The Carbonate Radical Is a Site-selective Oxidizing Agent of Guanine in Double-stranded Oligonucleotides*

The carbonate radical anion (CO⨪3) is believed to be an important intermediate oxidant derived from the oxidation of bicarbonate anions and nitrosoperoxocarboxylate anions (formed in the reaction of CO2 with ONOO−) in cellular environments. Employing nanosecond laser flash photolysis methods, we show that the CO⨪3 anion can selectively oxidize guanines in the self-complementary oligonucleotide duplex d(AACGCGAATTCGCGTT) dissolved in air-equilibrated aqueous buffer solution (pH 7.5). In these time-resolved transient absorbance experiments, the CO⨪3 radicals are generated by one-electron oxidation of the bicarbonate anions (HCO 3 − ) with sulfate radical anions (SO⨪4) that, in turn, are derived from the photodissociation of persulfate anions (S2O 8 2 − ) initiated by 308-nm XeCl excimer laser pulse excitation. The kinetics of the CO⨪3 anion and neutral guanine radicals, G(−H)⋅, arising from the rapid deprotonation of the guanine radical cation, are monitored via their transient absorption spectra (characteristic maxima at 600 and 315 nm, respectively) on time scales of microseconds to seconds. The bimolecular rate constant of oxidation of guanine in this oligonucleotide duplex by CO⨪3 is (1.9 ± 0.2) × 107 m −1 s−1. The decay of the CO⨪3 anions and the formation of G(−H)⋅radicals are correlated with one another on the millisecond time scale, whereas the neutral guanine radicals decay on time scales of seconds. Alkali-labile guanine lesions are produced and are revealed by treatment of the irradiated oligonucleotides in hot piperidine solution. The DNA fragments thus formed are identified by a standard polyacrylamide gel electrophoresis assay, showing that strand cleavage occurs at the guanine sites only. The biological implications of these oxidative processes are discussed.

The carbonate radical anion is a strong one-electron oxidant that oxidizes appropriate electron donors via electron transfer mechanisms (19). Detailed pulse radiolysis studies have shown that carbonate radicals can rapidly abstract electrons from aromatic amino acids (tyrosine and tryptophan). However, reactions of CO 3 . with sulfur-containing methionine and cysteine are less efficient (20 -22). Hydrogen atom abstraction by carbonate radicals is generally very slow (19), and their reactivities with other amino acids are negligible (20 -22). It is well established that carbonate radicals can play an important role in the modification of selective amino acids in proteins in cellular environments under conditions of oxidative stress, aging, and inflammatory processes (1,11,12). The role of HCO 3 Ϫ /CO 2 in potentiating oxidative DNA damage has received relatively little attention. It has been shown that the presence of HCO 3 Ϫ /CO 2 inhibits direct strand cleavage of DNA induced by ONOO Ϫ but enhances the formation of 8-nitroguanine, alkali-labile and formamidopyrimidine glycosylase-labile DNA lesions (23)(24)(25). Peroxynitrite causes direct DNA strand cleavage by oxidizing deoxyribose. However, in the presence of HCO 3 Ϫ /CO 2 there is a shift in product distribution from direct strand cleavage to the formation of oxidative modifications of guanines (26), suggesting that the carbonate radical anion could play an important role in this phenomenon (24). Although guanine is indeed the most easily oxidized base in DNA, the reactions of the carbonate radical anions with the different aromatic DNA residues have not yet been characterized.
In this work, we explore the electron transfer reactions from guanine electron donor residues embedded in the self-complementary hexadecanucleotide duplex d(AACGCGAATTC-GCGTT) to CO 3 . electron acceptor. Employing transient absorbance laser flash photolysis techniques, we monitored simultaneously the kinetics of disappearance of the CO 3 . anion radical and the appearance of guanine radicals by transient absorption spectroscopy techniques. The decay of the CO 3 . radical anion was followed by monitoring its absorbance ( max ϭ ϳ600 nm), whereas the concomitant formation of guanine neutral radicals, G(ϪH) ⅐ , was identified by their characteristic narrow absorption band at 315 nm. These observations constitute the first direct indication that carbonate radicals are site-selective one-electron oxidants of guanine residues in double-stranded DNA and do not react significantly with any of the other nucleic acid bases A, C, or T. The oxidation of the different guanines in the self-complementary DNA duplex results in alkali-labile guanine lesions; these are revealed by the site-selective DNA strand cleavage induced by a standard hot piperidine treatment and polyacrylamide gel electrophoresis assay. (8-oxo-dG), and all inorganic salts (Ͼ99% purity) were obtained from Sigma-Aldrich and were used as received. The d(AACGCGAATTCGCGTT) oligonucleotide was synthesized by standard automated phosphoramidite chemistry techniques. The tritylated oligonucleotide was removed from the solid support and deprotected with concentrated ammonium hydroxide. The crude oligonucleotide was purified by reversed phase high performance liquid chromatography and detritylated in 80% acetic acid according to standard protocols.

