Electron transfer in DNA duplexes containing 2-methyl-1,4-naphthoquinone

Abstract

2-Methyl-1,4-naphthoquinone (menadione, MQ) was linked to synthetic oligonucleotides and exposed to near-UV light to generate base radical cations in DNA. This model system of electron transfer induced alkali-labile breaks at GG doublets, similar to anthraquinone and metallointercalators systems. In sharp contrast to other systems, the photolysis of MQ–DNA duplexes gave interstrand cross-links and alkali-labile breaks at bases on the complementary strand opposite the MQ moiety. For sequences with an internal MQ, the formation of cross-links with A and C opposite the MQ moiety was 2- to 3-fold greater than that with G and T. The yield of cross-links was more than 10-fold greater than that of breaks opposite MQ, which in turn was more than 2-fold greater than breaks at GG doublets. The yield of damage at GG doublets greatly increased for a sequence with a terminal MQ. The distribution of base damage was measured by enzymatic digestion and HPLC analysis (dAdo > dThd > dGuo > dCyd). The formation of novel products in MQ–DNA duplexes was attributed to the ability of excited MQ to generate the radical cations of all four DNA bases; thus, this photochemical reaction provides an ideal model system to study the effects of ionizing radiation and one-electron oxidants.

INTRODUCTION

Model systems of electron transfer in DNA have been exploited to understand the biological effects of ionizing radiation, to probe the structure and biochemistry of DNA, and to develop molecular scale electronic devices (1). The majority of experimental systems involving DNA take advantage of photochemical electron transfer to inject electron-holes at specific sites into DNA. Three mechanisms have been proposed to account for the formation of DNA damage at a distance from the initial site of electron-hole injection: long-range tunneling, multi-step hopping, and polaron-mediated hopping. The mechanism of long-range tunneling is supported by the fast and efficient rates of electron transfer induced by tethered metallo-intercalators with DNA damage observed as far away as 60 bp (2–4). In contrast, the mechanism of multi-step hopping consists of a combination of superexchange and hopping events. It is supported by studies involving the photolysis of 4′-pivaloylated thymidine, stilbene hairpins, as well as by theoretical studies (5–7). The efficiency of DNA damage from G to a site of lower oxidation potential, GGG, separated by 1–3 AT base pairs falls off sharply with distance according to the superexchange hypothesis. The transport of electron holes is less efficient and dependent on separation distance when the bridge is composed of >3 AT base pairs, consistent with a mechanism of hopping (8,9). The insertion of GC base pairs within donor acceptor bridges containing multiple AT base pairs greatly enhances the efficiency of electron-hole transfer to remote GGG sites. Lastly, the mechanism of polaron-mediated hopping is supported by studies using anthraquinones tethered to DNA duplexes (10–13). This model system is characterized by the formation of DNA damage at long distances, which falls off slowly with distance (β < 0.2 Å⁻¹) and is sensitive to DNA primary and secondary structure, particularly the presence of water molecules and ionic charges (12). A major difference between the two models of hopping is that the radical cation is centralized at a single base in multi-step hopping but delocalized over several bases in polaron-mediated hopping.

The chemical reactions of DNA radical cations terminates electron transfer in DNA, leading to the formation of stable products. In the majority of model systems, the detection of damage at a distance from the initial site of the radical cation is based on product analysis involving ³²P-labeling and PAGE. Typically, the site of damage is revealed by heating DNA in hot piperidine to convert base and sugar modifications to breaks. This approach is very sensitive and permits the analysis of damage at specific sites in DNA. However, a drawback is that hot piperidine does not convert all damaged sites into breaks and very little information is obtained about the structure of the damage. An alternative method is to use DNA repair enzymes, such as formamidopyrimidine N-glycosylase and endonuclease III, to cleave DNA at sites of base damage. Again, this method is not quantitative because not all damage is converted to breaks and the activity of the enzymes varies from batch to batch. For example, some studies on electron transfer in DNA report several fold more formamidopyrimidine N-glycosylase-sensitive breaks compared to piperidine-induced breaks (14), whereas others report comparable levels (15–17). Because of the complex nature of DNA damage, and the multitude of intermediate species, it is important to further characterize this damage and to develop appropriate methods of analysis.

2-Methyl-1,4-naphthoquinone (menadione; MQ) is an excellent near-UV photosensitizer that induces one-electron oxidation of nucleic acid components (18–27). The mechanism of oxidation involves electron transfer from pyrimidine and purine bases to triplet excited MQ, which generates the corresponding radical cations of DNA bases and MQ radical anions, respectively (22). The rate of reaction of triplet MQ is nearly diffusion controlled (k > 10⁹ M⁻¹ s⁻¹) with several nucleic acid components, including thymidine (dThd), 2′-deoxycytidine (dCyd), 2′-deoxyadenosine (dAdo) and 2′-deoxyguanosine (dGuo), as well as the corresponding monophosphates (22). In addition, the yield of stable products from dThd was equal to the production of triplet MQ (66%), indicating that 100% of dThd radical cations transform to stable products (22). Previously, we characterized a large number of products from the photooxidation of dThd, dCyd and 5-methyl-2′-deoxycytidine using MQ and near-UV light (23,25–27). However, the photooxidation of DNA with free MQ is greatly limited because of the low rate of reaction of triplet MQ with DNA and the relatively high rate of self-quenching. In this work, we have overcome these problems by attaching the MQ moiety via a short chain amino linker to DNA. These studies demonstrate the formation of novel products, including DNA interstrand cross-links, piperidine-labile breaks opposite to the MQ moiety, and loss of dAdo, dThd, dGuo and dCyd, all of which are consistent with the formation of radical cations of all four DNA bases and their transformation to stable products.

