Structural study of DNA duplexes containing the (6–4) photoproduct by fluorescence resonance energy transfer

Abstract

Fluorescence resonance energy transfer (FRET) experiments have been performed to elucidate the structural features of oligonucleotide duplexes containing the pyrimidine(6–4)pyrimidone photoproduct, which is one of the major DNA lesions formed at dipyrimidine sites by UV light. Synthetic 32mer duplexes with and without the (6–4) photoproduct were prepared and fluorescein and tetramethylrhodamine were attached, as a donor and an acceptor, respectively, to the aminohexyl linker at the C5 position of thymine in each strand. Steady-state and time-resolved analyses revealed that both the FRET efficiency and the fluorescence lifetime of the duplex containing the (6–4) photoproduct were almost identical to those of the undamaged duplex, while marked differences were observed for a cisplatin-modified duplex, as a model of kinked DNA. Lifetime measurements of a series of duplexes containing the (6–4) photoproduct, in which the fluorescein position was changed systematically, revealed a small unwinding at the damage site, but did not suggest a kinked structure. These results indicate that formation of the (6–4) photoproduct induces only a small change in the DNA structure, in contrast to the large kink at the (6–4) photoproduct site reported in an NMR study.

INTRODUCTION

The UV component of sunlight, as well as ionizing radiation and chemicals, produces various damaged bases in DNA, which induce genetic mutations and cell death (for reviews see 1,2). One of the major forms of DNA damage caused by UV light is the pyrimidine(6–4)pyrimidone photoproduct [abbreviated as (6–4) photoproduct], formed between two adjacent pyrimidine bases (Fig. 1). This DNA lesion is extremely mutagenic and predominantly induces a T→C or C→T transition at the 3′ base (3–6). In eukaryotic cells, the (6–4) photoproducts are repaired by the nucleotide excision repair (NER) pathway, in which several repair factors are recruited (7). To date, the mechanism of NER has been investigated in relation to the human inherited disease xeroderma pigmentosum (XP), which is caused by a defect in NER. In a proposed model, a complex of the XPA protein with replication protein A (RPA) recognizes bulky DNA lesions, including the (6–4) photoproduct, at the initial step of NER (8,9). It was shown that the XPC–HR23B complex plays the role of damage recognition prior to the XPA–RPA step in the global genome NER pathway (10).

Figure 1.

Sequences and structures of the duplexes used for the FRET experiments.

In a previous study, we found that the UV-damaged DNA-binding (UV-DDB) protein, which is related to XPE, is the major factor that recognizes this DNA lesion in human cells and we characterized this protein to gain insight into its unknown biological function (11). Since circular permutation analysis revealed that the DNA helix was kinked at the damage site in the complex with UV-DDB protein, at an angle of ∼55°, we concluded that a kinked structure, or potential of the damaged DNA to become kinked, was recognized by this protein. The next question was whether the UV-DDB protein recognizes an intrinsic kink of the damaged DNA or the potential to become kinked as a result of the damage. In order to determine the recognition mechanism, the structure of the (6–4) photoproduct-containing DNA must be elucidated. This information is extremely important because other proteins involved in replication and repair may recognize this highly mutagenic damage in the same way.

Two structural studies of (6–4) photoproduct-containing DNA have been reported thus far. One is the NMR solution structure of d(CGCAT(6–4)TACGC)·d(GCGTAATGCG), which revealed an overall helix bending of 44° (12). Mutagenesis and repair were discussed on the basis of this kinked structure (13). On the other hand, an unrestrained molecular dynamics analysis of the same duplex (14) revealed a significantly smaller bend angle (5°). This discrepancy is in contrast to the structure of DNA duplexes modified with cis-diamminedichloroplatinum(II) (cisplatin), in which a large kink has been observed by several methods (15–19). To elucidate the actual structure of the (6–4) photoproduct-containing duplex, other experiments are required. In this paper, we describe the analysis of DNA duplexes by fluorescence resonance energy transfer (FRET). FRET is a process by which the excited state energy of a fluorescent dye is transferred to another dye, when the emission spectrum of the former overlaps the absorption of the latter. Since the FRET efficiency is dependent upon the distance between the two dyes when the relative orientations of the donor and acceptor dipoles are random, this technique has been used to determine the global structures of nucleic acids (20). The results of this study indicate that the duplex containing the (6–4) photoproduct is not kinked in solution, in contrast to the positive control cisplatin-modified duplex. The structural change caused by photoproduct formation was also analyzed systematically by FRET measurement.

