Site-specific oxidative DNA damage at polyguanosines produced by copper plus hydrogen peroxide.

Oxidative DNA damage has been implicated in diverse biological processes including mutagenesis, carcinogenesis, aging, radiation effects, and chemotherapy. We examined the in vitro effect of low concentrations of Cu(II) or H2O2 alone and in combination on supercoiled plasmid DNA. As much as 10(-2) M Cu(II) or 10(-2) M H2O2 alone did not break the DNA. However, a mixture of 10(-6) M Cu(II) plus 10(-5) M H2O2 produced strand breaks and inactivated transforming ability. Strand breakage was proportional to incubation time, temperature, and Cu(II) and H2O2 concentrations. Abasic sites were not detected. Strand breakage was inhibited by metal chelators, catalase, and by high levels of free radical scavengers implying that Cu(II), Cu(I), H2O2, and .OH were involved in the reaction. The extent of DNA strand breakage was not affected by superoxide dismutase indicating that superoxide was not a major contributor to the DNA damage. DNA sequence analysis demonstrated that hot piperidine-sensitive DNA lesions were produced preferentially at sites of 2 or more adjacent guanosine residues. This sequence specificity was observed with Cu(II) plus H2O2 but not with Cu(I) alone. Polyguanosine sequence specificity for DNA damage induction appears to be unique among simple chemical systems. This reaction may be important in mechanisms of oxidative damage in vivo.


in vivo.
Oxidative DNA damage from active oxygen species has been hypothesized to play a critical role in several diverse biological processes including mutagenesis, aging, carcinogenesis, radiation damage, and cancer chemotherapy (1-3). Cellular metabolism has been shown to generate such oxygen species as hydrogen peroxide (Hz02), hydroxyl radical (.OH), singlet oxygen, superoxide, and hydroperoxyl radical (2). Trace metals such as copper and iron which are present in biological systems may interact with active oxygen species, ionizing radiation, or microwaves to damage DNA (3-8). The DNA damage is thought to be produced by a Fenton-type mechanism in which transition metal ions are cycled by first being reduced by superoxide and then oxidized by Hz02. The . OH produced would then damage the DNA.
In order to extend our understanding of the role of oxidative copper reactions in damaging DNA and to further characterize the reaction mechanism, we examined the effect of low levels of copper and H20Z alone and in combination on supercoiled * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. plasmid DNA. This in vitro systems is an extremely sensitive indicator of DNA damage. Our results indicate that Cu(I1) in the presence of hydrogen peroxide can damage DNA through a mechanism that involves .OH radicals but not superoxide.
This damage inactivates transforming ability and produces lesions that include single and double strand breaks and alterations at sites of adjacent guanosines revealed by hot piperidine treatment.

MATERIALS AND METHODS
Source of DNA and Reagents"puC8c2 (5480 bp)' and pZ189 (5504 bp) DNA was prepared and purified as previously described (7,9, IO). Cupric chloride, catalase, and superoxide dismutase were purchased from Sigma. Cuprous chloride, and 30% Hz02 in water were obtained from Mallinckrodt or from Fisher. In some experiments, Hz02 was deionized by stirring for 10 min with 20% (w/v) mixed bed ionic exchange resin (RG 501-X8, Bio-Rad). Micrococcus luteus UV-endonuclease was obtained from Applied Genetics, Inc., (Long Island, NY) and cloned T4-endonuclease was a gift from Dr. E. Henderson, Temple University, Philadelphia, PA. Restriction enzymes were purchased from Bethesda Research Laboratories. All chemicals used were of analytical quality. Bacterial transformation was performed using routine procedures with Escherichia coli strain MBM7070 (9).