MATERIALS AND METHODS

General methods

Chemical reagents were obtained from Aldrich (Milwaukee, WI) unless otherwise stated. N-bromosuccinimide was recrystallized in hot water before use. T4 polynucleotide kinase was obtained from USB Corporation (Cleveland, Ohio) and [γ-³²P]ATP was purchased from Perkin Elmer (Norwalk, CT). Nuclease P1 and alkaline phosphatase were from Roche. Oligonucleotides were synthesized by AlphaDNA (Montreal, QC) and their mass was verified by MALDI-TOF analyses. Uni-link AminoModifier phosphoramidite was obtained from Clontech (Palo Alto, CA) and incorporated into an internal position during oligonucleotide synthesis. HPLC-UV analysis was carried out on Alliance systems (Waters Corp, Milford, MA; Models 2795 or 2690) connected to a dual wavelength UV detector (Model 2487) and Millenium workstation (Model version 4). ¹H-NMR spectra were recorded using a 300 MHz spectrometer (Bruker, Milton, ON). MALDI-TOF analysis was performed on a Biflex system (Bruker). A Model J810 spectropolarimeter was used to record UV melting curves and CD spectra (JASCO, Easton, MD).

2-Bromomethyl-6-methylnaphthalene

This synthesis was adapted from Futamura et al. (28). N-bromosuccinimide (NBS, 630 mg, 3.5 mmol) was added to a solution of 2,6-dimethylnaphthalene (500 mg, 3.2 mmol) of anhydrous benzene (64 ml). The solution was bubbled with nitrogen and irradiated for 40 min at room temperature with a 250 W medium pressure Hg lamp (Blak-Ray, Ultraviolet products, Upland, CA) equipped with an UV filter (λ ≥ 500 nm). The mixture was filtered, evaporated to dryness and purified by flash column chromatography (silica gel, toluene/hexane = 20:1). R_f = 0.7. Yield 526 mg (70%). ¹H-NMR (CDCl₃) δ 2.5 (s, 3H, −CH₃), 4.65 (s, 2H, −CH₂Br), 7.3–7.9 (m, 6H, H₃, H₄, H₅, H₇, H₈, H₉).

6-Bromomethyl-2-methyl-1,4-naphthoquinone

This synthesis was adapted from Antonini et al. (29). 2-Bromomethyl-6-methylnaphthalene (1 g, 4.2 mmol) was dissolved in 17 ml of 90% acetic acid and maintained in a melting ice bath at 10°C. Chromium (VI) oxide (1.3 g, 13 mmol) was dissolved in 3.6 ml of 50% acetic acid and added drop-wise to the solution of naphthalene while stirring for 40 min. The mixture was then heated at 50°C for 30 min, quenched with cold water and maintained on ice for 5 min to allow maximum recovery of brominated MQ. The crude product was collected by filtration, dissolved in ethyl acetate, dried with Na₂SO₄, evaporated to dryness and purified by flash column chromatography (silica gel, hexane/ethyl acetate = 9:1; R_f = 0.2). ¹H NMR (DMSO-d₆) δ 2.5 (s, 3H, –CH₃), 4.85 (s, 2H–CH₂Br), 7.0 (s, 1H, H₃), 7.85–8.05 (m, 3H, H₇, H₈, H₉).

6-Hydroxymethyl-2-methyl-1,4-naphthoquinone

6-Bromomethyl-2-methyl-1,4-naphtho-quinone (500 mg, 1.9 mmol) was refluxed in 15 ml water/dioxane (3:2) until the starting material was consumed as indicated by TLC (CH₂Cl₂/MeOH = 10:1). The solvent was evaporated and the product was subsequently used for the next step without further purification. ¹H NMR (CDCl₃) δ 2.3 (s, 1H, −CH₃), 2.7 (s, 1H, –OH), 4.85 (s, 1H, –CH₂), 6.85 (s, 1H, H₃), 7.7 (d, 1H, H₇), 8.0 (m, 2H, H₅, H₈).

2-Methyl-1,4-naphthoquino-6-carboxylic acid

6-Hydroxymethyl-2-methyl-1,4-naphthoquinone (190 mg, 0.95 mmol) was dissolved in 3.6 ml of 90% acetic acid and maintained in a melting ice bath at 10°C. Chromium (VI) oxide (215 mg, 2.15 mmol) was dissolved in 1.2 ml 50% acetic acid and added drop-wise to 6-hydroxymethyl-2-methyl-1,4-naphthoquinone while stirring for 40 min. The black mixture was then heated at 50°C for 30 min, quenched with cold water and maintained on ice for 5 min. The crude product was collected by filtration, dissolved in ethyl acetate, dried with Na₂SO₄, evaporated to dryness and purified by flash column chromatography (silica gel, CH₂Cl₂/MeOH = 10:1). R_f = 0.4. Yield 82 mg (40%). ¹H NMR (DMSO-d₆) δ 2.1 (s, 3H, –CH₃), 7.0 (s, 1H, H₃), 8.1 (d, 1H, H₇), 8.3 (d, 1H, H₈) 8.4 (s, 1H, H₅), 13.6 (s, 1H, acid).

