Structural Origins of Bulky Oxidative DNA Adducts - American

oxidation of DNA and test oligonucleotides in a sequence-dependent manner. ... Recently, some of the major oxidative DNA adducts characterized here we...
0 downloads 0 Views 259KB Size
Chem. Res. Toxicol. 1996, 9, 247-254

247

Structural Origins of Bulky Oxidative DNA Adducts (Type II I-Compounds) as Deduced by Oxidation of Oligonucleotides of Known Sequence Kurt Randerath,*,† Erika Randerath,† Charles V. Smith,‡ and Jian Chang† Division of Toxicology, Department of Pharmacology and Department of Pediatrics, Baylor College of Medicine, Houston, Texas 77030 Received May 15, 1995X

Bulky DNA adducts, previously termed type II I-compounds, are detected by 32P-postlabeling following treatment of DNA with several Fenton-type oxygen radical-generating reagents, i.e., mixtures of Fe(II) or Ni(II) and H2O2. In an attempt to characterize the chemical nature and mechanism(s) of formation of these novel adducts, 16 single-stranded deoxyribooligonucleotides (20- and 21-mers) of known sequence were oxidized with Fe(II) or Ni(II) and H2O2, and the products were analyzed by 32P-postlabeling. Eight adducts were obtained reproducibly by oxidation of DNA and test oligonucleotides in a sequence-dependent manner. One major adduct (2) was formed only if the test oligonucleotide contained two adjacent adenine residues. Similarly, adducts 3 and 8 specifically originated in AC and CA sequences, respectively. Adduct 6 required a 5′-C-purine-3′ sequence. On the other hand, GN sequences (where N is any normal nucleotide) gave rise to adduct 1, another major product, and adduct 7. Similarly, adducts 4 and 5 were produced by the oxidation of AN sequences. These observations are most readily explained if the oxidation reactions caused intrastrand cross-links between adjacent nucleotides, leading to dimer formation. The observation that adducts 1, 4, 5, and 7 did not require a specific 3′-nucleotide was consistent with the notion that these nucleotides lacked a 3′-base, suggesting the presence of a 5′f3′ purine-sugar cross-link in the oxidized products. The majority of the lesions came from AA and 5′-purine-N-3′ sequences. The effects of Fe(II) and Ni(II) were qualitatively similar; however, higher yields of products were observed with Fe(II) as the catalyst. The definition of the chemical origins of these bulky DNA modifications, which represent a new type of DNA damage, is expected to contribute to a better understanding of the mechanism of metal carcinogenesis and to shed light upon the origins of certain endogenous DNA lesions. Recently, some of the major oxidative DNA adducts characterized here were detected by 32P-postlabeling in the renal DNA of male rats treated with ferric nitrilotriacetate, a known potent prooxidative kidney carcinogen in these animals [Randerath, E., Watson, W. P., Zhou, G. D., Chang, J., and Randerath, K. (1995) Mutat. Res. 341, 265-279].

Introduction Much effort is being directed toward elucidating the effects of reactive oxygen species (ROS)1 on DNA (1-6). ROS, which comprise hydrogen peroxide, superoxide, singlet oxygen, and hydroxyl free radical, are generated as byproducts of normal metabolism, and their formation can be intensified by chemical exposures in vivo. The ubiquitous occurrence of ROS can lead to several forms of DNA damage, e.g., strand breaks (7), DNA-protein cross-links (8, 9), apurinic and apyrimidinic sites (10, 11), and various base and sugar lesions, such as 8-oxoguanine (8-oxo-G) (2, 6, 12-14), thymine glycol (2, 15), ringopened base products (2), and ribonucleotides (16). Previous work from our laboratory using 32P-postlabeling showed that bulky adducts are produced in vitro by Fenton-type reactions of DNA with Fe(II) or Ni(II) in the presence of H2O2 (17, 18). The major adducts (spots 1 * Correspondence and requests for reprints should be addressed to this author at the Division of Toxicology, Department of Pharmacology, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030. Telephone: (713) 798-4465. Fax: (713) 798-3145. † Department of Pharmacology. ‡ Department of Pediatrics. X Abstract published in Advance ACS Abstracts, December 1, 1995. 1 Abbreviations: ROS, reactive oxygen species; PEI, poly(ethylene imine); TLC, thin-layer chromatography; RAL, relative adduct labeling; MN, micrococcal nuclease; SPD, spleen phosphodiesterase; PNK, T4 polynucleotide kinase; Nu.P1, nuclease P1; PAP, prostatic acid phosphomonoesterase; 8-oxo-G, 8-oxoguanine; 8-oxo-dG, 8-oxo-2′-deoxyguanosine; N, a normal DNA nucleotide or nucleoside.

