 in a total volume of 10 μL under red light. The binding buffer conditions were 15 mM Tris·HCl (pH 7.5), 20 mM NaCl, 5 mM DTT, 50 μg·mL⁻¹ BSA, and 10% glycerol. Reactions were incubated for 30 min in the dark on ice and separated on nondenaturing 5% polyacrylamide gels in 25 mM Tris·HCl, 25 mM borate, and 0.6 mM EDTA (pH 8.4). Gels were run at 4 °C in the dark and afterward directly scanned and analyzed using the Odyssey Infrared Imaging System (Li-Cor Biosciences). The percentage of bound probe was expressed as the ratio of the intensity of shifted band to the sum of intensities of shifted and free probe bands × 100.

Repair of T<>T Dimers in Oligo(dT)₁₈.

Oligo(dT)₁₈ at 10 μM concentration representing about the same concentration of T<>T dimers was mixed with 10 nM cry3 that had been photoreduced immediately before the repair assay by illumination with 50 μmol·m⁻²·s⁻¹ blue light for 2 h to achieve its fully reduced active state. For control reactions, the same aliquot of buffer was used instead of cry3 solution. The final buffer conditions were 50 mM Tris·HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA, 10 mM DTT, and 10% glycerol. Reactions were placed in a Quartz Suprasil cell (Hellma GmbH & Co. KG) and incubated in the dark for 10 min on ice. Afterward, they were shifted to 10 °C and were exposed to 73 μmol·m⁻²·s⁻¹ of 365-nm UV-A light for 2 h or kept in the dark. At the desired time points, spectra in the 200- to 350-nm range were taken, and remaining T<>T dimers were quantified as described in the SI Text.

Repair of a Single T<>T Dimer Within Loop Structures in Double-Stranded Probes.

Concentrations of the probes and cry3 were the same as for the binding assay. Apart from the addition of 100 μg·ml⁻¹ BSA, the buffer conditions were identical to those for the repair of T<>T dimers in oligo(dT)₁₈. After binding was allowed for 30 min in the dark on ice, the reaction mixture was divided into 2 Quartz Suprasil cells and shifted to 10 °C. One cell was illuminated with 73 μmol·m⁻²·s⁻¹ of 365-nm UV-A; the other was kept in the dark. Aliquots were taken just before illumination (zero point), after 5 min of illumination, and at the end of a 2-h period. Cryptochrome 3 in the withdrawn aliquots was immediately inactivated by incubation at 95 °C for 10 min, and probes were allowed to reanneal. Buffer conditions were then adjusted to 25 mM Tris·HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA, 5 mM DTT, 100 μg·ml⁻¹ BSA, and 10% glycerol, and reactions were completed with 10 units of T4 pyrimidine dimer glycosylase (T4 PDG). Control reactions contained no enzyme. After 30 min at 37 °C, aliquots of the reactions were run on a denaturing 6% polyacrylamide gel in 100 mM Tris·HCl, 100 mM borate, and 2.5 mM EDTA, pH 8.4, at ambient temperature in the dark. Resulting gels were directly scanned and analyzed by using the Odyssey Infrared Imaging System.