2-Methyl-1,4-naphthoquinone-6-carboxylic acid N-hydroxysuccinimide ester (MQ-NHS ester)

This synthesis was adapted from Oswald et al. (30). The carboxylic acid derivative (200 mg, 0.9 mmol), was dissolved in 40 ml anhydrous N,N-dimethylformamide (DMF), followed by the addition of dicyclohexylcarbodiimide (DCC, 560 mg, 2.7 mmol) and N-hydroxysuccinimide (NHS, 320 mg, 2.8 mmol). The mixture was stirred overnight under argon and filtered. The filtrate was lyophilized, dissolved in 40 ml CH₂Cl₂, filtered again and evaporated to dryness. The product was then purified on reversed phase HPLC (ODS-AQ 25 cm × 10 mm i.d.; 2.5 ml/min, 50% acetonitrile/water, isocratic). Yield 112 mg (40%). ¹H NMR (DMSO-d₆) δ 2.1 (s, 3H, −CH₃), 3.3 (s, 4H, succinimide), 7.1 (s, 1H, H₃), 8.2 (d, 1H, H₇), 8.45–8.55 (m, 2H, H₈, H₅).

Conjugation of MQ-NHS ester with oligonucleotides

MQ-NHS ester was coupled to oligonucleotides by optimization of a standard protocol (31). The coupling reaction was carried out by mixing a solution of purified oligonucleotides containing amino modifier (200 μg, 19.7 nmol), precipitated and dissolved in 200 μl of KH₂PO₄ 20 mM buffer (pH 7.5), with a solution of MQ-NHS ester (0.5 mg, 3 μmol) in DMF (200 μl). The reaction was complete after 7 h at 37°C as monitored by HPLC with dual detection at 260 and 340 nm (Yield > 90%).

PAGE analyses

Gel purified oligonucleotides were labeled at the 5′ end with [γ-³²P]ATP using T4 polynucleotide kinase according to the standard procedure. Labeled DNA was immediately purified on Sephadex resin microcolumns and collected in 10 mM NaHPO₄ buffer (pH 7.0). The unlabeled MQ-conjugated and ³²P-labeled complementary strands were subsequently hybridized by heating equal molar solutions (1 μM) in 10 mM sodium phosphate buffer (pH 7.0) for 5 min at 65°C in a water bath and then allowing the mixture to cool to 30°C (1 h). The extent of hybridization was 100% as indicated by non-denaturing gel electrophoresis (10%; 19:1 acrylamide/bis-acrylamide). During photolysis, MQ–DNA duplex (7 μl) was withdrawn for PAGE analysis. The samples were divided into two equal aliquots for the analysis of DNA breaks, with and without piperidine treatment (1 M piperidine, 90°C, 30 min). All samples were evaporated to dryness in a vacuum concentrator and suspended in denaturing loading buffer, heated to 80°C for 10 min, quenched on ice and then subjected to PAGE analysis on a 20% 19:1 acrylamide/bis-acrylamide denaturing gel (7 M urea). The gels were exposed overnight on a phosphor screen cassette and scanned using a phosphorimager and a Storm Imagequant software (Molecular Dynamics, Sunnyvale, CA).

HPLC-UV analysis

MQ-conjugated and complementary strands (200 μg each) were dissolved in 800 μl of 10 mM phosphate buffer (pH 7.0). The mixture of oligonucleotides was heated for 5 min at 70°C in a water bath to allow for complete denaturation, and then the mixture was slowly cooled to room temperature over a period of 1 h. The resulting solution of MQ–DNA duplex (25 μM) was exposed to near-UV photolysis (365 nm) and aliquots (75 μl) of sample were collected at several time-points (0, 15, 30, 45, 60, 90, 120 and 180 min). For each aliquot, two HPLC-UV analyses were performed. The analysis of oligonucleotides was carried out on a reversed phase column (ODS-AQ 250 × 4.6 mm; YMC, Waters Corp.) using a gradient starting at 92% solvent A and 8% solvent B for 5 min and going to 80% solvent A and 20% solvent B in the next 24 min (0.5%/min), where solvent A is composed of 25 mM triethylammonium acetate (TEAA, pH = 7.0 and 2.5% acetonitrile) and solvent B is composed of 95% acetonitrile and 5% water. The flow rate was 1 ml/min and the column was maintained at a temperature of 60°C. The analysis of individual nucleosides in MQ–DNA duplexes was carried out by enzymatic digestion with P1 nuclease and alkaline phosphatase followed by HPLC-UV analysis of the mixture of nucleosides using the same column as above except that it was done under isocratic conditions with 10% methanol in 10 mM sodium phosphate buffer (pH = 7.0) at a flow rate of 1 ml/min. Oligonucleotides and nucleosides were detected at 260 and 340 nm. The analysis of 8-oxo-7,8-dihydro-2′-deoxyguanosine, 8-oxo-7,8-dihydro-2′-deoxyadenosine and 5-hydroxy-2′-deoxycytidine was carried by HPLC-UV (above) coupled to an electrochemical detection using a Coulochem detector (ESA Associates, Chelmsford, MA) with a Model 5011 flow cell (ESA Associates) set at 500 mV versus Pd reference electrode.