0893-228x/96/2709-0247$12.00/0

and 2) were also detected in normal tissue DNA, where they are significantly increased in the target organs of carcinogenesis following treatment with nickel(II) acetate (18) or ferric nitrilotriacetate (19), i.e., known prooxidant metal carcinogens (20, 21). According to Randerath et al. (17, 22), these DNA adducts fit the definition of I-compounds, as they represent bulky DNA modifications that increase by putative endogenous mechanisms with age in tissue DNA of untreated animals (17, 23-26). In addition to these bulky oxidative DNA adducts, which were termed type II I-compounds (19, 27) and apparently represent lesions associated with carcinogenesis, other endogenously formed bulky DNA adducts, called type I I-compounds, rather may play a protective role in aging and carcinogenesis, as suggested by a large body of indirect experimental evidence (19, 22, 25-27). Profiles of type II I-compounds provide a new class of biomarkers for oxidative DNA damage with utility in exploring the role of oxidative mechanisms in carcinogenesis, aging, and possibly other biological processes (19, 27, 28). The chemical nature as well as mechanism(s) of formation of the bulky oxidized nucleotide derivatives have not been determined so far, but the formation of bulky adducts from dinucleotides treated with Cu(II) or Fe(II) in the presence of H2O2 has been reported (29). The much stronger chromatographic conditions used in this work, however, suggest that those products are different from the ones described here (17, 18, 22). © 1996 American Chemical Society

248 Chem. Res. Toxicol., Vol. 9, No. 1, 1996

As shown previously (18), the 32P-postlabeled adduct profiles obtained by the oxidation of denatured DNA are qualitatively identical to those of oxidized native DNA, consistent with the interpretation that these bulky DNA modifications represent intrastrand rather than interstrand DNA cross-links. In the present study, their structural origins and mechanism(s) of formation were deduced by the oxidation of a selected series of chemically synthesized oligonucleotides (20- and 21-mers) of known sequences. The results showed that eight oxidation products were reproducibly obtained from specific dinucleotide sequences of these substrates, with purines appearing to be more susceptible to chemical alteration than pyrimidines. These adducts, therefore, represent a new class of oxidative DNA lesion, whose possible roles in aging, mutagenesis, prooxidant and spontaneous carcinogenesis, and other adverse health effects need further study.

Materials and Methods Materials. Iron(II) sulfate (ACS grade) and nickel(II) acetate tetrahydrate (99.999%) were purchased from Alfa Products (Danvers, MA). As Ni(II) salts are toxic and carcinogenic, appropriate precautions need to be taken when handling these chemicals (6-8). H2O2 (certified ACS, 30% solution) and sodium acetate (certified ACS) were from Fisher Scientific (Pittsburgh, PA). DNA used for oxidation in vitro was isolated from the lungs of 3-month-old female Sprague-Dawley rats (17). Terminally dephosphorylated oligonucleotides were obtained from DNA International, Inc. (Portland, OR). The sequences, which contained each of the 16 possible dinucleotide combinations, were (GAAC)5 ()5′-GAACGAACGAACGAACGAAC-3′), (A)20, (AAG)7, (AG)10, (AC)10, (ACG)7, (AAT)7, (GGT)7, (CCT)7, (ATT)7, (ATC)7, (CTG)7, (GTA)7, (GCC)7, (TTC)7, and (AGC)7. All reagent solutions for reactions in vitro were freshly prepared. The sources of materials for 32P-postlabeling have been previously reported (17, 30-32). Poly(ethylene imine) (PEI)-cellulose thin-layer chromatography (TLC) sheets were prepared on vinyl plastic sheets in the laboratory (33). Chemical Reaction Conditions. About 0.15 µg/µL DNA or oligonucleotide was oxidized in duplicate or triplicate in aqueous solution with 5 µM Fe(II) and 3 mM H2O2 or with 50 µM Ni(II) and 3 mM H2O2 at 37 °C for 1 h. As controls, 0.15 µg/µL DNA or oligonucleotide was incubated with H2O. Isolation and Purification of DNA or Oligonucleotides. Tissue DNA was isolated by solvent extraction and enzymatic digestion of protein and RNA (34). Incubated DNA or oligonucleotides were precipitated at -80 °C immediately after reaction in the presence of 0.1 vol of 3 M sodium acetate (pH 5.2) and 2 vol of 100% ethanol for 1 h and then spun at 4 °C and 14 000 rpm for 0.5 h. Pellets were washed twice with 70% (v/v) ethanol. DNA concentrations were determined spectrophotometrically on the basis of 1 mg of DNA ) 20 A260 units. Concentrations of oligonucleotides were estimated by spectrophotometry using A260/µg values provided by the supplier. 32P-Postlabeling Analysis of DNA or Oligonucleotides. The nuclease P1 (Nu.P1)-enhanced 32P-postlabeling assay was performed as developed by Reddy and Randerath (32). Briefly, 10 µg of DNA or oligonucleotide was hydrolyzed to normal and adducted 3′-nucleotides with micrococcal nuclease (MN, 50 munits/µL) and spleen phosphodiesterase (SPD, 0.4 µg/µL). Normal 3′-mononucleotides were converted to nucleosides by incubation with Nu.P1 (0.5 µg/µL), and then the Nu.P1resistant, modified 3′-nucleotides were converted to 5′-32Plabeled nucleoside 3′,5′-bisphosphate adducts by T4 polynucleotide kinase (PNK, 0.5 unit/µL) catalyzed transfer of [32P]phosphate from [γ-32P]ATP. In order to gain further insight into the structures of the oxidation products, in some experiments prostatic acid phosphomonoesterase (PAP, 20 munits/µL) was added during the Nu.P1 step. Labeled products were separated by multidirectional PEI-cellulose TLC, such that adducts were first purified with solvent I (Table 1). Nucleotides retained in

