Oxidative DNA Damage by Vitamin A and Its Derivative via Superoxide Generation*

Recent intervention studies revealed that β-carotene supplement to smokers resulted in a higher incidence of lung cancer. However, the causal mechanisms remain to be clarified. We reported here that vitamin A (retinol) and its derivative (retinal) caused cellular DNA cleavage detected by pulsed field gel electrophoresis. Retinol and retinal significantly induced 8-oxo-7,8-dihydro-2′-deoxyguanosine formation in HL-60 cells but not in H2O2-resistant HP100 cells, suggesting the involvement of H2O2 in cellular DNA damage. Experiments using 32P-labeled isolated DNA demonstrated that retinol and retinal caused Cu(II)-mediated DNA damage, which was inhibited by catalase. UV-visible spectroscopic and electron spin resonance-trapping studies revealed the generation of superoxide and carbon-centered radicals, respectively. The superoxide generation during autoxidation of retinoids was significantly correlated with the formation of 8-oxo-7,8-dihydro-2′-deoxyguanosine, although the yield of carbon-centered radicals was not necessarily related to the intensity of DNA damage. These findings suggest that superoxide generated by autoxidation of retinoids was dismutated to H2O2, which was responsible for DNA damage in the presence of endogenous metals. Retinol and retinal have prooxidant abilities, which might lead to carcinogenesis of the supplements of β-carotene.

Many studies have addressed the role of antioxidant vitamins A, C, and E in protection against cancers and cardiovascular diseases (1). It has been suggested that the antioxidant potency of vitamin A and ␤-carotene may scavenge oxygen radicals and protect against cancer occurrence (1)(2)(3). The Alpha-tocopherol, Beta-carotene Cancer Prevention (ATBC) study group (4) and the Beta-carotene and Retinol Efficacy Trial (CARET) (5) supplied ␤-carotene and/or vitamin A to smokers and asbestos-exposed workers, who were high risk groups for lung cancer. After following up for several years, surprisingly, higher incidences of lung cancer were observed in the intervention groups than the placebo groups. The CARET researchers have suggested that ␤-carotene can function as a prooxidant under the free-radical-rich atmosphere produced by the chemicals in cigarette smoke and the resultant inflammatory response in the lung (6). The ATBC researchers suspected that ␤-carotene acts as a promoter of pre-existing latent lung cancers in smokers (7). However, attempts to use retinoids and cartenoids for cancer chemoprevention and therapy are ongoing (8 -11). Therefore, the causal mechanisms should be elucidated to establish safe approaches in cancer chemoprevention.
Every antioxidant, including vitamin antioxidants, is in fact a redox (reduction-oxidation) agent, protecting against free radicals in some circumstances and promoting free radical generation in others (12). Several studies have revealed prooxidant effects of vitamin E (13,14) and vitamin C (15)(16)(17) under certain circumstances. Other antioxidants such as N-acetylcysteine (18) and quercetin (19,20) can cause oxidative damage to cellular and isolated DNA. ␤-Carotene may also act as both an antioxidant and a prooxidant under various oxygen partial pressures (21)(22)(23). Although vitamin A is a good acceptor and donor of electrons in chemical reactions, its properties appear to be very carefully protected by retinol-binding proteins and other endogenous antioxidants in vivo (24). However, pharmacological amounts of the supplements above physiological amounts may perturb key physiological processes.
We report here that low concentrations of vitamin A (retinol) and its derivative (retinal; vitamin A aldehyde) caused cellular DNA cleavage. Furthermore, induction of 8-oxo-7,8-dihydro-2Јdeoxyguanosine (8-oxodG) 1 by retinol and retinal was investigated in HL-60 cells and its H 2 O 2 -resistant clone HP100 cells. To clarify the mechanism of DNA damage, we performed experiments using 32 P-labeled DNA isolated from the human p53 tumor suppressor gene and c-Ha-ras-1 protooncogene. In addition, the time course of 8-oxodG formation was measured by HPLC-ECD, and the reactive species was also detected using UV-visible and ESR spectroscopies.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes (SmaI, EcoRI, StyI, and ApaI), calf intestine phosphatase, and proteinase K were purchased from Roche Molecular Biochemicals. Restriction enzymes (HindIII, AvaI, and PstI) and T 4 polynucleotide kinase were purchased from New England Biolabs.