Near-UV photolysis

Photolysis was carried out with a 1000 W Hg–Xe arc lamp (Spectra-Physics, Oriel Instruments, Stratford, CT) fitted with an infrared filter and diffracting monochromator (Spectral Energy Corporation, Chester, NY) set at 365 ± 2 nm. The beam of light was focused on a quartz cuvette (pathlength = 1 cm). The solutions were bubbled with oxygen for 20 min before photolysis and during photolysis. The number of incident photons impinging on the solution was determined by actinometry using the well-studied intermolecular photochemical reaction of MQ and dThd with the same reaction conditions (dThd, 1–10 mM); MQ (0.53 mM); phosphate buffer (10 mM, pH 7.0) (22). Similar to the results reported previously, the inverse of the rate of decomposition of dThd was linear with the inverse of the concentration of dThd (1–10 mM) (data not shown). This indicates that the quantum yield for the decomposition of dThd depends on the concentration of dThd according to competition kinetics. Therefore, the number of incident photons in experiments with MQ–DNA duplexes was estimated by measuring the rate of decomposition of dThd in solutions of dThd and MQ using the same sample cuvette and optical geometry. The rate of decomposition (0–60 min) of dThd was determined by HPLC-UV (10% methanol, 50 mM phosphate buffer, pH 5.5). From the rate of decomposition of dThd, the number of absorbed photons per unit time was calculated assuming a quantum yield (Φ) for the decomposition of dThd of 0.56 (this yield was estimated from analysis of 1/Φ versus 1/[dThd]; [Figure 3; (22)]. The number of incident photons was obtained from the optical density (OD) according to the following equation: photons (absorbed) = photons (incident) × (1−10^OD). With solutions containing 10 mM dThd and 0.53 mM MQ (OD = 0.68), the average number of incident photons was 1.26 × 10⁻⁶ Einstein/min (+/−20%). The quantum yield for the formation of products in MQ–DNA duplexes was calculated from the number of incident photons (above) and the fraction of light absorbed by MQ–DNA duplexes (1−10^−OD). For experiments involving HPLC analysis, the concentration of duplexes was 20–30 μM corresponding to an OD of ∼0.06 at 365 nm. For experiments involving PAGE analysis, it was not possible to measure the OD of the photolysis solution, and thus, the quantum yields of damage were estimated by assuming that the yields of cross-links were the same as those in HPLC-UV analysis.

Figure 3.

Detection of oligonucleotide products by HPLC-UV. Near-UV photolysis (365 nm) was carried out with non-labeled MQ–DNA duplex I (see Figure 1) at a concentration of 25 μM (400 μg/800 μl) in 10 mM sodium phosphate buffer (pH = 7.0). The formation of products was monitored from 0 to 180 min of photolysis. The solid line represents oligonucleotide products at initial times (0 min), which show the maximum level of complementary and MQ strands, and final exposure times (180 min), which show the maximum level of cross-links. The dotted lines shows the progression of damage at 30, 60 and 120 min with decreases in the complementary and MQ strands, and increases in cross-links at the longer exposure times.

Melting point and circular dichroism

To determine the melting point, complementary DNA sequences (7 μM; each strand) were annealed in 300 μl of 10 mM sodium phosphate buffer (pH 7.0). The solutions were placed in cuvettes with a 1 mm path length and the absorption was monitored at 260 nm from 20 to 90°C, heating at a rate of 1°C/min. The melting point was determined from the first derivative plot of absorption against temperature. For circular dichroism, the complementary DNA sequences (7 μM; each strand) were annealed under the same conditions as above. The solutions were placed in cuvettes with a 1 mm path length and the absorption was monitored between 220 and 340 nm at 20°C.

RESULTS

Synthesis and structure of MQ–DNA duplexes

The synthesis of oligonucleotides containing MQ was very challenging. The synthesis of modified phosphoramides and their assembly into oligonucleotides, used for the incorporation of anthraquinone into oligonucleotides (32), was not feasible in the case of MQ. The main problem was that the synthesis of the modified phosphoramidite via conjugation of 2-bromomethyl-1,4-naphthoquinone to N₃-benzoyl-dibutylstannylene uridine led to side reactions and extensive decomposition of the quinone moiety (yield < 10%). In addition, the MQ moiety was sensitive to alkali conditions required for the deprotection of oligonucleotides. Although we attempted to protect the quinone moiety by reduction and alkylation, subsequent deprotection and regeneration of the quinone led to modifications of DNA bases. Therefore, we developed an alternative approach involving the synthesis of the NHS ester of MQ and its coupling to oligonucleotides containing a reactive amino group. The coupling reaction gave yields of >90% using a 150-fold excess of the ester to DNA in DMF/H₂O (Scheme 1). This work reports the results for the photolysis of five oligonucleotide duplexes: four duplexes contain MQ attached to an internal position with A, T, C or G opposite MQ on the complementary strand; and one duplex contains MQ attached to the 5′-terminus (Chart ).

Scheme 1. Conjugation of menadione (MQ) with oligonucleotides.

Figure 1.

Chart 1. MQ-DNA duplexes (I-V).