Randerath et al. Table 1. Chromatographic Solvents for Separation of Oxidation Products solvent

composition

I II III IV

2.3 M sodium phosphate, pH 5.7 2.8 M lithium formate, 4.88 M urea, pH 3.35 0.53 M sodium phosphate, 5.3 M urea, pH 6.4 0.38 M boric acid, 0.38 M Tris-HCl, 7.5 mM EDTA, 0.98 M NaCl, 6 M urea, pH 8.0 V 0.56 M sodium phosphate, 0.35 M Tris-HCl, 5.95 M urea, pH 8.2 VI 0.56 M LiCl, 0.35 M Tris-HCl, 5.95 M urea, pH 8.0 VII 0.37 M sodium phosphate, 0.23 M Tris-HCl, 67 mM NaHCO3, 4.53 M urea, pH 8.2 VIII 1 M sodium phosphate, pH 6.0 IX isopropyl alcohol/4 N ammonium hydroxide (50:50, v/v) X 0.28 M ammonium sulfate, 50 mM sodium phosphate, pH 6.8

the origin area (2.8 × 1 cm) of the chromatogram were contacttransferred to a fresh sheet and resolved by two-dimensional TLC with solvent II in the first and solvent III in the second direction. 32P-Labeled adducts were located by screen-enhanced autoradiography employing DuPont Cronex-4 film at room temperature for 1-16 h. Re- and Cochromatography. In order to match adducts 1-8 from oxidized DNA (Figure 1, panel 1a) and oxidized oligonucleotides (Figure 1, all panels except 1a, 4c, and 4d), corresponding areas of chromatograms from different sources were excised separately, and 32P-labeled material was extracted by shaking the cut-outs, layer side facing down, with 900 µL of 6 N ammonium hydroxide/isopropyl alcohol (1:1, v/v) for 1520 min at room temperature (35). 32P-Labeled normal nucleotides were isolated similarly. Replicate extracts from the same source were combined and centrifuged twice at 10 000 rpm for 10 min, and then the supernates were evaporated to dryness. The eluate radioactivity was determined by Cerenkov spectrometry, and the pellets were dissolved in water to a final concentration of 100-200 cpm/µL. The eluates were separated individually or, for matching, mixed with another extract by onedimensional ascending PEI-cellulose TLC, and spots were located by autoradiography. Solvents I-III (Table 1) were used for the basic 32P-postlabeling procedure (see above), solvents II-X for re- and cochromatography, and solvent X, in addition, for the separation of oxidative adducts from normal nucleotides.

Results Oxidation Products in DNA and Oligonucleotides. As shown in panel 1a of Figure 1, the oxidation of denatured lung DNA with FeSO4 and H2O2 in vitro produced two strong adduct spots (denoted 1 and 2) and a number of weaker spots (3-13) upon PEI-cellulose TLC. This pattern was identical to that reported previously, with traces of spots 1 and 2 being detected in control lung DNA (17, 18). To determine the sequences giving rise to these adducts, oligonucleotides of known sequences were treated with FeSO4 and H2O2 and then 32P-postlabeled. Representative 2D maps from oligonucleotide oxidation reactions are illustrated in Figure 1 (all panels except 1a and 4d). Adducts 2, 4, and 5. Panels 1b, 1c, 1d, and 2a (Figure 1) depict maps of 32P-labeled oxidation products of oligonucleotides containing two adjacent adenine nucleotides [i.e., (A)20, (AAG)7, (AAT)7, and (GAAC)5]. The twodimensional profiles shown shared three spots (2, 4, and 5) among each other and with the oxidized DNA (panel 1a). This was confirmed by cochromatography in multiple conditions. As exemplified in Figure 2A, adduct 2 from DNA, (A)20, (AAG)7, (AAT)7, and (GAAC)5 was chromatographically identical in solvent VI (Table 1), as well as in seven additional solvents, i.e., II-V and VIIIX (autoradiograms not shown). Formation of spot 2 required the presence of two successive adenine residues, i.e., this adduct was not formed by the oxidation of (AC)10,