EXPERIMENTAL PROCEDURES
The oligonucleotide was dissolved in 20 mM phosphate buffer (pH 7.5) containing 100 mM Na 2 SO 4 . Annealing of the two strands was accomplished by heating the samples to 90°C for 10 min and then allowing the samples to cool slowly back to room temperature overnight.
Laser Flash Photolysis-The transient absorption spectra were recorded using a pulsed laser excitation source and a kinetic spectrometer system (response time, ϳ7 ns) described earlier (27)(28)(29)(30). Briefly, this system consists of a pulsed Lambda Physik EMG 160 MSC XeCl excimer laser (308 nm; full width at half-maximum pulse width, ϳ12 ns; energy, ϳ70 mJ pulse Ϫ1 cm 2Ϫ ; 0.1 Hz), and a 75 W pulsed xenon lamp to probe the transient absorption kinetics. The probe flash was passed through a McPherson monochromator and was monitored by a Hamamatsu R928 phomultiplier; its output was recorded and digitized using a Tektronix TDS 620 oscilloscope. All experiments, including data collection and analysis, were controlled by a computer.
The CO 3 . radical anions were generated by oxidation of HCO 3 Ϫ with SO 4 . radical anions (31,32). The SO 4 . ions were generated by the photodissociation of persulfate anions (31,33). Although persulfate ions exhibit negligible reactivity with nucleosides and oligonucleotides, the SO 4 . radical anions are known to react with the DNA bases. The triggered dissociation of persulfate ions into SO 4 . radical anions in timeresolved pulse radiolysis or flash photolysis experiments has been previously employed by Steenken and co-workers (34 -36), O'Neill and Davies (37), and Bachler and Hildenbrand (38), to study the kinetics of reactions of the SO 4 . radical anions with nucleic acids.
Air-equilibrated buffer solutions containing 20 -100 M oligonucleotide duplexes, 300 mM NaHCO 3 , and 25 mM Na 2 S 2 O 8 were placed in a 0.5-ml quartz flow cell (ϳ80% transmittance at 200 nm) from NSG Precision Cells, Inc. The optical pathlength of this cell was 1 cm, and the cell was aligned parallel to the direction of the xenon lamp probe light beam. After one laser shot, the irradiated solution was replaced by a fresh sample solution employing a computer-controlled flow system. All laser flash photolysis experiments were performed at 20°C.
Pseudo-first order rate constants (k n Ј) of the radical (R ⅐ ) ϩ substrate (S) reactions and the second order rate constants of the radical recombination were determined by least squares fits of the appropriate, indicated kinetic equations to the transient absorption profiles obtained in five different experiments with five different samples. The second order rate constants of the R ⅐ ϩ S reactions were calculated by a least square best fit to a linear dependence of the k n Ј versus [S] plots.
Steady-state Spectroscopy-Routine UV absorption spectra and UV melting profiles were measured using an HP 84453 diode array spectrophotometer with an HP 89090A Peltier temperature control unit (Hewlett Packard GMBH, Waldbronn, Germany).
Oxidative DNA Strand Cleavage Assay-The oligonucleotides were labeled at their 5Ј termini by using T4 polynucleotide kinase (Amersham Pharmacia Biotech) and [␥-32 P]ATP (PerkinElmer Life Sciences) as described previously (27). The 10-l samples containing 32 P 5Ј-endlabeled oligonucleotide strands in 2 ϫ 2-mm square pyrex capillary tubes (Vitrocom, Inc.) were irradiated with 308-nm laser light. After the irradiation, the reaction mixture was quenched by the addition of ␤-mercaptoethanol and was heated for 30 min at 90°C in 100 l of a 1 M piperidine solution. The samples were vacuum dried and assayed by a denaturing polyacrylamide gel electrophoresis (20% acrylamide, 19:1 bis ratio, 7 M urea), as described previously (27). The relative intensities of cleavage were analyzed using a Bio-Rad 250 imaging system.

