Cyclobutylpyrimidine dimer base flipping by DNA photolyase.

DNA Photolyase is a flavoprotein that uses light to repair cyclobutylpyrimidine dimers in DNA. From considerations of the crystal structure of the protein, it has been hypothesized that the dimer lesion is flipped out of the DNA double helix into the substrate binding pocket. We have used a fluorescent adenine analog, 2-aminopurine (2-Ap), as a probe of local double helical structure upon binding of the substrate to the protein. Our results show that the local structure around the thymidine lesion changes dramatically upon binding to Photolyase. This is consistent with base flipping of the lesion into the protein binding cavity with concomitant destacking of the opposing complementary 2-Ap nucleotide.

O 2 . The surface of the protein above the cavity incorporates a strip of positively charged amino acid residues that are thought to help orient the negatively charged phosphate backbone of the damaged DNA strand.
Unfortunately, no crystal structure of the enzyme-substrate complex is available, but a recent crystal structure of Thermus thermophilus complexed with thymine was published by Komori and co-workers (7), which shows the thymine residing in the putative substrate cavity. While thymine is not a substrate, it can be thought of as a partial product of the repair reaction. This is the strongest evidence regarding the substrate binding mode of photolyase for CPDs. However, it does not clarify whether one or both of the thymine bases of the CPD is bound in the cavity.
Other experimental evidence for the mode of substrate binding is available. Most notable are the ethylation and footprinting studies of Husain et al. (8). They showed that the ethylation of the phosphates 3Ј and 5Ј to the CPD inhibit the binding of the lesion strand to photolyase. Ethylation of phosphate groups on the complementary strand had no effect on binding. Footprinting experiments showed that only 6 -7 bases around the CPD were protected by the protein. This small footprint is significant in that it suggests that photolyase binds substrate in a structure-specific, as opposed to a sequence-specific manner.
Binding studies using specific nucleotide sequences have shown that distortions of the backbone caused by the CPD aid in the recognition of the lesion by PL (9). Alanine substitution mutagenesis studies by Berg and Sancar (10) also support this hypothesis. The question has sparked several computational studies as well, most of which support the base flipping hypothesis (11,12). Thermodynamic studies suggest that there is little energy cost for flipping the CPD into solution (13).
There are several DNA-binding proteins that appear to manipulate DNA by flipping out bases from the helical structure (14). For example, DNA methyltransferase is thought to base flip an adenine into the enzyme for methylation (15,16). Another repair protein, T4 endonuclease V, has been shown to flip an adenine opposing the CPD lesion out of the helix (17). Several of these studies have used a fluorescent adenine analog, 2-aminopurine (2-Ap), as a reporter of helical structure. 2-Ap can be incorporated into a DNA oligomer using automated synthesis and has a strong absorption band centered at 303 nm (⑀ ϭ 6000 M Ϫ1 cm Ϫ1 for the nucleotide) (18). High quantum yield emission peaks at about 370 nm for the nucleoside in aqueous solution (19). When 2-Ap is incorporated into oligomeric ss-DNA, the emission is significantly quenched and blueshifted (20). When a 2-Ap containing ss-DNA oligomer is annealed to its complement, the 2-Ap fluorescence is further quenched by base stacking (16). Thus, the quantum yield and emission maximum of 2-Ap are effective reporters of base stacking and solvent exposure (polarity), respectively.
We have utilized these properties to probe base flipping in photolyase by incorporated 2-Ap opposite a thymidine dimer.
Any structural changes that occur in the region of the lesion would be detected as changes in quantum yield and perhaps emission maximum. 2-Ap absorbs in the same spectral window as reduced photolyase. This will complicate the analysis of the experiment, because the 2-Ap emission will come from oligonucleotides in various stages of repair. The rate of repair of CPDs by DNA photolyase is in the nanosecond range (although the turnover rate is much slower (21)). To circumvent this difficulty the oxidized enzyme (PL ox ) is used, which can bind the CPD lesion without repairing it. The results using oxidized enzyme show that base flipping is operative in DNA photolyase. Experiments using the reduced enzyme demonstrate that the 2-Apcontaining CPD duplex is a substrate for photolyase.