Analysis of Strand Break Introduction-DNA (0.8-1.6 p g ) in 0.145 M NaCI, 0.01 M sodium phosphate, pH 7, was incubated for 30 min a t 24 "C with different Cu(I1) and H202 concentrations in a total volume of 12 pl. The reaction was stopped by addition of 1 pl of 0.2 M EDTA. Four pl of loading buffer (0.25% bromphenol blue, 0.25% xylene cyanole, 40% sucrose in 0.4 M Tris, 0.02 M EDTA, 0.2 M sodium acetate, pH 7.8 (10 X TEA buffer)) was added and samples analyzed by electrophoresis in 0.8% agarose in 1 X TEA buffer. The gel was stained with ethidium bromide, photographed, and scanned as described previously (7).
Enzymatic Sequencing-Dideoxy sequencing was performed on double-stranded DNA with avian myeloblastoma reverse transcriptase, as previously described (9). To assess DNA damage that would terminate enzymatic polymerization of DNA (stop assay) (11) 1 pg of pZ189 DNA was treated with copper plus H202, denatured with 0.4 M NaOH, and hybridized to about 2 ng of the pBR322 EcoRI clockwise site primer (d(GTATCACGAGGCCCT), New England BioLabs, Beverly, MA). After ethanol precipitation and lyophilization, DNA was dissolved in 11 pl of water and 4 p l of 0.3 M Tris-C1, pH 8.3, 37.5 mM MgC12, 2.5 mM dithiothreitol buffer. Then, 2 p1 of avian myeloblastoma reverse transcriptase (10 unitslpl, Promega, Madison, WI) and 40 pCi of [35S]dATP (1250 Ci/mmol, DuPont-New England Nuclear), were added. 4 p1 of the previous mix were incubated 20 min at 42°C with 1 p1 of dCTP, dTTP, dGTP all at 62 @M and dATP a t 3 pM (Pharmacia LKB Biotechnology Inc.) Chain termination was achieved by incubation for an additional 10 min with a solution containing 0.25 mM of each (nonradioactive) dNTP. After stopping the reaction with 6 pI of 95% deionized formamide, 10 mM EDTA, 0.2% bromphenol blue, 0.2% xylene cyanole, the sample was heated in boiling water for 3 min and loaded onto a denaturing gel containing 7% polyacrylamide with 8 M urea.
Chemical Sequencing-Conditions were established so that rapid preparation of a labeled pZ189 DNA fragment was achieved in a ' The abbreviation used is: bp, base pairs. 1729 single 1.5-ml tube without the need for phenol extraction. 5 pg of pZ189 DNA in 8 p1 of 50 M Tris, 10 mM MgC12, 0.1 M NaCl was linearized by treatment with 5 units of EcoRI for 30 min at 37 "C. The reaction was stopped by heating 15 min at 70 "C, and then NaCl concentration was diluted in half by adding an equal volume of 15 mM Tris buffer, pH 8. DNA was end-labeled with 32P by incubation with 6 units of the Klenow fragment of E. coli DNA polymerase I for 60 min at 12 "C in the presence of 70 pCi of [32P]dATP (3000 Ci/ mmol, Du Pont-New England Nuclear). The reaction was chased with 1 p1 of unlabeled dATP 100 mM for an additional 10 min at 12 "C and then heated for 15 min at 70 "C. A 3'-end-labeled 332-base pair fragment was generated by adding 5 units of AluI and incubating for 30 min at 37 "C. The fragment was separated by electrophoresis in a nondenaturing 5% acrylamide gel, its location in the gel determined by direct autoradiography, and the corresponding piece of gel cut out. The DNA fragment was eluted in 0.5 M ammonium acetate, 10 mM Mg acetate, 1 mM EDTA, 0.1% sodium dodecyl sulfate and gel fragments eliminated by use of filter units (cat no. 38-120, Rainin Inst. Co., Woburn, MA). This procedure yielded a pure 332-bp fragment beginning at the EcoI restriction site labeled with about 8 X lo5 cpm. Chemical sequencing was performed as detailed elsewhere (10). In order to detect oxidative base damage following treatment with copper plus HZ02, the 32P-labeled 332-bp fragment was heated for 30 min in 1 M piperidine at 90 "C in identical fashion to the samples for sequencing reactions (10).