The structure of oligonucleotides containing MQ was confirmed by enzymatic hydrolysis with P1 nuclease and alkaline phosphatase followed by analysis of the mixture by HPLC/UV and MS analyses. HPLC/UV analysis indicated the loss of a single A from the oligonucleotide composition of normal nucleosides and the appearance of a product with a longer retention time, i.e. more non-polar. The absorption of this product at 340 nm (260/340 = 10) pointed to an adduct composed of intact MQ and A moieties. Indeed, ESI/MS and MS/MS analyses depicted a molecular ion at m/z 703 (M + 2Na⁺) and fragmentation of this ion into two major ions at m/z −135 (loss of adenine) and m/z −251 (loss of dAdo). These analyses confirm the proposed structure in which MQ is covalently attached to A via a carbon linker (Scheme 2). Within these adducts (referred to as MQ–A), the A residue was 3′ to the MQ moiety because enzymatic hydrolysis of a 33mer containing 5′-T–MQ–A-3′ gave MQ–A whereas digestion of a 33mer containing 5′-A–MQ–T-3′ gave another compound with a different retention time. The structure of oligonucleotides containing an internal MQ moiety was also supported by MALDI-TOF analysis (m/z: measured, 10 138.9; calculated, 10 137.8). The melting point (T_m) of duplexes containing internal MQ was 2–3°C lower than that of non-modified duplexes, e.g. the T_m of duplex I containing A opposite MQ was 47.2°C, whereas the corresponding non-modified duplex was 49.8°C. Lastly, there was no apparent difference between the modified and non-modified duplexes by circular dichroism analysis.

Scheme 2. Proposed mechanism of MQ-sensitized photooxidation of DNA.

PAGE analysis of MQ–DNA duplexes

DNA damage was first examined by PAGE analyses (Figure 1a and b). These analyses revealed the formation of interstrand cross-links, piperidine-labile breaks opposite the MQ moiety, and piperidine-induced cleavage at GG doublets. Several distinct bands appeared on denaturing PAGE analyses, which migrated slower than the parent oligonucleotide, consistent with the formation of interstrand cross-links (see lane 2, Figure 1a). Upon treatment with hot piperidine, the majority of cross-links (>50%) converted to single strands that migrated with the same rate as the parent ³²P-labeled strand. Similar results were obtained for duplexes II–IV. Photolysis of duplex V in which MQ was attached to a terminal position also gave interstrand cross-links (Figure 1b). In control experiments, there was no detectable damage when ³²P-labeled duplex containing a naphthalene moiety in place of an MQ moiety was irradiated in the presence of non-labeled duplex containing a MQ moiety. Thus, the formation of damage in MQ–DNA duplexes takes place exclusively by the reaction of excited MQ in MQ–DNA duplexes and not by the reaction of small diffusing species generated by MQ photolysis.

Figure 1.

Detection of DNA damage by PAGE. (a) Near-UV photolysis (365 nm) was of ³²P-labeled MQ–DNA duplex I at a concentration of 1 μM (4 μg/200 μl) in 10 mM sodium phosphate buffer (pH = 7.0). PAGE analysis was carried out without further treatment [left side of gel; Piperidine (−)] and immediately following treatment with 1 M piperidine for 30 min at 90°C [right side; Piperidine (+)]. Lanes 1–7 correspond to progressively longer times of photolysis in the order of 0, 5, 10, 15, 30, 60 and 120 min. (b) same as panel (a) except with terminally tethered MQ–DNA duplex V. The numbering of GG doublets from (GG)₁ to (GG)₄ starts at the 3′ position of the complementary strand (see Chart ).

The yield of damage was estimated by autoradiography of ³²P-labeled oligonucleotides (Figure 2; Table 1). The formation of products was linear as a function of photolysis time except for cross-links, which appeared to reach a plateau after long exposures (Figure 2). The major products were cross-links with a quantum yield ranging from 4.6 × 10⁻⁴ to 13.7 × 10⁻⁴, depending on the sequence. The yield of cross-links was 2- to 3-fold greater for duplexes in which the complementary strand contained A and C compared to those with G and T opposite to the MQ moiety. The presence of A and C at this site also led to a greater percentage of cross-links that transformed into single strands upon treatment with hot piperidine. In comparison to cross-links, the quantum yield of piperidine-labile breaks opposite MQ was 0.18 × 10⁻⁴ to 0.66 × 10⁻⁴ (10- to 68-fold lower than cross-links). These breaks occurred at the base directly opposite MQ (A,G,T,C) as well as at neighboring bases [(T(3′) and T(5′)] with a bias toward damage at T(3′) for duplexes in which G, T and C were opposite MQ. Interestingly, the total damage was about 2-fold greater in duplexes containing pyrimidines (C = T) compared to those containing purines (G > A) opposite to the MQ moiety.

Figure 2.

Quantitation of damage by PAGE. DNA damage was based on the amount of radioactivity in ³²P-labeled oligonucleotides (Figure 1). (a) Interstrand cross-links before piperidine treatment (filled circles) and cross-links after piperidine treatment (filled squares). (b) Piperidine-induced cleavage at A opposite MQ (filled triangles) and T (3′-) opposite MQ (filled inverse triangles), and at the 5′-side (open triangles) and 3′-side (open inverse triangles) of GG-3 (5′). Control experiments (open circles) correspond to the photolysis of ³²P-labeled DNA duplex containing a naphthalene moiety in the presence of non-labeled MQ–DNA duplex.

Table 1. Yield of products from PAGE analysis^a.