EXPERIMENTAL PROCEDURES
Photolyase Protein-E. coli DNA Photolyase was overexpressed from JM109 cells transformed with the pMS969 plasmid containing the phr gene (the plasmid was a generous gift of Professor Aziz Sancar, University of North Carolina, Chapel Hill, NC). The purification of the protein was based on the method of Payne et al. (4) and modified as described in detail in a previous paper (22). Briefly, E. coli photolyase was isolated in the blue semiquinone form. The apoprotein was made by stripping the FAD and folate chromophores using low pH buffers containing saturated KBr (23). The apoprotein was reconstituted with oxidized FAD and purified on a phenyl-Sepharose column. The purity of the reconstituted protein was determined by the ratio of absorbances at ϭ 280 nm (protein) and 450 nm (FAD). The ratio of A 280 /A 450 was typically around 20 -25. Pure protein would have a ratio of about 12 based on the known extinction coefficients of the apoprotein and FADreconstituted protein (24), indicating that the protein purity was about 60%. Further purification was deemed unnecessary as apophotolyase will not bind substrate (25).
DNA Oligonucleotides: Irradiation and Purification-Two HPLCpurified 11-mer oligonucleotides were purchased from Integrated DNA Technologies, Inc. The 2-Ap-containing oligonucleotide with the sequence 5Ј-CTCCAACTTGC-3Ј ("2-Ap," where A ϭ 2-Ap) and its complement, 3Ј-GAGGTTGAACG-5Ј ("T-T") were resuspended in HPLC-grade water and used without further purification, though their purity was checked by reverse phase HPLC on a C 18 column (data not shown).
The T-T oligonucleotide was irradiated with UV-A light to produced the CPD-containing complementary oligonucleotide ("TϽϾT"). Irradiation of a solution of 8.4 ϫ 10 Ϫ2 mM T-T 11-mer and 1 mM acetophenone/ ethanol was performed in a 10-mm quartz fluorescence cell sealed with a septum. The cuvette was first purged with argon for 30 min to remove oxygen and subsequently placed on ice. The cell was covered with a Petri dish to filter out wavelengths below ϳ310 nm and irradiated for 4.25 h using 2 Philips Ultraviolet-B 40-watt bulbs at a distance of about 20 cm. Following irradiation, the solvent was removed using a Savant Speed Vac SC110. The pellet was then resuspended in 100 l of HPLC-grade water.
The TϽϾT oligonucleotide was purified using a Rainin HPLC with a Waters 8 ϫ 100 mm Radial Compression Nova-Pak C 18 column. A linear gradient of 7.8 -9.0% acetonitrile was used against 100 mM tetraethylammonium acetate over 20 min at 1 ml/min with detection at 254 nm (26). The TϽϾT oligonucleotide was collected and desiccated to dryness, resuspended in HPLC-grade water, and desalted using a Bio-Rad P6 mini-gel filtration column. It was stored at Ϫ20°C following desiccation until used.
UV-visible Photorepair Assay-The purified TϽϾT oligonucleotide was subjected to photorepair by photolyase to show that it was the substrate of the protein. A mixture of 0.50 M oxidized photolyase in buffer T (50 mM phosphate, 100 mM potassium chloride, 0.1 mM EDTA, and 10 mM ␤-mercaptoethanol, pH 7.5) was reduced under anoxic conditions with 1.2 ϫ 10 Ϫ2 mM sodium dithionite. Excess dithionite was allowed to react with water to produce bisulfate. 1.0 M TϽϾT oligonucleotide was added to the assay mixture. Repair of the TϽϾT oligonucleotide was initiated with exposure to 365 nm ultraviolet light from a Spectroline model ENF-240C hand lamp. Repair was monitored at 260 nm with a Hewlett-Packard 8452A UV-visible spectrometer following 5-s exposures to the UV light.