DNA Strand Break Induced by Copper Plus
H202-Untreated DNA showed a major band corresponding to the supercoiled form (form I) and a minor band corresponding to nicked circular form (form 11) (Fig. 1, control lane). No linear form (form 111) was evident. Plasmid DNA remained intact after incubation with M Cu(I1) or M H202 alone ( Fig.  1, lanes 3 and 4 ) . However, when DNA was incubated with a Hz02 as a function of incubation time was measured at different temperatures (Fig. 2). The decrease in proportion of form I DNA followed single hit kinetics as a function of incubation time with M Cu(I1) plus M H2Oz. An average of one single strand break/DNA molecule (37% form I molecules remaining) was obtained after incubation periods of 11, 52, and 300 min at 37,23, and 6.5 "C, respectively. This represents an exponential decrease in incubation period/strand break with increasing incubation temperature.
The effect of the concentration of copper plus H202 on the production of DNA strand breaks was studied (Fig. 3). With 30 min of incubation at 24"C, a mixture of 0.01 M Cu(I1) plus 0.01 mM HZ02 resulted in about 94% form I molecules remaining. 0.1 mM Cu(I1) plus 0.1 mM H202 resulted in 31% remaining form I molecules or slightly more than one single strand break/5400-bp molecule. This concentration corresponds to about 1 molecule of Cu(I1) and 1 molecule of H202/ DNA base pair. 10 mM Cu(I1) or 10 mM Hz02 alone did not produce any detectable loss of form I molecules (data not shown).
Inactivation of Transforming Ability-Low levels of Cu(I1) and Hz02 were used to treat pZ189 at 37 "C for increasing time intervals, and the ability to transform competent bacteria to ampicillin resistance was assayed (Fig. 4). Treatment with M Cu(I1) plus M HzOZ for 120 min reduced transforming ability to about 37% of that with untreated plasmid. This same treatment resulted in about 60% of form I molecules remaining. This implies that the loss of transforming ability cannot be solely explained by strand breakage and suggests that additional lesions may be biologically important. Treatment with Cu(1I) or H202 alone did not alter transforming ability or the proportion of plasmids in form I.
Abasic Sites-In an attempt to characterize further the type of damage produced by Cu(I1) plus H202 treatment, abasic sites which would appear as alkali labile lesions were investigated. As a positive control, apurinic sites were produced by acid plus heat treatment. 30 min of treatment produced about one apurinic site/DNA molecule. Treated DNA was intact at neutral pH, but apurinic sites were converted to breaks when the DNA was treated with alkali ( Fig. 5A). Incubation with DNase I produced strand breaks without alkali labile sites, as shown by the same curve following either neutral or alkali post-treatment (Fig. 5B). Alkali treatment of DNA treated with Cu(I1) plus HzOZ with either (i) 66 mM NaOH (final pH 12.6) for 15 min, (ii) 100 mM NaOH (final pH 12.8) for 2 h, or (iii) 2 M glycine/NaOH, pH 12.6 for 2.5 h (Fig. 5C) all failed to detect more single strand breaks after alkali treatment than after incubation in neutral conditions. This behavior resembles the pattern of damage obtained with DNase I (Fig. 5B), which does not produce alkali labile sites. Piperidine treatment of Cu(I1) plus HzOz-treated DNA, under the same conditions used in sequencing experiments, resulted in extensive DNA breakage (data not shown).
Possible formation of abasic sites was further investigated by incubating Cu(I1) plus H20z-treated DNA with either UVendonuclease or T4-endonuclease, both of which incise DNA at abasic sites. As a positive control both enzymes produced breaks in DNA exposed to 1000 J/m2 of UV radiation. These enzymes did not reveal an increased number of breaks in Cu(I1) plus H20z treated DNA (data not shown).