MATERIALS AND METHODS

Oligonucleotides for FRET experiments

The duplexes shown in Figure 1 were designed using molecular models constructed on a Silicon Graphics O₂ workstation with Insight II v.97.0 (Molecular Simulations). The crystal structure of a cisplatin-modified duplex (PDB code 1AIO) (17) was used for the kinked model (Fig. 2B). All oligonucleotides were synthesized on an Applied Biosystems 394 DNA/RNA synthesizer. For incorporation of the aminohexyl linker, Amino-Modifier C6 dT [5′-O-(4,4′-dimethoxytrityl)-5-[N-(6-trifluoroacetamidohexyl)-3-(E)-acrylamido]-2′-deoxyuridine 3′-(2-cyanoethyl)-N,N-diisopropylphosphoramidite] was purchased from Glen Research, and oligonucleotides containing the (6–4) photoproduct were synthesized using the previously described building block (21). After deprotection, the oligonucleotides were purified on a Waters µBondasphere C18 15 µm 300 Å column (7.8 × 300 mm) at a flow rate of 2.0 ml/min, with a linear gradient of acetonitrile (7–13% over 30 min) in 0.1 M triethylammonium acetate, pH 7.0. An intrastrand cisplatin adduct was formed as described previously (11). To attach the fluorescent dyes, the oligonucleotides bearing the aminohexyl group (50 nmol) were dissolved in 125 µl of 0.1 M NaHCO₃/Na₂CO₃, pH 9.0, and were mixed with 2 µl of a 0.17 M dimethylsulfoxide solution of the N-hydroxysuccinimide ester of 5-carboxyfluorescein (Berry & Associates) or 6-carboxytetramethylrhodamine (Glen Research). After incubation at 37°C for 2 h, the reagents were removed on a NAP-10 column (Amersham Pharmacia Biotech) and the labeled oligonucleotides were purified by reversed phase HPLC with a 10–23% acetonitrile gradient.

Figure 2.

Structure models of the DNA duplexes. (A) TT-L; (B) Pt-L. Fluorescein and tetramethylrhodamine attached to the linker are shown at the top and bottom of each duplex, respectively.

Duplex formation

For the steady-state and anisotropy measurements, the fluorescein-labeled complementary strand (240 pmol) was mixed with a 1.5-fold molar excess of the damaged strand, with and without tetramethylrhodamine, in a buffer (2.4 ml) containing 10 mM sodium phosphate, pH 7.9, 100 mM NaCl and 0.1 mM EDTA. Hybridization was carried out by heating the mixtures at 80°C and then cooling them gradually to 4°C. The anisotropy of tetramethylrhodamine was measured using the damaged strand (240 pmol) with a 1.5-fold excess of the complementary strand without fluorescein. The duplex concentration for the lifetime measurements was changed to 150 nM.

Steady-state FRET experiments

Steady-state fluorescence emission spectra were measured at 15°C in a Hitachi F-4500 fluorescence spectrophotometer. The excitation wavelength was 490 nm and emission data were collected at 1000 points between 500 and 700 nm. Absorption spectra were measured on a Beckman DU640 spectrophotometer. The critical transfer distance (R₀) (20) was determined to be 52.4 Å from the spectral properties of the donor and the acceptor. Fluorescence anisotropy (r), determined from the vertical and horizontal components of the emission (F_VV and F_VH, respectively) when a vertical excitation polarizer is used, was calculated using the equation r = (F_VV – GF_VH)/(F_VV + 2GF_VH). G is the correction factor given by F_HV/F_HH, where F_HV and F_HH represent the vertical and horizontal emission components, respectively, when a horizontal excitation polarizer is used. To determine the anisotropy, fluorescein and tetramethylrhodamine were excited at 490 and 550 nm and their emissions were detected at 520 and 580 nm, respectively.

Fluorescence lifetime measurement

Fluorescein emission decay curves were measured at 15°C in a Horiba NAES-700F fluorescence spectrophotometer. Toshiba Y-50 and KL-51 filters were used for excitation and detection, respectively, and data collection was performed up to 10 000 single photon counts for each duplex. The fluorescence lifetimes were obtained from single exponential fits of the decay curves, because the donor was found to have one major decay process. The extracted ranges (0–20 ns) are illustrated in Figures 3C and 4F. The donor–acceptor distance (R) was calculated using the equation E = 1 – (τ_DA/τ_D) = 1/{1 + (R/R₀)⁶}, where E is the efficiency of energy transfer and τ_DA and τ_D are the lifetimes of the donor fluorescence in the presence and absence of the acceptor, respectively.