Steady-state Fluorescence Emission-Samples were prepared fresh for each scan to 1 ml total volume in buffer T. Mixtures of oligonucleotides were placed in a sterile Eppendorf tube in minimal volume, typically less than 50 l. These samples were heated in a 60°C water bath for 6 min and allowed to cool to room temperature on the labora-tory bench for 30 min. The annealed samples were then brought to a total volume of 1 ml in buffer T, taking into account any extra volume incurred if PL was added to the sample. The fluorescence cuvette containing the sample was sonicated briefly in a small ultrasonic cleaner to remove bubbles.
Fluorescence emission spectra were obtained using a Spex Fluoro-Max-2 fluorimeter. Each emission spectrum are averages of four scans in the range of 330 -510 nm with excitation at 317 nm. The scans were set at an integration time of 0.2 s, increment of 2 nm, and the excitation and emission band pass of 8 and 4 nm, respectively. The excitation and emission light were unpolarized. Spectra are corrected for the background fluorescence and Raman scattering of the buffer or control sample. A 0.50 ϫ 1.0-cm fluorescence quartz cell was used. For temperature-dependent measurements the fluorimeter cuvette holder was connected to a temperature-variable circulating water bath. The cuvette was allowed to equilibrate in the holder for 10 min in the dark prior to a scan.

Preparation and Purification of Target Oligonucleotides-
The duplexes used in this study were as follows, where the underlined A indicates the position of the 2-aminopurine base in the 2-AP containing ss-DNA, T-T stands for the complementary strand, and TϽϾT indicates the cyclobutylthymidine dimer-containing oligonucleotide (the position of the lesion is indicated by TT). UV-A lamps were used to generate the CPD lesion in the TT-containing 11-mer oligonucleotide, and HPLC purification of the photoproduct was performed as described by Banerjee et al. (26). It should be noted that the complementary strand sequence is a subset of that used by Husain et al. (8) in their ethylation studies.
Activity of the Reconstituted Photolyase-To verify that the HPLC-purified DNA photoproduct contained the cis-syn dimer, a 1 M solution of this oligomer was mixed with 0.50 M "blue" photolyase (the holoenzyme containing both folate and flavin chromophores) in buffer T. The mixture was deoxygenated and then reduced with 12 M sodium dithionite. Irradiation of this mixture with 365 nm light from a UV hand lamp (ϳ500 W/ cm 2 ) resulted in photoreactivation of the dimer lesion as measured by the recovery of absorbance at 265 nm (data not shown) (23,25). The absorbance change at this wavelength corresponds to the reappearance of the C5 ϭ C6 and C5Ј ϭ C6Ј double bonds in the thymidine nucleotides, which were lost upon dimerization. The negative absorbance at ϳ230 nm is due to the greater extinction coefficient of the CPD relative to the repaired thymines (27).
Determination of T m for the 2-Ap Duplexes-It is important to ascertain whether changes in the fluorescence quantum yield of the 2-Ap probe are due to conformational changes of the helix produced by protein binding or a partial melting of the duplex due to perturbations introduced by the CPD and/or 2-Ap base. Fig. 1 shows the melting curves obtained for the 2-Ap/T-T and 2-Ap/TϽϾT duplexes by monitoring the 2-Ap fluorescence as a function of temperature. Melting points were derived by finding the temperature corresponding to the maximum of the derivative of the melting curve. The melting point of the 2-Ap/ T-T duplex is T m ϭ 48°C, while the 2-Ap/TϽϾT duplex has a T m ϭ 42°C (Ϯ2°C). The 6°C difference between the melting points agree well with that obtained for a decamer duplex with and without a CPD (56 and 48°C) (28), although the absolute T m is about 8°C lower than for the decamer without the 2-Ap. We ascribe this difference to the aminopurine itself, which has been shown to lead to lower melting points, as much as 10°C (29). However, the melting points for the duplexes used in this study are well above the temperature at which the quenching experiments were done (23°C) where any melting artifacts would be small. Other experiments performed at 15°C in our laboratory have given similar results (data not shown).