Chemical  H202 at room temperature action but not with heat-denatured catalase (Fig. 1). Native or denatured superoxide dismutase gave little or no protection (Fig. 1). Superoxide dismutase activity was tested for the ability to inhibit decoloration of nitro blue tetrazolium by potassium superoxide (12). As a negative control, boiling superoxide dismutase for 5 min irreversibly abolished about 90% of its native activity. As positive controls, superoxide dismutase retained about 50% of its original activity with M Cu(I1) plus M HzO, under conditions of Fig. 1 and also by treatment with low3 M HzOz for 30 min at 23 "C at pH 7.
More than 95% of the activity remained after incubation with M HzOz in agreement with the values reported (13). The fact that protection against DNA strand breakage was observed with catalase but not with superoxide dismutase, implies that Hz02 but not superoxide was involved in the reaction. This also indicates that a superoxide-forming system was not required for production of the observed DNA scission.
Identical strand breakage was observed when HzOz from two different commercial sources was used and also when the sodium stannate (up to 10 ng/ml) used as H202 stabilizer was removed by ionic exchange chromatography through a mixed bed resin (data not shown). The same results were also obtained when the reaction was carried out in saline in the absence of phosphate or Tris buffer.
The protective effect of two metal chelating agents, the divalent cation chelator, EDTA, and the specific Cu(1) chelator, bathocuproine (14), was studied. Both afforded almost complete protection (Table I) suggesting that, in addition to Cu(II), Cu(1) was also involved in production of DNA damage. We thus studied the direct effect of incubating supercoiled plasmid DNA with varying amounts of Cu(1) in the absence of HZO2 (Fig. 3). One single strand break/molecule was produced by incubation in 1.4 mM Cu(1) without Hz02 for 30 min at 24°C. At higher concentrations (5-10 mM Cu(I)), the plasmid integrity was highly disrupted (not shown). The extent of loss of form I molecules with Cu(1) was about 18 times less than with an equimolar mix of Cu(I1) and HzOz (Fig. 3).
The participation of .OH radicals in DNA damage was studied by examining the protective effect of radical scavengers (Table I). Sodium azide (0.1 M) was an efficient protector. At higher concentrations several other hydroxyl radical scavengers (tert-butyl alcohol, dimethyl sulfoxide, mannitol, ethanol, bovine serum albumin, and polyethylene glycol) were only partially protective. Complete protection with tert-butyl alcohol was only achieved at a concentration of 1 M. Potassium chloride, which is a poor free radical scavenger, produced only minimal protection even at 2 M concentration. Sequence Specificity of DNA Damage-Chemical sequencing with hot piperidine scission at damaged sites revealed specificity of the DNA damage. A 3'-end-labeled fragment of plasmid pZ189 was treated with Cu(I1) and H202, incubated with 1 M piperidine at 90 "C for 30 min, and analyzed by electrophoresis in a sequencing gel. Sequencing reactions run in parallel on the same gel allowed determination of the position of the scissions and the nucleotides involved (Fig. 6,  lanes 1-4). Polyguanosine sequences were preferred sites for the damage by Cu(I1) plus Hz02 (Fig. 6, lanes 7 and 8, arrows). Of 21 guanine residues sequenced in the labeled strand, 16 residues were sufficiently altered so as to result in a scission in the strand after hot piperidine treatment. These 16 residues  51-52,58-59,74-75,88-89, and 113-114) was not cut. The preferred sites of damage produced by Cu(I1) and H202 are indicated by underlining in the sequence written on the bottom of Fig. 6. The frequency of scission was greater at the 5' end of a run of Gs.
DNA treated with 10 mM Cu(I1) or 10 mM H202 alone, and subjected to hot piperidine treatment, failed to show any damage in sequencing gels (Fig. 6, lanes 10 and 11 ) in agreement with the studies of loss of supercoiling (Fig. 1). Cu(1)treated DNA subjected to hot piperidine produced bands lacking the sequence specific pattern observed with Cu(1I) plus Hz02 (Fig. 6, lanes 9 and 13). Cu(I1) plus Hz02 treatment without hot piperidine incubation yielded a smear without any discrete bands (not shown).