Figure 3.

FRET experiments using the cisplatin-modified duplex. (A and B) Steady-state fluorescence emission spectra of GG-L (A) and Pt-L (B). Thin lines are the spectra of the duplexes without tetramethylrhodamine. (C) Fluorescein emission decay curves of GG-L (circles) and Pt-L (squares). Open and closed markers represent the duplexes without and with tetramethylrhodamine, respectively. Single exponential curve fits are overlaid.

Figure 4.

FRET experiments using the (6–4) photoproduct-containing duplexes. (A–E) Steady-state fluorescence emission spectra of TT-L (A), (6–4)-L1 (B), (6–4)-L2 (C), TT-S (D) and (6–4)-S (E). (F) Fluorescein emission decay curves of TT-L (circles) and (6–4)-L1 (squares) with single exponential curve fits. Thin lines in (A–E) and open and closed markers in (F) are explained in the legend to Figure 3.

RESULTS

Design and synthesis of duplexes for FRET experiments

The duplexes designed for the FRET experiments are shown in Figure 1. Besides the (6–4) photoproduct, a duplex containing an intrastrand cisplatin adduct, which had been demonstrated to be kinked by several methods (15–19), was prepared as a positive control. Fluorescein and tetramethylrhodamine were used as the energy donor and acceptor, respectively. In FRET experiments, fluorescent dyes are usually attached at the ends of oligonucleotides, but in such cases a hypothesis is required to determine the geometric positions of the dyes. We simplified this determination by attaching them to an aminohexyl linker at the C5 position of an internal thymidine. When duplexes were formed, the linkers were expected to be stretched perpendicularly to the helix axis in the major groove of the DNA. The sites of dye attachment were chosen so that the two dyes reside on the same side of the helix and are closer to each other if DNA kinking occurs toward the major groove. For this purpose, molecular models of straight and kinked DNA were constructed, as shown in Figure 2, using the canonical B-form and the crystal structure of a cisplatin-modified duplex (17), respectively. At a distance of 23 bp, the vector from the donor to the acceptor was parallel to the helix axis in the unmodified DNA and was oblique in the cisplatin-modified duplex. For duplexes containing the (6–4) photoproduct, a similar or larger change of the donor–acceptor distance was expected if a kink occurred. In order to detect even minor structural changes caused by photoproduct formation, several (6–4) photoproduct-containing duplexes were designed with various donor–acceptor distances (see Fig. 5).

Figure 5.

Determination of the unwinding caused by (6–4) photoproduct formation. (A) Duplexes used for the systematic experiments. The positions of the fluorescein attachment are indicated by arrows with the distances from tetramethylrhodamine. TT represents the site of the (6–4) photoproduct. (B) The distances between the fluorescent dyes in the undamaged (open squares) and (6–4) photoproduct-containing (solid squares) duplexes. Each value is within ±1.2 Å. (C) DNA models used for the data fitting in (B). Unkinked models of the (6–4) photoproduct-containing duplex (left) and the B-DNA duplex (right) are illustrated. Fitting parameters are described in the text.

The oligonucleotides containing the (6–4) photoproduct were synthesized on a DNA synthesizer using a (6–4) photoproduct building block (21) and an aminohexyl linker was incorporated using a phosphoramidite of thymidine modified at the C5 position. After deprotection, all of the oligonucleotides were purified by reverse phase HPLC. A cisplatin adduct was formed between adjacent guanines in a 12mer and, after purification by HPLC, the oligonucleotides were ligated as described previously (11). Each oligonucleotide was treated with an N-hydroxysuccinimide ester of 5-carboxyfluorescein or 6-carboxytetramethylrhodamine and then again purified by HPLC. The duplexes shown in Figure 1 were prepared by annealing the complementary strand labeled with fluorescein to a 1.5-fold excess of the damaged strand, with or without tetramethylrhodamine.

Analysis of the cisplatin-modified duplex by FRET

To confirm that DNA kinks could be detected in the FRET experiments, a duplex containing an intrastrand cisplatin adduct was used as a positive control. Prior to measurement of the emission spectra, the anisotropy of each dye in the duplexes was measured to determine its rotational mobility. The values obtained for fluorescein and tetramethylrhodamine in each duplex were lower than 0.04 and between 0.19 and 0.20, respectively, which indicated that rotation of the dyes was free enough for structural analysis by FRET (22,23).