	A (I)	G (II)	T (III)	C (IV)	5′-MQ (V)
Cross-links (−P)	12.30	4.56	6.50	13.68	5.50
Cross-links (+P)	4.98	2.76	4.98	7.74	3.04
T (3′)	0.05	0.26	0.37	0.37	—
Base (A,G,T,C)	0.16	0.09	0.22	0.22	—
T (5′)	0.05	0.03	0.06	0.06	—
Sub-total	0.27	0.37	0.66	0.65	—
GG3 (3′)	0.05	0.01	0.07	0.04	0.42b
GG3 (5′)	0.11	0.03	0.04	0.04	1.91b
GG4 (3′)	0.03	0.01	0.02	0.03	0.01c
GG4 (5′)	0.02	(0.004)	0.02	0.03	0.18c
Sub-total	0.21	0.05	0.15	0.13	2.52

^aThe amount of damage was estimated by PAGE analysis (Figures 1 and 2). These values were converted into quantum yields (×10⁴) by assuming the same quantum yield of formation for cross-links by PAGE analysis and HPLC-UV analysis. Structures of duplexes I–V are shown in Chart .

^bStrand breaks at GG₁ in duplex V.

^cStrand breaks at GG₂ in duplex V.

The total yield of breaks at GG₃ and GG₄ ranged from 0.13 × 10⁻⁴ to 0.21 × 10⁻⁴ for duplexes in which MQ was attached to an internal position (I–IV). This damage was greater for proximal GG doublets (GG₃) than for distal GG doublets (GG₄). In addition, there was usually more damage on the 5′-side compared to the 3′-side of GG doublets although this difference did not persist for distal GG doublets in duplexes I–IV. Similar yields and trends were observed for GG doublets on the 3′-side of MQ (GG₁ and GG₂; see Figure 1); however, it was difficult to measure the yields of damage at GG₁ and GG₂ because of interference from parent ³²P-labeled oligonucleotides. In contrast to duplexes I–IV, the yield of damage at GG doublets dramatically increased for duplexes in which MQ was attached to the 5′-terminus. This damage was 12-fold higher for the proximal GG doublet (GG₁) compared to the distal one (GG₂), and was more pronounced for the 5′-side compared to the 3′-side of the GG doublet. No significant increase in breaks was observed for the remote GG doublets in duplex V (GG₃ and GG₄).

HPLC-UV analysis of MQ–DNA duplexes

The formation of damage in duplex I with A opposite the MQ moiety was examined in greater detail by HPLC-UV. These analyses showed losses in UV absorption of both the complementary and MQ-containing strands, with a concomitant increase in new oligonucleotide products (Figures 3 and 4; Table 2). The initial rates of decomposition indicated that the MQ-conjugated strands disappeared at a faster rate than the complementary strands (Figure 4a). This can be attributed to greater damage of the MQ-containing strand compared to the complementary strand: e.g. since the two strands contribute equally to the formation of the cross-linked oligonucleotide products, the loss of the MQ-strand, excluding cross-link formation, is ∼3-fold greater than the loss of the complementary strand, excluding cross-link formation (Table 2). Interestingly, the total absorption of oligonucleotides at 260 nm decreased by 20% during photolysis as estimated by HPLC-UV. This decrease of absorption was not due to decomposition of the oligonucleotides but rather to a hypochromatic effect resulting from rapid re-annealing of cross-linked strands before their detection by HPLC-UV. When non-modified duplexes were injected directly into the HPLC column, they eluted as well separated single strands under our conditions. Direct analysis of MQ–DNA duplexes during photolysis indicated no significant change in absorption of the solution at 260 nm (<5%). To verify that the new products detected by HPLC-UV were duplexes with an interstrand cross-links, the products were purified by HPLC-UV, labeled with ³²P-radioisotope, and then subjected to denaturing PAGE analyses. The resulting products migrated in the same region as cross-linked products observed in the analysis of irradiated ³²P-labeled MQ–DNA duplex (data not shown). The ratio of absorption at 260/340 nm for the interstrand cross-linked products (ratio = 88) was approximately half the value for the MQ-containing strand (ratio = 150), consistent with the presence of both the complementary and MQ strands.

Figure 4.

Quantitation of DNA damage by HPLC-UV. (a) shows the loss and gain of oligonucleotides based on UV absorption at 260 nm: complementary strand (filled triangles); MQ-containing strand (filled inverse triangles) and interstrand cross-links (filled circles); error bars (±SD) represent the average of three independent experiments with single analysis of each time-point. (b) shows the loss of dCyd (open circles) (insert) and the relative loss (normalized to dCyd) of dGuo (open squares), dThd (open inverse triangles) and dAdo (open triangles). The initial slope for the loss of each nucleoside was determined by fitting a second-order polynomial to the data (average of three experiments with multiple analyses for each time-point). The loss of dCyd was statistically significant (r² = 0.56; P = 0.016) as well as for the normalized losses of dAdo (r² = 0.91; P = 0.002), dThd (r² = 0.75; P = 0.001) and dGuo (r² = 0.80; P = 0.009).

Table 2. Yield of DNA damage by HPLC-UV analysis.

DNA damage	Quantum yield (×104)
GG-strand	−6.5
MQ-strand	−7.1
Cross-links	+12.3
dCyd	−10.8
dGuo	−12.4
dThd	−14.5
dAdo	−15.9
Total nucleosides	−53.6

The decomposition of individual nucleosides in irradiated MQ–DNA duplex I was investigated by enzymatic digestion with P1 nuclease and alkaline phosphatase followed by HPLC-UV analysis of dCyd, dThd, dGuo, dAdo (Figure 4b; Table 2). Initially, we attempted to purify the MQ-containing and complementary strands by HPLC-UV and then determine their nucleoside composition; however, this led to too much variation in the data due in part to cross-contamination of the peaks. Thus, the entire MQ–DNA duplex was taken for enzymatic digestion and HPLC-UV analysis. The quantities of dAdo, dThd and dGuo were normalized to that of dCyd, the nucleoside with the lowest rate of decomposition, in order to correct for variations in DNA precipitation and digestion between time-points. According to these results, the yield of damage for individual nucleosides was greatest for dAdo (dAdo > dThd > dGuo > dCyd) and the sum of this damage was 4.4-fold greater than the yield of cross-linked products.