RESULTS
Duplex Design-The duplex formed by the self-complementary 16-mer d(AACG 1 CG 2 AATTCG 3 CG 4 TT) sequence contains the 12-mer d(CG 1 CG 2 AATTCG 3 CG 4 ) core. The latter is the first oligonucleotide duplex containing at least one turn of helix for which structural information was obtained by single-crystal x-ray diffraction methods (39,40) and high resolution NMR spectroscopy (41,42). We added one AA and one TT doublet to the two ends of this 12-mer to diminish the known fraying (42) of the G:C base pairs at or near the ends of the oligonucleotide. The 16-mer duplex exhibits well defined cooperative melting curves. The average melting temperature (T m ϭ 67 Ϯ 1°C) and the hyperchromicity of ϳ17% were calculated from plots of the absorbance (260 nm) versus temperature obtained from heating-cooling cycles (43).
Generation of CO 3 . Radicals by the Oxidation of HCO 3 Ϫ by SO 4 .
Radicals-The kinetic parameters associated with the reactions of HCO 3 Ϫ and S 2 O 8 2Ϫ in aqueous solutions induced by laser pulse excitation are well documented (19). The excitation of S 2 O 8 2Ϫ in aqueous solutions with 308-nm XeCl excimer laser pulses generates the SO 4 . radical anions (   3 . radicals in aqueous buffer solutions (pH 7.5) Reaction schemes and rate constants are shown.
This work k 2 ϭ 1.35 ϫ 10 9 Ref. 33   , is a very strong acid (pK a Ͻ 0) and thus deprotonates rapidly in aqueous solutions (46). The recombination of the CO 3 . radicals is a slow process (  (47). The rate constants of reactions 2-4 measured during the course of this work are also summarized in Table I; these values differ only slightly from published values because of differences in ionic strength (45). Ϫ , recorded at various delay times (⌬t) after 308-nm laser pulse excitation, are shown in Fig. 1. The transient absorption spectrum recorded at ⌬t ϭ 60 s exhibiting a maximum at 600 nm corresponds to the well known spectrum of the CO 3 . radical (31,45,46); at this HCO 3 Ϫ concentration (300 mM), the decay of the SO 4 . anion radicals occurs rapidly (within ϳ0.7 s) and is not shown in Fig. 1.

Transient Absorption Studies of the Oxidation of Guanine by
The decay of the CO 3 . absorption band is accompanied by the rise of another, narrow transient absorption band centered at 315 nm. This spectrum, recorded at ⌬t ϭ 3.9 ms, is characteristic of both the guanine radical cation, G ⅐ ϩ , and the neutral guanine radical, G(ϪH) ⅐ (34). Thus, the carbonate radical is a site-selective one-electron oxidant of guanine residues in double-stranded DNA (Table I, Reaction 5). As shown below (DNA strand cleavage assay), the oxidation of the other three DNA bases by CO 3 . anion radicals is significantly less efficient than that of guanine. This is consistent with the relative rate constants of one-electron oxidation of 2Ј-deoxynucleotide monophosphates by CO 3 . radicals. In the case of dGMP, the rate constant is some 20 -40 times greater than the rate constant determined for dAMP, dCMP, and dTMP (Table I, Reactions 6 -9). However, the rate constant of oxidation of 8-oxo-dG is greater by a factor of ϳ10 (Table I, Reaction 10) than that of dGMP (36). Deprotonation of Guanine Radical Cations-The pK a of free guanine radical cations, G ⅐ ϩ , generated by the one-electron oxidation of dG or dGMP, is 3.9 (34). Therefore, in neutral aqueous solutions, deprotonation of G ⅐ ϩ occurs rapidly (rate constant k ϭ ϳ2 ϫ10 6 s Ϫ1 ). Candeias and Steenken (34) proposed that in double-stranded DNA, with the G ⅐ ϩ radical hydrogen-bonded to its partner Watson-Crick cytosine residue (pK a ϭ 4.45), the deprotonation rate constant of G ⅐ ϩ should be even greater than in the case of free G ⅐ ϩ . The neutral G(ϪH) ⅐ radical has indeed been detected by electron spin resonance techniques upon oxidation of double-stranded DNA in aqueous solutions at room temperature (48,49). Hence, the guanine radicals observed on millisecond time scales (Fig. 1) are neutral radicals, G(ϪH) ⅐ , rather than G ⅐ ϩ radical cations.