The structural information was obtained by both steady-state FRET measurement and fluorescence lifetime determination. The emission spectra of GG-L and Pt-L, excited at 490 nm, are shown in Figure 3A and B, respectively. The fluorescence intensity of fluorescein at 520 nm was obviously reduced by modification with cisplatin. Fluorescein emission decay curves were measured for these duplexes, as shown in Figure 3C. Since the donor was found to have one major decay process, fluorescence lifetimes were obtained from single exponential fits of these curves, and the results are listed in Table 1. The marked changes in both the FRET efficiency and the fluorescence lifetime observed for the cisplatin-modified duplex (Pt-L) indicated that formation of the cisplatin adduct obviously shortened the donor–acceptor distance, as a consequence of a helix kink at the modification site.

Table 1. Fluorescence lifetime analysis.

Duplex		Lifetime (ns)	χ2	Distance (Å)
With fluorescein
	TT-L	4.55 ± 0.05	1.42
	(6–4)-L1	4.44 ± 0.05	1.14
	GG-L	4.45 ± 0.05	1.26
	Pt-L	4.44 ± 0.05	1.47
With fluorescein and tetramethylrhodamine
	TT-L	4.12 ± 0.05	1.35	76.2 ± 1.1
	(6–4)-L1	4.12 ± 0.05	1.59	79.9 ± 1.1
	(6–4)-L2	4.11 ± 0.05	1.03	79.5 ± 1.1
	GG-L	4.07 ± 0.04	1.69	77.7 ± 1.1
	Pt-L	3.84 ± 0.04	1.23	71.5 ± 1.0

FRET analysis of the (6–4) photoproduct-containing duplexes

To analyze the structure of the (6–4) photoproduct-containing DNA, we prepared three duplexes in which the donor and the acceptor were separated by 23, 24 and 17 bp, as shown in Figure 1. Steady-state emission spectra of these duplexes are shown in Figure 4B, C and E, respectively. In contrast to the cisplatin-modified duplex, the fluorescence intensity was not reduced by photoproduct formation. No significant difference was observed in the fluorescence lifetimes (Table 1) derived from the fluorescein emission decay curves (Fig. 4F). Consequently, the donor–acceptor distances of the (6–4) photoproduct-containing duplexes were not shorter than that of the undamaged one. In order to examine the structural features of the (6–4) photoproduct-containing duplex in detail, a series of oligonucleotide duplexes in which the fluorescein position was changed systematically was prepared (Fig. 5A) and the distances between the dyes were measured. The results obtained from the lifetime measurements are shown in Figure 5B. The data were fitted to theoretical curves using the model shown in Figure 5C. In this model, which shows the undamaged B-form duplex, the dihedral angle between the two dye linkers is 180° at a distance of 18 bp and is 0° at 23 bp. The parameters obtained for the damaged duplex by the fitting were ΔH = 0.37 Å, L = 13 Å and ω = 56°, which represent a rise of the helical axis, the distance between the dye and the helical axis and the unwinding angle caused by photoproduct formation, respectively.

DISCUSSION

In this study, we have analyzed the structure of DNA duplexes containing the (6–4) photoproduct using FRET. Since structural studies on the (6–4) photoproduct have been limited by the difficulties of sample preparation, it should be noted that our method for the chemical synthesis of oligonucleotides containing this photoproduct (21,24) enabled us to perform the physicochemical experiments.

Although the FRET technique has been used to analyze nucleic acid structures (20,25–28), there have been only a few reports on duplex structures, and most of them described DNA bending induced by protein binding (29–32). Examples of the bending of DNA alone are studies on bulged nucleotides (33) and A₆ blocks (22). Therefore, we first verified whether the DNA kinks caused by nucleobase damage could be detected in our FRET experiments, using a duplex containing an intrastrand cisplatin adduct. Jamieson et al. (34) used FRET to study cisplatin-modified DNA. However, they investigated the structural changes induced by complex formation with a high mobility group domain protein and our system was quite different from theirs. As shown in Figure 3 and Table 1, clear differences were observed between the unmodified and cisplatin-modified duplexes. These differences could be attributed to the helix kink caused by modification with cisplatin. The distances were calculated using the FRET efficiencies in the lifetime measurements because the lifetime analysis is unaffected by duplex concentration and fluorescence of the donor was found to have one major decay process (35). The donor–acceptor distance for GG-L was calculated to be 77.7 ± 1.1 Å, using the equation described in Materials and Methods, and it is very close to the value expected for a distance of 23 bp (78 Å) in B-form DNA. This result demonstrated that the design of the labeled duplex illustrated in Figure 2A was suitable for the structural analysis. The shorter donor–acceptor distance calculated for Pt-L (71.5 ± 1.0 Å) can be explained by DNA kinking. Using the molecular model constructed for fixing the sites of dye attachment (Fig. 2B) and the value of L (Fig. 5C), the bend angle of Pt-L was estimated to be ∼29°. Takahara et al. reported that the bend angles of a cisplatin-modified duplex were 39° and 55° in their crystal structures and noted that a 34° bend was calculated for the NMR solution structure (36). Bend angles of 32–34° were reported in a study using gel electrophoresis (16). Since these reports, using several methods, suggested a bend angle of 30–40° and the angle derived from our FRET experiment was close to these values, the helix kink of damaged DNA can be observed by this method.