Emission Quantification of the changes in fluorescence in these scans are reported for the peak fluorescence intensity and the integrated intensity of the fluorescence spectrum between 350 -400 nm. This range was chosen to reduce the potential influence of Rayleigh scattering of the excitation wavelength into the emission spectrum. These quantities are ratioed to the 2-Ap oligonucleotide value in each case to get the relative quenching. The values are shown in Table I The slight increase in fluorescence yield compared with 2-Ap/ T-T can be ascribed to a reduction in base stacking at the 2-Ap site due to the perturbation introduced by the CPD on the opposing strand. This interpretation is consistent with the results from a solution state NMR structure of a CPD-containing dodecamer DNA duplex (28) in which the 3Ј-side of the CPD is perturbed relative normal B-DNA structure. In that study, a decided propeller twist of the T⅐ ⅐ ⅐ A planes across the helix was observed, resulting in a loss of base stacking of the 5Ј-adenine opposing the CPD. The hydrogen bond length between these two bases is also slightly longer than that for other base pairs in the dodecamer. Nordlund and co-workers (29) have explored the changes in duplex structure and rigidity due to the incorporation of a 2-Ap base in a decamer duplex. They found that 2-Ap is less rigidly constrained and more prone to spontaneous excursions out of the duplex, leading to less base stacking. Both a loss of base stacking and base pairing are consistent with the increase in fluorescence yield in our experiments.
Interestingly, the 2-Ap/TϽϾT duplex exhibited a 6 nm red shifted emission maximum compared with the 2-Ap 11-mer. In terms of solvatochromism, a red shift is expected if a chromophore goes from a less polar to a more polar solvent, as would be the case if 2-Ap suffered more solvent exposure (31). An additional requirement for the red shift is that the permanent dipole moment of the excited state is greater than the ground state dipole moment, which has been verified experimentally (32). However, this red shift is not apparent in the emission spectrum of the single-stranded 2-Ap oligonucleotide, which suggests that the 2-Ap base opposing the thymidine dimer has neither flipped out into solvent nor retained its normal interactions in the double helix. One possible explanation is that the 2-Ap undergoes enough of a rotation that its difference dipole moment relative to the helix electric field produces the observed red shift, although this is only a speculation.  Table II. Addition of photolyase to the ss-2-Ap oligonucleotide gave essentially no change in the fluorescence yield or wavelength. This shows that there is little interaction between the 2-Ap oligonucleotide and PL ox , which is consistent with the low association constant for a single stranded oligo without a CPD lesion (K A ϭ 10 3 M Ϫ1 ) (10).

Emission Spectra of 2-Ap Oligomers and Duplexes with Ox-
The fluorescence yield of the 2-Ap/T-T/PL ox sample is somewhat higher than without protein (I 2Ap/T-T/PL /I 2Ap/PL ϭ 0.19 Ϯ 0.06 versus 0.15), and the emission maximum is the same within experimental error. Photolyase is known to bind undamaged duplex DNA with higher affinity than ss-DNA (9). For the normal duplex 11-mer (K A ϭ 10 4 M Ϫ1 ) only about 0.003 M will bind (nonspecifically) to photolyase. The crystal structure shows that the substrate binding cavity is in the middle of a positively charged strip of amino acid residues (5). These residues undoubtedly attract and bind the negatively charged phosphate backbone of the duplex whether or not it has a CPD (9). The increased fluorescence may be the result of small structural changes to the helix bound to the protein surface.
When reconstituted oxidized photolyase was added to the 2-Ap/TϽϾT duplex, the fluorescence increased by a factor of about 5 relative to the 2-Ap/T-T/PL ox sample and is very close to the emission yield obtained for the single-stranded 2-Ap/ PL ox sample. The emission maximum of the enzyme-bound duplex shifted back to 368 Ϯ 2 nm from the 374 nm peak for 2-Ap/TϽϾT in buffer alone. Using an association constant of K A ϭ 10 9 M Ϫ1 for a DNA duplex containing a CPD (10), the concentration of protein-substrate complex will be about 0.49 M (or about 98%) of the duplex is bound. Such a large change in fluorescence yield indicates that a severe distortion of the local helical structure has occurred around the 2-Ap base.