Bulky Oxidative DNA Adducts (Type II I-Compounds)

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 249

Figure 1. Autoradiograms of 32P-labeled oxidation products from DNA or oligonucleotides induced by the Fenton reaction. About 0.15 µg/µL DNA or oligonucleotide was incubated in vitro at 37 °C for 1 h with 5 µM FeSO4 + 3 mM H2O2 and analyzed by 32Ppostlabeling and PEI-cellulose TLC. Chromatographic and other experimental conditions are given in Materials and Methods. Film exposure employed DuPont Cronex-4 film and Lightning Plus intensifying screens for 1-2 h at room temperature except for sample 1a, which was exposed for 16 h. Panels: 1a, DNA; 1b, (A)20; 1c, (AAG)7; 1d, (AAT)7; 2a, (GAAC)5; 2b, (AC)10; 2c, (ACG)7; 2d, (ATC)7; 3a, (CTG)7; 3b, (GTA)7; 3c, (GGT)7; 3d, (GCC)7; 4a, (AG)10; 4b, (ATT)7; 4c, (CCT)7; 4d, H2O control of (GAAC)5. The slight variation in the migration of spots in panels 2a and 2c was due to batch differences of the TLC sheets. Several circled weak spots (e.g., 7 in 3c and 6 in 3d) needed longer film exposure (6 h) for detection.

(ACG)7, (ATC)7, (GTA)7, (AG)10), and (ATT)7 (Figure 1). Oligonucleotides lacking adenine [i.e., (CTG)7, (GGT)7, (GCC)7, and (CCT)7] also did not give rise to adduct 2. In contrast, adducts 4 and 5 were detected not only in oxidized DNA, (A)20, (AAG)7, (AAT)7, and (GAAC)5 but also in oligonucleotides having an isolated adenine nucleotide [(AC)10, (ACG)7, (ATC)7, (GTA)7, (AG)10, (AGC)7, and (ATT)7]. Nevertheless, these adducts required the presence of adenine, i.e., they were not produced by the oxidation of (CTG)7, (GGT)7), (GCC)7, and (CCT)7. These results were consistent with adducts 4 and 5 being derived from NAN sequences, where N could be any

nucleotide (see Discussion). Similar to spot 2 (Figure 2A), the identities of these spots from oxidized DNA and oligonucleotides were confirmed by cochromatography (autoradiograms not shown). (AC)10, (ATC)7, (GTA)7, and (AG)10 gave several minor spots in the area of spot 2 (Figure 1), but none of them migrated with spot 2 upon rechromatography. Adducts 3, 6, and 8. The map from oxidized (AC)10 (Figure 1.2b) showed, in addition to spots 4 and 5, three products (3, 6, and 8) that chromatographed to locations similar to those observed for oxidized DNA (panel 1a). As exemplified in Figure 2B for adduct 3 and

250 Chem. Res. Toxicol., Vol. 9, No. 1, 1996

Randerath et al.

Figure 3. One-dimensional cochromatography of spots 1-8 from oxidized DNA (Figure 1.1a) or oligonucleotides (e.g., Figure 1.2a) alongside 32P-labeled normal mononucleotides in solvent X: lane 1, [32P]dpAp; lane 2, [32P]dpGp; lane 3, [32P]dpTp; lane 4, [32P]dpCp; lane 5, spot 1; lane 6, spot 2; lane 7, spots 3 and 5; lane 8, spot 4; lane 9, spot 6; lane 10, spot 7; lane 11, spot 8; lane 12, a mixture of the four [32P]dpNp’s.