DISCUSSION

The photooxidation of DNA components by MQ and near-UV light has been extensively studied by flash photolysis, electron spin resonance and detailed product analyses (21,22,25–27). For the intermolecular reaction between excited MQ and DNA components, the mechanism of oxidation involves the generation of MQ radical anions and DNA base radical cations by initial electron transfer photochemistry. It is reasonable to propose a similar mechanism for MQ–DNA duplexes. The low yield of damage in MQ–DNA duplexes compared to reactions of free MQ and nucleosides can be attributed to the overwhelming back-transfer of initial radical ion pairs in MQ–DNA duplexes. The reaction of MQ radical anions with oxygen enhances the formation of DNA damage because it competes with back-transfer of the initial radical pair. In the photolysis of MQ–DNA duplexes, we propose that the formation of products, including interstrand cross-links, breaks opposite MQ and breaks at GG doublets arise from DNA base radical cations (Scheme 2).

Interstrand cross-links in MQ–DNA duplexes

A major interstrand cross-link in duplex I appears to involve the formation of a covalent bond between two adenine residues (Bergeron,F., Klarskov,K., Hunting,O.H. and Wagner,J.R., manuscript submitted). This is consistent with the formation of cross-links in MQ–DNA duplexes by deprotonation of initial adenine radical cations, giving rise to adenine N6 radicals, which subsequently undergo addition to another adenine on the opposite strand. A surprising feature of these cross-links was their tendency to rupture into two non-modified fragments upon heating under neutral or alkali (piperidine) conditions. This accounts for the rupture of cross-links within duplex I upon treatment with hot piperidine. Changing the base opposite MQ also leads to cross-links, suggesting the formation of novel cross-links in which A may be covalently linked to C, T and G. Interestingly, the yield and lability of these cross-links is similar for duplexes with either A or C opposite MQ (Table 1). Thus, it is reasonable to propose that adenine N6 radicals generated within the MQ-conjugated strand attack cytosine in duplexes with C opposite MQ. Previously, the formation of cross-links with cytosine was proposed to involve the reaction of cytosine N4 radicals with another cytosine base in dinucleotides (33). On the other hand, duplexes with either G or T opposite MQ gave cross-links with a lower yield and cross-links that were less sensitive to piperidine treatment compared to duplexes with either A or C opposite to MQ. Although the reactions of carbon-centered radicals with guanine and adenine have been observed in dinucleotides and oligonucleotides (34,35), the same reactions with aminyl radicals have not been reported. Recently, interstrand cross-links were reported from the photolysis of tethered ruthenium tetraazaphenanthrene complexes, but these cross-links arise from the addition of phenanthrene ligands to guanine radical cations on the opposite strand (36). Interestingly, the near-UV photolysis of poly-A in the presence of MQ increases Rayleigh scattering, suggesting the formation of cross-links; however, the yield of these products was at least 50-fold less efficient for single-stranded DNA than for poly-A, and there was no significant increase of this damage for double-stranded DNA (37).

Piperidine-induced cleavage opposite MQ

The induction of damage at bases opposite the photosensitizer represents a new type of damage in model systems of electron transfer in DNA. Although similar damage was reported by the photolysis of DNA-rhodium complexes at 313 nm (38), the strand breaks in the rhodium system were produced immediately whereas those in the MQ–DNA duplexes were (>90%) revealed only after treatment with hot piperidine. Another example of sugar damage induced by base radicals involves the abstraction of H-atoms from the sugar moiety by uracil-5-yl radicals, generated by near-UV photolysis of 5-bromouracil in DNA (39). In the case of MQ–DNA duplexes, a similar reaction may be proposed involving initial H-atom abstraction from the sugar moiety by adenine N6 or cytosine N4 radicals within MQ–DNA duplexes.