Detection of Cellular DNA Damage by Pulsed Field Gel Electrophore-sis-HL-60 cells were grown in RPMI 1640 supplemented with 6% fetal bovine serum at 37°C under 5% CO 2 in a humidified atmosphere. Then the cells were washed, resuspended in 10 ml of RPMI 1640 containing 6% fetal bovine serum (1 ϫ 10 6 cells/ml), and incubated with retinoid, which was dissolved in ethanol (final concentration of ethanol, 0.05%) at 37°C for the indicated duration. Control conditions also contained 0.05% ethanol. After incubation, the medium was removed, and the cells were washed twice with phosphate-buffered saline and resuspended in phosphate-buffered saline. The cell suspension was solidified with agarose, followed by treatment with proteinase K according to the method described previously (25). Electrophoresis was performed in 0.5 ϫ TBE buffer (45 mM Tris, 45 mM boric acid, 1 mM EDTA (pH 8.0)) by a CHEF-Mapper pulsed field electrophoresis system (Bio-Rad) at 200 V at 14°C. Switch time was 60 s for 15 h followed by 90 s switch time for 9 h. The DNA in the gel was visualized using ethidium bromide. Analysis of 8-oxodG Formation in HL-60 and HP100 Cells Treated with Retinoids-HP100 cells have been derived from HL-60 cells by repeated exposure to H 2 O 2 followed by outgrowth of viable cells and are approximately 340-fold more resistant to H 2 O 2 than HL-60 cells (26,27). HL-60 and HP100 cells (1 ϫ 10 6 cells/ml) were incubated with retinoids in 5 ml of RPMI 1640 supplemented with 6% fetal bovine serum at 37°C. After the incubation, the medium was removed, and the cells were washed twice with phosphate-buffered saline. The cells were suspended in 0.05 mg/ml RNase A, 0.5 mg/ml proteinase K, and 500 l of DNA extractor and incubated for 60 min at 60°C. After ethanol precipitation, DNA was digested to nucleosides with nuclease P 1 and bacterial alkaline phosphatase and analyzed by HPLC-ECD, as described previously (28,29).
Preparation of 32 P 5Ј-end-labeled DNA Fragments-DNA fragments were obtained from the human p53 tumor suppressor gene. 2 Fragments from the p53 gene containing exons were prepared as described previously (31). The 5Ј-end-labeled 470-bp fragment (HindIII* 13038-EcoRI* 13507) was obtained by dephosphorylation with calf intestine phosphatase and rephosphorylation with [␥-32 P]ATP and T 4 polynucleotide kinase (the asterisk indicates 32 P-labeled). The 470-bp fragment was further digested with StyI to obtain the singly labeled 118-bp fragment (HindIII* 13038-StyI 13155), as described previously (31). A DNA fragment was also obtained from the human c-Ha-ras-1 protooncogene (32). A DNA fragment was prepared from plasmid pbcNI, which carries a 6.6-kilobase BamHI chromosomal DNA segment containing the c-Haras-1 gene, and the singly labeled 337-bp fragment (PstI 2345-AvaI* 2681) was obtained according to the method described previously (33). Nucleotide numbering starts with the BamHI site (32).
Detection of Damage to 32 P-5Ј-End-labeled DNA Fragments by Retinoids-The standard reaction mixture (in a microtube, 1.5 ml) contained retinoid, CuCl 2 (or FeCl 3, FeSO 4 , MnCl 2 ), 32 P-5Ј-end-labeled DNA fragments and sonicated calf thymus DNA in 200 l of 10 mM sodium phosphate buffer (pH 7.8) containing 2.5 M DTPA. DTPA was added to remove metal ions, which may have been contained in the phosphate buffer. After incubation at 37°C for the indicated durations, the DNA fragments were heated at 90°C in 1 M piperidine for 20 min where indicated.
The preferred cleavage sites were determined by direct comparison of the positions of the oligonucleotides with those produced by the chemical reactions of the Maxam-Gilbert procedure (34) using a DNA-sequencing system (LKB 2010 Macrophor). The relative amounts of oligonucleotides from the treated DNA fragments were measured with a laser densitometer (LKB 2222 UltroScan XL).
Analysis of 8-oxodG Formation in Calf Thymus DNA by Retinoids-Calf thymus DNA fragments were incubated with retinoid in the presence of CuCl 2 for the indicated duration at 37°C. DNA fragments were denatured at 90°C for 5 min and quickly chilled before incubation, where indicated. After ethanol precipitation, DNA was digested to the nucleosides using nuclease P 1 and calf intestine phosphatase and analyzed by HPLC-ECD, as described previously (28).
Detection POBN-The generation of radicals in the reaction system of retinol or retinal was detected using a trapping agent POBN with ESR spectroscopy (JEOL model JES-TE100) in a 100-kHz field modulation at room temperature. Spectra were recorded with a microwave power of 8 mW, a modulation width of 0.1 mT, and a receiver amplitude of 200.