Figure 2. Examples of one-dimensional cochromatography of isolated adducts 1, 2, and 3 obtained from oxidized DNA (Figure 1.1a) or oligonucleotides (e.g., Figure 1.2a) in solvent VI (see Table 1). For cochromatography conditions, see Materials and Methods. Panel A, spot 2: lane 1, (A)20; lane 2, (A)20 + DNA; lane 3, DNA; lane 4, (AAG)7 + DNA; lane 5, (AAG)7; lane 6, (AAT)7; lane 7, (AAT)7 + DNA; lane 8, DNA; lane 9, DNA + (GAAC)5; lane 10, (GAAC)5. Panel B, spot 3: lane 1, (AC)10; lane 2, (AC)10 + DNA; lane 3, DNA; lane 4, DNA + (ACG)7; lane 5, (ACG)7. Panel C, spot 1: lane 1, (GAAC)5; lane 2, (GAAC)5 + DNA; lane 3, DNA; lane 4, (GGT)7 + DNA; lane 5, (GGT)7; lane 6, (GGT)7 + (ACG)7; lane 7, (ACG)7; lane 8, (ACG)7 + (CTG)7; lane 9, (CTG)7; lane 10, (CTG)7 + (GTA)7; lane 11, (GTA)7. The chromatograms were developed to 14 cm past the origin, but autoradiograms were cut at 11 cm. Faint spots were from traces of oligonucleotide contaminants or decomposition products.

solvent VI, these adducts matched spots 3, 6, and 8 from oxidized DNA in solvents II-IX (autoradiograms for spots 6 and 8 and solvents II-V and VII-IX not shown). Spot 3 was obtained only from oligonucleotides having AC sequences [(GAAC)5, (AC)10, and (ACG)7]. The fact that adduct 3 was obtained from these sequences, but not from (ATC)7 and (AGC)7 (map not shown), indicated that it was not derived by oxidation of CA. Therefore, the formation of adduct 3 required AC. On the other hand, spot 8 was produced only from (AC)10, (ATC)7, and (AGC)7, but not from (GAAC)5 or (ACG)7; thus it was an oxidation product of CA. While in the two-dimensional system (Figure 1) spots 2 and 8 migrated similarly, they were clearly distinguished by rechromatography in solvent X (Figure 3). A faint double spot in the area of spots 2 and 8 (Figure 1.2c) from (ACG)7 migrated differently upon rechromatography in solvents VI and X. Spot 6 was given by CA sequences [i.e., (AC)10, (ATC)7, and (AGC)7] and also by CG sequences [i.e., (ACG)7, (GAAC)5, and (GCC)7], and thus its formation depended upon the presence of a 5′-C-purine-3′ sequence. CC or CT sequences [e.g., (CCT)7 and (TTC)7] did not give rise to any oxidation products under the conditions used.

Adducts 1 and 7. Adducts 1 and 7 required the presence of a single guanine or a guanine pair [i.e., (AAG)7, (GAAC)5, (ACG)7, (CTG)7, (GTA)7, (GGT)7, (GCC)7, and (AG)10]. These products were not formed from (A)20, (AAT)7, (AC)10, (ATC)7, (ATT)7, and (CCT)7. Spot 1 (see example in Figure 2C) from DNA, (GAAC)5, (GGT)7, (ACG)7, (CTG)7, and (GTA)7 was chromatographically identical in solvents II-IX (Table 1), and this was also observed for spot 7 and (AAG)7, (GCC)7, and (AG)10 (maps not shown). These data implied that adducts 1 and 7 originated in NGN sequences. Additional Labeled Products. Weak spots 9-13 (Figure 1.1a) were detected only in oxidized DNA, and therefore their structural origins were not identified. Besides the numbered adducts, additional, mostly minor spots were detected in some oxidized oligonucleotide samples, but were absent from the treated DNA sample (Figure 1.1a). As illustrated in Figure 1.4d, these spots were also observed in control incubations of oligonucleotides with H2O; thus, they were presumably due to contaminants, such as residues of protecting groups used in the synthesis of the oligonucleotides (36). Oxidation Products versus Normal Nucleotides. The isolated oxidation products were chromatographed alongside 32P-labeled normal deoxyribonucleoside 3′,5′monophosphates (dpNp’s) in solvent X in order to exclude the possibility that they represented unmodified DNA mononucleotides. As shown in Figure 3, all of the oxidation products migrated much more slowly than normal nucleotides upon anion-exchange TLC, remaining close to the origins in solvent X. Thus, it was apparent that the products analyzed here did not reflect normal mononucleotides. The possibility that they were unmodified oligonucleotides was excluded by their susceptibilities to hydrolysis by the various nucleases used in this study. Structural Features of Oxidation Products. As summarized in Table 2, all of the oxidation products (spots 1-8) appeared to be derivatives of two adjacent nucleotide residues, which presumably became covalently linked to each other upon oxidation via a base-base or base-sugar cross-link. When subjected to Nu.P1-enhanced 32P-postlabeling (32), these oxidation products behaved like bulky aromatic carcinogen-adducted nucle-

Bulky Oxidative DNA Adducts (Type II I-Compounds)