The damage produced by the mixture of Cu(I1) and H202 was also assessed by a polymerase chain termination assay. After treatment of plasmid pZ189 with Cu(I1) and H202, the  end-labeled fragment of pZ189 treated with Cu(I1) plus Hz02 both at 10 or 1 mM for 30 min at 23 "C followed by scission at damage by treatment with piperidine at 90 "C for 30 min showing specific cutting at sites of 2 or more adjacent G residues (arrows). The intensity of a band reflects the frequency of scissions at the corresponding base and hence the frequency of DNA damage at that site in the population of treated molecules. Cu(I), incubation with 1 or 10 mM Cu(1) for 30 min at room temperature followed by hot piperidine reveals less specific damage. Cu(II), 10 mM Cu(I1) alone followed by hot piperidine. H202,lO mM Hz02 alone followed by hot piperidine. B, incubation in buffer. P, incubation of untreated fragment with hot piperidine alone. G+A, guanine plus adenine Maxam-Gilbert sequencing reaction. T+C, thymine plus cytosine Maxam-Gilbert sequencing reaction. C, cytosine Maxam-Gilbert sequencing reaction. A X ' , adenine and cytosine Maxam-Gilbert sequencing reactions. Lower, sequence of fragment. Underlined runs of G indicate locations of adjacent guanines appearing as prominent bands in the lanes treated with Cu(I1) plus H202. This specificity is not seen in the lanes treated with Cu(I), Cu(II), or HzOz alone. DNA was incubated with avian myeloblastoma reverse transcriptase in the absence of ddNTP. Some types of DNA damage, such as cyclobutane dimer photoproducts, cause termination of the polymerase reaction (15). Results obtained in seven experiments including reaction conditions like those in Fig. 6 failed to detect polymerase termination at the sites of damage detected polymerase termination at the sites of damage detected in Fig. 6 (data not shown). In control reactions UV-treated plasmids showed terminations at pyrimidine photoproducts (data not shown).

DISCUSSION
DNA Damage-Although neither Cu(I1) nor Hz02 alone at concentration up to M was able to damage DNA, together they formed an active DNA damaging mixture. The sensitizing effect of copper was previously reported in phage inacti-vation after ionizing radiation (4) and in the decrease of sedimentation speed of T7 bacteriophage DNA after exposure to reduced oxygen species (16). High concentrations of Hz02 have been known for many years to produce single and double strand breaks (references in 17) due in part to sugar damage (3). However, using a sensitive supercoiled plasmid assay, we detected single strand breakage and loss of transforming ability induced by as little as M Cu(I1) plus M Hz02 with 30 min of incubation at 37 "C C (Fig. 4).
We found polyguanosine sequences to be preferentially damaged as revealed by scission after piperidine treatment. Feldberg et al. (17) reported that Cu(I1) plus H202 was able to damage poly(dG-C) but not poly(dA-T). Guanine has previously been shown to be the target for aflatoxin (la), nitrogen mustards (19), antitumor agents (bleomycin, neocarcinostatin, cis-platinum) (20), and antibiotic (tetracycline) action (21). However, our finding of the preference for polyguanosine sequence damage appears to be a unique feature of the Cu(I1) plus H202 system in comparison to that reported for other systems involving metals (14,20,22).
One type of guanosine damage might be loss of guanine as has been found in oxidative radiation damage to DNA or with 1,lO-phenanthroline-copper complex (3). This would result in an abasic site. We were unable to detect abasic sites (alkali labile sites or UVor T4-endonuclease sensitivity) after the damage produced in DNA by Cu(1I) plus HzOz (Fig. 5 ) . The fact that the polyguanosine lesion is converted to a strand break only after hot piperidine treatment (90 "C for 30 min), favors the idea of nucleotide lesion(s) stable under the usual alkali treatment, but made evident only under more drastic conditions. The guanosine damage thus might be a nucleotide modification such as the formamido pyrimidine derivative (23). Another feature of the guanosine damage is that it did not block in vitro DNA polymerization. This polymerase bypass has also been observed with the oxidative lesion, 8hydroxydeoxyguanosine, and was found to be an error-prone process (24).