Kinetics of Guanine Oxidation by CO 3 . Radicals in DNA Duplexes-In the absence of DNA, the bimolecular recombination of two CO 3 . radicals occurs, and the rate constant (k 4 ) of this bimolecular reaction has been reported ( The value of k 5 Ј was determined by fitting Equation 2 to the CO 3 . decay profiles measured at 600 nm and at different DNA concentrations. The value of k 4 was determined in the absence of the DNA duplex (Table I). As expected, k 5 Ј increases linearly with the concentration of the DNA duplex (Fig. 2), yielding a value of k 5 ϭ 1.9 ϫ 10 7 M Ϫ1 s Ϫ1 . The rate of oxidation of dGMP by CO 3 . radicals was determined by similar experiments (data not shown), and the rate constant k 6 ϭ 6.6 ϫ 10 7 M Ϫ1 s Ϫ1 (Table  I). The rate constant of oxidation of free dGMP (k 6 ) is ϳ3.5 times larger than k 5 , the value of the rate constant for the oxidation of G within the DNA duplex. The apparent decrease in the reactivity of guanines in the duplexes is probably associated with a decreased accessibility of guanines to the CO 3 .

radicals (see below).
Decay Kinetics of G(ϪH) ⅐ Radicals-The decay of the G(ϪH) ⅐ radicals in the double-stranded oligonucleotide is shown on two different time scales in Fig. 3. On the Ͻ100-ms time scale (Fig.  3, inset), the change in G(ϪH) ⅐ radical concentrations is too small to be measurable, but a slow decay is evident on a time scale of 0.8 s (Fig. 3).
Chemical Damage in Double-stranded DNA Induced by CO 3 .
Radicals-The rate constant of reaction of SO 4 . radicals with DNA is greater than that of CO 3 . radicals (Table I, Reactions 11 and 5, respectively). The value of k 11 ϭ 5 ϫ 10 9 M Ϫ1 s Ϫ1 (Table  I)  Ϫ with 308-nm excimer laser pulses, are depicted in the gel autoradiograph in Fig. 4A. Oligonucleotide strand cleavage is negligible in the unirradiated control samples with or without hot piperidine treatment (lanes 1 and 2, respectively). However, upon irradiation, cleavage is observed mostly at the guanine sites, and the extent of strand cleavage is more pronounced after hot piperidine treatment (lanes 4 and 6) than before (lanes 3 and 5). In lanes 3 and 4 the total energy dosage received by the sample was 0.4 J/cm 2 , whereas in lanes 4 and 6 the dosage was approximately ten times higher. The extent of damage increases with the laser energy dosage. It is evident that the sites of cleavage correlate well with the cleavage pattern obtained by the Maxam-Gilbert sequencing reaction (lane G), clearly indicating that strand cleavage occurs predominantly at the four different guanines in the self-complementary oligonucleotide duplex. The relative cleavage efficiencies at the different guanines are shown in Fig.  4B. The probability of strand cleavage at these sites differs, being higher at sites G 1 and G 2 , which are closer to the ends of the oligonucleotide, than at sites G 2 and G 3 , which are positioned closer to the center of the duplex. In these experiments the overall percentage of DNA strands cleaved was less than 5%, indicating that the DNA duplexes suffered, on average, less than one DNA strand cleavage event/molecule (0.4 mJ/cm 2 dosage). At the higher dosage, the fraction of cleaved DNA strands increased to 40 -45%. Nevertheless, strand cleavage at sites other than guanine remains small (Fig. 4C). These results clearly indicate that the CO 3 . radical anion is a site-selective oxidizing agent of guanines in double-stranded DNA. These hot piperidine-induced strand cleavage results are in good agreement with the spectroscopic laser photolysis data. The latter results suggest that carbonate radical anions react by oneelectron transfer mechanisms more readily with guanine than with the other three DNA bases.

Selective Oxidation of Guanines in DNA by CO 3 . Radicals-
The selective reactivity of the CO 3 . radicals with guanine rather than with any of the other three DNA bases, is a consequence of the thermodynamic and kinetic characteristics of these electron donor/acceptor reactions. The redox potential of the CO 3 . . radical anions. The oxidized guanine derivative 8-oxo-dG, a product of two-electron oxidation of guanine (50), is further oxidized by CO 3 . radicals with a greater rate constant than in the case of guanine (Table I). This enhanced reactivity of 8-oxo-dG with CO 3 . radicals is correlated with its lower redox potential, E 7 ϭ 0.74 V versus NHE (36), which is smaller than the redox potentials of the other nucleosides discussed.