Cellular DNA Damage Induced by
Retinoids-DNA strand breakage in cultured HL-60 cells treated with retinoids was detected by pulsed field gel electrophoresis (Fig. 1). Both retinol and retinal induced weak DNA strand breakage to produce 1000 -2000-kilobase fragments at 1 M and apparent breakage at 2 M and 5 M. Formation of 200 -300-kilobase fragments increased in cells treated with 5 M retinal but not retinol. This indicates that retinal has a stronger ability to damage DNA than retinol.
Formation of 8-oxodG in Cells Treated with Retinoids-As shown in Fig. 2, formation of 8-oxodG in HL-60 cells treated with 2-5 M of retinoids was significantly higher than that of the control. The content of 8-oxodG induced by retinal was significantly higher than that of retinol at 2 and 5 M. However, no significant increase in 8-oxodG was observed in H 2 O 2resistant HP100 cells treated with retinoids (data not shown). These findings suggest that the generation of H 2 O 2 plays a critical role in oxidative DNA damage by retinol and retinal.
Damage to 32 P-Labeled DNA Fragments Induced by Retinoids-Both retinol and retinal caused DNA damage in the presence of Cu(II), although retinoids alone did not (Fig. 3). The intensity of DNA damage increased depending on increasing concentrations of retinol and retinal (data not shown). Although the DNA damage also increased depending on incubation time, retinal caused DNA damage more efficiently than retinol after a 6-h incubation. Retinol and retinal did not cause DNA damage in the presence of other transition metals (Mn(II), Fe(II), or Fe(III)) (data not shown).
Effects of Scavengers and Bathocuproine on DNA Damage by Retinoids- Fig. 4 shows the effects of radical scavengers and a Cu(I)-specific chelator bathocuproine on retinol-induced DNA damage. Typical free hydroxyl radical (⅐OH) scavengers such as mannitol, sodium formate, and dimethyl sulfoxide showed little or no inhibitory effect on the DNA damage. However, the DNA damage was inhibited by methional, which can scavenge not only free ⅐OH but also other radicals like cryptohydroxyl radicals (35). SOD had little inhibitory effect on DNA damage. Catalase and bathocuproine completely inhibited Cu(II)-medi-2 P. Chumakov, EMBL data library, accession number X54156.  Site Specificity of DNA Damage by Retinoids-To examine the DNA damage site, 32 P-5Ј-end-labeled DNA fragments treated with retinol plus Cu(II) followed by piperidine treatment were electrophoresed. An autoradiogram was obtained and scanned with a laser densitometer to measure the relative intensity of DNA damage in the human p53 gene and the c-Ha-ras-1 protooncogene as shown in Fig. 5. Retinol and Cu(II) strongly generated piperidine-labile sites at cytosine and thymine residues, especially at the cytosine residues in CC sequences of double-stranded DNA from the p53 tumor suppressor gene (Fig. 5A). Retinol caused Cu(II)-dependent DNA cleavage at thymine residues in fragments from the c-Ha-ras-1 protooncogene (Fig. 5B). Also, the similar patterns were obtained with retinal (data not shown).
Formation of 8-oxodG in Calf Thymus DNA by Retinoids in the Presence of Cu(II)-We measured the time course of 8-ox-odG formation in calf thymus DNA incubated with retinoids in the presence of Cu(II) (Fig. 6). Within 2 h, the content of 8-oxodG formation by retinol was higher than that by retinal. No difference between retinol and retinal in 8-oxodG contents was observed at 3 h. After a 6-h incubation, 8-oxodG formation by retinal was higher than that by retinol.  (Fig. 7). Generation of Radicals Derived from Retinoids-ESR spectroscopic measurement with a trapping agent POBN detected retinoid-derived radicals with six-line signals (Fig. 8). The hyperfine splitting constants of the radical derived were a N ϭ 1.56 mT, a H ϭ 0.27 mT for retinol, and a N ϭ 1.56 mT, a H ϭ 0.25 mT for retinal. These radicals were both assigned as carboncentered radicals by reference to the reported hyperfine splitting constants (36). Retinol-derived radicals could be detected immediately after mixture, whereas retinal-derived radicals were detected only after the longer incubation. After a 6-h incubation, the radical signals from retinol were much larger than those from retinal. Under anaerobic conditions, the POBN-trapping radicals were much less accumulated (data not shown).