In contrast to the cisplatin-modified duplex, neither the steady-state FRET efficiency nor the fluorescence lifetime was changed by formation of the (6–4) photoproduct when similar duplexes were used (Fig. 4A–C and F). Steady-state analysis at larger FRET efficiencies [TT-S and (6–4)-S] also showed no difference between the undamaged and photoproduct-containing duplexes (Fig. 4D and E). The distances between the dyes were calculated by lifetime analysis of the systematic duplexes with and without the photoproduct (Fig. 5A). Both sets of data fit well with the unkinked model (Fig. 5C), indicating that the (6–4) photoproduct is unlikely to induce a kink in the duplex. The unwinding angle ω (56°) was somewhat larger than the value reported in the NMR study (30°) (12). The value of L (13 Å) was smaller than in the case where the dyes were attached at the end of the duplex (23), but the value was reasonable because the linker was attached to a thymine base at the bottom of the major groove in our case (Figs 1 and 2). It may be possible that photoproduct formation makes the duplex flexible. However, DNA kinking toward both the major and minor grooves, which would offset the FRET effects, is unlikely because this photoproduct has a rigid structure with covalently linked bases and DNA cyclization experiments (37) have ruled out this flexibility (Y.Fujiwara and S.Iwai, unpublished results). In contrast to structural studies of proteins, a set of only NOE data is insufficient to yield a definitive solution structure, such as global curvature, in the NMR study of DNA, although some local conformations have been defined in detail using inter-residual NOEs between neighboring base pairs (38,39). Therefore, the structural features should be verified by another method, as in the case of cisplatin-modified DNA (15–19). To our knowledge, there is no experimental demonstration besides a NMR study (12) and a molecular dynamics calculation (14) of the global structure of (6–4) photoproduct-containing DNA. Our present study can rectify the shortage of information of the structure of this damaged DNA, which is very important, especially in understanding repair of this damage. Our data support the structure obtained by unrestrained molecular dynamics (14) and we suppose that the kinked NMR structure (12) may reflect the initial model used for the molecular dynamics simulation and refinement.

From the viewpoint of damage recognition by proteins, the (6–4) photoproduct may resemble the abasic lesion. In a previous study (11), UV-DDB protein specifically bound DNA duplexes containing a (6–4) photoproduct or abasic site, and both of these duplexes were kinked in the complex at an angle of ∼55°. Solution structures of several DNA duplexes containing an abasic site have been reported. In most cases, a helix kink was not described (40,41) or the bend angle was small (∼10°) (42). Coppel et al. reported that a kink of ∼30° was induced by the abasic site (43), but the hydrogen bonding interaction found between the bases opposite and adjacent to the abasic site may have stabilized the kinked structure in their case. Our previous FRET experiment (11) supported a nearly straight structure. In the present study also, we have shown that the (6–4) photoproduct does not induce a large kink in DNA helices in solution. The kinked DNA structures found in complexes with UV-DDB protein suggest that its specific binding induces a kink in the DNA. This property is often observed for DNA-binding proteins. T4 endonuclease V recognizes a cyclobutane pyrimidine dimer site in DNA and kinks the helix at an angle of 60° (44), although the duplex is not significantly kinked by itself in solution (13,45,46). When EcoRV binds DNA in a non-specific manner, the DNA has a linear form. However, complex formation with its recognition sequence results in large helix bending (∼50°). The structural changes in the DNA backbone and the bases and the expulsion of water molecules from the interface, to allow more intimate contacts, are considered to trigger the conversion from a non-specific to a specific complex (47). Our present study suggests that UV-DDB protein does not recognize the kinked structure, but another feature of the damaged site. This protein may recognize the potential of the damaged DNA to become kinked, which could induce conformational changes in the specific complex, as in the cases of T4 endonuclease V and EcoRV.