The band shape of the emission spectra for the duplexes with PL ox was somewhat different from that of the duplexes alone. Part of the fluorescence increase occurs below and above the 2-Ap emission range (roughly 340 -460 nm). It is possible that this may be due to energy transfer from the excited 2Ap base to the FAD cofactor, which has significant absorption below 490 nm. However, oxidized photolyase alone show very little fluorescence when excited directly with 317 nm light (data not shown), and its fluorescence emission is peaked at about 520 nm. It is more likely that the broad fluorescence signal is due to a distribution of conformations experienced by 2-Ap as the duplexes bind non-specifically to the photolyase protein.
Reduced PL Repair Assay-Using the oxidized enzyme for these experiments has the advantage that base flipping can be observed without significant repair of the CPD during the fluorescence scans. However, to verify that the 2-Ap/TϽϾT duplex is a true substrate for photolyase, we performed fluorescence experiments after reduction of the photolyase by dithionite.
A sample consisting of 0.50 M 2-Ap, 0.50 M TϽϾT, 0.55 M PL ox was transferred to the 10-mm fluorescence quartz cuvette and subsequently purged with argon to achieve anoxic conditions. Four equivalents (in a 2-l volume) of sodium dithionite was added to reduce the photolyase. The sample was allowed to sit for 1 h to allow any excess dithionite to react to form bisulfate. All preparative steps were performed under yellow lights, and all subsequent measurements were done in the dark so that no unwanted photorepair occurred.
The cuvette containing the reduced enzyme-substrate was exposed to 365 nm light over 30 min, and fluorescence emission spectra were taken at 5-min intervals. The results can be seen in Fig. 3, where I(t) Ϫ I(0) is plotted versus emission wavelength. As repair proceeds the 2-Ap fluorescence is quenched. This is consistent with repair of the CPD lesion and subsequent dissociation of the protein-substrate complex. No shift in the emission maximum was observed. The 2-Ap fluorescence decreases to about 3.7 ϫ 10 5 at em max , which is roughly equal to the difference between the fluorescence yield of I(2-Ap/TϽϾT/PL ox ) and I(2-Ap/T-T/PL ox ) (ϭ3.4 ϫ 10 5 ). Reduction of the photolyase protein with dithionite in the presence of the 2-Ap/TϽϾT duplex showed no effect on the 2-Ap fluorescence as long as the sample was not exposed to blue light (data not shown). Dithionite treatment of a 2-Ap/TϽϾT sample had no effect on its fluorescence yield (data not shown).   These experiments show that binding of the CPD substrate by E. coli DNA photolyase is accompanied by a substantial conformational change to the duplex around the 2-Ap probe base. This local disruption of helical structure is strong evidence that local melting has occurred and suggests that the CPD has been flipped out of the helix into the cavity of the protein, exposing the complementary 2-Ap base to solvent and destroying base stacking interactions.
Other studies have suggested that base stacking will be disrupted by photolyase at the 5Ј-A opposite the 3Ј-T of a CPD. The work of Husain et al. (8) is particularly noteworthy in that it shows that the N3 of the complementary adenine opposing the 5Ј-T of the CPD can be methylated. The accessibility of this adenine points to the kind of duplex disorder observed in our fluorescence experiments. Unfortunately, no mention is made of whether the adenine opposing the 3ЈT shows similar accessibility, as this adenine has been replaced by 2-Ap in our experiments.
The mutagenesis and ethylation experiments of Berg and Sancar (10) were used by these authors to justify a model where the CPD is flipped into the photolyase cavity and stabilized by hydrophobic and electrostatic interactions. The authors predict that the 3Ј-T of the CPD interacts more strongly with the protein binding pocket. This prediction agrees with our observation that stacking interactions with 5Ј 2-Ap probe opposing this thymine base are disrupted. Our results are also consistent with an NMR study of a CPD-containing dodecamer duplex, which shows that the 3Ј-T of the CPD is more perturbed than the 5Ј-T, although binding to the protein will undoubtedly change the solution-state conformation. We are currently performing measurements using a 3Ј 2-Ap oligonucleotide to determine how far the structural distortion propagates along the duplex, particularly in the protein-substrate complex.