Chemical Species Involved in Inducing Damage-Both Cu(I1) and H202 are required for the DNA damage since EDTA and catalase both have a protective effect (Table I).
The protection by bathocuproine indicates that Cu(1) is also involved. This is further confirmed by the observation that Cu(1) alone produces single and double strand breaks in DNA in absence of H202 (Figs. 3 and 6). However, Cu(1) is about 18 times less effective in producing strand breaks than equimolar Cu(1I) plus H202 (Fig. 3) and also lacks the base sequence specificity for nucleotide damage (Fig. 6). The large differences existing in the relative stabilities of the complexes and chelates of Cu(I1) and Cu(I), together with the fact that Cu(I1) usually forms planar compounds while Cu(1) forms tetrahedral compounds (25) may account for the different base specificity observed in our experiments.
The ability of .OH scavengers to protect DNA from damage, indicates that the . OH participates in the mechanism of strand break formation produced by Cu(I1) and H202. However, only partial protection of DNA is conferred by .OH scavengers. This partial protection given by free radical scavengers suggests short action radii for the .OH, which hinder the protective efficiency of scavenger molecules. Hydroxyl radicals that possess this property by being complexed either to its precursive metal or to its substrate have been proposed in other systems (8). Superoxide radicals do not seem to play a major role in the DNA strand scission since active superoxide dimutase lacks any protective effect. This also suggests that direct reduction of Cu(I1) by H202 (17) is not a principal reaction since this reaction would generate superoxide. The lack of superoxide dismutase protection in our experiments agrees with similar results on the DNA damage produced either by ascorbate-Cu(I1) (4) or camphothecin-Cu(I1) (14). Singlet oxygen might also play a role in this mechanism since it has been reported to react specifically with guanine and is scavaged by azide (26). Model Mechanism-We hypothesize that DNA damage in the presence of Cu(I1) and Hz02 occurs in a multistage mechanism that involves one step favoring base sequence specificity, the binding of Cu(I1) (but not Cu(1)) to an electronegative region involving at least two guanosines. Guanosine residues are the most electronegative bases, and adjacent guanosine residues would be the most electronegative regions of the DNA molecule (19). This line of reasoning is further supported by previous evidence showing binding of Cu(I1) to guanine (27, 28). As a second step Cu(I1) would react with DNA, perhaps through proton transfer involving guanine (29), in a similar fashion as proposed for copper and other substances (25,27). This reduction of Cu(I1) could yield oxidation products of guanine which are piperidine sensitive but are not alkali sensitive and do not block DNA polymerization in vitro.
In the last step, H202 reacts with the Cu(1) formed, either still bound or in the proximity of DNA, generating . OH and regenerating Cu(I1). In turn, .OH produces (additional) piperidine-sensitive base damage and/or strand breaks at short range from the original Cu(I1)-binding site.
Biological Relevance-Our experiments show that DNA breaks and other lesions can be produced without the need of a superoxide-generating system or complex drugs, but with physiological concentrations (30) of simple substances such as Cu(I1) and H202 that are present in a wide range of biological systems. DNA damage involving copper would be relatively favored in diseases where copper concentration is elevated such as Menkes' syndrome (31), Wilson's disease, or certain neoplasms (32). In some neoplasms oxygen concentration is low as compared with normal cells (6). In normal cells Cu(1) would be oxidized mainly by O2 with little .OH formation and DNA damage. In tumor cells, the oxidation of Cu(1) by H202 would be favored, resulting in the DNA damage studied in this paper.
Better biochemical understanding of the mechanism involving Cu(I1) and H20z might result in the development of a simpler cancer therapy with fewer detrimental side reactions.