Piperidine-induced cleavage at GG doublets

The features of damage at GG doublets in MQ–DNA duplexes are similar in many respects to other model systems of electron transfer. There was generally a 3- to 5-fold difference in cleavage at the 5′-G compared to 3′-G sites of GG doublets (Table 1). In some duplexes, however, the yield of damage at GG doublets was low and did not show a clear bias toward damage at the 5′-side. The β value or fall-off of damage with distance can be estimated by comparing the intensity of total 3′ and 5′ damage between two GG doublets. For MQ–DNA duplexes in which the MQ moiety was attached at a central position (duplex I), the β value was 0.05 Å⁻¹ taking the average fall-off for GG₁/GG₂ and GG₄/GG₃. In contrast, the difference in damage between the first and second GG doublets in MQ–DNA duplexes containing terminally attached MQ (duplex V) indicated a steep decrease in damage with an estimated β value of 0.65 Å⁻¹. The low β value of 0.05 Å⁻¹ is very similar to that reported for anthraquinone duplexes (0.005–0.06 Å⁻¹), whereas the high value of 0.65 Å⁻¹ agrees with electron transfer in several other model systems, including those in which initial electron-hole injection takes place by photolysis of 4′-pivaloylated thymidine, stillbene and 2-aminopurine (40). The difference in β-values depending on the location of MQ–DNA duplexes (centrally or terminally located) is probably related to changes in the interaction of the MQ and base moieties. For example, the β value for the quenching of fluorescent analogs of adenine ranged from 0.1 to 1.0 depending on the ability of the analog to stack with other bases in duplex DNA (41). The yield of damage at GG doublets (GG₃ + GG₄) of MQ–DNA duplexes ranged from 0.05 × 10⁻⁴ to 0.21 × 10⁻⁴ for duplexes with centrally attached MQ to as much as 2.5 × 10⁻⁴ for duplexes with MQ attached at the terminus. The large difference between the two duplexes indicates that electron-hole injection at the terminus favors reactions leading to cleavage at GG doublets. In comparison, the yield of breaks at GG doublets for DNA duplexes containing metallo-intercalators (Rh and Ru) ranges from 0.02 × 10⁻⁴ to 0.9 × 10⁻⁴ and for anthraquinone-DNA duplexes from 0.2 × 10⁻⁴ to 500 × 10⁻⁴ (4,42).

Loss of individual nucleosides

The four nucleoside components (A, T, G and C) decreased in MQ–DNA duplex I as inferred by enzymatic hydrolysis with P1 nuclease and alkaline phosphatase, and analysis of the component nucleosides by HPLC-UV (Figure 1; Table 2). The loss of A (Φ = −15.9 × 10⁻⁴) can largely be explained by the formation of interstrand cross-links (Φ = 12.3 × 10⁻⁴). In particular, the formation of A–A cross-links in duplex I will result in the loss of two A residues as well as one non-modified T residue, which remains attached to the cross-link after enzymatic hydrolysis. The inability of P1 nuclease to cleave the phosphodiester bond on the 3′-side of damaged bases has been well-documented (43,44). The loss of dGuo (Φ = −12.4 × 10⁻⁴) can largely be accounted for by the formation of 8-oxo-7,8-dihydro-2′-deoxyguanosine (Φ = +9.2 × 10⁻⁴). In contrast, the analogous product for the oxidation of A (i.e. 8-oxo-7,8-dihydro-2′-deoxyadenosine) was formed in a much lower yield (Φ = +0.6 × 10⁻⁴). The formation of breaks at GG doublets probably arises from the presence of other dGuo oxidation products, such as oxazolone, which are piperidine labile (45). Finally, the loss of dCyd demonstrates that radical cations generated in MQ–DNA duplexes migrate to some extent to distant DNA bases, e.g. the closest C residue is at least 3 bp away from the site of initial electron-hole injection. Interestingly, there was no indication for the formation of 5-hydroxy-2′-deoxycytidine in our analysis. This product represents ∼10% of the total oxidation products resulting from the photooxidation of dCyd by MQ and near-UV light (27). The lack of formation of 5-hydroxy-2′-deoxycytidine suggests that C radical cations undergo alternative modes of decomposition in MQ–DNA duplexes. One possibility involves the deprotonation of C radical cations followed by coupling of the resulting nitrogen-centered radicals with DNA bases, which has been described in the photooxidation of d(CpC) (46).

Relevance to electron transfer in DNA

The formation of damage at all four bases of DNA is not consistent with current theories of electron transfer in DNA in which damage is exclusively observed at G. The formation of novel damage, including interstrand cross-links and piperidine breaks opposite the MQ moiety in duplex DNA indicates that there is a fundamental difference between the photochemistry of MQ and other photosensitizers used to study electron transfer in DNA. To explain the formation of unique products in MQ–DNA duplexes, we propose that excitation of MQ initially leads to the formation of the radical cations of all four DNA bases. In contrast, the excitation of other photosensitizers leads to damage at G residues probably because they only generate G radical cations, and once this intermediate is generated, the possibility of oxidizing other bases is highly unfavorable because of the relatively high oxidation potential of other DNA bases. Interestingly, the formation of damage at sites other than G has been documented in other systems of electron transfer in DNA, in particular, damage at A and T was reported in the photolysis of DNA using 193 nm light and high-intensity laser photolysis (47,48). It should also be noted that the photooxidation of DNA with free MQ in solution leads to damage at several DNA bases although there was a bias toward damage at G (47–49). Indeed, the ability of triplet-excited MQ to oxidize all four DNA bases has been demonstrated in studies with free MQ and nucleosides (18–23,25–27). In one study, the photooxidation of dCyd, dThd and dAdo was 100-fold more efficient than that of dGuo, indicating the low reactivity of dGuo radical cations, compared to that of the other base radical cations (49). The lack of dGuo decomposition may be attributed to the relatively long lifetime of dGuo radical cations leading to their efficient recombination with MQ radical anions. In contrast, the photooxidation of a mixture of all four nucleosides leads to greater oxidation of dGuo at the expense of the other nucleosides, indicating that the radical cations of dAdo, dCyd and dThd favorably accept an electron from dGuo ultimately giving dGuo radical cations. Therefore, the formation of damage in DNA duplexes covalently linked to MQ may be a balance between the fast kinetics of deprotonation and hydration of C, T and A radical cations, which leads to several novel products observed in this work, and the relatively slow kinetics of electron-hole migration to G, which leads to the 8-oxo-7,8-dihydroguanine product and piperidine labile breaks at GG.