DISCUSSION
We report here that both retinol and retinal caused oxidative damage to cellular and isolated DNA. Retinoids significantly induced 8-oxodG formation in HL-60 cells but did not significantly increase 8-oxodG in H 2 O 2 -resistant HP100 cells. It is suggested that the generation of H 2 O 2 plays a critical role in oxidative DNA damage by retinol and retinal. When the isolated DNA was used, the existence of Cu(II) was required for the DNA damage. Copper exists in the nucleus and is closely associated with chromosomes and DNA bases in vivo (37). In cell-free systems, catalase and bathocuproine inhibited Cu(II)mediated DNA damage, suggesting that H 2 O 2 reacts with Cu(I) to produce active species causing DNA damage. DNA-bound Cu(II) can undergo Cu(II)/Cu(I) redox cycling; also, O 2 is reduced to O 2 . , which is dismutated to H 2 O 2, resulting in the DNA-Cu(I)-H 2 O 2 complex (33). Typical free ⅐OH scavengers showed little or no inhibitory effect on the DNA damage. Therefore, it is considered that the DNA damage is caused by a reactive oxygen species such as Cu(I)OOH rather than free ⅐OH. Site-specific DNA cleavage supports this hypothesis, because it is known that free ⅐OH causes DNA damage without site specificity (38). Thus, it is speculated that H 2 O 2 reacts with endogenous metals to form a metal-oxygen complex causing oxidative DNA damage in vivo. ESR spectroscopic studies using a trapping agent POBN have demonstrated retinol-and retinal-derived radicals with six-line signals, assigned as carbon-centered radicals. Retinoic acid was shown to be oxidized to 5,6-epoxyretinoic acid via formation of carbon-centered radicals (a H ϭ 0.27 mT, a N ϭ 1.55 mT) trapped by POBN (39). In addition, Tesoriere et al. (40) suggest that retinol reacts with radical addition to the cyclohexenyl ring and was further oxidized to 5,6-epoxyretinol. These carbon-centered radicals from retinol and retinal do not appear to be the main species to react with DNA, because neither retinol nor retinal caused DNA damage even though it was detected using the ESR spin-trapping method when Cu(II) was omitted. Using the cytochrome c reduction method, generation of O 2 . derived from the autoxidation of retinoids was detected. Within a 2-h incubation, higher generation of O 2 . from retinol was observed than that from retinal, and the yield became very similar at 3-h. After a 6-h incubation, O 2 . from retinal was higher than that from retinol. The generation of  Halliwell (41) indicated that so-called "antioxidants" act as antioxidants in some circumstances but also act as prooxidants in other circumstances. Palozza et al. (21) report that ␤-carotene at high oxygen partial pressure lost its antioxidant activity in normal cells and exhibited a prooxidant effect in tumor cells. Caffeic acid, which has antioxidant properties and carcinogenicity, inhibited oxidative DNA damage, whereas caffeic acid by itself caused oxidative DNA damage at certain concentrations (42). Such antioxidant compounds like caffeic acid may have the dual function of carcinogenic and anticarcinogenic potentials. If antioxidants have a prooxidant property, they may have carcinogenic effects. Similarly, it was reported, using a simplified in vitro model, that an antioxidant vitamin E (␣-tocopherol) can act as a potent oxidative DNA-damaging agent in the presence of Cu(II) (14). Virtually all putative chemopreventive antioxidants may have potential carcinogenicity (18,20).
It is noteworthy to find that the addition of such low doses of retinoids can induce cellular DNA damage. ␤-Carotene is converted to two molecules of retinal principally by central cleavage. Retinal is further oxidized to retinoic acid or reduced to retinol. We confirmed using isolated DNA that other derivatives of vitamin A such as retinoic acid and ␤-ionone and ␤-carotene itself induced slight DNA damage, including 8-ox-odG in the presence of Cu(II), but not efficiently in comparison with retinol and retinal (data not shown). If excessive intake or supplements of vitamin A and ␤-carotene saturate binding protein, free compounds may have cytotoxicity. This has been suggested by several studies (43 44). Excess amounts of vitamin A increased chromosomal aberration in a lymphocyte culture system (43). Rats treated with retinol had increased incidence of pheochromocytomas in a dose-related manner (44). The supplements of ␤-carotene and retinol resulted in higher incidences of lung cancer in the intervention groups (4,5).
It is known that reactive oxygen species are related to tumorpromoting potencies (45,46). Reid and Loeb (30) indicated that oxidative stress caused tandem double CC 3 TT mutations. It is very interesting that retinoids strongly generated piperidinelabile sites at the CC sequences in the p53 tumor suppressor gene observed in our study with 32 P-labeled DNA fragment. On the basis of the finding that excessive retinol and retinal induced oxidative DNA damage via O 2 . generation, it is suggested that the oxidative DNA damage may be responsible for initiation and/or tumor promotion/progression in multistage carcinogenesis. Retinol and retinal, ␤-carotene metabolites, have a stronger ability to induce DNA damage and to generate O 2 .
than ␤-carotene itself. Therefore, it is suggested that both retinol and retinal play important roles in carcinogenesis in the intervention studies using excess amounts of ␤-carotene. Further studies will be required for evaluation of safety and efficacy before recommending use of retinoid supplements for cancer chemoprevention.