Rate Constants of Interaction of the CO 3 . Radicals with
Guanines-It is interesting to note that the rate constant of oxidation of dGMP by CO 3 . radical anions (k 6 ϭ 6.6 ϫ 10 7 M Ϫ1 s Ϫ1 ; Table I . radicals with guanines is lower than expected from differences in redox potentials alone. Schindler et al. (56) have shown that the CO 3 . /CO 3 2Ϫ system is characterized by a small self-exchange rate constant (k ϭ 0.4 M Ϫ1 s Ϫ1 ) and a high internal reorganization energy that results in a low reactivity of CO 3 . in bimolecular outer sphere electron transfer reactions. Furthermore, in doublestranded DNA the rate constant of oxidation of guanines is even lower than in the case of free dGMP. This value, k 6 ϭ 1.9 ϫ 10 7 M Ϫ1 s Ϫ1 (Table I), is averaged over the four different guanine sites G 1 , G 2 , G 3 , and G 4 . The cleavage efficiencies (Fig. 4B) are lower at the two central guanines G 2 and G 3 than at the outer guanines G 1 and G 4 . These variations in reactivities correlate roughly with the imino proton exchange rates and thus the base pair opening rates, in the different G:C base pairs in the 12-mer d(CG 1 CG 2 AATTCG 3 CG 4 ) studied by Patel and co-workers (42). Thus, the lower reactivities of the CO 3 . radical anion with guanines in double-stranded DNA are most likely due to site-depend-ent base pair opening rates (42) and probably to steric hindrance factors as well.
The Long Lifetimes of Guanine Radicals, G(ϪH) ⅐ , in Doublestranded DNA-It is interesting to note that the lifetimes of the G(ϪH) ⅐ radicals are significantly longer in the double-stranded oligonucleotide (Fig. 3) than the lifetimes of either dGMP(ϪH) ⅐ or dG(ϪH) ⅐ radicals in solution (27,29,57). Furthermore, the lifetime of the G(ϪH) ⅐ radicals in the d(AACGCGAAT-TCGCGTT) duplex (Fig. 3) is orders of magnitudes longer than in oligonucleotide duplexes in which these radicals are generated by one-electron transfer reactions from guanine to 2-aminopurine (2AP) radicals (28). In the latter case, as well as in the present experiments, the lifetime measurements were conducted in air-or oxygen-equilibrated solutions. Thus, differences in the decay rates caused by reactions of G(ϪH) ⅐ with O 2 can be ruled out. In the case of the 2AP modified duplexes, 2AP radicals were generated by two-photon photoionization of 2AP residues, thus generating hydrated electrons as well. The hydrated electrons are efficiently scavenged by molecular oxygen resulting in the formation of O 2 . radicals (58 (61)(62)(63). Hydroxyl radicals produced by the one-electron reduction of H 2 O 2 in Fenton-type reactions are well known species that can induce oxidative modifications and damage to the DNA (50,64). However, ⅐ OH radicals are extremely reactive and thus have a very limited range of diffusion (65). Therefore, ⅐ OH radicals induce damage to biomolecules only within the immediate locale where they are generated. Wolcott et al. (66) have proposed that ⅐ OH radicals can oxidize HCO 3 Ϫ anions and thus generate the CO 3 . radicals that are less reactive than the ⅐ OH radicals. Recently, Kalyanaraman and co-workers (3,4) have shown that the enhancement of peroxidase activity of copper-zinc superoxide dismutase in the presence of HCO 3 Ϫ (67) is associated with the generation of CO 3 . radicals by one-electron oxidation of HCO 3 Ϫ occurring at the enzyme active site. Still another source of carbonate anion radicals might be of importance. Inflammatory and infectious processes are associated with the generation of not only reactive oxygen species but also of nitric oxide ( ⅐ NO) catalyzed by ⅐ NO synthase type 2 (68,69). Beckman and co-workers (70)  Ϫ occurring on submicrosecond time scales (73) produces CO 3 . and ⅐ NO 2 radicals with a cage escape yield of ϳ0.3 (16, 18). Hence, because of the relatively high concentrations of HCO 3 Ϫ /CO 2 in intracellular and interstitial fluids of up to 30 mM (74), response to aging, chronic inflammation, and diverse infectious disorders might also include the enhanced generation of CO 3 . radicals that can damage not only proteins (1,12) but also DNA as shown in this report.