Stivers (33) and Rachofsky and co-workers (34) have shown that destabilization of base stacking occurs if a 2-Ap base is positioned opposing an abasic site (AB). The fluorescence increase for this case (I 2Ap-AB /I 2Ap-N , where N is A, G, T, or C) fell in a range from about 1.2 to 2.3 and was sequence-dependent (33). If base flipping occurs in photolyase then it would some-what mimic an abasic site for the opposing 2-Ap base. Our results show a factor of 5 increase in fluorescence when the target duplex is bound by photolyase. This suggests that the deformation of the duplex by base flipping is considerably greater than that generated by an abasic lesion, supporting the base flipping model.
The computational work of Sanders and Wiest (12) agrees well with our results. They performed molecular dynamics simulations of the protein-substrate complex, which predicts that a base-flipped CPD, is stabilized by the protein binding site over the duration of the simulation. Specifically, their simulation shows that the 3Ј-T will experience disruption of base pairing and stacking, whereas the 5Ј-T is much less perturbed.
Zhao et al. (35) used mismatched bases opposing the (6-4) DNA lesion as a target duplex to probe whether (6-4) photolyase base flips this substrate for repair. Since a mismatched sequence would give greater conformational flexibility to the lesion, it was expected that the binding affinity for this target would be higher than for the completely complementary sequence if base flipping were operative. They observed that the mismatch duplex had a factor of two higher affinity than the complementary duplex. Given the high homology between  and CPD photolyase (36), it is reasonable to presume that both  and CPD photolyases use base flipping for substrate binding.
Allan et al. (37) have also used mismatched base pairs to confirm that base flipping occurs in EcoRI DNA methyltransferase. This DNA-modifying enzyme transfers a methyl group from S-adennosyl-L-methionine to adenine N 6 . By replacing the target adenine with 2-Ap they showed that prior to the methylation step the target adenine is flipped out into a hydrophobic pocket of the protein. As in the work of Zhao et al. (35) they measured a 2-fold increase on equilibrium binding affinity for a 2-Ap-containing duplex versus the complementary duplex. They measured the dynamics of the base flipping reaction using fluorescence stopped-flow methods, obtaining a flipping rate of about 21 s Ϫ1 . It will be interesting to see whether the rate of flipping a CPD or (6-4) lesion differ from single nucleotide rate.
Finally, base flipping has been observed for T4 endonuclease V, which repairs CPDs in a light-independent manner (17). In this protein it is thought that the 3Ј-A base opposing a CPD is flipped into the endonuclease protein cavity, exactly the opposite behavior hypothesized for photolyase. McCullough and co-workers (17) observed no change in 2-Ap fluorescence when the 2-Ap opposed the 3Ј-T of the CPD, in sharp contrast to our results. We have preliminary data that show a similar fluorescence increase using a 3Ј-2-Ap probe as compared with the 5Ј-2-Ap case presented herein (data not shown). Quantitative analysis of the 3Ј-2-Ap results will provide more detailed information on how photolyase flips its substrate into the protein cavity, where it is repaired with nearly unit quantum efficiency.
Taken together, these data show that photolyase distorts the CPD-containing DNA duplex to such an extent as to disrupt base stacking around the lesion, strongly supporting the base flipping model originally proposed by Park et al. (5). The intriguing photolyase-thymine cocrystal structure introduces an ambiguity as to whether one or both thymines in the CPD actually enters the cavity. In another study, we have shown that both thymines are necessary to produce an apparent electrochromic shift in the PL ox -CPD absorption spectrum (22). Thus we expect that the entire CPD is accommodated by the protein cavity. Further experiments are planned to extend our knowledge of the extent to which the duplex is deformed to repair the CPD. The 2-Ap approach used in this study can be