Chem. Res. Toxicol., Vol. 9, No. 1, 1996 251

Table 2. Summary of Results Obtained from Oxidized Oligonucleotides reaction products

oligonucleotides that gave the adduct

sequence deduced

adducts 1 and 7 adduct 2 adduct 3 adducts 4 and 5 adduct 6 adduct 8

(AG)10, (AAG)7, (ACG)7, (GAAC)5, (GGT)7, (GCC)7, (CTG)7, (GTA)7, (AGC)7a (A)20, (AAG)7, (AAT)7, (GAAC)5 [not given by (CTG)7 or single A] (AC)10, (ACG)7, (GAAC)5 [not given by (ATC)7 or (AGC)7a] (A)20, (AAG)7, (AAT)7, (GAAC)5, (AC)10, (ACG)7, (GTA)7, (ATC)7, (AG)10, (ATT)7, (AGC)7a (AC)10, (ACG)7, (GAAC)5, (ATC)7, (GCC)7, (AGC)7a [not given by (CCT)7 or (TTC)7a] (AC)10, (ATC)7, (AGC)7a [not given by (ACG)7 or (GAAC)5]

GN AA AC AN C-purine CA

a

Data not shown in the results.

Figure 4. Schematic representation of the proposed sequence of enzymatic reactions leading from initial oxidized oligonucleotides to final 32P-labeled adducts, as exemplified for spots 1 (A) and 2 (B). For details, consult text. Zigzag lines indicate putative cross-links formed by oxidation between adjacent nucleotides. The dots in I and V signify the remaining polynucleotide sequence. Asterisks denote 32P-labeled phosphomonoester groups. Parentheses indicate phosphomonoester groups whose absence has not been demonstrated (see Discussion).

Figure 6. Quantitative estimation (31, 32) of oxidation products (RAL × 106) from oligonucleotides induced by Fe(II) or Ni(II) under conditions shown in Figures 1 and 5. Coefficients of variation for two independent oxidation reactions each were 6-14%. Table 3. Properties of Enzymes Used in the Characterization of Oxidized Products enzyme

Figure 5. Typical autoradiograms of 32P-labeled products induced by oxidation of DNA or oligonucleotides with Ni(II) and H2O2. About 0.15 µg/µL DNA or oligonucleotides were incubated in vitro at 37 °C for 1 h with 50 µM NiAc2 + 3 mM H2O2. Experimental conditions were the same as for Figure 1. Film exposure was for 5-6 h at room temperature except for 1a, which was for 16 h. Panels: 1a, DNA; 1b, (A)20; 1c, (AC)10; 2a, (ATC)7; 2b, (GGT)7; 2c, (CTG)7.

otides (17-19, 23-26) in terms of both their biochemical and chromatographic properties. For further information regarding the properties of the enzymes used, consult Table 3. The evidence presently available suggests to us that the most likely sequence of reactions leading to the major labeled adducts 1 and 2 was as illustrated in Figure 4 (for details, see Discussion).

properties

exceptions

MN

NpNpNp...

NpNp + Np...

few modified nucleotides

SPD

NpNpNp...

Np + NpNp...

few modified nucleotides and 5′-abasic sites

Nu.P1

NpNpNp...

N + pN

many modified nucleotides

PAP

pNpNpNp

PNK

Np

+ ATP

NpNpN + 2Pi pNp

abasic sites, nucleosides

Ni(II)-Catalyzed Formation of Oxidative Adducts. Previous work from our laboratory (18) has shown that Ni(II), a metal carcinogen, induces bulky DNA adducts in the presence of H2O2 in vitro, with most of them being chromatographically identical to those elicited by Fe(II) and H2O2, although Ni(II) is 12-25 times less effective than Fe(II) in this respect. To determine whether Ni(II) and Fe(II) induced similar oligonucleotide oxidation products in the presence of H2O2, DNA and the test oligonucleotides were treated with 5 µM Fe(II) and 3 mM H2O2 or with 50 µM Ni(II) and 3 mM H2O2 and then analyzed by 32P-postlabeling. Results obtained with DNA and five oligonucleotides are exemplified in Figure 5. A comparison with the examples given in Figure 1 revealed that Ni(II) and Fe(II) produced similar spot patterns; however, there were distinct quantitative differences (Figure 6). For instance, Fe(II), but not Ni(II), treatment of (A)20 resulted in high yields of spots 2 and 4, in spite of the 10-fold lower Fe(II) concentration. On the other hand, the two metal compounds gave similar overall yields of spots 3 and 4 from (AC)10 under these conditions, and this was also true for spot 1 from (GGT)7. In

252 Chem. Res. Toxicol., Vol. 9, No. 1, 1996

contrast, Fe(II) yielded much more adduct 1 from (CTG)7, but not from (GGT)7, than did Ni(II). These observations suggested that the extent of oxidation was affected by adjacent nucleotides.

Discussion Previous work employing 32P-postlabeling provided evidence for a new type of oxidative DNA damage comprising bulky adducts termed type II I-compounds (see Introduction). The main objective of this study has been to investigate the structural origins and possible mechanisms of formation of these adducts. As initial studies suggested to us that many of the type II Icompounds represented intrastrand DNA cross-links (18), different oligonucleotides of known sequence were oxidized under Fenton-type reaction conditions in order to determine whether these adducts were associated with specific sequences. The results showed that eight products (spots 1-8) detectable in oxidized DNA (Figure 1.1a) also arose by the oxidation of oligonucleotides, with the formation of individual adducts being sequence-dependent. The oxidation products were derivatives of adjacent nucleotides (Table 2). The condition for the formation of one of the two major type II I-compounds (adduct 1) was the presence of a guanine residue, irrespective of the 5′- or 3′-adjacent nucleotide (Figure 1.1c, 1.2a, 1.2c, 1.3a, 1.3b, 1.3c, 1.3d, and 1.4a). Cochromatography (Figure 2C) revealed the chromatographic identity of the adduct, independent of the source from which it was isolated. These results indicated the absence from the final product of a base at either the original 5′- or 3′-adjacent nucleotide positions, because the presence of a second base would have strongly affected the anion-exchange chromatographic behavior of the product (37). Loss of the 5′-adjacent base could be excluded as such a compound is not a substrate for hydrolysis by SPD or 32P-labeling by PNK (11, 38) (Table 3), and this was inconsistent with our results. In addition, adduct 1 was different from 32P-labeled normal 2′-deoxyguanosine 3′,5′-bisphosphate (Figure 3), which might have resulted from incomplete Nu.P1 digestion (Table 3). The corresponding 8-oxo-dG nucleotide could be excluded also, as this compound migrates in close proximity to normal dpGp (16). It would thus seem reasonable to postulate that adduct 1 simply contained a 3′,5′-phosphodiester link between a 5′-guanine nucleotide and the original 3′-adjacent nucleotide, which had lost its original base or undergone further oxidation to a phosphoglycolate derivative. Both of these structures can be generated by radiation and genotoxic agents, leading to the formation of ROS (11, 39). However, abasic sites and phosphoglycolates are highly sensitive to Nu.P1 cleavage (11, 38), in contrast to what was observed here for adduct 1. As the phosphodiester bond between the 5′-guanine nucleotide and the 3′-sugar was cleaved by SPD without the formation of 2′-deoxyguanosine 3′monophosphate (see Results), these two moieties must have been connected to each other via a covalent bond other than a phosphodiester bond. Therefore, our data are consistent with a structure where adduct 1 contained a cross-link between guanine and the 3′-adjacent sugar (product IV of Figure 4A). In support of this interpretation, noncovalent stacking interactions between successive bases and deoxyribose residues in DNA helices are well-known [e.g., ref 40 and references quoted therein]. Evidence for the right-hand 3′-phosphate in IV (Figure

Randerath et al.

4A) came from the fact that 3′-dephosphorylation by Nu.P1 would have required an intact base residue (11). All our data were in accord with the sequence of reactions depicted in Figure 4A, wherein the oxidized oligonucleide (I) was initially cleaved by MN and SPD to give product II, in which the remainder of the original right-hand 3′-nucleotide was linked covalently like a bulky adduct to the 5′-adjacent guanine. As 3′-phosphates of bulky adducts are usually resistant to cleavage by Nu.P1 (32, 41), the most likely product of Nu.P1 hydrolysis of II was III. Labeling of III then led to the formation of IV. In agreement with this scheme, when PAP was present during the Nu.P1 step (see Materials and Methods), no labeling of the product was detected. This observation excluded the presence of 3′,5′-phosphodiester bonds in the intermediates (II and III) as well as in the 32P-labeled final dimeric product (IV). The other major type II I-compound (adduct 2) required two neighboring adenine residues (Figure 1.1b, 1.1c, 1.1d, and 1.2a). Cochromatography ruled out that this compound was 32P-labeled normal dpApA (map not shown), which, in addition, would have been degraded by Nu.P1 prior to 32P-postlabeling (Table 3). The conversion of V to VI (Figure 4B) proceeded via the expected enzymatic conversion of each of the 3′, 5′-phosphodiester bonds of the oxidized oligonucleotide to a 3′-phosphomonoester bond (Table 3). Evidence for the presence of a 3′phosphomonoester bond in the oxidized dinucleotide (VI) came from the observation that 32P-labeling did not occur if PAP (Table 3) was present during the Nu.P1 step. On the basis of the chromatographic properties of the final labeled product, we believe that, upon incubation of VI with Nu.P1, only one of the two 3′-monophosphate groups was removed, as illustrated by structure VII, because if both groups had been retained, the resulting labeled product (VIII) would have contained four phosphomonoester groups. This possibility was inconsistent with the fast rate of migration of adduct 2 in a number of solvents upon PEI-cellulose anion-exchange TLC (see, for example, Figures 1, 2, and 5). It was not clear, however, which of the two phosphomonoester groups was resistant to Nu.P1, although for steric reasons, and in agreement with other observations on bulky DNA adducts (32, 41, 42), the one attached to the 5′-terminal deoxyadenosine residue was the more likely choice, as illustrated in Figure 4B. All of these observations, and the fact that formation of adduct 2 required an AA sequence, implied an oxidative cross-link between the successive adenine residues, i.e., the presence of an adenine dimer (V) in the original oxidized oligonucleotides. The origins of adducts 3-8, which were formed in smaller amounts than adducts 1 and 2 in oxidized DNA (Figure 1.1a), nevertheless could be identified (Table 2). In additional experiments (data not shown), spots 4-8 were not detected if PAP was present during the Nu.P1 hydrolysis step; therefore, these spots apparently were formed by mechanisms similar to those illustrated in Figure 4 for adducts 1 and 2. On the other hand, adduct 3 apparently contained a 3′,5′-phosphodiester bond resistant to both MN/SPD hydrolysis and Nu.P1 digestion (Table 3), as the presence of PAP during the Nu.P1 step did not interfere with the formation of this adduct (data not shown). As previously shown (17, 18), adducts 1 and 2 occur endogenously in unexposed rodents, and their levels are enhanced by exposure to prooxidant metal carcinogens

Bulky Oxidative DNA Adducts (Type II I-Compounds)

(18, 19). Small amounts of other oxidation products, e.g., spots 3 and 4, could also be detected under the latter conditions. This suggests that different amounts of multiple bulky oxidative DNA lesions can arise in DNA in vivo. The specific factors determining the relative amounts of these products, including tissue, species, gender, and the nature of the inducing agent, still remain to be determined. Possible modulations of these lesions by micronutrients, such as antioxidants, also deserve consideration. Because of the high reactivity of ROS and their capacity for modifying most biological molecules, including DNA, protein, and membrane lipids (43, 44), DNA damage not only can be generated by direct reaction of ROS with DNA, leading to both small and bulky adducts, but also may result from ROS-induced protein-DNA cross-links (8, 9) and the binding of lipid peroxidation products to DNA (43-45). The present data have established the bulky adducts as a new type of oxidative DNA damage produced directly by ROS, with each of the oxidation products (1-8) originating in a dinucleotide sequence containing at least one purine and three adducts (3, 6, and 8) containing cytosine. No thymine adducts were detected. Since oxygen free radicals and lipid peroxide-related radicals have the potential to damage any nucleotide sequence in DNA, giving rise to polar and nonpolar lesions, the adducts detected in this study do not reflect the totality of DNA oxidation products. Furthermore, the chromatographic conditions employed here would not have been suitable for the detection of even bulkier, highly nonpolar oxidative lesions, and some pyrimidine-containing DNA modifications, such as thymine dimers, are not substrates for PNK (11, 46) and thus presumably would not have been detected. The biological role of the type II I-compounds described here has not yet been clarified; however, their bulky adduct-like properties reflect a major alteration of the DNA sequence. The magnitude of the chemical alteration implies that the type II I-compounds represent biologically significant DNA lesions that may contribute to carcinogenesis through the mutation of oncogenes and tumor suppressor genes. In addition, as shown for pyrimidine dimers (47, 48), this type of DNA modification is expected to cause alterations of the secondary and tertiary structure of DNA, which may have important biological consequences. Recent studies have shown that TPA-induced oxidative DNA damage (49) may cause activation of the ras oncogene by converting CAA to CGA or CTA at codon 61 (50-52), consistent with our finding that the AA sequence is particularly susceptible to attack by ROS. This interpretation is further supported by the observation that this sequence is one of the hot-spots for mutations in X-irradiated mouse fibroblasts (53). Further studies are needed to determine the biological significance of the novel oxidation products, in relation to their chemical structures. As these DNA lesions exhibit low background levels in normal adult animals and are readily detected by 32P-postlabeling, they represent useful biomarkers of oxidative DNA damage. This aspect appears to be especially important as the classical biomarker of oxidative DNA damage, 8-oxo-dG, has recently been suggested to be not as specific as previously thought (54). The small amounts of biological material needed (